O26-4 Inverse associations between occupational organic dust exposure and incidence of chronic obstructive pulmonary disease (copd) – healthy worker survivor bias? [Conference abstract]

Background Organic dust exposure common in the farming and wood industry is a suggested risk factor for COPD.

Aims To investigate exposure-response relations between different exposure metrics for organic dust and the incidence of COPD in the farming and wood industry.

Methods We identified 35,957 blue-collar workers and followed them since their first employment in farming or wood industry 1997–2007. Information on exposure was obtained from two industry-based exposure matrices with time-dependent quantitative exposure estimates for organic dust in farming and wood industry. Incident COPD cases as defined by ICD-codes 1997–2013 were identified in the Danish national patient register. We analysed data with discrete-time hazard logistic regression with person-years as the unit of analysis and adjusted for age, sex, calendar year, and labour market attachment.

Results We observed an inverse association between levels of organic dust and COPD (P trend < 0.0001). The adjusted RRs (95% CIs) for cumulative exposure for the 2nd, 3rd, and 4th exposure quartiles compared with the 1st were; 1.01 (0.82; 1.26), 0.66 (0.51; 0.84), and 0.70 (0.53; 0.91). Comparable results were seen for average exposure and exposure duration. Analyses on a subset of the population with individual smoking information showed a slightly declining prevalence of smokers with increasing organic dust exposure, but adjusting for smoking did not affect the observed associations. Conclusions We found an inverse association between organic dust exposure and COPD. In this inception cohort our results should be less affected by a healthy hire effect or left truncation bias, although this can still not be ruled out completely. We find it unlikely that the declining risk of COPD reflects a gainful effect of organic dust exposure, but rather healthy worker survivor bias related to exposure level, smoking, or other underlying individual factors associated with exposure level even if our sub-analyses were unable to detect a confounding effect of smoking.