Article Text


0124 PM2.5 and Heart Disease in a Cohort of Aluminium Workers: An Application of Longitudinal Targeted Maximum Likelihood-based Estimation (TMLE)
  1. Daniel Brown1,
  2. Maya Petersen1,
  3. Mark van der Laan1,
  4. Sadie Costello1,
  5. Elizabeth Noth1,
  6. Katherine Hammond1,
  7. Mark Cullen2,
  8. Ellen Eisen1
  1. 1UC Berkeley, Berkeley, CA, USA
  2. 2Stanford University, Palo Alto, CA, USA


Objectives We estimated the effect of cumulative exposure to PM2.5 on the incidence of ischaemic heart disease (IHD) in aluminium workers followed for 15 years, adjusting for the healthy worker survivor effect. In previous analyses, higher cumulative exposure was found to be associated with lower mortality in this population.

Method We used longitudinal TMLE to estimate the cumulative risk of ischaemic heart disease in the cohort if constantly exposed above an exposure cut-off and compared it to the risk if constantly exposed below. We stratified all analyses by work process because exposures were an order of magnitude higher in smelters than fabrication facilities.

Results We selected cut-offs a priori at the median and 10th percentile exposure within each sub-cohort. Among the smelter workers, we estimated an increase in IHD risk of 2.1% (p = 0.22) after 15 years, comparing the always exposed to never exposed cohort using the median cut-off of 1.77 mg/m3. The difference was 2.9% (p = 0.01) using the 10th percentile cutoff of 0.10 mg/m3. For the fabrication workers, the differences were 0.1% (p = 0.47) using the median cut-off of 0.20 mg/m3 and 2.5% (p < 0.01) for the 10th percentile cut-off of 0.06 mg/m3. Results are presented as adjusted survival curves, describing the estimated cumulative risk for each cohort under each exposure regimen.

Conclusions The TMLE estimator allows us to observe an association between cumulative PM2.5 exposure and heart disease that was not visible using standard analytical techniques. This work represents the first application of longitudinal TMLE to the field of occupational epidemiology.

Statistics from

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.