Lung cancer is the leading cause of cancer-specific mortality for men and women in the U.S. [1,2,3]. The National Cancer Institute’s (NCI) Surveillance Epidemiology and End Results program estimated that 226,160 new lung cancer cases were diagnosed in the U.S. in 2012, and that 160,340 individuals died of this disease. Of all lung cancers, 99% consist of either small cell carcinomas (13.9%) or non-small cell carcinomas (85.1%). Histological subtypes of non-small cell lung cancer (NSCLC) include adenocarcinomas (37.5%), squamous and transitional cell carcinomas (19.8%), large cell carcinomas (3.3%), and other tumor subtypes (24.5%). Patients diagnosed with localized, early stage lung cancer have a 5-year survival rate of 49.5% and may be cured by surgical resection [4,5]. However, 77% of lung cancer patients have advanced stage disease at diagnosis, and their 5-year survival rate is only 20.6% for disease that has spread to the regional lymph nodes or 2.8% for disease that has metastasized to distant anatomical sites. Because patients who are diagnosed with early stage lung cancer have a significantly better prognosis, screening (i.e., early detection) represents a practical public health approach for decreasing lung cancer mortality.

Throughout the 1960s and 1970s, chest x-ray (CXR) alone or in conjunction with sputum cytology were evaluated to detect early stage lung cancer [4,5]. To assess the feasibility, validity, and efficacy of CXR for lung cancer screening, the NCI sponsored the Prostate, Lung, Colorectal, and Ovarian (PLCO) randomized screening trial, which was designed with a “no-screening” study arm and 89% power to detect a 10% reduction in lung cancer mortality [6,7]. While awaiting PLCO trial results [8,9], contemporary researchers have focused on developing low dose computed tomography and biomarkers as potential lung cancer screening modalities [10,11,12,13,14,15,16].

The human epidermal growth factor receptor 2 (HER2/ERBB2/neu) proto-oncogene encodes a cell surface receptor tyrosine kinase (RTK) that functions to regulate cell proliferation and survival [17]. HER2 amplification and protein overexpression have been implicated in the etiology and pathogenesis of several human malignancies including lung cancer [18]. HER2 is overexpressed in 10–20% of NSCLC cases and in perhaps as many as 30% of lung adenocarcinomas [19,20,21,22,23], where overexpression is associated with adverse tumor characteristics and poor patient prognosis [24,25]. In addition to full-length HER2, cells synthesize “soluble” HER2 (sHER2) isoforms [26,27,28,29,30]. These sHER2 isoforms are produced either by alternate mRNA splicing or by proteolytic cleavage, and, are either secreted or proteolytically shed from the plasma membrane into extracellular body fluids. Alternate splicing results in mRNA transcripts that encode 68-kDa [30] and 100-kDa [28] sHER2 isoforms; whereas, proteolytic cleavage results in 105-kDa [26] and 110-kDa [27] shed isoforms of sHER2 that encompass only extracellular subdomains of this RTK. While it is widely assumed that the 105-kDa sHER2 isoform represents the major constituent of human blood, a careful biochemical characterization of serum sHER2 isoforms has not yet been performed. Nonetheless, multiple assays to quantify sHER2 have been developed and used to assess the potential clinical utility of serum sHER2 in cancer patients across disease sites [31,32]. In particular, serum sHER2 has been examined in preliminary studies of lung cancer, but has not yet been rigorously validated as a potential risk assessment, screening, and/or diagnostic biomarker. Initial reports suggest that serum sHER2 is elevated in 5–64% of lung cancer cases [23,32,33,34,35,36] and may be elevated many months prior to clinical diagnosis [33].

In this study, we evaluated 244 primary NSCLC cases and 218 healthy controls using an unmatched, retrospective, case-control study design to determine whether age and/or gender are confounders or effect modifiers of the relationship between sHER2 and NSCLC. We observed that sHER2 concentrations are associated with age among healthy men, and differ between healthy men and women. Although sHER2 concentrations do not differ between healthy controls and patients with NSCLC overall, they are slightly higher in patients with adenocarcinoma regardless of gender, but not squamous cell carcinoma. Logistic regression models further demonstrate that sHER2 is a weak, independent classifier of adenocarcinoma but not of squamous cell carcinoma, and when adjusted for age and gender, the risk of adenocarcinoma increases with higher sHER2 concentrations. Moreover, when mutually adjusted for each other, sHER2, age, and gender distinguish healthy controls from patients with adenocarcinoma with 85.9% accuracy. These data suggest that albeit serum sHER2 is not a strong, stand-alone discriminatory classifier of adenocarcinoma, its independence of age and gender may confer some limited utility to sHER2 as a covariate in multivariate models for the risk assessment, screening, and/or diagnosis of lung adenocarcinoma.

Steroid hormones have been implicated in the regulation of HER2 expression; however, the age and gender specific effects of steroids on HER2 expression, as well as on sHER2 shedding is not completely understood. In human breast cancer cells in vitro, estradiol is a potent inhibitor of HER2 expression and sHER2 shedding [41,42,43,44]; whereas, progesterone positively regulates HER2 expression [45]. In contrast, neither estrogen nor progesterone regulates HER2 expression in ovarian cancer cells [46]. Among men with prostate cancer, high HER2 expression has been associated with androgen ablation therapy and androgen independent prostate carcinoma [47]. Among healthy women, oral contraceptive use and hormone replacement therapy have been associated with lower serum sHER2 concentrations [48]. In pregnant women, serum sHER2 concentrations are lower during the first and second trimester in pregnant women, which is when estrogen concentrations increase, but are higher during the third trimester [49,50,51]. Finally, elevated serum sHER2 concentrations among women with preeclampsia vs. a normal pregnancy suggest the possibility of fetomaternal transfer during the third trimester of pregnancy [49,52]. Together, these observations suggest that steroid hormones may regulate both the biosynthesis and shedding of sHER2 into circulatory fluids. Our observations that sHER2 concentrations decrease with age in healthy men, and that age and gender confound the relationship between sHER2 concentrations and lung adenocarcinoma are consistent with hormonal regulation of HER2 expression and/or shedding. These observations underscore the importance of adjusting for age, gender, and other potential confounding variables that may affect steroid hormone biosynthesis, such as body mass and smoking [53,54], when evaluating the utility of sHER2 for the risk assessment, early detection, and diagnosis of adenocarcinoma using epidemiological study designs and multivariate statistical methods.

Limitations and caveats of this retrospective study include: (a) an inability to assess the effect of long term storage at −80 °C on serum sHER2 stability, (b) an inability to assess potential differences in serum sHER2 concentrations between non-small cell adenocarcinomas compared to large or small cell carcinomas; (c) an inability to assess potential differences in serum sHER2 concentrations by disease stage or tumor grade; (d) lack of data on potential confounders or effect modifiers of sHER2 concentrations such as smoking status, body mass, personal and family cancer history, carcinogen exposure, menstrual cycle phase, parity, and exogenous hormone use among women; (e) potential ascertainment and cohort bias of the control group, which was not selected using the same enrollment criteria (e.g., age and smoking status) or at random during the same time period from the source population as the cases; and (f) an inability to assess the risk of developing lung adenocarcinoma among asymptomatic individuals with pre-diagnostic serum samples. Given these caveats and limitations, it will be important to conduct well-designed transitional studies to assess the effects of storage at −80 °C on serum sHER2 concentrations, and prospective case-control studies to further validate sHER2 as a weak, independent biomarker of lung adenocarcinoma using multivariate models. Such studies should include incident cases of lung adenocarcinoma, other malignant and benign lung neoplasms, and healthy controls selected randomly from the study population.

Several reports indicate that serum sHER2 concentrations have 5–64% sensitivity to detect lung cancer of various histological subtypes using cut-off thresholds that correctly classify healthy controls with 95% specificity [23,32,33,34,35,36]. Sensitivity was reported to be higher for patients with late stage lung cancer and for patients with adenocarcinoma compared to SCC, large cell carcinoma, and small cell carcinoma [23,32,34,35]. In general, these studies were small, used disparate immunoassays, or used previously established cut-off thresholds from an external control study group [23]. Additionally, most of these studies did not control for confounding by age, gender, and other variables such as smoking by using a matched, case-control study design or by adjusting for confounding in an unmatched, case-control study design with multivariate logistic regression methods [23,32,34,35,36]. Notably, the highest reported sensitivity (64%) and specificity (95%) of sHER2 concentrations to discern healthy controls from lung cancer cases was reported for a small age-, gender-, ethnic group-, and smoking-matched case-control study [33]. That age might be a potential confounder or effect modifier of the association between sHER2 and cancer is supported by our observation that sHER2 concentrations decrease with age in healthy men, and by a previous report showing that sHER2 concentrations increase with age in healthy women independently of menopausal status [55]. Here, it is observed that sHER2 concentrations are higher in primary NSCLC patients with adenocarcinoma than SCC, thus confirming that biological differences exist between these histological subtypes. Moreover, we report that the ability of sHER2 concentrations to discern healthy controls from patients with adenocarcinoma, but not SCC, is indeed confounded by age and gender in our dataset (Table 4 and Figure 4). When mutually adjusted for each other, sHER2, age, and gender yield 85.9% accuracy to classify controls from primary adenocarcinoma patients (Table 4). However, the strength to discern healthy controls from patients with adenocarcinoma is contributed mainly by age, not sHER2 and gender (AUC = 0.840 with age alone vs. AUC = 0.859 with age, gender, and sHER2 combined). The accuracy to discern adenocarcinoma patients from controls with age-adjusted sHER2 concentrations is 91% among men, but only 78% among women (Figure 4(I)), suggesting that additional confounders may be influencing the association between sHER2 and adenocarcinoma among women. Notably, at 95% specificity, the sensitivity to detect adenocarcinoma is 60% for men and 30% for women; these statistics are similar to those reported by Brandt-Rauf and colleagues, who used a matched case-control study design to control for confounding by age, gender, ethnicity, and smoking [33]. Importantly, the data reported here show that sHER2 is a weak, but statistically significant, independent classifier of adenocarcinoma. Although sHER2 is not a strong, stand-alone classifier of adenocarcinoma, its independence of age and gender may make sHER2 a useful covariate (with age, gender, and other biomarkers) in multivariate screening and diagnostic models of lung adenocarcinoma.

Risk prediction models that include smoking are useful for advising patients about their propensity to develop lung cancer. However, while smoking is a well-known risk factor for lung cancer, considered alone, smoking history has limited ability to predict which smokers will develop lung cancer. Other contributing epidemiological, clinical, and biological (i.e., biomarkers) factors may improve lung cancer risk prediction among smokers and even non-smokers [56,57]. Several groups have developed lung cancer risk models using the variables of age, smoking history, carcinogen exposure, and family history [58,59,60,61,62,63,64]. However, few of these studies included serum biomarkers, despite our understanding that biomarkers represent promising risk assessment tools [56,57,59,64]. Recently, Spitz and colleagues modestly improved their risk model, which included environmental tobacco smoke, family cancer history, dust exposure, prior respiratory disease, hay fever, and smoking history variables by adding two in vitro markers of DNA repair capacity (host-cell reactivation and mutagen sensitivity) [65]. Adding these markers significantly improved the model’s accuracy to discriminate between lung cancer cases and controls from 67% to 70% among former smokers (p = 0.006), and from 68% to 73% among current smokers (p = 0.0048). Similarly, Prindiville and colleagues found that sputum samples characterized by moderate to worse cytological atypia defined a cohort of high-risk patients who were at increased risk of developing lung cancer [66]. Notably, we observe that, adjusted for age, the relative risk (odds) of adenocarcinoma is 3.95 (95% CI: 1.22, 12.81) among women and 7.93 (95% CI: 2.26, 27.82) among men with high sHER2 concentrations (≥6.60 ng/mL) compared to women or men with low sHER2 concentrations (≤1.85 ng/mL; Table 5). These data are consistent, but not conclusive, with a role for serum sHER2 as a biological marker of risk for primary non-small cell lung adenocarcinoma.

Cancer treatment is undergoing a paradigm shift from the use of broad-spectrum cytotoxic chemotherapeutics to biologically targeted therapeutics [67], including HER-targeted drugs [68]. The HER family of receptors, EGFR/HER1, HER2, HER3, and HER4 are rational targets for anti-cancer drugs because they activate complex signal transduction pathways that lead to tumor cell proliferation, survival, and metastasis [69]. In patients with NSCLC, EGFR-targeted RTK inhibitors, such as gefitinib and erlotinib, have produced objective response rates in about 10% of patients with advanced stage disease and modest improvements in patient survival [70,71,72]. Clinical trials with HER2-targeted monoclonal antibodies, trastuzumab and pertuzumab, also have shown efficacy in some NSCLC patients [70,73,74]; however, trials with trastuzumab have been insufficiently powered to determine whether NSCLC patients with HER2 gene amplification or overexpression as determined by immunohistochemistry may benefit from treatment [75,76]. Notably, responsiveness and efficacy toward HER-targeted therapeutics is dependent on the selection of patients who present with the active target of choice, and accurate methods to identify these patients [77,78]. Here, we observe that serum sHER2 concentrations detected by ELISA are higher in a subset of primary NSCLC patients with adenocarcinoma (adjusted for age, 60% of men and 30% of women had sHER2 concentrations above the 95th cut-off threshold for healthy controls), suggesting that sHER2 may warrant investigation as a theragnostic biomarker to select a subset of patients with lung adenocarcinoma for treatment with HER2-targeted drugs.

In summary, we have observed that serum sHER2 concentrations are higher in a subset of patients with primary adenocarcinoma compared to patients with SCC or healthy controls, and weakly, but independently associated with adenocarcinoma even after adjusting for confounding by age and gender. These results are consistent with the hypothesis that sHER2 may have some limited clinical utility as a serum biomarker of non-small cell lung adenocarcinoma, if properly adjusted for confounding by age, gender, and variables that affect circulating and tissue steroid hormone concentrations (e.g., body mass, smoking, oral contraceptive use, and hormone replacement therapy) or circulating steroid hormones themselves. We conclude that even though serum sHER2 is not a strong, stand-alone discriminatory biomarker of adenocarcinoma, sHER2 may be a useful, independent covariate in multivariate regression models that warrants further investigation for the risk assessment, screening, and diagnosis of lung adenocarcinoma in well-designed epidemiological studies that control for confounding by age, gender, and other demographical, epidemiological, behavioral, and environmental variables using multivariate regression methods or a matched case-control study design.