Re: Mortality From Lymphohematopoietic Malignancies Among Workers in Formaldehyde Industries

The study described in the article byHauptmann et al. (1) has severalstrengths, including large size, long fol-low-up, and attempts to control for po-tentially important confounding factors.However, the study does not provideconclusive evidence of a causal associ-ation between formaldehyde exposureand leukemia for several reasons. In par-ticular, the large relative risks (RRs) re-ported for the internal comparison strat-ified by exposure category need to bereconciled with the external compari-sons in which the standardized mortalityratio (SMR) for all lymphohematopoi-etic neoplasms in formaldehyde-ex-posed workers is 0.8 (95% confidenceinterval [CI] 0.7 to 0.9). We haveestimated SMRs for leukemia and otherlympohematopoietic malignancies forthe peak formaldehyde exposure catego-ries shown in table 3 of Hauptmann etal. (1). These SMRs (Table 1) suggestthat mortality from lymphohematopoi-etic malignancies is not higher thanwould be expected in those workerswith a peak formaldehyde exposure of 4ppm or more, but rather is lower thanwould be expected in workers in thelowest exposure category of less than 2ppm (SMR 0.6, 95% CI 0.4 to 0.7).Similar conclusions may be drawn fromthe leukemia findings.Additionally, we point out that theassignment of peak exposure was basedprimarily on professional judgment, noton actual measurements. Such an as-signment makes this exposure the weak-est of the four exposure metrics used inthe study by Hauptmann et al. (1) rela-tive to supporting data, and this expo-sure is typically the most difficult expo-sure metric to estimate in the absence ofdetailed measurements.Hauptmann et al. (1) briefly discussthe biologic evidence relevant to theirhypothesis of formaldehyde-inducedlymphohematopoietic cancer, but theyconclude that the evidence is inconsis-tent. However, this conclusion is inconflict with substantial experimentaldata showing that, under controlledexposures, there is no increase in theconcentration of formaldehyde in theblood of humans (2 ppm), monkeys (6ppm), or rats (15 ppm) (2,3) and thatformaldehyde does not appear to in-duce cancer via inhalation at sitesother than the respiratory tract (4).These results strongly suggest that in-haled formaldehyde is rapidly metab-olized in the respiratory tract, does notreach the bone marrow, and is there-fore unlikely to induce distant-site tox-icity including leukemia.Finally, discrepancies in the datafrom available industrial studies sug-gest that the findings of Hauptmann etal. (1) may be due to chance, someuncontrolled confounding exposure, oran inappropriate comparison group.For example, a large study of workersexposed to formaldehyde in the U.K.(5) reports no increased risk for leuke-mia in the entire study cohort (SMR0.9, 95% CI 0.6 to 1.3) or amongworkers with the highest formalde-hyde exposures (SMR 0.7, 95% CI 0.3 to 1.4).Thus, we believe that the findingsof Hauptmann et al. (1) need to becritically assessed in light of externalcomparisons, existing biologic evi-dence, the findings of other studies,and a more complete understanding ofthe exposure metrics and classificationparameters used. Until these issues aremeaningfully addressed, questionswill continue to be raised about theoverall significance of the findingsreported.M