Accepted Manuscript Toxicokinetic and toxicodynamic considerations when deriving health-based exposure limits for pharmaceuticals John F. Reichard, M. Andrew Maier, Bruce D. Naumann, Alison M. Pecquet, Thomas Pfister, Reena Sandhu, Edward V. Sargent, Anthony J. Streeter, Patricia A. Weideman PII:
S0273-2300(16)30140-4
DOI:
10.1016/j.yrtph.2016.05.027
Reference:
YRTPH 3587
To appear in:
Regulatory Toxicology and Pharmacology
Received Date: 5 May 2016 Accepted Date: 19 May 2016
Please cite this article as: Reichard, J.F., Maier, M.A., Naumann, B.D., Pecquet, A.M., Pfister, T., Sandhu, R., Sargent, E.V., Streeter, A.J., Weideman, P.A., Toxicokinetic and toxicodynamic considerations when deriving health-based exposure limits for pharmaceuticals, Regulatory Toxicology and Pharmacology (2016), doi: 10.1016/j.yrtph.2016.05.027. 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
2
Toxicokinetic and Toxicodynamic Considerations when Deriving Health-Based Exposure Limits for
3
Pharmaceuticals
5
Authors
6
1
7
1
8
2
9
1
10
3
11
4
12
5
13
6
14
7
RI PT
4
Reichard, John F. Maier, M. Andrew
SC
Naumann, Bruce D. Pecquet, Alison M.
M AN U
Pfister, Thomas Sandhu, Reena Sargent, Edward V. Streeter, Anthony J.
Weideman, Patricia A.
15 1
Toxicology Excellence for Risk Assessment (TERA) at the University of Cincinnati
17
2
Merck & Co., Inc.
18
3
F. Hoffmann-La Roche Ltd
19
4
SafeDose Ltd
20
5
Rutgers University
21
6
22
7
EP
AC C
Johnson & Johnson
TE D
16
Genentech
1
ACCEPTED MANUSCRIPT
Contact Author
24
John F. Reichard
25
Toxicology Excellence for Risk Assessment (TERA) at the University of Cincinnati
26
Department of Environmental Health
27
College of Medicine
28
160 Panzeca Way
29
Cincinnati, OH 45267
30
[email protected]
AC C
EP
TE D
M AN U
SC
RI PT
23
2
ACCEPTED MANUSCRIPT
Abstract
32
The purpose of this paper is to describe the use of toxicokinetic (TK) and toxicodynamic (TD) data in
33
setting acceptable daily exposure (ADE) values and occupational exposure limits (OELs). Use of TK data
34
can provide a more robust exposure limit based on a rigorous evaluation of systemic internal dose.
35
Bioavailability data assist in extrapolating across different routes of exposure to be protective for multiple
36
routes of exposure. Bioaccumulation data enable extrapolation to chronic exposures when the point of
37
departure (PoD) is from a short-term critical study. Applied in the context of chemical-specific adjustment
38
factors (CSAFs), TK data partially replace traditional default adjustment factors for interspecies
39
extrapolation (extrapolation from studies conducted in animals to humans) and intraspecies variability (to
40
account for human population variability). Default adjustments of 10-fold each for interspecies and
41
intraspecies extrapolation are recommended in several guidelines, although some organization
42
recommend other values. Such default factors may overestimate variability for many APIs, while not being
43
sufficiently protective for variability with other APIs. For this reason, the use of chemical specific TK and
44
TD data is preferred. Making full use of existing TK and TD data reduces underlying uncertainties,
45
increases transparency, and ensures that resulting ADEs reflect the best available science.
46
TE D
M AN U
SC
RI PT
31
Key Words
48
Human health risk assessment, Toxicokinetic, Toxicodynamic, Pharmacokinetic, Pharmacodynamic,
49
Chemical Specific Adjustment Factor (CSAF), Bioaccumulation, Bioavailability, Cmax, Area under the
50
curve (AUC)
AC C
EP
47
3
ACCEPTED MANUSCRIPT
51
1. Introduction
52
In pharmaceutical manufacturing, acceptable daily exposure (ADE) and permitted daily exposure (PDE)
53
values are intended to protect patients against potential adverse effects from cross-contamination by
54
active pharmaceutical ingredients (APIs) that may be present in pharmaceutical products as a result of
55
the manufacturing process (EMA, 2012; ISPE, 2010; US FDA, 2008). Likewise, occupational exposure
56
limits (OELs) are used to protect workers who manufacture or process pharmaceuticals. In recent years,
57
several guidelines have advocated the use of health-based risk assessment methodologies for setting
58
ADEs and OELs (EMA, 2014; ICH, 2013, 2011; ISPE, 2010). However, these documents provide limited
59
technical guidance on how to use toxicokinetic (TK) or toxicodynamic (TD) data in the risk assessment
60
process. This contributes to differences among limits derived by different risk assessors and companies,
61
even for the same API (Snodin and McCrossen, 2012; Walsh, 2011a,b; Walsh et al., 2013). The objective
62
of this paper is to describe various approaches for using TK data in the ADE derivation process and
63
provide descriptive examples to illustrate concept application. It should be noted that these same
64
approaches can be applied for OEL derivation with specific considerations given to the inhalation route of
65
exposure, as described below.
TE D
66
M AN U
SC
RI PT
1
On a molecular level, toxic effects are not produced unless the biologically active agent (toxicant or drug)
68
reaches the appropriate target molecule(s) (such as receptors or enzymes) in the body at a concentration
69
and for a length of time sufficient to cause toxicity. The toxicity that results can be predicted on the basis
70
of two inputs: TK, which is the delivery of the drug (or its active metabolite) to the target molecule(s) (i.e.,
71
what the body does to the chemical); and TD, which is the relationship between drug concentration at the
72
target molecule(s) and the resulting adverse response, including the time course and intensity of
73
response (i.e., what the chemical does to the body). For toxicants having a non-mutagenic mode of action
74
(MOA), when the concentration of chemical is below the biological threshold of the target molecule(s)
75
there is insufficient perturbation to elicit a meaningful (i.e., clinically significant) response. In this respect,
76
consideration of available TD data is important for identifying the key TK parameters that need to be
77
evaluated in the analysis; for example, the critical effect may be a consequence of peak plasma
AC C
EP
67
1
The term ADE is synonymous with the PDE (EMA, 2014)
4
ACCEPTED MANUSCRIPT
concentrations (Cmax) or the total plasma concentration over time (AUC). The threshold below which the
79
critical effect does not occur is used as the point of departure (PoD), and demarcates the no-observed-
80
adverse-effect level (NOAEL) from the lowest-observed-adverse-effect level (LOAEL) (Bercu et al., 2016,
81
this issue). The main advantage of using TK and TD data for risk assessment is that it becomes possible
82
to integrate an internal measure of dose and response, which provides a more meaningful basis for
83
estimating risk and enables extrapolations across routes, species, and different durations of exposure.
RI PT
78
84
In toxicology and risk assessment, the terms TK and TD are often used in preference to the terms
86
pharmacokinetic (PK) and pharmacodynamic (PD). In general, PK and PD are used in relation to
87
therapeutic doses of pharmaceuticals where the intent is to provide a link between preclinical studies and
88
humans, or to characterize or tailor the therapeutic use. In contrast, TK and TD are more recent
89
extensions of the PK and PD concepts that describe adverse effects occurring at supra-therapeutic
90
doses, and carry the connotation of harm rather than therapeutic benefit (Welling, 1995). When setting an
91
ADE or OEL for a pharmaceutical, the critical effect may be one that is not considered adverse when the
92
exposure is a result of therapeutic clinical use. In patients, the potential for an adverse effect can be offset
93
by the presumed medical benefit, whereas ADEs and OELs do not reflect a presumed medical benefit
94
and, therefore, pharmacologic effects are considered adverse.
M AN U
TE D
95
SC
85
There are a range of approaches for utilizing TK/TD data in human health risk assessment; from the use
97
of bioavailability and bioaccumulation ratios to much more complex biologically-based dose-response
98
(BBDR) models and physiologically-based pharmacokinetic (PBPK) models. BBDR models are based on
99
animal-derived data that describe biological processes at the cellular and molecular level. BBDR models
100
link external exposure to an adverse apical response and estimate the probability of an adverse response
101
in humans. In a BBDR model, the linkage between exposure and adverse response is based on
102
quantifiable biological events involved in effect progression, such as apoptosis, protein induction, or cell
103
division (Crump et al., 2010). The linkage between external exposure and target site exposure (e.g.,
104
tissue concentration) can be predicted by a PBPK model. These structured, mathematical models
105
integrate in vitro, in silico, and in vivo data to evaluate the effects of various factors such as species
AC C
EP
96
5
ACCEPTED MANUSCRIPT
differences, variability within populations (e.g., age, sex, genetics, or organ dysfunction), and route of
107
administration differences on target tissue concentrations, and can be a valuable tool in predicting human
108
safety. While PBPK models are not often used for setting ADEs or OELs, their use in drug development is
109
becoming increasingly common and the resulting models can be used for risk assessment. Adaptation of
110
these advanced drug development models for risk assessment purposes lies beyond the scope of this
111
paper; however, in the absence of BBDR or PBPK models, TK/TD data can be utilized to replace default
112
uncertainties in the derivation of ADEs and OELs, as described below.
SC
113
RI PT
106
Compared to many industrial chemicals, many pharmaceutical APIs are ideally suited for use of TK/TD
115
data for risk assessment because they have rich datasets. Even in the drug discovery or development
116
phase, key TK/TD data are typically known; such as the identity of the target tissue and molecules,
117
mechanism of action, and receptor occupancy estimates. In addition, TK data on bioavailability, tissue
118
distribution, drug protein binding, metabolism, and excretion may also exist. Initially these data are
119
obtained from in vitro assays and well-designed animal studies. Later in development, TK data are also
120
available from clinical trials and post-marketing information.
TE D
121
M AN U
114
It should be noted that current guidelines for using TK and TD data for derivation of ADE limits (ICH,
123
2011; IPCS, 2005; ISPE, 2010) are predicated on a linear relationship between a dose metric (i.e., the
124
measure of a dose) and a biologic response. For the vast majority of small molecule APIs this relationship
125
is valid for exposures at and below the range of clinically-relevant doses. However, when the biologic
126
processes that regulate absorption, metabolism, and elimination are overwhelmed, the linear relationship
127
between the dose metric (i.e., AUC and Cmax, discussed below) and response is lost. Likewise, for some
128
biologic APIs, such as monoclonal antibodies, there can be nonlinearity in the sensitivity of the biologic
129
response. In either case, non-linearity in the range of clinically-relevant doses is problematic because it
130
results in unpredictable relationships between dose and response, which can include exaggerated on-
131
target pharmacology, off-target side effects, or therapeutic failure (if the API does not achieve adequate
132
systemic levels). Such non-linear effects are often observed in nonclinical animal studies, especially in
133
toxicokinetics, as dose levels increase towards maximally tolerated doses (MTD). When the relationship
AC C
EP
122
6
ACCEPTED MANUSCRIPT
between the dose metric and response lacks linearity, standard risk assessment methods for making
135
animal to human and inter-human adjustments cannot be used. However, careful evaluation of TK/TD
136
data can identify a more appropriate dose metric (Kirman et al., 2003) or a lower dosing range where
137
linearity exists. There are some APIs, such as biologics, where non-linearity exists in either the kinetics or
138
response behavior at clinically-relevant doses due to the biology associated with their MOA or kinetic
139
disposition (Muller et al., 2009). In these situations, applications of more sophisticated PBPK models may
140
be useful. One such approach is to use kinetic and dynamic data to computationally determine the
141
minimum-anticipated-biological-effect level (MABEL) from which a human equivalent dose (HED) at the
142
predicted NOEL can be obtained (Muller et al., 2009).
144
M AN U
143
SC
RI PT
134
2. Background and History: The Move from Default Adjustment Factors to Data-Derived Approaches
145
The most commonly used adjustment factors in ADE derivations are those used to account for human
147
variability (interindividual or intraspecies variability - AFH) for the toxic response and those used for
148
interspecies extrapolation (AFA) when applying animal toxicity data to humans. In the absence of data, a
149
default adjustment factor of 10 is commonly used to account for human variability. Likewise, in the
150
absence of data, a default adjustment factor of either 10, or allometric scaling factors , are used to
151
extrapolate from an experimental animal study to humans. These consensus default factors of 10 evolved
152
from the earliest risk assessments performed by the Food and Drug Administration (FDA) where
153
biological data were limited and a factor of 10 for both interspecies extrapolation and human
154
(intraspecies) variability were combined, culminating in a “100-fold safety factor” (Lehman and Fitzhugh,
155
1954). Over the decades, as the scientific knowledge base grew and evolved, the paradigm shifted from
156
application of these default factors or allometric scaling factors to application of quantitative chemical-
157
specific adjustments, which provide a more robust scientific rationale for the overall risk assessment
158
(IPCS, 2005; Renwick, 1993, 1991; Sussman et al., 2016, this issue).
2
AC C
EP
TE D
146
159 2
Allometric scaling factors account for changes in basal metabolic rate that occur with changes in scaling animal size.
7
ACCEPTED MANUSCRIPT
The shift away from using default factors began with the proposal of Renwick (1991) to subdivide the AFH
161
and AFA default values of 10 (each) into two subfactors, one that accounts for differences in TK and the
162
other accounting for differences in TD. This subdivision serves a useful function when data are available
163
because the TK and TD data can be separately utilized to replace portions of the default value, as
164
appropriate. In a subsequent publication, Renwick (1993) determined that TK differences are typically
165
greater than TD differences and, therefore, the AFH and AFA safety factors should be unequally divided
166
into factors of 4 for TK and 2.5 for TD. The International Programme on Chemical Safety (IPCS) adopted
167
the principles set forward by Renwick (1993) but suggested that only the AFA (interspecies) factor be
168
divided unequally into 4-fold (TK) and 2.5-fold (TD), while the AFH (intraspecies) factor should be split
169
evenly (3.16 for both TK and TD) because the database analyzed was insufficient to justify an uneven
170
subdivision of the 10-fold factor for human variability (Figure 1) (IPCS, 2005).
M AN U
SC
RI PT
160
171
Based on this concept, the IPCS produced guidelines on the data needed to replace defaults with
173
empirically determined data. These chemical-specific adjustment factors (CSAF) quantify inter- and
174
intraspecies differences and variability associated with TK and TD parameters. More recently, the US
175
Environmental Protection Agency (US EPA, 2014) also introduced data-derived extrapolation factors
176
(DDEF) for intraspecies and interspecies extrapolation. The DDEF approach also moves away from the
177
10-fold framework of individual default factors (Figure 1). There are numerous papers illustrating how to
178
use and apply TK/TD data to replace default values for pharmaceuticals (Gould et al., 2013; Naumann et
179
al., 2004, 2001, 1997; Sargent et al., 2002; Silverman et al., 1999), and the guidance documents for
180
CSAF and DDEF derivation provide useful examples for the application of such data. Since the focus of
181
this communication relates to pharmaceuticals, and since the DDEF guidelines are a recent development
182
for application to environmental chemicals, the emphasis herein is on the CSAF guidelines.
EP
AC C
183
TE D
172
8
SC
RI PT
ACCEPTED MANUSCRIPT
184
Default adjustment factors, by definition, should only be used in the absence of relevant suitable data.
186
When sufficient chemical-specific data are available, CSAFs should be used to replace portions of the
187
default factors for AFA and AFH. TK and TD data are used to derive CSAFs in accordance with the
188
published guidelines (IPCS, 2005; Meek et al., 2002; Sussman et al., 2016, this issue; US EPA, 2014).
M AN U
185
189
3. CSAF Values for Intra- and Interspecies Adjustment
191
Developing CSAF values using TK data requires knowledge of the relationship between the external dose
192
and the internal target concentrations at the site of action for the critical effect. This knowledge applies to
193
the AFA differences between humans and animals and the AFH variability resulting from susceptible
194
human subpopulations. The ideal internal dose (internal exposure) would be the free drug or
195
toxicologically active metabolite concentration at the critical site of action in both humans and animals,
196
although such complete datasets are not typically available. Instead, drug concentrations in the plasma,
197
or other appropriate fluids, from either nonclinical or clinical studies, are commonly used as surrogates for
198
estimating the drug exposure at the critical site of action.
199
AC C
EP
TE D
190
200
3.1.
Choosing the Appropriate CSAF Parameter
201
The most commonly used internal dose metric is the area under the (concentration-time) curve (AUC) in
202
the fluid sampled (Figure 2). Especially for toxicants that bind covalently or cause irreversible damage, an
203
integrated measure of dose over time such as AUC is considered to be the most appropriate (US EPA,
204
2014). This parameter implicitly assumes that the effect of the chemical on the target tissue is linear over
9
ACCEPTED MANUSCRIPT
both concentration and time (Clewell et al., 2002). Less commonly, adjustments are based on an effect
206
for which peak plasma concentration is critical (e.g., interference with ion channels that can result in
207
arrhythmias or central nervous system effects), in which case maximum observed plasma concentration
208
(Cmax) is a better choice for dose metric. Other TK metrics used for CSAF determination include clearance
209
rates, glomerular filtration rates, and others that, while not direct measures of exposure, are correlated
210
with the critical target tissue concentration. Selection of the most appropriate dose metric is not always
211
easy, as it requires knowledge of the relationship between the TK profile and occurrence of the critical
212
effect. Guidance for selection of the most appropriate TK parameter and dose metric can be found in the
213
CSAF and DDEF guidance documents (IPCS, 2005; US EPA, 2014).
214
AC C
EP
TE D
M AN U
SC
RI PT
205
215
Indications that Cmax may be a good dose metric include adverse effects that are reported shortly after
216
dosing, or those that are transient and reversible, such as headache, nausea, and acute phase reactions
217
after parenteral administration. Similarly, effects such as irritancy and asphyxia are more likely to be
218
related to Cmax rather than AUC. In such instances, the concentration of chemical at the target site
219
exceeds a biological threshold concentration that acutely elicits the adverse effect (Clewell et al., 2002).
220
This could be attributable to interactions with essential cell or tissue targets that result in acute disruption
221
of biological activity, including binding with receptors (e.g., opioid receptor), enzymes (e.g., glutamine
10
ACCEPTED MANUSCRIPT
synthetase), critical cellular function proteins (e.g., mitochondrial electron transport proteins), or ion
223
channel disruption (e.g., sodium/potassium ion channels). At chemical concentrations below the critical
224
threshold, the stoichiometric interaction of the chemical with the target is insufficient to elicit a response
225
(e.g., depolarization); while at sufficiently high concentrations a sufficient proportion of the target
226
molecules (e.g., ion transport channels) are perturbed such that a response is initiated leading to the
227
observed adverse effect (e.g., seizure).
RI PT
222
228
The peak observed concentration, Cmax, is a function of the rate of absorption, volume of distribution (Vd),
230
and the rate of elimination. In the simplest theoretical circumstance, when a substance is given by rapid
231
bolus injection, Cmax can be described as the quotient of dose/volume of distribution. In reality, following
232
exposure to a substance, two processes act to lower the Cmax: elimination (metabolic, renal, etc.) which
233
removes the substance, and distribution which dilutes the substance by partitioning it to various distal
234
compartments and by binding to plasma or other proteins. For this reason, the magnitude of Cmax
235
depends on the rate at which the substance reaches systemic circulation (e.g., via absorption or infusion).
236
The time period required for a substance to reach Cmax is defined as Tmax (Figure 2). The importance of
237
the relationship between Cmax and Tmax is that a longer Tmax will typically result in a lower Cmax.
238
Conversely, a shorter Tmax can equate to a higher Cmax because there is less time for the substance to be
239
distributed and eliminated. This is particularly relevant to gavage studies where a substance is given as a
240
single oral bolus dose which results in a much shorter Tmax and much higher Cmax than if the same dose of
241
the substance were dispensed over time, such as in a dietary feeding study. For this reason, both Cmax
242
and Tmax must be considered when there is concern for a peak-related effect, and should be evaluated in
243
the context of ceiling limits as specified by the Threshold Limit Value guidelines for occupational
244
exposures established by the American Conference of Governmental Industrial Hygienists (ACGIH,
245
2016).
246
AC C
EP
TE D
M AN U
SC
229
247
3.2
Intraspecies Toxicokinetic Variability (AFH)
248
As described earlier, in the absence of data a default factor of 10 has been applied historically for AFH
249
[alternatively referred to as the F2 factor in the European Medicines Agency (EMA) guidance documents]
11
ACCEPTED MANUSCRIPT
based on a scientific rationale documented in a number of reviews (Dourson and Stara, 1983; Lehman
251
and Fitzhugh, 1954). This adjustment factor accounts for both TK and TD variations within the human
252
population (interindividual variability) and is intended to protect subpopulations that may be more
253
susceptible to toxic effects from chemical exposure due to their age, sex, genetics, pre-existing diseases,
254
etc. (IPCS, 2005). However, empirical data suggest that in some cases the application of a default factor
255
of 10 is greater than needed and thus overly stringent. When overly protective assumptions are applied
256
throughout the risk assessment process for pharmaceutical manufacturing, the cumulative consequences
257
can lead to unnecessary production costs, increased efforts related to safe handling by workers, and
258
unnecessary use of specialized production facilities. Conversely, the default 10-fold factor may not be
259
sufficiently protective. This possibility exists when known genetic polymorphisms in metabolic pathways
260
(poor metabolizers) result in higher plasma levels of the substance in those individuals than the general
261
population without the polymorphism (extensive metabolizers). Naumann et al. (2004) used timolol
262
maleate to illustrate a situation where a default factor of 10 is insufficiently protective. In this example, a
263
polymorphism in CYP2D6 results in a bimodal distribution of the human population as either poor or
264
extensive metabolizers, yielding an AFH of 9.8 for timolol maleate. When combined with the TD subfactor,
265
the resulting CSAF is 12-31 for interindividual variability, depending on the whether the default TD
266
subfactor of 3.2 is used, or a TD subfactor of 1.2 based on beta-adrenergic receptor binding data.
SC
M AN U
TE D
267
RI PT
250
The extent to which the study population is representative of the general population relies on the ratio of
269
the tail of the distribution to the central tendency for the appropriate TK parameter (IPCS, 2005; Meek et
270
al., 2002; US EPA, 2014) (Equation 1). This approach is based on the premise that if a subpopulation
271
(e.g., tail of the population distribution) is sufficiently different from the general population (i.e., a
272
significantly higher AUC or Cmax for the same dose), the acceptable exposure should to be reduced
273
(adjusted downward) to coincide with the exposure in normal (average) healthy individuals (Sussman et
274
al., 2016, this issue). Where two or more distinct subpopulations exist (bimodal distribution), the ratio of
275
the upper tail of the most sensitive subpopulation over the central tendency of the remainder of the
276
population is used to derive an appropriate adjustment factor (Naumann et al., 1997; IPCS, 2005; US
277
EPA, 2014). In either case the subfactor replaces the default AFH of 3.16 for the TK fraction, after an
AC C
EP
268
12
ACCEPTED MANUSCRIPT
278
evaluation of the sensitivity of the study population to be sure it covers the full variability for the target
279
population. =
280
~
Equation 1.
282
3.3
283
The AFA accounts for TK and TD variations across different species and takes into account differences in
284
metabolic rates and physiological parameters between species for extrapolation. Default values for AFA
285
are variable: some guidelines recommend a factor of 10 be used, which is subdivided into factors of 4.0
286
for TK and 2.5 for TD (e.g., IPCS, 2005) (Figure 1). Other guidelines (e.g., ICH Q3C) recommend an AFA
287
adjustment factor based on allometric scaling and refer to it as the “F1” factor (ICH, 2011).
M AN U
SC
Interspecies Toxicokinetic Extrapolation (AFA)
RI PT
281
288
Allometric scaling describes the relationship between body weight or surface area and metabolic or
290
physiological differences (e.g., breathing rates, blood flows, etc.) between species. Scaling can be
291
performed based on comparative body surface area (W0.67) or on basal metabolic rate (W 0.75) (EFSA,
292
2012; ICH, 2011; Rennen et al., 2001; US EPA, 2011; US FDA, 2005). Selection of a scaling approach
293
largely depends on the regulatory agency guidelines being followed. The proposed update to the Risk-
294
MaPP guidelines (ISPE, 2010) does not specify a preference between allometric scaling based on W
295
an approach in widespread use for determining dose levels for first-in-human clinical trials (US FDA,
296
2005), or by W
297
area (W
298
basal metabolic rate approach (W
299
metabolic rate and the 0.75 exponent provides better scaling and is the preferred relationship in
300
veterinary medicine. However, the motivation for using the 0.67 exponent for human clinical trials is
301
related to safety; it provides a lower dose estimate and therefore a larger safety margin. In the absence of
302
satisfactory toxicokinetic studies, the default approach that scales an animal dose to an HED can be
303
based on body surface area, as shown in Equation 2.
,
as recommended by the US EPA (2011). The ICH Q3C guidelines specify the surface
AC C
) approach, whereas the EMA guidelines for first-in-human clinical trials refer to the US FDA 0.75
). Scientifically, body surface area is an imperfect estimate of
304 305
0.67
EP
0.67
0.75
TE D
289
The default allometric scaling values based on body surface area are described mathematically as:
13
ACCEPTED MANUSCRIPT
306
HED = !"# $ %&'( "! )#*/,*-. × 0
1 234 )3- 78.:;. 451 234 )3-
6
Equation 2.
Allometric scaling provides a default factor that accounts only for TK differences in body surface area or
308
metabolic rate. In some cases, it is not appropriate to scale doses on the basis of body mass alone, since
309
other TK factors override the effects of scaling. Highly protein-bound drugs should be adjusted for the
310
unbound fraction before scaling; blood volume and lung surface area increase in direct proportion to body
311
mass and scale with an exponent close to 1. Chemicals that undergo active transport and those
312
undergoing significant biliary excretion or renal secretion are not likely amenable to simple allometric
313
scaling (Sharma and McNeill, 2009).
SC
RI PT
307
314
A consequence of the fact allometric adjustments capture only physiologic factors that scale with
316
metabolism (e.g., uptake and non-metabolic clearance processes) means that it may not account for
317
species-specific differences in TD (e.g., differences in target expression, receptor binding, pathway
318
activities). Allometric scaling assumes that all species are equally sensitive to the adverse effects of a
319
compound when adjusted for the metabolic rate (or body surface area) of the organisms; however, this
320
may not be the case. For this reason, there are instances where both metabolic rate and toxicodynamic
321
differences contribute to differences in sensitivity between species (Rhomberg and Lewandowski, 2004).
322
For example, if the PoD was based on toxicity studies conducted in mice and it is shown that humans are
323
substantially more sensitive to the toxicodynamic effect at the same internal exposure, it would be most
324
appropriate to apply a CSAF to account for both TK and TD factors.
TE D
EP
325
M AN U
315
When sufficient chemical-specific data are available, deriving a CSAF for AFA provides a more robust and
327
scientific approach for deriving non-default values than allometric scaling factors (IPCS, 2005). The
328
assumption is that humans are more sensitive than the animal species from which the PoD was derived,
329
and by employing an adjustment factor to address AFA differences the ADE is reduced accordingly. In
330
some cases, animals may be more sensitive, as is the case when higher enzyme activity results in
331
greater bioactivation of a less toxic parent substance to more toxic metabolites.
AC C
326
332 333
As with the AFH, an initial decision is made with regard to whether the critical effect is likely to be related
14
ACCEPTED MANUSCRIPT
to the Cmax or to the AUC. In general, it is reasonable to assume that effects that emerge over time (i.e.,
335
with subchronic or chronic exposure) are related to the AUC, whereas acute effects could be related to
336
either AUC or Cmax. The adequacy of the available data is considered by taking into account aspects of
337
the critical studies such as the routes of administration, nature of the population, administered doses, and
338
sample size. The numbers of animals and humans in the dataset should be sufficient to ensure that the
339
data allow a reliable estimate of the central tendency for each species. As a guide, the standard error
340
[standard deviation (SD) of the sample divided by the square root of the sample size] should be less than
341
approximately 20% of the mean. The CSAF is calculated by the ratio of the kinetic parameter in animals
342
divided by the same kinetic parameter in humans (Equation 3), assuming the chosen parameter is
343
properly adjusted for confounding factors such as dose, administration frequency, period of integration
344
(length of time used in calculating the AUC), etc. In some instances, the toxicokinetics dataset may not
345
allow satisfactory evaluation of Cmax and AUC parameters, but does establish clearance rate. In such
346
cases, clearance can also be used to calculate a CSAF; however, it is important to note that clearance
347
rates are inversely related to AUC and Cmax [clearance = (bioavailability x dose)/AUC], therefore, the
348
greater the clearance, the lower the AUC and Cmax. For this reason, the clearance ratio is the reciprocal
349
(animal / human) to AUC and Cmax, as shown in Equation 3.
351
SC
M AN U
TE D
350
RI PT
334
< =
=>? @
F @
×
@HIJK
@JKLIJM
Equation 3.
4. Bioavailability and Route Specific Adjustments Using TK/TD Data
353
The relative bioavailability of a substance via the administration route in the critical study and the route of
354
interest can be used to adjust the PoD to reflect differences in route-specific bioavailability. Bioavailability
355
is a measure of internal exposure that describes the fraction of a chemical or its active moiety that is
356
systemically absorbed and distributed in the body. Route-to-route extrapolation is necessary when
357
attempting to derive an ADE from a study conducted by one route (e.g., oral) that is different from the
358
potential route of exposure (e.g., intravenous). Although direct studies using the relevant route of
359
administration are preferred, extrapolation using a bioavailability correction factor may be useful to ensure
360
adequate protection in these situations (Naumann et al., 2009). To illustrate the importance of
361
bioavailability corrections, consider the situation where an ADE for product A (an oral drug) was used to
AC C
EP
352
15
ACCEPTED MANUSCRIPT
support cleaning limits for product B (an intravenous drug) in an API manufacturing facility. The oral
363
bioavailability of product A is 1%, i.e., only 1% of the administered dose is systemically available in the
364
blood. If product B is administered parenterally but this difference in route is not accounted for, the
365
amount of allowable carryover would be two orders of magnitude higher for the parenteral dosage form
366
(given that product A will have 100% bioavailability from the parenteral route). For this reason, the ADE
367
for product A that is protective for parenteral exposure is 100-fold lower than the ADE for the oral route.
RI PT
362
368
Route-to-route extrapolation is most relevant for relating non-local effects to systemic exposure levels
370
(Sargent et al., 2013). To relate external dose to internal dose, a correction factor is used to account for
371
differences in bioavailability by different routes, particularly for pharmaceutical industry OELs where the
372
critical effect is typically derived from oral or parenteral studies and the primary exposure of concern is
373
inhalation. Assessing the ratio of the same toxicity endpoint via multiple routes of administration for the
374
same chemical can also assist in establishing whether toxicity is greater via one route versus another. For
375
example, if the NOAEL for a study conducted via the oral route is 30 mg/kg-day, and the NOAEL is 3
376
mg/kg-day for a similar study conducted via the intravenous route, decreased bioavailability via the oral
377
route may account for this 10-fold difference. In the absence of data, a protective default assumption of
378
100% bioavailability is often assumed for the inhalation, subcutaneous, dermal, and all parenteral
379
exposure routes for small molecules, unless there is evidence indicating otherwise. There are a number
380
of route-specific metabolic considerations related to bioavailability, as discussed in the following sections.
EP
TE D
M AN U
SC
369
381 4.1
Default Adjustments for Individual Route of Exposure
383
4.1.1
384
Lower respiratory tract deposition is considered to be more relevant to overall bioavailability of inhaled
385
chemical aerosols than the total inhaled dose, and intra-nasal deposition or instillation often (but not
386
always) has limited bioavailability. Including data on the particle size distribution, potential regional
387
distribution from inhalation, and solubility in PBPK modeling could greatly increase the accuracy of OELs
388
and ADEs for inhalation exposure over the highly precautionary practice of assuming 100%, which is
389
currently practiced by many organizations (Naumann et al., 2009; RIVM, 2002). Dosimetry models
AC C
382
Inhalation
16
ACCEPTED MANUSCRIPT
390
available to assist in estimating deposition of aerosols are available for risk assessment (Kuempel et al.,
391
2015). Although dosimetry methods such as computation fluid dynamics are available for assessing gas
392
phase exposures, such exposures are less common for pharmaceuticals.
RI PT
393 Inhalation differs from other routes of exposure because determining the portion of a dose delivered to
395
the deep lungs, where most absorption occurs, is a complicated matter. Depending on the size and
396
aerodynamics of the inhaled particles, inhaled aerosols can deposit in the nose, mouth, throat, upper and
397
lower airways, and deep lung, and a fraction of the inhaled mass typically escapes during exhalation.
398
Particles >10 µm can be expected to deposit via sedimentation or impaction in the nasopharynx and
399
tracheobronchial passages. In humans, particles that deposit in the upper airway are generally cleared
400
via the mucocilliary escalator and subsequently swallowed, which can, for some molecules, provide a
401
secondary (oral) route of exposure. Particles with sizes in the ranges of 1–5 µm in diameter are most
402
efficiently deposited in the deep lungs (pulmonary region), with >50% of the 3 µm diameter particles being
403
deposited in the alveolar region (Labiris and Dolovich, 2003). With the growing interest in biologic
404
pharmaceuticals, occupational exposure to inhaled polypeptides and proteins has become a growing
405
concern. For large therapeutic proteins (> 40 kD), there is evidence that the bioavailability by inhalation is
406
maximally 5% (Pfister et al., 2014), while the oral bioavailability is negligible in the gastrointestinal tract
407
due to proteolysis.
409
4.1.2
410
The gastrointestinal (GI) tract subjugates substances to a combination of conditions not associated with
411
any other route of administration. GI flora and enzymes released into the GI tract mediate chemical
412
changes including hydrolysis, oxidation/reduction, and conjugation; lipophilic substances are emulsified
413
for increased intestinal absorption; a host of transport systems are present that efficiently move
414
substances across membrane barriers; and well perfused microvilli dramatically increase the surface area
415
available for absorption. In addition, local conditions in various regions of the GI tract can alter the
416
ionization state of substances (e.g., weak acids and bases), which can facilitate absorption in the
417
stomach where pH levels are below the pKa values of many weak acids (ECHA, 2010).
AC C
Oral
EP
408
TE D
M AN U
SC
394
17
ACCEPTED MANUSCRIPT
In the absence of specific data on oral bioavailability, defaults are often suggested. ECHA (2010) has
419
recommended a bioavailability default of 50% in the absence of data when there is evidence that there is
420
no first-pass effect, reabsorption, or bolus effect (Sargent et al., 2013). Surrogate data and potential ways
421
to predict oral absorption and bioavailability are still proving difficult to model, with few, if any, simple rules
422
showing promise in offering a qualitative way to set defaults. Tian et al. (2011) showed that rule-based
423
systems [e.g., Lipinski’s rule of five; Lipinski et al., (2001)] used for predicting absorption based on
424
molecular properties [e.g., molecular weight, partition coefficient (Log P), water-solubility, etc.] are not
425
particularly good predictors of oral bioavailability because they do not account for the influence of
426
important metabolic processes, such as liver metabolism. Rather, such simple rules only show good
427
performance for classifying intestinal absorption, not for classifying oral bioavailability. In fact, molecules
428
with low metabolism are more closely correlated with model predictions than those with higher metabolic
429
reactivity. This potentially underscores the importance of first-pass metabolism in any model that hopes to
430
predict bioavailability via the oral route (Tian et al., 2011).
M AN U
SC
RI PT
418
431
First-pass metabolism is a form of pre-systemic clearance that can be a critical determinant of systemic
433
bioavailability for parent molecules. First-pass metabolism decreases the apparent absorption of a
434
chemical before the chemical reaches the systemic circulation. In some cases, however, first-pass
435
metabolism can rapidly transform a less toxic parent molecule into the active toxicant (metabolite)
436
responsible for adverse effects in tissues distant from the site of metabolism (e.g., cyclophosphamide). In
437
theory, first-pass metabolism can occur via any route of administration where enzyme systems are
438
present in sufficient quantity to prevent absorbed substances from reaching systemic distribution,
439
including the lungs, skin, and other tissues. However, the liver is the primary site for first-pass metabolism
440
due to its importance as a site of biotransformation and as the portal of entry for substances carried by
441
the mesenteric blood flow from the GI tract. Pulmonary metabolism of some substances also occurs
442
(Borlak et al., 2005), but few substances are reported to undergo a quantitatively important pulmonary
443
first-pass metabolism (ECHA, 2010).
AC C
EP
TE D
432
444
18
ACCEPTED MANUSCRIPT
First-pass metabolism is a major consideration impacting bioavailability from the oral [and intraperitoneal
446
(IP)] routes of exposure that is not typically associated with other routes of exposure. Liver first-pass
447
metabolism can dramatically reduce systemic exposure to substances (or increase exposure to
448
metabolites) by the oral and IP routes of administration. When a substance is administered by either of
449
these routes, it is absorbed primarily via the mesenteric circulation, which, after perfusing the small
450
intestines and a portion of the colon, empties to the portal vein and is delivered to the liver. In this way,
451
100% of the mesenteric blood flow, and all substances absorbed from the intestines or peritoneal space
452
pass through the liver before entering the general circulation (as parent compound or metabolites). As a
453
result, the same dose administered via another route (e.g., inhalation, intravenous) will result in
454
substantially higher systemic exposures, as measured by Cmax and AUC. In general, oral studies do not
455
provide sufficient data to distinguish the relative contributions of absorption vs. first-pass metabolism to
456
decreased bioavailability, and both are observed jointly as a relative decrease in systemic exposure. It is
457
also important to recognize that the liver can be a target of toxicity in the absence of systemic exposure.
458
Since the liver receives 100% of the mesenteric blood flow, a chemical substance can be absorbed from
459
the GI and delivered to the liver where it exerts its toxicological effect. This can be an effect of the parent
460
chemical or as a result of an active metabolite. Administration of the same chemical via other routes could
461
likely produce a similar effect; however, systemic delivery to the liver would be required, albeit at lower
462
plasma concentrations due to dilution, and in such cases, other target tissue effects might also be
463
observed.
SC
M AN U
TE D
EP
464
RI PT
445
Liver metabolism may be either a low extraction (capacity-limited) process, in which the enzymatic
466
capacity of the liver is the limiting factor, or a high extraction (blood flow-limited) process, where the
467
limiting factor for the rate of metabolism is the rate at which the chemical is delivered to the liver. The
468
expression and activity of many hepatic metabolizing enzymes can be perturbed by a host of factors
469
including age, environmental factors (e.g., diet, lifestyle), gender, pregnancy, and concurrent
470
pharmaceutical use. When enzyme systems mediating first-pass metabolism are perturbed by any such
471
factors, the effect is usually in the direction of a reduction in metabolizing capacity for high extraction
472
chemicals since the responsible metabolic pathways are already performing with high efficiency;
AC C
465
19
ACCEPTED MANUSCRIPT
therefore, perturbations that increase activity have little additional effect on clearance while inhibition can
474
dramatically reduce efficiency. Conversely, perturbations of the activity of metabolizing enzymes (e.g.,
475
enzyme induction) have the greatest impact on low extraction chemicals that undergo capacity-limited
476
clearance since these perturbations can both increase or decrease the metabolizing activity of these
477
systems (Fan et al., 2015). Once a chemical is systemically available, the kinetic disposition is the same
478
regardless of exposure route.
RI PT
473
479
It should be noted that chemicals with a PoD that is derived locally at the initial site of contact will have
481
route-specific toxicity. In these cases, route-to-route extrapolation is not considered appropriate. For
482
example, when irritancy is the concern, the effect is often dependent on concentration at the site of
483
contact rather than systemic dose. When route-specific toxicity exists, it may be necessary to establish
484
both a concentration limit for the chemical in question as well as a systemic ADE. Another situation in
485
which route-to-route extrapolation is not considered appropriate is hypersensitivity, where the types of
486
reactions tend to be route specific (Gould et al., 2016, this issue).
M AN U
SC
480
487 4.1.3
489
There are a number of parenteral routes of administration including: intravenous, intramuscular,
490
subcutaneous, intraosseous, intraventricular, epidural, intracardiac, intraarticular, intracavernous,
491
intravitreal, and others. Only a few of these are common administration routes, but similar considerations
492
on adjusting for route of exposure should be applied to each. In many cases, standards exist for daily
493
intake of substances in food, water, air, and occupational exposure but without information for parenteral
494
administration. Where data are available but not considered sufficient for a safety assessment, the ICH
495
Q3D guideline for elemental impurities in pharmaceutical products lays out the following guidelines for
496
using oral standards for parental ADEs (ICH, 2013):
AC C
EP
Parenteral Routes
TE D
488
497
•
498
•
Oral bioavailability < 50% - divide by a modifying factor of 10
499
•
Oral bioavailability between 50% and 90% - divide by a modifying factor of 2
500
•
Oral bioavailability > 90% - divide by a modifying factor of 1
Oral bioavailability < 1% - divide by a modifying factor of 100
20
ACCEPTED MANUSCRIPT
501
However, there are some limitations associated with these default values for general application. The
502
guidelines do not address what to do when there are no data indicating absolute bioavailability via the
503
oral route, or how potential oral bioavailability should be estimated.
505 506
4.2
RI PT
504 Bioavailability Correction Factor (BCF) to Replace Default Adjustments for Route of Exposure
Absolute bioavailability (F) is defined as the dose-normalized bioavailable fraction by any extravascular
508
route (e.g., oral, dermal, etc.) divided by the dose-normalized intravenous bioavailable fraction (presumed
509
to be 100%) (Equation 4). = 100 ×
×
@LS
@PQHR
Equation 4.
M AN U
510
SC
507
Relative to direct intravenous administration, other routes can have some degree of absorption
512
inefficiency resulting in either decreased or delayed systemic bioavailability. In some instances, when
513
data from the relevant route are not available, bioavailability adjustments are useful to ensure adequate
514
exposure protection. If the differences are quantitatively relevant, a bioavailability correction factor (BCF)
515
should be applied (Naumann et al., 2009), resulting in route-specific ADE values. A BCF, also referred to
516
as alpha (α) in some references (e.g., Sargent and Kirk, 1988), is defined as the bioavailability via the
517
exposure route of interest divided by the bioavailability via the route used in the critical study (Naumann
518
et al., 2009), as shown in Equation 5. The BCF is equivalent to 1/F and is included in the denominator of
519
the limit derivation.
EP
520
TE D
511
TU &V $Wℎ Y. =
% [@\] ^ \ 4 E@5 @5 .3.,4 @. % [@\] ^ \ 4 5`^ @5 .3.,@.
Equation 5.
In practice, the ADE is derived from the PoD based on the route used in the critical study by placing the
522
BCF (α) in the denominator of the typical ADE equation adapted from Naumann et al. (2009). As an
523
example, consider an ADE derived based on an effect observed from an oral route of administration. If
524
the systemic bioavailability for the oral route is 10% and the bioavailability by the inhalation route is
525
assumed to be 100%; the resulting BCF would be 100/10 or a factor of 10. Thus, application of this factor
526
lowers the ADE by a factor of 10 to account for the greater bioavailability via the inhalation route. This
527
adjustment estimates the intravenous equivalent of the oral PoD on an internal dose basis, and would be
528
relevant to a manufacturing scenario where a manufacturing line was used to produce an oral drug which
AC C
521
21
ACCEPTED MANUSCRIPT
529
was followed by production of a parenteral drug. Likewise, a similar situation would occur if a PoD is
530
based on oral dosing studies but the concern is for occupational inhalation exposures.
531
533
There are a number of special considerations to be taken into account when applying a BCF:
RI PT
532
1. When extrapolating from oral or intraperitoneal animal data to other routes (e.g., inhalation,
534
dermal) between species, consideration should be given to differences in the rate and extent of
535
liver metabolism (e.g., first-pass clearance) between humans and the study species.
2. The rate of absorption can be very different for different routes of exposure. For example, in
537
humans, oral and dermal exposure tends to be slower than inhalation. For this reason,
538
consideration should be given to effects from inhalation associated with peak plasma
539
concentrations (Cmax), since quicker absorption increases peak systemic concentrations despite
540
comparable AUC values.
M AN U
541
SC
536
3. Rates of absorption can be limited by saturation due to limited capacity of enzymes and transporters involved in the biotransformation or absorption processes. As a result, the PoD may
543
be based on a non-linear dose-response relationship (ECHA, 2010).
544
TE D
542
Initially, an ADE should only be calculated for the route with the best available data. This initial ADE
546
should be the starting point for calculating ADEs for alternative routes of exposure, if such ADEs are
547
needed. As route-specific bioavailability may also differ between species, ADE calculations for these
548
alternative routes of exposure should be made based on comparative human exposure data (e.g., results
549
from an absolute bioavailability study), wherever possible. There have been a number of compilations for
550
drugs of both animal and human oral bioavailabilities which may prove useful (Akabane et al., 2010; Cao
551
et al., 2006; Chiou and Buehler, 2002; Musther et al., 2014; Sietsema, 1989). If experimental human data
552
are missing, route-specific differences in bioavailability may be estimated using read across from
553
chemicals with closely related structure, physiochemical properties, and toxicity profiles.
AC C
EP
545
554 555
Several investigations and reviews have provided background on the use of BCFs and guidance on when
556
and how they should be applied for OEL development (Naumann et al., 2009; Pfister et al., 2014; Sargent
22
ACCEPTED MANUSCRIPT
et al., 2013). These same approaches are relevant to setting ADEs. In many cases, the absolute oral
558
bioavailability of a compound may not be known, as it may not have been administered to humans by the
559
intravenous route. However, it is always preferable to use whatever reliable data are available for a
560
compound, rather than apply a default correction factor.
RI PT
557
561
Interspecies differences in oral bioavailability are most commonly a result of differences in first pass
563
metabolism in the gut and liver (Clark and Smith, 1984; Musther et al., 2014). Although it is well
564
established that the metabolic enzymes responsible for the biotransformation of drugs can be very
565
different between species, it does not necessarily mean that oral bioavailability will always be very
566
different. The percentage of drugs for which the human and animal oral bioavailabilities differed by more
567
than 10-fold for the same drug were lowest for dogs at 2% and rats/rodents at 2%/8%, while monkeys
568
demonstrated the greatest differences at 11%, 38%, and 40%, as measured by three different groups
569
(Akabane et al., 2010; Chiou and Buehler, 2002; Sietsema, 1989). The large difference between human
570
and primate bioavailability may, however, reflect the much more limited datasets (N = 35, 13, and 10,
571
respectively) available for primates compared with the datasets available for other species. A recent
572
reanalysis using the largest bioavailability dataset to date, inclusive of most of the individual datasets
573
indicated above, confirmed the poor overall correlation between human and animal oral bioavailability
574
(Musther et al., 2014), strongly supporting application of BCFs when data are available. In this way, the
575
BCF adjustment reduces uncertainty by increasing or decreasing the PoD as dictated by the available
576
data. It is noteworthy that the predictive value of animal bioavailability data is improved for substances
577
that have good membrane permeability, are not drug transporter substrates, and undergo little or no in
578
vivo metabolism (Akabane et al., 2010).
M AN U
TE D
EP
AC C
579
SC
562
580
5. Time Weighted Averaging, Study Duration, and Bioaccumulation Adjustments
581
Frequently, the critical study used to identify the PoD and derive the ADE has either an infrequent dosing
582
schedule or is from a subchronic duration study. As the ADE is to be protective of daily lifetime exposure,
583
the PoD derived from a study with infrequent dosing (e.g., once weekly administration) should be
584
adjusted for everyday exposures. In the absence of data, Haber’s Rule (see below) has been used by
23
ACCEPTED MANUSCRIPT
convention. However, when data are available, the preferred approach is to use TK data to assess steady
586
state and bioaccumulation through derivation of a steady state “S” factor (Sargent and Kirk, 1988).
587
Likewise, when the critical study used is not from a chronic study, a subchronic to chronic adjustment
588
factor (AFS) is typically applied on the assumption that the adverse effects of chemical exposure will occur
589
at lower exposure dose levels or with greater severity as the exposure duration increases. This can be
590
due to both the accumulation of the compound and/or the damage that it produces. While the AFS cannot
591
always be replaced by application of an S factor, it may be appropriate to use a reduced default AFS
592
pertaining to study duration (subchronic to chronic) when an S factor is used. These factors are related
593
and not independent, as the S factor differs from the AFS in that it only addresses TK considerations.
SC
RI PT
585
M AN U
594 5.1
596
Haber’s rule (Fritz Haber, 1868 - 1964) states that the concentration (C) of a substance multiplied by the
597
length of time (t) it is administered produces a fixed level of effect for a given endpoint, and thus predicts
598
the severity and incidence of the adverse health effect. Haber’s rule is a standard convention used for
599
dose averaging chemicals to make duration adjustments to the PoD when the administration frequency
600
used in the critical study is different from the exposure scenario. Haber’s rule provides an approximation
601
relating exposure to effect; however, there are many exceptions to the rule, such as when the adverse
602
effect results from Cmax or when exposure is for less than the biological half-life. Thus, understanding of
603
the chemical’s mechanism of action is an important consideration when deciding to use Haber’s rule.
EP
604
Time Weighted Averaging for Dose Frequency - Haber’s Rule
TE D
595
For intermittent exposures or exposures with variable concentrations, the average concentration of
606
exposure over the time period is used, such that concentration x time is equivalent to the AUC (Gaylor,
607
2000). For interpolation between different short-term exposures, this method may prove to be more
608
appropriately protective than when extrapolating to longer times [where t (actual exposure) / t (study
609
exposure) is greater than one]. However, in the absence of data for different durations of exposure,
610
extrapolating to shorter exposure may be inappropriate (Gaylor, 2000). A very important exception to
611
Haber’s rule is when toxicity is associated with peak concentration levels (Cmax).
AC C
605
612
24
ACCEPTED MANUSCRIPT
5.1.2
Examples
614
The PoD of 20 mg/day is from an IV infusion study in which patients received an anti-neoplastic drug 6
615
hours a day for 3 days during the first week, followed by 3 weeks without treatment (total 4 weeks or 28
616
days for the entire regimen). Developing an ADE using dose averaging, Haber’s Law states to multiply
617
the PoD by 3/28 (3 days exposure in the study/28 day exposure scenario). For example: 20
618
13
`^
×0
b `^
c `^
6 = 2.1
13
`^
de
RI PT
613
Haber’s law offers an estimated duration adjustment in the absence of data; however, the preferred
620
approach is to use TK data to assess steady state bioaccumulation, and MOA data to assess concerns
621
associated with peak effects (as described in section 3.1). In this respect, the anti-neoplastic regimen
622
described above might provide an AUC of 1500 ng•hr/mL. A similar operation to that described above is
623
used to adjust for the dosing frequency and duration. For example: 1500
M AN U
SC
619
!* ∙ ℎV 3 % k' !* ∙ ℎV ×i m = 375 de #h 28 % k' #h
In both instances either the dose or AUC could serve as the PoD. Additional adjustments for route of
625
administration (i.e., BCF) and steady state accumulation (described below in section 5.3) would be
626
applied if required.
627
TE D
624
5.2
Study Duration Adjustment Factor (AFS)
629
According to the Risk-MaPP guidance, ADEs are calculated for a potential lifetime exposure (ISPE,
630
2010). Although this is a highly protective assumption in most circumstances, the guidelines nevertheless
631
state that if the PoD is from a study that is less than lifetime, an adjustment factor may need to be applied
632
(AFS) (Dourson et al., 1996; Sussman et al., 2016, this issue). This duration adjustment for the critical
633
study is based on the assumption that the adverse effects of chemical exposure will occur at lower dose
634
levels as the exposure duration increases. For example, one could use a short-term animal study as the
635
basis for a chronic exposure ADE in humans. Considering that some drugs are often prescribed to
636
patients with intermittent dosing, such as once weekly, it is plausible that a worker or patient could be
637
exposed on a daily basis to a drug that is intended for weekly, biweekly, or monthly human dosing. But
AC C
EP
628
25
ACCEPTED MANUSCRIPT
638
the guidance documents available for these conversions are not robust enough to cover all extrapolation
639
scenarios.
640 The ICH Q3C guidelines for solvent impurities in drug products sets specific default factors for use when
642
adjusting from a shorter term study to a lifetime exposure, based on study species and the fraction of
643
animal lifetime covered in the study (ICH, 2011). No guidance document specifically addresses what data
644
are necessary to replace duration adjustment defaults or of what magnitude they should be adjusted
645
when certain TK data are available. The Risk-MaPP baseline guide states that in the absence of TK data
646
for duration adjustments, a factor of 3 is sufficient to account for the possibility that a lower PoD would
647
potentially have resulted from a longer-term study (ISPE, 2010). Risk-MAPP also suggests that when
648
longer studies are used to set shorter term limits, an upward adjustment to the ADE using an adjustment
649
factor of less than 1 may be appropriate. It does not, however, give guidance on how to calculate a
650
fractional adjustment factor or what is most appropriate (ISPE, 2010).
M AN U
SC
RI PT
641
651
Naumann and Weideman (1995) noted that high quality toxicology studies support an adjustment factor
653
of 3 to account for the possibility that a lower NOAEL could have resulted if longer-duration studies were
654
conducted (Kadry et al., 1995; Lewis and Nessel, 1994; McNamara, 1976; Weil and McCollister, 1963;
655
Woutersen et al., 1984). A recent study that tried to minimize chemical- and study-specific variability
656
identified an appropriate adjustment factor of 3 to 4 for orally administered compounds in the 90
657
percentile, with similar values for inhalation studies (Batke et al., 2011). Since these evaluations are
658
based primarily on data derived from studies where steady state was obtained, effect progression largely
659
reflects TD and TK effects other than accumulation (i.e., redistribution, latency before damage, age-
660
related changes, etc.). Hence, if an adjustment factor of 3 or 4 covers effect progression, then, in the
661
context of a traditional default factor framework where a full factor is 10x, the remaining factor for TK
662
variability is approximately 3x. A 3x factor is appropriate for accumulation for chemicals where the half-life
663
is about 2-fold greater than their administration frequency; a longer half-life or shorter administration
664
frequency will result in greater than 3-fold accumulation.
th
AC C
EP
TE D
652
665
26
ACCEPTED MANUSCRIPT
5.3
Data-Based Steady State “S” Factor to Account for Bioaccumulation and Portions of AFS
667
A bioaccumulation adjustment is required when an ADE is developed on the basis of single-dose or short-
668
duration studies and applied to a more protracted exposure scenario, and can be based on available
669
chemical-specific TK data. The steady state internal concentration can be calculated for one compartment
670
models using elimination half-lives (or from the half-life of the beta-phase of the elimination curve in two
671
compartment models) and other TK data. Steady state systemic concentrations are considered to be
672
achieved in multiple-dose studies when the period over which they are conducted exceeds 3 to 5
673
elimination half-lives (~90% to ~97% of steady state, respectively) of a chemical substance, after which
674
little additional accumulation will occur with additional administrations (as long as the dosing remains
675
consistent) (Rowland and Tozer, 1980a).
M AN U
SC
RI PT
666
676
For the development of ADEs, it can be assumed that the substance is effectively cleared if repeated
678
exposure occurs every 1.44 half-lives or less (e.g., daily exposures or work shifts) (Rowland and Tozer,
679
1980b). If the PoD is taken from a short or single-dose study and the substance has a long t½, then
680
bioaccumulation can be expected. To reduce the dose at the PoD in order to account for accumulation
681
from repeated administration, a bioaccumulation factor or steady state factor is applied (Sargent et al.,
682
2013). The S factor is a ratio of the plasma levels of the substance after steady state is achieved in a
683
repeat-dose study to the plasma levels following a single dose (ISPE, 2010). Since the S factor is used in
684
the denominator, the ADE is reduced by the incremental increase in blood levels following repeated vs.
685
single exposure. For example, a 30-day study is used as the critical study and the chemical has a 15-day
686
half-life. With daily administration, steady state would not be achieved for 2-3 months and the
687
accumulation ratio would be predicted to be approximately 21-fold. This adjustment is not needed when
688
the PoD is from a study in which steady state has been reached.
EP
AC C
689
TE D
677
690
If the t½ of the drug is known from a single-dose or short-term study, but its steady state profile is not, the
691
extent of drug accumulation can be estimated (Rowland and Tozer, 1980b) using Equation 6:
692
7.ss E t½
oop#p$ q"&! V q"& r. = @3 \
693 27
Equation 6.
ACCEPTED MANUSCRIPT
It is acknowledged that the above equation may have some degree of error since it is most applicable for
695
substances described by single compartment kinetics, whereas most substances are described by a TK
696
model with two or more compartments (Gibaldi and Perrier, 1982). However, for most drugs the
697
distribution phase is much shorter than the elimination phase, and so the extent of drug accumulation will
698
be mainly dependent on the elimination t½ and the calculated accumulation ratio will be a good
699
approximation (Dettli, 1982).
RI PT
694
700
While the AFS is applied when the PoD is based on a less than lifetime critical studies, it is intended to
702
generally account for TD considerations (the potential for the effects to have inadequate repair time) and
703
drug accumulation. In the presence of actual data (e.g., clearance), it is more systematic to apply an S
704
factor to account for bioaccumulation at steady state. It is important to stress that caution must be taken
705
not to overlap the AFS with the application of the S factor; otherwise accumulation would be accounted for
706
twice. However, S factor differs from the AFS default in that it the S factor only addresses TK
707
considerations, not TD-related changes. Notwithstanding, these factors are related and not independent.
708
The discrimination between the accumulation of systemic exposure up to steady state and the
709
accumulation of the related biological effect is often not possible. In practice, both aspects are taken into
710
account by applying the factor for study duration. Carefully describing the considerations involved in
711
applying each factor allows for greater transparency and will reduce the likelihood of inadvertently
712
accounting for the same factor twice.
713
EP
TE D
M AN U
SC
701
6.
Summary and Conclusions
715
The focus of this paper was how TK/TD data can be used to derive ADEs and other safe limits. It is
716
expected that this paper be read in the context of the series of papers in general and closely along with
717
the AF paper (Sussman et al., 2016, this issue), as the larger context for these adjustments is presented
718
there. Overall, ADEs and OELs are not meant to be developed by non-professionals, thus this paper is
719
not meant to be standard operating procedure for the derivation of ADEs. Also, every ADE should be
720
accompanied with full documentation of the data used, the rationale for the decisions made, and the
721
calculations at each step.
AC C
714
28
ACCEPTED MANUSCRIPT
Health-based risk assessment methodologies are now the standard of practice for setting ADEs for APIs
723
and other impurities (EMA, 2014; ICH, 2013, 2011; ISPE, 2010). Guidance documents advocate the use
724
of chemical-specific health-based approaches for setting exposure limits beginning with a rigorous critical
725
evaluation and synthesis of all pharmacology and toxicology data. Default factors have traditionally been
726
used to account for interindividual variability adjustment (AFH) and uncertainties associated with
727
extrapolation from animals to humans (AFA). In the absence of toxicokinetic data, a default adjustment
728
factor of 10 is commonly used to account for each of these such that the total adjustment factor applied to
729
the ADE is 100-fold (10 AFA x 10 AFH). Allometric scaling approaches have also been used to extrapolate
730
from animals to humans; however, such an approach accounts only for differences in how body surface
731
area or body weight scale to metabolic rate. Recall that this latter factor assumes humans are more
732
sensitive than animals, which, while a protective assumption, may not be true for all compounds. Also,
733
some factors, such as blood volume and renal secretion, do not predictably scale with body size (Sharma
734
and McNeill, 2009).
M AN U
SC
RI PT
722
735
When sufficient data are available, data-driven CSAF approaches have been described and provide
737
detailed guidance on how to use TK and TD data in place of default AFA and AFH adjustment factors
738
(IPCS, 2005; Meek et al., 2002; US EPA, 2014). The CSAF guidance subdivides the AFH factor into two
739
subfactors that separately account for toxicokinetic and toxicodynamic variability. These factors account
740
for variability in the dose metric selected as most relevant for the critical effect, and guidance on selection
741
of the appropriate dose metric is also available (IPCS, 2005; US EPA, 2014).
EP
AC C
742
TE D
736
743
Derived ADE limits can also consider adjustmenting for route-specific differences in bioavailability.
744
Although it is preferable to use studies most relevant to the expected route of exposure, such studies are
745
not always available. In these situations, it is useful to apply a BCF to extrapolate the PoD from the route
746
used in the critical study to the route of the anticipated exposure. Once a chemical is systemically
747
available, the kinetic disposition is the same regardless of route of exposure, although there are many
748
physiological factors that impact systemic availability.
749
29
ACCEPTED MANUSCRIPT
The risk assessment process also takes into consideration the duration of the critical study in which the
751
critical effect is identified, and the dosing schedule and frequency of dose administration in the critical
752
study. Haber’s Rule has been traditionally used to average infrequent dosing schedules over time.
753
However, when data are available, the steady state S factor can be used instead to account for potential
754
accumulation from a different dosing scenario. In instances where short-term and subchronic studies are
755
used as the basis of the ADE, an adjustment factor (AFS) is applied to account for the short length of the
756
study and the possibility that a longer study might yield a lower PoD or an effect of greater severity at the
757
same dose. When the PoD is from a short-term study (i.e., acute, subchronic) or when the substance has
758
a half-life that is significantly longer than the dosing frequency, the preferred approach is to calculate the
759
accumulation ratio (S factor) to adjust the PoD for predicted bioaccumulation. In doing so, a reduced
760
adjustment factor for subchronic to chronic extrapolation (AFS) should be applied. The application of a
761
separate factor to account for steady state accumulation with longer durations of exposure has several
762
advantages:
M AN U
SC
RI PT
750
1. There is no established guidance on how to subdivide the adjustment factor for study length on
764
the basis of TD and TK components. In the absence of an accepted authoritative guideline,
765
current practices are ad hoc and inconsistent.
TE D
763
2. Accounting for accumulation with a separate bioaccumulation ratio increases transparency and
767
can easily permit the use of longer studies to set shorter term limits, and the upward adjustment
768
of the ADE in the absence of accumulation for drugs that are cleared rapidly (Sargent et al.,
769
2013).
771 772 773 774 775 776
3. Reserving the adjustment factor for subchronic to chronic extrapolation for toxicodynamic effects
AC C
770
EP
766
permits greater flexibility to consider reductions in the NOAEL/LOAEL boundary and increasing effect severity with increasing exposure duration. The average differences between subchronic and chronic values are in the range of 3-fold, whereas a small percentage of chemicals have ratios exceeding 10-fold (Naumann and Weideman, 1995).
4. Application of a separate bioaccumulation ratio helps to minimize overlap among adjustment factors and supports independence among adjustment factors.
777
30
ACCEPTED MANUSCRIPT
778
All of these factors are tied together into a single formula used for derivation of ADE limits, as shown in
779
Equation 7: x@.[y.
780
vw =
781
where the PoD is in mg/kg; BW is in kg; AFC is the composite adjustment factor including all CSAF
782
adjustments and other adjustments); BCF is the unitless bioavailability correction factor; S is the steady-
783
state adjustment ratio; and MF is any additional required modifying factor(s) (Naumann et al., 2009). The
784
use of TK and TD data in the derivation of the ADE is the most data-based and robust approach to the
785
protection of human health.
SC
RI PT
Equation 7.
786 7. Acknowledgements
788
The statements and conclusions in this paper reflect the opinions of the authors and do not necessarily
789
represent official policies of the organizations as listed on the title page. The authors would like to
790
acknowledge Patricia Weideman, Andrew Maier, and Alison Pecquet for organizing and facilitating the
791
workshop that served as the basis for developing this manuscript. The authors would also like to thank all
792
of the participants of the workshop for their contributions at the workshop and subsequent reviews of this
793
manuscript, including: Joel Bercu, Courtney Callis, David Dolan, Andreas Flueckiger, Janet Gould, Eileen
794
Hayes, Robert Jolly, Ester Lovsin Barle, Wendy Luo, Eric Morinello, Lance Molnar, Michael Olson,
795
Christopher Seaman, Claudia Sehner, Bryan Shipp, Brad Stanard, Robert Sussman, and Andrew Walsh.
796
The manuscript was developed in part with funding from Genentech Inc. for organizational and editorial
797
staff activities.
798
8. References
799
ACGIH, 2016. TLV/BEI Development Process. American Conference of Governmental Industrial
801
TE D
EP
AC C
800
M AN U
787
Hygienists. Available at: http://www.acgih.org/tlv-bei-guidelines/policies-procedurespresentations/tlv-bei-development-process.
802
Akabane, T., Tabata, K., Kadono, K., Sakuda, S., Terashita, S., Teramura, T., 2010. A Comparison of
803
Pharmacokinetics between Humans and Monkeys. Drug Metab. Dispos. 38, 308–316.
804
doi:10.1124/dmd.109.028829
31
ACCEPTED MANUSCRIPT
805
Batke, M., Escher, S., Hoffmann-Doerr, S., Melber, C., Messinger, H., Mangelsdorf, I., 2011. Evaluation of
806
time extrapolation factors based on the database RepDose. Toxicol. Lett. 205, 122–129.
807
doi:10.1016/j.toxlet.2011.05.1030 Bercu, J.P., Morinello, E., Sehner, C., Shipp, B., Weideman, P., 2016. Point of departure (PoD) selection
RI PT
808 809
for the derivation of acceptable daily exposure (ADE) values for active pharmaceutical ingredients
810
(APIs). Manuscr. Accept. Regul. Toxicol. Pharmacol.
811
Borlak, J., Blickwede, M., Hansen, T., Koch, W., Walles, M., Levsen, K., 2005. Metabolism of verapamil in cultures of rat alveolar epithelial cells and pharmacokinetics after administration by intravenous
813
and inhalation routes. Drug Metab. Dispos. Biol. Fate Chem. 33, 1108–1114.
814
doi:10.1124/dmd.105.003723
M AN U
815
SC
812
Cao, X., Gibbs, S.T., Fang, L., Miller, H.A., Landowski, C.P., Shin, H.-C., Lennernas, H., Zhong, Y.,
816
Amidon, G.L., Yu, L.X., Sun, D., 2006. Why is it challenging to predict intestinal drug absorption
817
and oral bioavailability in human using rat model. Pharm. Res. 23, 1675–1686.
818
doi:10.1007/s11095-006-9041-2
821 822 823
monkey and human. Pharm. Res. 19, 868–874.
TE D
820
Chiou, W.L., Buehler, P.W., 2002. Comparison of oral absorption and bioavailablity of drugs between
Clark, B., Smith, D.A., 1984. Pharmacokinetics and toxicity testing. Crit. Rev. Toxicol. 12, 343–385. doi:10.3109/10408448409044214
Clewell, H.J., Andersen, M.E., Barton, H.A., 2002. A consistent approach for the application of
EP
819
pharmacokinetic modeling in cancer and noncancer risk assessment. Environ. Health Perspect.
825
110, 85–93.
AC C
824
826
Crump, K.S., Chen, C., Chiu, W.A., Louis, T.A., Portier, C.J., Subramaniam, R.P., White, P.D., 2010.
827
What role for biologically based dose-response models in estimating low-dose risk? Environ.
828
Health Perspect. 118, 585–588. doi:10.1289/ehp.0901249
829
Dettli, L., 1982. Pharmacokinetic parameters of interest to the clinician, in: Bolzer, G., van Rossum, J.M.
830
(Eds.), Pharmacokinetics During Drug Development: Data Analysis and Evaluation Techniques.
831
Fischer, Stuttgart, New York, pp. 18– 26.
32
ACCEPTED MANUSCRIPT
832
Dourson, M.L., Felter, S.P., Robinson, D., 1996. Evolution of science-based uncertainty factors in
833
noncancer risk assessment. Regul. Toxicol. Pharmacol. RTP 24, 108–120.
834
doi:10.1006/rtph.1996.0116
836 837
Dourson, M.L., Stara, J.F., 1983. Regulatory history and experimental support of uncertainty (safety) factors. Regul. Toxicol. Pharmacol. RTP 3, 224–238.
RI PT
835
ECHA, 2010. Chapter R.8: Characterisation of dose [concentration]-response for human health., in: Guidance on Information Requirements and Chemical Safety Assessment. Version 2. European
839
Chemicals Agency (ECHA), Helsinki, Finland. Available at: https://echa.europa.eu, Helsinki,
840
Finland.
EFSA, 2012. Guidance on selected default values to be used by the EFSA Scientific Committee,
M AN U
841
SC
838
842
Scientific Panels and Units in the absence of actual measured data. Eur. Food Saf. J. 10, 2579–
843
2611. doi:10.2903/j.efsa.2012.2579
844
EMA, 2014. Guideline on Setting Health Based Exposure Limits for Use in Risk Identification in the Manufacture of Different Medicinal Products in Shared Facilities. EMA/CHMP/ CVMP/
846
SWP/169430/2012. European Medicine Agency (EMA), London. Available at:
847
http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2014/11/WC50017
848
7735.pdf.
849
TE D
845
EMA, 2012. Guideline on the Approach to Establish a Pharmacological ADI. EMA/CVMP/SWP/355689/2006. European Medicine Agency (EMA), London. Available at:
851
http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2012/01/WC50012
852
0832.pdf.
854 855 856 857 858 859
AC C
853
EP
850
Fan, J., Teng, X., Liu, L., Mattaini, K.R., Looper, R.E., Vander Heiden, M.G., Rabinowitz, J.D., 2015. Human phosphoglycerate dehydrogenase produces the oncometabolite D-2-hydroxyglutarate. ACS Chem. Biol. 10, 510–516. doi:10.1021/cb500683c
Gaylor, D.W., 2000. The use of Haber’s law in standard setting and risk assessment. Toxicology 149, 17– 19. Gibaldi, M., Perrier, D. (Eds.), 1982. Pharmacokinetics, Second Edition. CRC Press, New York, pp. 121– 125.
33
ACCEPTED MANUSCRIPT
860
Gould, J.C., Kasichayanula, S., Shepperly, D.C., Boulton, D.W., 2013. Use of low-dose clinical pharmacodynamic and pharmacokinetic data to establish an occupational exposure limit for
862
dapagliflozin, a potent inhibitor of the renal sodium glucose co-transporter 2. Regul. Toxicol.
863
Pharmacol. RTP 67, 89–97. doi:10.1016/j.yrtph.2013.07.002
864
RI PT
861
ICH, 2013. Draft Consensus Guideline. Guideline for Elemental Impurities Q3D. International Conference on Harmonisation (ICH), Geneva. Available at:
866
http://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Quality/Q3D/Q3D_Step2
867
b.pdf.
868
SC
865
ICH, 2011. Impurities: Guideline for Residual Solvents Q3C(R5) Step 4 Version. International Conference on Harmonisation (ICH), Geneva. Available at:
870
http://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Quality/Q3C/Step4/Q3C_
871
R5_Step4.pdf.
872
M AN U
869
IPCS, 2005. Chemical-specific adjustment factors for interspecies differences and human variability: Guidance document for use of data in dose/concentration-response assessment. International
874
Programme on Chemical Safety (IPCS). World Health Organization (WHO), Geneva. Available at:
875
http://www.inchem.org/documents/harmproj/harmproj/harmproj2.pdf.
876
TE D
873
ISPE, 2010. Baseline Pharmaceutical Engineering Guide, Volume 7, Risk-Based Manufacture of Pharmaceutical Products: A Guide to Managing Risks Associated with Cross Contamination.
878
International Society for Pharmaceutical Engineering (ISPE), Tampa, FL. Available at:
879
http://www.ispe.org/baseline-guides/risk-mapp.
881 882 883 884 885
Kadry, A.M., Skowronski, G.A., Abdel-Rahman, M.S., 1995. Evaluation of the use of uncertainty factors in
AC C
880
EP
877
deriving RfDs for some chlorinated compounds. J. Toxicol. Environ. Health 45, 83–95. doi:10.1080/15287399509531982
Kirman, C.R., Sweeney, L.M., Meek, M.E., Gargas, M.L., 2003. Assessing the dose-dependency of allometric scaling performance using physiologically based pharmacokinetic modeling. Regul Toxicol. Pharmacol. 38(3), 345-67.
34
ACCEPTED MANUSCRIPT
886
Kuempel, E.D., Sweeney, L.M., Morris, J.B., Jarabek, A.M., 2015. Advances in Inhalation Dosimetry
887
Models and Methods for Occupational Risk Assessment and Exposure Limit Derivation. J. Occup.
888
Environ. Hyg. 12 Suppl 1, S18–40. doi:10.1080/15459624.2015.1060328 Labiris, N.R., Dolovich, M.B., 2003. Pulmonary drug delivery. Part II: the role of inhalant delivery devices
890
and drug formulations in therapeutic effectiveness of aerosolized medications. Br. J. Clin.
891
Pharmacol. 56, 600–612.
RI PT
889
Lehman, A.J., Fitzhugh, 0. G., 1954. 100-Fold margin of safety. Assoc. Food Drug US Q. Bull. 18, 33–35.
893
Lewis, S.C., Nessel, C.S., 1994. Extrapolating subchronic test results to estimate chronic no-observed-
895
adverse-effect levels: Factors of 10 are larger than necessary. Toxicologist 14, 401. Lipinski, C.A., Lombardo, F., Dominy, B.W., Feeney, P.J., 2001. Experimental and computational
M AN U
894
SC
892
896
approaches to estimate solubility and permeability in drug discovery and development settings.
897
Adv. Drug Deliv. Rev. 46, 3–26.
898
McNamara, B.P., 1976. Estimates of the toxicity of hydrocyanic acid vapors in man. Edgewood Arsenal Technical Report, EB-TR-76023. Edgewood Arsenal, U.S. Department of the Army, Aberdeen
900
Proving Ground, MD. Available at: http://www.dtic.mil/cgi-
901
bin/GetTRDoc?Location=U2&doc=GetTRDoc.pdf&AD=ADA028501.
TE D
899
Meek, M.E., Renwick, A., Ohanian, E., Dourson, M., Lake, B., Naumann, B.D., Vu, V., 2002. Guidelines
903
for application of chemical-specific adjustment factors in dose/concentration–response
904
assessment. Toxicology 181, 115–120.
906 907 908
Muller, P.Y., Milton, M., Lloyd, P., Sims, J., Brennan, F.R., 2009. The minimum anticipated biological effect level (MABEL) for selection of first human dose in clinical trials with monoclonal antibodies.
AC C
905
EP
902
Curr. Opin. Biotechnol., Chemical biotechnology ● Pharmaceutical biotechnology 20, 722–729. doi:10.1016/j.copbio.2009.10.013
909
Musther, H., Olivares-Morales, A., Hatley, O.J.D., Liu, B., Rostami Hodjegan, A., 2014. Animal versus
910
human oral drug bioavailability: do they correlate? Eur. J. Pharm. Sci. Off. J. Eur. Fed. Pharm.
911
Sci. 57, 280–291. doi:10.1016/j.ejps.2013.08.018
35
ACCEPTED MANUSCRIPT
912
Naumann, B.D., Dolan, D.G., Sargent, E.V., 2004. Rationale for the Chemical-Specific Adjustment
913
Factors Used to Derive an Occupational Exposure Limit for Timolol Maleate. Hum. Ecol. Risk
914
Assess. Int. J. 10, 99–111. doi:10.1080/10807030490280990 Naumann, B.D., Silverman, K.C., Dixit, R., Faria, E.C., Sargent, E.V., 2001. Case Studies of Categorical
RI PT
915 916
Data-Derived Adjustment Factors. Hum. Ecol. Risk Assess. Int. J. 7, 61–105.
917
doi:10.1080/20018091094213
918
Naumann, B.D., Weideman, P.A., 1995. Scientific basis for uncertainty factors used to establish
occupational exposure limits for pharmaceutical active ingredients. Hum. Ecol. Risk Assess. Int.
920
J. 1, 590–613. doi:10.1080/10807039509380049
Naumann, B.D., Weideman, P.A., Dixit, R., Grossman, S.J., Shen, C.F., Sargent, E.V., 1997. Use of
M AN U
921
SC
919
922
toxicokinetic and toxicodynamic data to reduce uncertainties when setting occupational exposure
923
limits for pharmaceuticals. Hum. Ecol. Risk Assess. Int. J. 3, 555–565.
924
doi:10.1080/10807039709383711
925
Naumann, B.D., Weideman, P.A., Sarangapani, R., Hu, S.-C., Dixit, R., Sargent, E.V., 2009. Investigations of the use of bioavailability data to adjust occupational exposure limits for active
927
pharmaceutical ingredients. Toxicol. Sci. Off. J. Soc. Toxicol. 112, 196–210.
928
doi:10.1093/toxsci/kfp195
929
TE D
926
Pfister, T., Dolan, D., Bercu, J., Gould, J., Wang, B., Bechter, R., Barle, E.L., Pfannkuch, F., Flueckiger, A., 2014. Bioavailability of therapeutic proteins by inhalation--worker safety aspects. Ann. Occup.
931
Hyg. 58, 899–911. doi:10.1093/annhyg/meu038
933 934 935 936 937 938
Rennen, M.A.J., Hakkert, B.C., Stevenson, H., Bos, P.M.J., 2001. Data-base derived values for the
AC C
932
EP
930
interspecies extrapolation : a quantitative analysis of historical toxicity data. Comments Toxicol 7, 423–436.
Renwick, A.G., 1993. Data‐derived safety factors for the evaluation of food additives and environmental contaminants. Food Addit. Contam. 10, 275–305. doi:10.1080/02652039309374152
Renwick, A.G., 1991. Safety factors and establishment of acceptable daily intakes. Food Addit. Contam. 8, 135–149. doi:10.1080/02652039109373964
36
ACCEPTED MANUSCRIPT
939
Rhomberg, R.L., Lewandowski, T.A., 2004. Methods for identifying a default cross-species scaling factor.
940
Risk Assessment Forum, U.S. Environmental Protection Agency (EPA), Washington, D.C.
941
Available at: http://www.epa.gov/raf/publications/pdfs/RHOMBERGSPAPER.PDF. RIVM, 2002. Multiple Path Particle Dosimetry Model (MPPD v 1.0): A Model for Human and Rat Airway
943
Particle Dosimetry. RIVA Report 650010030. National Institute for Public Health and the
944
Environment (RIVM), Bilthoven, The Netherlands.
947 948 949
and Febiger, Philadelphia, PA, p. 100.
SC
946
Rowland, M., Tozer, T.N., 1980a. Clinical Pharmacokinetics: Concepts and Applications. 1st Edition. Lea
Rowland, M., Tozer, T.N., 1980b. Clinical Pharmacokinetics: Concepts and Applications. 1st Edition. Lea and Febiger, Philadelphia, PA, pp. 174–179.
M AN U
945
RI PT
942
Sargent, E.V., Faria, E., Pfister, T., Sussman, R.G., 2013. Guidance on the establishment of acceptable
950
daily exposure limits (ADE) to support risk-based manufacture of pharmaceutical products. Regul.
951
Toxicol. Pharmacol. 65, 242–250.
952
Sargent, E.V., Kirk, G.D., 1988. Establishing airborne exposure control limits in the pharmaceutical
953
industry. Am. Ind. Hyg. Assoc. J. 49, 309–313. doi:10.1080/15298668891379792 Sargent, E.V., Naumann, B.D., Dolan, D.G., Faria, E.C., Schulman, L., 2002. The Importance of Human
TE D
954 955
Data in the Establishment of Occupational Exposure Limits. Hum. Ecol. Risk Assess. Int. J. 8,
956
805–822. doi:10.1080/20028091057213
959 960 961 962 963
EP
958
Sharma, V., McNeill, J.H., 2009. To scale or not to scale: the principles of dose extrapolation. Br. J. Pharmacol. 157, 907–921. doi:10.1111/j.1476-5381.2009.00267.x Sietsema, W.K., 1989. The absolute oral bioavailability of selected drugs. Int. J. Clin. Pharmacol. 27,
AC C
957
179–211.
Silverman, K.C., Naumann, B.D., Holder, D.J., Dixit, R., Faria, E.C., Sargent, E.V., Gallo, M.A., 1999. Establishing Data-Derived Adjustment Factors from Published Pharmaceutical Clinical Trial Data. Hum. Ecol. Risk Assess. Int. J. 5, 1059–1089. doi:10.1080/10807039991289347
964
Snodin, D.J., McCrossen, S.D., 2012. Guidelines and pharmacopoeial standards for pharmaceutical
965
impurities: overview and critical assessment. Regul. Toxicol. Pharmacol. RTP 63, 298–312.
966
doi:10.1016/j.yrtph.2012.03.016
37
ACCEPTED MANUSCRIPT
967
Sussman, R.G., Naumann, B.D., Pfister, T., E.P., Sehner, C., Weideman, P., 2016. A harmonization effort
968
for exposure methodology – considerations for application of adjustment factors. Manuscr.
969
Accept. Regul. Toxicol. Pharmacol. Tian, S., Li, Y., Wang, J., Zhang, J., Hou, T., 2011. ADME evaluation in drug discovery. 9. Prediction of
RI PT
970 971
oral bioavailability in humans based on molecular properties and structural fingerprints. Mol.
972
Pharm. 8, 841–851. doi:10.1021/mp100444g
US EPA, 2014. Guidance for Applying Quantitative Data to Develop Data-Derived Extrapolation Factors
974
for Interspecies and Intraspecies Extrapolation. Office of the Science Advisor, Risk Assessment
975
Forum, U.S. Environmental Protection Agency (EPA), Washington, DC. Available at:
976
http://www2.epa.gov/sites/production/files/2015-01/documents/ddef-final.pdf.
M AN U
SC
973
977
US EPA, 2011. Recommended Use of Body Weight ¾ as the Default Method in Derivation of the Oral
978
Reference Dose, EPA/100/R11/0001. Office of the Science Advisor, Risk Assessment Forum,
979
U.S. Environmental Protection Agency (EPA), Washington, DC. Available at:
980
http://www.epa.gov/raf/publications/pdfs/recommended-use-of-bw34.pdf.
981
US FDA, 2008. Guidance for Industry: Genotoxic and Carcinogenic Impurities in Drug Substances and Products: Recommended Approaches. U.S. Food and Drug Administration (FDA), Silver Spring,
983
MD. Available at:
984
http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UC
985
M347725.pdf.
988 989 990 991 992 993 994
EP
987
US FDA, 2005. Guidance for Industry Estimating the Maximum Safe Starting Dose in Initial Clinical Trials for Therapeutics in Adult Healthy Volunteers. U.S. Food and Drug Administration (FDA), Center
AC C
986
TE D
982
for Drug Evaluation and Research, Silver Spring, MD. Available at: http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/ucm 078932.pdf.
Walsh, A., 2011a. Cleaning validation for the 21st century: Acceptance limits for active pharmaceutical ingredients (APIs): Part I. Pharm. Eng. 31, 74–83. Walsh, A., 2011b. Cleaning validation for the 21st century: Acceptance limits for active pharmaceutical ingredients (APIs): Part II. Pharm. Eng. 31, 44–49.
38
ACCEPTED MANUSCRIPT
997 998 999 1000 1001 1002
Acceptance limits for cleaning agents. Pharm. Eng. 12–24. Weil, C.S., McCollister, D.D., 1963. Relationship between short and long term feeding studies in designing an effective toxicity test. Agr. Food Chem 11, 486–491.
RI PT
996
Walsh, A., Ovais, M., Altmann, T., Sargent, E.V., 2013. Cleaning validation for the 21st century:
Welling, P.G., 1995. Differences between pharmacokinetics and toxicokinetics. Toxicol. Pathol. 23, 143– 147.
Woutersen, R.A., Til, H.P., Feron, V.J., 1984. Sub-acute versus sub-chronic oral toxicity study in rats: comparative study of 82 compounds. J. Appl. Toxicol. JAT 4, 277–280.
SC
995
1003 9. Figure Captions
1005
Figure 1. Illustration of the breakdown of the AFH and AFA 10-fold adjustment factors into their TK and TD
1006
components, as suggested by CSAF (IPCS, 2005) or DDEF guidelines (US EPA, 2014). Figure as
1007
adapted from IPCS (2005).
M AN U
1004
1008
Figure 2. Illustration of the time course of oral exposure as it relates to available dose-metrics for use in
1010
CSAF development. Concentration axis is presented on a logarithmic scale axis. Cmax – peak
1011
concentration; AUC – area under the curve; t½ - half-life; Tmax – time to peak concentration.
AC C
EP
TE D
1009
39
ACCEPTED MANUSCRIPT
1
Highlights
2
•
A number of TK parameters can be used to replace default adjustment factors.
3
•
TK data can also be used to adjust the point-of-departure (PoD) for intermittent dosing and to
Selection of the best TK parameter requires knowledge of the predictors of target tissue
TE D
M AN U
SC
concentrations and expert judgement.
EP
6
•
AC C
5
RI PT
account for steady state.
4