Mis-specified and non-robust mortality risk models for nasopharyngeal cancer in the National Cancer Institute formaldehyde worker cohort study.

An IARC (International Agency for Research on Cancer) working group categorized formaldehyde as a human carcinogen (Group 1) in 2004. A major component of the epidemiologic evidence evaluated by IARC to reach this decision was the analysis published by Hauptmann et al. [Hauptmann, M., Lubin, J. H., Stewart, P. A., Hayes, R. B., Blair, A. 2004. Mortality from solid cancers among workers in formaldehyde industries. Am. J. Epidemiol. 159, 1117-1130.] of the National Cancer Institute (NCI) historical cohort comprising industrial workers exposed to formaldehyde in 10 U.S. plants. The NCI authors emphasized the relationship found between highest formaldehyde peak exposure and death from nasopharyngeal cancer (NPC). We performed two additional types of re-analyses of the NCI cohort data with focus on peak exposure and NPC mortality. The analyses were aimed at (1) investigating whether the model specification chosen by Hauptmann et al. (2004) was appropriate (interaction assessment) and (2) exploring the degree of instability of the risk estimates for NPC in relation to highest peak exposure (sensitivity analysis). Hauptmann et al. (2004) failed to account for an important interaction structure between plant group and the exposure variable that prohibits a generalization of formaldehyde effects within the NCI cohort and, in particular, beyond the NCI cohort. In addition, our sensitivity analysis demonstrates considerable uncertainties in the risk estimates and points convincingly to instability problems particularly related to Plant 1. Even a simple sensitivity model taking only one additional death into account produced a variation of the risk estimates beyond the instability conveyed by standard confidence intervals. The results of our current reanalysis of the NCI study do not support NCI's suggestion of a causal association with formaldehyde exposure and nasopharyngeal cancer. The decision by the IARC working group to reclassify formaldehyde as a Group 1 substance was clearly premature considering: (1) the missing evidence of an NPC excess from the large British and NIOSH cohort studies; (2) the absence of an association with formaldehyde and NPC in the independent and expanded study of Plant 1; and (3) the mis-specified and non-robust internal analysis of the NCI cohort study brought to light in our current re-analysis. Thus, the 2004 IARC decision to reclassify formaldehyde as a Group 1 substance should be reconsidered.