- Email: [email protected]
Impact assessment of WHO TobReg proposals for mandated lowering of selected mainstream cigarette smoke toxicants
Accepted Manuscript Impact assessment of WHO TobReg proposals for mandated lowering of selected mainstream cigarette smoke toxicants Alison C. Eldridge, Kevin G. McAdam, Tatiana R. Betson, Marcos V. Gama, Christopher J. Proctor PII:
S0273-2300(17)30053-3
DOI:
10.1016/j.yrtph.2017.02.022
Reference:
YRTPH 3782
To appear in:
Regulatory Toxicology and Pharmacology
Received Date: 5 October 2016 Revised Date:
23 January 2017
Accepted Date: 27 February 2017
Please cite this article as: Eldridge, A.C., McAdam, K.G., Betson, T.R., Gama, M.V., Proctor, C.J., Impact assessment of WHO TobReg proposals for mandated lowering of selected mainstream cigarette smoke toxicants, Regulatory Toxicology and Pharmacology (2017), doi: 10.1016/j.yrtph.2017.02.022. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT 1
Impact assessment of WHO TobReg proposals for mandated lowering of selected mainstream
2
cigarette smoke toxicants
3
Authors
4
Christopher J. Proctor
Alison C. Eldridge, Kevin G. McAdam, Tatiana R. Betson, Marcos V. Gama, and
AC C
EP
TE D
M AN U
SC
RI PT
5
1
ACCEPTED MANUSCRIPT 6
Abstract
7
The WHO Tobacco Product Regulation Study Group (TobReg) has proposed three regulatory
8
models for cigarettes, each creating mandatory limits for emissions of nine smoke toxicants. One
9
approach proposes country-specific limits, using median or 1.25x median toxicant/nicotine emission ratios. A second model provides fixed toxicant-ratio limits. The third model limits were
11
three times the lowest toxicant emission on a market. Currently, the practical implications of
12
these models are largely unknown.
13
An impact assessment was conducted using cigarette data from 79 countries to identify four
14
diverse test markets. We sampled all products from each market but limited product availability
15
led to incomplete (80-97%) sourcing. Analysis showed that the country-specific model led to
16
diverse (up to threefold) toxicant limits across the four markets. 70%–80% of products were
17
non-compliant, rising to 100% in some countries with the second and the third models. With
18
each regulatory model the main drivers of non-compliance were the tobacco-specific
19
nitrosamines, the simultaneous application of limits for nine poorly correlated smoke toxicants,
20
and analytical variability. Use of nicotine ratios led to compliance of some high toxicant emission
21
products due to high nicotine emissions.
22
Our findings suggest that these proposals would have greater impact on global markets than
23
TobReg’s stated aims.
25
SC
M AN U
TE D
EP
AC C
24
RI PT
10
2
ACCEPTED MANUSCRIPT
28
WHO – World Health Organisation
29
TobReg – WHO Tobacco Product Regulation Study Group
30
HPHC – Harmful or Potentially Harmful Compounds
31
ISO – International Organization for Standardization
32
FCTC – Framework Convention on Tobacco Control
33
NNN – N’-nitrosonornicotine
34
NNK – 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone
35
B[a]P – Benzo[a]pyrene
36
PCA - Principal components analysis
37
HCI – Health Canada Intense
38
CO – Carbon monoxide
39
BAT – British American Tobacco
40
ANVISA – Brazilian Medical Device Regulations
41
TSNAs – Tobacco Specific Nitrosamines
SC
Abbreviations
AC C
EP
TE D
M AN U
27
RI PT
26
42
3
ACCEPTED MANUSCRIPT 43
1. INTRODUCTION
44 The mortality and morbidity associated with cigarette use is a hazard currently facing over a billion
46
smokers (WHO 2016a). Health risks have been found to diminish after cessation and disease
47
progression may slow substantially (US Department of Health and Human Services, 2014). However,
48
despite significant global Public Health efforts towards smoking cessation, WHO projections suggest
49
that if current trends continue annual deaths from smoking are likely to increase by 2030 (WHO
50
2011).
SC
RI PT
45
51
Since the 1950s scientists have sought to explain the risks associated with cigarette smoking by
53
identifying and quantifying compounds in tobacco and smoke that have toxic properties. These
54
compounds are often referred to as toxicants and several of them have been identified, the number
55
of which has grown longer as toxicological understanding and analytical techniques have improved.
56
The current reference point for cigarette smoke toxicants is the established list of over 90 Harmful or
57
Potentially Harmful Compounds (HPHC) identified by a Technical Advisory group to the FDA in 2011
58
(FDA, 2011). The US Institute of Medicine has expressed the view that some of the harm caused by
59
tobacco use may potentially be reduced through introduction of products that might result in
60
substantial reduction in exposure to one or more tobacco toxicants (Stratton et al., 2001).
TE D
EP
AC C
61
M AN U
52
62
Cigarette smoke toxicants have been the focus of increased regulatory interest since this time (Liu et
63
al. 2013). Traditional approaches for the regulation of cigarette products, based upon reporting and
64
limiting cigarette emissions of tar, nicotine and carbon monoxide measured under the ISO smoking
65
regime have been replaced or supplemented by reporting requirements on cigarette smoke
66
toxicants and at additional smoking regimes. Starting with the mandated measurement and
67
reporting of toxicant emissions in Canada (Health Canada 2000) and Brazil (ANVISA 2007), the
4
ACCEPTED MANUSCRIPT requirement to measure and report emissions has gradually spread to other countries such as
69
Taiwan (Taiwan 2010), and the USA (FDA 2012). Cigarette emission regulations currently show a
70
great deal of inconsistency from country to country, ranging from ceilings on tar, nicotine and
71
carbon monoxide in the European Union and other jurisdictions to detailed, but differing,
72
requirements for annual per-product emission reporting of 7 constituents in Taiwan, 18 constituents
73
in the USA, and more than 40 in Brazil, Canada and Venezuela (Liu et al, 2013).
RI PT
68
74
In 2003 the World Health Organization (WHO) adopted the Framework Convention on Tobacco
76
Control (FCTC), to which 168 countries globally are signatories and 180 are parties (WHO 2016b).
77
The FCTC has an objective of “providing a framework for tobacco control measures to be
78
implemented by the Parties at the national, regional and international levels in order to reduce
79
continually and substantially the prevalence of tobacco use and exposure to tobacco smoke”. FCTC
80
represents a mechanism for global deployment of tobacco control initiatives (WHO 2005).
M AN U
SC
75
TE D
81
Articles 9 and 10 of the FCTC focus on tobacco product regulation (WHO 2005). As a step towards
83
this, in 2008, a WHO advisory group on Tobacco Product Regulation (TobReg) published a new
84
strategy for tobacco product regulation (WHO 2008, Burns et al 2008). The proposed strategy
85
focused on use of standardised measures of cigarette smoke toxicity that characterised, as far as
86
possible, potential differences in harm caused by different cigarettes. Central to TobReg’s strategy is
87
the mandated lowering of nine separate toxicants in cigarette smoke emissions. The toxicants
88
identified for mandated reduction were chosen based on toxicity, observations of differences in
89
yields across brands, availability of technology or other approaches to reduce yields, and the
90
existence of markets for low yield products. The proposed approach was therefore viewed as an
91
example of the precautionary principle often deployed in public health with parallels drawn to
92
strategies used with other consumer products, where the focus is also to reduce levels of known
93
toxicants present in the product.
AC C
EP
82
5
ACCEPTED MANUSCRIPT The nine selected toxicants: 1,3-butadiene, acetaldehyde, acrolein, benzene, benzo[a]pyrene (B[a]P),
95
carbon monoxide, formaldehyde, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK) and N’-
96
nitrosonornicotine (NNN), represent a range of chemical classes found in cigarette smoke, covering
97
both the vapour phase and the particulate phase, and were viewed as implicated in carcinogenicity,
98
cardiovascular and pulmonary toxicity; they also commonly appear in various other existing lists of
99
identified smoke toxicants. As TobReg’s proposals were founded upon the widely accepted
RI PT
94
regulatory practice of reducing toxicants in products intended for human use TobReg did not regard
101
it as necessary to have specific proof of a link between a lower level of these toxicants in cigarette
102
smoke and a lower level of human disease. This is an important point, as links between individual
103
cigarette smoke toxicants and human disease are currently incomplete (Burns et al., 2008).
M AN U
SC
100
104
The 2008 TobReg proposals recommended two models, both of which established emission limits for
106
each of the 9 toxicants, as measured using the Health Canada Intense smoking regime, and
107
expressed per milligram of nicotine. Use of the ratio to nicotine was proposed on the basis that
108
machine-measured per-cigarette smoke emissions were unreliable estimates of smoker’s exposure
109
to cigarette smoke. TobReg noted that individual smokers “seek to achieve nicotine intakes
110
sufficient to satisfy their addiction”. Use of a toxicant/nicotine ratio was therefore described as
111
shifting focus away from quantity of smoke generated per cigarette and preventing interpretation
112
that the values obtained represent a measure of smoker exposure to toxicants (WHO 2008).
EP
AC C
113
TE D
105
114
Under the TobReg proposals failure of a cigarette brand to meet each of the nine separate limits
115
would lead to its withdrawal from sale on a market. The strategy was recommended to be
116
implemented in phases, starting with annual reporting of toxicant levels for 2-3 years, setting
117
toxicant limits from these data, enforcing limits two years after they are set, and potentially lowering
118
limits progressively over time. The aim in setting the limits was to balance the need to regulate a
6
ACCEPTED MANUSCRIPT 119
range of toxicants, to mandate lowering those toxicants to the greatest extent and yet not to
120
eliminate most brands from the market in their current form (WHO 2008).
121 Under the first TobReg model (“model 1”) the toxicant emission limits for a country were to be set
123
by measuring emission ratios to nicotine for all products on a market, calculating the median value
124
for each toxicant, setting this value as the limit for NNN and NNK, and setting the limits for the other
125
seven toxicants as 125% of their respective median values (Burns et al., 2008).
RI PT
122
SC
126
The second TobReg model (“model 2”) was proposed for countries possessing limited laboratory
128
capacity. This model provided two sets of fixed regulatory limits that could be deployed across
129
multiple countries. One set of limits, labelled ‘international brands’, was calculated from an
130
International measurement survey of 49 Philip Morris cigarette brands, conducted in 2001, (Counts
131
et al 2005) and were proposed for use in countries whose cigarette products were predominately
132
US- or American blend products, or where the style of cigarettes on the market are not readily
133
identifiable. The second set of limits, labelled ‘Canadian brands’, were calculated from a set of 91
134
Canadian brands reported to Health Canada in 2004, reduced to exclude brands with levels of NNN
135
per mg nicotine > 0.1 ng, which eliminates most US and Gauloise brands (12 brands), and duplicate
136
and erroneous values (31 brands), leaving a database of 48 Canadian products representing
137
“unblended cigarettes containing predominantly flue-cured (bright) tobacco”; these limits were
138
intended for use in countries whose cigarette products reflected more flue-cured or “Virginia” blend
139
style products (WHO 2008).
TE D
EP
AC C
140
M AN U
127
141
TobReg conducted a limited impact assessment of the proposed models using these datasets, and
142
concluded that regulation of brands for toxicants other than NNK and NNN would result in 40-41% of
143
brands failing to meet the set of limits without modification (WHO 2008). This form of analysis was
144
based on the belief that there is existing technology for dramatically lowering the nitrosamine 7
ACCEPTED MANUSCRIPT contents of tobaccos. This position arose from analysis showing wide differences in nitrosamine
146
emissions between brands and geographic sources, and reports that modifying North American flue-
147
curing approaches from direct-heating to indirect heating mechanisms were lowering nitrosamine
148
levels in USA and Canada tobaccos and cigarettes (Peele et al 2001, IARC 2004, Gray and Boyle 2004).
149
TobReg also concluded that the values of individual toxicant ratios found across the datasets suggest
150
that mandated reductions would have a substantial effect on the emissions from brands remaining
151
on the market.
RI PT
145
SC
152
During the present study, a second WHO technical report on toxicant levels in smoke was published
154
(WHO 2015). The technical report includes an independent commentary by the late Dr Nigel Gray
155
which, whilst not necessarily representing the view of the WHO or TobReg, was unanimously
156
recommended for inclusion in the report due to ‘the thought-provoking nature of its content and
157
goals’. Gray set forth a third model (model 3) of toxicant emission limits based on a ‘generous’
158
upper limit set at three times the lowest toxicant emission level achieved on the market, reviewable
159
after 2 years and then, where practical, set lower.
TE D
160
M AN U
153
A limitation in the TobReg approach, identified by TobReg [WHO 2008], was the lack of available full-
162
market data with which to assess the impact of TobRegs proposals and consequent reliance on the
163
relatively small available dataset of publically available toxicant emission values. TobReg noted the
164
limitations involved in the limited dataset covering few geographic sources and cigarette brands, and
165
recognised that the performance of charcoal filter cigarettes were not well characterised by the
166
approach of normalising yields to nicotine (a limitation that might also apply to products with
167
different circumferences). In the eight years following publication of the TobReg proposals the
168
limited database of cigarette toxicant yields has not grown significantly, and therefore the impact of
169
TobRegs proposed toxicant-reduction regulations on real world cigarette markets, together with the
AC C
EP
161
8
ACCEPTED MANUSCRIPT practical realities of adhering to the proposals, have not been tested and their implications remain
171
unclear.
172
The aim of the present study was to fill this knowledge-gap, and thereby assess the real-world
173
impact of TobReg’s proposals. Four diverse cigarette markets, identified initially by principal
174
components analysis (PCA) of an extensive database of smoke emission values (Camacho et al 2015),
175
were chosen to understand the performance of the TobReg proposals under the widest possible
176
range of testable conditions. Cigarette products were sampled from these countries following the
177
TobReg approach, and smoke emissions of the nine toxicants and nicotine were determined in these
178
products under the Health Canada intense (HCI) smoking regime. These data were then used to test
179
the three TobReg models, in order to characterise their impact on these four diverse cigarette
180
markets.
M AN U
SC
RI PT
170
181 2. METHODS
183
2.1 Selection of study markets
184
Principal components analysis (PCA) was used to select four diverse markets with which to assess
185
the impact of the TobReg proposals for mandated lowering of emissions of nine smoke toxicants. An
186
in-house database containing recent (2006-2011) measurements on tobacco filler blend components,
187
mainstream smoke toxicant emissions, and various physical parameters for 811 commercial
188
cigarette products from 79 markets was used in the PCA to choose four markets with distinctly
189
different product styles.
190
Smoke toxicant emission yields were not used as variables in the PCA because the majority of the in-
191
house data had been generated under the ISO smoking regime, rather than HCI regime upon which
192
the TobReg recommendations are based [WHO, 2008; Burns et al 2008]. Instead, the cigarette filler
193
blend components NNK and NNN, which are direct precursors of NNK and NNN in smoke, were used,
AC C
EP
TE D
182
9
ACCEPTED MANUSCRIPT together with total sugar, a precursor of the smoke toxicant formaldehyde. A fourth variable, filter
195
charcoal loading was included in the PCA due to its ability to affect the levels of volatile smoke
196
toxicants, such as acetaldehyde, acrolein, benzene and 1,3-butadiene. The variables entered in the
197
PCA include factors affecting eight of the nine toxicants proposed for mandated lowering, carbon
198
monoxide (CO) being the exception. The PCA was conducted via JMP Pro version 10 (SAS Institute,
199
Cary, NC, USA).
200
2.2 Study products
201
After the study markets had been selected by PCA, a list of all current cigarette products (regardless
202
of manufacturer) on sale in each of those markets was obtained by local BAT sales-force employees
203
or in the case of Brazil, where product registration is a regulatory requirement, the product list was
204
obtained from ANVISA. The Brazilian market was sampled in Q1 2012; Romania in Q4 2012; Australia
205
in Q1 2013; and Germany in Q4 2013. The products were either sourced directly from BAT factories
206
or purchased from the market place in the case of brands from other manufacturers. The
207
acquisition of products took several months for each market. Once acquired product samples were
208
sent directly to the analysis laboratory.
209
2.3 Product analysis
210
All products were analysed at a single, ISO 17025 accredited laboratory in Brazil. The four markets
211
were analysed sequentially in a series of batches wherein each market was measured as a whole
212
prior to commencement of the next study market. This process lasted more than 2 years, owing to
213
the large number of study products.
214
Smoke toxicant emissions of acetaldehyde, acrolein, 1,3-butadiene, benzo[a]pyrene (B[a]P), benzene,
215
CO, formaldehyde, NNN, NNK, tar, and nicotine were determined under the HCI regime [Health
216
Canada, 1999]. If the product contained a flavour capsule then analysis was carried out after the
217
capsule had been crushed. Five replicate measurements were conducted per product for
AC C
EP
TE D
M AN U
SC
RI PT
194
10
ACCEPTED MANUSCRIPT acetaldehyde, acrolein, 1,3-butadiene, B[a]P, benzene, formaldehyde, NNN and NNK, whereas eight
219
replicates were conducted for tar, nicotine and CO, in compliance with ISO sampling requirements
220
for these toxicants [ISO 8243:2013].
221
The mainstream smoke toxicants were analysed by standard methods that have been internally
222
validated for repeatability and reproducibility [AOAC 2002; ISO 1994]. The methods follow, or are
223
based on, internationally standardised or recognised protocols (i.e. ISO, CORESTA or official Health
224
Canada methods). The methods are multi-analyte, whereby members of a group of toxicants (e.g.
225
volatiles or carbonyls or tobacco specific nitrosamines, etc.) are analysed simultaneously from the
226
same cigarettes. Details of the analytical methods are given in the Supplementary Information.
M AN U
SC
RI PT
218
227 228 2.4 Data analysis
230
Smoke emissions data were collated and toxicant-to-nicotine emission ratios were calculated using
231
Base SAS version 9.3 (SAS Institute). Product assessment against potential toxicant emission ratio
232
limits was performed with Microsoft Excel 2010 (Microsoft Corporation, Redmond, WA, USA). The
233
four markets were analysed separately.
234
In analysing TobReg model 1, for each market, the limit for each of the nine toxicants was calculated
235
in accordance with TobReg [WHO, 2008] as follows. For each toxicant, the mean yields of toxicant
236
and nicotine for each product were used to calculate the mean emission level as a ratio to the mean
237
yield of nicotine: the emission ratio value. The median emission ratio value per market was then
238
determined across the range of toxicant ratios for all products measured in that market. The toxicant
239
emission limits for each market were determined as the median emission ratio value for NNN and
240
NNK, and 125% of the median emission ratio value for formaldehyde, acetaldehyde, acrolein, B[a]P,
241
CO, 1,3-butadiene and benzene. With TobReg model 2, the limits used were those published by
AC C
EP
TE D
229
11
ACCEPTED MANUSCRIPT TobReg [WHO 2008, Burns et al 2008]. With TobReg model 3 (WHO 2015) the limits were calculated
243
for each market by establishing the minimum per-cigarette emission level for each toxicant, and
244
multiplying this value by three. The measured toxicant emission ratio values, or toxicant emission
245
levels, were compared to the limits for each model, and the number of compliant and non-compliant
246
products calculated with regard to each toxicant limit individually and also when all nine limits were
247
applied simultaneously.
RI PT
242
SC
248 3. RESULTS
250
3.1 Selection of study markets using Principal Component Analysis
251
PCA was used to identify four markets with diverse products. The four input variables, filler blend
252
content of NNN, NNK and total sugar, and filter charcoal loading (Supplementary Table 1), were
253
chosen on the basis of data measured by BAT on 811 products from 79 countries (Supplementary
254
Table 2). The PCA used parameters representing the majority of the nine TobReg toxicants. B[a]P in
255
filler blend was also considered as a variable for the PCA, as it is a driver for B[a]P smoke emissions,
256
however this data was not available for all products in the in-house database. Other physical
257
measures, such as circumference, were not included in the PCA as the majority of in-house data was
258
from king-size circumference cigarettes.
259
The PCA defined a variable for blend character (i.e., TSNA to formaldehyde (sugar)) as principal
260
component 1 (PC1), accounting for ~45% of the variation in product measurements, and a variable
261
for the amount of charcoal included in the cigarette filter as principal component 2 (PC2), accounting
262
for ~26% of the variation associated within these measurements. The PC1 and PC2 scores of the
263
products were plotted to identify diverse markets (Figure 1).
264
AC C
EP
TE D
M AN U
249
12
ACCEPTED MANUSCRIPT From the score plot, we identified four groupings that described the most diverse market scenarios
266
within our database. These four markets were Brazil (mixed blend products, low filter charcoal
267
prevalence; Figure 2A) ; Romania (mixed blend products, high filter charcoal prevalence; Figure 2B) ;
268
Australia (predominantly flue-cured Virginia blended products; Figure 2C) ; Germany (predominantly
269
US-blended products; Figure 2D).
270
271
After selection of the four study markets, current products on sale in each market were obtained for
272
analysis. Sourcing the test products in each market proved logistically very challenging, for reasons
273
such as delisting of product and geographical limitations in distribution. In addition, it was not
274
possible to source a sufficient volume of some products (a minimum of 400 cigarettes per product
275
were required for analysis, with 800 preferred) due to the limited availability of low-market share
276
cigarette brands and product de-listing during the sampling exercise. Across the four markets, the
277
percentage of products collected in sufficient quantity for testing was 80% (132 out of 166 products)
278
in Brazil, 93% (138 out of 148 products) in Romania, 97% (172 out of 177 products) in Australia, and
279
92% (339 out of 367 products) in Germany. Given the difficulty encountered with the sampling
280
exercise, multiple time-point sampling of the type recommended by TobReg was not possible.
281
Nevertheless, the products sampled in this single point-in-time exercise were considered sufficient
282
for the analysis to be conducted as planned.
283
3.2 Mainstream smoke toxicant emissions
284
HCI mainstream smoke emissions of the nine TobReg toxicants plus nicotine and tar were measured
285
for each test product, and the distribution of toxicant emission yields per market, both on a per
286
cigarette basis and as a ratio to nicotine emission yield, were summarized (Table 1).
287
Regarding the toxicant yields per cigarette, fairly normal distributions were observed with mean and
288
median values being similar in each case, and no markedly different values comparing the minimum
AC C
EP
TE D
M AN U
SC
RI PT
265
13
ACCEPTED MANUSCRIPT yield to the lower quartile nor maximum yield to the upper quartile, (Table 1, Supplementary Figure
290
1). In general, standard deviation measurements were ~20% of mean or median values however
291
standard deviations were a larger percentage of the mean or median values for NNN and NNK,
292
particularly for NNN in Australia and NNK in Brazil.
293
There were a few exceptions due to the inclusion of atypical products. In the Australian market, zero
294
yields of nicotine (and TSNA) were determined for two herbal cigarettes; as a result, these two
295
products were excluded from the yield determinations per mg of nicotine, although their yields of
296
the seven non-TSNA toxicants from these non-tobacco cigarettes were similar to those in
297
conventional cigarettes. The Australian market survey also included a kretek-style product (a blend
298
of tobacco, cloves, and other flavours), which was responsible for the notably high maximum yields
299
of nicotine (3.58 mg/cig), tar (61.5 mg/cig), B[a]P (40.9 ng/cig), formaldehyde (253 µg/cig), acrolein
300
(187 µg/cig) and benzene (135 µg/cig) in that market. The German market included a cigarette
301
wrapped in a reconstituted tobacco sheet instead of cigarette paper, which produced the notably
302
high maximum yields determined for acetaldehyde (2567 µg/cig), 1,3-butadiene (385µg/cig),
303
benzene (262µg/cig) and CO (72.1 mg/cig) in that market.
304
Comparing the emissions of each toxicant among the four sampled markets, the Australian market
305
exhibited lower TSNA yields and higher formaldehyde yields, which is consistent with the
306
predominantly flue-cured characteristics of products in this market. B[a]P yields were lower in the
307
Romanian market, whereas nicotine and 1,3-butadiene yields were lower in the Brazilian market
308
(Table 1).
309
Considering toxicant yields per mg nicotine (Table 1, Supplementary Figure 2), summary statistics
310
for yields per mg nicotine were lower as compared with the results per cigarette because, on
311
average, nicotine yields were greater than 1 mg/cig. The exception to this, besides the two herbal
312
Australian products already mentioned, was a locally produced Brazilian product that had the
313
highest levels of NNN (392 ng/cig) and NNK yields (670 ng/cig) on a per cigarette basis, in that
AC C
EP
TE D
M AN U
SC
RI PT
289
14
ACCEPTED MANUSCRIPT market, and the lowest nicotine yield (0.82 mg/cig) of all the Brazilian products surveyed, and
315
therefore produced notably high TSNA yields per mg nicotine (NNN 479 ng/mg nicotine, NNK 818
316
ng/mg nicotine). This product also produced the maximum yield ratios in Brazil for B[a]P (28.4 ng per
317
mg nicotine), acetaldehyde (1847 µg per mg nicotine), acrolein (170 µg per mg nicotine), benzene
318
(93 µg per mg nicotine) and CO (30.2 mg per mg nicotine), as a consequence of the low nicotine yield.
319
320
3.3 Toxicant limits
321
Model 1:
322
Market-specific toxicant limits were calculated by using the method proposed by TobReg (Table 2).
323
The toxicants with the widest range of TobReg limits were the two TSNAs (NNK, 21–78 ng/mg
324
nicotine; NNN, 27–90 ng/mg nicotine), where the range of limits (difference between highest and
325
lowest limit expressed as a % of the lowest limit) was 230-270%. In contrast, six of the other seven
326
toxicants showed much smaller ranges, an order of magnitude lower, at 17 - 30%; the range of limits
327
for B[a]P was also comparatively low, at around 50%. Brazil had the highest limits for seven of the
328
nine toxicants with only the 1,3-butadiene limit being relatively low (second lowest when compared
329
across the 6 datasets). Romanian products were notable for having the lowest calculated limits for
330
six toxicants. The lowest TSNA limits were found with Australian products under model 1.
331
Model 2:
332
Comparing the limits calculated under model 1 with the fixed limits provided by TobReg under
333
model 2 (Table 2), showed that the ‘Canadian brands’ limits set by TobReg were lower than those
334
calculated in the current study for five of the nine toxicants, but higher for formaldehyde.
335
Comparing the limits for the two flue-cured datasets (Australia in this study, model 1 vs Canadian
336
brands limits set by TobReg, model 2), the largest discrepancy was in NNK limits, where the Canadian
AC C
EP
TE D
M AN U
SC
RI PT
314
15
ACCEPTED MANUSCRIPT brands limit set by TobReg was more than twice the market-limit determined for Australia in this
338
work; the difference between the limits for the other toxicants ranged from –27% to 20%. For the
339
two US-blended datasets (Germany in this study, model 1 vs International brands limits set by
340
TobReg, model 2), the largest discrepancy was in TSNA limits, where the International brands limits
341
for both NNK and NNN were more than 50% higher than those determined with German brands in
342
this work; otherwise, the differences ranged between –6% and 23% per toxicant. The International
343
brands limit (model 2) for NNN was higher than any of the four markets surveyed in the present
344
study, model 1.
345
Market-specific limits were also calculated according to model 3 (Table 3). The toxicants with the
346
widest ranges of limits were again the two TSNAs, with a 170% range of limits for NNK, and a 270%
347
range of limits for NNN. These ranges of TSNA limits with model 3 are similar to the range of limits
348
found with TobReg model 1. The range of model 3 limits for the other seven toxicants are 36-96%,
349
which is a greater range than found under model 1.
350
351
3.4 Product compliance with limits
352
3.4.1 Model 1 - Market-specific limits
353
For each market, the proportions of non-compliant products were determined individually for each
354
toxicant (Figure 3A) and cumulatively when applying all nine toxicant limits simultaneously (Figure
355
3B). Notably, between 72% and 79% of products within each market would be non-compliant
356
following the TobReg model 1 approach (Table 4), i.e. after the application of all nine market-specific
357
toxicant limits [WHO 2008]. The proportions of non-compliant product were also determined
358
cumulatively when applying to both TSNAs (median limits) and cumulatively for the non-TSNA
359
toxicants (125% of median limits), (Figure 3B). The consistency of the proportions of non-compliant
AC C
EP
TE D
M AN U
SC
RI PT
337
16
ACCEPTED MANUSCRIPT product across the four markets for each comparison in Figure 3 emphasises the mathematical
361
nature of this approach.
362
363
The main driver of non-compliance in all markets were the limits for the two TSNAs, where, by
364
definition, the TobReg proposed use of median values ensures that 50% of products would be
365
immediately non-compliant with limits for both NNN and NNK. Imperfect correlation between the
366
two TSNAs resulted in a combined non-compliance rate of 56-66% (Figure 3B) for these two
367
toxicants. The non-TSNA toxicant limits, being based on 125% of the market median, produced
368
fewer non-compliant products than the two TSNAs. The individual non-compliance rates for the non-
369
TSNA toxicants ranged from 4% (benzene in Romania) to 24% (1,3-butadiene in Brazil), with a
370
combined non-compliance rate of 34-45% for these 7 toxicants (Figure 3B).
371
Regarding the number of toxicants for which each individual product was non-complaint (Figure 3C),
372
it was most common for products to be non-compliant for 0, 1 or 2 toxicants (20%–30% of products
373
for each), followed by 3 toxicants (~10%), and then 4–9 toxicants (<5% for each). The two TSNAs
374
accounted for most products with 1 or 2 toxicant non-compliances, whereas the third most common
375
non-compliant toxicant varied by market: 1,3-butadiene in Brazil and Australia; formaldehyde in
376
Romania and Germany (Figure 3A).
377
3.4.2 Model 2 - TobReg fixed toxicant limits
378
The application of limits based on ‘International brands’ or ‘Canadian brands’, set by TobReg under
379
model 2 for use in the absence of market specific information [WHO 2008], resulted in an increase in
380
overall numbers of non-compliant products, as compared with model 1’s calculated market-specific
381
limits, for most market comparisons (Table 4). This increase was often substantial—for example,
382
100% of products in Brazil were non-compliant against both of TobReg’s fixed limits (Figure 4A and
383
4B).
AC C
EP
TE D
M AN U
SC
RI PT
360
17
ACCEPTED MANUSCRIPT Application of the International brands limits to the predominantly US-blended German market
385
(Table 4, Figure 4A), increased the percentage of non-compliant products from 80% to 89%. Similarly,
386
application of the Canadian brands limits to the predominantly flue-cured Australian market resulted
387
in an increase in non-compliant products from 72% to 80%. Only the Romanian market (mixed blend;
388
high incidence of charcoal in the filter) showed a reduction in non-compliant products from 78% to
389
71% when the International brands limits were compared with the market-specific limits of model 1.
390
391
Comparing product non-compliance rates for NNN and NNK when using the two sets of fixed limits
392
compared to the market specific limits of model 1 (Figures 3A and 4), showed that use of the
393
international brands limits resulted in a marked decrease in non-compliance levels against the TSNA
394
limits for 3 of the markets surveyed in this study, whilst for the Brazilian market non-compliance
395
rates were similar (~60%). Applying the Canadian brands fixed limits for NNN and NNK increased
396
non-compliance rates for 3 of the markets substantially (>90%), whereas for the Australian market
397
there was a small decrease in non-compliance rates, from 56 to 50%.
398
Product non-compliance rates for non-TSNA toxicants when using TobReg fixed limits showed a
399
marked increase to 66%-99% non-compliance, compared to ~40% non-compliance when using the
400
market-specific limits of model 1 (Figure 3B and Figure 4). Formaldehyde limits produced the
401
greatest number of non-compliances under the ‘International brands’ set of limits (59-88% non-
402
compliance), but a negligible level of non-compliance under the ‘Canadian brands’ limits (1-2% non-
403
compliance). Conversely, acetaldehyde emissions produced high levels of non-compliance under the
404
‘Canadian brands’ limits (45%-88%), but lower levels of non-compliance under the ‘International
405
brands’ limits (1%-36%). Formaldehyde was the main driver of non-compliance when applying the
406
‘International brands’ limits in each of the four markets whilst NNN, followed by NNK (except for the
407
Australian market products), acetaldehyde, benzene (except for the Romanian market products) and
408
1,3-butadiene were the main drivers when applying the ‘Canadian brands’ limits. The Brazilian
AC C
EP
TE D
M AN U
SC
RI PT
384
18
ACCEPTED MANUSCRIPT market in particular was negatively affected by the use of the fixed limits with the highest levels of
410
non-compliance for each toxicant individually, apart from 1,3-butadiene, and 100% non-compliance
411
when applying all 9 limits simultaneously under either set of limits.
412
3.4.3 Model 3 limits:
413
The level of product non-compliances obtained when applying the model 3 limits to each of the
414
appropriate markets are summarized in Table 4. The total level of non-compliances varied
415
significantly by market, ranging from 68% in Australia to 99% in Germany. The impact on the
416
Australian and Romanian markets were comparable to that found with the market-specific limit
417
model. However, the model 3 limits produced significantly higher levels of non-compliances for the
418
Brazilian (86%) and German markets (99%).
419
The levels of non-compliances against the model 3 limits were driven strongly by NNN, and to a
420
lesser degree, by NNK, Figure 5. The other seven toxicants resulted in few non-compliances, other
421
than with formaldehyde in Romania (35% non-compliances) and Germany (14%), and with B[a]P in
422
Romania (17%) and Brazil (15%).
423
EP
TE D
M AN U
SC
RI PT
409
425
4. DISCUSSION
426
The principal goal of TobReg’s proposed strategy is to reduce the emissions of nine selected
427
mainstream smoke toxicants in commercial cigarettes by excluding products with the highest
428
smoking-machine measured levels of these toxicants. TobReg’s proposals are based on a complex
429
trade-off of considerations that they believe will result in substantial lowering of toxicant emissions,
430
while not resulting in the elimination of most of the brands sold on a market. This study represents
431
the first full evaluation of this global proposal for tobacco product regulation, and has explored the
AC C
424
19
ACCEPTED MANUSCRIPT impact of the proposed regulations on real-world, whole-market data. A number of important
433
learnings have emerged from this analysis.
434
4.1 Toxicant ceilings arising from TobReg models
435
4.1.1 Comparison of model 1 and model 2 limits.
436
The suitability of TobReg proposed fixed limits for global deployment can be assessed by comparison
437
with the individual market limits calculated in this work, Table 2. Very few of the individual limits
438
from the four markets examined in the current work gave a set of limits that matched the TobReg
439
proposals. Comparing the Australian market limits to the TobReg limits calculated from the
440
Canadian brands showed that while NNN and B[a]P limits were in exact agreement, CO and acrolein
441
limits differed by 8-10%, benzene limits by 14%, acetaldehyde, formaldehyde and 1,3-butadiene
442
limits differed by 20-30% and NNK limits differed by 55%.
443
Similarly, comparing the TobReg limits calculated from the International Brands dataset with the
444
limits obtained in this study from the German, Romanian and Brazilian datasets also showed
445
significant disagreements. CO limits differed by 3-12%, acetaldehyde limits by 4-16%, 1,3-butadiene
446
limits by 3-19%, acrolein limits by 5-28%, NNK limits by 8-36%, B[a]P limits by 18-27%, benzene limits
447
by 10-44%, NNN limits by 21-37% and formaldehyde limits by 30-57%.
448
Given the sensitivity of compliance/non-compliance to small changes in limits or emission ratios
449
around the median and 125% median values (Section 4.3) these observations clearly demonstrate
450
that the TobReg proposed fixed limits are inappropriate for use in these four markets, and most
451
likely on a global basis.
452
In understanding why the TobReg proposed fixed limits do not reflect actual market limits, it should
453
be noted that TobReg did not use full-market data to calculate either set of limits: the 2004 Canadian
454
emissions dataset contained measured toxicant emissions data for only a subset of the products
455
listed (60 out of 249 brands sold in Canada in 2004), on top of which data from US and French blend
AC C
EP
TE D
M AN U
SC
RI PT
432
20
ACCEPTED MANUSCRIPT 456
cigarettes were removed (products where NNN levels > 100 ng/mg of nicotine), leaving a final set of
457
48 products. For the ‘International brands’, the 49 products reported by Counts et al (2008) were
458
from a single manufacturer, with few charcoal filtered or reduced circumference products. Neither
459
dataset reflects the breadth of products available in any one market, let alone worldwide.
RI PT
460 The substantial differences in non-compliance rates when using either of the TobReg fixed limits,
462
together with the lack of fit with limits calculated from actual market data, highlights how important
463
it would be to have accurate, contemporary, market-specific data if the proposed approach were to
464
be enacted, particularly given the potentially high rates of product non-compliance observed in the
465
present study.
M AN U
SC
461
466 4.1.2 Model 3 limits
468
With model 3 the toxicant limits (Table 3) are defined by the emissions of one product for each
469
toxicant in each market, and therefore interpretation of these observations is likely to be challenging,
470
as the product defining the market limit may be unrepresentative of the overall market. For
471
example, Germany, a US-blended market, has the lowest TSNA limits of the four markets, despite
472
having one of the widest distributions of TSNA emissions of the markets examined in this survey
473
(Supplementary Figure 1). In contrast, the Australian formaldehyde limit is the highest of the four
474
markets, which is a representative reflection of the predominantly flue-cured nature of the tobacco
475
products sold on this market. Another example of the potential for unrepresentative products
476
defining market limits under model 3 is provided by the Brazil market. Brazil provided five of the
477
lowest model 3 limits (defined by a single product for each toxicant) in the overall dataset. This is in
478
direct contrast to the findings from application of model 1 (defined by all products on the market),
479
where Brazil had seven of the highest limits. Clearly, the method for calculation of toxicant limits
480
has a profound impact on the standards that products would have to meet if these models were
481
enacted.
AC C
EP
TE D
467
21
ACCEPTED MANUSCRIPT 4.2 Product compliance with toxicant ceilings:
483
4.2.1 Model 1 - Individual market limits
484
Our survey of four markets that differ in tobacco blend style, cigarette design and toxicant profiles
485
has shown that the impact of this proposed regulatory strategy would be consistently severe: 72%–
486
80% of the products analysed in each market were non-compliant with the market-specific limits
487
(Figure 3B). There were two main drivers of the high non-compliance levels: the primary driver being
488
use of median market yields as the ceiling for NNN and NNK emissions which in itself resulted in 56-
489
66% failures; and a secondary driver being the simultaneous application of 9 separate toxicant
490
ceilings. The end result, where most products in a market cannot comply with proposed ceilings,
491
clearly does not meet the stated intention of TobReg of providing a regulatory framework that does
492
not cause elimination of most brands in their current form from the market.
493
4.2.2 Model 2 - Fixed limits
494
Application of the two sets of ‘fixed’ TobReg proposed limits was associated, in almost all cases, with
495
substantially higher non-compliance rates than found with market-specific limits (model 1). In some
496
cases 100% non-compliance was observed. The extent of product failures with the International
497
Brand limits was 70-100%, and 80-100% with the Canadian brands limits (Figures 4A and B).
498
On application of the ‘International brands’ limits the increase in non-compliances was driven by the
499
non-TSNA toxicants, particularly formaldehyde, rather than the two TSNAs. Brands failing the
500
‘International’ formaldehyde limits reached as high as 88%, and up to 83% failed the benzene limits;
501
conversely in most cases failures against TSNAs were <30%.
502
In contrast, for the ‘Canadian brands limits’ the increase in non-compliance rates arose from a
503
number of toxicants and varied by market: there were large increases in NNN and NNK non-
504
compliances for Brazil, Romania and Germany (NNN only) and in non-compliances with
505
acetaldehyde and 1,3-butadiene, but the non-compliance rate for formaldehyde was negligible.
AC C
EP
TE D
M AN U
SC
RI PT
482
22
ACCEPTED MANUSCRIPT TobReg’s impact assessment estimated 40-41% product failures, driven by the 7 toxicants other than
507
NNN and NNK, an estimation based upon the belief that TSNA emissions could be lowered by
508
cigarette manufacturers without difficulty. Our analysis clearly demonstrates much higher levels of
509
non-compliances when actual products are examined, reaching 100% in some cases. Even the
510
impact of the seven limits for toxicants other than the TSNAs was more severe than estimated by
511
TobReg, with 66-99% failures against the International Brands limits, and 66-90% non-compliances
512
with the Canadian brands limits. Therefore, use of the proposed fixed limits does not meet the
513
TobReg intention of toxicant reduction without severe disruption to a market.
514
4.2.3 Model 3 limits
515
Application of the model 3 limits (3 x market minimum levels) also caused greater levels of non-
516
compliance than the TobReg model 1 limits (based on market median) in 2 of the 4 markets
517
surveyed in this work (Table 4). This was predominately driven by the limits for TSNAs and NNN in
518
particular; with the other seven toxicants having relatively little impact on compliance (Figure 5).
519
The model 3 limits were described as generous (WHO, 2016), however when their impact is
520
examined on typical cigarette markets, their impact has been shown to be severe, with 99% non-
521
compliances in Germany.
522
Limits calculated under model 3 are highly susceptible to low yield, atypical products. Germany was
523
by far the largest market surveyed and had an overall TSNA distribution comparable to Brazil and
524
Romania, and much higher than Australia (Supplementary Figure 1). Nevertheless, due to the low
525
TSNA emissions of one product on the German market, the NNN and NNK limits for the German
526
market were the lowest of the four markets examined in this work, which resulted in 99% non-
527
compliances with the products sold in Germany.
528
Two factors may present difficulties for the model 3 approach. First, low level quantification of
529
smoke toxicant emissions can be particularly sensitive to analytical imprecision. At low levels, such
AC C
EP
TE D
M AN U
SC
RI PT
506
23
ACCEPTED MANUSCRIPT as at the limit of quantification, analytical errors can have a substantial impact on the measured
531
value. Second, no guidance was provided (WHO 2016) on the most appropriate procedure under
532
model 3 when the lowest emission value on a market are below the limits of quantification, and this
533
represents a serious shortcoming of the proposals.
534
4.2.4 Toxicant yields as a ratio to nicotine
535
With the first two TobReg models, the recommendation was to express toxicant emission data as a
536
ratio to the measured smoke yield of nicotine in order that the machine measured values are not
537
misleadingly judged to represent a measure of human exposure and therefore risk [WHO, 2008]. In
538
contrast, the later model 3 proposals, published in WHO Technical Report 989, proposed use of per-
539
cigarette yields rather than values calculated as a ratio to nicotine emissions. To understand the
540
importance of these differences in model approaches, we therefore examined the data obtained in
541
this study to assess the impact of using nicotine ratios on product compliance.
542
543
Analysis of the data showed that setting an emission limit for a toxicant on the basis of its yield ratio
544
to nicotine resulted in some products with high absolute toxicant emissions being compliant with
545
toxicant/nicotine limits when they had a relatively high nicotine yield. Figure 6 provides an example
546
demonstrating the relationship between nicotine and NNN yields, and NNN-to-nicotine yields for the
547
German market products when compliance is categorised using the model 1 approach, and the
548
calculated market median for NNN-to-nicotine emissions of 72 ng/mg (Figure 6A).
549
Figure 6B shows that some compliant products have higher NNN emissions than non-compliant
550
products, which seems counter-intuitive. For example, 80 out of the 169 ‘non-compliant’ products
551
(in terms of NNN-to-nicotine yields) have absolute NNN yields that are lower than that of the
552
‘compliant’ product with the highest NNN yield (171 ng/cig). Conversely 92 out of 169 ‘compliant’
553
products (in terms of NNN-to-nicotine yields) have absolute NNN yields that are higher than that of
AC C
EP
TE D
M AN U
SC
RI PT
530
24
ACCEPTED MANUSCRIPT the ‘non-compliant’ product with the lowest NNN yield (98 ng/cig). This is a point of major concern
555
as Clinical studies have demonstrated that products with higher NNN emissions lead to greater NNN
556
exposure amongst smokers (as measured by biomarkers of exposure) than products with lower NNN
557
emissions [Shepperd et al, 2013], and similarly for NNK [Czoli & Hammond, 2014]. Figure 6C shows
558
that nicotine emission levels make a significant contribution to whether products are compliant to
559
the NNN/nicotine limit, with non-compliant products associated with lower level nicotine emissions,
560
however this contribution is much lower than the magnitude and differences in NNN emissions.
561
Figure 7 shows similar plots to Figure 6B for the remaining 8 toxicants for the German market
562
products, i.e. toxicant yields categorised by compliance to the model 1 market specific calculated
563
toxicant-to-nicotine ceiling. The analysis shows that the volatile species are even more prone to this
564
effect, with significant overlap in per-cigarette emissions between compliant and non-compliant
565
products. A similar full set of plots for the Brazilian, Romanian and Australian market survey data are
566
included in Supplementary Figures 3-5.
567
These observations highlight an important concern over the TobReg models based on use of
568
emission values as a ratio to nicotine levels, as the practical consequences are contrary to TobReg’s
569
stated intention of excluding products with the highest toxicant emissions, as well as having
570
implications for human exposure.
571
572
4.3 Impact of Measurement Error on TobReg proposals
573
All analytical measurements are associated with uncertainty, imprecision or inaccuracy, arising from
574
a combination of analytical error and variation in the manufactured product. Measurement of
575
mainstream smoke toxicant emissions in particular are associated with a significant amount of
576
variation [Hyodo et al., 2006; Morton and Laffoon 2008; Gaworski et al 2011; Intorp et al., 2009;
577
Teillet et al., 2013; Purkis et al., 2014; Eldridge et al. 2015]. By definition, the median values that are
AC C
EP
TE D
M AN U
SC
RI PT
554
25
ACCEPTED MANUSCRIPT pivotal to the TobReg model lie in the middle of the data series, which for normal or near normal
579
distributions (such as that seen for cigarette smoke toxicant data, see Supplementary Figures 1 and 2)
580
is the most densely populated portion of the data distribution. This region is most susceptible to
581
these errors (Figure 8), with consequential uncertainty over the relative positions of individual
582
products in a ranked population. We therefore investigated whether measurement error
583
significantly influences product compliance with TobReg proposed ceilings.
584
585
To assess the impact of measurement error on product compliance, the potential error in smoke
586
toxicant ratio emission measurements, calculated as ±2 times the coefficient of variation (2CV), has
587
been determined when using this single ISO 17025 accredited laboratory in terms of both the
588
repeatability of the measurement method, based on a reference cigarette, and the variation
589
observed when analysing repeatedly manufactured commercial cigarette products [Eldridge et al
590
2015]. Applying the potential measurement error to the NNK-to-nicotine emissions for the German
591
market products demonstrates the practical consequences of these sources of variability (Figure 8):
592
a hypothetical product with a “true” NNK/nicotine value at the measured median limit of 46 ng/mg
593
has an uncertainty of measurement that means that measured values between 31 – 62 ng/mg of
594
actual NNK / nicotine can be obtained from the measurement laboratory. Hence the compliance or
595
non-compliance of this hypothetical product would be a question of chance. This range of possible
596
measurement values is similar to the interquartile range of all NNK-to-nicotine measurements in the
597
German market (34-63 ng NNK/ mg nicotine, Table 1). Therefore, to guarantee compliance with a
598
median limit, and minimise the impact of measurement uncertainty, a true value around the lower
599
quartile value would be the necessary performance target for a brand. In this study only 20%, or 67
600
out of 339, of the German products measured could be reliably determined as compliant for
601
NNK/Nicotine due to the impact of measurement uncertainty, and in the same way only 26%, or 89
602
out of 339 products, could be reliably determined as non-compliant.
AC C
EP
TE D
M AN U
SC
RI PT
578
26
ACCEPTED MANUSCRIPT By contrast, the effect of likely measurement error on toxicant emission ratios when limits are based
604
on 125% of the median was, in general, less pronounced in terms of the number of products
605
affected. The impact of measurement error will depend both on where the 125% of the median limit
606
lies within the distribution of emission data and on the magnitude of the potential error. Using the
607
typical measurement error reported by Eldridge et al. [2015], the best- and worst-case scenarios
608
from this study are shown in Figure 9: the best case scenario being CO/Nicotine in Romania (Figure
609
9A), where the potential measurement error affects ~20 out of the 138 products, and the worst
610
case 1,3-butadiene/nicotine in Romania (Figure 9B), where the potential measurement error
611
encompasses all 138 products measured. The uncertainty in 1,3-butadiene/nicotine product
612
compliance arises due to substantial longitudinal analytical variation previously observed with 1,3-
613
butadiene analysis; over a 10-month period 2CV values of 46% were reported [Eldridge et al. 2015]
614
Cumulatively, this more stringent performance standard introduced by real-world measurement
615
uncertainty would reduce the number of products that could be guaranteed to be compliant to
616
around zero when 9 simultaneous limits are considered, (see supplementary Tables 3 and 4 for the
617
impact on non-compliance rates, including measurement error, for each toxicant in each of the four
618
markets).
619
It might be considered that increasing the product sampling frequency may reduce the impact of
620
measurement uncertainty. However, the effectiveness of this strategy is contingent on
621
measurement variability arising solely or predominately from product rather than analytical sources.
622
In the case where analytical errors are a major contribution to the overall measurement variability
623
then further measurements cannot be guaranteed to reduce measurement error.
624
625
All the present data were acquired in a single laboratory and for one-point-in-time samples, and
626
therefore does not include product variability nor the variability observed between results arising
AC C
EP
TE D
M AN U
SC
RI PT
603
27
ACCEPTED MANUSCRIPT from different laboratories. Significant levels of between laboratory variation have been previously
628
reported — up to 18% CV between 3 laboratories (Hyodo et al 2006) and up to 100% CV (70% CV
629
excluding 1,3-butadiene) between 15 laboratories (Intorp et al 2009), for the 9 toxicants from this
630
study. This suggests that meeting toxicant ratio emission limits through measurements carried out at
631
a number of different laboratories might be even more strongly influenced by measurement
632
uncertainty than measurements in a single laboratory.
633
Effective product regulation is fundamentally dependent on accurate and precise determination of
634
smoke toxicants, as well as the availability of technically feasible technologies for their reduction.
635
Current analytical methods are capable of producing precise single-point-in-time measurements;
636
however, a lack of standardised methods and certified reference materials means that accuracy is
637
compromised, owing to substantial variation over time, within and particularly between laboratories
638
[Morton et al 2008; Intorp et al 2009; Oldham et al 2014, Eldridge et al 2015]. Our study
639
demonstrates that for the majority of products on-sale on a market, compliance with TobReg
640
proposals, would be strongly influenced by random measurement variability; management of
641
product toxicant emissions under these circumstances would be a highly challenging activity.
TE D
M AN U
SC
RI PT
627
EP
642 4.4 Multiple toxicant limits
644
TobReg conducted an assessment of correlations between different toxicant emissions from
645
products in their database (WHO 2008), and found evidence for both positive correlations between
646
toxicants, such as between NNN and NNK, as well as negative correlations such as those between
647
formaldehyde and TSNAs or between benzo[a]pyrene and carbonyls. To further understand the
648
impact of the simultaneous application of multiple limits on numbers of product non-compliances
649
we examined the extent to which the nine toxicants were correlated. This exercise is of value
650
because the current study generated a significantly larger database than that used by TobReg. To
AC C
643
28
ACCEPTED MANUSCRIPT achieve this, Pearson’s correlation coefficients (r) were determined among the toxicant emission
652
yields for each market (Table 5), after excluding the four atypical products discussed in Section 3.2.
653
Toxicant emissions that were predominantly driven by precursors in the tobacco filler blend (NNN,
654
NNK, formaldehyde and B[a]P) showed the poorest levels of correlation both amongst themselves
655
and against the volatile and gaseous toxicant yields. B[a]P correlations varied widely across toxicants
656
and also against the same toxicant across the four markets. Formaldehyde correlations were positive
657
and weak across the majority of toxicant comparisons but negative against both TSNAs. Even yields
658
of the chemically similar NNN and NNK showed varying degrees of correlation, ranging from
659
moderate (r=0.475), for Romanian market products, to very strong (r=0.864) for Australian market
660
products. This analysis demonstrates that toxicant yields are inter-related to varying degrees, with
661
evidence of both positive (e.g. between CO and volatile toxicants) and negative (e.g. between TSNAs
662
and formaldehyde, or between TSNAs and 1,3-butadiene) correlations (Table 5).
663
The generally weak and inconsistent correlations are a concern for compliance with the TobReg
664
proposals, as it results in a greater number of non-compliances when applying limits to each of the
665
nine toxicants simultaneously: for example, the 50% non-compliance rate observed for an individual
666
TSNA rose to approximately 60% for both TSNAs (Figure 3B). Furthermore, as these relationships
667
are not totally consistent across different markets, effective management of toxicant yields by
668
cigarette manufacturers across multiple markets under this regulatory proposal would be a highly
669
complex affair. In particular, the negative correlations suggest that reducing yields of nitrosamines
670
could lead to an increase in yields of formaldehyde and 1,3-butadiene to some degree, as seen when
671
comparing the Australian market data to the other markets examined in this study.
672
673
4.4 Logistical Considerations
AC C
EP
TE D
M AN U
SC
RI PT
651
29
ACCEPTED MANUSCRIPT The present study highlighted the logistical challenges in obtaining samples of all currently available
675
commercial cigarette products for a given market. Commercial cigarettes are “fast-moving consumer
676
goods” and there is continuing evolution in the identity of products in a market place, many of which
677
will have low market share or use by smokers. As a result, in our experience, sourcing all products
678
on-sale in a market is highly challenging, and we were not able to do so in any of the four markets
679
studied. Overall, 80%–97% of the ‘snapshot list’ of commercial products were obtained for the
680
markets in this study. Even in Brazil, where the identity of all products is clear, due to a paid product
681
registration process with the regulatory authority (ANVISA), it was not possible to sample all
682
products. Although it was straightforward to generate a list of products, it took many weeks to
683
source all of the available products in some markets, and in that time a number of products were
684
removed from the market. In addition, for some competition products with a small sales volume, it
685
was not possible to obtain sufficient cigarettes (>400) on the open market for analysis. Because all
686
products contribute equally to the setting of TobReg proposed limits, limits created in this way may
687
reflect a point-in-time picture of a market, rather than a long term stable view. The extent to which
688
market dynamics may vary in overall toxicant profiles over time is unclear and requires further
689
examination.
690
Given the logistical challenges of sampling all products from a market, the most efficient method of
691
collecting all products would be to require the cigarette manufacturers to analyse or provide
692
samples for analysis. However, TobReg have expressed concerns over manufacturers providing
693
sample products for analysis and instead recommend obtaining product directly from the market or
694
through unannounced collection during manufacture or distribution. This latter route may be the
695
only practical approach for sampling a complete market, but also relies upon frequent or continuous
696
production or availability of a product. With small sales volume products this may not be the case.
AC C
EP
TE D
M AN U
SC
RI PT
674
697 698
4.5 Practicality of reducing TSNA toxicant emissions:
30
ACCEPTED MANUSCRIPT TobReg’s basis for using the market median as a limit for TSNAs are: the observation that a
700
comparatively wide range of NNN and NNK emissions are observed for cigarette products around
701
the world; the view that one of the two main styles of tobacco (flue-cured Virginia or “bright”
702
tobacco) is naturally lower in TSNA levels than the other (air-cured Burley); and reports that
703
nitrosamine contents of US and Canadian flue-cured tobaccos were declining due to changes in flue-
704
curing approaches in these countries (IARC 2004, Gray and Boyle 2004). TobReg therefore regarded
705
reducing TSNA levels via the proposed limits as straightforward to achieve by changing the tobacco
706
blend for those with lower toxicant potential.
707
However, use of the median as a toxicant ratio emission limit is intrinsically problematic: no matter
708
how high or low, or widespread the data or its distribution, half of the products on a market will fail
709
to meet a median limit by definition. Therefore, for markets with the lowest toxicant/nicotine
710
distribution, the practical mechanism for compliance with limits is unclear. For example, with
711
Australian products where NNN/nicotine levels were the lowest of the countries in the current
712
survey, substitution by brands from other markets is unlikely to be a practical solution, and
713
compliance with limits may rely upon the availability of a limited pool of low toxicant precursor
714
tobaccos.
715
The negative correlation between TSNA/nicotine and formaldehyde/nicotine suggests that the
716
simple framework of reducing each of the 9 toxicant/nicotine emissions from cigarettes may be
717
more complex than previously considered, with the potential for unintended impact on emissions of
718
other toxicants. The scope of this was examined by comparing the distribution of toxicant/nicotine
719
emissions across the four markets (Supplementary Figure 2). With NNN/nicotine, the data shows
720
significant commonality in values, other than the Australian products, which showed a lower
721
distribution (but still with significant overlap) to the other three markets. In contrast, with
722
formaldehyde/nicotine Australian products (together with Brazilian products) showed wider
723
distribution of values than the other two markets. Consequently, reductions in NNN/nicotine might
AC C
EP
TE D
M AN U
SC
RI PT
699
31
ACCEPTED MANUSCRIPT be achievable in the Romanian and German markets by changing to an Australian product profile but
725
at the cost of increased formaldehyde/nicotine, and consequent potential for failure against this
726
limit. The consumer acceptability of these changes in product styles is also unclear.
727
TobReg expressed the view that compliance to the levels mandated for NNN and NNK will be readily
728
achievable through established blending and curing practices [WHO 2008]; however, the practical
729
mechanisms by which this could be achieved, and their feasibility on national or global scales is
730
unknown. As noted by Gray and Boyle (2004) “the international tobacco trade is complicated, often
731
based on an auction system, and the introduction of nitrosamine assays on unmanufactured tobacco
732
would be difficult. Certainly regulations could enforce a prohibition on manufacture and sale of high
733
nitrosamine tobacco, but both import and export regulation would be required by many countries,
734
and would need to be applied to tobacco as well as tobacco products. While all this is possible in
735
theory, in practice it is likely that the developed countries will protect themselves best and others
736
probably not at all”.
TE D
737
M AN U
SC
RI PT
724
5. CONCLUSION
739
The present study has provided the first full assessment of the practicalities and impact of three
740
WHO TobReg regulatory proposals for the emissions of nine toxicants in mainstream cigarette
741
smoke.
742
Our study sampled products according to TobReg’s proposals, and we measured the emissions of all
743
available products from four countries with different product styles. The TobReg process of
744
sampling and analysing all products on a market was found to be logistically challenging, and our
745
experience suggests it may not be possible to achieve in practise unless products are sampled from
746
manufacturing sites. When the TobReg model of market-specific limits are applied, 70%–80% of
747
products sold in each of the four countries (Australia, Brazil, Germany and Romania) were found to
AC C
EP
738
32
ACCEPTED MANUSCRIPT be non-compliant with the toxicant limits, and thus would require substantial redesign to continue
749
to be sold in the subsequent regulated market. Application of the second TobReg model of fixed
750
limits (both ‘International brands’ or ‘Canadian brands’ set limits) resulted in, often substantial,
751
increases in non-compliance rates both at an individual toxicant level and when applying multiple
752
toxicant limits simultaneously. In some cases, all products on a market failed to meet the TobReg-
753
provided limits. A third, recently proposed model, of setting limits based on three-times the lowest
754
emission levels of products on a market, shows similar challenging results, with 68-99% non-
755
compliances. This latter model is particularly sensitive to the impact of atypical products on a
756
market, and may be challenged by analytical errors and emission levels too low to be quantified.
757
Use of the market median to set toxicant limits is technically challenging to comply with: first, half of
758
the products on the market are automatically non-compliant; and second, when the likely
759
measurement error is taken into account, compliance with the proposed limits is highly influenced
760
by product and analytical variability for a large proportion of products.
761
Simultaneous application of all nine toxicant limits also contributes substantially to the high levels of
762
non-compliance owing to the lack of close positive correlation between many toxicant emission
763
levels. Evidence of negative correlation (between TSNAs and formaldehyde, or between 1,3-
764
butadiene and NNN, NNK or formaldehyde) suggests that reducing yields of some toxicants would
765
lead to an increase in yields of other toxicants. Very different limits for each toxicant were calculated
766
across the four markets, with limits for TSNAs ranging more than threefold. Furthermore, regulating
767
toxicant emission levels as a ratio to nicotine yields increases the complexity of the regulation and
768
allows some products with relatively high toxicant emissions to register as compliant if their nicotine
769
yields are also high.
770
A fundamental finding of our study is that each of the three TobReg proposed toxicant limit models
771
go far beyond TobReg’s aims of reducing toxicant levels in cigarette smoke whilst not eliminating the
772
majority of cigarette brands in their current form from a market.
AC C
EP
TE D
M AN U
SC
RI PT
748
33
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
773
34
ACCEPTED MANUSCRIPT 6. Acknowledgements
775
The staff of British American Tobaccos analytical laboratories for product and smoke analysis. Lucy
776
Evans for editorial support in drafting the manuscript
777
7. Key References
778
ANVISA 2007, Brazil Resolution RDC No. 90 of the Federal Sanitation Agency effective 27 December
779
2007 (re-published), (www.anvisa.gov.br).
780
AOAC. Guidelines for single laboratory validation of chemical methods for dietary supplements and
781
botanicals. AOAC International, 2002.
782
Australian DOH. Australian cigarette emissions data. Australian Government, Department of Health
783
2002.
784
April 2014).
785
Burns DM, Dybing E, Gray N, Hecht S, Anderson C, Sanner T, O’Connor R, Djordjevic M, Dresler C,
786
Hainaut P, Jarvis M, Opperhuizen A, Straif K (2008). Mandated lowering of toxicants in cigarette
787
smoke: a description of the World Health Organization TobReg proposal. Tob. Control 2008 17:132–
788
141. doi:10.1136/tc.2007.024158
789
Counts ME, Morton MJ, Laffoon SW, Cox RH, Lipowicz PJ. Smoke composition and predicting
790
relationships for international commercial cigarettes smoked with three machine-smoking
791
conditions. Regul. Toxicol. Pharmacol. 2005 41(3):185-227.
792
Czoli C., Hammond D. TSNA Exposure: Levels of NNAL among Canadian tobacco users. Nicotine Tob
793
Res (2014) doi: 10.1093/ntr/ntu251 Article in press.
794
Eldridge A, Betson TR, Vinicius Gama M, McAdam K. Variation in tobacco and mainstream smoke
795
toxicant yields from selected commercial cigarette products. Regul. Toxicol. Pharmacol. 71 (2015)
796
409–427
M AN U
SC
RI PT
774
(accessed
AC C
EP
TE D
http://www.health.gov.au/internet/main/publishing.nsf/Content/tobacco-emis
35
ACCEPTED MANUSCRIPT FDA. Harmful and potentially harmful constituents in tobacco products and tobacco smoke:
798
Established list. US Food and Drug Administration, 2012.
799
http://www.fda.gov/TobaccoProducts/GuidanceComplianceRegulatoryInformation/ucm297786.htm
800
Gaworski CL, Wagner K, Morton MJ, Oldham MJ. Insights from a multi-year program designed to test
801
the impact of ingredients on mainstream cigarette smoke toxicity. Inhalation Toxicology, 2011; 23
802
(s1); 172-183 DOI: 10.3109/08958378.2010.546440
RI PT
797
803
Gregg E, Hill C, Hollywood M, Kearney M, McAdam K, Purkis S, McLaughlin D,
805
Williams M. The UK smoke constituents testing study. Summary of
806
results and comparison with other studies. Beit. Tabakforsch. Intl. 2004 21:117–118.
807
Health Canada. Determination of “tar”, nicotine and carbon monoxide in mainstream tobacco
808
smoke-official method. Health Canada, Ottawa, 1999.
809
Health Canada. Tobacco Reporting Regulations SOR /2000-273. .
810
Published 2000. Accessed April 2014.
811
Health Canada. Constituents and emissions reported for cigarettes sold in
812
Canada – 2004. Unpublished data received on request from [email protected].
813
Health Canada. Official method: determination of benzo (a) pyrene in mainstream tobacco smoke.
814
No. T103
815
Health Canada. Official method: determination of TSNAs in mainstream tobacco smoke. No. T111
816
Hyodo T., Inoue O., Katagiri H., Mikita A., Fujiwara M. Long-term inter-laboratory comparisons of
817
selected analytes in 2R4F mainstream smoke. 2006 CORESTA Conference Paper SS7.
AC C
EP
TE D
M AN U
SC
804
818
36
ACCEPTED MANUSCRIPT International Organization for Standardization. ISO 5725-2:1994. Accuracy (trueness and precision)
820
of measurements methods and results – part 2. Basic method for determination of repeatability and
821
reproducibility of a standard measurement method. ISO, Geneva, 1994.
822
International Organization for Standardization. ISO 4387:2000 (E). Cigarettes - Determination of
823
Total and Dry Particulate Matter Using a Routine Analytical Smoking Machine.
824
International Organization for Standardization. ISO 10315:2000 (E). Cigarettes - Determination of
825
Nicotine in Smoke Condensates - Gas Chromatographic Method
826
International Organization for Standardization. ISO 8454:2007 (E). Cigarettes - Determinations of
827
Carbon Monoxide in the Vapour Phase of Smoke - NDIR Method
828
International Organization for Standardization. ISO 3308:2012. Routine analytical cigarette-smoking
829
machine — Definitions and standard conditions. ISO, Geneva, 2012.
830
International Organization for Standardization. ISO 8243:2013. Cigarettes – Sampling. ISO, Geneva,
831
2013.
832
Intorp, M., Purkiss, S., Whittaker, M., Wright, W., Determination of ‘Hoffmann Analytes’ in cigarette
833
mainstream smoke. The CORESTA 2006 joint experiment. Beit. Tabakforsch. Intl. 2009 23:161-202
SC
M AN U
TE D
EP AC C
834
RI PT
819
835
Intorp, M., Purkis, S.W., Wagstaff, W., 2011. Determination of Selected Volatiles in Cigarette
836
Mainstream Smoke. The CORESTA 2009 Collaborative Study and Recommended Method. Beitrage
837
zur Tabakforsch., 24(5), 243-251.
838 839
Morton, M.J., Laffoon, S.W. Cigarette smoke chemistry market maps under Massachusetts
840
Department of Public Health smoking conditions. Regul. Toxicol. Pharmacol. 2008 51:1-30.
37
ACCEPTED MANUSCRIPT Oldham, M.J., Desoi, D.J., Rimmer, L.T., Wagner, K.A., Morton, M.J., Insights from analysis for
842
harmful and potentially harmful constituents (HPHCs) in tobacco products. Regul. Toxicol. Pharmacol.
843
2014 70 (1) 138-48, doi 10.1016/j.yrtph.2014.06.017.
844
Purkis S., Intorp M. Analysis of reference cigarette smoke yield data from 21 laboratories for 28
845
selected analytes as a guide to selection of new CORESTA Recommended Methods; Beit.
846
Tabakforsch. Intl. 2014 26(2): 57-73.
847
Rodgman A, Perfetti TA: The Chemical Components of Tobacco and Tobacco Smoke. CRC Press: New
848
York; 2013.
849
Shepperd C.J., Eldridge A., Camacho O.M., McAdam K., Proctor C. Changes in levels of biomarkers of
850
exposure observed in a controlled study of smokers switched from conventional to reduced toxicant
851
prototype cigarettes. Regul. Toxicol. Pharmacol. 66 (2013) 147–162
852
Taiwan, 2010, http://www.health99.doh.gov.tw/box2/smokefreelife/law.aspx, accessed 14/6/2010
853
Teillet, B., Cahours, X., Verron, T., Colard, S., Purkis, S., Comparison of smoke yield data collected
854
from different laboratories; Beit. Tabakforsch. Intl. 2013 25(8): 662-670.
855 856
World Health Organization. WHO Framework Convention on Tobacco Control. WHO Geneva 2005. ISBN 978 92 4 159101 0; whqlibdoc.who.int/publications/2003/9241591013.pdf
857
World Health Organization. The Scientific Basis of Tobacco Product Regulation. WHO Technical
858
Report Series 951. WHO, Geneva, 2008.
859
World Health Organization. WHO report on the global tobacco epidemic, 2011: warning about the
860
dangers of tobacco. WHO Geneva 2011
861
World Health Organisation. Report on the Scientific Basis of Tobacco Product Regulations: Fifth
862
Report of a WHO Study Group. WHO Technical Report Series 989. WHO, Geneva, 2015.
AC C
EP
TE D
M AN U
SC
RI PT
841
38
ACCEPTED MANUSCRIPT 863
World Health Organization (2016a). Tobacco- Fact Sheet No339, June 2016.
864
http://www.who.int/mediacentre/factsheets/fs339/en/ (accessed 16
865
January 2017). World health Organisation (2016b). http://www.who.int/fctc/signatories_parties/en/ (accessed 16
867
January 2017)
RI PT
866
868
AC C
EP
TE D
M AN U
SC
869
39
ACCEPTED MANUSCRIPT Figure Captions
871 872 873
Figure 1. PCA of global product data. (A) Score plot for PC1 and PC2– crosses indicate individual products in the database. Inset shows the percentage variation accounted for by the first four principal components. (B) Loadings plot for PC1 and PC2.
874 875
Figure 2. Identification of four diverse market scenarios. (A) Brazil; (B) Romania, (C) Australia, and (D) Germany. Red crosses, products from market of interest; grey crosses, remaining market data.
876 877 878
Figure 3. Proportion of non-compliant products by market using market-specific ceilings (model 1). (A) Separate application to toxicants (B) Simultaneous application of toxicant limits (C) Number of non-compliant toxicants per product.
879
Figure 4. Percentage of non-compliant products in each market using TobReg set limits (model 2)
880
A) ‘International brands’ limits B) ‘Canadian brands’ limits.
881
Figure 5: Percentage of non-compliant products in each market using model 3
882 883 884 885
Figure 6. Comparison of German market product data categorised by compliance to TobReg limit (market median of NNN-to-nicotine emissions calculated via model 1): A) NNN-to-nicotine, reference line showing market median B) NNN emissions, reference lines showing highest level for compliant and lowest level for non-compliant products and C) nicotine emissions.
886 887
Figure 7. Toxicant emissions for German market survey products categorised by compliance to TobReg calculated limits (model 1)
888 889 890 891
Figure 8. Rank order of the 339 German market products for NNK-to-nicotine emissions. Effect of the median limit, as proposed by TobReg model 1 [WHO 2008], and the potential error associated with toxicant ratio determination. Considerably fewer products fall within the variability of the analysis.
892 893 894 895 896
Figure 9. Toxicant-to-nicotine ratio emissions in the Romanian market survey. (A) Rank order of products for CO. (B) Rank order of products for 1,3-butadiene. Graphs indicate both the TobReg proposed limit (model 1) based on 125% of the median and the potential error associated with toxicant ratio determination.
AC C
EP
TE D
M AN U
SC
RI PT
870
40
ACCEPTED MANUSCRIPT
Table 1. Market summary of smoke toxicant emissions (HCI) per cigarette and as a ratio to nicotine yield.
Nicotine
mg Brazil Romania Australia Germany
132 138 172 339
1.52 1.80 1.93 1.88
0.25 0.32 0.44 0.33
per cigarette Lower Quartile 0.82 1.39 1.11 1.57 0.00 1.65 1.12 1.65
Tar
mg Brazil Romania
132 138
24.1 25.0
3.6 4.2
13.5 15.2
22.0 22.0
24.1 24.2
26.4 27.8
34.5 33.5
132 138
16.0 13.9
2.5 1.1
12.2 10.8
14.3 13.3
15.2 13.9
17.2 14.6
27.6 17.6
Australia Germany
172 339
27.0 27.8
5.1 4.1
16.7 17.2
23.4 24.6
27.2 27.7
29.7 31.2
61.5 37.5
170 339
a
14.0 15.0
1.6 2.1
9.2 9.3
13.0 13.9
14.0 14.8
15.0 16.0
19.7 24.0
Brazil Romania
132 138
138 150
51 78
34 39
105 104
139 132
163 173
392 496
132 138
94 85
47 45
16 17
70 59
90 74
112 97
479 283
Australia Germany
172 339
68 144
57 66
0 11
30 100
51 134
83 181
338 424
170 339
a
37 79
32 41
7 5
16 55
27 72
45 96
181 294
Brazil Romania
132 138
155 100
96 34
35 37
91 77
122 96
194 120
670 183
132 138
Australia Germany
172 339
54 95
36 47
0 13
27 61
40 88
75 118
178 318
170 339
Brazil Romania
132 138
18.7 13.1
4.8 3.9
7.8 5.7
15.3 10.4
18.2 12.7
21.8 15.1
38.3 27.2
132 138
Australia Germany
172 339
18.0 19.9
4.5 5.1
9.0 9.4
14.7 16.1
17.8 19.1
20.4 22.9
40.9 36.4
170 339
Brazil Romania
132 138
92 89
19 21
48 32
78 76
91 89
102 103
143 154
132 138
Australia Germany
172 339
119 99
32 27
60 42
97 82
112 97
138 114
Brazil Romania
132 138
1236 1167
163 226
593 626
1150 972
1263 1188
1343 1316
Australia Germany
172 339
1288 1374
122 191
932 854
1223 1264
1296 1367
1362 1509
Brazil Romania
132 138
129 117
17 23
59 61
122 99
130 116
141 136
ng
Formaldehyde µg
Acetaldehyde µg
Acrolein
µg
1,3-Butadiene µg
Benzene
CO
µg
Min
1.51 1.75 1.92 1.85
Upper Quartile 1.66 2.04 2.24 2.09
2.12 2.55 3.58 3.21
Median
Australia Germany
172 339
136 146
17 21
78 81
126 132
136 146
Brazil Romania
132 138
71 99
15 20
34 60
61 84
68 102
Australia Germany
172 339
107 104
13 24
67 54
Brazil Romania
132 138
85 75
14 18
37 42
Australia Germany
172 339
88 94
11 17
56 55
mg Brazil Romania
132 138
23.1 23.4
3.5 5.2
11.1 11.9
Australia Germany
172 339
25.8 25.8
3.0 4.9
18.5 13.9
Max
N
Mean
Std Dev
a
a
per mg nicotine Lower Min Quartile
Median
Upper Quartile
Max
RI PT
Std Dev
SC
Mean
M AN U
B[a ]P
ng
N
109 57
93 20
17 18
61 42
78 55
124 69
818 110
29 53
20 31
7 6
14 34
21 46
42 63
125 221
12.5 7.2
3.7 1.5
6.5 4.0
10.1 6.2
11.5 7.2
13.7 8.1
28.4 13.5
9.3 10.8
1.2 2.9
6.4 5.4
8.4 8.8
9.0 10.4
10.0 12.0
13.1 23.0
61 51
13 16
35 26
52 40
59 49
69 56
107 101
TE D
NNK
ng
Market
253 204
170 339
a
62 54
14 16
30 17
51 42
61 52
71 63
104 126
1563 1803
132 138
829 655
168 104
495 394
732 587
801 658
907 721
1847 915
1629 2567
170 339
a
683 749
135 151
328 385
592 666
671 739
767 829
1239 1762
172 169
132 138
86 66
16 11
50 42
77 58
85 66
94 74
170 100
147 158
187 295
a
170 339
72 80
15 17
43 42
62 70
69 78
80 87
147 180
81 113
116 145
132 138
47 55
12 10
27 36
39 49
44 55
54 61
97 92
a
EP
NNN
Unit
AC C
Toxicant
98 94
107 104
116 114
134 385
170 339
57 57
11 16
35 31
48 49
54 55
64 62
91 265
79 62
85 77
93 85
118 116
132 138
56 42
11 7
31 28
50 36
55 42
62 46
93 62
a
81 84
89 95
95 104
135 262
170 339
46 51
8 12
31 30
41 44
45 50
52 56
74 180
21.7 19.2
23.3 23.4
25.3 27.1
30.1 39.1
132 138
15.5 13.0
3.2 2.1
9.4 7.9
13.6 12.0
15.2 13.0
17.1 14.5
30.2 18.1
35.4 72.1
a
13.6 14.1
2.4 3.6
6.9 6.4
11.9 12.2
13.4 13.9
15.2 15.7
20.4 49.5
24.1 22.9
25.6 25.6
27.9 29.0
170 339
two herbal products with zero nicotine emissions excluded
b
minimum nicotine value for Australian tobacco products was 1.14 mg/cigarette
c
minimum NNN value for Australian tobacco products was 13.6 ng/cigarette
d
minimum NNK value for Australian tobacco products was 17.1 ng/cigarette
AC C
EP
TE D
M AN U
SC
a
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
Table 2. Market-specific toxicant emissions limits calculated as per the TobReg model 1, compared to set limits from model 2.
RI PT
Germany US Blended 924 97 63 13 69 17.3 65 46 72
SC
µg µg µg ng µg mg µg ng ng
M AN U
Acetaldehyde Acrolein Benzene B[a]P 1,3-Butadiene CO Formaldehyde NNK NNN
Market specific limits Brazil Romania Australia Mixed (low charcoal) Mixed (High charcoal) Flue-Cured 1001 822 842 106 82 87 69 53 57 14 9 11 54 69 68 19.0 16.2 16.7 74 61 76 78 55 21 90 74 27
AC C
EP
*TobReg's set limits identified by TobReg as part of model 2.
TE D
Toxicant (per mg nicotine)
TobReg set limits* International Canadian Brands Brands 860 670 83 97 48 50 11 11 67 53 18.4 15.4 47 97 72 47 114 27
ACCEPTED MANUSCRIPT
RI PT
Table 3: Limits calculated according to model 3
Per cigarette
Brazil
Romania
Australia
Germany
Highest/lowest Limit ratio
Acetaldehyde Acrolein Benzene B[a]P 1,3-Butadiene CO Formaldehyde NNK NNN
µg µg µg ng µg mg µg ng ng
1780 178 110 23.3 103 33.3 145 103.8 101.4
1878 184 127 17.0 179 35.6 95.2 110.1 118.1
2797 234 169 26.9 202 55.5 181 51.2 40.8
2561 243 164 28.2 163 41.8 124 37.9 32.3
1.6 1.4 1.5 1.6 2.0 1.7 1.9 2.7 3.5
AC C
EP
TE D
M AN U
SC
Toxicant
ACCEPTED MANUSCRIPT
TobReg model 1, Market Specific limits
72
93
Brazil
77
100
Germany
79
89
Romania
78
71
AC C
EP
TE D
Australia
TobReg model 3, 3x lowest value
“Canadian Brands”
SC
“International Brands”
M AN U
Country
TobReg model 2, fixed limits
RI PT
Table 4: % non-compliance rates for cigarette brands, by market, for each of the different regulatory models
80
68
100
86
97
99
98
80
ACCEPTED MANUSCRIPT Table 5. Correlations among absolute toxicant emission yields: matrix of Pearson correlation coefficients (r). Strong correlations (>│0.6│) are shown in bold. Negative correlations are highlighted in red. Tar
Acetaldehyde
Acetaldehyde Acetaldehyde Acetaldehyde Acetaldehyde
Brazil Romania Australia Germany
0.326 0.631 0.305 0.284
0.658 0.800 0.510 0.644
Acrolein Acrolein Acrolein Acrolein
Brazil Romania Australia Germany
0.369 0.570 0.332 0.180
0.647 0.714 0.385 0.492
0.758 0.929 0.583 0.777
Benzene Benzene Benzene Benzene
Brazil Romania Australia Germany
0.444 0.696 0.584 0.395
0.501 0.815 0.627 0.536
0.696 0.868 0.645 0.696
B[a ]P B[a ]P B[a ]P B[a ]P
Brazil Romania Australia Germany
0.280 0.738 0.830 0.373
0.504 0.823 0.659 0.505
1,3-Butadiene 1,3-Butadiene 1,3-Butadiene 1,3-Butadiene
Brazil Romania Australia Germany
0.300 0.646 0.398 0.398
0.269 0.732 0.377 0.448
CO CO CO CO
Brazil Romania Australia Germany
0.307 0.689 0.460 0.244
0.731 0.865 0.672 0.634
Formaldehyde Formaldehyde Formaldehyde Formaldehyde
Brazil Romania Australia Germany
0.338 0.008 0.502 0.193
NNK NNK NNK NNK
Brazil Romania Australia Germany
-0.331 0.145 -0.037 -0.113
NNN NNN NNN NNN
Brazil Romania Australia Germany
Toxicant Benzene B[a ]P
1,3-Butadiene
CO
Formaldehyde
0.571 0.799 0.565 0.493 0.336 0.686 0.400 0.158
0.544 0.813 0.629 0.334
0.403 0.835 0.570 0.604
0.374 0.740 0.390 0.528
0.837 0.936 0.539 0.677
0.340 0.784 0.337 0.121
0.885 0.925 0.809 0.839
0.656 0.796 0.468 0.675
0.731 0.890 0.715 0.774
0.566 0.793 0.402 0.203
0.460 0.847 0.561 0.598
0.480 0.187 0.517 0.250
0.140 0.167 0.162 0.138
0.524 0.299 0.463 0.409
0.055 0.040 0.428 0.129
0.018 0.044 0.487 0.111
0.019 -0.094 0.230 0.197
0.071 0.064 0.113 0.123
0.325 0.367 0.203 0.337
0.377 0.453 0.066 0.303
0.204 0.425 -0.201 0.060
0.009 0.304 0.019 0.061
0.469 0.267 -0.050 0.251
-0.120 0.235 -0.320 -0.024
0.377 0.469 0.244 0.311
-0.014 0.093 -0.346 -0.340
0.212 0.175 0.146 0.192
0.104 0.270 0.004 0.181
0.176 0.304 -0.210 -0.050
-0.184 0.224 -0.002 0.020
0.127 0.060 -0.096 0.088
-0.270 0.132 -0.270 -0.060
0.078 0.255 0.233 0.169
-0.035 -0.114 -0.388 -0.467
TE D
0.524 0.793 0.331 0.265
EP
AC C
-0.089 0.148 -0.043 -0.123
Acrolein
NNK
RI PT
Nicotine 0.592 0.880 0.798 0.624
SC
Market Brazil Romania Australia Germany
M AN U
Toxicant Tar Tar Tar Tar
0.593 0.475 0.864 0.749
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
Nicotine
mg
Brazil Romania Australia Germany
132 138 172 339
1.52 1.80 1.93 1.88
0.25 0.32 0.44 0.33
per cigarette Lower Min Quartile 0.82 1.39 1.11 1.57 0.00b 1.65 1.12 1.65
Tar
mg
Brazil Romania Australia Germany
132 138 172 339
24.1 25.0 27.0 27.8
3.6 4.2 5.1 4.1
13.5 15.2 16.7 17.2
22.0 22.0 23.4 24.6
24.1 24.2 27.2 27.7
26.4 27.8 29.7 31.2
34.5 33.5 61.5 37.5
NNN
ng
Brazil Romania Australia Germany
132 138 172 339
138 150 68 144
51 78 57 66
34 39 0c 11
105 104 30 100
139 132 51 134
163 173 83 181
392 496 338 424
NNK
ng
Brazil Romania Australia Germany
132 138 172 339
155 100 54 95
96 34 36 47
35 37 0d 13
91 77 27 61
122 96 40 88
194 120 75 118
B[a ]P
ng
Brazil Romania Australia Germany
132 138 172 339
18.7 13.1 18.0 19.9
4.8 3.9 4.5 5.1
7.8 5.7 9.0 9.4
15.3 10.4 14.7 16.1
18.2 12.7 17.8 19.1
Formaldehyde
µg
Brazil Romania Australia Germany
132 138 172 339
92 89 119 99
19 21 32 27
48 32 60 42
78 76 97 82
Acetaldehyde
µg
Brazil Romania Australia Germany
132 138 172 339
1236 1167 1288 1374
163 226 122 191
593 626 932 854
Acrolein
µg
Brazil Romania Australia Germany
132 138 172 339
129 117 136 146
17 23 17 21
1,3-Butadiene
µg
Brazil Romania Australia Germany
132 138 172 339
71 99 107 104
15 20 13 24
1.51 1.75 1.92 1.85
2.12 2.55 3.58 3.21
Median
Max
N
Mean
RI PT
Std Dev
Std Dev
per mg nicotine Lower Min Quartile
Median
Upper Quartile
Max
132 138 170a 339
16.0 13.9 14.0 15.0
2.5 1.1 1.6 2.1
12.2 10.8 9.2 9.3
14.3 13.3 13.0 13.9
15.2 13.9 14.0 14.8
17.2 14.6 15.0 16.0
27.6 17.6 19.7 24.0
132 138 170a 339
94 85 37 79
47 45 32 41
16 17 7 5
70 59 16 55
90 74 27 72
112 97 45 96
479 283 181 294
670 183 178 318
132 138 170a 339
109 57 29 53
93 20 20 31
17 18 7 6
61 42 14 34
78 55 21 46
124 69 42 63
818 110 125 221
21.8 15.1 20.4 22.9
38.3 27.2 40.9 36.4
132 138 170a 339
12.5 7.2 9.3 10.8
3.7 1.5 1.2 2.9
6.5 4.0 6.4 5.4
10.1 6.2 8.4 8.8
11.5 7.2 9.0 10.4
13.7 8.1 10.0 12.0
28.4 13.5 13.1 23.0
91 89 112 97
102 103 138 114
143 154 253 204
132 138 170a 339
61 51 62 54
13 16 14 16
35 26 30 17
52 40 51 42
59 49 61 52
69 56 71 63
107 101 104 126
1150 972 1223 1264
1263 1188 1296 1367
1343 1316 1362 1509
1563 1803 1629 2567
132 138 170a 339
829 655 683 749
168 104 135 151
495 394 328 385
732 587 592 666
801 658 671 739
907 721 767 829
1847 915 1239 1762
59 61 78 81
122 99 126 132
130 116 136 146
141 136 147 158
172 169 187 295
132 138 170a 339
86 66 72 80
16 11 15 17
50 42 43 42
77 58 62 70
85 66 69 78
94 74 80 87
170 100 147 180
34 60 67 54
61 84 98 94
68 102 107 104
81 113 116 114
116 145 134 385
132 138 170a 339
47 55 57 57
12 10 11 16
27 36 35 31
39 49 48 49
44 55 54 55
54 61 64 62
97 92 91 265
SC
Mean
M AN U
N
Upper Quartile 1.66 2.04 2.24 2.09
TE D
Market
EP
Unit
AC C
Toxicant
ACCEPTED MANUSCRIPT
µg
Brazil Romania Australia Germany
132 138 172 339
85 75 88 94
14 18 11 17
37 42 56 55
79 62 81 84
85 77 89 95
93 85 95 104
118 116 135 262
132 138 170a 339
56 42 46 51
11 7 8 12
31 28 31 30
50 36 41 44
55 42 45 50
62 46 52 56
93 62 74 180
CO
mg
Brazil Romania Australia Germany
132 138 172 339
23.1 23.4 25.8 25.8
3.5 5.2 3.0 4.9
11.1 11.9 18.5 13.9
21.7 19.2 24.1 22.9
23.3 23.4 25.6 25.6
25.3 27.1 27.9 29.0
30.1 39.1 35.4 72.1
132 138 170a 339
15.5 13.0 13.6 14.1
3.2 2.1 2.4 3.6
9.4 7.9 6.9 6.4
13.6 12.0 11.9 12.2
15.2 13.0 13.4 13.9
17.1 14.5 15.2 15.7
30.2 18.1 20.4 49.5
a
two herbal products with zero nicotine emissions excluded minimum nicotine value for Australian tobacco products was 1.14 mg/cigarette c minimum NNN value for Australian tobacco products was 13.6 ng/cigarette d minimum NNK value for Australian tobacco products was 17.1 ng/cigarette
AC C
EP
TE D
M AN U
SC
b
RI PT
Benzene
µg µg µg ng µg mg µg ng ng
Brazil Mixed (low charcoal) 1001 106 69 14 54 19.0 74 78 90
Romania Mixed (High charcoal) 822 82 53 9 69 16.2 61 55 74
Australia Flue-Cured 842 87 57 11 68 16.7 76 21 27
Germany US Blended 924 97 63 13 69 17.3 65 46 72
EP
TE D
M AN U
SC
*TobReg's set limits identified by TobReg as part of model 2
AC C
Acetaldehyde Acrolein Benzene B[a]P 1,3-Butadiene CO Formaldehyde NNK NNN
ACCEPTED MANUSCRIPT Market specific limits
TobReg set limits* International Canadian Brands Brands 860 670 83 97 48 50 11 11 67 53 18.4 15.4 47 97 72 47 114 27
RI PT
Toxicant (per mg nicotine)
ACCEPTED MANUSCRIPT
Toxicant Nicotine 0.592 0.880 0.798 0.624
Acetaldehyde
Acrolein
Benzene
1,3-Butadiene
CO
Formaldehyde
Acetaldehyde Acetaldehyde Acetaldehyde Acetaldehyde
Brazil Romania Australia Germany
0.326 0.631 0.305 0.284
0.658 0.800 0.510 0.644
Acrolein Acrolein Acrolein Acrolein
Brazil Romania Australia Germany
0.369 0.570 0.332 0.180
0.647 0.714 0.385 0.492
0.758 0.929 0.583 0.777
Benzene Benzene Benzene Benzene
Brazil Romania Australia Germany
0.444 0.696 0.584 0.395
0.501 0.815 0.627 0.536
0.696 0.868 0.645 0.696
0.571 0.799 0.565 0.493
B[a ]P B[a ]P B[a ]P B[a ]P
Brazil Romania Australia Germany
0.280 0.738 0.830 0.373
0.504 0.823 0.659 0.505
0.524 0.793 0.331 0.265
0.336 0.686 0.400 0.158
0.544 0.813 0.629 0.334
1,3-Butadiene 1,3-Butadiene 1,3-Butadiene 1,3-Butadiene
Brazil Romania Australia Germany
0.300 0.646 0.398 0.398
0.269 0.732 0.377 0.448
0.403 0.835 0.570 0.604
0.374 0.740 0.390 0.528
0.837 0.936 0.539 0.677
0.340 0.784 0.337 0.121
CO CO CO CO
Brazil Romania Australia Germany
0.307 0.689 0.460 0.244
0.731 0.865 0.672 0.634
0.885 0.925 0.809 0.839
0.656 0.796 0.468 0.675
0.731 0.890 0.715 0.774
0.566 0.793 0.402 0.203
0.460 0.847 0.561 0.598
Formaldehyde Formaldehyde Formaldehyde Formaldehyde
Brazil Romania Australia Germany
0.338 0.008 0.502 0.193
0.480 0.187 0.517 0.250
0.140 0.167 0.162 0.138
0.524 0.299 0.463 0.409
0.055 0.040 0.428 0.129
0.018 0.044 0.487 0.111
0.019 -0.094 0.230 0.197
0.071 0.064 0.113 0.123
NNK
Brazil
-0.331
0.325
0.377
0.204
0.009
0.469
-0.120
0.377
EP
TE D
M AN U
SC
RI PT
Market Brazil Romania Australia Germany
AC C
Tar
B[a ]P
Toxicant Tar Tar Tar Tar
-0.014
NNK
ACCEPTED MANUSCRIPT
Romania Australia Germany
0.145 -0.037 -0.113
0.367 0.203 0.337
0.453 0.066 0.303
0.425 -0.201 0.060
0.304 0.019 0.061
0.267 -0.050 0.251
0.235 -0.320 -0.024
0.469 0.244 0.311
0.093 -0.346 -0.340
NNN NNN NNN NNN
Brazil Romania Australia Germany
-0.089 0.148 -0.043 -0.123
0.212 0.175 0.146 0.192
0.104 0.270 0.004 0.181
0.176 0.304 -0.210 -0.050
-0.184 0.224 -0.002 0.020
0.127 0.060 -0.096 0.088
-0.270 0.132 -0.270 -0.060
0.078 0.255 0.233 0.169
-0.035 -0.114 -0.388 -0.467
AC C
EP
TE D
M AN U
SC
RI PT
NNK NNK NNK
0.593 0.475 0.864 0.749
ACCEPTED MANUSCRIPT Title: Impact assessment of WHO TobReg proposals for mandated lowering of selected mainstream cigarette smoke toxicants Figure 1:
B)
AC C
EP
TE D
M AN U
SC
RI PT
A)
ACCEPTED MANUSCRIPT Figure 2: A) Brazil:
M AN U
SC
RI PT
B) Romania:
C) Australia:
AC C
EP
TE D
D) Germany:
ACCEPTED MANUSCRIPT Figure 3:
90 80 70
RI PT
60 50 40 30 20 10 0
B)
M AN U
SC
Non-Compliant Product (%)
A) 100
100
Non-Compliant Product (%)
90 80 70 60
Brazil
50
Romania
40
TE D
Australia
30 20 10 0
C)
100
EP
All 9 Toxicants TSNAs only
Germany
Non-TSNAs only
Non-Compliant Product (%)
AC C
90 80 70 60 50 40 30 20 10 0
1
2
3
4
5
6
7
No. non-compliant toxicants
8
9
ACCEPTED MANUSCRIPT Figure 4: A) 100 80 70 60 Brazil
50
RI PT
Non-Compliant Product (%)
90
40
Romania
30
Australia
20
Germany
10
M AN U
SC
0
B) 100 80
50 40 30 20 10 0
EP
60
TE D
70
AC C
Non-Compliant Product (%)
90
Brazil Romania Australia Germany
ACCEPTED MANUSCRIPT Figure 5:
100 90 80
RI PT
70 60 50 40 30
SC
20 10
AC C
EP
TE D
M AN U
0
Australia Brazil Germany Romania
ACCEPTED MANUSCRIPT Figure 6:
A)
B) 450
300
C)
3.5
400 250 3.0
RI PT
200
100
Nicotine (mg/cig)
150
250 200 171 150
72
98
50 0 Non-compliant
Compliant
EP
TE D
Compliant
0
M AN U
100 50
2.5
2.0
SC
NNN (ng/cig)
300
AC C
NNN / Nicotine (ng/mg)
350
Non-compliant
1.5
1.0 Compliant
Non-compliant
ACCEPTED MANUSCRIPT Figure 7: Acetaldehyde
Acrolein
Benzene 240
1600
Benzene µg/cig
2400
Acrolein µg/cig
Acetaldehyde µg/cig
300
200
160
80
100 800 NC
C
B[a]P
NC
C
1,3-Butadiene
CO
270
150
20
10 C
NC
C
NC
NNK 300
150
60 0 C
NC
M AN U
180
NNK ng/cig
Formaldehyde µg/cig
Formaldehyde
120
C
EP
TE D
C – compliant product; NC – non-compliant product
AC C
40
NC
C
SC
20
60
CO mg/cig
1,3-Butadiene µg/cig
B[a]P µg/cig
390
30
NC
RI PT
C
NC
ACCEPTED MANUSCRIPT Figure 8.:
RI PT
200
150
50 Method Variability
0
50
100
M AN U
Product & Method Variability
0
150
200
Product Rank
TobReg Ceiling Method variability (±2CV reference cigarette)
EP
TE D
Product + Method variability (±2CV commercial cigarette)
62 55 Median 37 31
SC
100
AC C
Germany: NNK/Nicotine (ng/mg)
250
250
300
350
ACCEPTED MANUSCRIPT Figure 9: B)
A)
110
108
18 17.6 17.3
14
12
10
100
90 80
RI PT
15.1 14.9
1,3-Butadiene/N icotine
125% Median
16
70 60 50 40 30 20
40
60
80
100
120
0
140
20
40
60
80
Product Rank
Product Rank
M AN U
TobReg Ceiling Method variability (±2CV reference cigarette) Product + Method variability (±2CV commercial cigarette)
TE D
20
EP
0
SC
8
AC C
CO /N icotine
100
100
120
125% median
37 29
140
ACCEPTED MANUSCRIPT
RI PT
Highlights WHO TobReg proposals to reduce levels of 9 cigarette smoke toxicants were tested.
•
All available (80-97%) cigarette products from 4 diverse countries were analysed.
•
70-100% of cigarette products failed to meet the proposed WHO regulatory models.
•
These proposals would have greater impact on global markets than WHO’s stated aims.
AC C
EP
TE D
M AN U
SC
•
S0273-2300(17)30053-3
DOI:
10.1016/j.yrtph.2017.02.022
Reference:
YRTPH 3782
To appear in:
Regulatory Toxicology and Pharmacology
Received Date: 5 October 2016 Revised Date:
23 January 2017
Accepted Date: 27 February 2017
Please cite this article as: Eldridge, A.C., McAdam, K.G., Betson, T.R., Gama, M.V., Proctor, C.J., Impact assessment of WHO TobReg proposals for mandated lowering of selected mainstream cigarette smoke toxicants, Regulatory Toxicology and Pharmacology (2017), doi: 10.1016/j.yrtph.2017.02.022. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT 1
Impact assessment of WHO TobReg proposals for mandated lowering of selected mainstream
2
cigarette smoke toxicants
3
Authors
4
Christopher J. Proctor
Alison C. Eldridge, Kevin G. McAdam, Tatiana R. Betson, Marcos V. Gama, and
AC C
EP
TE D
M AN U
SC
RI PT
5
1
ACCEPTED MANUSCRIPT 6
Abstract
7
The WHO Tobacco Product Regulation Study Group (TobReg) has proposed three regulatory
8
models for cigarettes, each creating mandatory limits for emissions of nine smoke toxicants. One
9
approach proposes country-specific limits, using median or 1.25x median toxicant/nicotine emission ratios. A second model provides fixed toxicant-ratio limits. The third model limits were
11
three times the lowest toxicant emission on a market. Currently, the practical implications of
12
these models are largely unknown.
13
An impact assessment was conducted using cigarette data from 79 countries to identify four
14
diverse test markets. We sampled all products from each market but limited product availability
15
led to incomplete (80-97%) sourcing. Analysis showed that the country-specific model led to
16
diverse (up to threefold) toxicant limits across the four markets. 70%–80% of products were
17
non-compliant, rising to 100% in some countries with the second and the third models. With
18
each regulatory model the main drivers of non-compliance were the tobacco-specific
19
nitrosamines, the simultaneous application of limits for nine poorly correlated smoke toxicants,
20
and analytical variability. Use of nicotine ratios led to compliance of some high toxicant emission
21
products due to high nicotine emissions.
22
Our findings suggest that these proposals would have greater impact on global markets than
23
TobReg’s stated aims.
25
SC
M AN U
TE D
EP
AC C
24
RI PT
10
2
ACCEPTED MANUSCRIPT
28
WHO – World Health Organisation
29
TobReg – WHO Tobacco Product Regulation Study Group
30
HPHC – Harmful or Potentially Harmful Compounds
31
ISO – International Organization for Standardization
32
FCTC – Framework Convention on Tobacco Control
33
NNN – N’-nitrosonornicotine
34
NNK – 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone
35
B[a]P – Benzo[a]pyrene
36
PCA - Principal components analysis
37
HCI – Health Canada Intense
38
CO – Carbon monoxide
39
BAT – British American Tobacco
40
ANVISA – Brazilian Medical Device Regulations
41
TSNAs – Tobacco Specific Nitrosamines
SC
Abbreviations
AC C
EP
TE D
M AN U
27
RI PT
26
42
3
ACCEPTED MANUSCRIPT 43
1. INTRODUCTION
44 The mortality and morbidity associated with cigarette use is a hazard currently facing over a billion
46
smokers (WHO 2016a). Health risks have been found to diminish after cessation and disease
47
progression may slow substantially (US Department of Health and Human Services, 2014). However,
48
despite significant global Public Health efforts towards smoking cessation, WHO projections suggest
49
that if current trends continue annual deaths from smoking are likely to increase by 2030 (WHO
50
2011).
SC
RI PT
45
51
Since the 1950s scientists have sought to explain the risks associated with cigarette smoking by
53
identifying and quantifying compounds in tobacco and smoke that have toxic properties. These
54
compounds are often referred to as toxicants and several of them have been identified, the number
55
of which has grown longer as toxicological understanding and analytical techniques have improved.
56
The current reference point for cigarette smoke toxicants is the established list of over 90 Harmful or
57
Potentially Harmful Compounds (HPHC) identified by a Technical Advisory group to the FDA in 2011
58
(FDA, 2011). The US Institute of Medicine has expressed the view that some of the harm caused by
59
tobacco use may potentially be reduced through introduction of products that might result in
60
substantial reduction in exposure to one or more tobacco toxicants (Stratton et al., 2001).
TE D
EP
AC C
61
M AN U
52
62
Cigarette smoke toxicants have been the focus of increased regulatory interest since this time (Liu et
63
al. 2013). Traditional approaches for the regulation of cigarette products, based upon reporting and
64
limiting cigarette emissions of tar, nicotine and carbon monoxide measured under the ISO smoking
65
regime have been replaced or supplemented by reporting requirements on cigarette smoke
66
toxicants and at additional smoking regimes. Starting with the mandated measurement and
67
reporting of toxicant emissions in Canada (Health Canada 2000) and Brazil (ANVISA 2007), the
4
ACCEPTED MANUSCRIPT requirement to measure and report emissions has gradually spread to other countries such as
69
Taiwan (Taiwan 2010), and the USA (FDA 2012). Cigarette emission regulations currently show a
70
great deal of inconsistency from country to country, ranging from ceilings on tar, nicotine and
71
carbon monoxide in the European Union and other jurisdictions to detailed, but differing,
72
requirements for annual per-product emission reporting of 7 constituents in Taiwan, 18 constituents
73
in the USA, and more than 40 in Brazil, Canada and Venezuela (Liu et al, 2013).
RI PT
68
74
In 2003 the World Health Organization (WHO) adopted the Framework Convention on Tobacco
76
Control (FCTC), to which 168 countries globally are signatories and 180 are parties (WHO 2016b).
77
The FCTC has an objective of “providing a framework for tobacco control measures to be
78
implemented by the Parties at the national, regional and international levels in order to reduce
79
continually and substantially the prevalence of tobacco use and exposure to tobacco smoke”. FCTC
80
represents a mechanism for global deployment of tobacco control initiatives (WHO 2005).
M AN U
SC
75
TE D
81
Articles 9 and 10 of the FCTC focus on tobacco product regulation (WHO 2005). As a step towards
83
this, in 2008, a WHO advisory group on Tobacco Product Regulation (TobReg) published a new
84
strategy for tobacco product regulation (WHO 2008, Burns et al 2008). The proposed strategy
85
focused on use of standardised measures of cigarette smoke toxicity that characterised, as far as
86
possible, potential differences in harm caused by different cigarettes. Central to TobReg’s strategy is
87
the mandated lowering of nine separate toxicants in cigarette smoke emissions. The toxicants
88
identified for mandated reduction were chosen based on toxicity, observations of differences in
89
yields across brands, availability of technology or other approaches to reduce yields, and the
90
existence of markets for low yield products. The proposed approach was therefore viewed as an
91
example of the precautionary principle often deployed in public health with parallels drawn to
92
strategies used with other consumer products, where the focus is also to reduce levels of known
93
toxicants present in the product.
AC C
EP
82
5
ACCEPTED MANUSCRIPT The nine selected toxicants: 1,3-butadiene, acetaldehyde, acrolein, benzene, benzo[a]pyrene (B[a]P),
95
carbon monoxide, formaldehyde, 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK) and N’-
96
nitrosonornicotine (NNN), represent a range of chemical classes found in cigarette smoke, covering
97
both the vapour phase and the particulate phase, and were viewed as implicated in carcinogenicity,
98
cardiovascular and pulmonary toxicity; they also commonly appear in various other existing lists of
99
identified smoke toxicants. As TobReg’s proposals were founded upon the widely accepted
RI PT
94
regulatory practice of reducing toxicants in products intended for human use TobReg did not regard
101
it as necessary to have specific proof of a link between a lower level of these toxicants in cigarette
102
smoke and a lower level of human disease. This is an important point, as links between individual
103
cigarette smoke toxicants and human disease are currently incomplete (Burns et al., 2008).
M AN U
SC
100
104
The 2008 TobReg proposals recommended two models, both of which established emission limits for
106
each of the 9 toxicants, as measured using the Health Canada Intense smoking regime, and
107
expressed per milligram of nicotine. Use of the ratio to nicotine was proposed on the basis that
108
machine-measured per-cigarette smoke emissions were unreliable estimates of smoker’s exposure
109
to cigarette smoke. TobReg noted that individual smokers “seek to achieve nicotine intakes
110
sufficient to satisfy their addiction”. Use of a toxicant/nicotine ratio was therefore described as
111
shifting focus away from quantity of smoke generated per cigarette and preventing interpretation
112
that the values obtained represent a measure of smoker exposure to toxicants (WHO 2008).
EP
AC C
113
TE D
105
114
Under the TobReg proposals failure of a cigarette brand to meet each of the nine separate limits
115
would lead to its withdrawal from sale on a market. The strategy was recommended to be
116
implemented in phases, starting with annual reporting of toxicant levels for 2-3 years, setting
117
toxicant limits from these data, enforcing limits two years after they are set, and potentially lowering
118
limits progressively over time. The aim in setting the limits was to balance the need to regulate a
6
ACCEPTED MANUSCRIPT 119
range of toxicants, to mandate lowering those toxicants to the greatest extent and yet not to
120
eliminate most brands from the market in their current form (WHO 2008).
121 Under the first TobReg model (“model 1”) the toxicant emission limits for a country were to be set
123
by measuring emission ratios to nicotine for all products on a market, calculating the median value
124
for each toxicant, setting this value as the limit for NNN and NNK, and setting the limits for the other
125
seven toxicants as 125% of their respective median values (Burns et al., 2008).
RI PT
122
SC
126
The second TobReg model (“model 2”) was proposed for countries possessing limited laboratory
128
capacity. This model provided two sets of fixed regulatory limits that could be deployed across
129
multiple countries. One set of limits, labelled ‘international brands’, was calculated from an
130
International measurement survey of 49 Philip Morris cigarette brands, conducted in 2001, (Counts
131
et al 2005) and were proposed for use in countries whose cigarette products were predominately
132
US- or American blend products, or where the style of cigarettes on the market are not readily
133
identifiable. The second set of limits, labelled ‘Canadian brands’, were calculated from a set of 91
134
Canadian brands reported to Health Canada in 2004, reduced to exclude brands with levels of NNN
135
per mg nicotine > 0.1 ng, which eliminates most US and Gauloise brands (12 brands), and duplicate
136
and erroneous values (31 brands), leaving a database of 48 Canadian products representing
137
“unblended cigarettes containing predominantly flue-cured (bright) tobacco”; these limits were
138
intended for use in countries whose cigarette products reflected more flue-cured or “Virginia” blend
139
style products (WHO 2008).
TE D
EP
AC C
140
M AN U
127
141
TobReg conducted a limited impact assessment of the proposed models using these datasets, and
142
concluded that regulation of brands for toxicants other than NNK and NNN would result in 40-41% of
143
brands failing to meet the set of limits without modification (WHO 2008). This form of analysis was
144
based on the belief that there is existing technology for dramatically lowering the nitrosamine 7
ACCEPTED MANUSCRIPT contents of tobaccos. This position arose from analysis showing wide differences in nitrosamine
146
emissions between brands and geographic sources, and reports that modifying North American flue-
147
curing approaches from direct-heating to indirect heating mechanisms were lowering nitrosamine
148
levels in USA and Canada tobaccos and cigarettes (Peele et al 2001, IARC 2004, Gray and Boyle 2004).
149
TobReg also concluded that the values of individual toxicant ratios found across the datasets suggest
150
that mandated reductions would have a substantial effect on the emissions from brands remaining
151
on the market.
RI PT
145
SC
152
During the present study, a second WHO technical report on toxicant levels in smoke was published
154
(WHO 2015). The technical report includes an independent commentary by the late Dr Nigel Gray
155
which, whilst not necessarily representing the view of the WHO or TobReg, was unanimously
156
recommended for inclusion in the report due to ‘the thought-provoking nature of its content and
157
goals’. Gray set forth a third model (model 3) of toxicant emission limits based on a ‘generous’
158
upper limit set at three times the lowest toxicant emission level achieved on the market, reviewable
159
after 2 years and then, where practical, set lower.
TE D
160
M AN U
153
A limitation in the TobReg approach, identified by TobReg [WHO 2008], was the lack of available full-
162
market data with which to assess the impact of TobRegs proposals and consequent reliance on the
163
relatively small available dataset of publically available toxicant emission values. TobReg noted the
164
limitations involved in the limited dataset covering few geographic sources and cigarette brands, and
165
recognised that the performance of charcoal filter cigarettes were not well characterised by the
166
approach of normalising yields to nicotine (a limitation that might also apply to products with
167
different circumferences). In the eight years following publication of the TobReg proposals the
168
limited database of cigarette toxicant yields has not grown significantly, and therefore the impact of
169
TobRegs proposed toxicant-reduction regulations on real world cigarette markets, together with the
AC C
EP
161
8
ACCEPTED MANUSCRIPT practical realities of adhering to the proposals, have not been tested and their implications remain
171
unclear.
172
The aim of the present study was to fill this knowledge-gap, and thereby assess the real-world
173
impact of TobReg’s proposals. Four diverse cigarette markets, identified initially by principal
174
components analysis (PCA) of an extensive database of smoke emission values (Camacho et al 2015),
175
were chosen to understand the performance of the TobReg proposals under the widest possible
176
range of testable conditions. Cigarette products were sampled from these countries following the
177
TobReg approach, and smoke emissions of the nine toxicants and nicotine were determined in these
178
products under the Health Canada intense (HCI) smoking regime. These data were then used to test
179
the three TobReg models, in order to characterise their impact on these four diverse cigarette
180
markets.
M AN U
SC
RI PT
170
181 2. METHODS
183
2.1 Selection of study markets
184
Principal components analysis (PCA) was used to select four diverse markets with which to assess
185
the impact of the TobReg proposals for mandated lowering of emissions of nine smoke toxicants. An
186
in-house database containing recent (2006-2011) measurements on tobacco filler blend components,
187
mainstream smoke toxicant emissions, and various physical parameters for 811 commercial
188
cigarette products from 79 markets was used in the PCA to choose four markets with distinctly
189
different product styles.
190
Smoke toxicant emission yields were not used as variables in the PCA because the majority of the in-
191
house data had been generated under the ISO smoking regime, rather than HCI regime upon which
192
the TobReg recommendations are based [WHO, 2008; Burns et al 2008]. Instead, the cigarette filler
193
blend components NNK and NNN, which are direct precursors of NNK and NNN in smoke, were used,
AC C
EP
TE D
182
9
ACCEPTED MANUSCRIPT together with total sugar, a precursor of the smoke toxicant formaldehyde. A fourth variable, filter
195
charcoal loading was included in the PCA due to its ability to affect the levels of volatile smoke
196
toxicants, such as acetaldehyde, acrolein, benzene and 1,3-butadiene. The variables entered in the
197
PCA include factors affecting eight of the nine toxicants proposed for mandated lowering, carbon
198
monoxide (CO) being the exception. The PCA was conducted via JMP Pro version 10 (SAS Institute,
199
Cary, NC, USA).
200
2.2 Study products
201
After the study markets had been selected by PCA, a list of all current cigarette products (regardless
202
of manufacturer) on sale in each of those markets was obtained by local BAT sales-force employees
203
or in the case of Brazil, where product registration is a regulatory requirement, the product list was
204
obtained from ANVISA. The Brazilian market was sampled in Q1 2012; Romania in Q4 2012; Australia
205
in Q1 2013; and Germany in Q4 2013. The products were either sourced directly from BAT factories
206
or purchased from the market place in the case of brands from other manufacturers. The
207
acquisition of products took several months for each market. Once acquired product samples were
208
sent directly to the analysis laboratory.
209
2.3 Product analysis
210
All products were analysed at a single, ISO 17025 accredited laboratory in Brazil. The four markets
211
were analysed sequentially in a series of batches wherein each market was measured as a whole
212
prior to commencement of the next study market. This process lasted more than 2 years, owing to
213
the large number of study products.
214
Smoke toxicant emissions of acetaldehyde, acrolein, 1,3-butadiene, benzo[a]pyrene (B[a]P), benzene,
215
CO, formaldehyde, NNN, NNK, tar, and nicotine were determined under the HCI regime [Health
216
Canada, 1999]. If the product contained a flavour capsule then analysis was carried out after the
217
capsule had been crushed. Five replicate measurements were conducted per product for
AC C
EP
TE D
M AN U
SC
RI PT
194
10
ACCEPTED MANUSCRIPT acetaldehyde, acrolein, 1,3-butadiene, B[a]P, benzene, formaldehyde, NNN and NNK, whereas eight
219
replicates were conducted for tar, nicotine and CO, in compliance with ISO sampling requirements
220
for these toxicants [ISO 8243:2013].
221
The mainstream smoke toxicants were analysed by standard methods that have been internally
222
validated for repeatability and reproducibility [AOAC 2002; ISO 1994]. The methods follow, or are
223
based on, internationally standardised or recognised protocols (i.e. ISO, CORESTA or official Health
224
Canada methods). The methods are multi-analyte, whereby members of a group of toxicants (e.g.
225
volatiles or carbonyls or tobacco specific nitrosamines, etc.) are analysed simultaneously from the
226
same cigarettes. Details of the analytical methods are given in the Supplementary Information.
M AN U
SC
RI PT
218
227 228 2.4 Data analysis
230
Smoke emissions data were collated and toxicant-to-nicotine emission ratios were calculated using
231
Base SAS version 9.3 (SAS Institute). Product assessment against potential toxicant emission ratio
232
limits was performed with Microsoft Excel 2010 (Microsoft Corporation, Redmond, WA, USA). The
233
four markets were analysed separately.
234
In analysing TobReg model 1, for each market, the limit for each of the nine toxicants was calculated
235
in accordance with TobReg [WHO, 2008] as follows. For each toxicant, the mean yields of toxicant
236
and nicotine for each product were used to calculate the mean emission level as a ratio to the mean
237
yield of nicotine: the emission ratio value. The median emission ratio value per market was then
238
determined across the range of toxicant ratios for all products measured in that market. The toxicant
239
emission limits for each market were determined as the median emission ratio value for NNN and
240
NNK, and 125% of the median emission ratio value for formaldehyde, acetaldehyde, acrolein, B[a]P,
241
CO, 1,3-butadiene and benzene. With TobReg model 2, the limits used were those published by
AC C
EP
TE D
229
11
ACCEPTED MANUSCRIPT TobReg [WHO 2008, Burns et al 2008]. With TobReg model 3 (WHO 2015) the limits were calculated
243
for each market by establishing the minimum per-cigarette emission level for each toxicant, and
244
multiplying this value by three. The measured toxicant emission ratio values, or toxicant emission
245
levels, were compared to the limits for each model, and the number of compliant and non-compliant
246
products calculated with regard to each toxicant limit individually and also when all nine limits were
247
applied simultaneously.
RI PT
242
SC
248 3. RESULTS
250
3.1 Selection of study markets using Principal Component Analysis
251
PCA was used to identify four markets with diverse products. The four input variables, filler blend
252
content of NNN, NNK and total sugar, and filter charcoal loading (Supplementary Table 1), were
253
chosen on the basis of data measured by BAT on 811 products from 79 countries (Supplementary
254
Table 2). The PCA used parameters representing the majority of the nine TobReg toxicants. B[a]P in
255
filler blend was also considered as a variable for the PCA, as it is a driver for B[a]P smoke emissions,
256
however this data was not available for all products in the in-house database. Other physical
257
measures, such as circumference, were not included in the PCA as the majority of in-house data was
258
from king-size circumference cigarettes.
259
The PCA defined a variable for blend character (i.e., TSNA to formaldehyde (sugar)) as principal
260
component 1 (PC1), accounting for ~45% of the variation in product measurements, and a variable
261
for the amount of charcoal included in the cigarette filter as principal component 2 (PC2), accounting
262
for ~26% of the variation associated within these measurements. The PC1 and PC2 scores of the
263
products were plotted to identify diverse markets (Figure 1).
264
AC C
EP
TE D
M AN U
249
12
ACCEPTED MANUSCRIPT From the score plot, we identified four groupings that described the most diverse market scenarios
266
within our database. These four markets were Brazil (mixed blend products, low filter charcoal
267
prevalence; Figure 2A) ; Romania (mixed blend products, high filter charcoal prevalence; Figure 2B) ;
268
Australia (predominantly flue-cured Virginia blended products; Figure 2C) ; Germany (predominantly
269
US-blended products; Figure 2D).
270
271
After selection of the four study markets, current products on sale in each market were obtained for
272
analysis. Sourcing the test products in each market proved logistically very challenging, for reasons
273
such as delisting of product and geographical limitations in distribution. In addition, it was not
274
possible to source a sufficient volume of some products (a minimum of 400 cigarettes per product
275
were required for analysis, with 800 preferred) due to the limited availability of low-market share
276
cigarette brands and product de-listing during the sampling exercise. Across the four markets, the
277
percentage of products collected in sufficient quantity for testing was 80% (132 out of 166 products)
278
in Brazil, 93% (138 out of 148 products) in Romania, 97% (172 out of 177 products) in Australia, and
279
92% (339 out of 367 products) in Germany. Given the difficulty encountered with the sampling
280
exercise, multiple time-point sampling of the type recommended by TobReg was not possible.
281
Nevertheless, the products sampled in this single point-in-time exercise were considered sufficient
282
for the analysis to be conducted as planned.
283
3.2 Mainstream smoke toxicant emissions
284
HCI mainstream smoke emissions of the nine TobReg toxicants plus nicotine and tar were measured
285
for each test product, and the distribution of toxicant emission yields per market, both on a per
286
cigarette basis and as a ratio to nicotine emission yield, were summarized (Table 1).
287
Regarding the toxicant yields per cigarette, fairly normal distributions were observed with mean and
288
median values being similar in each case, and no markedly different values comparing the minimum
AC C
EP
TE D
M AN U
SC
RI PT
265
13
ACCEPTED MANUSCRIPT yield to the lower quartile nor maximum yield to the upper quartile, (Table 1, Supplementary Figure
290
1). In general, standard deviation measurements were ~20% of mean or median values however
291
standard deviations were a larger percentage of the mean or median values for NNN and NNK,
292
particularly for NNN in Australia and NNK in Brazil.
293
There were a few exceptions due to the inclusion of atypical products. In the Australian market, zero
294
yields of nicotine (and TSNA) were determined for two herbal cigarettes; as a result, these two
295
products were excluded from the yield determinations per mg of nicotine, although their yields of
296
the seven non-TSNA toxicants from these non-tobacco cigarettes were similar to those in
297
conventional cigarettes. The Australian market survey also included a kretek-style product (a blend
298
of tobacco, cloves, and other flavours), which was responsible for the notably high maximum yields
299
of nicotine (3.58 mg/cig), tar (61.5 mg/cig), B[a]P (40.9 ng/cig), formaldehyde (253 µg/cig), acrolein
300
(187 µg/cig) and benzene (135 µg/cig) in that market. The German market included a cigarette
301
wrapped in a reconstituted tobacco sheet instead of cigarette paper, which produced the notably
302
high maximum yields determined for acetaldehyde (2567 µg/cig), 1,3-butadiene (385µg/cig),
303
benzene (262µg/cig) and CO (72.1 mg/cig) in that market.
304
Comparing the emissions of each toxicant among the four sampled markets, the Australian market
305
exhibited lower TSNA yields and higher formaldehyde yields, which is consistent with the
306
predominantly flue-cured characteristics of products in this market. B[a]P yields were lower in the
307
Romanian market, whereas nicotine and 1,3-butadiene yields were lower in the Brazilian market
308
(Table 1).
309
Considering toxicant yields per mg nicotine (Table 1, Supplementary Figure 2), summary statistics
310
for yields per mg nicotine were lower as compared with the results per cigarette because, on
311
average, nicotine yields were greater than 1 mg/cig. The exception to this, besides the two herbal
312
Australian products already mentioned, was a locally produced Brazilian product that had the
313
highest levels of NNN (392 ng/cig) and NNK yields (670 ng/cig) on a per cigarette basis, in that
AC C
EP
TE D
M AN U
SC
RI PT
289
14
ACCEPTED MANUSCRIPT market, and the lowest nicotine yield (0.82 mg/cig) of all the Brazilian products surveyed, and
315
therefore produced notably high TSNA yields per mg nicotine (NNN 479 ng/mg nicotine, NNK 818
316
ng/mg nicotine). This product also produced the maximum yield ratios in Brazil for B[a]P (28.4 ng per
317
mg nicotine), acetaldehyde (1847 µg per mg nicotine), acrolein (170 µg per mg nicotine), benzene
318
(93 µg per mg nicotine) and CO (30.2 mg per mg nicotine), as a consequence of the low nicotine yield.
319
320
3.3 Toxicant limits
321
Model 1:
322
Market-specific toxicant limits were calculated by using the method proposed by TobReg (Table 2).
323
The toxicants with the widest range of TobReg limits were the two TSNAs (NNK, 21–78 ng/mg
324
nicotine; NNN, 27–90 ng/mg nicotine), where the range of limits (difference between highest and
325
lowest limit expressed as a % of the lowest limit) was 230-270%. In contrast, six of the other seven
326
toxicants showed much smaller ranges, an order of magnitude lower, at 17 - 30%; the range of limits
327
for B[a]P was also comparatively low, at around 50%. Brazil had the highest limits for seven of the
328
nine toxicants with only the 1,3-butadiene limit being relatively low (second lowest when compared
329
across the 6 datasets). Romanian products were notable for having the lowest calculated limits for
330
six toxicants. The lowest TSNA limits were found with Australian products under model 1.
331
Model 2:
332
Comparing the limits calculated under model 1 with the fixed limits provided by TobReg under
333
model 2 (Table 2), showed that the ‘Canadian brands’ limits set by TobReg were lower than those
334
calculated in the current study for five of the nine toxicants, but higher for formaldehyde.
335
Comparing the limits for the two flue-cured datasets (Australia in this study, model 1 vs Canadian
336
brands limits set by TobReg, model 2), the largest discrepancy was in NNK limits, where the Canadian
AC C
EP
TE D
M AN U
SC
RI PT
314
15
ACCEPTED MANUSCRIPT brands limit set by TobReg was more than twice the market-limit determined for Australia in this
338
work; the difference between the limits for the other toxicants ranged from –27% to 20%. For the
339
two US-blended datasets (Germany in this study, model 1 vs International brands limits set by
340
TobReg, model 2), the largest discrepancy was in TSNA limits, where the International brands limits
341
for both NNK and NNN were more than 50% higher than those determined with German brands in
342
this work; otherwise, the differences ranged between –6% and 23% per toxicant. The International
343
brands limit (model 2) for NNN was higher than any of the four markets surveyed in the present
344
study, model 1.
345
Market-specific limits were also calculated according to model 3 (Table 3). The toxicants with the
346
widest ranges of limits were again the two TSNAs, with a 170% range of limits for NNK, and a 270%
347
range of limits for NNN. These ranges of TSNA limits with model 3 are similar to the range of limits
348
found with TobReg model 1. The range of model 3 limits for the other seven toxicants are 36-96%,
349
which is a greater range than found under model 1.
350
351
3.4 Product compliance with limits
352
3.4.1 Model 1 - Market-specific limits
353
For each market, the proportions of non-compliant products were determined individually for each
354
toxicant (Figure 3A) and cumulatively when applying all nine toxicant limits simultaneously (Figure
355
3B). Notably, between 72% and 79% of products within each market would be non-compliant
356
following the TobReg model 1 approach (Table 4), i.e. after the application of all nine market-specific
357
toxicant limits [WHO 2008]. The proportions of non-compliant product were also determined
358
cumulatively when applying to both TSNAs (median limits) and cumulatively for the non-TSNA
359
toxicants (125% of median limits), (Figure 3B). The consistency of the proportions of non-compliant
AC C
EP
TE D
M AN U
SC
RI PT
337
16
ACCEPTED MANUSCRIPT product across the four markets for each comparison in Figure 3 emphasises the mathematical
361
nature of this approach.
362
363
The main driver of non-compliance in all markets were the limits for the two TSNAs, where, by
364
definition, the TobReg proposed use of median values ensures that 50% of products would be
365
immediately non-compliant with limits for both NNN and NNK. Imperfect correlation between the
366
two TSNAs resulted in a combined non-compliance rate of 56-66% (Figure 3B) for these two
367
toxicants. The non-TSNA toxicant limits, being based on 125% of the market median, produced
368
fewer non-compliant products than the two TSNAs. The individual non-compliance rates for the non-
369
TSNA toxicants ranged from 4% (benzene in Romania) to 24% (1,3-butadiene in Brazil), with a
370
combined non-compliance rate of 34-45% for these 7 toxicants (Figure 3B).
371
Regarding the number of toxicants for which each individual product was non-complaint (Figure 3C),
372
it was most common for products to be non-compliant for 0, 1 or 2 toxicants (20%–30% of products
373
for each), followed by 3 toxicants (~10%), and then 4–9 toxicants (<5% for each). The two TSNAs
374
accounted for most products with 1 or 2 toxicant non-compliances, whereas the third most common
375
non-compliant toxicant varied by market: 1,3-butadiene in Brazil and Australia; formaldehyde in
376
Romania and Germany (Figure 3A).
377
3.4.2 Model 2 - TobReg fixed toxicant limits
378
The application of limits based on ‘International brands’ or ‘Canadian brands’, set by TobReg under
379
model 2 for use in the absence of market specific information [WHO 2008], resulted in an increase in
380
overall numbers of non-compliant products, as compared with model 1’s calculated market-specific
381
limits, for most market comparisons (Table 4). This increase was often substantial—for example,
382
100% of products in Brazil were non-compliant against both of TobReg’s fixed limits (Figure 4A and
383
4B).
AC C
EP
TE D
M AN U
SC
RI PT
360
17
ACCEPTED MANUSCRIPT Application of the International brands limits to the predominantly US-blended German market
385
(Table 4, Figure 4A), increased the percentage of non-compliant products from 80% to 89%. Similarly,
386
application of the Canadian brands limits to the predominantly flue-cured Australian market resulted
387
in an increase in non-compliant products from 72% to 80%. Only the Romanian market (mixed blend;
388
high incidence of charcoal in the filter) showed a reduction in non-compliant products from 78% to
389
71% when the International brands limits were compared with the market-specific limits of model 1.
390
391
Comparing product non-compliance rates for NNN and NNK when using the two sets of fixed limits
392
compared to the market specific limits of model 1 (Figures 3A and 4), showed that use of the
393
international brands limits resulted in a marked decrease in non-compliance levels against the TSNA
394
limits for 3 of the markets surveyed in this study, whilst for the Brazilian market non-compliance
395
rates were similar (~60%). Applying the Canadian brands fixed limits for NNN and NNK increased
396
non-compliance rates for 3 of the markets substantially (>90%), whereas for the Australian market
397
there was a small decrease in non-compliance rates, from 56 to 50%.
398
Product non-compliance rates for non-TSNA toxicants when using TobReg fixed limits showed a
399
marked increase to 66%-99% non-compliance, compared to ~40% non-compliance when using the
400
market-specific limits of model 1 (Figure 3B and Figure 4). Formaldehyde limits produced the
401
greatest number of non-compliances under the ‘International brands’ set of limits (59-88% non-
402
compliance), but a negligible level of non-compliance under the ‘Canadian brands’ limits (1-2% non-
403
compliance). Conversely, acetaldehyde emissions produced high levels of non-compliance under the
404
‘Canadian brands’ limits (45%-88%), but lower levels of non-compliance under the ‘International
405
brands’ limits (1%-36%). Formaldehyde was the main driver of non-compliance when applying the
406
‘International brands’ limits in each of the four markets whilst NNN, followed by NNK (except for the
407
Australian market products), acetaldehyde, benzene (except for the Romanian market products) and
408
1,3-butadiene were the main drivers when applying the ‘Canadian brands’ limits. The Brazilian
AC C
EP
TE D
M AN U
SC
RI PT
384
18
ACCEPTED MANUSCRIPT market in particular was negatively affected by the use of the fixed limits with the highest levels of
410
non-compliance for each toxicant individually, apart from 1,3-butadiene, and 100% non-compliance
411
when applying all 9 limits simultaneously under either set of limits.
412
3.4.3 Model 3 limits:
413
The level of product non-compliances obtained when applying the model 3 limits to each of the
414
appropriate markets are summarized in Table 4. The total level of non-compliances varied
415
significantly by market, ranging from 68% in Australia to 99% in Germany. The impact on the
416
Australian and Romanian markets were comparable to that found with the market-specific limit
417
model. However, the model 3 limits produced significantly higher levels of non-compliances for the
418
Brazilian (86%) and German markets (99%).
419
The levels of non-compliances against the model 3 limits were driven strongly by NNN, and to a
420
lesser degree, by NNK, Figure 5. The other seven toxicants resulted in few non-compliances, other
421
than with formaldehyde in Romania (35% non-compliances) and Germany (14%), and with B[a]P in
422
Romania (17%) and Brazil (15%).
423
EP
TE D
M AN U
SC
RI PT
409
425
4. DISCUSSION
426
The principal goal of TobReg’s proposed strategy is to reduce the emissions of nine selected
427
mainstream smoke toxicants in commercial cigarettes by excluding products with the highest
428
smoking-machine measured levels of these toxicants. TobReg’s proposals are based on a complex
429
trade-off of considerations that they believe will result in substantial lowering of toxicant emissions,
430
while not resulting in the elimination of most of the brands sold on a market. This study represents
431
the first full evaluation of this global proposal for tobacco product regulation, and has explored the
AC C
424
19
ACCEPTED MANUSCRIPT impact of the proposed regulations on real-world, whole-market data. A number of important
433
learnings have emerged from this analysis.
434
4.1 Toxicant ceilings arising from TobReg models
435
4.1.1 Comparison of model 1 and model 2 limits.
436
The suitability of TobReg proposed fixed limits for global deployment can be assessed by comparison
437
with the individual market limits calculated in this work, Table 2. Very few of the individual limits
438
from the four markets examined in the current work gave a set of limits that matched the TobReg
439
proposals. Comparing the Australian market limits to the TobReg limits calculated from the
440
Canadian brands showed that while NNN and B[a]P limits were in exact agreement, CO and acrolein
441
limits differed by 8-10%, benzene limits by 14%, acetaldehyde, formaldehyde and 1,3-butadiene
442
limits differed by 20-30% and NNK limits differed by 55%.
443
Similarly, comparing the TobReg limits calculated from the International Brands dataset with the
444
limits obtained in this study from the German, Romanian and Brazilian datasets also showed
445
significant disagreements. CO limits differed by 3-12%, acetaldehyde limits by 4-16%, 1,3-butadiene
446
limits by 3-19%, acrolein limits by 5-28%, NNK limits by 8-36%, B[a]P limits by 18-27%, benzene limits
447
by 10-44%, NNN limits by 21-37% and formaldehyde limits by 30-57%.
448
Given the sensitivity of compliance/non-compliance to small changes in limits or emission ratios
449
around the median and 125% median values (Section 4.3) these observations clearly demonstrate
450
that the TobReg proposed fixed limits are inappropriate for use in these four markets, and most
451
likely on a global basis.
452
In understanding why the TobReg proposed fixed limits do not reflect actual market limits, it should
453
be noted that TobReg did not use full-market data to calculate either set of limits: the 2004 Canadian
454
emissions dataset contained measured toxicant emissions data for only a subset of the products
455
listed (60 out of 249 brands sold in Canada in 2004), on top of which data from US and French blend
AC C
EP
TE D
M AN U
SC
RI PT
432
20
ACCEPTED MANUSCRIPT 456
cigarettes were removed (products where NNN levels > 100 ng/mg of nicotine), leaving a final set of
457
48 products. For the ‘International brands’, the 49 products reported by Counts et al (2008) were
458
from a single manufacturer, with few charcoal filtered or reduced circumference products. Neither
459
dataset reflects the breadth of products available in any one market, let alone worldwide.
RI PT
460 The substantial differences in non-compliance rates when using either of the TobReg fixed limits,
462
together with the lack of fit with limits calculated from actual market data, highlights how important
463
it would be to have accurate, contemporary, market-specific data if the proposed approach were to
464
be enacted, particularly given the potentially high rates of product non-compliance observed in the
465
present study.
M AN U
SC
461
466 4.1.2 Model 3 limits
468
With model 3 the toxicant limits (Table 3) are defined by the emissions of one product for each
469
toxicant in each market, and therefore interpretation of these observations is likely to be challenging,
470
as the product defining the market limit may be unrepresentative of the overall market. For
471
example, Germany, a US-blended market, has the lowest TSNA limits of the four markets, despite
472
having one of the widest distributions of TSNA emissions of the markets examined in this survey
473
(Supplementary Figure 1). In contrast, the Australian formaldehyde limit is the highest of the four
474
markets, which is a representative reflection of the predominantly flue-cured nature of the tobacco
475
products sold on this market. Another example of the potential for unrepresentative products
476
defining market limits under model 3 is provided by the Brazil market. Brazil provided five of the
477
lowest model 3 limits (defined by a single product for each toxicant) in the overall dataset. This is in
478
direct contrast to the findings from application of model 1 (defined by all products on the market),
479
where Brazil had seven of the highest limits. Clearly, the method for calculation of toxicant limits
480
has a profound impact on the standards that products would have to meet if these models were
481
enacted.
AC C
EP
TE D
467
21
ACCEPTED MANUSCRIPT 4.2 Product compliance with toxicant ceilings:
483
4.2.1 Model 1 - Individual market limits
484
Our survey of four markets that differ in tobacco blend style, cigarette design and toxicant profiles
485
has shown that the impact of this proposed regulatory strategy would be consistently severe: 72%–
486
80% of the products analysed in each market were non-compliant with the market-specific limits
487
(Figure 3B). There were two main drivers of the high non-compliance levels: the primary driver being
488
use of median market yields as the ceiling for NNN and NNK emissions which in itself resulted in 56-
489
66% failures; and a secondary driver being the simultaneous application of 9 separate toxicant
490
ceilings. The end result, where most products in a market cannot comply with proposed ceilings,
491
clearly does not meet the stated intention of TobReg of providing a regulatory framework that does
492
not cause elimination of most brands in their current form from the market.
493
4.2.2 Model 2 - Fixed limits
494
Application of the two sets of ‘fixed’ TobReg proposed limits was associated, in almost all cases, with
495
substantially higher non-compliance rates than found with market-specific limits (model 1). In some
496
cases 100% non-compliance was observed. The extent of product failures with the International
497
Brand limits was 70-100%, and 80-100% with the Canadian brands limits (Figures 4A and B).
498
On application of the ‘International brands’ limits the increase in non-compliances was driven by the
499
non-TSNA toxicants, particularly formaldehyde, rather than the two TSNAs. Brands failing the
500
‘International’ formaldehyde limits reached as high as 88%, and up to 83% failed the benzene limits;
501
conversely in most cases failures against TSNAs were <30%.
502
In contrast, for the ‘Canadian brands limits’ the increase in non-compliance rates arose from a
503
number of toxicants and varied by market: there were large increases in NNN and NNK non-
504
compliances for Brazil, Romania and Germany (NNN only) and in non-compliances with
505
acetaldehyde and 1,3-butadiene, but the non-compliance rate for formaldehyde was negligible.
AC C
EP
TE D
M AN U
SC
RI PT
482
22
ACCEPTED MANUSCRIPT TobReg’s impact assessment estimated 40-41% product failures, driven by the 7 toxicants other than
507
NNN and NNK, an estimation based upon the belief that TSNA emissions could be lowered by
508
cigarette manufacturers without difficulty. Our analysis clearly demonstrates much higher levels of
509
non-compliances when actual products are examined, reaching 100% in some cases. Even the
510
impact of the seven limits for toxicants other than the TSNAs was more severe than estimated by
511
TobReg, with 66-99% failures against the International Brands limits, and 66-90% non-compliances
512
with the Canadian brands limits. Therefore, use of the proposed fixed limits does not meet the
513
TobReg intention of toxicant reduction without severe disruption to a market.
514
4.2.3 Model 3 limits
515
Application of the model 3 limits (3 x market minimum levels) also caused greater levels of non-
516
compliance than the TobReg model 1 limits (based on market median) in 2 of the 4 markets
517
surveyed in this work (Table 4). This was predominately driven by the limits for TSNAs and NNN in
518
particular; with the other seven toxicants having relatively little impact on compliance (Figure 5).
519
The model 3 limits were described as generous (WHO, 2016), however when their impact is
520
examined on typical cigarette markets, their impact has been shown to be severe, with 99% non-
521
compliances in Germany.
522
Limits calculated under model 3 are highly susceptible to low yield, atypical products. Germany was
523
by far the largest market surveyed and had an overall TSNA distribution comparable to Brazil and
524
Romania, and much higher than Australia (Supplementary Figure 1). Nevertheless, due to the low
525
TSNA emissions of one product on the German market, the NNN and NNK limits for the German
526
market were the lowest of the four markets examined in this work, which resulted in 99% non-
527
compliances with the products sold in Germany.
528
Two factors may present difficulties for the model 3 approach. First, low level quantification of
529
smoke toxicant emissions can be particularly sensitive to analytical imprecision. At low levels, such
AC C
EP
TE D
M AN U
SC
RI PT
506
23
ACCEPTED MANUSCRIPT as at the limit of quantification, analytical errors can have a substantial impact on the measured
531
value. Second, no guidance was provided (WHO 2016) on the most appropriate procedure under
532
model 3 when the lowest emission value on a market are below the limits of quantification, and this
533
represents a serious shortcoming of the proposals.
534
4.2.4 Toxicant yields as a ratio to nicotine
535
With the first two TobReg models, the recommendation was to express toxicant emission data as a
536
ratio to the measured smoke yield of nicotine in order that the machine measured values are not
537
misleadingly judged to represent a measure of human exposure and therefore risk [WHO, 2008]. In
538
contrast, the later model 3 proposals, published in WHO Technical Report 989, proposed use of per-
539
cigarette yields rather than values calculated as a ratio to nicotine emissions. To understand the
540
importance of these differences in model approaches, we therefore examined the data obtained in
541
this study to assess the impact of using nicotine ratios on product compliance.
542
543
Analysis of the data showed that setting an emission limit for a toxicant on the basis of its yield ratio
544
to nicotine resulted in some products with high absolute toxicant emissions being compliant with
545
toxicant/nicotine limits when they had a relatively high nicotine yield. Figure 6 provides an example
546
demonstrating the relationship between nicotine and NNN yields, and NNN-to-nicotine yields for the
547
German market products when compliance is categorised using the model 1 approach, and the
548
calculated market median for NNN-to-nicotine emissions of 72 ng/mg (Figure 6A).
549
Figure 6B shows that some compliant products have higher NNN emissions than non-compliant
550
products, which seems counter-intuitive. For example, 80 out of the 169 ‘non-compliant’ products
551
(in terms of NNN-to-nicotine yields) have absolute NNN yields that are lower than that of the
552
‘compliant’ product with the highest NNN yield (171 ng/cig). Conversely 92 out of 169 ‘compliant’
553
products (in terms of NNN-to-nicotine yields) have absolute NNN yields that are higher than that of
AC C
EP
TE D
M AN U
SC
RI PT
530
24
ACCEPTED MANUSCRIPT the ‘non-compliant’ product with the lowest NNN yield (98 ng/cig). This is a point of major concern
555
as Clinical studies have demonstrated that products with higher NNN emissions lead to greater NNN
556
exposure amongst smokers (as measured by biomarkers of exposure) than products with lower NNN
557
emissions [Shepperd et al, 2013], and similarly for NNK [Czoli & Hammond, 2014]. Figure 6C shows
558
that nicotine emission levels make a significant contribution to whether products are compliant to
559
the NNN/nicotine limit, with non-compliant products associated with lower level nicotine emissions,
560
however this contribution is much lower than the magnitude and differences in NNN emissions.
561
Figure 7 shows similar plots to Figure 6B for the remaining 8 toxicants for the German market
562
products, i.e. toxicant yields categorised by compliance to the model 1 market specific calculated
563
toxicant-to-nicotine ceiling. The analysis shows that the volatile species are even more prone to this
564
effect, with significant overlap in per-cigarette emissions between compliant and non-compliant
565
products. A similar full set of plots for the Brazilian, Romanian and Australian market survey data are
566
included in Supplementary Figures 3-5.
567
These observations highlight an important concern over the TobReg models based on use of
568
emission values as a ratio to nicotine levels, as the practical consequences are contrary to TobReg’s
569
stated intention of excluding products with the highest toxicant emissions, as well as having
570
implications for human exposure.
571
572
4.3 Impact of Measurement Error on TobReg proposals
573
All analytical measurements are associated with uncertainty, imprecision or inaccuracy, arising from
574
a combination of analytical error and variation in the manufactured product. Measurement of
575
mainstream smoke toxicant emissions in particular are associated with a significant amount of
576
variation [Hyodo et al., 2006; Morton and Laffoon 2008; Gaworski et al 2011; Intorp et al., 2009;
577
Teillet et al., 2013; Purkis et al., 2014; Eldridge et al. 2015]. By definition, the median values that are
AC C
EP
TE D
M AN U
SC
RI PT
554
25
ACCEPTED MANUSCRIPT pivotal to the TobReg model lie in the middle of the data series, which for normal or near normal
579
distributions (such as that seen for cigarette smoke toxicant data, see Supplementary Figures 1 and 2)
580
is the most densely populated portion of the data distribution. This region is most susceptible to
581
these errors (Figure 8), with consequential uncertainty over the relative positions of individual
582
products in a ranked population. We therefore investigated whether measurement error
583
significantly influences product compliance with TobReg proposed ceilings.
584
585
To assess the impact of measurement error on product compliance, the potential error in smoke
586
toxicant ratio emission measurements, calculated as ±2 times the coefficient of variation (2CV), has
587
been determined when using this single ISO 17025 accredited laboratory in terms of both the
588
repeatability of the measurement method, based on a reference cigarette, and the variation
589
observed when analysing repeatedly manufactured commercial cigarette products [Eldridge et al
590
2015]. Applying the potential measurement error to the NNK-to-nicotine emissions for the German
591
market products demonstrates the practical consequences of these sources of variability (Figure 8):
592
a hypothetical product with a “true” NNK/nicotine value at the measured median limit of 46 ng/mg
593
has an uncertainty of measurement that means that measured values between 31 – 62 ng/mg of
594
actual NNK / nicotine can be obtained from the measurement laboratory. Hence the compliance or
595
non-compliance of this hypothetical product would be a question of chance. This range of possible
596
measurement values is similar to the interquartile range of all NNK-to-nicotine measurements in the
597
German market (34-63 ng NNK/ mg nicotine, Table 1). Therefore, to guarantee compliance with a
598
median limit, and minimise the impact of measurement uncertainty, a true value around the lower
599
quartile value would be the necessary performance target for a brand. In this study only 20%, or 67
600
out of 339, of the German products measured could be reliably determined as compliant for
601
NNK/Nicotine due to the impact of measurement uncertainty, and in the same way only 26%, or 89
602
out of 339 products, could be reliably determined as non-compliant.
AC C
EP
TE D
M AN U
SC
RI PT
578
26
ACCEPTED MANUSCRIPT By contrast, the effect of likely measurement error on toxicant emission ratios when limits are based
604
on 125% of the median was, in general, less pronounced in terms of the number of products
605
affected. The impact of measurement error will depend both on where the 125% of the median limit
606
lies within the distribution of emission data and on the magnitude of the potential error. Using the
607
typical measurement error reported by Eldridge et al. [2015], the best- and worst-case scenarios
608
from this study are shown in Figure 9: the best case scenario being CO/Nicotine in Romania (Figure
609
9A), where the potential measurement error affects ~20 out of the 138 products, and the worst
610
case 1,3-butadiene/nicotine in Romania (Figure 9B), where the potential measurement error
611
encompasses all 138 products measured. The uncertainty in 1,3-butadiene/nicotine product
612
compliance arises due to substantial longitudinal analytical variation previously observed with 1,3-
613
butadiene analysis; over a 10-month period 2CV values of 46% were reported [Eldridge et al. 2015]
614
Cumulatively, this more stringent performance standard introduced by real-world measurement
615
uncertainty would reduce the number of products that could be guaranteed to be compliant to
616
around zero when 9 simultaneous limits are considered, (see supplementary Tables 3 and 4 for the
617
impact on non-compliance rates, including measurement error, for each toxicant in each of the four
618
markets).
619
It might be considered that increasing the product sampling frequency may reduce the impact of
620
measurement uncertainty. However, the effectiveness of this strategy is contingent on
621
measurement variability arising solely or predominately from product rather than analytical sources.
622
In the case where analytical errors are a major contribution to the overall measurement variability
623
then further measurements cannot be guaranteed to reduce measurement error.
624
625
All the present data were acquired in a single laboratory and for one-point-in-time samples, and
626
therefore does not include product variability nor the variability observed between results arising
AC C
EP
TE D
M AN U
SC
RI PT
603
27
ACCEPTED MANUSCRIPT from different laboratories. Significant levels of between laboratory variation have been previously
628
reported — up to 18% CV between 3 laboratories (Hyodo et al 2006) and up to 100% CV (70% CV
629
excluding 1,3-butadiene) between 15 laboratories (Intorp et al 2009), for the 9 toxicants from this
630
study. This suggests that meeting toxicant ratio emission limits through measurements carried out at
631
a number of different laboratories might be even more strongly influenced by measurement
632
uncertainty than measurements in a single laboratory.
633
Effective product regulation is fundamentally dependent on accurate and precise determination of
634
smoke toxicants, as well as the availability of technically feasible technologies for their reduction.
635
Current analytical methods are capable of producing precise single-point-in-time measurements;
636
however, a lack of standardised methods and certified reference materials means that accuracy is
637
compromised, owing to substantial variation over time, within and particularly between laboratories
638
[Morton et al 2008; Intorp et al 2009; Oldham et al 2014, Eldridge et al 2015]. Our study
639
demonstrates that for the majority of products on-sale on a market, compliance with TobReg
640
proposals, would be strongly influenced by random measurement variability; management of
641
product toxicant emissions under these circumstances would be a highly challenging activity.
TE D
M AN U
SC
RI PT
627
EP
642 4.4 Multiple toxicant limits
644
TobReg conducted an assessment of correlations between different toxicant emissions from
645
products in their database (WHO 2008), and found evidence for both positive correlations between
646
toxicants, such as between NNN and NNK, as well as negative correlations such as those between
647
formaldehyde and TSNAs or between benzo[a]pyrene and carbonyls. To further understand the
648
impact of the simultaneous application of multiple limits on numbers of product non-compliances
649
we examined the extent to which the nine toxicants were correlated. This exercise is of value
650
because the current study generated a significantly larger database than that used by TobReg. To
AC C
643
28
ACCEPTED MANUSCRIPT achieve this, Pearson’s correlation coefficients (r) were determined among the toxicant emission
652
yields for each market (Table 5), after excluding the four atypical products discussed in Section 3.2.
653
Toxicant emissions that were predominantly driven by precursors in the tobacco filler blend (NNN,
654
NNK, formaldehyde and B[a]P) showed the poorest levels of correlation both amongst themselves
655
and against the volatile and gaseous toxicant yields. B[a]P correlations varied widely across toxicants
656
and also against the same toxicant across the four markets. Formaldehyde correlations were positive
657
and weak across the majority of toxicant comparisons but negative against both TSNAs. Even yields
658
of the chemically similar NNN and NNK showed varying degrees of correlation, ranging from
659
moderate (r=0.475), for Romanian market products, to very strong (r=0.864) for Australian market
660
products. This analysis demonstrates that toxicant yields are inter-related to varying degrees, with
661
evidence of both positive (e.g. between CO and volatile toxicants) and negative (e.g. between TSNAs
662
and formaldehyde, or between TSNAs and 1,3-butadiene) correlations (Table 5).
663
The generally weak and inconsistent correlations are a concern for compliance with the TobReg
664
proposals, as it results in a greater number of non-compliances when applying limits to each of the
665
nine toxicants simultaneously: for example, the 50% non-compliance rate observed for an individual
666
TSNA rose to approximately 60% for both TSNAs (Figure 3B). Furthermore, as these relationships
667
are not totally consistent across different markets, effective management of toxicant yields by
668
cigarette manufacturers across multiple markets under this regulatory proposal would be a highly
669
complex affair. In particular, the negative correlations suggest that reducing yields of nitrosamines
670
could lead to an increase in yields of formaldehyde and 1,3-butadiene to some degree, as seen when
671
comparing the Australian market data to the other markets examined in this study.
672
673
4.4 Logistical Considerations
AC C
EP
TE D
M AN U
SC
RI PT
651
29
ACCEPTED MANUSCRIPT The present study highlighted the logistical challenges in obtaining samples of all currently available
675
commercial cigarette products for a given market. Commercial cigarettes are “fast-moving consumer
676
goods” and there is continuing evolution in the identity of products in a market place, many of which
677
will have low market share or use by smokers. As a result, in our experience, sourcing all products
678
on-sale in a market is highly challenging, and we were not able to do so in any of the four markets
679
studied. Overall, 80%–97% of the ‘snapshot list’ of commercial products were obtained for the
680
markets in this study. Even in Brazil, where the identity of all products is clear, due to a paid product
681
registration process with the regulatory authority (ANVISA), it was not possible to sample all
682
products. Although it was straightforward to generate a list of products, it took many weeks to
683
source all of the available products in some markets, and in that time a number of products were
684
removed from the market. In addition, for some competition products with a small sales volume, it
685
was not possible to obtain sufficient cigarettes (>400) on the open market for analysis. Because all
686
products contribute equally to the setting of TobReg proposed limits, limits created in this way may
687
reflect a point-in-time picture of a market, rather than a long term stable view. The extent to which
688
market dynamics may vary in overall toxicant profiles over time is unclear and requires further
689
examination.
690
Given the logistical challenges of sampling all products from a market, the most efficient method of
691
collecting all products would be to require the cigarette manufacturers to analyse or provide
692
samples for analysis. However, TobReg have expressed concerns over manufacturers providing
693
sample products for analysis and instead recommend obtaining product directly from the market or
694
through unannounced collection during manufacture or distribution. This latter route may be the
695
only practical approach for sampling a complete market, but also relies upon frequent or continuous
696
production or availability of a product. With small sales volume products this may not be the case.
AC C
EP
TE D
M AN U
SC
RI PT
674
697 698
4.5 Practicality of reducing TSNA toxicant emissions:
30
ACCEPTED MANUSCRIPT TobReg’s basis for using the market median as a limit for TSNAs are: the observation that a
700
comparatively wide range of NNN and NNK emissions are observed for cigarette products around
701
the world; the view that one of the two main styles of tobacco (flue-cured Virginia or “bright”
702
tobacco) is naturally lower in TSNA levels than the other (air-cured Burley); and reports that
703
nitrosamine contents of US and Canadian flue-cured tobaccos were declining due to changes in flue-
704
curing approaches in these countries (IARC 2004, Gray and Boyle 2004). TobReg therefore regarded
705
reducing TSNA levels via the proposed limits as straightforward to achieve by changing the tobacco
706
blend for those with lower toxicant potential.
707
However, use of the median as a toxicant ratio emission limit is intrinsically problematic: no matter
708
how high or low, or widespread the data or its distribution, half of the products on a market will fail
709
to meet a median limit by definition. Therefore, for markets with the lowest toxicant/nicotine
710
distribution, the practical mechanism for compliance with limits is unclear. For example, with
711
Australian products where NNN/nicotine levels were the lowest of the countries in the current
712
survey, substitution by brands from other markets is unlikely to be a practical solution, and
713
compliance with limits may rely upon the availability of a limited pool of low toxicant precursor
714
tobaccos.
715
The negative correlation between TSNA/nicotine and formaldehyde/nicotine suggests that the
716
simple framework of reducing each of the 9 toxicant/nicotine emissions from cigarettes may be
717
more complex than previously considered, with the potential for unintended impact on emissions of
718
other toxicants. The scope of this was examined by comparing the distribution of toxicant/nicotine
719
emissions across the four markets (Supplementary Figure 2). With NNN/nicotine, the data shows
720
significant commonality in values, other than the Australian products, which showed a lower
721
distribution (but still with significant overlap) to the other three markets. In contrast, with
722
formaldehyde/nicotine Australian products (together with Brazilian products) showed wider
723
distribution of values than the other two markets. Consequently, reductions in NNN/nicotine might
AC C
EP
TE D
M AN U
SC
RI PT
699
31
ACCEPTED MANUSCRIPT be achievable in the Romanian and German markets by changing to an Australian product profile but
725
at the cost of increased formaldehyde/nicotine, and consequent potential for failure against this
726
limit. The consumer acceptability of these changes in product styles is also unclear.
727
TobReg expressed the view that compliance to the levels mandated for NNN and NNK will be readily
728
achievable through established blending and curing practices [WHO 2008]; however, the practical
729
mechanisms by which this could be achieved, and their feasibility on national or global scales is
730
unknown. As noted by Gray and Boyle (2004) “the international tobacco trade is complicated, often
731
based on an auction system, and the introduction of nitrosamine assays on unmanufactured tobacco
732
would be difficult. Certainly regulations could enforce a prohibition on manufacture and sale of high
733
nitrosamine tobacco, but both import and export regulation would be required by many countries,
734
and would need to be applied to tobacco as well as tobacco products. While all this is possible in
735
theory, in practice it is likely that the developed countries will protect themselves best and others
736
probably not at all”.
TE D
737
M AN U
SC
RI PT
724
5. CONCLUSION
739
The present study has provided the first full assessment of the practicalities and impact of three
740
WHO TobReg regulatory proposals for the emissions of nine toxicants in mainstream cigarette
741
smoke.
742
Our study sampled products according to TobReg’s proposals, and we measured the emissions of all
743
available products from four countries with different product styles. The TobReg process of
744
sampling and analysing all products on a market was found to be logistically challenging, and our
745
experience suggests it may not be possible to achieve in practise unless products are sampled from
746
manufacturing sites. When the TobReg model of market-specific limits are applied, 70%–80% of
747
products sold in each of the four countries (Australia, Brazil, Germany and Romania) were found to
AC C
EP
738
32
ACCEPTED MANUSCRIPT be non-compliant with the toxicant limits, and thus would require substantial redesign to continue
749
to be sold in the subsequent regulated market. Application of the second TobReg model of fixed
750
limits (both ‘International brands’ or ‘Canadian brands’ set limits) resulted in, often substantial,
751
increases in non-compliance rates both at an individual toxicant level and when applying multiple
752
toxicant limits simultaneously. In some cases, all products on a market failed to meet the TobReg-
753
provided limits. A third, recently proposed model, of setting limits based on three-times the lowest
754
emission levels of products on a market, shows similar challenging results, with 68-99% non-
755
compliances. This latter model is particularly sensitive to the impact of atypical products on a
756
market, and may be challenged by analytical errors and emission levels too low to be quantified.
757
Use of the market median to set toxicant limits is technically challenging to comply with: first, half of
758
the products on the market are automatically non-compliant; and second, when the likely
759
measurement error is taken into account, compliance with the proposed limits is highly influenced
760
by product and analytical variability for a large proportion of products.
761
Simultaneous application of all nine toxicant limits also contributes substantially to the high levels of
762
non-compliance owing to the lack of close positive correlation between many toxicant emission
763
levels. Evidence of negative correlation (between TSNAs and formaldehyde, or between 1,3-
764
butadiene and NNN, NNK or formaldehyde) suggests that reducing yields of some toxicants would
765
lead to an increase in yields of other toxicants. Very different limits for each toxicant were calculated
766
across the four markets, with limits for TSNAs ranging more than threefold. Furthermore, regulating
767
toxicant emission levels as a ratio to nicotine yields increases the complexity of the regulation and
768
allows some products with relatively high toxicant emissions to register as compliant if their nicotine
769
yields are also high.
770
A fundamental finding of our study is that each of the three TobReg proposed toxicant limit models
771
go far beyond TobReg’s aims of reducing toxicant levels in cigarette smoke whilst not eliminating the
772
majority of cigarette brands in their current form from a market.
AC C
EP
TE D
M AN U
SC
RI PT
748
33
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
773
34
ACCEPTED MANUSCRIPT 6. Acknowledgements
775
The staff of British American Tobaccos analytical laboratories for product and smoke analysis. Lucy
776
Evans for editorial support in drafting the manuscript
777
7. Key References
778
ANVISA 2007, Brazil Resolution RDC No. 90 of the Federal Sanitation Agency effective 27 December
779
2007 (re-published), (www.anvisa.gov.br).
780
AOAC. Guidelines for single laboratory validation of chemical methods for dietary supplements and
781
botanicals. AOAC International, 2002.
782
Australian DOH. Australian cigarette emissions data. Australian Government, Department of Health
783
2002.
784
April 2014).
785
Burns DM, Dybing E, Gray N, Hecht S, Anderson C, Sanner T, O’Connor R, Djordjevic M, Dresler C,
786
Hainaut P, Jarvis M, Opperhuizen A, Straif K (2008). Mandated lowering of toxicants in cigarette
787
smoke: a description of the World Health Organization TobReg proposal. Tob. Control 2008 17:132–
788
141. doi:10.1136/tc.2007.024158
789
Counts ME, Morton MJ, Laffoon SW, Cox RH, Lipowicz PJ. Smoke composition and predicting
790
relationships for international commercial cigarettes smoked with three machine-smoking
791
conditions. Regul. Toxicol. Pharmacol. 2005 41(3):185-227.
792
Czoli C., Hammond D. TSNA Exposure: Levels of NNAL among Canadian tobacco users. Nicotine Tob
793
Res (2014) doi: 10.1093/ntr/ntu251 Article in press.
794
Eldridge A, Betson TR, Vinicius Gama M, McAdam K. Variation in tobacco and mainstream smoke
795
toxicant yields from selected commercial cigarette products. Regul. Toxicol. Pharmacol. 71 (2015)
796
409–427
M AN U
SC
RI PT
774
(accessed
AC C
EP
TE D
http://www.health.gov.au/internet/main/publishing.nsf/Content/tobacco-emis
35
ACCEPTED MANUSCRIPT FDA. Harmful and potentially harmful constituents in tobacco products and tobacco smoke:
798
Established list. US Food and Drug Administration, 2012.
799
http://www.fda.gov/TobaccoProducts/GuidanceComplianceRegulatoryInformation/ucm297786.htm
800
Gaworski CL, Wagner K, Morton MJ, Oldham MJ. Insights from a multi-year program designed to test
801
the impact of ingredients on mainstream cigarette smoke toxicity. Inhalation Toxicology, 2011; 23
802
(s1); 172-183 DOI: 10.3109/08958378.2010.546440
RI PT
797
803
Gregg E, Hill C, Hollywood M, Kearney M, McAdam K, Purkis S, McLaughlin D,
805
Williams M. The UK smoke constituents testing study. Summary of
806
results and comparison with other studies. Beit. Tabakforsch. Intl. 2004 21:117–118.
807
Health Canada. Determination of “tar”, nicotine and carbon monoxide in mainstream tobacco
808
smoke-official method. Health Canada, Ottawa, 1999.
809
Health Canada. Tobacco Reporting Regulations SOR /2000-273.
810
Published 2000. Accessed April 2014.
811
Health Canada. Constituents and emissions reported for cigarettes sold in
812
Canada – 2004. Unpublished data received on request from [email protected].
813
Health Canada. Official method: determination of benzo (a) pyrene in mainstream tobacco smoke.
814
No. T103
815
Health Canada. Official method: determination of TSNAs in mainstream tobacco smoke. No. T111
816
Hyodo T., Inoue O., Katagiri H., Mikita A., Fujiwara M. Long-term inter-laboratory comparisons of
817
selected analytes in 2R4F mainstream smoke. 2006 CORESTA Conference Paper SS7.
AC C
EP
TE D
M AN U
SC
804
818
36
ACCEPTED MANUSCRIPT International Organization for Standardization. ISO 5725-2:1994. Accuracy (trueness and precision)
820
of measurements methods and results – part 2. Basic method for determination of repeatability and
821
reproducibility of a standard measurement method. ISO, Geneva, 1994.
822
International Organization for Standardization. ISO 4387:2000 (E). Cigarettes - Determination of
823
Total and Dry Particulate Matter Using a Routine Analytical Smoking Machine.
824
International Organization for Standardization. ISO 10315:2000 (E). Cigarettes - Determination of
825
Nicotine in Smoke Condensates - Gas Chromatographic Method
826
International Organization for Standardization. ISO 8454:2007 (E). Cigarettes - Determinations of
827
Carbon Monoxide in the Vapour Phase of Smoke - NDIR Method
828
International Organization for Standardization. ISO 3308:2012. Routine analytical cigarette-smoking
829
machine — Definitions and standard conditions. ISO, Geneva, 2012.
830
International Organization for Standardization. ISO 8243:2013. Cigarettes – Sampling. ISO, Geneva,
831
2013.
832
Intorp, M., Purkiss, S., Whittaker, M., Wright, W., Determination of ‘Hoffmann Analytes’ in cigarette
833
mainstream smoke. The CORESTA 2006 joint experiment. Beit. Tabakforsch. Intl. 2009 23:161-202
SC
M AN U
TE D
EP AC C
834
RI PT
819
835
Intorp, M., Purkis, S.W., Wagstaff, W., 2011. Determination of Selected Volatiles in Cigarette
836
Mainstream Smoke. The CORESTA 2009 Collaborative Study and Recommended Method. Beitrage
837
zur Tabakforsch., 24(5), 243-251.
838 839
Morton, M.J., Laffoon, S.W. Cigarette smoke chemistry market maps under Massachusetts
840
Department of Public Health smoking conditions. Regul. Toxicol. Pharmacol. 2008 51:1-30.
37
ACCEPTED MANUSCRIPT Oldham, M.J., Desoi, D.J., Rimmer, L.T., Wagner, K.A., Morton, M.J., Insights from analysis for
842
harmful and potentially harmful constituents (HPHCs) in tobacco products. Regul. Toxicol. Pharmacol.
843
2014 70 (1) 138-48, doi 10.1016/j.yrtph.2014.06.017.
844
Purkis S., Intorp M. Analysis of reference cigarette smoke yield data from 21 laboratories for 28
845
selected analytes as a guide to selection of new CORESTA Recommended Methods; Beit.
846
Tabakforsch. Intl. 2014 26(2): 57-73.
847
Rodgman A, Perfetti TA: The Chemical Components of Tobacco and Tobacco Smoke. CRC Press: New
848
York; 2013.
849
Shepperd C.J., Eldridge A., Camacho O.M., McAdam K., Proctor C. Changes in levels of biomarkers of
850
exposure observed in a controlled study of smokers switched from conventional to reduced toxicant
851
prototype cigarettes. Regul. Toxicol. Pharmacol. 66 (2013) 147–162
852
Taiwan, 2010, http://www.health99.doh.gov.tw/box2/smokefreelife/law.aspx, accessed 14/6/2010
853
Teillet, B., Cahours, X., Verron, T., Colard, S., Purkis, S., Comparison of smoke yield data collected
854
from different laboratories; Beit. Tabakforsch. Intl. 2013 25(8): 662-670.
855 856
World Health Organization. WHO Framework Convention on Tobacco Control. WHO Geneva 2005. ISBN 978 92 4 159101 0; whqlibdoc.who.int/publications/2003/9241591013.pdf
857
World Health Organization. The Scientific Basis of Tobacco Product Regulation. WHO Technical
858
Report Series 951. WHO, Geneva, 2008.
859
World Health Organization. WHO report on the global tobacco epidemic, 2011: warning about the
860
dangers of tobacco. WHO Geneva 2011
861
World Health Organisation. Report on the Scientific Basis of Tobacco Product Regulations: Fifth
862
Report of a WHO Study Group. WHO Technical Report Series 989. WHO, Geneva, 2015.
AC C
EP
TE D
M AN U
SC
RI PT
841
38
ACCEPTED MANUSCRIPT 863
World Health Organization (2016a). Tobacco- Fact Sheet No339, June 2016.
864
http://www.who.int/mediacentre/factsheets/fs339/en/ (accessed 16
865
January 2017). World health Organisation (2016b). http://www.who.int/fctc/signatories_parties/en/ (accessed 16
867
January 2017)
RI PT
866
868
AC C
EP
TE D
M AN U
SC
869
39
ACCEPTED MANUSCRIPT Figure Captions
871 872 873
Figure 1. PCA of global product data. (A) Score plot for PC1 and PC2– crosses indicate individual products in the database. Inset shows the percentage variation accounted for by the first four principal components. (B) Loadings plot for PC1 and PC2.
874 875
Figure 2. Identification of four diverse market scenarios. (A) Brazil; (B) Romania, (C) Australia, and (D) Germany. Red crosses, products from market of interest; grey crosses, remaining market data.
876 877 878
Figure 3. Proportion of non-compliant products by market using market-specific ceilings (model 1). (A) Separate application to toxicants (B) Simultaneous application of toxicant limits (C) Number of non-compliant toxicants per product.
879
Figure 4. Percentage of non-compliant products in each market using TobReg set limits (model 2)
880
A) ‘International brands’ limits B) ‘Canadian brands’ limits.
881
Figure 5: Percentage of non-compliant products in each market using model 3
882 883 884 885
Figure 6. Comparison of German market product data categorised by compliance to TobReg limit (market median of NNN-to-nicotine emissions calculated via model 1): A) NNN-to-nicotine, reference line showing market median B) NNN emissions, reference lines showing highest level for compliant and lowest level for non-compliant products and C) nicotine emissions.
886 887
Figure 7. Toxicant emissions for German market survey products categorised by compliance to TobReg calculated limits (model 1)
888 889 890 891
Figure 8. Rank order of the 339 German market products for NNK-to-nicotine emissions. Effect of the median limit, as proposed by TobReg model 1 [WHO 2008], and the potential error associated with toxicant ratio determination. Considerably fewer products fall within the variability of the analysis.
892 893 894 895 896
Figure 9. Toxicant-to-nicotine ratio emissions in the Romanian market survey. (A) Rank order of products for CO. (B) Rank order of products for 1,3-butadiene. Graphs indicate both the TobReg proposed limit (model 1) based on 125% of the median and the potential error associated with toxicant ratio determination.
AC C
EP
TE D
M AN U
SC
RI PT
870
40
ACCEPTED MANUSCRIPT
Table 1. Market summary of smoke toxicant emissions (HCI) per cigarette and as a ratio to nicotine yield.
Nicotine
mg Brazil Romania Australia Germany
132 138 172 339
1.52 1.80 1.93 1.88
0.25 0.32 0.44 0.33
per cigarette Lower Quartile 0.82 1.39 1.11 1.57 0.00 1.65 1.12 1.65
Tar
mg Brazil Romania
132 138
24.1 25.0
3.6 4.2
13.5 15.2
22.0 22.0
24.1 24.2
26.4 27.8
34.5 33.5
132 138
16.0 13.9
2.5 1.1
12.2 10.8
14.3 13.3
15.2 13.9
17.2 14.6
27.6 17.6
Australia Germany
172 339
27.0 27.8
5.1 4.1
16.7 17.2
23.4 24.6
27.2 27.7
29.7 31.2
61.5 37.5
170 339
a
14.0 15.0
1.6 2.1
9.2 9.3
13.0 13.9
14.0 14.8
15.0 16.0
19.7 24.0
Brazil Romania
132 138
138 150
51 78
34 39
105 104
139 132
163 173
392 496
132 138
94 85
47 45
16 17
70 59
90 74
112 97
479 283
Australia Germany
172 339
68 144
57 66
0 11
30 100
51 134
83 181
338 424
170 339
a
37 79
32 41
7 5
16 55
27 72
45 96
181 294
Brazil Romania
132 138
155 100
96 34
35 37
91 77
122 96
194 120
670 183
132 138
Australia Germany
172 339
54 95
36 47
0 13
27 61
40 88
75 118
178 318
170 339
Brazil Romania
132 138
18.7 13.1
4.8 3.9
7.8 5.7
15.3 10.4
18.2 12.7
21.8 15.1
38.3 27.2
132 138
Australia Germany
172 339
18.0 19.9
4.5 5.1
9.0 9.4
14.7 16.1
17.8 19.1
20.4 22.9
40.9 36.4
170 339
Brazil Romania
132 138
92 89
19 21
48 32
78 76
91 89
102 103
143 154
132 138
Australia Germany
172 339
119 99
32 27
60 42
97 82
112 97
138 114
Brazil Romania
132 138
1236 1167
163 226
593 626
1150 972
1263 1188
1343 1316
Australia Germany
172 339
1288 1374
122 191
932 854
1223 1264
1296 1367
1362 1509
Brazil Romania
132 138
129 117
17 23
59 61
122 99
130 116
141 136
ng
Formaldehyde µg
Acetaldehyde µg
Acrolein
µg
1,3-Butadiene µg
Benzene
CO
µg
Min
1.51 1.75 1.92 1.85
Upper Quartile 1.66 2.04 2.24 2.09
2.12 2.55 3.58 3.21
Median
Australia Germany
172 339
136 146
17 21
78 81
126 132
136 146
Brazil Romania
132 138
71 99
15 20
34 60
61 84
68 102
Australia Germany
172 339
107 104
13 24
67 54
Brazil Romania
132 138
85 75
14 18
37 42
Australia Germany
172 339
88 94
11 17
56 55
mg Brazil Romania
132 138
23.1 23.4
3.5 5.2
11.1 11.9
Australia Germany
172 339
25.8 25.8
3.0 4.9
18.5 13.9
Max
N
Mean
Std Dev
a
a
per mg nicotine Lower Min Quartile
Median
Upper Quartile
Max
RI PT
Std Dev
SC
Mean
M AN U
B[a ]P
ng
N
109 57
93 20
17 18
61 42
78 55
124 69
818 110
29 53
20 31
7 6
14 34
21 46
42 63
125 221
12.5 7.2
3.7 1.5
6.5 4.0
10.1 6.2
11.5 7.2
13.7 8.1
28.4 13.5
9.3 10.8
1.2 2.9
6.4 5.4
8.4 8.8
9.0 10.4
10.0 12.0
13.1 23.0
61 51
13 16
35 26
52 40
59 49
69 56
107 101
TE D
NNK
ng
Market
253 204
170 339
a
62 54
14 16
30 17
51 42
61 52
71 63
104 126
1563 1803
132 138
829 655
168 104
495 394
732 587
801 658
907 721
1847 915
1629 2567
170 339
a
683 749
135 151
328 385
592 666
671 739
767 829
1239 1762
172 169
132 138
86 66
16 11
50 42
77 58
85 66
94 74
170 100
147 158
187 295
a
170 339
72 80
15 17
43 42
62 70
69 78
80 87
147 180
81 113
116 145
132 138
47 55
12 10
27 36
39 49
44 55
54 61
97 92
a
EP
NNN
Unit
AC C
Toxicant
98 94
107 104
116 114
134 385
170 339
57 57
11 16
35 31
48 49
54 55
64 62
91 265
79 62
85 77
93 85
118 116
132 138
56 42
11 7
31 28
50 36
55 42
62 46
93 62
a
81 84
89 95
95 104
135 262
170 339
46 51
8 12
31 30
41 44
45 50
52 56
74 180
21.7 19.2
23.3 23.4
25.3 27.1
30.1 39.1
132 138
15.5 13.0
3.2 2.1
9.4 7.9
13.6 12.0
15.2 13.0
17.1 14.5
30.2 18.1
35.4 72.1
a
13.6 14.1
2.4 3.6
6.9 6.4
11.9 12.2
13.4 13.9
15.2 15.7
20.4 49.5
24.1 22.9
25.6 25.6
27.9 29.0
170 339
two herbal products with zero nicotine emissions excluded
b
minimum nicotine value for Australian tobacco products was 1.14 mg/cigarette
c
minimum NNN value for Australian tobacco products was 13.6 ng/cigarette
d
minimum NNK value for Australian tobacco products was 17.1 ng/cigarette
AC C
EP
TE D
M AN U
SC
a
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
Table 2. Market-specific toxicant emissions limits calculated as per the TobReg model 1, compared to set limits from model 2.
RI PT
Germany US Blended 924 97 63 13 69 17.3 65 46 72
SC
µg µg µg ng µg mg µg ng ng
M AN U
Acetaldehyde Acrolein Benzene B[a]P 1,3-Butadiene CO Formaldehyde NNK NNN
Market specific limits Brazil Romania Australia Mixed (low charcoal) Mixed (High charcoal) Flue-Cured 1001 822 842 106 82 87 69 53 57 14 9 11 54 69 68 19.0 16.2 16.7 74 61 76 78 55 21 90 74 27
AC C
EP
*TobReg's set limits identified by TobReg as part of model 2.
TE D
Toxicant (per mg nicotine)
TobReg set limits* International Canadian Brands Brands 860 670 83 97 48 50 11 11 67 53 18.4 15.4 47 97 72 47 114 27
ACCEPTED MANUSCRIPT
RI PT
Table 3: Limits calculated according to model 3
Per cigarette
Brazil
Romania
Australia
Germany
Highest/lowest Limit ratio
Acetaldehyde Acrolein Benzene B[a]P 1,3-Butadiene CO Formaldehyde NNK NNN
µg µg µg ng µg mg µg ng ng
1780 178 110 23.3 103 33.3 145 103.8 101.4
1878 184 127 17.0 179 35.6 95.2 110.1 118.1
2797 234 169 26.9 202 55.5 181 51.2 40.8
2561 243 164 28.2 163 41.8 124 37.9 32.3
1.6 1.4 1.5 1.6 2.0 1.7 1.9 2.7 3.5
AC C
EP
TE D
M AN U
SC
Toxicant
ACCEPTED MANUSCRIPT
TobReg model 1, Market Specific limits
72
93
Brazil
77
100
Germany
79
89
Romania
78
71
AC C
EP
TE D
Australia
TobReg model 3, 3x lowest value
“Canadian Brands”
SC
“International Brands”
M AN U
Country
TobReg model 2, fixed limits
RI PT
Table 4: % non-compliance rates for cigarette brands, by market, for each of the different regulatory models
80
68
100
86
97
99
98
80
ACCEPTED MANUSCRIPT Table 5. Correlations among absolute toxicant emission yields: matrix of Pearson correlation coefficients (r). Strong correlations (>│0.6│) are shown in bold. Negative correlations are highlighted in red. Tar
Acetaldehyde
Acetaldehyde Acetaldehyde Acetaldehyde Acetaldehyde
Brazil Romania Australia Germany
0.326 0.631 0.305 0.284
0.658 0.800 0.510 0.644
Acrolein Acrolein Acrolein Acrolein
Brazil Romania Australia Germany
0.369 0.570 0.332 0.180
0.647 0.714 0.385 0.492
0.758 0.929 0.583 0.777
Benzene Benzene Benzene Benzene
Brazil Romania Australia Germany
0.444 0.696 0.584 0.395
0.501 0.815 0.627 0.536
0.696 0.868 0.645 0.696
B[a ]P B[a ]P B[a ]P B[a ]P
Brazil Romania Australia Germany
0.280 0.738 0.830 0.373
0.504 0.823 0.659 0.505
1,3-Butadiene 1,3-Butadiene 1,3-Butadiene 1,3-Butadiene
Brazil Romania Australia Germany
0.300 0.646 0.398 0.398
0.269 0.732 0.377 0.448
CO CO CO CO
Brazil Romania Australia Germany
0.307 0.689 0.460 0.244
0.731 0.865 0.672 0.634
Formaldehyde Formaldehyde Formaldehyde Formaldehyde
Brazil Romania Australia Germany
0.338 0.008 0.502 0.193
NNK NNK NNK NNK
Brazil Romania Australia Germany
-0.331 0.145 -0.037 -0.113
NNN NNN NNN NNN
Brazil Romania Australia Germany
Toxicant Benzene B[a ]P
1,3-Butadiene
CO
Formaldehyde
0.571 0.799 0.565 0.493 0.336 0.686 0.400 0.158
0.544 0.813 0.629 0.334
0.403 0.835 0.570 0.604
0.374 0.740 0.390 0.528
0.837 0.936 0.539 0.677
0.340 0.784 0.337 0.121
0.885 0.925 0.809 0.839
0.656 0.796 0.468 0.675
0.731 0.890 0.715 0.774
0.566 0.793 0.402 0.203
0.460 0.847 0.561 0.598
0.480 0.187 0.517 0.250
0.140 0.167 0.162 0.138
0.524 0.299 0.463 0.409
0.055 0.040 0.428 0.129
0.018 0.044 0.487 0.111
0.019 -0.094 0.230 0.197
0.071 0.064 0.113 0.123
0.325 0.367 0.203 0.337
0.377 0.453 0.066 0.303
0.204 0.425 -0.201 0.060
0.009 0.304 0.019 0.061
0.469 0.267 -0.050 0.251
-0.120 0.235 -0.320 -0.024
0.377 0.469 0.244 0.311
-0.014 0.093 -0.346 -0.340
0.212 0.175 0.146 0.192
0.104 0.270 0.004 0.181
0.176 0.304 -0.210 -0.050
-0.184 0.224 -0.002 0.020
0.127 0.060 -0.096 0.088
-0.270 0.132 -0.270 -0.060
0.078 0.255 0.233 0.169
-0.035 -0.114 -0.388 -0.467
TE D
0.524 0.793 0.331 0.265
EP
AC C
-0.089 0.148 -0.043 -0.123
Acrolein
NNK
RI PT
Nicotine 0.592 0.880 0.798 0.624
SC
Market Brazil Romania Australia Germany
M AN U
Toxicant Tar Tar Tar Tar
0.593 0.475 0.864 0.749
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
Nicotine
mg
Brazil Romania Australia Germany
132 138 172 339
1.52 1.80 1.93 1.88
0.25 0.32 0.44 0.33
per cigarette Lower Min Quartile 0.82 1.39 1.11 1.57 0.00b 1.65 1.12 1.65
Tar
mg
Brazil Romania Australia Germany
132 138 172 339
24.1 25.0 27.0 27.8
3.6 4.2 5.1 4.1
13.5 15.2 16.7 17.2
22.0 22.0 23.4 24.6
24.1 24.2 27.2 27.7
26.4 27.8 29.7 31.2
34.5 33.5 61.5 37.5
NNN
ng
Brazil Romania Australia Germany
132 138 172 339
138 150 68 144
51 78 57 66
34 39 0c 11
105 104 30 100
139 132 51 134
163 173 83 181
392 496 338 424
NNK
ng
Brazil Romania Australia Germany
132 138 172 339
155 100 54 95
96 34 36 47
35 37 0d 13
91 77 27 61
122 96 40 88
194 120 75 118
B[a ]P
ng
Brazil Romania Australia Germany
132 138 172 339
18.7 13.1 18.0 19.9
4.8 3.9 4.5 5.1
7.8 5.7 9.0 9.4
15.3 10.4 14.7 16.1
18.2 12.7 17.8 19.1
Formaldehyde
µg
Brazil Romania Australia Germany
132 138 172 339
92 89 119 99
19 21 32 27
48 32 60 42
78 76 97 82
Acetaldehyde
µg
Brazil Romania Australia Germany
132 138 172 339
1236 1167 1288 1374
163 226 122 191
593 626 932 854
Acrolein
µg
Brazil Romania Australia Germany
132 138 172 339
129 117 136 146
17 23 17 21
1,3-Butadiene
µg
Brazil Romania Australia Germany
132 138 172 339
71 99 107 104
15 20 13 24
1.51 1.75 1.92 1.85
2.12 2.55 3.58 3.21
Median
Max
N
Mean
RI PT
Std Dev
Std Dev
per mg nicotine Lower Min Quartile
Median
Upper Quartile
Max
132 138 170a 339
16.0 13.9 14.0 15.0
2.5 1.1 1.6 2.1
12.2 10.8 9.2 9.3
14.3 13.3 13.0 13.9
15.2 13.9 14.0 14.8
17.2 14.6 15.0 16.0
27.6 17.6 19.7 24.0
132 138 170a 339
94 85 37 79
47 45 32 41
16 17 7 5
70 59 16 55
90 74 27 72
112 97 45 96
479 283 181 294
670 183 178 318
132 138 170a 339
109 57 29 53
93 20 20 31
17 18 7 6
61 42 14 34
78 55 21 46
124 69 42 63
818 110 125 221
21.8 15.1 20.4 22.9
38.3 27.2 40.9 36.4
132 138 170a 339
12.5 7.2 9.3 10.8
3.7 1.5 1.2 2.9
6.5 4.0 6.4 5.4
10.1 6.2 8.4 8.8
11.5 7.2 9.0 10.4
13.7 8.1 10.0 12.0
28.4 13.5 13.1 23.0
91 89 112 97
102 103 138 114
143 154 253 204
132 138 170a 339
61 51 62 54
13 16 14 16
35 26 30 17
52 40 51 42
59 49 61 52
69 56 71 63
107 101 104 126
1150 972 1223 1264
1263 1188 1296 1367
1343 1316 1362 1509
1563 1803 1629 2567
132 138 170a 339
829 655 683 749
168 104 135 151
495 394 328 385
732 587 592 666
801 658 671 739
907 721 767 829
1847 915 1239 1762
59 61 78 81
122 99 126 132
130 116 136 146
141 136 147 158
172 169 187 295
132 138 170a 339
86 66 72 80
16 11 15 17
50 42 43 42
77 58 62 70
85 66 69 78
94 74 80 87
170 100 147 180
34 60 67 54
61 84 98 94
68 102 107 104
81 113 116 114
116 145 134 385
132 138 170a 339
47 55 57 57
12 10 11 16
27 36 35 31
39 49 48 49
44 55 54 55
54 61 64 62
97 92 91 265
SC
Mean
M AN U
N
Upper Quartile 1.66 2.04 2.24 2.09
TE D
Market
EP
Unit
AC C
Toxicant
ACCEPTED MANUSCRIPT
µg
Brazil Romania Australia Germany
132 138 172 339
85 75 88 94
14 18 11 17
37 42 56 55
79 62 81 84
85 77 89 95
93 85 95 104
118 116 135 262
132 138 170a 339
56 42 46 51
11 7 8 12
31 28 31 30
50 36 41 44
55 42 45 50
62 46 52 56
93 62 74 180
CO
mg
Brazil Romania Australia Germany
132 138 172 339
23.1 23.4 25.8 25.8
3.5 5.2 3.0 4.9
11.1 11.9 18.5 13.9
21.7 19.2 24.1 22.9
23.3 23.4 25.6 25.6
25.3 27.1 27.9 29.0
30.1 39.1 35.4 72.1
132 138 170a 339
15.5 13.0 13.6 14.1
3.2 2.1 2.4 3.6
9.4 7.9 6.9 6.4
13.6 12.0 11.9 12.2
15.2 13.0 13.4 13.9
17.1 14.5 15.2 15.7
30.2 18.1 20.4 49.5
a
two herbal products with zero nicotine emissions excluded minimum nicotine value for Australian tobacco products was 1.14 mg/cigarette c minimum NNN value for Australian tobacco products was 13.6 ng/cigarette d minimum NNK value for Australian tobacco products was 17.1 ng/cigarette
AC C
EP
TE D
M AN U
SC
b
RI PT
Benzene
µg µg µg ng µg mg µg ng ng
Brazil Mixed (low charcoal) 1001 106 69 14 54 19.0 74 78 90
Romania Mixed (High charcoal) 822 82 53 9 69 16.2 61 55 74
Australia Flue-Cured 842 87 57 11 68 16.7 76 21 27
Germany US Blended 924 97 63 13 69 17.3 65 46 72
EP
TE D
M AN U
SC
*TobReg's set limits identified by TobReg as part of model 2
AC C
Acetaldehyde Acrolein Benzene B[a]P 1,3-Butadiene CO Formaldehyde NNK NNN
ACCEPTED MANUSCRIPT Market specific limits
TobReg set limits* International Canadian Brands Brands 860 670 83 97 48 50 11 11 67 53 18.4 15.4 47 97 72 47 114 27
RI PT
Toxicant (per mg nicotine)
ACCEPTED MANUSCRIPT
Toxicant Nicotine 0.592 0.880 0.798 0.624
Acetaldehyde
Acrolein
Benzene
1,3-Butadiene
CO
Formaldehyde
Acetaldehyde Acetaldehyde Acetaldehyde Acetaldehyde
Brazil Romania Australia Germany
0.326 0.631 0.305 0.284
0.658 0.800 0.510 0.644
Acrolein Acrolein Acrolein Acrolein
Brazil Romania Australia Germany
0.369 0.570 0.332 0.180
0.647 0.714 0.385 0.492
0.758 0.929 0.583 0.777
Benzene Benzene Benzene Benzene
Brazil Romania Australia Germany
0.444 0.696 0.584 0.395
0.501 0.815 0.627 0.536
0.696 0.868 0.645 0.696
0.571 0.799 0.565 0.493
B[a ]P B[a ]P B[a ]P B[a ]P
Brazil Romania Australia Germany
0.280 0.738 0.830 0.373
0.504 0.823 0.659 0.505
0.524 0.793 0.331 0.265
0.336 0.686 0.400 0.158
0.544 0.813 0.629 0.334
1,3-Butadiene 1,3-Butadiene 1,3-Butadiene 1,3-Butadiene
Brazil Romania Australia Germany
0.300 0.646 0.398 0.398
0.269 0.732 0.377 0.448
0.403 0.835 0.570 0.604
0.374 0.740 0.390 0.528
0.837 0.936 0.539 0.677
0.340 0.784 0.337 0.121
CO CO CO CO
Brazil Romania Australia Germany
0.307 0.689 0.460 0.244
0.731 0.865 0.672 0.634
0.885 0.925 0.809 0.839
0.656 0.796 0.468 0.675
0.731 0.890 0.715 0.774
0.566 0.793 0.402 0.203
0.460 0.847 0.561 0.598
Formaldehyde Formaldehyde Formaldehyde Formaldehyde
Brazil Romania Australia Germany
0.338 0.008 0.502 0.193
0.480 0.187 0.517 0.250
0.140 0.167 0.162 0.138
0.524 0.299 0.463 0.409
0.055 0.040 0.428 0.129
0.018 0.044 0.487 0.111
0.019 -0.094 0.230 0.197
0.071 0.064 0.113 0.123
NNK
Brazil
-0.331
0.325
0.377
0.204
0.009
0.469
-0.120
0.377
EP
TE D
M AN U
SC
RI PT
Market Brazil Romania Australia Germany
AC C
Tar
B[a ]P
Toxicant Tar Tar Tar Tar
-0.014
NNK
ACCEPTED MANUSCRIPT
Romania Australia Germany
0.145 -0.037 -0.113
0.367 0.203 0.337
0.453 0.066 0.303
0.425 -0.201 0.060
0.304 0.019 0.061
0.267 -0.050 0.251
0.235 -0.320 -0.024
0.469 0.244 0.311
0.093 -0.346 -0.340
NNN NNN NNN NNN
Brazil Romania Australia Germany
-0.089 0.148 -0.043 -0.123
0.212 0.175 0.146 0.192
0.104 0.270 0.004 0.181
0.176 0.304 -0.210 -0.050
-0.184 0.224 -0.002 0.020
0.127 0.060 -0.096 0.088
-0.270 0.132 -0.270 -0.060
0.078 0.255 0.233 0.169
-0.035 -0.114 -0.388 -0.467
AC C
EP
TE D
M AN U
SC
RI PT
NNK NNK NNK
0.593 0.475 0.864 0.749
ACCEPTED MANUSCRIPT Title: Impact assessment of WHO TobReg proposals for mandated lowering of selected mainstream cigarette smoke toxicants Figure 1:
B)
AC C
EP
TE D
M AN U
SC
RI PT
A)
ACCEPTED MANUSCRIPT Figure 2: A) Brazil:
M AN U
SC
RI PT
B) Romania:
C) Australia:
AC C
EP
TE D
D) Germany:
ACCEPTED MANUSCRIPT Figure 3:
90 80 70
RI PT
60 50 40 30 20 10 0
B)
M AN U
SC
Non-Compliant Product (%)
A) 100
100
Non-Compliant Product (%)
90 80 70 60
Brazil
50
Romania
40
TE D
Australia
30 20 10 0
C)
100
EP
All 9 Toxicants TSNAs only
Germany
Non-TSNAs only
Non-Compliant Product (%)
AC C
90 80 70 60 50 40 30 20 10 0
1
2
3
4
5
6
7
No. non-compliant toxicants
8
9
ACCEPTED MANUSCRIPT Figure 4: A) 100 80 70 60 Brazil
50
RI PT
Non-Compliant Product (%)
90
40
Romania
30
Australia
20
Germany
10
M AN U
SC
0
B) 100 80
50 40 30 20 10 0
EP
60
TE D
70
AC C
Non-Compliant Product (%)
90
Brazil Romania Australia Germany
ACCEPTED MANUSCRIPT Figure 5:
100 90 80
RI PT
70 60 50 40 30
SC
20 10
AC C
EP
TE D
M AN U
0
Australia Brazil Germany Romania
ACCEPTED MANUSCRIPT Figure 6:
A)
B) 450
300
C)
3.5
400 250 3.0
RI PT
200
100
Nicotine (mg/cig)
150
250 200 171 150
72
98
50 0 Non-compliant
Compliant
EP
TE D
Compliant
0
M AN U
100 50
2.5
2.0
SC
NNN (ng/cig)
300
AC C
NNN / Nicotine (ng/mg)
350
Non-compliant
1.5
1.0 Compliant
Non-compliant
ACCEPTED MANUSCRIPT Figure 7: Acetaldehyde
Acrolein
Benzene 240
1600
Benzene µg/cig
2400
Acrolein µg/cig
Acetaldehyde µg/cig
300
200
160
80
100 800 NC
C
B[a]P
NC
C
1,3-Butadiene
CO
270
150
20
10 C
NC
C
NC
NNK 300
150
60 0 C
NC
M AN U
180
NNK ng/cig
Formaldehyde µg/cig
Formaldehyde
120
C
EP
TE D
C – compliant product; NC – non-compliant product
AC C
40
NC
C
SC
20
60
CO mg/cig
1,3-Butadiene µg/cig
B[a]P µg/cig
390
30
NC
RI PT
C
NC
ACCEPTED MANUSCRIPT Figure 8.:
RI PT
200
150
50 Method Variability
0
50
100
M AN U
Product & Method Variability
0
150
200
Product Rank
TobReg Ceiling Method variability (±2CV reference cigarette)
EP
TE D
Product + Method variability (±2CV commercial cigarette)
62 55 Median 37 31
SC
100
AC C
Germany: NNK/Nicotine (ng/mg)
250
250
300
350
ACCEPTED MANUSCRIPT Figure 9: B)
A)
110
108
18 17.6 17.3
14
12
10
100
90 80
RI PT
15.1 14.9
1,3-Butadiene/N icotine
125% Median
16
70 60 50 40 30 20
40
60
80
100
120
0
140
20
40
60
80
Product Rank
Product Rank
M AN U
TobReg Ceiling Method variability (±2CV reference cigarette) Product + Method variability (±2CV commercial cigarette)
TE D
20
EP
0
SC
8
AC C
CO /N icotine
100
100
120
125% median
37 29
140
ACCEPTED MANUSCRIPT
RI PT
Highlights WHO TobReg proposals to reduce levels of 9 cigarette smoke toxicants were tested.
•
All available (80-97%) cigarette products from 4 diverse countries were analysed.
•
70-100% of cigarette products failed to meet the proposed WHO regulatory models.
•
These proposals would have greater impact on global markets than WHO’s stated aims.
AC C
EP
TE D
M AN U
SC
•