Familial Cancer History and Lung Cancer Risk in United States Nonsmoking Men and Women¹

A number of studies find that cases with lung cancer report having a significantly greater number of first-degree relatives with lung cancer compared to controls without lung cancer (1, 2, 3, 4, 5, 6, 7, 8, 9). Many studies also find an excess of total cancers in first-degree relatives of lung cancer cases compared to controls (2, 4, 10, 11, 12). The rather consistent finding of familial risk of lung cancer could be due to genetic factors or shared environmental factors. As most studies of familial lung cancer are done in cases who smoked cigarettes, familial aggregation of smoking habits could potentially explain the excess number of lung cancers, and possibly total cancers, in case families relative to control families. Some investigators have attempted to examine the smoking habits of family members–Tokuhata and Lilienfeld (7) conducted a careful analysis of familial risk of lung cancer while taking into account smoking habits of the family members. Their results indicated that the excess risk of lung cancer in case relatives compared to control relatives occurred irrespective of the relative’s smoking history. Also, the familial risk estimate in their study was greater for nonsmokers than for smokers, suggesting that the risk was not simply due to shared smoking habits in families of lung cancer patients.

Another approach for estimating familial risk for lung cancer while minimizing the potential impact of shared tobacco habits in families is to do case-control studies in nonsmokers. Given the difficulty in assembling a large group of lung cancer cases who are nonsmokers, only a few such studies are available. The largest such study is that of Wu et al. (13), a multicenter study of lung cancer in 646 female lung cancer patients in the United States (lifetime nonsmokers) and 1252 population controls (also lifetime nonsmokers). Lung cancer occurred nonsignificantly more frequently in first-degree relatives of lung cancer patients than in comparable relatives of population controls (adjusted OR³, 1.29; 95% CI, 0.9–1.9). The familial risk was slightly greater for cases with adenocarcinoma of the lung (adjusted OR, 1.50; 95% CI, 1.0–2.2). Schwartz et al. (1) also conducted a population-based study of familial lung cancer risk in 257 nonsmoking lung cancer cases (72% female) and 277 nonsmoking controls from Detroit. Nonsmoking cases were more likely (nonsignificantly) than nonsmoking controls to have a first-degree relative with lung cancer (OR, 1.4; 95% CI, 0.8–2.5). The risk was greater for female cases than male cases (ORs, 1.7 and 0.5, respectively) and slightly greater for cases with adenocarcinomas (OR, 1.7; 95% CI, 0.9–3.3). Wang et al. (12) reported on the family history of any cancer in 135 lifetime nonsmoking female lung cancer cases and an equal number of matched controls in China. Cases were significantly more likely than controls to report a positive family history of any cancer (including lung cancer; OR, 2.29; 95% CI, 1.01–5.17).

We completed a population-based case-control study of lung cancer in a relatively large number of nonsmoking men and women in the United States; some of the findings from this study focusing on other risk factors have been published previously (14, 15, 16). This study provided an opportunity to examine the hypothesis that a family history of lung and total cancers is associated with an increased risk of lung cancer in nonsmoking men and women using a case-control approach. Moreover, as we also collected data on all cancers occurring in first-degree relatives of both cases and controls and age and smoking status of the relatives, we were able to more fully evaluate a family history of other cancers in first-degree relatives as a risk factor using cohort techniques.

A population-based, individually matched case-control study of lung cancer in nonsmokers was conducted in New York State from 1982 to 1984. The methods for the data collection have been described previously (14, 15, 16). The procedures used were approved by the Institutional Review Board of the New York State Department of Health. Informed consent was obtained from all participants.

The study area was comprised of eight standard metropolitan statistical areas (23 counties) in Upstate New York. A special system for the rapid ascertainment of lung cancer cases was established in this region so that patients could be identified and interviewed as soon after diagnosis as possible. Patients were typically identified via the Medical Records Department, the Pathology Department, or the Tumor Registry at the institutions in this region.

To be included as a case in the study, a patient had to reside in the 23-county region, be between 20 and 80 years of age, never have smoked >100 cigarettes (nonsmokers), or have smoked at some time but not have smoked >100 cigarettes in the 10 years before diagnosis (former smokers). Cases also must have been given a diagnosis of primary lung cancer between July 1, 1982, and December 31, 1984. Initial diagnoses were confirmed by reexamination of pathological specimens and clinical records. Slides or blocks of tissues were available for all but five of the cases. All materials were reviewed by investigators who were masked with respect to the patient’s initial diagnosis, smoking history, and other risk factors. Interviews were conducted with 76% of the eligible patients or their closest available relatives or friends (surrogates).

One population control was selected for each case. Controls were selected randomly from the New York State Department of Motor Vehicles’ file of licensed drivers. For each case, six potential controls were identified on the basis of age (±5 years), sex, and county of residence. Potential controls were called and screened for smoking status. The first potential control who was found to match the case on smoking status and who agreed to participate in the study was included. On the average, two potential controls had to be contacted to obtain a control who matched the case on smoking status and who agreed to participate in the study.

Face-to-face interviews were conducted by the field staff in the homes of all cases and controls. Information was collected using a precoded questionnaire that took ≈1 h to complete. Cases and controls were interviewed in the same manner with the same questionnaire. No attempt was made to mask interviewers to case-control status due to the often obvious evidence of disease among lung cancer patients.

Despite efforts to rapidly identify and interview all cases, surrogate interviews were needed for 141/437 (32%) of all cases. When a surrogate respondent was used for the case, surrogate respondents were also used for the matched control (with the exception of nine matched controls where surrogate controls interviews were not matched for surrogate case interviews). Thus, cases and controls were matched for sex, age, smoking history, county of residence, and type of interview.

Cases and controls were asked to report the health histories of their immediate biological family, i.e., their parents, brothers, sisters, or biological children. More specifically, for each relative, the country of birth, year of birth, year of death, smoking status (yes, no), presence or absence of cancer (yes, no), type of cancer, and year of diagnosis were queried.

Descriptive statistics were first conducted to characterize the study population. Cancers were grouped together for analyses as follows (International Classification of Diseases 9 codes): lung 162–163; aerodigestive tract 140–149 and 160–165 (includes lung); digestive 150–159; female breast 174; and prostate 185. ORs for family history of total cancers and selected other common cancers were then estimated by conditional logistic regression as described by Holford et al. (17). This method is based on the linear logistic model described by Cornfield et al. (18) and takes into account pairwise matching in case-control studies. For case-control analyses of cancers occurring in brothers, sisters, and children of cases versus controls, conditional logistic regression resulted in an unacceptable loss of data and unreliable risk estimates due to one member of the matched pair being uninformative (i.e., not having a brother, a sister, or child). In these instances, unconditional logistic regression was used to calculate risk estimates, with adjustment for case/control age and number of relatives (brothers, sisters, children).

We then performed a cohort analysis on all of the relatives of the cases and controls. That is, cohorts of fathers, mothers, sisters, and brothers were constructed, including variables for age of the cohort member and smoking status of the cohort member. A children’s cohort was not constructed due to an insufficient number of end points for analysis. An indicator variable was included to indicate if the relative was related to a case or control. Logistic models were then performed to predict the occurrence of various cancers in the cohort of interest. We also performed Cox proportional hazards modeling on the cohort, but due to some missing data on the time until failure variables in several of the relatives, we chose to present the cohort analysis rather than a survival analysis of the relatives.

The characteristics of the 437 case-control pairs used in this analysis are summarized in Table 1. Approximately one-half of the cases were females, and 45% of the cases were never smokers. Adenocarcinoma was the predominant histological subtype of the cases, accounting for 51% of the cases. The reported prevalence of smoking in case relatives was similar to that in control relatives (none of the proportions differed significantly). Cases had more brothers (P = 0.001) and sisters (P = 0.055) and slightly fewer children (P = 0.17) than their matched controls. Mothers (P = 0.04), brothers (P = 0.01), and sisters (P = 0.0004) of cases were younger than the corresponding relatives of controls.

The associations between paternal and maternal history of cancer and lung cancer risk based on the case-control approach are shown in Table 2. Cases were more likely than their matched controls to report a paternal history of any cancer (P = 0.01) and aerodigestive tract cancer (P = 0.009). Cases and controls reported a similar maternal history of cancer, with the exception of breast cancer, which reportedly occurred two times more frequently in mothers of cases than in mothers of matched controls (P = 0.049).

The associations between history of cancer in sisters, brothers, and children and lung cancer risk based on the case-control analyses are shown in Table 3. Cases were significantly more likely than their matched controls to report having had sisters with any cancer (P = 0.01). The odds of having a sister with lung cancer (OR, 4.14), aerodigestive tract cancer (OR, 3.50), and breast cancer (OR, 2.07) were also notably increased, although of borderline statistical significance. Cases were significantly more likely than controls to report having had brothers with any cancer (P = 0.03). Only 17 children of cases had cancer as compared to 12 children of controls. The breakdown of the 29 cancers in offspring was unremarkable.

Combining all first-degree relatives together (data not shown), a positive family history of any cancer (P = 0.0001), lung cancer (P = 0.0098), aerodigestive tract cancer (P = 0.003), and breast cancer (P = 0.007) was statistically significantly associated with an approximate doubling of the risk of lung cancer. The risk associated with a family history of any cancer, lung cancer, and aerodigestive tract cancer was highest for cases with adenocarcinoma of the lung, whereas the risk associated with a family history of breast cancer was greatest for cases with histologies other than squamous cell carcinoma or adenocarcinoma (data not shown). A positive family history of any cancer, lung cancer, aerodigestive tract cancer, or breast cancer remained a statistically significant risk factor for lung cancer when proxy interviews were excluded from the analysis.

Cohort analyses of relatives of cases and controls are shown in Tables 4 and 5. Fathers of cases were at greater risk of developing any cancer (RR = 1.90; P = 0.002) and aerodigestive tract cancers (RR = 2.77; P = 0.006) compared to fathers of controls. Mothers of cases were at greater risk of developing any cancer (RR = 1.42; P = 0.07) and breast cancer (RR = 2.52, P = 0.01) compared to mothers of controls. The risk estimate for sisters to have any cancer was lower in the cohort analysis than in the case-control analysis (RR = 1.25 versus OR = 1.66) and no longer statistically significant, which was most likely due to the fact that the cohort analysis takes into account the fact that cases had more sisters than their matched controls. However, the RR for lung cancer (RR = 3.42) and aerodigestive tract cancer (RR = 2.91) for case sisters as compared to control sisters remained quite high in the cohort analysis, although it was not significant. Similarly, the risk estimate for brothers to have any cancer was slightly lower in the cohort analysis than in the case-control analysis (RR = 1.45, P = 0.056 versus OR = 1.58, P = 0.03), most likely due to the fact that the cohort analysis takes into account the fact that cases had more brothers than their matched controls.

This study is one of the largest studies to date of family cancer history as a risk factor for lung cancer in nonsmoking men and women. Higher risk estimates for total cancers were observed in case relatives compared to control relatives for fathers (P = 0.002), mothers (P = 0.07), and brothers (P = 0.056). This excess risk does not appear to be due to a greater prevalence of smoking in first-degree relatives of cases compared to controls because we adjusted the estimates for the smoking status of the relatives, although we did not collect data on amount smoked in relatives.

Our risk estimates are in the range reported by other investigators, with ORs for a family history of lung cancer in a first-degree relative in the literature ranging from 1.3 to 7.4 (Refs. 13 and 2, respectively), and ORs for a family history of any cancer in a first degree relative ranging from 1.3 to 2.9 (Refs. 3 and 2, respectively). A family history of any cancer, lung cancer, and aerodigestive tract cancer appeared to confer the greatest risk for cases with adenocarcinoma as compared to squamous cell carcinoma and other histologies in this study (data not shown). A positive family history has been reported to be more strongly associated with adenocarcinoma of the lung than with squamous/oat cell cancers in studies involving both smokers (4, 8) and nonsmokers (1, 13).

A family history of digestive cancers and prostate cancers was not associated consistently with an increased risk of lung cancer in this population, arguing against a generalized recall bias in reporting of family history by cases relative to controls. Also, the relationships with family history seen in this data set remained statistically significant when proxy interviews were excluded, adding further credibility to the findings.

The finding that a family history of breast cancer was significantly associated with an increased risk of lung cancer in nonsmokers is intriguing. The association was significant for both mothers and sisters of cases in the case-control analysis, although the association was not significant in the cohort analysis of sisters. We are unaware of other studies demonstrating such a relationship, but most studies of family history as a risk factor for lung cancer generally only inquired about total cancers and/or lung cancer in relatives. Wu et al. (13) reported that lifetime nonsmoking women in the United States with lung cancer had a 10% excess of breast cancer in first-degree relatives compared to controls (OR, 1.10; 95% CI, 0.8–1.6). The risk increased to 1.17 among cases with adenocarcinoma, but this was still not significant (95% CI, 0.8–1.7). A mutational analysis of tissue blocks from nonsmokers with lung cancer and with a positive family history of breast cancer in first-degree relatives might be of interest to help clarify this potential association.

The strength of this study lies in the relatively large number of nonsmoking cases with lung cancer for analysis, particularly nonsmoking men. Also, the population-based study design is an important strength. Detailed tobacco exposure histories were obtained for cases and controls, and cases were carefully classified by histological subtype. Smoking data on relatives were also obtained. However, the exposure of interest, family history of cancer, was based upon self report; no effort was made to verify family history in any of the subjects. Thus, recall bias is of concern; however, as previously stated, the increased risks associated with only some of these cancers but not all argues against a strong influence of recall bias in these data.

In conclusion, this population-based case-control study demonstrates an excess of certain cancers, particularly lung cancer, aerodigestive tract cancers, and female breast cancer in first-degree relatives of nonsmokers with lung cancer compared to nonsmokers without lung cancer.