Leukaemia incidence in people living close to an oil refinery.

OBJECTIVES: To assess the incidence of leukaemia in an area downwind of a large oil refinery emitting carcinogenic volatile organic compounds (VOCs) including benzene.
METHODS: Using a dispersion model and the prevailing wind direction, two parishes with about 5000 inhabitants were a priori considered to be exposed to VOCs from the refinery. Numbers of observed and expected leukaemia cases in 1975-2004 were calculated using regional sex- and age-specific incidence rates. In addition, five nearby parishes (12000 inhabitants), considered unaffected by the emissions, served as a local reference area. Based on emission data, dispersion modelling and VOC measurements, the refinery's contribution to the population's exposure to carcinogenic VOCs was estimated. Published "unit risks" for carcinogenic VOCs were used to estimate the expected excess leukaemia risk.
RESULTS: The incidence of leukaemia in the "exposed parishes" was significantly increased in 1975-2004 (33 cases v. 22 expected cases), owing to an increase in the last 10-year period, from 1995 to 2004 (19 cases v. 8.5 expected cases). The leukaemia incidence in the local control area was normal (50 cases v. 56 expected cases). The estimated contribution from the refinery to VOC concentrations was, however, only about 2 microg/m(3) (yearly average) for benzene, 2 microg/m(3) for ethylene, 0.5 microg/m(3) for 1,3-butadiene and 5 microg/m(3) for propene. Calculations of expected excess risk using published risk estimates would indicate a much lower excess risk in the exposed parishes.
CONCLUSIONS: Using risk estimates extrapolated from high-level exposure, we would not expect an increase of leukaemia at low exposure to VOC emissions. Nevertheless, the clear elevation of leukaemia in the priori selected, exposed parishes was remarkable. Our finding may reflect a causal association due to emissions, but it could also be due to unknown confounding, or chance.