The 93 families that have been studied include 489 persons affected with lung cancer of whom 45 are unrelated (marrying-in to the pedigree) and 444 are related to other affected family members, and informative for linkage analysis. From these families we have accrued 1,156 blood samples, 24 buccal cell samples, 58 sputum samples, and 274 archival blocks containing normal tissue. Archival tumor blocks of lung cancer-affected subjects have been collected from 186 persons and 88 blocks from other tissues. When other sources of DNA were not available, we used archival tissue blocks for genotyping. Where possible, since we are interested in studying the coinheritance of lung cancer with genetic markers present in the germline, we have performed analyses on tumor blocks from non lung cancer specimens. Otherwise, when lung cancers have been studied, one of us (AG) has retrieved normal tissue from the tumor margins. Of the 93 families, three are African-American and 1 family has mixed racial composition (African-American, Creole, and Caucasian); the remaining 89 families are Caucasian.
Lung cancer affected individuals are 63.4% male, 81.8% deceased at the time of data collection, and 86.3% ever smokers, with a median value of 50 pack years. For the unaffected individuals who reported cigarette smoking history data, 73.2% were ever smokers with median pack year value of 26, a generally higher level of smokers than in the general population (20
). However, because these persons come from families with a strong history of lung cancer among a smoking relative, and smoking aggregates in families, they are more likely to be smokers. Smoking histories for deceased individuals were obtained from surrogates. Numerous studies have reported that surrogate reported data are about 90%-95% accurate for smoking status, but usually underestimate pack years (21
). Cancer status has been verified with medical records, cancer registry data or death certificates on 417 (85.3%) of the 489 lung affected persons. Pathology reports were obtained whenever possible, i.e., the tissue sample was obtained for diagnosis, medical records could be located and patient or family had signed a medical record release. The distribution of cell type of lung cancer was similar to that reported in the past for the general population (26
). In 59 families studied by Bailey-Wilson et al. (8
) of 224 lung cancer- affected persons on whom we have pathology reports, 75 (33.5%) had adenocarcinoma, 69 (30.8%) had squamous cell carcinoma, and 22 (9.85%) had small cell carcinoma. Seven families presented with predominantly either adenocarcinoma (N=3) or squamous cell carcinoma (N=4).
Two pedigree characteristics that affect informativeness for linkage analysis are the number of affected persons in the family and number of generations with affected persons (). For assessing informativeness, we count only affected persons who have at least a third-degree relationship to another affected. In bilineal families, we count only those in the predominant lineage with lung cancer when both parents are lung cancer affected (), Because at least some of the families with only 2 and 3 affected relatives may not segregate effects from a major susceptibility factor but may rather reflect chance clustering of lung cancer, we have separated this group into subset 1. Similarly, families that include 5 or more affected relatives in 2 or more generations are most likely to segregate a dominantly inherited locus that increases susceptibility and these families have been denoted as subset 2. Families that include 4 or more relatives in a sibship are denoted subset 3. The median number of affected persons per family is 5. In the 93 families, there are 66 families with 4 or more affected and 57 of these families have affected persons in more than 1 generation (subset 4, results in supplementary figures
). Of the 50 families with 5 or more affected persons, 47 have affected persons in multiple generations. Linkage analyses of chromosome 6 show that families with 5 or more affected persons in multiple generations exhibited linkage to chromosome 6q.
Number of lung cancer affected individuals in families, having at least a third-degree relationship to each other
Linkage and haplotype analyses of risk
Maximal heterogeneity LOD (HLOD) scores from genomewide linkage analyses are presented in . Results from linkage analyses are presented in and Supplementary Figure 1
. In we present the results of linkage analysis on chromosome 6, while the supplementary figure 1
provides results for other chromosomes that yielded an HLOD score > 1.0 in any subset. The proportions of families estimated from heterogeneity LOD score analysis was 0.53 for the entire dataset and for subsets 1-3 the heterogeneity estimates were 0.74, 1.0, 0.35 respectively. Of the entire set of 93 families, 10 had a LOD score over 0.3 on chromosome 6q at 158 cM.
Maximum heterogeneity LOD scores over 1.0 in linkage analysis of any subset
Figure 1 Heterogeneity LOD scores from analysis of chromosome 6 for 93 families selected to include multiple relatives with lung cancer. Subset 1 includes families with 2 or 3 individuals affected by lung cancer, Subset 2 includes families with 5 or more individuals (more ...)
Further analysis of the impact of smoking on risk for cancer was carried out as indicated above by first defining carrier status and then by performing Cox regression modeling treating the intensity of smoking as an ordinal variable. There were 292 individuals who carried a risk haplotype, 441 who were in families segregating a risk haplotype who were noncarriers of that haplotype, and 2248 individuals for whom carrier status could not be derived and were classified as unknown carrier status. results from Kaplan-Meier analysis showing that among carriers the overall risk for lung cancer was higher than among noncarriers. There is also significantly higher risk for lung cancer among ever compared with never smokers, as assessed by the log-rank test. However, among smoking carriers there was no evidence for increasing risk with an increasing exposure level to cigarette smoke (p=0.36). On the other hand, among noncarriers (p=0.085) and individuals with unknown carrier status (p=0.0008) a more usual dose-effect relationship between smoking and lung cancer risk is observed. These findings suggest that any level of tobacco exposure increases risk among those with inherited lung cancer susceptibility, suggesting that such individuals should be heavily targeted for smoking prevention and monitored by early detection procedures. Comparing with the risk in never smokers (), carriers had higher hazard ratios of 3.44 (95% CI= [1.40,8.48], p=0.007) for light smokers, 4.91 (95% CI = [2.46, 9.8], p <0.0001) for moderate smokers and 5.18 (95% CI =[2.81, 9.56], p<0.0001) for heavy smokers. Among noncarriers, no events occurred in never smokers, so that hazards ratios could not be estimated. For unknown carrier status there was a much stronger effect of smoking, with all groups having highly significant differences from never smokers (p<0.0001). For those light smokers with unknown carrier status, the hazards ratio compared to never smokers was 4.25 (95% CI = [2.11, 8.54]), for moderate smokers the hazards ratio was 9.77 (95% CI = [5.9, 16.20]) and for heavy smokers the hazards ratio was 11.89 (95% CI = 7.59, 18.61). When the analyses were adjusted for excess selection for affected individuals (), we found very little trend in carriers, with the hazards ratios in carriers being 2.67 (95% CI=[1.22, 5.86}), 2.34 (95% CI = [1.37, 3.98]), and 2.75 (95% CI = 1.74, 4.37]) in light, moderate, and heavy smokers respectively, while for those with unknown carrier status the hazard ratios were 3.00 (95% CI = [1.64, 5.88]), 5.20 (95% CI = [3.67, 7.58]) and 7.32 (95% CI = [5.28, 10.14]) respectively for light, moderate and heavy smokers.
Figure 2 Time to lung cancer among carriers (left panel), noncarriers (middle panel), and individuals with unknown carrier status (right panel). Smoking strata are shown with the black line reserved for nonsmokers, the red line for light smokers (1-19 pack years), (more ...)
Table 4a. Comparsion of risks for light, moderate, and heavy smokers versus nonsmokers stratified by carrier status, without adjustment for sampling through multiple affected relatives.
An alternative approach to evaluating risk compares risk among carrier groups, conditioning on smoking behavior (supplemental table 2a
). Using individuals with unknown carrier status as the referrent, for never smokers the hazards ratio for for non-carriers was 0 (no events, p=0.99) and 4.71 for carriers (95% CI = 2.35, 9.43], p<0.0001). For light smokers the hazards ratio was 1.08 for noncarriers (95% CI = 0.31, 3.83], p=0.90) and 4.34 for carriers (95% CI = [1.76, 10.7], p=0.0001). For moderate smokers the hazards ratios were 0.83 for noncarriers (95% CI = [0.41, 1.165],p=0.59) and 2.51 for carriers (95% CI = [1.53, 4.13], p=0.0003). For heavy smokers the hazards ratio was 0.83 for noncarriers (95% CI = 0.54, 1.29], p=0.41) and 2.21 for carriers (95% CI = [1.65, 2.97], p<0.0001). Thus, comparing noncarriers and those with no known haplotype (unknown carrier status) there is no significant difference in risk between these two groups according to smoking behavior. However, among those who are carriers the increased risk is most prominent in never smokers. However, as shown in , any degree of smoking confers a marked increased in risk beyond this baseline. Decreasing hazards ratios according to increasing smoking reflect the higher risks among the noncarriers of risk haplotypes according to increased effects from smoking, but comparable risks for lung cancer among carriers who have any degree of smoking exposure.