Objective: To determine whether the National Cancer Institute's (NCI) recent suggestion of a causal association between formaldehyde exposure and mortality from nasopharyngeal cancer (NPC) is robust with respect to alternative methods of data analysis and alternative categorizations of formaldehyde exposure.
Methods: The original authors provided the cohort data. We computed U.S. and local county (regional) rate-based standardized mortality ratios (SMRs) and internal cohort rate-based relative risks (RR) by categories of four formaldehyde exposure metrics (highest peak, average intensity, cumulative, and duration of exposure), using both NCI categories and an alternative categorization based on tertiles of all NPC deaths among exposed subjects. We computed SMRs and RRs for each of 10 study plants and by plant group (Plant 1 (n = 4261) vs. Plants 2-10 (n = 21,358)).
Results: Six of 10 NPC deaths observed in the NCI study occurred in only one plant (Plant 1) and the remaining four cases occurred individually in four of the other nine plants studied. A large, statistically significant, regional rate-based NPC SMR of 10.32 (95% CI = 3.79-22.47) among formaldehyde-exposed workers in Plant 1 contrasted sharply with a 35% deficit in NPC deaths (SMR = .65, 95% CI = .08-2.33) among exposed workers in Plants 2-10 combined. The statistically significant exposure-response relationship with formaldehyde and NPC reported in the NCI study for highest peak exposure was driven entirely by a large, statistically significant excess NPC risk in Plant 1 for the highest peak exposure category (4+ ppm). For the remaining nine plants, RRs for all non-baseline highest peak exposure categories were less than 1.0, and we observed no evidence of an exposure-response relationship. Most of the observed NPC excesses for the non-baseline categories of the other exposure metrics (average intensity, cumulative, and duration of formaldehyde exposure) were concentrated in Plant 1, and by contrast to the NCI findings, none of the corresponding exposure-response relationships was statistically significant.
Conclusions: Overall, our reanalysis provided little evidence to support NCI's suggestion of a causal association between formaldehyde exposure and mortality from NPC. NCI's conclusion of a possible causal association was driven heavily by anomalous findings in one study plant (Plant 1). An independent and larger study of Plant 1 by the current authors concluded the NPC excess was not associated with formaldehyde exposure. Our findings cast considerable additional uncertainty regarding the validity of NCI's suggested causal association.