Formaldehyde and Myeloid Leukemia: Diverging Tracks of Human Health Science and Regulation

In parallel with the evolution of epidemiological evidence over nearly two decades, various advisory and regulatory groups in Europe and the United States have conducted several hazard assessments on inhalation exposure to formaldehyde and risk of leukemia. All of the hazard assessments addressed all or the myeloid leukemias (MLs) combined, and none focused on the acute type, etiologically distinct and shown to have a chemical cause (e.g., benzene). However, conclusions regarding formaldehyde as a human leukemogen have been conflicting, although largely based on the same modest body of evidence. As there are no known animal models for formaldehyde and leukemia (any type), good evidence that formaldehyde cannot reach the bone marrow, and no demonstrated mechanism whereby formaldehyde induces leukemia, the hazard assessment ultimately rests on about ten epidemiological studies and their interpretation. Beginning in 2009 with the International Agency for Research on Cancer (IARC), followed by other groups in the United States and Europe, formaldehyde has been classified as causing human ML. On the other hand, the EU Scientific Committee on Occupational Exposure Limits (SCOEL) and the European Chemicals Agency (ECHA) concluded that formaldehyde unlikely causes ML. Multiple classifications of formaldehyde as leukemogenic were motivated by claims of "new studies," including Beane Freeman et al., Hauptmann et al., and Zhang et al. Closer evaluation of these reveals no increased occurrence of ML in Beane Freeman et al. and several important criticisms in the others that limit their value for classifying formaldehyde. Nevertheless, the hazard assessments summarized here varied with some observing consistent positive findings and some others-including recent meta-analyses-finding no association. That similar hazard assessment approaches applied to the same modest collection of epidemiological studies can lead to different-and even conflicting-causal conclusions is problematic. One might reasonably conclude that the research methods for evaluating epidemiological data are unreliable, leading to non-replicable and sometimes unexplained findings and conclusions. Assuming that frameworks for critical review and synthesis of epidemiological evidence are reasonably valid-and faithfully followed-it remains possible that they are methodologically inadequate, inadvertently increasing the potential for subjective elements to be incorporated and allowed to influence interpretations and conclusions. A serious international effort to develop robust and replicable hazard assessment methods based on epidemiological evidence and promote universal adoption is overdue.