WHAT PROPORTION OF LUNG CANCER IN NEVER-SMOKERS CAN BE ATTRIBUTED TO KNOWN RISK FACTORS?

Abstract

Though tobacco smoking is the primary risk factor for lung cancer, a significant fraction of lung cancer deaths occur in lifetime non-smokers. In this paper, we calculate the burden of lung cancer in never-smokers attributable to previously identified risk factors in North America, Europe, and China, using population-based estimates of exposure prevalence and estimates of relative risk derived from recently published meta-analyses. Population attributable fractions (PAFs) for individual risk factors ranged from 0.40% to 19.93%. Due to differences in the prevalence of exposures, the PAFs associated with several of the risk factors varied greatly by geographical region. Exposure to the selected risk factors appeared to explain a much larger proportion of lung cancer cases in never-smokers in China than in Europe and North America. Our results demonstrate the geographic variability of the epidemiology of lung cancer in never-smokers, and highlight the need for further research in this area, particularly in Europe and North America.

INTRODUCTION

Lung cancer is a serious public health problem, accounting for about 16 percent of all incident cancers and 18 percent of all cancer deaths in 2008.¹ While the majority of lung cancer can be attributed to tobacco smoking, approximately 10 to 15% of cases occur in individuals who have never smoked.^{2, 3} Although the corresponding range ofglobal deaths, 180,000 to 270,000 per year, accounted for a measurable fraction of worldwide cancer mortality in 2008,¹ lung cancer in never-smokers has been studied far less extensively than tobacco-related lung cancer.

Despite relatively little research in this area, several risk factors for lung cancer in never-smokers have been identified, including secondhand smoke (SHS), previous lung diseases, indoor radon, occupational exposures, household coal smoke, and family history of lung cancer. The prevalence of these risk factors, as well as the prevalence of active smoking, varies widely across the globe, resulting in regional differences in the epidemiology of lung cancer in never-smokers. The aim of the present study is to estimate the proportion of lung cancer in 2008 that occurred in never-smokers in three different geographic regions that can be attributed to known risk factors.

METHODS

This study aimed to estimate the population attributable fraction (PAF) of lung cancer in never-smokers attributable to several known risk factors in three different geographic regions: North America, Europe, and China.

Population Attributable Fraction

The population attributable fraction (PAF) is an estimate of the disease burden in a population that would be avoided if population-level exposure to a specific risk factor were to be reduced or eliminated. In this report, all PAFs for each risk factor were calculated under the alternative scenario of no exposure. PAFs were calculated using the following formula ⁴:

$$\text{AF}=\text{P}\times (\text{RR}-1)∕[(\text{P}\times \text{RR}-1)+1]$$

where RR is the relative risk of lung cancer in never-smokers associated with exposure to a risk factor, and P is the prevalence of exposure to the risk factor in the never-smoking population. Greenland has previously demonstrated that this formula may not be optimal when using adjusted relative risks,⁵ and modified formulas for adjusted estimates are available.⁶ However, we were limited to the use of Levin’s formula, as our estimates of relative risk were largely derived from pooled meta-analyses, without detailed information on the adjustment variables used in each study. We used the assumption of a linear non-threshold relationship between residential radon exposure and lung cancer risk;^{7, 8} since radon was measured continuously, we derived risk using the following equation:

$$\text{Risk}=ex{p}^{\text{In}\left(\text{Risk}\text{per}\text{unit}\right)\times \text{average}\text{level}\text{of}\text{exposure}]}$$

and used the following formula to calculate the PAF:

$$\text{PAF}=(\text{Risk}-1)∕\text{Risk}$$

In addition to calculating separate PAFs for SHS exposure that occurred in the home and at work, a combined PAF was calculated for both sources of exposure. Under the assumption that the two sources of exposure are independent and uncorrelated, the following equation can be used:⁹

$${\text{PAF}}_{\text{total}}={\text{PAF}}_{\text{home}}\times {\text{PAF}}_{\text{workplace}}+{\text{PAF}}_{\text{home}}\times (1-{\text{PAF}}_{\text{workplace}})+{\text{PAF}}_{\text{workplace}}\times (1-{\text{PAF}}_{\text{home}})$$

Non-smoking individuals may be exposed to two or more risk factors that increase their risk of lung cancer, possibly in a synergistic manner. However, there is essentially no published data on either the joint effect of exposure to multiple risk factors on lung cancer risk in never-smokers or on the concomitance of the selected risk factors in these populations, so all PAFs were calculated under the assumption that each risk factor affects lung cancer risk independently of the others. However, because overlapping exposures may jointly contribute to lung cancer, summing the PAFs due to multiple risk factors in a region will likely lead to an overestimate of the proportion of lung cancer deaths that are attributable to these causes.

Selection of Risk Factors

We chose to calculate the PAF of risk factors that had been conclusively demonstrated to increase the risk of lung cancer in never-smokers. Environmental tobacco smoke, radon, and use of coal for household heating and cooking have all been classified as Group 1 carcinogens by the International Agency for Research on Cancer (IARC),^10-12 indicating sufficient evidence of carcinogenicity in humans. We also evaluated two other risk factors that have been consistently associated with increased risks of lung cancer in studies of never-smokers: a family history of lung cancer¹³ and a personal history of previous lung diseases, including chronic obstructive pulmonary disease (COPD), pneumonia, and tuberculosis.¹⁴ Because few attempts have been made to quantify the risk of lung cancer associated with household coal use in North America or Europe, a PAF for this risk factor was calculated only for China. No estimate for the PAF of COPD was calculated in European populations, as the relative risk for this factor was below one.

We chose not to assess the impact of diet on the risk of lung cancer in this report. Although an inverse association between high intakes of fruits and vegetables and lung cancer has been hypothesized, results in this area have been mixed. While results from some large cohort studies have suggested a protective effect of these foods in never-smokers,^15-17 in other studies, this relationship was observed exclusively in current and former smokers.^18-20 One large, international cohort study with 25 years of follow-up reported null findings for both food groups.²¹ The magnitude of the possible inverse association is also a matter of debate. While some studies have reported reductions in lung cancer risk of greater than 30% for fruits^22-24 and 40% for vegetables,¹⁶ estimates from a recent, large cohort study were a mere 2% for fruits and vegetables combined, with no effect seen for fruits alone.²⁵ In its 2007 report, the WCRF/AIRC concluded that the evidence for an inverse association with lung cancer was “probable” for fruits, and “limited” for non-starchy vegetables, though this report did not distinguish according to smoking status.²⁶

Results of several studies have provided some evidence that exposure to airborne particulate matter, including traffic emissions, may be risk factor for lung cancer. During 26 years of follow-up in the Cancer Prevention II study (CPS-II), an increase of 10-μg/m³ in fine particulate matter (PM_2.5) was associated with a 15-27% increase in lung cancer risk among never-smokers.²⁷ Conversely, a case-control study nested within the European Prospective Investigation into Cancer and Nutrition study (EPIC), found no association between increasing levels of particulate matter less than 10 (PM₁₀) and lung cancer risk in never- and ex-smokers, though high levels of NO₂ were significantly associated with higher.²⁸ However, these studies used aggregate-level data to quantify particulate levels, not individual measurements of exposure, so ecological fallacy may be an issue. Other studies have attempted to estimate individual-level exposure by interpolating concentrations of particulate matter from air-monitoring stations to zip codes of study participants,²⁹ or by using dispersion models.³⁰ A recent study using the validated Danish AirGIS dispersion modeling system to assess individual exposure showed a significantly increased risk of lung cancer among nonsmokers exposed to high levels of nitrous oxides (>29.7 μg/m³), though no linear trend was observed with increasing concentrations.³⁰ In addition to being vulnerable to measurement error, many air pollution studies have not used consistent definitions of particulate matter or reference groups. While there may be some increased risk of lung cancer associated with exposure to particulate matter, further studies are needed before any PAF can be calculated for this risk factor.

Occupational exposures to a number of substances, including asbestos, silica, and arsenic,^31-33 have been classified by IARC as carcinogenic. Some authors have also classified several specific occupations and industries as “List A or B”, reflecting workers’ potential exposure to known or suspected carcinogens, respectively.³⁴ Unfortunately, many studies of occupational exposures and lung cancer either do not stratify their results by smoking status, or include very small numbers of never-smokers, leading to imprecise estimates of risk. A small number of studies have been conducted in lifetime non-smokers, but often have few cases who report in employment in any of the occupations entailing carcinogenic exposure, but results of these studies have not been consistent. One European case-control study reported increased risks for men and women employed in list-A occupations, though not for men employed in list-B occupations. ³⁵ Zeka et al found that risk was increased in women employed in list B, but not list A, occupations,³⁶ while a cohort study of female never-smokers found no elevated risk for either list, but only for specific occupations.³⁷ Because these issues prevent the calculation of any reliable calculation of the relative risks of lung cancer in never-smokers, we did not calculate a PAF for occupational exposures, but present a summary of relevant studies published since 2000 in Table 1.

Table 1. Results of Selected Studies on Occupations and Lung Cancer in Never-Smokers.

Reference, study location, study period	Study Design	Exposure Category	Relative Risk	Adjustment Variables
IARC List A and B Occupations
Pohlabeln et al.35 Europe, multi-center;1988-1994	Population and hospital-based case-control; Male and female never-smokers; 650 cases, 1542 controls	Ever worked in a List A occupationa
		Men	1.52 (0.78, 2.97)	Age, study center
		Women	1.50 (0.49, 4.53)	Age, study center
		Ever worked in a List B occupationa
		Men	1.05 (0.60, 1.83)	Age, study center
		Women	1.69 (1.09, 2.63)	Age, study center
		Ever worked in a List A or List B occupationa
		Men	1.20 (0.76, 1.92)	Age, study center
		Women	1.67 (1.10,2.52)	Age, study center
Kreuzer et al.22 Germany, multi- center; 1990-1996	Population-based case-control; Never-smoking men; 58 cases, 803 controls	Ever worked in a List A occupationa	2.4 (1.06, 5.43)	Age, region
		Ever worked in a List B occupationa	1.4 (0.71, 2.79)	Age, region
Kreuzer et al.38; Germany, multicenter; 1991-1996	Population-based case-control; Never-smoking women; 234 cases, 535 controls	Ever worked in a List A occupation	0.77 (0.29, 2.50)	Age, region
		Ever worked in a List B occupationa	1.51 (0.84, 2.71)	Age, region
		Ever worked in a List A or List B occupationa	1.32 (0.78, 2.23)	Age, region
Zeka et al.36; Europe, multi-center; 1998-2002	Population and hospital-based case-control; Male and female never-smokers; 223 cases, 1039 controls	Ever worked in a List A occupationa
		Men	0.43 (0.09, 2.06)	Age, study center
		Women	0.72 (0.26, 2.01)	Age, study center
		Ever worked in a List B occupationa
		Men	0.85 (0.37, 1.98)	Age, study center
		Women	1.37 (0.66, 2.84)	Age, study center
		Ever worked in a List A or List B occupationa
		Men	0.74 (0.34, 1.61)	Age, study center
		Women	1.09 (0.60, 2.01)	Age, study center
Pronk et al.37; Shanghai, China; 1996-2000	Cohort study; 71,067 lifetime non-smoking women; 219 lung cancer cases	Ever worked in a List B occupationa	0.8 (0.3, 1.9)	Passive smoking, education level, family history of lung cancer
		Ever worked in a List A or List B occupationa	0.6 (0.3, 1.6)	Passive smoking, education level, family history of lung cancer
Asbestos
Neuberger et al.39; Iowa, USA; 1994-1997	Population-based case control; Never-smoking women; 37 cases, 413 controls	Occupational exposure to asbestos	4.38 (1.10, 17.45)	Age, education, radon exposure
Brenner et al.40; Canada; 1997-2002	Population and hospital-based case control; Never-smoking men and women; 156 cases, 466 controls	Occupational exposure to asbestos	1.0 (0.3, 3.0)	Age, sex, ethnicity, education
Frost et al.41; Great Britain; 1971-2005	Cohort of asbestos workers; 35 never-smoking cases	Occupational exposure to asbestos (medium v. low)	1.9 (0.8, 4.3)	Age, calendar period, sex main occupation
		Occupational exposure to asbestos (high v. low)	1.6 (0.6, 4.2)	Age, calendar period, sex main occupation
Tse et al.42; Hong Kong; 2004-2006	Community-based case-control; Lifetime non-smoking men; 132 cases, 536 controls	Occupational exposure to asbestos dust (v. no exposure to any listed carcinogens)	0.99 (0.30, 3.32)	Age, place of birth, education levels, residential radon exposure, previous lung disease, any cancer in first-degree relatives intake of meat
Silica
Zeka et al.36; Europe, multi-center; 1998-2002	Population and hospital-based casecontrol; Male and female never-smokers; 223 cases, 1039 controls	Ever exposed to silica	1.76 (0.97, 3.21)	Age, sex, study center
Cassidy et al.43; Europe, multicenter; 1998-2002	Community-based case-control; Never-smoking men and women; 274 cases, 1110 controls	Occupational exposure to silica	1.41 (0.79, 2.49)	Age, sex, study center, education, exposure to wood dust and insulation dust
Tse et al.42; Hong Kong; 2004-2006	Community-based case-control; Lifetime non-smoking men; 132 cases, 536 controls	Occupational exposure to silica dust (v. no exposure to any listed carcinogens)	3.09 (1.30, 7.37)	Age, place of birth, education levels, residential radon exposure, previous lung disease, any cancer in first-degree relatives intake of meat
Vida et al.44; Canada; 1979-1986, 1996-2001	2 pooled population-based case control studies; Never-smoking men; 38 cases, 482 controls	Any occupational exposure to silica	1.28 (0.93, 1.75)	Age, race, proxy respondent, education, region, occupational exposure to other carcinogens, study
		Substantial occupational exposure to silica	1.77 (1.05, 2.98)	Age, race, proxy respondent, education, region, occupational exposure to other carcinogens, study

^aReference category is “Never worked in List A or List B occupation” for all estimates

Emissions from high-temperature frying have been shown to contain a number of cancer-causing organic compounds, including polycyclic aromatic hydrocarbons (PAHs), and have been classified as probably carcinogenic by the IARC working group.⁴⁵ Exposure to cooking fumes is prevalent in many Asian countries, where deep-frying is common, and kitchens are often poorly ventilated. Increased levels of several carcinogens were found in the urine of non-smoking Chinese women who reported frequent wok-style cooking.⁴⁶ While several case-control studies have consistently demonstrated elevated risks in never-smoking Chinese women who are exposed to cooking fumes,^47-53 reliable data on the prevalence of high-temperature frying is not widely available, and definitions of exposure vary widely between studies, precluding the estimation of the PAF for this risk factor.

Data used to calculate the RRs

Table 2 provides a summary of the relative risks used in our calculations. For each risk factor, estimates of relative risk were derived from the most recently published meta-analysis of observational epidemiological studies. We sought separate estimates of relative risk for each of the geographical regions in our study; when these were not immediately available, we calculated them by combining estimates from the studies included in the most recent meta-analysis using the metaan command in STATA software version 11 (STATA, College Station TX). Separate estimates of risk were used for SHS exposure occurring in the home and in the workplace. When relative risks were not presented separately by gender, the increases in risk associated with all factors were assumed to be the same for men and women in each region. Few studies estimated the risk associated with household coal smoke in never-smoking men, so this calculation was performed only for women, who are likely more exposed as they traditionally perform most cooking in Chinese households.

Table 2. Relative Risks for Selected Risk Factors.

	Relative Risk (95% Confidence Interval)	Reference
North America
SHS exposure at home	1.15 (1.04, 1.26)	54
SHS exposure at workplace	1.24 (1.03, 1.49)	54
Tuberculosis	1.45 (1.13, 1.85)	55
COPD	1.35 (1.00, 1.83)	14
Pneumonia	1.36 (1.10, 1.69)	14
Family history of lung cancer	1.31 (1.05, 1.65)	13
Residential radona	0.10 (−0.09, 0.42)	56
Europe
SHS exposure at home	1.16 (1.03, 1.30)	54
SHS exposure at workplace	1.13 (0.96, 1.34)	54
Tuberculosis	1.45 (1.13, 1.85)	55
COPD	0.88 (0.59, 1.31)	14
Family history of lung cancer	1.29 (0.90, 1.84)	13
Residential radona	0.106 (0.003, 0.28)	57
China
SHS exposure at home	1.43 (1.24, 1.66)	54
SHS exposure at workplace	1.32 (1.13, 1.55)	54
Tuberculosis	2.10 (1.47, 3.01)	55
COPD	1.18 (0.72, 1.94)	14
Family history of lung cancer	1.67 (0.95, 2.94)	13
Residential radona	0.124	58
Household use of coalb	2.93 (1.40, 6.12)	59

^aExcess risk per 100 Bq/m³ increase

^bFemales only

Data used to estimate the prevalence

The prevalence estimates for each risk factor are listed in Table 3. We assumed a lag time of approximately 15 years between the relevant carcinogenic exposure and development of lung cancer; therefore, we sought prevalence data that reflected exposure in the early to mid-1990s. When possible, we used prevalence data from large, population-based surveys. If no such data was available, we used estimates from recent meta-analyses, cohort studies or from population-based case-control studies.

Table 3. Prevalence Estimates for Selected Risk Factors.

	Prevalence of exposure (%) (95% Confidence Interval)	Data Source	Reference
North America
SHS exposure at home		Third National Health and Nutrition Examination Survey (NHANES III)	66
Men	14.3 (13.3, 15.4)
Women	14.9 (14.0, 15.8)
SHS exposure at workplace			66
Men	27.7 (26.4, 29.0)
Women	15.2 (14.3, 16.1)
Tuberculosis	2.56 (2.30, 2.86)	World Health Organization (WHO), meta-analysisa	62, 67
COPD	1.15 (0.70, 1.89)	Meta-analysis, NHANES IIIa	64, 68
Pneumonia	0.476 (0.475, 0.477)	National Hospital Discharge Survey	69
Family history of lung cancer	6.61 (5.29, 8.14)	Population-based case control study	70
Residential radonb	25 (3.1)	United Nations Scientific Committeee on the Effects of Atomic Radiation (UNSCEAR 2000)	71
Europe
SHS exposure at home		Population-based case control study	72
Men	12.8 (10.1, 16.0)
Women	62.7 (59.6, 65.6)
SHS exposure at workplace			72
Men	70.1 (67.5, 75.3)
Women	47.0 (43.9, 50.2)
Tuberculosis	5.49 (4.90, 6.10)	WHO, meta-analysisa	62, 67
COPD	1.73 (1.38, 2.18)	Meta-analysis, IBERPOCa	63, 68
Family history of lung cancer	4.23 (3.10, 5.64)	Population-based case control study	13
Residential radonb	59	UNSCEAR 2000	71
China
SHS exposure at home		International Collaborative Study of Cardiovascular Disease in Asia (InterASIA)	60
Men	12.1 (11.2, 13.0)
Women	51.3 (50.5, 52.1)
SHS exposure at workplace			60
Men	26.7 (25.5, 27.9)
Women	26.2 (25.5, 26.9)
Tuberculosis	13.2 (11.8, 14.6)	WHO, meta-analysisa	62, 67
COPD	3.47 (1.16, 9.30)	Meta-analysis, Chinese Epidemiological Survey of COPD (CESPOD)a	65, 68
Household use of coal	12.9 (7.9, 17.9)	WHO Global Burden of Disease	73
Family history	4.5 (4.06, 4.97)	Cohort study	48
Residential radonb	34.4 (1.95)	UNSCEAR 2006	74

^aUsed to calculate proportion of COPD/tuberculosis cases among never-smokers

^bGeometric mean (Bq/m³) (SD)

Exposure to SHS at home and in the workplace were the only risk factors for which prevalence was calculated separately in men and women. We estimated prevalence of SHS from surveys that used self-report questionnaires, not biomarkers (such as serum cotinine level), to assess exposure. SHS data from China was taken from a survey conducted in 2004,⁶⁰ since a 1996 study in the region did not provide detailed estimates of SHS exposure for non-smokers.⁶¹

Estimates of the prevalence of COPD and tuberculosis were available only for the entire population, though these diseases are more frequently found in smokers. A recent meta-analysis reported that the relative risk of tuberculosis for smokers compared to non-smokers is 1.73 (95% CI: 1.46, 2.04);⁶² this estimate was used to calculate the prevalence of tuberculosis in the never-smoking population. Similarly, large, population-based studies were used to estimate the prevalence of COPD in never-smokers in each geographical area.^63-65 Because no reliable data was available on the incidence of pneumonia in China or Europe, a PAF for this risk factor was calculated only for the United States.

Exposure to residential radon was reported as the geometric mean of indoor radon levels in each region.

RESULTS

The PAFs of lung cancer attributable to the selected risk factors varied by geographical region (Table 4). These risk factors appeared to be responsible for a significant proportion of lung cancer in Chinese never-smokers, but accounted for a smaller proportion of cases in Europe and North America. In China, known risk factors accounted for a larger proportion of lung cancer cases in women than in men. This difference is due, in part, to the PAF due to coal smoke, which was high among women (19.93%), but was not calculated for men.

Table 4. Lung Cancers Attributable to Selected Risk Factors in 2008.

	Population Attributable Fraction (%)
	North America	Europe	China
SHS exposure at home
Men	2.10	2.01	4.95
Women	2.19	9.12	18.07
SHS exposure at workplace
Men	6.23	8.50	7.87
Women	3.52	5.76	7.74
SHS exposure at home or at workplace
Men	8.20	10.34	12.43
Women	5.63	14.28	24.11
Tuberculosis	1.14	2.41	12.67
COPD	0.40		0.62
Family history	2.01	1.21	2.93
Residential radon	2.35	5.77	3.72
Pneumonia	0.17
Household use of coala			19.93

^aCalculated for women only

SHS exposure was an important risk factor in all regions. The combined PAF for SHS exposure occurring at home and in the workplace was highest in Chinese women (24.11%), and lowest among North American women (5.63%). Among never-smoking Chinese women, exposure to SHS that occurred in the household was responsible for nearly 20% of the lung cancer burden in that region.

Of the three chronic lung diseases included in our analysis, tuberculosis appeared to contribute to the largest proportion of lung cancer cases. In China, where tuberculosis is prevalent, the PAF for this infection was 12.67%, compared with a PAF of 0.62% for COPD. Even in North America, where prevalence of tuberculosis infection was lowest, the PAF was 1.14%, compared with PAFs of 0.40% and 0.17% for COPD and pneumonia, respectively.

The PAF associated with a family history of lung cancer was relatively low in all regions, ranging from 1.21% in Europe to 2.93% in China. The PAF for residential radon exposure was highest in Europe (5.77%), where the highest levels of indoor radon where reported.

DISCUSSION

This study provides an estimate of the contributions of several known risk factors to the burden of lung cancer in never-smokers in three distinct geographic regions. The proportion of lung cancer that could be attributed to these risk factors was highest in China, and lowest in North America, largely reflecting the different prevalence of exposures in these regions. Known risk factors seem to be important contributors to the high rates of lung cancer in Chinese never-smokers, as they appear to explain a significant proportion of the cases in this region. In comparison, only a relatively small percentage of lung cancer in never-smokers in Europe and North America could be attributed to these risk factors, suggesting there may be several important risk factors in these populations that have yet to be identified.

A large proportion of the lung cancers in never-smoking Chinese women could be attributed to the selected risk factors. In absolute terms, this translates to a large number of preventable cases, since Chinese women have notably high rates of lung cancer¹ and a smoking prevalence of less than 7%.⁶⁰ SHS exposure from the home or the workplace accounted for a quarter of cases, while use of coal for household heating and cooking was found to be responsible for nearly 20% of lung cancer in this population. Combustion of coal causes release of several airborne particles, including carcinogenic PAHs.^{75, 76} PAHs have been found in high concentrations in air of rural Chinese homes that use coal, as well as in the urine of residents.⁷⁷ While we used a relative risk of 2.93 in our calculations, another meta-analysis reported estimates of relative risk exceeding 5.0,⁷⁸ suggesting that the PAF for this risk factor may be even greater than our current estimate. Additionally, while we were unable to calculate a PAF of coal use for males, indoor coal smoke does likely contribute to some percentage of lung cancer in never-smoking Chinese males.

A significant proportion of lung cancers in Chinese never-smokers were also attributable to previous lung diseases. This was driven largely by the prevalence of tuberculosis in this population, which was estimated to be greater than one-third by a WHO-convened panel.⁶⁷ Tuberculosis accounted for a far greater percentage of the lung cancer burden in China than COPD and pneumonia, which were much less prevalent, and were less strongly associated with developing lung cancer. Tuberculosis induces persistent inflammation of the lung tissue, which may lead, in turn, to tumorigenesis.⁷⁹ No PAF for pneumonia could be calculated for Europe or China, as no reliable estimates of incidence exist for those regions. In European populations, a history of pneumonia has been shown to be a predictor of individual lung cancer risk.⁸⁰ Estimates from an Italian population-based case-control study showed that 12% of never-smoking controls reported at least one pneumonia diagnosis during their lifetime, indicating pneumonia may contribute to the burden of lung cancer in Europe. However, the PAF for pneumonia in Europe and China is likely to be modest, as only 0.17% of lung cancers in North America were attributable to this infection.

The PAF of SHS at home tended to be higher in women than in men. While the difference was negligible for North American populations, it was more than four times higher for Chinese women than Chinese men, reflecting the vast gender differences in smoking prevalence in that region.⁶⁰ Our combined estimate of the PAF for SHS was limited to exposures that occurred in the home or in the workplace, not exposures that may have occurred in public places, and therefore may underestimate the proportion of lung cancer due to passive smoking.

The validity of our estimates is limited by the quality of the available data that was used to estimate the prevalence of the risk factors and the relative risks associated with them. While we sought prevalence estimates from large, population-based surveys, these estimates were not always available, and several did not stratify by smoking status. With the exception of data on exposure to SHS, most estimates of prevalence and relative risk were not stratified by sex, so we were unable to calculate PAFs for men and women separately.

Prevalence data from the time period of interest was not always available. National surveys of tobacco use in China were conducted in 1984⁸¹ and 1996;⁶¹ however, published reports of these surveys did not provide estimates for passive smoking in male non-smokers. Instead, we opted to use data from a nationally representative survey of 15,540 subjects that was conducted between 2000-2001.⁶⁰ The more recent study reported estimates of active smoking that were slightly lower for men (60.2% v. 63.0%) and higher for women (6.9% v. 3.8%) than the 1996 survey, so our estimates of passive smoking may not fully capture the extent of SHS exposure during the mid-1990s. We used estimates of the worldwide tuberculosis burden in 1997; however, the risk of lung cancer following tuberculosis has been shown to increase as early as 1 to 5 years following infection, and remains elevated for more than 20 years.⁵⁵

For SHS exposure, we opted to use self-reported measures of exposure to calculate prevalence. However, population-based surveys of non-smokers in the US and Europe which use serum cotinine as a biomarker of SHS exposure have yielded estimates of prevalence which differ from ours. Between 1988-1994, NHANES reported that the proportion of the US population with detectable levels of serum cotinine was 83.9 percent (95% CI: 81.4 – 86.2),⁸² suggesting that we may have underestimated SHS exposure in our analysis. However, use of this biomarker does not allow us to differentiate the source of exposure, and may only reflect exposure to SHS in the 3 to 4 days prior to specimen collection.⁸³

We were unable to calculate PAFs for a small number of known risk factors for lung cancer in never-smokers. Our study does not provide estimates of the PAF for occupational exposures, which likely contribute to rates of lung cancer mortality in all three regions. The lack of available data regarding the prevalence of occupational exposures and their associated risks in never-smokers underscores the need for further research in this area, though the PAF for occupational exposures may decline in coming years, due to the enactment of workplace safety legislation and the use of personal protective equipment at worksites. We were also unable to calculate an estimate of the PAF for exposure to cooking fumes in Chinese women, owing to a lack of data on the prevalence of high-temperature frying. Similarly, no estimate of the PAF for pneumonia could be calculated in China or Europe, as no prevalence data for the relevant time period was available.

We chose not to calculate a PAF for the role of diet, as evidence in this area has been largely inconsistent. An inverse association between fruit and vegetable consumption and risk of lung cancer in never-smokers has been frequently reported. However, other findings, including several from large cohort studies, have been null. No valid estimate of the lung cancer burden attributable to low fruit and vegetable consumption can be calculated until an inverse relationship is more conclusively proven. Similarly, we did not calculate a PAF for outdoor air pollution. Until further studies are conducted using individual-level data and standardized definitions of exposure, no PAF for this risk factor can be reliably estimated.

We did not calculate separate PAFs for lung cancer incidence and mortality. It is possible that some risk factors may be more strongly associated with lung cancer mortality than with lung cancer incidence. However, because lung cancer is a highly fatal disease, with five-year survival rates ranging from about 9% to 15% globally,⁸⁴ the PAFs associated with lung cancer incidence and mortality are unlikely to be very different.

Our study provides an estimate of the burden of lung cancer in 2008 attributable to several known risk factors for lung cancer in never-smokers in three distinct geographical regions. Passive smoking, indoor radon, previous lung disease, family history, and household coal smoke were shown to make substantial contributions to the worldwide burden of lung cancer in lifetime non-smokers. As efforts are undertaken to control the tobacco epidemic, particularly in developing countries, the epidemiology of lung cancer in never-smokers is sure to evolve, warranting further research in this area.

Abbreviations

SHS

secondhand smoke

PAF

population attributable fraction

RR

relative risk

COPD

chronic obstructive pulmonary disease

PAHs

polycyclic aromatic hydrocarbons

IARC

International Agency for Research on Cancer

NHANES

National Health and Nutrition Examination Survey

CPS-II

Cancer Prevention II study

EPIC

European Prospective Investigation into Cancer and Nutrition study