Abstract Introduction Although myelodysplastic syndrome (MDS), acute myeloid leukemia (AML), myeloproliferative neoplasms (MPN) – including chronic myeloid leukemia (CML) – and myelodysplastic/myeloproliferative neoplasms (MDS/MPN) are largely clinically distinct myeloid malignancies, epidemiological studies rarely examine them separately and often combine them with lymphoid malignancies, limiting possible etiological interpretations for specific myeloid malignancies. Methods We systematically evaluated the epidemiological literature on the four chemical agents (1,3-butadiene, formaldehyde, benzene, and tobacco smoking, excluding pharmaceutical, microbial and radioactive agents, and pesticides) classified by the International Agency for Research on Cancer as having sufficient epidemiological evidence to conclude that each causes “myeloid malignancies.” Literature searches of IARC Monographs and PubMed identified 85 studies that we critically assessed, and for appropriate subsets, summarized results using meta-analysis. Results Only two epidemiological studies on 1,3-butadiene were identified, but reported findings were inadequate to evaluate specific myeloid malignancies. Studies on formaldehyde reported results for AML and CML – and not for MDS or MPN – but reported no increased risks. For benzene, several specific myeloid malignancies were evaluated, with consistent associations reported with AML and MDS and mixed results for CML. Studies of tobacco smoking examined all major myeloid malignancies, demonstrating consistent relationships with AML, MDS and MPN, but not with CML. Conclusions Surprisingly few epidemiological studies present results for specific myeloid malignancies, and those identified were inconsistent across studies of the same exposure, as well as across chemical agents. This exercise illustrates that even for agents classified as having sufficient evidence of causing “myeloid malignancies,” the epidemiological evidence for specific myeloid malignancies is generally limited and inconsistent. Future epidemiological studies should report findings for the specific myeloid malignancies, as combining them post hoc – where appropriate – always remains possible, whereas disaggregation may not. Furthermore, combining results across possibly discrete diseases reduces the chances of identifying important malignancy-specific causal associations.
