Public health and chronic low chlordecone exposure in Guadeloupe, Part 1: hazards, exposure-response functions, and exposures

Table 3 BMD₁₀ for liver carcinomas in female mice (from NCI, 1976)

Model	Total number of parameters in the model	AIC	Goodness of fit test P-value	Scaled residual near BMR	BMD₁₀ (mg/kg/d)	BMD_10-HED (mg/kg/d)
Gamma-Restricted	3	147.182	0.014	0	0.671	0.1005
Gamma	3	141.305	0.568	0	2E-35	na
Logistic	2	166.361	0.000	3.47	1.824	0.2734
LogLogistic-Restricted	3	143.927	0.081	0	0.464	0.0695
LogLogistic	3	139.233	0.881	0	computation failed
LogProbit-Restricted	3	149.575	0.004	0	1.149	0.1722
LogProbit	3	139.233	0.881	0	computation failed
Multistage-Restricted	2	147.182	0.014	0	0.671	0.1005
Multistage	2	140.979	1.000	0	0.275	0.0412
Multistage-Cancer 1	2	147.182	0.014	0	0.671	0.1005
Multistage-Cancer 2	3	147.182	0.014	0	0.671	0.1005
Multistage-Cancer 3	4	147.182	0.014	0	0.671	0.1005
Probit	2	164.975	0.000	3.509	1.704	0.2553
Weibull-Restricted	3	147.182	0.014	0	0.671	0.1005
Weibull	3	147.182	0.014	0	0.671	0.1005
Quantal-Linear	3	147.182	0.014	0	0.671	0.1005
Minimal BMD						na
Maximal BMD						na
Ratio max/min						na
BMD from best AIC model						0.0412
Geometric mean of BMDs for acceptable models						na

The benchmark response is 10 %. AIC: Akaïke’s Information Criterion. BMD₁₀: Benchmark dose for 10 % excess risk. Results in bold are acceptable models according to these criteria: total number of parameters does not exceed the number of groups modeled (n = 3); and p-value > 0.1; and scaled residual < │2│ and no computation error or unrealistic graphical BMD₁₀. If ratio max/min is less than 3, then geometric mean of acceptable models is selected, else the best AIC model is selected. “na”: not applicable

ISSN: 1476-069X