Carbon black and lung cancer—testing a novel exposure metric by multi‐model inference

BACKGROUND: In a recent analysis of a UK cohort Sorahan and Harrington [2007: Am J Ind Med 50: 555-564] assessed the most recent 15 years of exposure ("lugging") to support their hypothesis that carbon black acts as a late stage lung carcinogen. We tested this metric in a German cohort of 1,528 carbon black workers. METHODS: We used a multi-model Cox regression approach (720 models) to explore the impact of duration and cumulative exposure to carbon black "lugged" by 0, 5, 10, 15, and 20 years. This approach covered four sub-cohorts, including an inception cohort, different exposure scenarios and varying combinations of confounders. RESULTS: Seven hundred nineteen models returned negative coefficients. Only one model estimated a small positive, but clearly non-significant coefficient (P = 0.8). CONCLUSIONS: Despite extensive searching, no exposure scenario suggested an adverse effect of "lugged" carbon black exposure on lung cancer mortality. Our analysis does not support the hypothesis of carbon black being a late stage carcinogen.