Article Text

O27-1 Analysis of straight metalworking fluids and end-stage renal disease in autoworkers, adjusted for healthy worker survivor bias
  1. Sally Picciotto1,
  2. Ellen A Eisen1,
  3. Katie M Applebaum2
  1. 1University of California, Berkeley, School of Public Health, Berkeley, USA
  2. 2George Washington University, Milken Institute School of Public Health, Washington, USA


Background Few studies have examined occupational risk factors for kidney disease. Oil-based metalworking fluids contain polycyclic aromatic hydrocarbons, which are suspected causes of end-stage renal disease. This study analysed a cohort of autoworkers, with special attention to healthy worker survivor bias.

Methods Workers who remain employed accumulate more occupational exposure and also experience adverse outcomes later than workers who leave work for health-related reasons. This can result in underestimates of the harmful effects of exposure if proper methods are not used to adjust for time-varying confounding. G-estimation of an accelerated failure time model overcomes this problem. A cohort of workers in three General Motors plants in Michigan (USA) was followed for end-stage renal disease from 1973, when the state’s registry for dialysis patients is considered complete, through 2009. Exposure was lagged by 15 years, because end-stage renal disease develops over time and because exposure data were not available after 1994. The analysis adjusted for sex, race, age, calendar time, plant, employment status, intermittent time off work, and previous exposures to both oil- and water-based metalworking fluids. Confidence intervals were computed using the 2.5th and 97.5th percentiles from 500 bootstrap samples.

Results Exposed workers were estimated to have reached end-stage renal disease an average of 2.0 years earlier (95% CI: −0.3, 5.1) than they would have if they had not been exposed to straight metalworking fluids.

Conclusions Oil-based metalworking fluids appear to speed the development of end-stage renal disease. Further analyses using models that estimate relative risk (rather than time to event among those who fail) should supplement these results, with continued attention to time-varying confounding affected by prior exposure.

Statistics from

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.