Association between lung cancer somatic mutations and occupational exposure in never-smokers

Discussion

We demonstrated the mutational pattern of NSCLCs in never-smokers to be slightly associated with occupational exposure. More specifically, exposure to asbestos appeared to be correlated with a lower frequency of EGFR mutations (20% versus 44% in the unexposed) and a higher frequency of HER2/ERBB2 mutations. Exposure to primary occupational carcinogens suggests a particular distribution of KRAS, BRAF, PI3K and ALK alterations.

Only a few studies have dealt with the molecular profile of lung adenocarcinomas in relation to asbestos exposure. Andujar et al. [13] performed a comparative study of 50 cases of asbestos-exposed NSCLC and 50 unexposed cases. Overall, the authors observed a higher EGFR mutation rate in unexposed patients (12% versus 4% in exposed patients; p=ns). In never-smokers, the respective proportions were 50% and 14% (p=ns). In addition, KRAS mutations were found in 10% and 16% of exposed and unexposed cases, respectively (0% and 17% for never-smokers). In a cohort of 510 Finnish NSCLC patients, Mäki-Nevala et al. [11] observed no differences in terms of EGFR mutations between 46 asbestos-exposed and 198 unexposed patients (10.6% versus 9.6%, respectively). The same team performed exome sequencing of lung adenocarcinoma (n=26, nine of which were asbestos-exposed) [12]. In the adenocarcinoma cases, 42% exhibited a KRAS mutation in both exposed and unexposed patients, with no activating EGFR mutation observed. Unlike our findings, only two BRAF mutations were found in that study, both of which were observed in unexposed patients. In a recent paper, the same team reported no association between asbestos exposure and mutation patterns (for EGFR, KRAS, BRAF, ALK, HER2, PI3K and others like, MET, TP53, PTEN or NRAS) in 425 Finnish NSCLC patients, including 8.9% never-smokers and 29 subjects (6.8%) exposed to asbestos [23]. In another study conducted in the USA, 84 male patients (95% of whom were smokers) with NSCLC, who were also subjected to asbestos exposure, exhibited a higher frequency of KRAS mutations (crude OR 4.8, 95% CI 1.5–15.4) [24]. No association was established however, between KRAS mutations and asbestos exposure in 105 NSCLC patients (all of whom were smokers) [25]. The ALK alterations were investigated in relation to asbestos exposure in only one malignant mesothelioma (MM) cohort (n=63), with none of the samples exhibiting ALK alteration [26]. Overall, these data highlight the scarcity of literature regarding mutational patterns of driver oncogenes in relation to occupational exposure. Whereas the study of Andujar et al. [13] reported similar results to the present study concerning EGFR mutations and asbestos exposure, other articles reported no such correlation in a similar setting. These studies nonetheless, used relatively small samples, and were not conducted specifically in never-smokers. For these reasons, active smoking status is possibly too strong a factor to determine differences in EGFR mutation frequency.

Our study has some limitations. The main limitation is that of the very low number of subjects in some subgroups. The number of patients exposed to certain occupational carcinogens was therefore very small (ranging from 27 exposed to PAH, to six each exposed to paints and chrome). In addition, the number of alterations observed in this cohort proved to be very small for some driver genes (ranging from 20 ALK alterations to four PI3K mutations, with the exclusion of EGFR mutations). Altogether, we obtained some very small subgroups, such as four EGFR mutations in asbestos-exposed patients and three BRAF wild type in DEF-exposed patients (see table 2). The interpretation of patterns by carcinogen (or biomarker) must therefore be done with caution. Moreover, the majority of patients (21 out of 40, 53%) had been affected by simultaneous exposure to at least two occupational carcinogens (ranging from two to five, table 1 and figure 2), and it is therefore difficult to differentiate the role of each occupational exposure to carcinogens. Although nonsignificant, multivariate models adjusted for age, gender, BMI and passive smoking did not affect the trends observed in our primary results. Thus, our findings seem to be independent of these adjusting factors. Another limitation arose from the definition of occupational exposure that was based exclusively on a self-administered questionnaire, without biological evidence, and without measured metrological data of such exposure. Nonetheless, mineralogical analyses are not required for MM management [27, 28]. A third limitation of the current analysis is the lack of consideration of the exposure to environmental radon, which is a leading risk factor of lung cancer, and which might be associated with EGFR and ALK molecular patterns [29]. Another limitation arose from the heterogeneity of assays and techniques used for biomarker assessment. Indeed, each platform is itself able to determine which panel (although a minimal is expected), and which assay is to be used. The IFCT-ERMETIC study [19] was conducted to investigate the accuracy of EGFR and KRAS mutation detection in 15 platforms. That study found a favourable agreement between centres, underlining the accuracy of such analysis [19]. In addition, as clinicians were allowed to forego biomarker testing if one mutation was found, this might have introduced a selection bias in the case of multiple mutations. However, multiple mutations in lung cancer are uncommon. The IFCT-Biomarker France study (n=17 664 patients), found only 1% of patients with multiple mutations. Patients with multiple alterations were more likely to be never-smokers [30]. In the American Lung Cancer Mutation Consortium Experience, the rate of multiple mutations appeared to be slightly higher (2.9% among 1007 specimens), but the panel used was the widest in comparison to other studies [31]. In the current analysis, we found an intermediate rate of 1.9%, which illustrates that selection bias was probably not a major factor. Finally, the design of our study might have also generated selection bias. It is possible that some physicians were more alert to track the never-smoker status in particular settings, such as with younger subjects or women. However, our overall results were comparable to those in the literature within a similar setting [9]. In addition, our study is a case-only single cohort study without comparison to smokers, and without an independent cohort to validate our findings. Our study also exhibits certain strengths. To the best of our knowledge, this is the only cohort to address six driver oncogenes and six occupational carcinogens simultaneously in a unique cohort of lifelong never-smokers with NSCLC. We therefore report fully comprehensive findings. In particular, smoking did not interfere as a confounder, unlike most studies addressing occupational lung cancers. In addition, we used a standardised questionnaire, delivered during a phone interview by dedicated staff, to limit redaction bias, as well as memorisation bias, and to minimise the occurrence of missing values. Finally, we used an internationally recognised definition of never-smoker to avoid contamination bias [9].

Whereas exposure to passive smoking does not appear to affect the molecular pattern in a French never-smoker cohort [21], occupational exposure seems to be slightly associated with specific patterns. In particular, EGFR and HER2/ERBB2 appear to have opposite levels of association with asbestos exposure; although these findings were limited by the small sample size. Such results highlight the crucial step of assessing occupational exposure in lung cancer patients, especially in male never-smokers [9]. These original results could contribute to the formulation of hypotheses to design further studies and have a better understanding of the oncogenic pathways driven by occupational carcinogens.