Lung Cancer Etiology: Title and subTitle BreakIndependent and Joint Effects of Genetics, Tobacco, and Arsenic

Lung cancer is the number one neoplasm in the world, both in terms of incidence and mortality.¹ The incidence of lung cancer differs by geographic area, sex, age, and over time,^{1 - 2} reflecting the effect of the underlying distribution and trend in use of its principal determinant, tobacco smoking. Although 80% to 90% of lung cancer cases occur in current or past tobacco smokers, only a small fraction of smokers (1%-15%) develop lung cancer,² depending on how much and how long an individual has smoked and the presence of other causes of lung cancer. Clearly, because all lung cancers do not occur in smokers and the vast majority of smokers do not develop lung cancer, other etiological factors can independently (in the absence of smoking) or jointly (in conjunction with smoking) cause lung cancer, beyond the purely stochastic nature of the disease process. These factors include genetics (measured as family history),^{3 - 4} arsenic exposure,^{5 - 8} radiation exposure, and other environmental carcinogens.² Although genetic factors probably contribute in all populations, the contribution of other factors is population-specific. For example, in all areas of the world lung cancer shows a modest level of familial aggregation,^{2 ,9 - 12} whereas only in specific environmental, occupational, and therapeutic settings do arsenic and ionizing radiation contribute to lung cancer etiology.²

Given that tobacco smoking is the overwhelming cause of lung cancer and is prevalent throughout the world, examination of the causes of lung cancer other than tobacco smoking should take tobacco into consideration. In this issue of JAMA, the articles by Chen and colleagues¹³ and Jonsson and colleagues¹⁴ attempt to examine the contribution of arsenic exposure and family history, respectively, in lung cancer while also taking tobacco smoking into consideration. In doing so, Jonsson et al¹⁴ are primarily concerned with smoking as a confounder (whether familial aggregation of smoking can explain the familial aggregation of lung cancer), whereas Chen et al¹³ are primarily concerned with whether there is a synergistic effect of smoking and arsenic.

Although major genes similar to those for single-gene disorders have yet to be identified for lung cancer, an aggregation of lung cancer in families has long been recognized.^{3 - 4 ,15} Between 3% and 6% of all patients with lung cancer have a positive family history of lung cancer.^{2 ,9 - 12} Familial aggregation in the context of lung cancer may be due to shared exposure to tobacco and other lung carcinogens or shared preventive factors within families; major genes (eg, chromosome 6p locus)¹⁶ ; modifier genes that enhance the carcinogenic potential of tobacco smoking and other occupational and environmental lung carcinogens, including arsenic; addiction genes (eg, nicotine addiction genes) that may predispose individuals to initiate or continue tobacco smoking^{17 - 18} ; and chance. Both shared genes and shared environment can produce detectable familial aggregation in lung cancer (Figure). If familial aggregation of lung cancer can be established as an independent phenomenon from familial aggregation of smoking or other lung carcinogens, a genetic etiology for at least a subset of lung cancer cases aggregating within families could be inferred. However, separating the effect of family history from the effect of shared tobacco or other environmental carcinogens is challenging, especially among populations in which such exposures are common.

In certain exposure settings within specific populations, other exposures become prominent or at least as important as tobacco in lung cancer incidence. For example, millions of people have been chronically exposed to arsenic from drinking water in Taiwan,¹⁹ Argentina,²⁰ and Chile,²¹ parts of United States,²² and most recently Bangladesh,²³ West Bengal, India,²⁴ and parts of China.²⁵ In these settings in which tobacco smoking is also prevalent, the issue of separating the effects of tobacco from arsenic effects or examining their synergisms also becomes an important consideration.

The studies by Chen et al¹³ and Jonsson et al¹⁴ attempt to disentangle the effects of smoking and other risk factors (arsenic and family history) but in quite different ways. Both investigations can be loosely classified as cohort studies, although neither perfectly follows the traditional mold. Chen et al¹³ identified individuals based on arsenic exposure with a follow-up period of 8 years to ascertain their lung cancer experience. However, the authors combined several groups of individuals with very different exposure characteristics in 1 analysis: household members who drank water from their own contaminated wells until 1990s, as well as community residents and participants of an earlier case-control study of blackfoot disease who drank water from shared contaminated community wells until the early 1960s. Although lung cancer was assessed at the individual level for all participants, the arsenic exposure for the first group was assessed at a village level, making the study to some extent similar to a “semi-individual” study.²⁶

The family history study from Jonsson et al¹⁴ began with complete ascertainment of all lung cancer probands from January 1, 1955, to February 28, 2002, linked them to the Icelandic genealogical database to identify various types of relatives who are at risk, and determined how many of these relatives also appeared as cases in the registry. By comparing the risk of lung cancer in various types of relatives with the general population rates, the risk ratio (RR) for family history was estimated. To address ascertainment bias, individuals were counted multiple times each time they appeared in the genealogic database as a relative of a different proband. This is a standard technique to correct for single ascertainment, but its use in this setting of complete ascertainment could lead to bias. The analysis of smoking within families was similar, except that the ascertainment was based on a separate random sample of smokers. This situation ismore like the usual “single-ascertainment” model, in which the probability of ascertainment is small, leading to multiple-smoker families being overrepresented, as they have more opportunities to be identified. Because the smoking status of all relatives was not available, only relatives who also appear in the sample were ascertained, an event that would occur with probability equal to the square of the ascertainment probability under the null hypothesis of no familial aggregation of smoking.

Because smoking and family history information was not available on the same individuals, Jonsson et al¹⁴ could not adjust the estimate of familial aggregation of lung cancer for smoking. Instead, they relied on a mathematical argument to show that the familial aggregation of lung cancer cannot be entirely explained by the familial correlation in smoking. This argument follows from the well-known epidemiological principle that for a potential confounder to explain a relative risk of size RR, the association of that factor with both exposure and disease must be at least as large as RR.²⁷ Here the argument is somewhat more complex, because 4 factors are involved: lung cancer in the proband and relatives, and smoking in the probands and relatives. The observation that the familial RR of lung cancer is larger than the familial RR for smoking would strongly support the claim that the familial aggregation of lung cancer cannot be completely explained by the familial aggregation of smoking, were it not for the possibility that the 2 RRs could be differently distorted by their different ascertainment schemes. Nevertheless, earlier analyses of this question²⁸ support this interpretation.

On the other hand, Chen et al¹³ had the ability more directly to take smoking into consideration in their analyses since information about arsenic exposure and smoking was obtained from the same person. Rather than evaluating whether the arsenic effect was confounded by smoking, the investigators evaluated their combined effects. Their assessment of combined effects showed a more than additive effect, similar to previous analyses of smoking and arsenic.^{29 - 30} It is unclear whether the observed joint effect is more or less than multiplicative, which could be interpreted as evidence that these effects act at different stages of a multistage process of carcinogenesis.^{31 - 35} The biological implication of this finding thus remains unanswered, although the synergistic effect has potential public policy implications.

Both of these studies have major strengths. The study by Jonsson et al¹⁴ is an example of how a familial aggregation study, even in the postgenomic era, may be useful if a unique population is studied. The authors’ observation of an increased risk among first-, second-, and even third-degree relatives and also a stronger effect among blood relatives (vs spouses), in younger individuals (vs older persons), and for lung adenocarcinoma (vs other histological subtypes) suggests a hereditary component underlying the familial aggregation. By using a total population coverage, the study avoids recall biases to which standard familial aggregation studies are often susceptible. Chen et al¹³ used cancer registry data, which other countries with large arsenic-exposed populations currently lack. In fact, other than the studies in occupational settings, this is the first study to our knowledge to investigate the joint effects of environmental arsenic ingestion through drinking water and tobacco smoking in relation to lung cancer incidence using a prospective cohort design.

Several aspects of lung cancer etiology with respect to family history, tobacco smoking, and other environmental carcinogens become apparent from these 2 studies. Numerous epidemiological studies have established and characterized the relationship between tobacco smoking and lung cancer.^{2 ,36 - 37} The pattern of the effects of tobacco smoking on lung cancer by age groups and histological subtypes provided initial clues to the possible genetic etiology of subsets of lung cancer, and the Jonsson et al¹⁴ study exploited these clues. Future studies can exploit these and other aspects (Figure) for isolating the effects of heritable factors. For example, early-onset lung adenocarcinoma in nonsmokers causing familial aggregation may be the logical target for studies of major gene discovery. Nicotine addiction genes exert their effects secondary to the tobacco effect and so would not contribute to familial aggregation in nonsmoking families. Similarly, by virtue of their low penetrance and indirect roles in lung cancer, modifier genes are unlikely to contribute much to familial aggregation in nonsmoking families.

An effective strategy for separating the effects of family history from the environment would be to exploit these aspects in designing future studies. To minimize the mathematical complexities in separating different effects in a combined analysis, one simple strategy would be to sample study individuals based on tobacco smoking and family history information. For example, a study among nonsmokers would effectively eliminate the influence of active tobacco smoking.³⁸ For isolating the effect of non-tobacco environmental carcinogens, it is often useful to exploit unique occupational and environmental exposure settings. In the study by Chen et al¹³ in Taiwan, the arsenic exposure had ceased many years earlier making it difficult to assess synergistic effects of concurrent exposure to arsenic and tobacco smoking.³⁹ Such opportunities now exist in populations currently exposed to arsenic in Bangladesh,²³ West Bengal, India,²⁴ and China.²⁵

Irrespective of the roles of familial aggregation and environmental exposures to arsenic or other carcinogens, lung cancer is primarily caused by tobacco smoking—an exposure that is largely preventable. If nicotine addiction genes or modifier genes play roles in subsets of patients with lung cancer, such cases can be prevented by preventing tobacco smoking. Although there are no addiction genes for arsenic or other environmental exposures, modifier genes could modulate the effects of these nontobacco carcinogens.^{40 - 41} Such genes, in combination with major genes, could lead to familial aggregation. Innovative epidemiological studies to detect and separate these effects, taking the lead from studies like those of Jonsson et al¹⁴ and Chen et al,¹³ need to be designed in the future. Even for individuals with such a familial risk, or for those who are already chronically exposed to arsenic⁴² or other lung carcinogens, avoiding tobacco smoking remains the most feasible option for reducing lung cancer risk.