- Email: [email protected]
The relationship of snus use to diabetes and allied conditions
Accepted Manuscript The relationship of snus use to diabetes and allied conditions Peter N. Lee, Alison J. Thornton PII:
S0273-2300(17)30332-X
DOI:
10.1016/j.yrtph.2017.10.017
Reference:
YRTPH 3964
To appear in:
Regulatory Toxicology and Pharmacology
Received Date: 21 June 2017 Revised Date:
20 September 2017
Accepted Date: 19 October 2017
Please cite this article as: Lee, P.N., Thornton, A.J., The relationship of snus use to diabetes and allied conditions, Regulatory Toxicology and Pharmacology (2017), doi: 10.1016/j.yrtph.2017.10.017. 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
TITLE
The relationship of snus use to diabetes and allied
2
conditions
3 Peter N Lee1 and Alison J Thornton2
5 6 7 8 9 10
1
P.N. Lee Statistics and Computing Ltd., 17 Cedar Road, Sutton, Surrey SM2 5DA
2
Independent Consultant, Oak Cottage, Beer Farm, Okehampton, Devon, EX20 1SG
RI PT
Authors
SC
4
Corresponding author
Peter N Lee P.N. Lee Statistics and Computing Ltd., 17 Cedar Road, Sutton, Surrey SM2 5DA Tel: +44 (0) 20 8642 8265 Fax: +44 (0) 20 8642 2135 Email: Peter [email protected]
18 19 20 21
Word Counts
Abstract Text References
M AN U
11 12 13 14 15 16 17
AC C
EP
TE D
193 3754 729
(C:\pdfconversion\WORK\authordoc\Snus and diabetes_red
lined changes to text.docx)
ACCEPTED MANUSCRIPT Abstract A recent meta-analysis reported smoking to be associated with a 37% higher risk of type 2 diabetes in current smokers, rising to a 57% increase in heavy smokers, which declines on quitting. If the increase results from nicotine exposure, it is possible that
RI PT
using Swedish moist snuff (“snus”), which provides at least equivalent nicotine doses, might also increase diabetes risk. Following a recent publication reporting pooled results from five cohorts, we present a detailed meta-analysis of data from 18 studies.
SC
Based on covariate-adjusted estimates, no significant increased risk was seen in never
M AN U
smokers with RRs (95% CIs) of 1.08 (0.86-1.34), 0.93 (0.79-1.11) and 1.05 (0.941.18) for current, former and ever snus users. Significant increases were also not seen in the whole population, the corresponding RR estimates being 1.18 (0.94-1.48), 0.69 (0.49-0.96) and 0.95 (0.81-1.11). Nor was there an association of snus use with related endpoints, such as impaired glucose tolerance.
However, dose-response
TE D
analyses showed a relationship, with the highest levels of snus exposure associated with a diabetes RR of 1.65 (1.25-2.18) in never smokers. The evidence relating snus to type 2 diabetes is somewhat limited, requiring further studies to confirm any
AC C
EP
possible relationship.
Keywords Diabetes Tobacco Smokeless
ACCEPTED MANUSCRIPT
Introduction1
RI PT
1.
In 2011, in this journal, one of us (PNL) reviewed the epidemiological evidence relating snus to health (Lee, 2011), noting that there was “scant support for any major adverse health effect of snus”. Although it was noted that “snus users have increased weight” the review
SC
concluded that diabetes seems to be “unaffected”, and a lack of relationship with heart disease risk was also demonstrated. The section
M AN U
on diabetes cited 10 relevant publications (Attvall et al., 1993; Eliasson et al., 1995; Eliasson et al., 2004; Eliasson et al., 1991; Hergens et al., 2005; Johansson et al., 2005; Norberg et al., 2006; Persson et al., 2000; Wallenfeldt et al., 2001; Wändell et al., 2008) which noted that, despite occasional reports of significant associations, there was generally no clear or consistently observed relationship of snus use
TE D
with diabetes or a range of associated endpoints, such as glucose intolerance, fasting glucose levels, fasting insulin levels, or metabolic syndrome. However, meta-analyses were not attempted.
EP
Elsewhere, it has also been reported that the fasting blood glucose levels of snus users are not significantly different from those for never
1
AC C
users (Foulds et al., 2003). However, our conclusion that an association of snus with diabetes was “not clearly established” contrasts with BMI, body mass index; CI, confidence interval; ETS, environmental tobacco smoke; HR, hazard ratio; LADA, latent autoimmune diabetes of adulthood; OR, odds ratio; RR, relative risk; SES, socioeconomic status; T2D, type 2 diabetes; VIP, Våsterbotten Intervention Programme
Page 3 of 21
ACCEPTED MANUSCRIPT
that from a pooled analysis of results from five cohorts (Carlsson et al., 2017) which concluded that “high consumption of snus is a risk
RI PT
factor for type 2 diabetes”. This analysis estimated that, compared to never users the confounder-adjusted hazard ratio (HR) was 1.15
SC
(95% CI 1.00-1.32) in current users of snus, rising to 1.68 (1.17-2.41) in users of seven or more boxes per week.
M AN U
Preliminary inspection of the results of the analysis suggested that the significantly increased risk in current users derived principally from the results in one of the five cohorts considered, the Våsterbotten Intervention Programme (VIP), where the HR was 1.28 (95% confidence interval [CI] 1.08-1.52), no significant increase being seen in one of the other cohorts, and no increase at all in the other three, though two of these estimates were based on less than 10 diabetes cases in current users. However, the results for heavy consumers were
TE D
more consistent, with an elevated risk seen in all four cohorts with dose-response data, though statistically significantly only for the VIP.
EP
While Carlsson et al. (2017) refer to some of the publications we cited (Eliasson et al., 1995; Eliasson et al., 2004; Persson et al., 2000)
AC C
and to results of some more recent studies on diabetes (Östenson et al., 2012; Rasouli et al., 2016), they did not attempt a meta-analysis of the overall literature. The objective of this publication is to present such a meta-analysis.
Page 4 of 21
2.
Materials and methods
2.1
Literature searching
RI PT
ACCEPTED MANUSCRIPT
Relevant papers were derived from PubMed searches of (snuff OR snus OR smokeless tobacco) AND (diabetes OR glucose intolerance
SC
OR insulin) initially conducted on 14th February 2017, and updated on 5th September, 2017. Abstracts were studied initially, with papers
M AN U
of potential interest then looked at. Relevant reviews were also studied for relevant references as were reference lists in papers identified. The objective was to identify papers that presented evidence on the relationship of snus use, in never smokers and/or in the whole population, to diabetes or to associated endpoints, such as impaired glucose tolerance or fasting glucose or insulin resistance or response.
EP
Data extraction
For each study identified, information was extracted on the study name and location, relevant references, study design (prospective, casecontrol or cross-sectional), subjects included, sexes studied, endpoints studied and confounding variables adjusted for. For each endpoint
AC C
2.2
TE D
A similar Embase search was conducted on 8th September, looking for additional relevant papers.
for which results were available, relative risks (RRs), odds ratios (ORs) or HRs were extracted, or derived from data presented using standard methods (Gardner and Altman, 1989; Hamling et al., 2008; Lee et al., 2012). Estimates were extracted for current, former and ever snus users relative to never snus users, separately for never smokers and for the whole population. Where available, estimates were
Page 5 of 21
ACCEPTED MANUSCRIPT
extracted (or derived) adjusted for the maximum number of potential confounding variables for which results were available, and also
RI PT
unadjusted for any variable (for case-control and cross-sectional studies) or for age only (for prospective studies). Estimates were also extracted (or derived) by extent of exposure to snus. For diabetes in never smokers, estimates were also made of the increase in RR per
M AN U
SC
box used per week.
Exceptionally, some studies only reported results for continuous endpoints, such as blood glucose or insulin. Here the findings were
Meta-analyses
Meta-analyses were only carried out for diabetes, separately for current, former and ever snus use, based on results either for never
EP
smokers or the whole population. Meta-analysis was mainly restricted to effect estimates which were adjusted for confounding variables. Standard methods were used to carry out the meta-analyses (Fleiss and Gross, 1991) with fixed-effect and random-effects estimates
AC C
2.3
TE D
summarized from the relevant publications.
computed, together with a test for heterogeneity between individual estimates. Estimates were derived using all the available data for a particular analysis, and also for subsets of the data by time of publication (to 2010, 2011 onwards), study design (case-control or crosssectional, prospective), type of diabetes (type 2, unspecified), number of potential confounding variables adjusted for (<4, 5+), and
Page 6 of 21
ACCEPTED MANUSCRIPT
number of cases of diabetes in exposed subjects (<10, 11+ for analyses of current and former snus use, and <20, 21+ for analyses of ever
AC C
EP
TE D
M AN U
SC
RI PT
snus use). Publication bias was also investigated using Egger’s test (Egger et al., 1997).
Page 7 of 21
3.
Results
3.1.
Literature searches
RI PT
ACCEPTED MANUSCRIPT
SC
The first PubMed search produced 100 hits. Based on the abstracts, 83 were rejected as irrelevant. The remaining 17 papers were obtained for scrutiny, with four rejected as providing no additional primary data and one being a review. The remaining 12 papers were
M AN U
supplemented by an additional three publications identified in reviews or in reference lists of papers identified as relevant. The second PubMed search produced an additional five hits, four papers being rejected based on abstracts, with one recent relevant paper added to the 16 previously identified. The Embase search produced 179 hits, with only two additional papers obtained, both rejected as providing
EP
The studies
The 16 references described eighteen studies which provided evidence on the association between snus use and diabetes and related
AC C
3.2.
TE D
no additional primary data.
conditions. Table 1 gives details of the main characteristics of the studies, a more detailed description of each study being given in Appendix A. Two studies were published in the 1990s, seven between 2000 and 2010, and nine after 2011. Except for one study in Norway, all the studies were in Sweden. Seven studies were cross-sectional, while two were of a case-control design, both drawing
Page 8 of 21
ACCEPTED MANUSCRIPT
controls from the general population. The remaining nine studies were prospective, with follow-up ranging from 10 to 27 years. Fifteen
RI PT
studies were in males, while two involved both sexes. One study only included women for certain endpoints.
SC
Various endpoints were considered. Type 2 diabetes (T2D) was investigated in 12 studies, although one of these included impaired
M AN U
fasting glucose in their definition. A further three studies provided information on diabetes, but did not specify the type. Four studies used an endpoint of impaired glucose tolerance and/or pathological glucose tolerance, the latter being a combination of diabetes or impaired glucose tolerance. One study considered latent autoimmune diabetes of adulthood (LADA), and another study the metabolic
resistance and response levels.
TE D
syndrome. Blood glucose level was examined by four studies, with insulin level considered by three. One study also examined insulin
EP
Three studies (Attvall et al., 1993; Eliasson et al., 1991; Wallenfeldt et al., 2001), none of which provided information on diabetes, did
AC C
not adjust their results for any potential confounding variables. In the remaining 15 studies (see Table 2), all but three considered age, although one of these studies did carry out a series of analyses in which they evaluated various endpoints at different ages. Various other confounders were considered, the most common being waist circumference, body mass index, alcohol consumption, smoking status, physical activity, education and family history of diabetes.
Page 9 of 21
ACCEPTED MANUSCRIPT
RI PT
Diabetes
Table 3 summarizes the evidence relating snus use in never smokers to covariate-adjusted risk of type 2 diabetes. Corresponding
SC
unadjusted (or for prospective studies age-adjusted results) are given in Appendix B, Table B1. These tables also include overall meta-
M AN U
analysis results, with adjusted results separated by time of publication, study design, type of diabetes, number of confounding variables, and number of exposed cases shown in Appendix C, Table C1. In these tables, the effects estimates are referred to as RRs, though they may have been estimated as ORs or HRs.
TE D
It can be seen from Table 3 that a number of the individual study effect estimates are based on quite small numbers of snus-exposed cases, with, for example, five of the 11 estimates for current use based on less than 10 cases. Also, the number of snus-exposed cases in
EP
the VIP study form over half the total number in all the studies combined. None of the random-effect estimates, whether for current, former or ever snus use show significant (p<0.05) evidence of an effect, with the combined estimates being, respectively, 1.08 (95% CI
AC C
3.3.
0.86-1.34), 0.93 (0.79-1.11) and 1.05 (0.94-1.18). Nor was there any significant evidence of heterogeneity between estimates, despite two of the estimates for current use (from the SDPP1 and VIP studies) being statistically significantly increased, or of publication bias (data not shown). There was an indication that case-control or cross sectional studies produced somewhat higher results for current snus
Page 10 of 21
ACCEPTED MANUSCRIPT
use, with the overall estimate for these four studies being 1.40 (0.86-2.31), compared to 1.14 (0.99-1.30) for the seven prospective studies
RI PT
(see Table C1). However, this difference was not statistically significant (p = 0.42), and was scarcely evident for former or ever snus use. There was no evidence that estimates varied significantly by year of publication, type of diabetes, number of confounding variables
SC
considered, or number of exposed cases. The overall unadjusted estimates (see Table B1) tended to be somewhat lower, at 0.93 (0.82-
M AN U
1.05), 0.65 (0.39-1.09) and 0.69 (0.52-0.94) for, respectively, current, former and ever snus use. The low estimates for former and ever snus use, and the substantial heterogeneity for these estimates, are largely due to particularly low unadjusted estimates from the ESTRID study.
TE D
Tables 4, B2 and C2 similarly summarize the evidence relating snus use in the whole population to risk of type 2 diabetes. The numbers of snus-exposed cases shown in Table 4 remain quite small, except in the ESTRID and HUNT studies. Again, none of the overall
EP
random-effect estimates, 1.18 (95% CI 0.94-1.48) for current use, 0.69 (0.49-0.96) for former use and 0.95 (0.81-1.11) for ever use show
AC C
any significant evidence of an increased risk with snus use, the estimate for former use indicating a significantly decreased risk, though relying primarily on the results from the ESTRID study. There was no significant evidence of heterogeneity for current, former or ever use. The tests for publication bias were not significant except for current use, where five estimates above 1.0 from relatively small studies were countered by a slightly reduced estimate from the ESTRID study. Results based on case-control or cross-sectional studies
Page 11 of 21
ACCEPTED MANUSCRIPT
did not vary significantly from those based on prospective studies, for current, ex- or ever snus use (see Table C2). There was also little
RI PT
indication that the effect estimates varied by level of the other factors considered, though for ever use there was marginally significant (0.01
SC
studies, though these comparisons were not independent of each other. The overall unadjusted estimates (see Table B2) were here quite
M AN U
similar to the adjusted ones, being 1.16 (0.90-1.51), 0.82 (0.63-1.09) and 0.96 (0.71-1.30) for, respectively, current, former and ever use. The significant heterogeneity for ever use derives from the low effect estimate for the HUNT study.
Table 5 summarizes available dose-response results from 11 studies, eight presenting results for never smokers and five for the whole
TE D
population. The results presented are for various dose metameters – boxes/wk, cans/wk, box-years, duration (in years) and periods of follow-up – and for various dose levels. Significant increases, compared to never users, at the highest level of exposure are seen for
EP
never smokers in two studies (SDPP1 for 3+ boxes/wk and VIP for 7+ boxes/wk) and for the whole population in one study (SDDP2 for
AC C
5+ boxes/wk). Meta-analyses of the effect at the highest level were carried out both for never smokers and for the whole population, preferring estimates for boxes/wk, where a single study gave results for multiple dose metameters. For never smokers, the randomeffects analysis showed clear evidence of an association, with a combined estimate of 1.65 (95% CI 1.25-2.18) based on eight independent estimates and no significant heterogeneity. The largest contributors to the overall estimate were the estimates of 2.70 (1.30-
Page 12 of 21
ACCEPTED MANUSCRIPT
5.50) from the SDPP1 study and of 1.79 (1.14-2.80) from the VIP study, the combined estimate becoming non-significant, at 1.31 (0.87-
RI PT
1.98), based on the other six estimates. For the whole population, the combined evidence from the highest level of exposure was a nonsignificant 1.20 (0.71-2.04), based on four somewhat variable estimates (heterogeneity p = 0.0528). For diabetes in never smokers,
SC
where RRs are available by the number of boxes of snus used per week, Table 5 shows the estimated increase in RR per box per week.
M AN U
Based on these seven estimates, a combined fixed-effect estimate of 1.05 (95% CI 1.02-1.08) was derived. Although there was no significant heterogeneity between the individual increaes, the overall estimate was again heavily dependent on the increases seen in the
Other endpoints
The evidence relating snus use to other endpoints in never smokers which is summarized in Table 6 is sparse, but is not at all indicative
EP
of a positive relationship with snus use, with none of the ten estimates statistically significant, and only two of them above 1. Although one of the 17 estimates for other endpoints in the whole population shown in Table 7 shows a significant increase (RR 2.2, 95% CI 1.1-
AC C
3.4
TE D
SDPP1 and VIP studies.
4.4) this, for insulin response, is for former snus users, with no significant increase seen in this study for current users. The overall evidence shown in the Table is not suggestive of an effect of snus, with 12 of the 17 estimates below 1.
Page 13 of 21
ACCEPTED MANUSCRIPT
Not included in Tables 6 and 7 are results from four studies which reported results relating snus use to blood glucose and insulin levels.
RI PT
In the first study (Eliasson et al., 1991), snus use had no effect on blood glucose levels, but serum insulin was higher in snus users (5.5 mU/L) than in non-tobacco users (3.6 mU/L) and this difference was statistically significant (p<0.01). The second study (Attvall et al.,
SC
1993) also reported no difference in blood glucose levels or in fasting insulin levels before or after normoglycaemic clamps. The third
M AN U
study (Wallenfeldt et al., 2001) reported no significant difference in fasting blood glucose or plasma insulin between current, ex- or never snus users. The MONICA I study (Eliasson et al., 1995) found no differences between non-tobacco users and snus users in fasting glucose levels, fasting insulin or post-load insulin.
EP AC C
endpoints studied.
TE D
The limited dose-response evidence for other endpoints, shown in Table 8, provides no real indication of an effect of snus on any of the
Page 14 of 21
ACCEPTED MANUSCRIPT
RI PT
Discussion A recent meta-analysis of prospective studies relating smoking to diabetes risk (Pan et al., 2015) reported combined results from 88 prospective studies involving almost 300,000 incident cases of type 2 diabetes. Significant combined estimates were noted for current
SC
smokers (RR 1.37, 95% CI 1.33-1.42), former smokers (1.14, 1.10-1.18) and never smokers exposed to environmental tobacco smoke
M AN U
(1.22, 1.10-1.35). A dose-response relationship was also noted for current smoking, with an estimate of 1.57 (1.47-1.66) for heavy smokers. Risk estimates declined with time quit. Given that snus users have levels of nicotine exposure that are similar to, and perhaps somewhat greater than, that from smoking (Lee, 2011), the possibility arises that snus use might, if nicotine is relevant to risk of diabetes,
TE D
also increase risk of diabetes.
The evidence that snus use increases the risk of type 2 diabetes is clearly much weaker for snus than for smoking. Thus, there is no
EP
significant evidence of an increased risk, compared to never snus users, in current, former or ever snus users, whether restricting attention to never smokers or considering the whole population. Nor is there any clear evidence of an association of snus use with endpoints
AC C
4.
related to diabetes, such as glucose or insulin levels. The only real evidence for a possible effect of snus on diabetes arises from the dose-response data, where in never smokers (though not in the whole population) there is an increased RR of 1.65 (95% CI 1.25-2.18) for
Page 15 of 21
ACCEPTED MANUSCRIPT
the highest level of current snus use, and a RR of 1.05 (1.02-1.08) can be derived for the increase in RR per extra box of snus used per
RI PT
week.
SC
In considering the evidence on snus, various issues should be considered. Firstly, only 15 studies provide data on snus and diabetes, far
M AN U
less than the 88 studies considered in the recent review of smoking and diabetes (Pan et al., 2015). Furthermore, many of the estimates are based on rather small numbers of snus-exposed cases, with the VIP study on its own providing more than half the exposed cases for never smokers (see Table 3) and the ESTRID and HUNT studies contributing largely to the results for the whole population (see Table 4). With the single exception of the HUNT study in Norway, the studies are all conducted in Sweden, so that variation between countries
TE D
in the product used is unlikely to be an issue. Many, but not all of the studies, adjusted for a range of potential confounding variables, though little evidence was provided on which specific variables materially affected risk estimates. While 12 of the 15 studies provided
EP
results specifically for type 2 diabetes, three studies (SHEP/VHEP, SALLS and Stockholm County) only referred to diabetes. However,
AC C
there was no evidence that results varied significantly by definition of diabetes. Nor was there evidence that results varied by other factors considered, including time of publication, study design, number of potential confounding variables adjusted for, or number of cases in snus-exposed subjects. This suggests that it is reasonable to take the overall estimates shown in Tables 3 and 4 as indicative of the relationship between snus and diabetes. The extent to which these estimates might be affected by various biases is difficult to assess.
Page 16 of 21
ACCEPTED MANUSCRIPT
We found no evidence of publication bias using the Egger test, although this test has limited power when few studies are available. We
SC
though it seems doubtful whether these would have any major effect.
RI PT
were unable to formally take into account possible biases arising from misclassification of disease or of exposure (to snus or smoking),
M AN U
As all but one of the studies was conducted in Sweden, some overlap between studies is possible. However, we consider it unlikely to be is a major issue. Studies sharing a location tended to differ either in the time period during which they were conducted or the endpoint considered, or the population was more narrowly defined in some studies, meaning that only a small number of subjects were likely to have participated in multiple studies. Only for the SDPP and SHEP/VHEP studies does it appear possible that more significant overlap
TE D
may be present, as the studies included a similar age range and took place at the same time. SDPP I and II were limited to four municipalities in Stockholm county whereas SHEP/VHEP appears to include the whole county but there is no indication from the authors
EP
that the SDPP studies are a subset of SHEP/VHEP. There are enough differences between them in age range, study period and
AC C
geographical location to consider them as separate studies, any overlap probably being quite small.
The main evidence of a relationship of snus to diabetes comes from the dose-response analyses. Though, for the whole population, the combined estimate for the highest level of exposure was not significant (RR 1.39, 95% CI 0.79-2.46) this was based on four rather
Page 17 of 21
ACCEPTED MANUSCRIPT
variable estimates. Evidence of a dose-response was clearer for never smokers, with the combined RR estimate being 1.65 (95% CI 1.25-
RI PT
2.18), based on eight estimates, with no significant heterogeneity. As different studies reported dose-response results in different ways, four of the estimates were for 7+ boxes/wk, one for 5+ boxes/wk, two for 3+ boxes/wk, and one for 30+ years duration. Of these eight
M AN U
SC
estimates, seven were above 1.00, with two (from the SDPP1 and VIP studies) statistically significant.
Based on pooled results from the VIP, SPHC, SALT and MDCS studies, a recent review (Carlsson et al., 2017) reported a significant increased risk of type 2 diabetes for 7+ boxes/wk (RR 1.68, 95% CI 1.17-2.41) and 5-6 boxes/wk (1.42, 1.07-1.87), and concluded that “high consumption of snus is a risk factor for type 2 diabetes”, with a risk “similar to that in smokers” and supporting “the notion that
TE D
nicotine increases the risk of type 2 diabetes.” Although the findings of a dose-response in never smokers is indicative of a possible relationship of snus to diabetes risk, the general lack of association seen in the other analyses reported in this review suggests to us that
EP
more evidence is required before any firm conclusion can be reached. We note that the increased risk for 7+ boxes/wk reported in the
estimate of 1.49 (0.80-2.75).
AC C
review is heavily dependent on results from the VIP study – the combined results from the other three studies giving a non-significant
Page 18 of 21
ACCEPTED MANUSCRIPT
RI PT
Conclusions We found no evidence of an overall increased risk of type 2 diabetes in current, former or ever snus users, either in never smokers or the whole population. Nor is snus use associated with endpoints related to diabetes. There is some evidence of a dose-response relationship
EP
TE D
M AN U
SC
with diabetes risk, in never smokers, though not the whole population, but this requires confirmation from additional studies.
AC C
5.
Page 19 of 21
ACCEPTED MANUSCRIPT
RI PT
Conflict of interest P N Lee is a long-term consultant to the tobacco industry and A J Thornton is a consultant to P N Lee’s company, P N Lee Statistics and
SC
Computing Ltd. However, this is an independent scientific assessment, the views expressed being those of the author alone.
M AN U
Acknowledgements
We thank Mrs. Y. Cooper and Mrs. D. Morris for their assistance in typing the various drafts of this report, and their help in assembling the
AC C
EP
TE D
relevant literature. We also thank colleagues for providing helpful comments, and Swedish Match for financial support.
Page 20 of 21
ACCEPTED MANUSCRIPT
RI PT
References
Attvall, S., Fowelin, J., Lager, I., Von Schenck, H., Smith, U., 1993. Smoking induces insulin resistance - a potential link with the insulin resistance syndrome. J. Intern. Med. 233, 4, 327-332.
M AN U
SC
Byhamre, M.L., Gustafsson, P.E., Jansson, J.H., Wennberg, M., Hammarstrom, A., Wennberg, P., 2017. Snus use during the life-course and risk of the metabolic syndrome and its components. Scand. J. Public Health. Epub ahead of print Jul 1, 1403494817706631. DOI:10.1177/1403494817706631. Carlsson, S., et al., 2017. Smokeless tobacco (snus) is associated with an increased risk of type 2 diabetes: results from five pooled cohorts. J. Intern. Med. 281, 4, 398-406. DOI:10.111/joim.12592.
TE D
Egger, M., Davey Smith, G., Schneider, M., Minder, C., 1997. Bias in meta-analysis detected by a simple, graphical test. BMJ. 315, 629-634. DOI:10.1136/bmj.315.7109.629. Eliasson, M., Asplund, K., Evrin, P.-E., Lundblad, D., 1995. Relationship of cigarette smoking and snuff dipping to plasma fibrinogen, fibrinolytic variables and serum insulin. The Northern Sweden MONICA study. Atherosclerosis. 113, 41-53.
EP
Eliasson, M., Asplund, K., Nasic, S., Rodu, B., 2004. Influence of smoking and snus on the prevalence and incidence of type 2 diabetes amongst men: the northern Sweden MONICA study. J. Intern. Med. 256, 101-110.
AC C
Eliasson, M., Lundblad, D., Hägg, E., 1991. Cardiovascular risk factors in young snuff-users and cigarette smokers. J. Intern. Med. 230, 1, 1722. Fleiss, J.L., Gross, A.J., 1991. Meta-analysis in epidemiology, with special reference to studies of the association between exposure to environmental tobacco smoke and lung cancer: a critique. J. Clin. Epidemiol. 44, 2, 127-139. DOI:10.1016/0895-4356(91)90261-7.
Page 21 of 21
ACCEPTED MANUSCRIPT
RI PT
Foulds, J., Ramstrom, L., Burke, M., Fagerström, K., 2003. Effect of smokeless tobacco (snus) on smoking and public health in Sweden. Tob. Control. 12, 349-359. Gardner, M.J., Altman, D.G. (Eds.), 1989. Statistics with confidence. Confidence intervals and statistical guidelines. British Medical Journal, London.
M AN U
SC
Gustafsson, P.E., Persson, M., Hammarström, A., 2011. Life course origins of the metabolic syndrome in middle-aged women and men: the role of socioeconomic status and metabolic risk factors in adolescence and early adulthood. Ann. Epidemiol. 21, 2, 103-10. DOI:10.1016/j.annepidem.2010.08.012. Hamling, J., Lee, P., Weitkunat, R., Ambühl, M., 2008. Facilitating meta-analyses by deriving relative effect and precision estimates for alternative comparisons from a set of estimates presented by exposure level or disease category. Stat. Med. 27, 954-970. DOI:10.1002/sim.3013.
TE D
Hergens, M.-P., Ahlbom, A., Andersson, T., Pershagen, G., 2005. Swedish moist snuff and myocardial infarction among men. Epidemiology. 16, 1, 12-16. Johansson, S.-E., Sundquist, K., Qvist, J., Sundquist, J., 2005. Smokeless tobacco and coronary heart disease a 12-year follow-up study. Eur. J. Cardiovasc. Prev. Rehabil. 12, 4, 387-392.
EP
Lee, P.N., 2011. Summary of the epidemiological evidence relating snus to health. Regul. Toxicol. Pharmacol. 59, 2, 197-214. DOI:10.1016/j.yrtph.2010.12.002.
AC C
Lee, P.N., Forey, B.A., Coombs, K.J., 2012. Systematic review with meta-analysis of the epidemiological evidence in the 1900s relating smoking to lung cancer. BMC Cancer. 12, 385. DOI:10.1186/1471-2407-12-385. Neumann, A., Norberg, M., Schoffer, O., Norstrom, F., Johansson, I., Klug, S.J., Lindholm, L., 2013. Risk equations for the development of worsened glucose status and type 2 diabetes mellitus in a Swedish intervention program. BMC Public Health. 13, 1014. DOI:10.1186/1471-2458-13-1014.
Page 22 of 21
ACCEPTED MANUSCRIPT
RI PT
Norberg, M., Stenlund, H., Lindahl, B., Boman, K., Weinehall, L., 2006. Contribution of Swedish moist snuff to the metabolic syndrome: a wolf in sheep's clothing? Scand. J. Public Health. 34, 6, 576-583. Östenson, C.G., Hilding, A., Grill, V., Efendic, S., 2012. High consumption of smokeless tobacco ("snus") predicts increased risk of type 2 diabetes in a 10-year prospective study of middle-aged Swedish men. Scand. J. Public Health. 40, 8, 730-737.
SC
Pan, A., Wang, Y., Talaei, M., Hu, F.B., Wu, T., 2015. Relation of active, passive, and quitting smoking with incident type 2 diabetes: a metaanalysis and systematic review. Lancet Diabetes Endocrinol. 3, 12, 958-67. DOI:10.1016/s2213-8587(15)00316-2.
M AN U
Persson, P.G., Carlsson, S., Svanström, L., Östenson, C.G., Efendic, S., Grill, V., 2000. Cigarette smoking, oral moist snuff use and glucose intolerance. J. Intern. Med. 248, 2, 103-110. Rasouli, B., et al., 2016. Use of Swedish smokeless tobacco (snus) and the risk of Type 2 diabetes and latent autoimmune diabetes of adulthood (LADA). Diabet. Med. 34, 4, 514-521. DOI:10.1111/dme.13179.
TE D
Wallenfeldt, K., Hulthe, J., Bokemark, L., Wikstrand, J., Fagerberg, B., 2001. Carotid and femoral atherosclerosis, cardiovascular risk factors and C-reactive protein in relation to smokeless tobacco use or smoking in 58-year-old men. J. Intern. Med. 250, 492-501.
AC C
EP
Wändell, P.E., Bolinder, G., de Faire, U., Hellénius, M.-L., 2008. Association between metabolic effects and tobacco use in 60-year-old Swedish men. Eur. J. Epidemiol. 23, 6, 431-434.
Page 23 of 21
ACCEPTED MANUSCRIPT
Table 1:
Details of studies examining the risk of diabetes and snus use Yeara
Location
Study typeb
Subjects
Sexc
No. of casesd
Umeå (Eliasson et al., 1991)
1991
Sweden, Umeå
CS
University students and teachers/ respondents to advertisements
M
-
Göteborg (Attvall et al., 1993)
1993
Sweden, Göteborg
CS
Healthy current smokers
M+F
-
SDPP I (Persson et al., 2000)
2000
Sweden, Stockholm
CS
General population
M
52
AIR (Wallenfeldt et al., 2001)
2001
Sweden, Göteborg
CS
General population
M
-
MONICA I (Eliasson et al., 1995; Eliasson et al., 2004)
2004
Sweden, Norrbotten and Våsterbotten
CS
General population
M
127
MONICA II (Eliasson et al., 2004)
2004
SHEP/VHEP (Hergens et al., 2005)
2005
SALLS (Johansson et al., 2005)
M AN U
SC
RI PT
Study name (ref)
P (13)
General population
M
38
Sweden, Stockholm and Våsternorrland
CC
General population
M
92
2005
Sweden
P (12)
General population
M
53
Stockholm County (Wändell et al., 2008)
2008
Sweden, Stockholm county
CS
60 year olds
M
78
SDPP II (Östenson et al., 2012)
2012
Sweden, Stockholm County
P (12)
General population
M
99
ESTRID (Rasouli et al., 2016)
2016
Sweden, Scania and Uppsala
CC
General population
M
724
HUNT (Rasouli et al., 2016)
2016
Norway, NordTrøndelag county
CS
General population
M
829
VIP (Carlsson et al., 2017; Neumann et al., 2013; Norberg et al., 2006)
2017
Sweden, Våsterbotten County
P (21)
General population
Mg
1033
SPHC (Carlsson et al., 2017)
2017
Sweden, Stockholm County
P (10)
General population
M
237
SALT (Carlsson et al., 2017)
2017
Sweden, nationwide
P (12)
Twins
M
405
MDCS (Carlsson et al., 2017)
2017
Sweden, Malmö
P (21)
General Population
M
461
AC C
EP
TE D
Sweden, Norrbotten and Våsterbotten
Page 1 of 7
ACCEPTED MANUSCRIPT Study name (ref)
Yeara
Location
NMC (Carlsson et al., 2017)
2017
Sweden, nationwide
NSC (Byhamre et al., 2017; Gustafsson et al., 2011)
2017
Sweden, Luleå
g h
M+F
21h
Year of most relevant publication CC = Case control, CS = Cross-sectional, P = Prospective (length of follow-up in years) F = Female, M = Male Total exposed and non-exposed cases for diabetes DTNS = diabetes type not specified, T2D = Type 2 diabetes BGL = blood glucose, IFG = impaired fasting glucose, IGT = impaired glucose tolerance, INP = plasma insulin, INS = serum insulin, IRESIS = insulin resistance, IRESP = insulin response, LADA = latent autoimmune diabetes of adulthood, MS = metabolic syndrome, PGT = pathological glucose tolerance (defined as diabetes or IGT at oral glucose tolerance test) Analyses for IFG and IGT are based on male and female participants Total number of cases for whole study, includes current and ex- cigarette smokers
RI PT
f
Adolescents in 9th grade
SC
e
M
M AN U
d
Participants in fundraising event
No. of casesd 305
TE D
c
Sexc
EP
b
P (27)
Subjects
AC C
a
Study typeb P (13)
Page 2 of 7
ACCEPTED MANUSCRIPT Table 2:
Confounding variables used in the studies Age
BMI
Physical activity
SDPP I (Persson et al., 2000)
Yes
Yes
Yes
MONICA I (Eliasson et al., 1995; Eliasson et al., 2004)
Yes
MONICA II (Eliasson et al., 2004)
Yes
SHEP/VHEP (Hergens et al., 2005)
Yes
Yes
ESTRID (Rasouli et al., 2016)
Yes
HUNT (Rasouli et al., 2016)
Yes
VIP (Carlsson et al., 2017; Neumann et al., 2013; Norberg et al., 2006)
Yes
Yes
Alcohol
Family history of diabetes
Yes
Yes
Year
Yes
Yes Yes
Yes
Yes
Yes
Yes
Yes
Yes
TE D
SDPP II (Östenson et al., 2012)
Yes
M AN U
Yes
SC
SALLS (Johansson et al., 2005) Stockholm County (Wändell et al., 2008)
Education
RI PT
Study name (ref)
Yes
Yes
Yes
Yes
SPHC (Carlsson et al., 2017)
Yes
Yes
Yes
Yes
Yes
Yes
SALT (Carlsson et al., 2017)
Yes
Yes
Yes
Yes
Yes
Yes
MDCS (Carlsson et al., 2017)
Yes
Yes
Yes
Yes
Yes
Yes
NMC (Carlsson et al., 2017)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
BMI = body mass index, SES = socioeconomic status
AC C
Note:
EP
NSC (Byhamre et al., 2017; Gustafsson et al., 2011)
Page 3 of 7
Sm to
ACCEPTED MANUSCRIPT Table 3:
Risk of type 2 diabetes and snus use in never smokers (adjusted for
variables given in Table 2 except smoking) Current use n RR (95% CI)
MONICA I (Eliasson et al., 2004) MONICA II (Eliasson et al., 2004)
a
Stockholm County (Wändell et al., 2008)
b
SDPP II (Östenson et al., 2012) ESTRID (Rasouli et al., 2016)
4
3.90 (1.10-14.30)
6
1.06 (0.43-2.64)
5
1.45 (0.54-3.87)
0
-
1
1.72 (0.20-14.8)
1
2.12 (0.25-17.71)
1
3.10 (0.36-26.84)
-
2.30 (0.50-9.80)
27
1.17 (0.58-2.37)
HUNT (Rasouli et al., 2016)
-
VIP (Carlsson et al., 2017)
11
SALT (Carlsson et al., 2017)
0.53 (0.20-1.39) -
165
1.28 (1.08-1.52)
74
0.84 (0.66-1.07)
25
1.09 (0.70-1.69)
19
1.10 (0.67-1.08)
13
0.66 (0.38-1.16)
M AN U
SPHC (Carlsson et al., 2017)
-
RI PT
SDPP I (Persson et al., 2000)
Former use n RR (95% CI
SC
Study name (ref)
43
0.95 (0.69-1.33)
8
0.60 (0.30-1.22)
7
0.93 (0.43-2.03)
-
0.38 (0.12-1.16)
-
Overall random-effects estimate (95% CI)
1.08 (0.86-1.34)
0.93 (0.79-1.11)
Heterogeneity chi-squared (degrees of freedom, p)
14.79 (10 df, p=0.14)
7.61 (7 df, p=0.37
MDCS (Carlsson et al., 2017) NMC (Carlsson et al., 2017)
a
b
Also available are results for diabetes diagnosed at oral glucose tolerance test: current use n = 1 RR = 0.91(0.10-8.01); former use n = 3 RR = 3.97 (0.86-18.33); ever use n = 4 RR = 2.03 (0.55-7.55)c Type of diabetes not specified Estimated from data given
EP
c
12
TE D
NSC (Byhamre et al., 2017)
-
Risk of type 2 diabetes and snus use in the whole population
AC C
Table 4:
(adjusted for variables given in Table 2)
Study name (ref)
Current use n RR (95% CI)
Former use n RR (95% CI
Ever use n RR (95% CI)
SDPP I (Persson et al., 2000)
13
1.50 (0.803.00)
5
0.80 (0.302.00)
18
1.20 (0.672.16)a
SHEP/VHEP (Hergens et al., 2005)ab
11
1.50 (0.762.90)
4
1.10 (0.403.30)
15
1.37 (0.762.46)a
2
1.35 (0.68-
SALLS (Johansson et
-
-
Page 4 of 7
0.94 (0.51-1.71)
ACCEPTED MANUSCRIPT al., 2005)b
2.67)
Stockholm County (Wändell et al., 2008)b
11
1.48 (0.72 3.03)
1
2.46 (0.3020.43)
12
1.54 (0.783.06)a
SDPP II (Östenson et al., 2012)
16
1.10 (0.602.00)
6
0.50 (0.201.20)
22
0.82 (0.491.39)a
ESTRID (Rasouli et al., 2016)
129
0.96 (0.671.37)
80
0.63 (0.410.95)
209
0.81 (0.601.08)a
157
0.91 (0.751.10)
-
-
Overall randomeffects estimate (95% CI)
1.18 (0.941.48)
0.69 (0.490.96)
0.95 (0.811.11)
Heterogeneity chisquared (degrees of freedom, p)
2.86 (5 df, p=0.72)
2.91 (4 df, p=0.57 )
5.61 (5 df, p=0.35)
SC M AN U
TE D
Estimated from data given Type of diabetes not specified
EP
b
AC C
a
RI PT
HUNT (Rasouli et al., 2016)
Page 5 of 7
ACCEPTED MANUSCRIPT Table 5: Dose-response results for the risk of type 2 diabetes and snus use among current users Population
Exposure category
Levels
SDPP I (Persson et al., 2000) Stockholm County (Wändell et al., 2008)b SDPP II (Östenson et al., 2012) ESTRID (Rasouli et al., 2016)
Never smokers All
Boxes/wk Cans/wk
0,1-2,3+ 0,<3,3+
VIP (Carlsson et al., 2017)
All All Never smokers All Never smokers All Never smokers Never smokers
Boxes/wk Boxes/wk Boxes/wk Box-years Box-years Boxes/wk Boxes/wk Boxes/wk
SPHC (Carlsson et al., 2017)
Never smokers
Duration (yr) Boxes/wk
SALT (Carlsson et al., 2017)
Never smokers
Duration (yr) Boxes/wk
MDCS (Carlsson et al., 2017)
Never smokers
Duration (yr) Boxes/wk
NMC (Carlsson et al., 2017) NSC (Byhamre et al., 2017)
Never smokers All
0,1-5,5+ 0,<5,5+ 0,<5,5+ 0,<10,10+ 0,<10, 10+ 0,<3,3+ 0,<3,3+ 0,1-2,3-4, 5-6,7+ 0,<30,30+ 0,1-2,3-4, 5-6,7+ 0,<30,30+ 0,1-2,3-4, 5-6,7+ 0,<30,30+ 0,1-2,3-4, 5-6,7+ 0,<30,30+ 0,1,2, 3,4
a
0.20 (0.00-2.00) 1.30 (0.49-3.40)
SC
M AN U Duration (yr) Follow-up periods
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
0.60 (0.20-1.40) 0.78 (0.56-1.09) 0.83 (0.41-1.71) 0.74 (0.52-1.06) 0.74 (0.31-1.77) 0.88 (0.72-1.08) 1.15 (0.72-1.82) 1.12 (0.81-1.54) 1.48 (1.09-2.02) 1.40 (1.06-1.84) No cases 2.62 (0.62-11.02) 1.96 (0.45-8.58) 1.17 (0.64-2.15) 1.02 (0.42-2.50) 0.82 (0.36-1.87) 1.19 (0.37-3.78) 0.71 (0.17-2.89) No cases 0.80 (0.39-1.64) 1.96 (0.44-2.12)
Estimated for never smokers only, based on a regression analysis in which the mean boxes/week used was taken as the midpoint of the range for closed dose intervals and as the geometric mean of the low value and 12 for open dose intervals. Using 20 rather than 12 had little effect on the overall estimate. Type of diabetes not stated.
AC C
EP
b
1.00 1.00
TE D
HUNT (Rasouli et al., 2016)
Adjusted RR (95% CI) by leve
RI PT
Study name (ref)
Page 6 of 7
2.70 (1.30-5 1.80 (0.67-4
3.30 (1.40-8 0.95 (0.57-1 1.01 (0.42-2 1.05 (0.67-1 1.00 (0.47-2 0.92 (0.46-1 0.89 (0.21-3 1.17 (0.89-1 1.79 (1.14-2 1.22 (1.00-1 1.50 (0.45-5 1.50 (0.20-1 1.17 (0.41-3 0.74 (0.45-1 1.51 (0.76-2 0.97 (0.68-1 0.32 (0.08-1 1.30 (0.18-9 1.50 (0.65-3 1.40 (0.67-2 0.57 (0.22-1
ACCEPTED MANUSCRIPT Table 6:
Risk of other endpoints and snus use in never smokers (adjusted
for variables given in Table 2 except smoking)
ESTRID (Rasouli et al., 2016)
LADA
13
0.98 (0.452.11)
SDPP I (Persson et al., 2000)
IGT
6
0.90 (0.402.10)
MONICA I (Eliasson et al., 2004)
IGT
5
0.78 (0.292.09)
MONICA II (Eliasson et al., 2004)
IGT
1
0.23 (0.031.80)
NSC (Byhamre et al., 2017)
MS
12
4
0.46 (0.151.43) -
1.15 (0.522.51)
Ever use n RR (95% CI) 17
0.75 (0.391.44)b -
6
1.48 (0.573.80)
11
1.05 (0.512.17)
2
0.75 (0.163.57)
3
0.42 (0.121.48)b
-
-
TE D
a
Former use n RR (95% CI
RI PT
Current use n RR (95% CI)
SC
Endpointa
M AN U
Study name (ref)
Abbreviations used: IGT = impaired glucose tolerance, LADA = latent autoimmune diabetes of adulthood, MS = metabolic syndrome Estimated from data given
AC C
EP
b
Page 7 of 7
ACCEPTED MANUSCRIPT Table 7:
Risk of other endpoints and snus use in the whole population
(adjusted for variables given in Table 2)
Endpointa
ESTRID (Rasouli et al., 2016)
LADA
VIP (Neumann et al., 2013
Former use n RR (95% CI
1.01 (0.671.54)
19
IFG
-
0.92 (0.821.03)
-
IFG/IGT
-
0.79 (0.591.05)
26
0.80 (0.501.30)
IRESIS
9
IRESP
8
-
Ever use n RR (95% CI) 64
0.83 (0.581.19)b -
-
-
19
0.70 (0.401.20)
45
0.75 (0.511.11)b
0.90 (0.402.00)
3
0.40 (0.101.30)
12
0.68 (0.331.40)b
1.20 (0.502.80)
12
2.20 (1.104.40)
20
1.70 (0.933.11)b
M AN U
IGT
0.60 (0.341.04)
SC
45
TE D
SDPP I (Persson et al., 2000)
Current use n RR (95% CI)
RI PT
Study name (ref)
IRESP
-
3.30 (0.912.10)
-
-
NSC (Gustafsson et al., 2011)c
MS
-
0.79 (0.331.86)d
-
-
-
0.96 (0.581.56)d
-
-
EP
SDPP II (Östenson et al., 2012)
a
b c d
e
MS
AC C
NSC (Gustafsson et al., 2011)e
Abbreviations used: IFG = impaired fasting glucose, IGT = impaired glucose tolerance, IRESIS = insulin resistance, IRESP = insulin response, LADA = latent autoimmune diabetes of adulthood, MS = metabolic syndrome Estimated from data given Males Adjusted for socioeconomic status at ages 16, 21 and 43, body mass index, systolic and diastolic blood pressure at ages 16 and 21, daily smoking, daily snuff use, alcohol consumption and physical inactivity at age 43 Females
Page 8 of 7
ACCEPTED MANUSCRIPT Table 8: Dose-response results for the risk of other endpoints and snus use among current users Level
SDPP I (Persson et al., 2000)
IGT (never smokers) IRESIS (never smokers) IRESP (never smokers) ESTRID (Rasouli et LADA (all) al., 2016) LADA (never smokers) LADA (all)
Boxes/wk
0,1-2,3+
Boxes/wk
0,1-2,3+
Boxes/wk
0,1-2,3+
Boxes/wk
0,<5,5+
Boxes/wk
0,<5,5+
Box-years
0,<10,10+
LADA (never smokers)
Box-years
0,<10,10+
1.00 0.70 (0.401.40) 1.00 0.50 (0.201.60) 1.00 2.10 (1.104.10) 1.00 0.71 (0.461.09) 1.00 0.75 (0.341.67) 1.00 0.56 (0.350.91) 1.00 0.46 (0.161.31)
0.80 (0.401.40) 0.70 (0.301.70) 1.20 (0.502.90) 1.18 (0.672.10) 0.67 (0.241.86) 1.45 (0.892.37) 1.01 (0.452.29)
EP
TE D
M AN U
Abbreviations used: IGT = impaired glucose tolerance, IRESIS = insulin resistance, IRESP = insulin response, LADA = latent autoimmune diabetes of adulthood
AC C
a
Adjusted RR (95% CI) by level
RI PT
Exposure category
SC
Study name (ref) Endpointa (population)
Page 9 of 7
ACCEPTED MANUSCRIPT Highlights
RI PT
We reviewed 18 studies that examined snus use and diabetes or related endpoints
Overall, current, ex or ever snus use showed no significant association with diabetes
SC
This was true for never smokers and the whole population
M AN U
Only heavy snus users who were never smokers had an increased risk of diabetes
AC C
EP
TE D
Other diabetes-related endpoints also failed to show any association with snus use
C:\pdfconversion\WORK\authordoc\Highlights.docx
S0273-2300(17)30332-X
DOI:
10.1016/j.yrtph.2017.10.017
Reference:
YRTPH 3964
To appear in:
Regulatory Toxicology and Pharmacology
Received Date: 21 June 2017 Revised Date:
20 September 2017
Accepted Date: 19 October 2017
Please cite this article as: Lee, P.N., Thornton, A.J., The relationship of snus use to diabetes and allied conditions, Regulatory Toxicology and Pharmacology (2017), doi: 10.1016/j.yrtph.2017.10.017. 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
TITLE
The relationship of snus use to diabetes and allied
2
conditions
3 Peter N Lee1 and Alison J Thornton2
5 6 7 8 9 10
1
P.N. Lee Statistics and Computing Ltd., 17 Cedar Road, Sutton, Surrey SM2 5DA
2
Independent Consultant, Oak Cottage, Beer Farm, Okehampton, Devon, EX20 1SG
RI PT
Authors
SC
4
Corresponding author
Peter N Lee P.N. Lee Statistics and Computing Ltd., 17 Cedar Road, Sutton, Surrey SM2 5DA Tel: +44 (0) 20 8642 8265 Fax: +44 (0) 20 8642 2135 Email: Peter [email protected]
18 19 20 21
Word Counts
Abstract Text References
M AN U
11 12 13 14 15 16 17
AC C
EP
TE D
193 3754 729
(C:\pdfconversion\WORK\authordoc\Snus and diabetes_red
lined changes to text.docx)
ACCEPTED MANUSCRIPT Abstract A recent meta-analysis reported smoking to be associated with a 37% higher risk of type 2 diabetes in current smokers, rising to a 57% increase in heavy smokers, which declines on quitting. If the increase results from nicotine exposure, it is possible that
RI PT
using Swedish moist snuff (“snus”), which provides at least equivalent nicotine doses, might also increase diabetes risk. Following a recent publication reporting pooled results from five cohorts, we present a detailed meta-analysis of data from 18 studies.
SC
Based on covariate-adjusted estimates, no significant increased risk was seen in never
M AN U
smokers with RRs (95% CIs) of 1.08 (0.86-1.34), 0.93 (0.79-1.11) and 1.05 (0.941.18) for current, former and ever snus users. Significant increases were also not seen in the whole population, the corresponding RR estimates being 1.18 (0.94-1.48), 0.69 (0.49-0.96) and 0.95 (0.81-1.11). Nor was there an association of snus use with related endpoints, such as impaired glucose tolerance.
However, dose-response
TE D
analyses showed a relationship, with the highest levels of snus exposure associated with a diabetes RR of 1.65 (1.25-2.18) in never smokers. The evidence relating snus to type 2 diabetes is somewhat limited, requiring further studies to confirm any
AC C
EP
possible relationship.
Keywords Diabetes Tobacco Smokeless
ACCEPTED MANUSCRIPT
Introduction1
RI PT
1.
In 2011, in this journal, one of us (PNL) reviewed the epidemiological evidence relating snus to health (Lee, 2011), noting that there was “scant support for any major adverse health effect of snus”. Although it was noted that “snus users have increased weight” the review
SC
concluded that diabetes seems to be “unaffected”, and a lack of relationship with heart disease risk was also demonstrated. The section
M AN U
on diabetes cited 10 relevant publications (Attvall et al., 1993; Eliasson et al., 1995; Eliasson et al., 2004; Eliasson et al., 1991; Hergens et al., 2005; Johansson et al., 2005; Norberg et al., 2006; Persson et al., 2000; Wallenfeldt et al., 2001; Wändell et al., 2008) which noted that, despite occasional reports of significant associations, there was generally no clear or consistently observed relationship of snus use
TE D
with diabetes or a range of associated endpoints, such as glucose intolerance, fasting glucose levels, fasting insulin levels, or metabolic syndrome. However, meta-analyses were not attempted.
EP
Elsewhere, it has also been reported that the fasting blood glucose levels of snus users are not significantly different from those for never
1
AC C
users (Foulds et al., 2003). However, our conclusion that an association of snus with diabetes was “not clearly established” contrasts with BMI, body mass index; CI, confidence interval; ETS, environmental tobacco smoke; HR, hazard ratio; LADA, latent autoimmune diabetes of adulthood; OR, odds ratio; RR, relative risk; SES, socioeconomic status; T2D, type 2 diabetes; VIP, Våsterbotten Intervention Programme
Page 3 of 21
ACCEPTED MANUSCRIPT
that from a pooled analysis of results from five cohorts (Carlsson et al., 2017) which concluded that “high consumption of snus is a risk
RI PT
factor for type 2 diabetes”. This analysis estimated that, compared to never users the confounder-adjusted hazard ratio (HR) was 1.15
SC
(95% CI 1.00-1.32) in current users of snus, rising to 1.68 (1.17-2.41) in users of seven or more boxes per week.
M AN U
Preliminary inspection of the results of the analysis suggested that the significantly increased risk in current users derived principally from the results in one of the five cohorts considered, the Våsterbotten Intervention Programme (VIP), where the HR was 1.28 (95% confidence interval [CI] 1.08-1.52), no significant increase being seen in one of the other cohorts, and no increase at all in the other three, though two of these estimates were based on less than 10 diabetes cases in current users. However, the results for heavy consumers were
TE D
more consistent, with an elevated risk seen in all four cohorts with dose-response data, though statistically significantly only for the VIP.
EP
While Carlsson et al. (2017) refer to some of the publications we cited (Eliasson et al., 1995; Eliasson et al., 2004; Persson et al., 2000)
AC C
and to results of some more recent studies on diabetes (Östenson et al., 2012; Rasouli et al., 2016), they did not attempt a meta-analysis of the overall literature. The objective of this publication is to present such a meta-analysis.
Page 4 of 21
2.
Materials and methods
2.1
Literature searching
RI PT
ACCEPTED MANUSCRIPT
Relevant papers were derived from PubMed searches of (snuff OR snus OR smokeless tobacco) AND (diabetes OR glucose intolerance
SC
OR insulin) initially conducted on 14th February 2017, and updated on 5th September, 2017. Abstracts were studied initially, with papers
M AN U
of potential interest then looked at. Relevant reviews were also studied for relevant references as were reference lists in papers identified. The objective was to identify papers that presented evidence on the relationship of snus use, in never smokers and/or in the whole population, to diabetes or to associated endpoints, such as impaired glucose tolerance or fasting glucose or insulin resistance or response.
EP
Data extraction
For each study identified, information was extracted on the study name and location, relevant references, study design (prospective, casecontrol or cross-sectional), subjects included, sexes studied, endpoints studied and confounding variables adjusted for. For each endpoint
AC C
2.2
TE D
A similar Embase search was conducted on 8th September, looking for additional relevant papers.
for which results were available, relative risks (RRs), odds ratios (ORs) or HRs were extracted, or derived from data presented using standard methods (Gardner and Altman, 1989; Hamling et al., 2008; Lee et al., 2012). Estimates were extracted for current, former and ever snus users relative to never snus users, separately for never smokers and for the whole population. Where available, estimates were
Page 5 of 21
ACCEPTED MANUSCRIPT
extracted (or derived) adjusted for the maximum number of potential confounding variables for which results were available, and also
RI PT
unadjusted for any variable (for case-control and cross-sectional studies) or for age only (for prospective studies). Estimates were also extracted (or derived) by extent of exposure to snus. For diabetes in never smokers, estimates were also made of the increase in RR per
M AN U
SC
box used per week.
Exceptionally, some studies only reported results for continuous endpoints, such as blood glucose or insulin. Here the findings were
Meta-analyses
Meta-analyses were only carried out for diabetes, separately for current, former and ever snus use, based on results either for never
EP
smokers or the whole population. Meta-analysis was mainly restricted to effect estimates which were adjusted for confounding variables. Standard methods were used to carry out the meta-analyses (Fleiss and Gross, 1991) with fixed-effect and random-effects estimates
AC C
2.3
TE D
summarized from the relevant publications.
computed, together with a test for heterogeneity between individual estimates. Estimates were derived using all the available data for a particular analysis, and also for subsets of the data by time of publication (to 2010, 2011 onwards), study design (case-control or crosssectional, prospective), type of diabetes (type 2, unspecified), number of potential confounding variables adjusted for (<4, 5+), and
Page 6 of 21
ACCEPTED MANUSCRIPT
number of cases of diabetes in exposed subjects (<10, 11+ for analyses of current and former snus use, and <20, 21+ for analyses of ever
AC C
EP
TE D
M AN U
SC
RI PT
snus use). Publication bias was also investigated using Egger’s test (Egger et al., 1997).
Page 7 of 21
3.
Results
3.1.
Literature searches
RI PT
ACCEPTED MANUSCRIPT
SC
The first PubMed search produced 100 hits. Based on the abstracts, 83 were rejected as irrelevant. The remaining 17 papers were obtained for scrutiny, with four rejected as providing no additional primary data and one being a review. The remaining 12 papers were
M AN U
supplemented by an additional three publications identified in reviews or in reference lists of papers identified as relevant. The second PubMed search produced an additional five hits, four papers being rejected based on abstracts, with one recent relevant paper added to the 16 previously identified. The Embase search produced 179 hits, with only two additional papers obtained, both rejected as providing
EP
The studies
The 16 references described eighteen studies which provided evidence on the association between snus use and diabetes and related
AC C
3.2.
TE D
no additional primary data.
conditions. Table 1 gives details of the main characteristics of the studies, a more detailed description of each study being given in Appendix A. Two studies were published in the 1990s, seven between 2000 and 2010, and nine after 2011. Except for one study in Norway, all the studies were in Sweden. Seven studies were cross-sectional, while two were of a case-control design, both drawing
Page 8 of 21
ACCEPTED MANUSCRIPT
controls from the general population. The remaining nine studies were prospective, with follow-up ranging from 10 to 27 years. Fifteen
RI PT
studies were in males, while two involved both sexes. One study only included women for certain endpoints.
SC
Various endpoints were considered. Type 2 diabetes (T2D) was investigated in 12 studies, although one of these included impaired
M AN U
fasting glucose in their definition. A further three studies provided information on diabetes, but did not specify the type. Four studies used an endpoint of impaired glucose tolerance and/or pathological glucose tolerance, the latter being a combination of diabetes or impaired glucose tolerance. One study considered latent autoimmune diabetes of adulthood (LADA), and another study the metabolic
resistance and response levels.
TE D
syndrome. Blood glucose level was examined by four studies, with insulin level considered by three. One study also examined insulin
EP
Three studies (Attvall et al., 1993; Eliasson et al., 1991; Wallenfeldt et al., 2001), none of which provided information on diabetes, did
AC C
not adjust their results for any potential confounding variables. In the remaining 15 studies (see Table 2), all but three considered age, although one of these studies did carry out a series of analyses in which they evaluated various endpoints at different ages. Various other confounders were considered, the most common being waist circumference, body mass index, alcohol consumption, smoking status, physical activity, education and family history of diabetes.
Page 9 of 21
ACCEPTED MANUSCRIPT
RI PT
Diabetes
Table 3 summarizes the evidence relating snus use in never smokers to covariate-adjusted risk of type 2 diabetes. Corresponding
SC
unadjusted (or for prospective studies age-adjusted results) are given in Appendix B, Table B1. These tables also include overall meta-
M AN U
analysis results, with adjusted results separated by time of publication, study design, type of diabetes, number of confounding variables, and number of exposed cases shown in Appendix C, Table C1. In these tables, the effects estimates are referred to as RRs, though they may have been estimated as ORs or HRs.
TE D
It can be seen from Table 3 that a number of the individual study effect estimates are based on quite small numbers of snus-exposed cases, with, for example, five of the 11 estimates for current use based on less than 10 cases. Also, the number of snus-exposed cases in
EP
the VIP study form over half the total number in all the studies combined. None of the random-effect estimates, whether for current, former or ever snus use show significant (p<0.05) evidence of an effect, with the combined estimates being, respectively, 1.08 (95% CI
AC C
3.3.
0.86-1.34), 0.93 (0.79-1.11) and 1.05 (0.94-1.18). Nor was there any significant evidence of heterogeneity between estimates, despite two of the estimates for current use (from the SDPP1 and VIP studies) being statistically significantly increased, or of publication bias (data not shown). There was an indication that case-control or cross sectional studies produced somewhat higher results for current snus
Page 10 of 21
ACCEPTED MANUSCRIPT
use, with the overall estimate for these four studies being 1.40 (0.86-2.31), compared to 1.14 (0.99-1.30) for the seven prospective studies
RI PT
(see Table C1). However, this difference was not statistically significant (p = 0.42), and was scarcely evident for former or ever snus use. There was no evidence that estimates varied significantly by year of publication, type of diabetes, number of confounding variables
SC
considered, or number of exposed cases. The overall unadjusted estimates (see Table B1) tended to be somewhat lower, at 0.93 (0.82-
M AN U
1.05), 0.65 (0.39-1.09) and 0.69 (0.52-0.94) for, respectively, current, former and ever snus use. The low estimates for former and ever snus use, and the substantial heterogeneity for these estimates, are largely due to particularly low unadjusted estimates from the ESTRID study.
TE D
Tables 4, B2 and C2 similarly summarize the evidence relating snus use in the whole population to risk of type 2 diabetes. The numbers of snus-exposed cases shown in Table 4 remain quite small, except in the ESTRID and HUNT studies. Again, none of the overall
EP
random-effect estimates, 1.18 (95% CI 0.94-1.48) for current use, 0.69 (0.49-0.96) for former use and 0.95 (0.81-1.11) for ever use show
AC C
any significant evidence of an increased risk with snus use, the estimate for former use indicating a significantly decreased risk, though relying primarily on the results from the ESTRID study. There was no significant evidence of heterogeneity for current, former or ever use. The tests for publication bias were not significant except for current use, where five estimates above 1.0 from relatively small studies were countered by a slightly reduced estimate from the ESTRID study. Results based on case-control or cross-sectional studies
Page 11 of 21
ACCEPTED MANUSCRIPT
did not vary significantly from those based on prospective studies, for current, ex- or ever snus use (see Table C2). There was also little
RI PT
indication that the effect estimates varied by level of the other factors considered, though for ever use there was marginally significant (0.01
SC
studies, though these comparisons were not independent of each other. The overall unadjusted estimates (see Table B2) were here quite
M AN U
similar to the adjusted ones, being 1.16 (0.90-1.51), 0.82 (0.63-1.09) and 0.96 (0.71-1.30) for, respectively, current, former and ever use. The significant heterogeneity for ever use derives from the low effect estimate for the HUNT study.
Table 5 summarizes available dose-response results from 11 studies, eight presenting results for never smokers and five for the whole
TE D
population. The results presented are for various dose metameters – boxes/wk, cans/wk, box-years, duration (in years) and periods of follow-up – and for various dose levels. Significant increases, compared to never users, at the highest level of exposure are seen for
EP
never smokers in two studies (SDPP1 for 3+ boxes/wk and VIP for 7+ boxes/wk) and for the whole population in one study (SDDP2 for
AC C
5+ boxes/wk). Meta-analyses of the effect at the highest level were carried out both for never smokers and for the whole population, preferring estimates for boxes/wk, where a single study gave results for multiple dose metameters. For never smokers, the randomeffects analysis showed clear evidence of an association, with a combined estimate of 1.65 (95% CI 1.25-2.18) based on eight independent estimates and no significant heterogeneity. The largest contributors to the overall estimate were the estimates of 2.70 (1.30-
Page 12 of 21
ACCEPTED MANUSCRIPT
5.50) from the SDPP1 study and of 1.79 (1.14-2.80) from the VIP study, the combined estimate becoming non-significant, at 1.31 (0.87-
RI PT
1.98), based on the other six estimates. For the whole population, the combined evidence from the highest level of exposure was a nonsignificant 1.20 (0.71-2.04), based on four somewhat variable estimates (heterogeneity p = 0.0528). For diabetes in never smokers,
SC
where RRs are available by the number of boxes of snus used per week, Table 5 shows the estimated increase in RR per box per week.
M AN U
Based on these seven estimates, a combined fixed-effect estimate of 1.05 (95% CI 1.02-1.08) was derived. Although there was no significant heterogeneity between the individual increaes, the overall estimate was again heavily dependent on the increases seen in the
Other endpoints
The evidence relating snus use to other endpoints in never smokers which is summarized in Table 6 is sparse, but is not at all indicative
EP
of a positive relationship with snus use, with none of the ten estimates statistically significant, and only two of them above 1. Although one of the 17 estimates for other endpoints in the whole population shown in Table 7 shows a significant increase (RR 2.2, 95% CI 1.1-
AC C
3.4
TE D
SDPP1 and VIP studies.
4.4) this, for insulin response, is for former snus users, with no significant increase seen in this study for current users. The overall evidence shown in the Table is not suggestive of an effect of snus, with 12 of the 17 estimates below 1.
Page 13 of 21
ACCEPTED MANUSCRIPT
Not included in Tables 6 and 7 are results from four studies which reported results relating snus use to blood glucose and insulin levels.
RI PT
In the first study (Eliasson et al., 1991), snus use had no effect on blood glucose levels, but serum insulin was higher in snus users (5.5 mU/L) than in non-tobacco users (3.6 mU/L) and this difference was statistically significant (p<0.01). The second study (Attvall et al.,
SC
1993) also reported no difference in blood glucose levels or in fasting insulin levels before or after normoglycaemic clamps. The third
M AN U
study (Wallenfeldt et al., 2001) reported no significant difference in fasting blood glucose or plasma insulin between current, ex- or never snus users. The MONICA I study (Eliasson et al., 1995) found no differences between non-tobacco users and snus users in fasting glucose levels, fasting insulin or post-load insulin.
EP AC C
endpoints studied.
TE D
The limited dose-response evidence for other endpoints, shown in Table 8, provides no real indication of an effect of snus on any of the
Page 14 of 21
ACCEPTED MANUSCRIPT
RI PT
Discussion A recent meta-analysis of prospective studies relating smoking to diabetes risk (Pan et al., 2015) reported combined results from 88 prospective studies involving almost 300,000 incident cases of type 2 diabetes. Significant combined estimates were noted for current
SC
smokers (RR 1.37, 95% CI 1.33-1.42), former smokers (1.14, 1.10-1.18) and never smokers exposed to environmental tobacco smoke
M AN U
(1.22, 1.10-1.35). A dose-response relationship was also noted for current smoking, with an estimate of 1.57 (1.47-1.66) for heavy smokers. Risk estimates declined with time quit. Given that snus users have levels of nicotine exposure that are similar to, and perhaps somewhat greater than, that from smoking (Lee, 2011), the possibility arises that snus use might, if nicotine is relevant to risk of diabetes,
TE D
also increase risk of diabetes.
The evidence that snus use increases the risk of type 2 diabetes is clearly much weaker for snus than for smoking. Thus, there is no
EP
significant evidence of an increased risk, compared to never snus users, in current, former or ever snus users, whether restricting attention to never smokers or considering the whole population. Nor is there any clear evidence of an association of snus use with endpoints
AC C
4.
related to diabetes, such as glucose or insulin levels. The only real evidence for a possible effect of snus on diabetes arises from the dose-response data, where in never smokers (though not in the whole population) there is an increased RR of 1.65 (95% CI 1.25-2.18) for
Page 15 of 21
ACCEPTED MANUSCRIPT
the highest level of current snus use, and a RR of 1.05 (1.02-1.08) can be derived for the increase in RR per extra box of snus used per
RI PT
week.
SC
In considering the evidence on snus, various issues should be considered. Firstly, only 15 studies provide data on snus and diabetes, far
M AN U
less than the 88 studies considered in the recent review of smoking and diabetes (Pan et al., 2015). Furthermore, many of the estimates are based on rather small numbers of snus-exposed cases, with the VIP study on its own providing more than half the exposed cases for never smokers (see Table 3) and the ESTRID and HUNT studies contributing largely to the results for the whole population (see Table 4). With the single exception of the HUNT study in Norway, the studies are all conducted in Sweden, so that variation between countries
TE D
in the product used is unlikely to be an issue. Many, but not all of the studies, adjusted for a range of potential confounding variables, though little evidence was provided on which specific variables materially affected risk estimates. While 12 of the 15 studies provided
EP
results specifically for type 2 diabetes, three studies (SHEP/VHEP, SALLS and Stockholm County) only referred to diabetes. However,
AC C
there was no evidence that results varied significantly by definition of diabetes. Nor was there evidence that results varied by other factors considered, including time of publication, study design, number of potential confounding variables adjusted for, or number of cases in snus-exposed subjects. This suggests that it is reasonable to take the overall estimates shown in Tables 3 and 4 as indicative of the relationship between snus and diabetes. The extent to which these estimates might be affected by various biases is difficult to assess.
Page 16 of 21
ACCEPTED MANUSCRIPT
We found no evidence of publication bias using the Egger test, although this test has limited power when few studies are available. We
SC
though it seems doubtful whether these would have any major effect.
RI PT
were unable to formally take into account possible biases arising from misclassification of disease or of exposure (to snus or smoking),
M AN U
As all but one of the studies was conducted in Sweden, some overlap between studies is possible. However, we consider it unlikely to be is a major issue. Studies sharing a location tended to differ either in the time period during which they were conducted or the endpoint considered, or the population was more narrowly defined in some studies, meaning that only a small number of subjects were likely to have participated in multiple studies. Only for the SDPP and SHEP/VHEP studies does it appear possible that more significant overlap
TE D
may be present, as the studies included a similar age range and took place at the same time. SDPP I and II were limited to four municipalities in Stockholm county whereas SHEP/VHEP appears to include the whole county but there is no indication from the authors
EP
that the SDPP studies are a subset of SHEP/VHEP. There are enough differences between them in age range, study period and
AC C
geographical location to consider them as separate studies, any overlap probably being quite small.
The main evidence of a relationship of snus to diabetes comes from the dose-response analyses. Though, for the whole population, the combined estimate for the highest level of exposure was not significant (RR 1.39, 95% CI 0.79-2.46) this was based on four rather
Page 17 of 21
ACCEPTED MANUSCRIPT
variable estimates. Evidence of a dose-response was clearer for never smokers, with the combined RR estimate being 1.65 (95% CI 1.25-
RI PT
2.18), based on eight estimates, with no significant heterogeneity. As different studies reported dose-response results in different ways, four of the estimates were for 7+ boxes/wk, one for 5+ boxes/wk, two for 3+ boxes/wk, and one for 30+ years duration. Of these eight
M AN U
SC
estimates, seven were above 1.00, with two (from the SDPP1 and VIP studies) statistically significant.
Based on pooled results from the VIP, SPHC, SALT and MDCS studies, a recent review (Carlsson et al., 2017) reported a significant increased risk of type 2 diabetes for 7+ boxes/wk (RR 1.68, 95% CI 1.17-2.41) and 5-6 boxes/wk (1.42, 1.07-1.87), and concluded that “high consumption of snus is a risk factor for type 2 diabetes”, with a risk “similar to that in smokers” and supporting “the notion that
TE D
nicotine increases the risk of type 2 diabetes.” Although the findings of a dose-response in never smokers is indicative of a possible relationship of snus to diabetes risk, the general lack of association seen in the other analyses reported in this review suggests to us that
EP
more evidence is required before any firm conclusion can be reached. We note that the increased risk for 7+ boxes/wk reported in the
estimate of 1.49 (0.80-2.75).
AC C
review is heavily dependent on results from the VIP study – the combined results from the other three studies giving a non-significant
Page 18 of 21
ACCEPTED MANUSCRIPT
RI PT
Conclusions We found no evidence of an overall increased risk of type 2 diabetes in current, former or ever snus users, either in never smokers or the whole population. Nor is snus use associated with endpoints related to diabetes. There is some evidence of a dose-response relationship
EP
TE D
M AN U
SC
with diabetes risk, in never smokers, though not the whole population, but this requires confirmation from additional studies.
AC C
5.
Page 19 of 21
ACCEPTED MANUSCRIPT
RI PT
Conflict of interest P N Lee is a long-term consultant to the tobacco industry and A J Thornton is a consultant to P N Lee’s company, P N Lee Statistics and
SC
Computing Ltd. However, this is an independent scientific assessment, the views expressed being those of the author alone.
M AN U
Acknowledgements
We thank Mrs. Y. Cooper and Mrs. D. Morris for their assistance in typing the various drafts of this report, and their help in assembling the
AC C
EP
TE D
relevant literature. We also thank colleagues for providing helpful comments, and Swedish Match for financial support.
Page 20 of 21
ACCEPTED MANUSCRIPT
RI PT
References
Attvall, S., Fowelin, J., Lager, I., Von Schenck, H., Smith, U., 1993. Smoking induces insulin resistance - a potential link with the insulin resistance syndrome. J. Intern. Med. 233, 4, 327-332.
M AN U
SC
Byhamre, M.L., Gustafsson, P.E., Jansson, J.H., Wennberg, M., Hammarstrom, A., Wennberg, P., 2017. Snus use during the life-course and risk of the metabolic syndrome and its components. Scand. J. Public Health. Epub ahead of print Jul 1, 1403494817706631. DOI:10.1177/1403494817706631. Carlsson, S., et al., 2017. Smokeless tobacco (snus) is associated with an increased risk of type 2 diabetes: results from five pooled cohorts. J. Intern. Med. 281, 4, 398-406. DOI:10.111/joim.12592.
TE D
Egger, M., Davey Smith, G., Schneider, M., Minder, C., 1997. Bias in meta-analysis detected by a simple, graphical test. BMJ. 315, 629-634. DOI:10.1136/bmj.315.7109.629. Eliasson, M., Asplund, K., Evrin, P.-E., Lundblad, D., 1995. Relationship of cigarette smoking and snuff dipping to plasma fibrinogen, fibrinolytic variables and serum insulin. The Northern Sweden MONICA study. Atherosclerosis. 113, 41-53.
EP
Eliasson, M., Asplund, K., Nasic, S., Rodu, B., 2004. Influence of smoking and snus on the prevalence and incidence of type 2 diabetes amongst men: the northern Sweden MONICA study. J. Intern. Med. 256, 101-110.
AC C
Eliasson, M., Lundblad, D., Hägg, E., 1991. Cardiovascular risk factors in young snuff-users and cigarette smokers. J. Intern. Med. 230, 1, 1722. Fleiss, J.L., Gross, A.J., 1991. Meta-analysis in epidemiology, with special reference to studies of the association between exposure to environmental tobacco smoke and lung cancer: a critique. J. Clin. Epidemiol. 44, 2, 127-139. DOI:10.1016/0895-4356(91)90261-7.
Page 21 of 21
ACCEPTED MANUSCRIPT
RI PT
Foulds, J., Ramstrom, L., Burke, M., Fagerström, K., 2003. Effect of smokeless tobacco (snus) on smoking and public health in Sweden. Tob. Control. 12, 349-359. Gardner, M.J., Altman, D.G. (Eds.), 1989. Statistics with confidence. Confidence intervals and statistical guidelines. British Medical Journal, London.
M AN U
SC
Gustafsson, P.E., Persson, M., Hammarström, A., 2011. Life course origins of the metabolic syndrome in middle-aged women and men: the role of socioeconomic status and metabolic risk factors in adolescence and early adulthood. Ann. Epidemiol. 21, 2, 103-10. DOI:10.1016/j.annepidem.2010.08.012. Hamling, J., Lee, P., Weitkunat, R., Ambühl, M., 2008. Facilitating meta-analyses by deriving relative effect and precision estimates for alternative comparisons from a set of estimates presented by exposure level or disease category. Stat. Med. 27, 954-970. DOI:10.1002/sim.3013.
TE D
Hergens, M.-P., Ahlbom, A., Andersson, T., Pershagen, G., 2005. Swedish moist snuff and myocardial infarction among men. Epidemiology. 16, 1, 12-16. Johansson, S.-E., Sundquist, K., Qvist, J., Sundquist, J., 2005. Smokeless tobacco and coronary heart disease a 12-year follow-up study. Eur. J. Cardiovasc. Prev. Rehabil. 12, 4, 387-392.
EP
Lee, P.N., 2011. Summary of the epidemiological evidence relating snus to health. Regul. Toxicol. Pharmacol. 59, 2, 197-214. DOI:10.1016/j.yrtph.2010.12.002.
AC C
Lee, P.N., Forey, B.A., Coombs, K.J., 2012. Systematic review with meta-analysis of the epidemiological evidence in the 1900s relating smoking to lung cancer. BMC Cancer. 12, 385. DOI:10.1186/1471-2407-12-385. Neumann, A., Norberg, M., Schoffer, O., Norstrom, F., Johansson, I., Klug, S.J., Lindholm, L., 2013. Risk equations for the development of worsened glucose status and type 2 diabetes mellitus in a Swedish intervention program. BMC Public Health. 13, 1014. DOI:10.1186/1471-2458-13-1014.
Page 22 of 21
ACCEPTED MANUSCRIPT
RI PT
Norberg, M., Stenlund, H., Lindahl, B., Boman, K., Weinehall, L., 2006. Contribution of Swedish moist snuff to the metabolic syndrome: a wolf in sheep's clothing? Scand. J. Public Health. 34, 6, 576-583. Östenson, C.G., Hilding, A., Grill, V., Efendic, S., 2012. High consumption of smokeless tobacco ("snus") predicts increased risk of type 2 diabetes in a 10-year prospective study of middle-aged Swedish men. Scand. J. Public Health. 40, 8, 730-737.
SC
Pan, A., Wang, Y., Talaei, M., Hu, F.B., Wu, T., 2015. Relation of active, passive, and quitting smoking with incident type 2 diabetes: a metaanalysis and systematic review. Lancet Diabetes Endocrinol. 3, 12, 958-67. DOI:10.1016/s2213-8587(15)00316-2.
M AN U
Persson, P.G., Carlsson, S., Svanström, L., Östenson, C.G., Efendic, S., Grill, V., 2000. Cigarette smoking, oral moist snuff use and glucose intolerance. J. Intern. Med. 248, 2, 103-110. Rasouli, B., et al., 2016. Use of Swedish smokeless tobacco (snus) and the risk of Type 2 diabetes and latent autoimmune diabetes of adulthood (LADA). Diabet. Med. 34, 4, 514-521. DOI:10.1111/dme.13179.
TE D
Wallenfeldt, K., Hulthe, J., Bokemark, L., Wikstrand, J., Fagerberg, B., 2001. Carotid and femoral atherosclerosis, cardiovascular risk factors and C-reactive protein in relation to smokeless tobacco use or smoking in 58-year-old men. J. Intern. Med. 250, 492-501.
AC C
EP
Wändell, P.E., Bolinder, G., de Faire, U., Hellénius, M.-L., 2008. Association between metabolic effects and tobacco use in 60-year-old Swedish men. Eur. J. Epidemiol. 23, 6, 431-434.
Page 23 of 21
ACCEPTED MANUSCRIPT
Table 1:
Details of studies examining the risk of diabetes and snus use Yeara
Location
Study typeb
Subjects
Sexc
No. of casesd
Umeå (Eliasson et al., 1991)
1991
Sweden, Umeå
CS
University students and teachers/ respondents to advertisements
M
-
Göteborg (Attvall et al., 1993)
1993
Sweden, Göteborg
CS
Healthy current smokers
M+F
-
SDPP I (Persson et al., 2000)
2000
Sweden, Stockholm
CS
General population
M
52
AIR (Wallenfeldt et al., 2001)
2001
Sweden, Göteborg
CS
General population
M
-
MONICA I (Eliasson et al., 1995; Eliasson et al., 2004)
2004
Sweden, Norrbotten and Våsterbotten
CS
General population
M
127
MONICA II (Eliasson et al., 2004)
2004
SHEP/VHEP (Hergens et al., 2005)
2005
SALLS (Johansson et al., 2005)
M AN U
SC
RI PT
Study name (ref)
P (13)
General population
M
38
Sweden, Stockholm and Våsternorrland
CC
General population
M
92
2005
Sweden
P (12)
General population
M
53
Stockholm County (Wändell et al., 2008)
2008
Sweden, Stockholm county
CS
60 year olds
M
78
SDPP II (Östenson et al., 2012)
2012
Sweden, Stockholm County
P (12)
General population
M
99
ESTRID (Rasouli et al., 2016)
2016
Sweden, Scania and Uppsala
CC
General population
M
724
HUNT (Rasouli et al., 2016)
2016
Norway, NordTrøndelag county
CS
General population
M
829
VIP (Carlsson et al., 2017; Neumann et al., 2013; Norberg et al., 2006)
2017
Sweden, Våsterbotten County
P (21)
General population
Mg
1033
SPHC (Carlsson et al., 2017)
2017
Sweden, Stockholm County
P (10)
General population
M
237
SALT (Carlsson et al., 2017)
2017
Sweden, nationwide
P (12)
Twins
M
405
MDCS (Carlsson et al., 2017)
2017
Sweden, Malmö
P (21)
General Population
M
461
AC C
EP
TE D
Sweden, Norrbotten and Våsterbotten
Page 1 of 7
ACCEPTED MANUSCRIPT Study name (ref)
Yeara
Location
NMC (Carlsson et al., 2017)
2017
Sweden, nationwide
NSC (Byhamre et al., 2017; Gustafsson et al., 2011)
2017
Sweden, Luleå
g h
M+F
21h
Year of most relevant publication CC = Case control, CS = Cross-sectional, P = Prospective (length of follow-up in years) F = Female, M = Male Total exposed and non-exposed cases for diabetes DTNS = diabetes type not specified, T2D = Type 2 diabetes BGL = blood glucose, IFG = impaired fasting glucose, IGT = impaired glucose tolerance, INP = plasma insulin, INS = serum insulin, IRESIS = insulin resistance, IRESP = insulin response, LADA = latent autoimmune diabetes of adulthood, MS = metabolic syndrome, PGT = pathological glucose tolerance (defined as diabetes or IGT at oral glucose tolerance test) Analyses for IFG and IGT are based on male and female participants Total number of cases for whole study, includes current and ex- cigarette smokers
RI PT
f
Adolescents in 9th grade
SC
e
M
M AN U
d
Participants in fundraising event
No. of casesd 305
TE D
c
Sexc
EP
b
P (27)
Subjects
AC C
a
Study typeb P (13)
Page 2 of 7
ACCEPTED MANUSCRIPT Table 2:
Confounding variables used in the studies Age
BMI
Physical activity
SDPP I (Persson et al., 2000)
Yes
Yes
Yes
MONICA I (Eliasson et al., 1995; Eliasson et al., 2004)
Yes
MONICA II (Eliasson et al., 2004)
Yes
SHEP/VHEP (Hergens et al., 2005)
Yes
Yes
ESTRID (Rasouli et al., 2016)
Yes
HUNT (Rasouli et al., 2016)
Yes
VIP (Carlsson et al., 2017; Neumann et al., 2013; Norberg et al., 2006)
Yes
Yes
Alcohol
Family history of diabetes
Yes
Yes
Year
Yes
Yes Yes
Yes
Yes
Yes
Yes
Yes
Yes
TE D
SDPP II (Östenson et al., 2012)
Yes
M AN U
Yes
SC
SALLS (Johansson et al., 2005) Stockholm County (Wändell et al., 2008)
Education
RI PT
Study name (ref)
Yes
Yes
Yes
Yes
SPHC (Carlsson et al., 2017)
Yes
Yes
Yes
Yes
Yes
Yes
SALT (Carlsson et al., 2017)
Yes
Yes
Yes
Yes
Yes
Yes
MDCS (Carlsson et al., 2017)
Yes
Yes
Yes
Yes
Yes
Yes
NMC (Carlsson et al., 2017)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
BMI = body mass index, SES = socioeconomic status
AC C
Note:
EP
NSC (Byhamre et al., 2017; Gustafsson et al., 2011)
Page 3 of 7
Sm to
ACCEPTED MANUSCRIPT Table 3:
Risk of type 2 diabetes and snus use in never smokers (adjusted for
variables given in Table 2 except smoking) Current use n RR (95% CI)
MONICA I (Eliasson et al., 2004) MONICA II (Eliasson et al., 2004)
a
Stockholm County (Wändell et al., 2008)
b
SDPP II (Östenson et al., 2012) ESTRID (Rasouli et al., 2016)
4
3.90 (1.10-14.30)
6
1.06 (0.43-2.64)
5
1.45 (0.54-3.87)
0
-
1
1.72 (0.20-14.8)
1
2.12 (0.25-17.71)
1
3.10 (0.36-26.84)
-
2.30 (0.50-9.80)
27
1.17 (0.58-2.37)
HUNT (Rasouli et al., 2016)
-
VIP (Carlsson et al., 2017)
11
SALT (Carlsson et al., 2017)
0.53 (0.20-1.39) -
165
1.28 (1.08-1.52)
74
0.84 (0.66-1.07)
25
1.09 (0.70-1.69)
19
1.10 (0.67-1.08)
13
0.66 (0.38-1.16)
M AN U
SPHC (Carlsson et al., 2017)
-
RI PT
SDPP I (Persson et al., 2000)
Former use n RR (95% CI
SC
Study name (ref)
43
0.95 (0.69-1.33)
8
0.60 (0.30-1.22)
7
0.93 (0.43-2.03)
-
0.38 (0.12-1.16)
-
Overall random-effects estimate (95% CI)
1.08 (0.86-1.34)
0.93 (0.79-1.11)
Heterogeneity chi-squared (degrees of freedom, p)
14.79 (10 df, p=0.14)
7.61 (7 df, p=0.37
MDCS (Carlsson et al., 2017) NMC (Carlsson et al., 2017)
a
b
Also available are results for diabetes diagnosed at oral glucose tolerance test: current use n = 1 RR = 0.91(0.10-8.01); former use n = 3 RR = 3.97 (0.86-18.33); ever use n = 4 RR = 2.03 (0.55-7.55)c Type of diabetes not specified Estimated from data given
EP
c
12
TE D
NSC (Byhamre et al., 2017)
-
Risk of type 2 diabetes and snus use in the whole population
AC C
Table 4:
(adjusted for variables given in Table 2)
Study name (ref)
Current use n RR (95% CI)
Former use n RR (95% CI
Ever use n RR (95% CI)
SDPP I (Persson et al., 2000)
13
1.50 (0.803.00)
5
0.80 (0.302.00)
18
1.20 (0.672.16)a
SHEP/VHEP (Hergens et al., 2005)ab
11
1.50 (0.762.90)
4
1.10 (0.403.30)
15
1.37 (0.762.46)a
2
1.35 (0.68-
SALLS (Johansson et
-
-
Page 4 of 7
0.94 (0.51-1.71)
ACCEPTED MANUSCRIPT al., 2005)b
2.67)
Stockholm County (Wändell et al., 2008)b
11
1.48 (0.72 3.03)
1
2.46 (0.3020.43)
12
1.54 (0.783.06)a
SDPP II (Östenson et al., 2012)
16
1.10 (0.602.00)
6
0.50 (0.201.20)
22
0.82 (0.491.39)a
ESTRID (Rasouli et al., 2016)
129
0.96 (0.671.37)
80
0.63 (0.410.95)
209
0.81 (0.601.08)a
157
0.91 (0.751.10)
-
-
Overall randomeffects estimate (95% CI)
1.18 (0.941.48)
0.69 (0.490.96)
0.95 (0.811.11)
Heterogeneity chisquared (degrees of freedom, p)
2.86 (5 df, p=0.72)
2.91 (4 df, p=0.57 )
5.61 (5 df, p=0.35)
SC M AN U
TE D
Estimated from data given Type of diabetes not specified
EP
b
AC C
a
RI PT
HUNT (Rasouli et al., 2016)
Page 5 of 7
ACCEPTED MANUSCRIPT Table 5: Dose-response results for the risk of type 2 diabetes and snus use among current users Population
Exposure category
Levels
SDPP I (Persson et al., 2000) Stockholm County (Wändell et al., 2008)b SDPP II (Östenson et al., 2012) ESTRID (Rasouli et al., 2016)
Never smokers All
Boxes/wk Cans/wk
0,1-2,3+ 0,<3,3+
VIP (Carlsson et al., 2017)
All All Never smokers All Never smokers All Never smokers Never smokers
Boxes/wk Boxes/wk Boxes/wk Box-years Box-years Boxes/wk Boxes/wk Boxes/wk
SPHC (Carlsson et al., 2017)
Never smokers
Duration (yr) Boxes/wk
SALT (Carlsson et al., 2017)
Never smokers
Duration (yr) Boxes/wk
MDCS (Carlsson et al., 2017)
Never smokers
Duration (yr) Boxes/wk
NMC (Carlsson et al., 2017) NSC (Byhamre et al., 2017)
Never smokers All
0,1-5,5+ 0,<5,5+ 0,<5,5+ 0,<10,10+ 0,<10, 10+ 0,<3,3+ 0,<3,3+ 0,1-2,3-4, 5-6,7+ 0,<30,30+ 0,1-2,3-4, 5-6,7+ 0,<30,30+ 0,1-2,3-4, 5-6,7+ 0,<30,30+ 0,1-2,3-4, 5-6,7+ 0,<30,30+ 0,1,2, 3,4
a
0.20 (0.00-2.00) 1.30 (0.49-3.40)
SC
M AN U Duration (yr) Follow-up periods
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
0.60 (0.20-1.40) 0.78 (0.56-1.09) 0.83 (0.41-1.71) 0.74 (0.52-1.06) 0.74 (0.31-1.77) 0.88 (0.72-1.08) 1.15 (0.72-1.82) 1.12 (0.81-1.54) 1.48 (1.09-2.02) 1.40 (1.06-1.84) No cases 2.62 (0.62-11.02) 1.96 (0.45-8.58) 1.17 (0.64-2.15) 1.02 (0.42-2.50) 0.82 (0.36-1.87) 1.19 (0.37-3.78) 0.71 (0.17-2.89) No cases 0.80 (0.39-1.64) 1.96 (0.44-2.12)
Estimated for never smokers only, based on a regression analysis in which the mean boxes/week used was taken as the midpoint of the range for closed dose intervals and as the geometric mean of the low value and 12 for open dose intervals. Using 20 rather than 12 had little effect on the overall estimate. Type of diabetes not stated.
AC C
EP
b
1.00 1.00
TE D
HUNT (Rasouli et al., 2016)
Adjusted RR (95% CI) by leve
RI PT
Study name (ref)
Page 6 of 7
2.70 (1.30-5 1.80 (0.67-4
3.30 (1.40-8 0.95 (0.57-1 1.01 (0.42-2 1.05 (0.67-1 1.00 (0.47-2 0.92 (0.46-1 0.89 (0.21-3 1.17 (0.89-1 1.79 (1.14-2 1.22 (1.00-1 1.50 (0.45-5 1.50 (0.20-1 1.17 (0.41-3 0.74 (0.45-1 1.51 (0.76-2 0.97 (0.68-1 0.32 (0.08-1 1.30 (0.18-9 1.50 (0.65-3 1.40 (0.67-2 0.57 (0.22-1
ACCEPTED MANUSCRIPT Table 6:
Risk of other endpoints and snus use in never smokers (adjusted
for variables given in Table 2 except smoking)
ESTRID (Rasouli et al., 2016)
LADA
13
0.98 (0.452.11)
SDPP I (Persson et al., 2000)
IGT
6
0.90 (0.402.10)
MONICA I (Eliasson et al., 2004)
IGT
5
0.78 (0.292.09)
MONICA II (Eliasson et al., 2004)
IGT
1
0.23 (0.031.80)
NSC (Byhamre et al., 2017)
MS
12
4
0.46 (0.151.43) -
1.15 (0.522.51)
Ever use n RR (95% CI) 17
0.75 (0.391.44)b -
6
1.48 (0.573.80)
11
1.05 (0.512.17)
2
0.75 (0.163.57)
3
0.42 (0.121.48)b
-
-
TE D
a
Former use n RR (95% CI
RI PT
Current use n RR (95% CI)
SC
Endpointa
M AN U
Study name (ref)
Abbreviations used: IGT = impaired glucose tolerance, LADA = latent autoimmune diabetes of adulthood, MS = metabolic syndrome Estimated from data given
AC C
EP
b
Page 7 of 7
ACCEPTED MANUSCRIPT Table 7:
Risk of other endpoints and snus use in the whole population
(adjusted for variables given in Table 2)
Endpointa
ESTRID (Rasouli et al., 2016)
LADA
VIP (Neumann et al., 2013
Former use n RR (95% CI
1.01 (0.671.54)
19
IFG
-
0.92 (0.821.03)
-
IFG/IGT
-
0.79 (0.591.05)
26
0.80 (0.501.30)
IRESIS
9
IRESP
8
-
Ever use n RR (95% CI) 64
0.83 (0.581.19)b -
-
-
19
0.70 (0.401.20)
45
0.75 (0.511.11)b
0.90 (0.402.00)
3
0.40 (0.101.30)
12
0.68 (0.331.40)b
1.20 (0.502.80)
12
2.20 (1.104.40)
20
1.70 (0.933.11)b
M AN U
IGT
0.60 (0.341.04)
SC
45
TE D
SDPP I (Persson et al., 2000)
Current use n RR (95% CI)
RI PT
Study name (ref)
IRESP
-
3.30 (0.912.10)
-
-
NSC (Gustafsson et al., 2011)c
MS
-
0.79 (0.331.86)d
-
-
-
0.96 (0.581.56)d
-
-
EP
SDPP II (Östenson et al., 2012)
a
b c d
e
MS
AC C
NSC (Gustafsson et al., 2011)e
Abbreviations used: IFG = impaired fasting glucose, IGT = impaired glucose tolerance, IRESIS = insulin resistance, IRESP = insulin response, LADA = latent autoimmune diabetes of adulthood, MS = metabolic syndrome Estimated from data given Males Adjusted for socioeconomic status at ages 16, 21 and 43, body mass index, systolic and diastolic blood pressure at ages 16 and 21, daily smoking, daily snuff use, alcohol consumption and physical inactivity at age 43 Females
Page 8 of 7
ACCEPTED MANUSCRIPT Table 8: Dose-response results for the risk of other endpoints and snus use among current users Level
SDPP I (Persson et al., 2000)
IGT (never smokers) IRESIS (never smokers) IRESP (never smokers) ESTRID (Rasouli et LADA (all) al., 2016) LADA (never smokers) LADA (all)
Boxes/wk
0,1-2,3+
Boxes/wk
0,1-2,3+
Boxes/wk
0,1-2,3+
Boxes/wk
0,<5,5+
Boxes/wk
0,<5,5+
Box-years
0,<10,10+
LADA (never smokers)
Box-years
0,<10,10+
1.00 0.70 (0.401.40) 1.00 0.50 (0.201.60) 1.00 2.10 (1.104.10) 1.00 0.71 (0.461.09) 1.00 0.75 (0.341.67) 1.00 0.56 (0.350.91) 1.00 0.46 (0.161.31)
0.80 (0.401.40) 0.70 (0.301.70) 1.20 (0.502.90) 1.18 (0.672.10) 0.67 (0.241.86) 1.45 (0.892.37) 1.01 (0.452.29)
EP
TE D
M AN U
Abbreviations used: IGT = impaired glucose tolerance, IRESIS = insulin resistance, IRESP = insulin response, LADA = latent autoimmune diabetes of adulthood
AC C
a
Adjusted RR (95% CI) by level
RI PT
Exposure category
SC
Study name (ref) Endpointa (population)
Page 9 of 7
ACCEPTED MANUSCRIPT Highlights
RI PT
We reviewed 18 studies that examined snus use and diabetes or related endpoints
Overall, current, ex or ever snus use showed no significant association with diabetes
SC
This was true for never smokers and the whole population
M AN U
Only heavy snus users who were never smokers had an increased risk of diabetes
AC C
EP
TE D
Other diabetes-related endpoints also failed to show any association with snus use
C:\pdfconversion\WORK\authordoc\Highlights.docx