Sir,

We appreciate the thoughtful comments by Rylander and Jacobs (2014) on our paper (Checkoway et al, 2014). The absence of an inverse exposure–response relation for endotoxin and lung cancer in the extended follow-up was somewhat unexpected in view of the reported consistent findings from numerous prior studies, including our initial follow-up of the Shanghai textile worker cohort (Astrakianakis et al, 2007). Although neither the modest excess relative risks observed nor the exposure–response trend for exposures >15 years since first exposure (Table 3) were statistically significant, the findings are somewhat suggestive of a possible late pro-carcinogenic effect. We do not believe that our observations on endotoxin exposure and lung cancer risk necessarily challenge a well-established association. Instead, we would argue that the exposure–response association may change over time owing to complex, yet poorly understood, underlying mechanisms. We are also not the first to report that an inverse association between endotoxin and lung cancer risk may be time varying, diminishing over time (Mastrangelo et al, 2005).

We have acknowledged the absence of data on risk factors other than active smoking, such as indoor air pollution from cooking fuels and diet. However, it is highly unlikely that either indoor air pollution or diet was correlated with endotoxin exposure in this cohort, and thus were probably not important confounders. Socio-economic status was relatively homogenous in the cohort, and also was unlikely to have been a confounder. Our exposure assessment for endotoxin (Astrakianakis et al, 2006) did take into account temporal changes in exposure levels during the cohort’s relevant work experience, to the extent that available historical data permitted. Endotoxin is a highly variable exposure, and as we noted in the paper, some exposure misclassification was inevitable.

We encourage analyses that consider temporal patterns of association in other endotoxin-exposed study populations, which can provide valuable insights into disease aetiology and pathogenesis.