efsa-opinion-chromium-food-drinking-water

efsa-opinion-chromium-food-drinking-water efsa-opinion-chromium-food-drinking-water

damienvanherp
from damienvanherp More from this publisher
16.04.2014 Views

Chromium in food and bottled water The graphs in Figure J15 show the fit of the logprobit, the multistage (with not acceptable fit) and the Weibull model (unrestricted with acceptable and restricted with unacceptable fit, p < 10 -7 ), respectively. The restricted Weibull model resulted, as expected, in a higher BMDL 10 value of 0.27 mg/kg b.w. per day and was not used for dose-response assessment. The restricted log-logistic model showed an acceptable fit (p = 0.05) with a BMDL 10 = 0.09 mg/kg b.w. per day and was also not used. No BMDL -10 was determined from the dose-response data of this endpoint in this study since the BMD/BMDL ratios ranges between a factor of 6 and 13 and the range of the BMDL values of the acceptable models was larger than one order of magnitude. Using a different modelling approach not following EFSA (2009) ADTSR reported a BMDL 10 = 0.09, which corresponds to the BMDL 10 value of the restricted log-logistic model, which was not used by the CONTAM Panel. Table J15: Diffuse epithelial hyperplasia in the duodenum in female mice. The benchmark dose (BMD 10 ) and the 95 % benchmark dose lower confidence limit (BMDL 10 ) values are given for a BMR of 10 % extra risk with characteristics of the model fit. Data Dose 0 0.38 1.4 3.1 8.7 Response 0/50 16/50 35/50 31/50 42/50 Models Restriction BMD Minus 10 BMDL 10 N of P- (mg/kg (mg/kg Loglikelihood Accepted parameters value b.w. per b.w. per day) day) Full model na 5 117.1 – – – – Null (reduced) model na 1 173.3 - - - - Probit na 2 145.0

Chromium in food and bottled water a) b) LogProbit Model, with BMR of 10% Extra Risk for the BMD and 0.95 Lower Confidence Limit for the BMDL Multistage Model, with BMR of 10% Extra Risk for the BMD and 0.95 Lower Confidence Limit for the BMDL 1 LogProbit BMD Lower Bound 1 Multistage BMD Lower Bound 0.8 0.8 Fraction Affected 0.6 0.4 Fraction Affected 0.6 0.4 0.2 0.2 0 BMDLBMD 0 1 2 3 4 5 6 7 8 9 dose 09:46 01/21 2014 0 BMDLBMD 0 1 2 3 4 5 6 7 8 9 dose 09:51 01/21 2014 c) d) Weibull Model, with BMR of 10% Extra Risk for the BMD and 0.95 Lower Confidence Limit for the BMDL Weibull Model, with BMR of 10% Extra Risk for the BMD and 0.95 Lower Confidence Limit for the BMDL 1 Weibull BMD Lower Bound 1 Weibull BMD Lower Bound 0.8 0.8 Fraction Affected 0.6 0.4 Fraction Affected 0.6 0.4 0.2 0.2 0 BMDLBMD 0 1 2 3 4 5 6 7 8 9 dose 09:52 01/21 2014 0 BMDL BMD 0 1 2 3 4 5 6 7 8 9 dose 09:53 01/21 2014 Figure J15: Fits of the Logprobit model (a), Multistage model (b), unrestricted Weibull model (c) and the restricted Weibull model (d) to the dose-response data on diffuse epithelial hyperplasia in the duodennum of female mice. J 2.4. Hystiocytic cellular infiltration in mesenteric lymph nodes in male mice. No acceptable model fitted for the dose-response analysis of the incidence of hystiocytic cellular infiltration in mesenteric lymph nodes in male mice, see Table J16 and the graph of the multistage model (Figure J16). Also the restricted loglogist, logprobit, multistage, Weibull showed no acceptable fit. No BMDL 10 was determined from the dose-response data of this endpoint for male, since no model was acceptable including restricted models. Using a different modelling approach not following EFSA (2009) ADTSR reported also no BMDL. Table J16: Histiocytic cellular infiltration in mesenteric lymph nodes in male mice. The benchmark dose (BMD 10 ) and the 95 % benchmark dose lower confidence limit (BMDL 10 ) values are given for a BMR of 10 % extra risk with characteristics of the model fit. EFSA Journal 2014;12(3):3595 247

Chromium in <strong>food</strong> and bottled <strong>water</strong><br />

a) b)<br />

LogProbit Model, with BMR of 10% Extra Risk for the BMD and 0.95 Lower Confidence Limit for the BMDL<br />

Multistage Model, with BMR of 10% Extra Risk for the BMD and 0.95 Lower Confidence Limit for the BMDL<br />

1<br />

LogProbit<br />

BMD Lower Bound<br />

1<br />

Multistage<br />

BMD Lower Bound<br />

0.8<br />

0.8<br />

Fraction Affected<br />

0.6<br />

0.4<br />

Fraction Affected<br />

0.6<br />

0.4<br />

0.2<br />

0.2<br />

0<br />

BMDLBMD<br />

0 1 2 3 4 5 6 7 8 9<br />

dose<br />

09:46 01/21 2014<br />

0<br />

BMDLBMD<br />

0 1 2 3 4 5 6 7 8 9<br />

dose<br />

09:51 01/21 2014<br />

c) d)<br />

Weibull Model, with BMR of 10% Extra Risk for the BMD and 0.95 Lower Confidence Limit for the BMDL<br />

Weibull Model, with BMR of 10% Extra Risk for the BMD and 0.95 Lower Confidence Limit for the BMDL<br />

1<br />

Weibull<br />

BMD Lower Bound<br />

1<br />

Weibull<br />

BMD Lower Bound<br />

0.8<br />

0.8<br />

Fraction Affected<br />

0.6<br />

0.4<br />

Fraction Affected<br />

0.6<br />

0.4<br />

0.2<br />

0.2<br />

0<br />

BMDLBMD<br />

0 1 2 3 4 5 6 7 8 9<br />

dose<br />

09:52 01/21 2014<br />

0<br />

BMDL BMD<br />

0 1 2 3 4 5 6 7 8 9<br />

dose<br />

09:53 01/21 2014<br />

Figure J15: Fits of the Logprobit model (a), Multistage model (b), unrestricted Weibull model (c) and<br />

the restricted Weibull model (d) to the dose-response data on diffuse epithelial hyperplasia in the<br />

duodennum of female mice.<br />

J 2.4. Hystiocytic cellular infiltration in mesenteric lymph nodes in male mice.<br />

No acceptable model fitted for the dose-response analysis of the incidence of hystiocytic cellular<br />

infiltration in mesenteric lymph nodes in male mice, see Table J16 and the graph of the multistage<br />

model (Figure J16). Also the restricted loglogist, logprobit, multistage, Weibull showed no acceptable<br />

fit.<br />

No BMDL 10 was determined from the dose-response data of this endpoint for male, since no model was<br />

acceptable including restricted models. Using a different modelling approach not following EFSA<br />

(2009) ADTSR reported also no BMDL.<br />

Table J16: Histiocytic cellular infiltration in mesenteric lymph nodes in male mice. The benchmark<br />

dose (BMD 10 ) and the 95 % benchmark dose lower confidence limit (BMDL 10 ) values are given for a<br />

BMR of 10 % extra risk with characteristics of the model fit.<br />

EFSA Journal 2014;12(3):3595 247

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!