- Email: [email protected]
Evaluation of the Uncertainty Factor for Subchronic-to-Chronic Extrapolation: Statistical Analysis of Toxicity Data
REGULATORY TOXICOLOGY AND PHARMACOLOGY ARTICLE NO.
27, 108–111 (1998)
RT971196
Evaluation of the Uncertainty Factor for Subchronic-to-Chronic Extrapolation: Statistical Analysis of Toxicity Data M. N. Pieters, H. J. Kramer, and W. Slob National Institute of Public Health and the Environment (RIVM), P.O. Box 1, 3720 BA Bilthoven, The Netherlands Received March 31, 1997
For the extrapolation of a subchronic no-observedadverse-effect level (NOAEL) to a chronic NOAEL an uncertainty factor of 10 is routinely applied. We evaluated this uncertainty factor by statistically analyzing a database comprising 149 oral NOAELsubchronic/ NOAELchronic ratios. Since any database forms only a limited sample of all existent chemicals, we believe that estimation errors should be taken into account. Therefore, in addition to the 95th percentile (P95) of the ratio distribution, we calculated the 95% confidence interval (CI) of this percentile. The geometric mean (GM), geometric standard deviation (GSD), P95 , and CI were 1.7, 5.6, 29, and 20 –46, respectively. These data do not support a lowering of the uncertainty factor as has been suggested by others. Furthermore, we analyzed the LOAEL/NOAEL ratio for subacute, subchronic, and chronic toxicity studies. However, the size of the LOAEL/NOAEL ratio only depends on the spacing between doses of the studies reviewed. Therefore, though the results may be considered supportive for an uncertainty factor of 10, we believe that there is no justification for the use of such a factor. Instead, we recommend the use of dose – response modeling which would make LOAEL-to-NOAEL extrapolation redundant. q 1998 Academic Press
INTRODUCTION
In the absence of adequate human toxicity data, health-based acceptable exposure limits such as acceptable daily intake (ADI) or reference dose (RfD) are based on (an array of) toxicity studies in experimental animals. For regulatory purposes the no-observed-adverse-effect levels (NOAELs) of relevant toxicity parameters are used. In practice the lowest of the NOAELs assessed is often used as a starting point for establishing human exposure limits. In many cases the lowest NOAEL is derived (or expected to be derived) in a chronic, i.e., life-time, toxicity study. If chronic toxicity data are not available, regulatory agencies may rely on less-than-chronic toxicity data. In these cases an
108
0273-2300/98 $25.00 Copyright q 1998 by Academic Press All rights of reproduction in any form reserved.
AID
RTP 1196
/
6e18$$$261
uncertainty factor of 10 is applied as a default. Subsequently, the estimated chronic NOAEL is divided by uncertainty factors accounting for interspecies and intraspecies variation. Though an uncertainty factor of 10 is generally used for extrapolating a subchronic NOAEL to a chronic NOAEL, only limited data are available to support this factor. In 1963, Weil and McCollister compared data of 33 different compounds. The geometric mean (GM) of the ratio distribution was smaller than 2, while the 97.5 percentile was approximately 9. This study was cited by Dourson and Stara (1983) to support the uncertainty factor of 10. Lewis et al. (1990) referred to the same data of Weil and McCollister (1963) and noted that 90% of the ratios were 5 or less and 73% were 3 or less. McNamara (1976) investigated the subchronic/ chronic NOAEL ratio in 41 chemicals and combined his data with the data of Weil and McCollister (1963). He showed that in 95% of the cases the ratio was 5 or less while 85% of the ratios were smaller than around 3. Lewis and Nessel (1994) reported a geometric mean of 2 by using a database of 32 compounds. Based on these studies with a limited amount of compounds studied, ECETOC recommended a decrease of the default value of 10 to a provisional default factor of 2–3 (ECETOC, 1995). Recently, we constructed a database of approximately 350 compounds which had been evaluated by regulatory agencies. We reported of a statistical analysis on the ratios LD50/NOAELchronic and NOAELsubacute/ NOAELchronic (Kramer et al., 1996). In the present paper we used the same database for the evaluation of subchronic-to-chronic extrapolation. If a relevant toxicological effect is still present at the lowest dose tested, regulatory agencies may base their risk evaluation on this LOAEL. To support the LOAELto-NOAEL uncertainty factor, several studies on LOAEL/NOAEL ratios have been published [e.g., Weil and McCollister (1963), Kadry et al. (1995), Naumann and Weidermann (1995)]. However, since a LOAEL/ NOAEL ratio distribution only reflects the variability in study design, it is irrelevant to base an uncertainty
06-19-98 13:48:00
rtpa
AP: RTP
EVALUATION OF SUBCHRONIC-TO-CHRONIC UNCERTAINTY FACTOR
109
factor on these data. Though the latter is more and more recognized, the application of an uncertainty factor on a LOAEL is still current practice. For this reason and because a provisional factor of 2–3 for LOAEL-toNOAEL extrapolation has been recommended based on the analyses mentioned above (ECETOC, 1995), we also analyzed LOAEL/NOAEL ratios. PROCEDURE
Composition of the Database The database used was composed of toxicological data evaluated by international agencies. Toxicological evaluations carried out at RIVM’s Center for Compounds and Risks are stored in TOXbank. In this database only studies meeting international quality standards are included. Though the database mostly comprises primary sources, additional secondary sources may have been used for the risk evaluations. For our analyses, we selected the lowest NOAEL when multiple toxicity data (e.g., four 28-day toxicity studies) were available. The latter was carried out regardless species or strain. As a result, we ended up comparing, for example, a subchronic NOAEL estimated in mice with a chronic NOAEL estimated in rats. Though the latter may appear scientifically unjustified, one should be aware that in current regulatory practice information on species (and strains) is not considered when applying an uncertainty factor. A factor of 10 is applied on the lowest NOAEL assessed in whatever species (perhaps with the exception of monkeys, though the interspecies variation between monkeys and humans may be considerable and not necessarily smaller than, for example, between rats and humans). When we regard NOAELs estimated in several species and strains as a distribution of all possible NOAELs, regulators establishing exposure limits are thus interested in a low percentile of this distribution. To obtain a database as large as possible, we combined data of TOXbank with evaluations carried out by the International Program on Chemical Safety (IPCS, Environmental Health Criteria) and by the Agency of Toxic Substances and Disease Registration (ATSDR, Toxicological Profiles). Though some differences in risk evaluation may exist between these agencies, these differences would not interfere with our purposes. The final database contained 425 records comprising 332 different compounds for several test species. Depending on species and study design, the absolute exposure time varied: subacute (3–6 weeks), subchronic (10–26 weeks), and chronic (1–2 years). Pesticides formed the major part of the database (approximately 50%), followed by solvents (approximately 25%), metalcontaining compounds (approximately 5%), phthalates (approximately 3%), and other compounds (approximately 17%).
AID
RTP 1196
/
6e18$$$262
06-19-98 13:48:00
FIG. 1. Relationship between the percentile of the NOAELsubchronic/ NOAELchronic ratio distribution and the associated percentage of compounds. The solid circles denote the point estimates of various percentiles of the NOAELsubchronic/NOAELchronic ratio distribution (n Å 149 ratios). The 95% confidence interval of P95 (20–46) is indicated. Notice that the current factor of 10 corresponds with the point estimate of the 85th percentile.
Statistical Analysis The statistical analysis performed was the same as previously described (Kramer et al., 1996). In short, we calculated the NOAELsubchronic/NOAELchronic and the LOAEL/NOAEL ratios for the individual compounds and examined the ratio distribution. Ratios were determined for the complete database comprising various animal species, as well as for a selection from the database comprising rat/rat data only. Based on the linearity of the data in a log-probability plot, it was assumed that the ratios were log-normally distributed. We calculated the GM as representative of the median ratio estimate of the corresponding distribution and the geometric standard deviation (GSD) as a measure for the variation of the distribution. In addition to point estimates of percentiles, we calculated their 95% confidence limits (see Kramer et al., 1996) to have an indication of the precision of the estimated percentiles. RESULTS
Estimation of a Chronic NOAEL from Subchronic Data The GM of the NOAELsubchronic/NOAELchronic was 1.7, while the GSD was 5.6 (n Å 149 compounds). Figure 1 shows the relationship between the percentage chosen and the point estimate of the associated percentile. The point estimate of the 95th percentile of the
rtpa
AP: RTP
110
PIETERS, KRAMER, AND SLOB
TABLE 1 GM, GSD, Sample Size, 95th Percentile (P95), and 95% Confidence Intervals (CI) of Oral NOAELsubchronic/ NOAELchronic Ratios Ratio NOAELsubchronic NOAELchronic
GM 1.7
GSD
n
5.6
P95
149
29
95% CI 20–46
NOAELsubchronic/NOAELchronic ratio distribution is 29 (see Table 1). The 95% confidence interval associated with P95 ranges from 20 to 46. To gain insight in the predictive value of the NOAELsubchronic for the NOAELchronic , we performed linear regression analysis. This analysis revealed that the NOAELsubchronic explained 55.2% of the variance of the NOAELchronic . The predictive value of the subchronic NOAEL may thus be considered to be rather poor. It may be expected that the use of NOAELs of various animal species introduces a large variation in the results. We therefore performed the same analysis also on rat/rat ratios only (n Å 70). The GM, GSD, and the P95 were 1.5, 6.3, and 31, respectively. Surprisingly, the variance increased rather than decreased. Defining Pesticides as a Subgroup In an attempt to reduce the variance of the distribution NOAELsubchronic/NOAELchronic , we analyzed pesticides as a subgroup. Since a similar mode of toxic action may result in more homogeneous data, we further subdivided the pesticides in cholinesterase inhibitors and remaining pesticides. Table 2 shows that by defining subgroups the GSD of the ratio distribution decreases, indicating an increase in homogeneity. LOAEL-to-NOAEL Extrapolation Though we do not support the idea that observed LOAEL/NOAEL ratios have any supportive value for the uncertainty factor for LOAEL-to-NOAEL extrapolation, we did perform a statistical analysis on LOAEL/ NOAEL ratios derived from subacute, subchronic, and chronic studies. The results are presented in Table 3.
TABLE 3 GM, GSD, Sample Size, 95th Percentile (P95), and 95% Confidence Intervals (CI) of Oral LOAEL/NOAEL Ratios Ratio LOAEL/NOAEL
GM
GSD
n
P95
95% CI
Subacute Subchronic Chronic
3.5 4.3 4.5
1.8 2.2 1.7
95 226 175
9 16 11
8–11 14–19 10–12
DISCUSSION
In this study we analyzed NOAEL ratios to evaluate the validity of the subchronic-to-chronic uncertainty factor, as used in practical risk assessments. Since these risk assessments aim at establishing ‘‘safe’’ exposure limits, the lowest reported NOAEL of a relevant toxicological parameter is used as a starting point for extrapolation to humans. The latter is carried out irrespective of the species (mostly rat or mouse) or strain used. Analogous to this current practice we constructed a database of lowest NOAELs of 332 compounds. For the evaluation of the subchronic-tochronic uncertainty factor, 149 NOAEL ratios were available. We calculated the GM and GSD of the distribution, as well as the 95th percentile of the ratio distribution. Though we analyzed a number of ratios higher than ever reported before, it still forms only a limited sample of all compounds available. Therefore, we believe it is necessary to calculate not only percentiles of the ratio distributions, but also the associated confidence intervals. The confidence intervals give an indication of the precision of the estimated percentile. In this way estimation errors caused by database composition are taken into account. The GM of the NOAELsubchronic/NOAELchronic ratio distribution was similar to those previously reported by Weil and McCollister (1963) and McNamara (1976). Based on these data Dourson and Stara (1983) concluded that for most of the compounds (96%) an uncertainty factor of 10 would suffice. We reanalyzed the data of Weil and McCollister (1963), McNamara (1976), and Rulis and Hattan (1985) and compared these with our
TABLE 4 Comparison of Data on the NOAELsubchronic/ NOAELchronic Ratio with the Literature
TABLE 2 Influence of Defining Subgroups on the Variance (GSD) of the NOAELsubchronic/NOAELchronic Ratio NOAELsubchronic NOAELchronic
GM
GSD
n
P95
95% CI
All compounds Pesticides Cholinesterase inhibitor Remaining pesticides
1.7 1.7 1.8 1.7
5.6 4.4 3.5 4.6
149 105 16 89
29 19 14 21
20–46 13–32 7–56 14–36
AID
RTP 1196
/
6e18$$$262
06-19-98 13:48:00
Present analysis Weil and McCollister (1963) McNamara (1976) Rulis and Hattan (1985)
GM
GSD
n
P95
95% CI
1.7 2.3 1.4 1.9
5.6 2.4 2.2 3.0
149 29 73 20
29 10 5 12
20–46 6.4–18 4.0–7 6.5–34
Note. Figures are obtained by reanalyzing the data provided in manuscripts.
rtpa
AP: RTP
EVALUATION OF SUBCHRONIC-TO-CHRONIC UNCERTAINTY FACTOR
own (Table 4). It is clear that the variation in our data set is considerably larger (GSD is 5.6) and consequently the P95 and UCL are higher. These differences are probably caused by differences in database composition. Weil and McCollister (1963) analyzed a relatively small (n Å 29) database. McNamara (1976) added data to those of Weil and McCollister which were very homogeneous: 28 of the 41 added compounds had a short-term/long-term ratio of unity. Rulis and Hattan analyzed food additives only. Though the outcome of database analyses thus highly depends on the database construction, the analysis of existing toxicity data as such may form an important contribution to the discussion on the numerical values of uncertainty factors. Based on the study of Weil and McCollister (1963) and McNamara (1976), ECETOC (1995) recommended to decrease the default value of 10 to a default factor of 2–3. However, our data do not support this recommendation. On the contrary, if the 95th percentile is considered appropriate as a subchronic-to-chronic uncertainty factor, the current default value of 10 would even be too low: the 95% confidence interval of the 95th percentile ranged from 20 to 46 and thus did not include the factor of 10. Figure 1 shows that an uncertainty factor of 10 would correspond with approximately the estimated 85th percentile, while a factor of 2–3 would be close to the geometric mean. In this study we also analyzed LOAEL/NOAEL ratios (subacute, n Å 95; subchronic, n Å 226; and chronic, n Å 175). The GMs of the LOAEL/NOAEL ratios are more or less in coherence with the results reported by Dourson and Stara (1963). Our results indicated GMs of 4.3 and 4.5 for subchronic and chronic studies, respectively, while Dourson and Stara reported values ranging from 2 to 4. Dourson and Stara (1983) referred to data of Weil and McCollister (1963) to show that the uncertainty factor for LOAEL-toNOAEL extrapolation varied between 1 and 10. For 96% of the 52 LOAEL/NOAEL ratios (27 subchronic and 25 chronic comparisons), the ratio was 5 or less. Recently, Kadry et al. (1995) showed that for 23 LOAEL/NOAEL ratios of six chlorinated compounds a majority of the ratios (87%) were less than 5. Based on these data, ECETOC (1995) recommended a provisional factor of 2–3 for LOAEL-to-NOAEL extrapolation. A higher factor was considered justified in the case of severe effects at the LOAEL. Though Naumann and Weidemann (1995) preferred the use of a benchmark approach, they suggested a factor of 3 as a ‘‘best estimate’’ to adjust a LOAEL into a NOAEL. We believe that for LOAEL-to-NOAEL extrapolation observed ratios of LOAEL to NOAEL are irrelevant since they only give information on applied intervals between dose levels in the reviewed studies. Furthermore, if a particular study results in a LOAEL, there is no guarantee whatsoever that at one dose interval lower the effect would
AID
RTP 1196
/
6e18$$$262
06-19-98 13:48:00
111
be statistically nonsignificant. Instead, we recommend the use of dose–response modeling which would make the use of LOAEL-to-NOAEL extrapolation redundant. CONCLUSIONS
1. Data analysis may contribute to the discussion on the numerical values of uncertainty factors. However, one should take into account that database composition and sample size highly influence the statistical outcomes and conclusions. 2. Based on the statistical analysis of 149 NOAELsubchronic/NOAELchronic ratios we found a 95th percentile of 29 (point estimate) with a confidence interval of 20–46. These results do not support lowering the uncertainty factor of 10 used for subchronic-tochronic extrapolation. 3. Though the analysis of the LOAEL/NOAEL ratios resulted in a 95th percentile close to 10, we believe this does not justify the use of such a factor. Rather, we recommend the use of dose–response modeling which would make LOAEL-to-NOAEL extrapolation redundant. REFERENCES Dourson, M. L., and Stara, J. F. (1983). Regulatory history and experimental support of uncertainty (safety) factors. Regul. Toxicol. Pharmacol. 15, 224–238. ECETOC (1995). Assessment factors in human health risk assessment. ECETOC Technical Report No. 68. Kadry, A. M., Skowronski, G. A., and Abdel-Rahman, M. S. (1995). Evaluation of uncertainty factors in deriving RfDs for some chlorinated compounds. J. Toxicol. Environ. Health 44, 83–94. Kramer, H. J., Van den Ham, W. A., Slob, W., and Pieters, M. N. (1996). Conversion factors estimating indicative chronic no-observed-adverse-effect levels from short-term toxicity data. Regul. Toxicol. Pharmacol. 23, 249–255. Lewis, S. C., Lynch, J. R., and Nikiforov, A. I. (1990). A new approach to deriving community exposure guidelines from ‘no-observed-adverse-effect levels.’ Regul. Toxicol. Pharmacol. 11, 314–330. Lewis, S. C., and Nessel, C. S. (1994). Extrapolating subchronic test results to estimate chronic no-observed-adverse-effect levels: Factors of 10 are larger than necessary. Toxicologist 14, 401. McNamara, B. P. (1976). Concepts in health evaluation of commercial and industrial chemicals. In Advances in Modern Technology, Vol. 1, Part 1: New concepts in Safety Evaluation (M. A. Mehlman, R. E. Shapiro, and H. Blumenthal, Eds.), pp. 61–140. Hemisphere, Washington, DC. Naumann, B. D., and Weidemann, P. A. (1995). Scientific basis for uncertainty factors used to establish occupational exposure limits for pharmaceutical active ingredients. Hum. Ecol. Risk Assess. 1, 590–613. Rulis, A. M., and Hattan, D. G. (1985). FDA’s priority-based assessment of food additives. II. General toxicity parameters. Regul. Toxicol. Pharmacol. 5, 152–174. Weil, C. S., and McCollister, D. D. (1963). Relationship between short- and long-term feeding studies in designing an effective toxicity test. J. Agric. Food Chem. 11, 486–491. Weil, C. S., Woodside, M. D., Bernard, J. R., and Carpenter, C. P. (1969). Relationship between single per-oral, one-week, and ninety-day rat feeding studies. Toxicol. Appl. Pharmacol. 14, 426– 431.
rtpa
AP: RTP
27, 108–111 (1998)
RT971196
Evaluation of the Uncertainty Factor for Subchronic-to-Chronic Extrapolation: Statistical Analysis of Toxicity Data M. N. Pieters, H. J. Kramer, and W. Slob National Institute of Public Health and the Environment (RIVM), P.O. Box 1, 3720 BA Bilthoven, The Netherlands Received March 31, 1997
For the extrapolation of a subchronic no-observedadverse-effect level (NOAEL) to a chronic NOAEL an uncertainty factor of 10 is routinely applied. We evaluated this uncertainty factor by statistically analyzing a database comprising 149 oral NOAELsubchronic/ NOAELchronic ratios. Since any database forms only a limited sample of all existent chemicals, we believe that estimation errors should be taken into account. Therefore, in addition to the 95th percentile (P95) of the ratio distribution, we calculated the 95% confidence interval (CI) of this percentile. The geometric mean (GM), geometric standard deviation (GSD), P95 , and CI were 1.7, 5.6, 29, and 20 –46, respectively. These data do not support a lowering of the uncertainty factor as has been suggested by others. Furthermore, we analyzed the LOAEL/NOAEL ratio for subacute, subchronic, and chronic toxicity studies. However, the size of the LOAEL/NOAEL ratio only depends on the spacing between doses of the studies reviewed. Therefore, though the results may be considered supportive for an uncertainty factor of 10, we believe that there is no justification for the use of such a factor. Instead, we recommend the use of dose – response modeling which would make LOAEL-to-NOAEL extrapolation redundant. q 1998 Academic Press
INTRODUCTION
In the absence of adequate human toxicity data, health-based acceptable exposure limits such as acceptable daily intake (ADI) or reference dose (RfD) are based on (an array of) toxicity studies in experimental animals. For regulatory purposes the no-observed-adverse-effect levels (NOAELs) of relevant toxicity parameters are used. In practice the lowest of the NOAELs assessed is often used as a starting point for establishing human exposure limits. In many cases the lowest NOAEL is derived (or expected to be derived) in a chronic, i.e., life-time, toxicity study. If chronic toxicity data are not available, regulatory agencies may rely on less-than-chronic toxicity data. In these cases an
108
0273-2300/98 $25.00 Copyright q 1998 by Academic Press All rights of reproduction in any form reserved.
AID
RTP 1196
/
6e18$$$261
uncertainty factor of 10 is applied as a default. Subsequently, the estimated chronic NOAEL is divided by uncertainty factors accounting for interspecies and intraspecies variation. Though an uncertainty factor of 10 is generally used for extrapolating a subchronic NOAEL to a chronic NOAEL, only limited data are available to support this factor. In 1963, Weil and McCollister compared data of 33 different compounds. The geometric mean (GM) of the ratio distribution was smaller than 2, while the 97.5 percentile was approximately 9. This study was cited by Dourson and Stara (1983) to support the uncertainty factor of 10. Lewis et al. (1990) referred to the same data of Weil and McCollister (1963) and noted that 90% of the ratios were 5 or less and 73% were 3 or less. McNamara (1976) investigated the subchronic/ chronic NOAEL ratio in 41 chemicals and combined his data with the data of Weil and McCollister (1963). He showed that in 95% of the cases the ratio was 5 or less while 85% of the ratios were smaller than around 3. Lewis and Nessel (1994) reported a geometric mean of 2 by using a database of 32 compounds. Based on these studies with a limited amount of compounds studied, ECETOC recommended a decrease of the default value of 10 to a provisional default factor of 2–3 (ECETOC, 1995). Recently, we constructed a database of approximately 350 compounds which had been evaluated by regulatory agencies. We reported of a statistical analysis on the ratios LD50/NOAELchronic and NOAELsubacute/ NOAELchronic (Kramer et al., 1996). In the present paper we used the same database for the evaluation of subchronic-to-chronic extrapolation. If a relevant toxicological effect is still present at the lowest dose tested, regulatory agencies may base their risk evaluation on this LOAEL. To support the LOAELto-NOAEL uncertainty factor, several studies on LOAEL/NOAEL ratios have been published [e.g., Weil and McCollister (1963), Kadry et al. (1995), Naumann and Weidermann (1995)]. However, since a LOAEL/ NOAEL ratio distribution only reflects the variability in study design, it is irrelevant to base an uncertainty
06-19-98 13:48:00
rtpa
AP: RTP
EVALUATION OF SUBCHRONIC-TO-CHRONIC UNCERTAINTY FACTOR
109
factor on these data. Though the latter is more and more recognized, the application of an uncertainty factor on a LOAEL is still current practice. For this reason and because a provisional factor of 2–3 for LOAEL-toNOAEL extrapolation has been recommended based on the analyses mentioned above (ECETOC, 1995), we also analyzed LOAEL/NOAEL ratios. PROCEDURE
Composition of the Database The database used was composed of toxicological data evaluated by international agencies. Toxicological evaluations carried out at RIVM’s Center for Compounds and Risks are stored in TOXbank. In this database only studies meeting international quality standards are included. Though the database mostly comprises primary sources, additional secondary sources may have been used for the risk evaluations. For our analyses, we selected the lowest NOAEL when multiple toxicity data (e.g., four 28-day toxicity studies) were available. The latter was carried out regardless species or strain. As a result, we ended up comparing, for example, a subchronic NOAEL estimated in mice with a chronic NOAEL estimated in rats. Though the latter may appear scientifically unjustified, one should be aware that in current regulatory practice information on species (and strains) is not considered when applying an uncertainty factor. A factor of 10 is applied on the lowest NOAEL assessed in whatever species (perhaps with the exception of monkeys, though the interspecies variation between monkeys and humans may be considerable and not necessarily smaller than, for example, between rats and humans). When we regard NOAELs estimated in several species and strains as a distribution of all possible NOAELs, regulators establishing exposure limits are thus interested in a low percentile of this distribution. To obtain a database as large as possible, we combined data of TOXbank with evaluations carried out by the International Program on Chemical Safety (IPCS, Environmental Health Criteria) and by the Agency of Toxic Substances and Disease Registration (ATSDR, Toxicological Profiles). Though some differences in risk evaluation may exist between these agencies, these differences would not interfere with our purposes. The final database contained 425 records comprising 332 different compounds for several test species. Depending on species and study design, the absolute exposure time varied: subacute (3–6 weeks), subchronic (10–26 weeks), and chronic (1–2 years). Pesticides formed the major part of the database (approximately 50%), followed by solvents (approximately 25%), metalcontaining compounds (approximately 5%), phthalates (approximately 3%), and other compounds (approximately 17%).
AID
RTP 1196
/
6e18$$$262
06-19-98 13:48:00
FIG. 1. Relationship between the percentile of the NOAELsubchronic/ NOAELchronic ratio distribution and the associated percentage of compounds. The solid circles denote the point estimates of various percentiles of the NOAELsubchronic/NOAELchronic ratio distribution (n Å 149 ratios). The 95% confidence interval of P95 (20–46) is indicated. Notice that the current factor of 10 corresponds with the point estimate of the 85th percentile.
Statistical Analysis The statistical analysis performed was the same as previously described (Kramer et al., 1996). In short, we calculated the NOAELsubchronic/NOAELchronic and the LOAEL/NOAEL ratios for the individual compounds and examined the ratio distribution. Ratios were determined for the complete database comprising various animal species, as well as for a selection from the database comprising rat/rat data only. Based on the linearity of the data in a log-probability plot, it was assumed that the ratios were log-normally distributed. We calculated the GM as representative of the median ratio estimate of the corresponding distribution and the geometric standard deviation (GSD) as a measure for the variation of the distribution. In addition to point estimates of percentiles, we calculated their 95% confidence limits (see Kramer et al., 1996) to have an indication of the precision of the estimated percentiles. RESULTS
Estimation of a Chronic NOAEL from Subchronic Data The GM of the NOAELsubchronic/NOAELchronic was 1.7, while the GSD was 5.6 (n Å 149 compounds). Figure 1 shows the relationship between the percentage chosen and the point estimate of the associated percentile. The point estimate of the 95th percentile of the
rtpa
AP: RTP
110
PIETERS, KRAMER, AND SLOB
TABLE 1 GM, GSD, Sample Size, 95th Percentile (P95), and 95% Confidence Intervals (CI) of Oral NOAELsubchronic/ NOAELchronic Ratios Ratio NOAELsubchronic NOAELchronic
GM 1.7
GSD
n
5.6
P95
149
29
95% CI 20–46
NOAELsubchronic/NOAELchronic ratio distribution is 29 (see Table 1). The 95% confidence interval associated with P95 ranges from 20 to 46. To gain insight in the predictive value of the NOAELsubchronic for the NOAELchronic , we performed linear regression analysis. This analysis revealed that the NOAELsubchronic explained 55.2% of the variance of the NOAELchronic . The predictive value of the subchronic NOAEL may thus be considered to be rather poor. It may be expected that the use of NOAELs of various animal species introduces a large variation in the results. We therefore performed the same analysis also on rat/rat ratios only (n Å 70). The GM, GSD, and the P95 were 1.5, 6.3, and 31, respectively. Surprisingly, the variance increased rather than decreased. Defining Pesticides as a Subgroup In an attempt to reduce the variance of the distribution NOAELsubchronic/NOAELchronic , we analyzed pesticides as a subgroup. Since a similar mode of toxic action may result in more homogeneous data, we further subdivided the pesticides in cholinesterase inhibitors and remaining pesticides. Table 2 shows that by defining subgroups the GSD of the ratio distribution decreases, indicating an increase in homogeneity. LOAEL-to-NOAEL Extrapolation Though we do not support the idea that observed LOAEL/NOAEL ratios have any supportive value for the uncertainty factor for LOAEL-to-NOAEL extrapolation, we did perform a statistical analysis on LOAEL/ NOAEL ratios derived from subacute, subchronic, and chronic studies. The results are presented in Table 3.
TABLE 3 GM, GSD, Sample Size, 95th Percentile (P95), and 95% Confidence Intervals (CI) of Oral LOAEL/NOAEL Ratios Ratio LOAEL/NOAEL
GM
GSD
n
P95
95% CI
Subacute Subchronic Chronic
3.5 4.3 4.5
1.8 2.2 1.7
95 226 175
9 16 11
8–11 14–19 10–12
DISCUSSION
In this study we analyzed NOAEL ratios to evaluate the validity of the subchronic-to-chronic uncertainty factor, as used in practical risk assessments. Since these risk assessments aim at establishing ‘‘safe’’ exposure limits, the lowest reported NOAEL of a relevant toxicological parameter is used as a starting point for extrapolation to humans. The latter is carried out irrespective of the species (mostly rat or mouse) or strain used. Analogous to this current practice we constructed a database of lowest NOAELs of 332 compounds. For the evaluation of the subchronic-tochronic uncertainty factor, 149 NOAEL ratios were available. We calculated the GM and GSD of the distribution, as well as the 95th percentile of the ratio distribution. Though we analyzed a number of ratios higher than ever reported before, it still forms only a limited sample of all compounds available. Therefore, we believe it is necessary to calculate not only percentiles of the ratio distributions, but also the associated confidence intervals. The confidence intervals give an indication of the precision of the estimated percentile. In this way estimation errors caused by database composition are taken into account. The GM of the NOAELsubchronic/NOAELchronic ratio distribution was similar to those previously reported by Weil and McCollister (1963) and McNamara (1976). Based on these data Dourson and Stara (1983) concluded that for most of the compounds (96%) an uncertainty factor of 10 would suffice. We reanalyzed the data of Weil and McCollister (1963), McNamara (1976), and Rulis and Hattan (1985) and compared these with our
TABLE 4 Comparison of Data on the NOAELsubchronic/ NOAELchronic Ratio with the Literature
TABLE 2 Influence of Defining Subgroups on the Variance (GSD) of the NOAELsubchronic/NOAELchronic Ratio NOAELsubchronic NOAELchronic
GM
GSD
n
P95
95% CI
All compounds Pesticides Cholinesterase inhibitor Remaining pesticides
1.7 1.7 1.8 1.7
5.6 4.4 3.5 4.6
149 105 16 89
29 19 14 21
20–46 13–32 7–56 14–36
AID
RTP 1196
/
6e18$$$262
06-19-98 13:48:00
Present analysis Weil and McCollister (1963) McNamara (1976) Rulis and Hattan (1985)
GM
GSD
n
P95
95% CI
1.7 2.3 1.4 1.9
5.6 2.4 2.2 3.0
149 29 73 20
29 10 5 12
20–46 6.4–18 4.0–7 6.5–34
Note. Figures are obtained by reanalyzing the data provided in manuscripts.
rtpa
AP: RTP
EVALUATION OF SUBCHRONIC-TO-CHRONIC UNCERTAINTY FACTOR
own (Table 4). It is clear that the variation in our data set is considerably larger (GSD is 5.6) and consequently the P95 and UCL are higher. These differences are probably caused by differences in database composition. Weil and McCollister (1963) analyzed a relatively small (n Å 29) database. McNamara (1976) added data to those of Weil and McCollister which were very homogeneous: 28 of the 41 added compounds had a short-term/long-term ratio of unity. Rulis and Hattan analyzed food additives only. Though the outcome of database analyses thus highly depends on the database construction, the analysis of existing toxicity data as such may form an important contribution to the discussion on the numerical values of uncertainty factors. Based on the study of Weil and McCollister (1963) and McNamara (1976), ECETOC (1995) recommended to decrease the default value of 10 to a default factor of 2–3. However, our data do not support this recommendation. On the contrary, if the 95th percentile is considered appropriate as a subchronic-to-chronic uncertainty factor, the current default value of 10 would even be too low: the 95% confidence interval of the 95th percentile ranged from 20 to 46 and thus did not include the factor of 10. Figure 1 shows that an uncertainty factor of 10 would correspond with approximately the estimated 85th percentile, while a factor of 2–3 would be close to the geometric mean. In this study we also analyzed LOAEL/NOAEL ratios (subacute, n Å 95; subchronic, n Å 226; and chronic, n Å 175). The GMs of the LOAEL/NOAEL ratios are more or less in coherence with the results reported by Dourson and Stara (1963). Our results indicated GMs of 4.3 and 4.5 for subchronic and chronic studies, respectively, while Dourson and Stara reported values ranging from 2 to 4. Dourson and Stara (1983) referred to data of Weil and McCollister (1963) to show that the uncertainty factor for LOAEL-toNOAEL extrapolation varied between 1 and 10. For 96% of the 52 LOAEL/NOAEL ratios (27 subchronic and 25 chronic comparisons), the ratio was 5 or less. Recently, Kadry et al. (1995) showed that for 23 LOAEL/NOAEL ratios of six chlorinated compounds a majority of the ratios (87%) were less than 5. Based on these data, ECETOC (1995) recommended a provisional factor of 2–3 for LOAEL-to-NOAEL extrapolation. A higher factor was considered justified in the case of severe effects at the LOAEL. Though Naumann and Weidemann (1995) preferred the use of a benchmark approach, they suggested a factor of 3 as a ‘‘best estimate’’ to adjust a LOAEL into a NOAEL. We believe that for LOAEL-to-NOAEL extrapolation observed ratios of LOAEL to NOAEL are irrelevant since they only give information on applied intervals between dose levels in the reviewed studies. Furthermore, if a particular study results in a LOAEL, there is no guarantee whatsoever that at one dose interval lower the effect would
AID
RTP 1196
/
6e18$$$262
06-19-98 13:48:00
111
be statistically nonsignificant. Instead, we recommend the use of dose–response modeling which would make the use of LOAEL-to-NOAEL extrapolation redundant. CONCLUSIONS
1. Data analysis may contribute to the discussion on the numerical values of uncertainty factors. However, one should take into account that database composition and sample size highly influence the statistical outcomes and conclusions. 2. Based on the statistical analysis of 149 NOAELsubchronic/NOAELchronic ratios we found a 95th percentile of 29 (point estimate) with a confidence interval of 20–46. These results do not support lowering the uncertainty factor of 10 used for subchronic-tochronic extrapolation. 3. Though the analysis of the LOAEL/NOAEL ratios resulted in a 95th percentile close to 10, we believe this does not justify the use of such a factor. Rather, we recommend the use of dose–response modeling which would make LOAEL-to-NOAEL extrapolation redundant. REFERENCES Dourson, M. L., and Stara, J. F. (1983). Regulatory history and experimental support of uncertainty (safety) factors. Regul. Toxicol. Pharmacol. 15, 224–238. ECETOC (1995). Assessment factors in human health risk assessment. ECETOC Technical Report No. 68. Kadry, A. M., Skowronski, G. A., and Abdel-Rahman, M. S. (1995). Evaluation of uncertainty factors in deriving RfDs for some chlorinated compounds. J. Toxicol. Environ. Health 44, 83–94. Kramer, H. J., Van den Ham, W. A., Slob, W., and Pieters, M. N. (1996). Conversion factors estimating indicative chronic no-observed-adverse-effect levels from short-term toxicity data. Regul. Toxicol. Pharmacol. 23, 249–255. Lewis, S. C., Lynch, J. R., and Nikiforov, A. I. (1990). A new approach to deriving community exposure guidelines from ‘no-observed-adverse-effect levels.’ Regul. Toxicol. Pharmacol. 11, 314–330. Lewis, S. C., and Nessel, C. S. (1994). Extrapolating subchronic test results to estimate chronic no-observed-adverse-effect levels: Factors of 10 are larger than necessary. Toxicologist 14, 401. McNamara, B. P. (1976). Concepts in health evaluation of commercial and industrial chemicals. In Advances in Modern Technology, Vol. 1, Part 1: New concepts in Safety Evaluation (M. A. Mehlman, R. E. Shapiro, and H. Blumenthal, Eds.), pp. 61–140. Hemisphere, Washington, DC. Naumann, B. D., and Weidemann, P. A. (1995). Scientific basis for uncertainty factors used to establish occupational exposure limits for pharmaceutical active ingredients. Hum. Ecol. Risk Assess. 1, 590–613. Rulis, A. M., and Hattan, D. G. (1985). FDA’s priority-based assessment of food additives. II. General toxicity parameters. Regul. Toxicol. Pharmacol. 5, 152–174. Weil, C. S., and McCollister, D. D. (1963). Relationship between short- and long-term feeding studies in designing an effective toxicity test. J. Agric. Food Chem. 11, 486–491. Weil, C. S., Woodside, M. D., Bernard, J. R., and Carpenter, C. P. (1969). Relationship between single per-oral, one-week, and ninety-day rat feeding studies. Toxicol. Appl. Pharmacol. 14, 426– 431.
rtpa
AP: RTP