Results for diabetes prevalence by back extrapolated TCDD (to end of exposure), logistic regression, restricted to exposed subjects with >10 ppt TCDD at time of examination
| Study |
Coefficient for TCDD (95% CI)4-150 |
Coefficient for log TCDD (95% CI)4-150 | OR (95% CI) by quartile TCDD >10 ppt, referent <10 ppt4-151 |
| Ranch Hand | 0.0013 (0.0005 to 0.0020) | 0.3174 (0.0575 to 0.5773) | 1.02 (0.57 to 1.85), 1.60 (0.83 to 3.06), 2.30 (1.07 to 4.94) |
| NIOSH | 0.0002 (0.0000 to 0.0003) | 0.1796 (−0.2044 to 0.5635) | NA4-152 |
↵4-150 The test for heterogeneity of exposure effect between studies gave p=0.01 for untransformed TCDD, and p=0.41 for log transformed TCDD.
↵4-151 Quartiles are 10–64, 64–140, 140–376, ⩾376 ppt, common cut off points based on the combined data.
↵4-152 Quartile ORs could not be estimated with common cut off points because there were no NIOSH cases in the 2nd quartile. Collapsing the 2nd and 3rd quartiles, the ORs for the NIOSH data for quartiles 2/3v 1 is 0.76 and for quartile 4v 1 is 0.65. With NIOSH specific cut off points (10–243, 243–584, 584–1515, ⩾1515), ORs by quartiles were 0.94, 1.06, 0.75 respectively for the upper three quartilev the first.