THE HEALTHY WORKER SURVIVOR EFFECT AND MORTALITY AT 2 AUTOMOTIVE ENGINE MANUFACTURING PLANTS

Attributes such as rime since hire or length of follow up may be impor tant in occupational mortality due to the ''healthy worker survivor ef fect.'' In a regression analysis of mortality odds ratios at two autom otive engine plants, strong effects of overall employment duration (la tency weighted) were observed in addition to effects for (similarly we ighted) cumulative exposures. The duration effects were negative for s everal cancer and noncancer outcomes, mid confounded exposure effects. The lung cancer odds ratio declined to 0.68 (95% CI = 0.51, 0.90) at the mean duration of employment. With control for employment duration, adjusted lung cancer odds ratios for work as millwrights increased fr om 3.0 to 3.8, and for work in cylinder head production, from 3.3 to 3 .9. Several causes of death with strong duration effects were smoking- related, suggesting diminished smoking risk factors with increasing em ployment duration. Similarly, trends for cirrhosis of the liver mortal ity suggested the alcohol risk factor is smaller in long-duration work ers. If personal risk factors are an important component of the health y worker survivor effect, they could be powerful negative confounders of exposure-response for related outcomes. Including a term for employ ment duration in regression models appears to partially correct for he althy worker survivor bias. (C) 1996 Wiley-Liss, Inc.