Indirect adjustment of tobacco smoking in occupational studies of lung cancer: A systematic review of the available methods and their applications

Abstract

Tobacco smoking is an important risk factor and potentially a major confounding factor in occupational lung cancer studies. However, as individual information on tobacco smoking is often not available, indirect adjustment methods may be used to account for potential confounding from smoking. Therefore, we aimed at providing an overview of the available indirect adjustment methods for smoking in studies of occupational exposures and lung cancer risk. We conducted a systematic search of relevant studies that applied statistical methods for indirect adjustment of tobacco smoking and were published between 1-Jan-2000 and 2-Apr-2025 to capture developments in recent decades. Studies were retrieved from Embase, MEDLINE, and Web of Science. Fifteen studies fulfilled our inclusion criteria and were included. We grouped the studies into four methods of indirect smoking adjustment: (1) without distributions for adjusted data; (2) distributions for adjusted data; (3) negative control outcomes; (4) factor analysis models. For studies with an external comparison group, percentage change in estimates from before to after indirect adjustment ranged -36.1 %_to_+ 17.3 %, while the corresponding range for those with internal comparison was -16.2 %_to_+ 47.8 %. The choice of indirect adjustment method depends on the use of reference group (external vs. internal) and the data available. Adjustment methods 1 and 2 use partial cohort data or ancillary data from other similar workers and may be preferable over methods 3 and 4, if such data are available. Methods 3 and 4 may be well suited if such data are lacking but have stronger assumptions.

Keywords: Confounding; Correction; Factor analysis models; Negative control outcomes.