Asthma and asthma symptom control in relation to incidence of lung cancer in the HUNT study

Large prospective studies on asthma, especially asthma symptom control, as a potential risk factor for lung cancer are limited. We followed up 62,791 cancer-free Norwegian adults from 1995–1997 to 2017. Self-reported doctor-diagnosed asthma was categorized into active and non-active asthma. Levels of asthma symptom control were classified into controlled and partially controlled (including partly controlled and uncontrolled) according to the Global Initiative for Asthma guidelines. Incident lung cancer cases were ascertained from the Cancer Registry of Norway. Cox regression models were used to estimate hazard ratios (HRs) with 95% confidence intervals (CIs) for possible associations. Totally, 984 participants developed lung cancer during a median follow-up of 21.1 years. After adjustment for smoking and other potential confounders, an increased incidence of lung cancer was found for adults with partially controlled asthma (HR 1.39, 95% CI 1.00–1.92) compared with those without asthma at baseline. Adults with active asthma had a tendency of increased lung cancer incidence (HR 1.29, 95% CI 0.95–1.75). Sensitivity analyses indicated that the observed associations were less likely resulted from reverse causation or residual confounding by smoking. Our findings suggested that proper control of asthma symptoms might contribute to a reduced incidence of lung cancer.

. Distribution of baseline characteristics according to levels of asthma symptom control in the HUNT2 Study, 1995-1997 (n=62791 1 Age was used as the time scale in the crude model. 2 Adjusted for sex, body mass index, smoking [(never, former (<10, 10-20, and >20 pack-years (pyrs)), current (<10, 10-20, and >20 pyrs)], passive smoking, alcohol consumption, physical activity, total sitting time daily, education, economic difficulties, family history of cancer and allergic rhinitis. Age was used as the time scale. Tvc option of the stcox command in Stata was used to model the non-proportional hazards for sex, smoking and economic difficulties in the adjusted models. 3 An "unknown" level of asthma symptom control with limited lung cancer cases (n=4) is not shown.  1 Age was used as the time scale in the crude model. 2 Adjusted for sex, body mass index, smoking [(never, former (<10, 10-20, and >20 pack-years (pyrs)), current (<10, 10-20, and >20 pyrs)], passive smoking, alcohol consumption, physical activity, total sitting time daily, education, economic difficulties, family history of cancer and allergic rhinitis. Age was used as the time scale. Tvc option of the stcox command in Stata was used to model the non-proportional hazards for sex, smoking and economic difficulties in the adjusted model.
Supplementary Table S4. The associations of asthma overall, asthma status and levels of asthma symptom control with lung cancer incidence after excluding participants with asthma who also had post bronchodilator FEV1/FVC z score <-  1 Age was used as the time scale in the crude model. 2 Adjusted for sex, body mass index, smoking [(never, former (<10, 10-20, and >20 pack-years (pyrs)), current (<10, 10-20, and >20 pyrs)], passive smoking, alcohol consumption, physical activity, total sitting time daily, education, economic difficulties, family history of cancer and allergic rhinitis. Age was used as the time scale. Tvc option of the stcox command in Stata was used to model the non-proportional hazards for sex, smoking and economic difficulties in the adjusted model.
Supplementary The aim of analysis using a negative control exposure is to identify residual confounding that may have resulted in invalid causal inference for the main exposure-outcome association 1 . In the current study, we used "migraine" as the negative control exposure to detect residual confounding by smoking in the observed asthma-lung cancer association (Supplementary Figure). "Migraine" was chosen as the negative control exposure was because it is associated with the confounder (smoking) 2,3 , but not causally associated with the outcome (lung cancer). We expected to observe a null association between migraine and lung cancer after adjustment for smoking, suggesting that the observed asthma-lung cancer association was less likely biased by the residual confounding of smoking.
Based on our study population (n=62791), we excluded participants without information on headache. This left 49945 participants to study the relationship between migraine and lung cancer incidence. Participants with migraine were those who answered yes to the question "Have you suffered from headache during the last 12 months?" and specified the type of headache as "migraine", and the rest were regarded as no migraine.
We adjusted for smoking status in model 1. In model 2, we adjusted for the same confounders as in our primary study (Supplementary Table  S6).
The negative control exposure analysis showed that fewer participants with migraine were heavy smokers (> 20 pack-years) than participants without migraine (6.1% vs 10.3%). There was an inverse association between migraine and lung cancer incidence without adjustment for smoking (crude HR 0.61, 95% CI 0.43-0.86). After adjustment for smoking the association between migraine and lung cancer became less clear (HR 0.75, 95% CI 0.53-1.06), and additional adjustment for the same confounders as in the primary study did not have material changes in the result. This indicated that our observed associations of active asthma and partially controlled asthma with increased lung cancer incidence were less likely biased by residual confounding due to smoking.
Supplementary Figure. DAG for asthma (as the main exposure), migraine (as the negative control exposure) and incidence of lung cancer (as the outcome) Figure legend. This DAG was created by using DAGitty V 3.0 4 (http://www.dagitty.net/dags.html). Supplementary