Risk factors for lung cancer in never-smokers: Multi-cohort study

Introduction

Lung cancer is a leading cause of death worldwide and it is very well documented that cigarette smoking is its most powerful risk factor. Even with only 10% of lung cancer cases arising in people with no history of smoking,¹ the absolute number of disease events is high, such that, if lung cancer in neversmokers was a single disease entity, it would be the sixth most commonly occurring malignancy.² While this brings into sharp focus the need to identify the causes of lung cancer in never smokers, the process is methodologically challenging because large studies are required and electronic health records, the most obvious source of data, often lack information on smoking habit. The current evidence base implicates a history of non-cancer respiratory disease and exposure to second-hand smoke, asbestos, and radon as being among the few modifiable risk factors in the occurrence of lung cancer in people who never smoked.¹ We therefore examined the role of an array of other determinants by pooling individualparticipant data from a series of well-characterised, identical, field-based cohort studies.

Methods

Described in detail elsewhere,³ we used individual-participant (raw) data from 18 cohort studies with identical methodology: the Health Survey for England (15 studies) and the Scottish Health Surveys (3 studies). Ethical approval for data collection was granted by the London Research Ethics Council and local research ethics committees, and study members consented to participate.

A total of 222,829 people participated in a home-based biomedical survey between 1994 and 2008 when aged 16-102 years; 193,873 consented to being linked to a national mortality registry (figure 1). Of these, 94,456 people reported that they had never smoked. The further omission of study members with a plasma or salivary cotinine >1.0 ng/mL to identify so called smoking deceivers;⁴ those reporting use of nicotine replacement therapy; and study members with a history of cancer, resulted in 91,588 (55,452 women) remaining individuals and this was our analytical sample.

• Download figure
• Open in new tab

Figure 1. Flow of study members into the analytical sample:

Follow-up of study members in the Health Survey for England and the Scottish Health Surveys

HSE, Health Survey for England; SHS, Scottish Health Survey; NRT, nicotine replacement therapy

Results

A maximum of 17 years of mortality surveillance (mean 9.7) of 91,588 study members gave rise to 85 lung cancer deaths. Being older (hazard ratio; 95% confidence interval per 10 year increase: 2.45; 2.11, 2.85) and male (versus female: 2.25; 1.46, 3.48) was associated with an elevated risk of lung cancer (table 1). While there was also a suggestion of associations with some of the others study member characteristics – higher body mass index and alcohol consumption – these did not reach conventional levels of statistical significance. The only exception was higher intake of vegetable and fruit where there was a more than doubling in the rate of lung cancer mortality (2.29; 1.25, 4.17).

Discussion

Our main finding was that, with the exception of risk factor associations for higher age and being male -- both well-replicated in this context¹ -- there was little evidence of a clear association for the remaining 17 potential determinants of never-smoking lung cancer. Our finding that higher fruit and vegetable consumption appeared to confer elevated risk has been reported in other studies⁷ but is not a universal observation.⁸ Given the paucity of known risk factors for lung cancer in never smokers, we included as many potential determinants as possible from our dataset, however, rarely, some of these (e.g., psychological distress), were not hypothesis-driven.

In conclusion, despite the number and breadth of potential environmental risk factors described, there was no clear emergence of new determinants for never-smoking lung cancer. It may be that future research should consider the role of genetic indices.