Toxicological Effects of Methylmercury (2000) / Chapter Skim
Currently Skimming:
Dose-Response Assessment
Dose-Response Assessment
Pages 271-303
The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 271... ...
After exploring the possibility that differences in power might explain some of this discrepancy, the committee presents and discusses several approaches that could be used to provide a benchmark dose based on data from the three studies. Among the methods considered are traditional approaches based on selecting a single outcome from a single study and an integrative analysis that combines information from different studies and outcomes.
|
From page 272... ...
uses minimum risk level (MRL)
|
From page 273... ...
over the background rate. Because the benchmark dose generally occurs within the range of experimental data, Crump and others have argued that its estimation is relatively robust to model choice.
|
From page 274... ...
From the figure, it is easy to see that the further left we move the curve corresponding to the exposed group, the higher the percentage of the exposed population that falls below the cutoff point. Gaylor and Slikker's suggestion simply involves finding the exposure level that leads to a specific increase in the proportion of the population falling below the cutoff point.
|
From page 275... ...
As described above, one option is to fix PO at a specified percentile of performance in the unexposed population (e.g., 0.05 or 0.10~. Assuming that the data follow a linear model (Yi = aO + apex i + ci, where X i represents the exposure level for the ith subject, aO and al are unknown regression coefficients, and £i is random error, assumed to be normally distributed with variance of)
|
From page 276... ...
Keiding, Copenhagen University, and P Grandjean, University of Southern Denmark, unpublished material, May 5, 2000~.
|
From page 277... ...
. it is straightforward to compute the power to detect specific values of the dose-response parameter al, but comparing such calculations across studies is complicated, because the computed power depends on the distributions of exposure levels and outcomes within each study (see Zar Statistical power refers to the probability of correctly rejecting the null hypothesis of no association when, in fact, a true association exists (see Zar 1998~.
|
From page 278... ...
Figure 7-2 graphs the power that each study would have had to detect various values of the standardized regression coefficients. The Farce Islands study, being the largest study, has the highest power, and the New Zealand study has the lowest.
|
From page 279... ...
levels and, in particular, on whether there are sufficient observations at high exposure levels to characterize the true shape of the dose response in that region. in fact, all three studies had fairly skewed exposure distributions, with a large number of subjects clustered at low exposure levels, along with a few subjects exposed at moderate to high levels.
|
From page 280... ...
1me Digit Span Boston Naming Test-no cues Boston Naming Test-cues 6.15 0.54 80 1.5 5.3 5.3 -1.1 0.12 40.3 -.27 -1.77 -1.91 -0.07 0.08 0.18 -0.06 -0.12 -0.13 CVLT-Short-term 3.1 - 0.57 - 0.06 CVLT-Long-term 3.8 - 0.55 - 0.05 New 3.31 TOLD-Language 16 -0.6 -0.12 Zealand c Development WISC-R:PIQ 16 -0.54 -0.11 WISC-R:FSIQ 16 - 0.55 - 0.11 McCarty 10 -0.53 -0.17 Perceptual Performance McCarthy Motor 0.15 Testb -0.007 -0.15 Exposures measured on the log-scale. Exposure SD and regression coefficients provided by study investigators (Grandjean et al.
|
From page 281... ...
However, the idea makes more sense if we think of a benchmark dose as simply a transformation of the estimated dose response (as we saw in Eq.
|
From page 282... ...
, language (Boston Naming Test) , and short-term memory (California Verbal Learning Test)
|
From page 283... ...
The committee requested these artalyses only for the outcomes measured when the children were 5 to 7 years old, because that age period was the only one available from the Faroe Islands study, and data from that age group have better predictive ability than data from earlier ages (E. Budtz-10rgensen, Copenhagen University, N
|
From page 284... ...
. 22 Farce IslandsC Finger Tapping 20 12 CPT Reaction Time 17 10 Bender Copying Errors 28 15 Boston Naming Test 15 10 CVLT: Delayed Recall 27 14 New Zealand TOLD Language Development 12 6 WISC-R:PIQ 12 6 WISC-R:FSIQ 13 6 McCarthy Perceptual Performance 8 McCarty Motor Test 4 13 6 aBMDs are calculated from the K-power model under the assumption that 5% of the responses will be abnormal in unexposed subjects (P0= 0.05)
|
From page 285... ...
Abbreviations: Wl, Woodcock-lohnson Tests of Achievement; CPT, Continuous Performance Test; CVET, California Verbal Learning Test; TOLD, Test of Language Development; WISC-R:P1Q, Wechsler Intelligence Scale for Children-Revised Performance {Q; WTSC-R:FSIQ, Wechsler Intelligence Scale for Children-Revised Full-Scale ~Q. appropriate to choose the lowest BMDL as the point of departure because that could easily result in choosing an unreliable "nd po~nt (i.e., one with great uncertainty)
|
From page 286... ...
The committee concluded that it would be inappropriate to pick the Seychelles study as the basis for risk assessment, given the available evidence for positive effects in the New Zealand and Faroe Islands studies, as well as in the Seychelles own pilot study. The committee felt that a good argument can be made for choosing the Farce Islands study as the basis for the risk assessment.
|
From page 287... ...
. Despite several strong arguments in favor of choosing a point of departure based on the Faroe Islands study, there was some concern that the estimated BMDs and corresponding BMDT~s could be confounded by PCB exposure, which was not adjusted statistically In the benchmark analysis.
|
From page 288... ...
Comparing the low PCB subset with the full cohort results, for example, the BMDs for Finger Tapping and CPT Reaction Time were 5-13 ppm lower for maternal hair and 19-99 ppb lower for cord blood, and the BMDs for Boston Naming and Delayed Recall were 5-6 ppm higher for maternal hair and 42-147 ppb higher for cord blood. Thus, the BMDs for the low-PCB-exposed subset for the two end points that were related to PCB exposure—Boston Naming and California Verbal Learning—did not differ from the BMDs for the total sample by any more than the BMDs for the two end points that did not relate to PCB exposure.
|
From page 289... ...
AN INTEGRATIVE ANALYSIS Although the committee felt comfortable with recommencing that risk assessment be based on the Boston Naming Test from the Faroe Islands study, it also explored a weight-of-evidence approach based on an integrative analysis that allows a quantitative synthesis of informa
|
From page 290... ...
. Although the technical details can be complicated, the hierarchical modeling basically serves to smooth away some of the random variation that complicates the interpretation of Me data presented in Table 7-2.4 The approach also provides a way to quantify study-to-study and outcome-to-outcome variability.
|
From page 291... ...
Abbreviations: W] , Woodcock-Johnson Tests of Achievement; CPT, Continuous Performance Test; CVLT, California Verbal Learning Test; TOLD, Test of Language Development; WISC-R:PIQ, Wechsler Intelligence Scale for Children-Revised Performance IQ; WISC-R:FSIQ, Wechsler Intelligence Scale for Children-Revised Full-Scale IQ.
|
From page 292... ...
One could also argue for using the estimate of central tendency derived from the hierarchical modeling approach. The comm~ttee's analysis based on the K-power model suggests a mean BMD of 21 ppm, which coincidentally corresponds precisely to the mean of the smoothed BMDs from the Faroe islands study.
|
From page 293... ...
The committee wondered, for example, if the influence of a few highly exposed individuals on the estimated dose response might explain the large model-to-model variations. The Faroe Islands study research group conducted sensitivity analyses repeating the regression models after omitting some of the highest observations.
|
From page 294... ...
The models can yield very different results for BMD calculations, however, because these calculations necessitate extrapolating to estimate the mean response at zero exposure level. Both the square-root and the log models take on a supralinear shape at low doses, that is, they postulate a steeper slope at low doses.
|
From page 295... ...
Keiding, Copenhagen University, and P Grandjean, University of Southern Denmark, unpublished material, May 5, 2000.
|
From page 296... ...
What becomes clear from Figure 7-5 is that variations in estimated BMDs are not explained by differences in how tvell the models fit the bulk of the data but rather what the models predict for the mean response for unexposed individuals.
|
From page 297... ...
would predict that the change in mean outcome associated with any doubling of MeHg concentration should be the same as the change observed In the mean outcome as concentration increases from 10 to 20 ppm. Thus, the log model would predict that the same magnitude change in outcome would be expected as the concentration goes from ~ to 2 ppm or from 4 to ~ ppm as that observed for the concentration going from 10 to 20 ppm—that is, the extrapolation down to zero exposure will predict a very steep slope at low doses.
|
From page 298... ...
. Specification of PO in the context of the MeHg studies, however, is somewhat problematic because of the absence of subjects with true zero exposure.
|
From page 299... ...
The Boston Naming Test scores are the most sensitive, reliable end point. · The potential for confounding by PCB exposure is of some concern for the Faroe Islands study.
|
From page 300... ...
· Given the available data, risk assessment should be based on the Boston Naming Test from the Farce Islands study using MeHg measured in cord blood. · Despite some potential for PCB exposures to bias BMD estimates based on the Farce Islands study, the committee recommends using estimates based on the full cohort and not adjusting for PCB exposure, mostly because the larger sample size is believed to result in more reliable estimates.
|
From page 301... ...
2000. Benchmark dose concentrations for methylmercury obtained from the Seychelles Child Development Study.
|
From page 302... ...
1993. Upper confidence limits on excess risk for quantitative responses.
|
From page 303... ...
DOSE-RESPONSE ASSESSMENT 303 Rice, I)
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.