Interstitial lung disease is associated with an increased risk of lung cancer in systemic sclerosis: Longitudinal data from the Canadian Scleroderma Research Group

Abstract

Objective:

The literature supports an increased risk of malignancy in systemic sclerosis, including lung cancer. Our objective was to identify potential independent predictors of lung cancer risk in systemic sclerosis.

Methods:

We used a cohort of 1560 systemic sclerosis patients from the Canadian Scleroderma Research Group, enrolled from 2004 and followed for a maximum of 11 years. Time to lung cancer was calculated from the onset of the first non-Raynaud’s symptoms. Baseline demographic, clinical, and serological characteristics of patients with and without lung cancer were compared. Cox proportional hazards models were used to estimate the effects of demographic variables, exposure to smoking, disease duration, disease subset (diffuse vs limited), immunosuppressant drug exposure, and presence of interstitial lung disease on the risk of lung cancer.

Results:

Over the 5519 total person-years of follow-up, 18 SSc patients were diagnosed with lung cancer after cohort entry (3.2 cancers per 1000 person-years). In univariate comparisons, cancer cases were more likely to be male, to have a smoking history, and to have interstitial lung disease than non-cases. In multivariate analysis, interstitial lung disease was independently associated with the risk of lung cancer (hazard ratio: 2.95, 95% confidence interval: 1.10–7.87).

Conclusion:

In addition to known demographic (male sex) and lifestyle risk factors (smoking), interstitial lung disease is an independent risk factor for lung cancer in systemic sclerosis. These results have implications for lung cancer screening in systemic sclerosis.

Introduction

Systemic sclerosis (SSc) is a multisystem connective tissue disease characterized by vasculopathy, autoimmunity, and extensive fibrosis involving the skin and visceral organs including the lung, heart, and gastrointestinal tract. ¹ Lung cancer is the most common malignancy reported in association with SSc: three meta-analyses of observational studies^2–4 have shown a threefold to fourfold increase in the risk of lung cancer risk in SSc patients compared to controls. Predictive factors of lung cancer in SSc have been previously investigated, but studies have shown conflicting results.^2,4–10 The aim of this study was to identify the demographic, clinical, and serological risk factors for lung cancer in SSc.

Materials and Methods

Data collection and case ascertainment

Subjects included in this analysis were identified from the Canadian Scleroderma Research Group (CSRG) registry, a multicenter cohort followed annually since 2004. Over 1500 subjects are currently enrolled from 15 centers across Canada, with a mean follow-up of approximately 5 years. Subjects must be at least 18 years of age, have a diagnosis of SSc confirmed by a participating rheumatologist, and be fluent in either English or French. The vast majority of subjects in the cohort meet the 2013 ACR/EULAR (American College of Rheumatology/European League Against Rheumatism) classification criteria for SSc. ¹¹ This study included subjects recruited between cohort inception and June 2015.

The CSRG data collection protocol has been approved by the McGill University Institutional Review Board, and by all participating sites. All patients provided written informed consent.

Lung cancer diagnoses occurring at any time after cohort entry were included in our analyses. Ascertainment of lung cancer cases was based on physician reports at each annual follow-up visit. Time to lung cancer diagnosis was calculated from the onset of the first non-Raynaud symptoms, with left censoring to account for time between onset of first non-Raynaud symptoms and cohort entry. Pathology reports were retrieved to confirm lung cancer diagnosis and subtype. If pathology reports were not available, cause of death reported by a study physician was retrieved.

Study variables

Baseline data extracted from the registry included self-reported demographic and smoking status, physician-reported disease duration (defined from onset of first non-Raynaud symptoms attributed to SSc to cohort entry), and cutaneous disease subset (limited vs diffuse). Exposure to immunosuppressive medications (cyclophosphamide, methotrexate, mycophenolate, and azathioprine) was recorded by a study physician as “in the past” or “currently” at baseline study visit and “in the past year” or “currently” at every subsequent study visit.

The presence of interstitial lung disease (ILD) was determined using a published clinical decision rule. ¹² Using this algorithm, ILD was considered present if a high-resolution computed tomography (HRCT) scan of the lung was interpreted by an experienced radiologist as showing ILD or, in the case where no HRCT was available, if either a chest X-ray was reported as showing either increased interstitial markings (not thought to be due to congestive heart failure) or fibrosis, and/or if a study physician reported the presence of typical “Velcro-like crackles” on physical examination.

Pulmonary function tests were obtained at each site in accordance with the American Thoracic Society standards. The percentage predicted forced vital capacity (FVC) and single breath diffusing capacity of the lung for carbon monoxide (DLCO) corrected for hemoglobin were extracted from reports.

Baseline sera was collected and sent to a central laboratory at the University of Calgary, Canada, where samples were aliquoted and stored at −80°C until analyzed. Anti-centromere (CENP-A and CENP-B), anti-topoisomerase I, and anti-RNA polymerase III (RP11 and RP155) antibodies were detected by Euroline SSc profile line immunoassay (Euroimmun, Lubeck, Germany) according to manufacturer’s instructions. With the intent of optimizing specificity, antibodies were reported as absent (negative, equivocal, and low titers) and present (moderate and high titers).

Statistical methods

Baseline demographic, clinical, and serological characteristics were compared between SSc patients who developed lung cancer versus those who remained cancer free, using descriptive statistics. Cox proportional hazards models were used to estimate hazard ratios (HRs) with 95% confidence intervals (CIs), with the outcome of lung cancer adjusted for sex, age, race, number of pack-years smoked (compared to never smoking), presence of ILD, disease duration, and disease subset (diffuse vs limited). In sensitivity analyses, immunosuppressant drug exposure and alternatively cyclophosphamide exposure alone, modeled as time-dependent exposures, were also included in the model.

Results

A total of 1560 patients with SSc were enrolled in the CSRG cohort from 2004 and followed for a maximum of 11 years. Over the total 5519 person-years of follow-up, 18 SSc patients were diagnosed with lung cancer (3.2 cancers/1000 person-years). Lung cancer histopathology obtained from pathology reports or reported as “cause of death” consisted of adenocarcinoma in 10, squamous cell carcinoma in 5, non-small cell carcinoma not otherwise specified in 1, and “metastatic lung cancer” in 2 patients.

As seen in Table 1, patients who developed lung cancer were older at cohort entry (58.8 ± 11.2 vs 55.2 ± 12.3 years), more likely to be male (33.3% vs 13.9%), more likely to have a past or current smoking history (94.4% vs 59.1%), more likely to have diffuse cutaneous SSc (50.0% vs 36.6%), and more likely to have ILD (58.8% vs 30.3%), with lower FVC (85.5% vs 91.6%) and DLCO (59.5% vs 72.6%) compared to those without lung cancer.

Table 1.

Baseline characteristics of lung cancer cases and cancer-free controls in a multicenter SSc cohort.

Characteristics	Lung cancer (N = 18)		Lung cancer free (N = 1535)		95% CI for the difference
	Mean (SD) or N (%)	Missing	Mean (SD) or N (%)	Missing
Age, years	58.8 (11.2)	0	55.2 (12.3)	3	−2.0, 9.4
Female	12 (66.7%)	0	1321 (86.1%)	1	2.3, 42.4
Ever smoker	17 (94.4%)	0	844 (59.1%)	107	15.0, 40.6
Caucasian	15 (83.3%)	0	1238 (86.6%)	105	−7.8, 25.9
Disease duration, years	11.9 (7.0)	0	10.2 (9.4)	21	−2.6, 6.1
Diffuse cutaneous subset	9 (50%)	0	555 (36.6%)	19	−7.8, 34.5
Interstitial lung disease	10 (58.8%)	1	452 (30.3%)	41	5.6, 48.3
FVC predicted (%)	85.5 (19.4)	2	91.6 (19.5)	236	−15.7, 3.5
DLCO predicted (%)	59.5 (22.1)	5	72.6 (36.1)	421	−32.8, 6.6
Pulmonary hypertension	1 (6.3%)	2	140 (10.8%)	236	−17.6, 10.0
Ever cyclophosphamide	0	0	100 (6.6%)	20	−11.0–8.0
Ever methotrexate	2 (11.1%)	0	292 (19.3%)	20	−13.6, 16.4
Ever azathioprine	1 (5.6%)	0	99 (6.5%)	20	−19.3, 5.7
Ever mycophenolate	0	0	47 (3.1%)	20	−14.5, 4.1
Anti-centromere antibodies	5 (29.4%)	1	507 (38.7%)	224	−14.6, 25.6
Anti-topoisomerase I antibodies	3 (17.7%)	1	209 (15.9%)	224	−9.9, 25.2
Anti-RNA polymerase III antibodies	1 (5.9%)	1	192 (14.7%)	224	−12.4, 14.0

SD: standard deviation; CI: confidence interval; FVC: forced vital capacity; DLCO: diffusing capacity of the lungs for carbon monoxide.

In a Cox proportional hazard model, male sex (HR: 3.14, 95% CI: 1.08–9.18) and smoking (compared to non-smokers at <10 pack-years (HR: 2.36, 95% CI: 0.51–10.97); 10–20 pack-years (HR: 5.04, 95% CI: 1.11–22.85); 20–30 pack-years (HR: 3.54, 95% CI: 0.55–23.00); and >30 pack-years (HR: 6.17, 95% CI: 1.29–29.49)) were independently associated with lung cancer (Table 2). Similarly, the presence of ILD (HR: 2.95, 95% CI: 1.10–7.87) was an independent risk factor for lung cancer.

Table 2.

Cox proportional hazard model to estimate the effect of interstitial lung disease and other potential risk factors on risk of lung cancer in SSc.

	Unadjusted model HR (95% CI)	Adjusted model HR (95% CI)	Sensitivity model 1 HR (95% CI)	Sensitivity model 2 HR (95% CI)
Interstitial lung disease	3.41 (1.30–8.96)	2.95 (1.10–7.87)	2.71 (1.00–7.37)	2.80 (1.05–7.50)
Male	3.73 (1.40–9.95)	3.14 (1.08–9.18)	3.16 (1.07–9.28)	3.15 (1.07–9.28)
Age (per year)	1.03 (0.99–1.08)	1.03 (0.98–1.09)	1.03 (0.98–1.09)	1.03 (0.98–1.09)
Caucasian	0.61 (0.18–2.11)	0.41 (0.11–1.48)	0.41 (0.11–1.51)	0.40 (0.11–1.47)
Smoking (reference non-smokers)
<10 pack-years	2.17 (0.49–9.69)	2.36 (0.51–10.97)	2.33 (0.50–10.84)	2.35 (0.50–10.96)
10–20 pack-years	4.49 (1.01–20.08)	5.04 (1.11–22.85)	5.14 (1.13–23.36)	5.11 (1.12–23.33)
20–30 pack-years	4.83 (0.97–23.99)	3.54 (0.55–23.00)	3.60 (0.55–23.41)	3.57 (0.55–23.21)
>30 pack-years	6.30 (1.41–28.17)	6.17 (1.29–29.49)	6.37 (1.33–30.51)	6.23 (1.30–29.89)
Disease duration (per year)	1.02 (0.97–1.07)	1.02 (0.97–1.08)	1.03 (0.98–1.08)	1.03 (0.97–1.08)
Diffuse cutaneous subset	1.70 (0.67–4.28)	1.45 (0.54–3.93)	1.50 (0.55–4.05)	1.50 (0.55–4.05)
Cyclophosphamide	–	–	1.65 (0.20–13.95)	–
Immunosuppressants	–	–	–	1.04 (0.28–3.87)

HR: hazard ratio; CI: confidence interval.

In sensitivity analyses adjusting for either cyclophosphamide exposure alone (sensitivity analysis model 1) or, alternatively, immunosuppressant exposure (sensitivity analysis model 2) modeled as time-dependent exposures, the above reported associations persisted (Table 2).

Over the follow-up period, 11 of 18 (61.1%) lung cancer patients died, at a mean of 3.0 years after lung cancer diagnosis. Among patients free of lung cancer, there were 215/1535 (14%) deaths over a mean of 3.0 years of follow-up.

Discussion

According to the Canadian Cancer Statistics, the incidence rate of lung cancer in the Canadian general population is 0.58 per 1000 among males, and 0.48 per 1000 among females. The incidence of lung cancer in our SSc cohort was about fivefold higher (3.2 per 1000 patient/years), which may support an increased risk of lung malignancy in this disease. However, this comparison needs to be interpreted with caution. First, our SSc cohort was not age-matched to the Canadian general population and second, we did not examine differences in lung cancer rates across provinces or by race/ethnicity. Moreover, direct comparison with general population cancer rates is problematic, since general population cancer rates are obtained from cancer registry data, and the SSc cancer cases in this study were not ascertained using this method. Nevertheless, our lung cancer cases were confirmed by pathology.

Our findings suggesting an increased risk of lung cancer in SSc are consistent with previously reported studies.^5,13 Our findings that males with SSc have an increased risk of lung cancer have also been reported by two meta-analyses.^2,4 This is also consistent with the trend in the general population data for more lung cancers in males versus females, though this is thought to be driven largely by smoking patterns, and there is an emerging trend in the general population for lung cancer rates in females to approach that of males.

Data in most prior studies (and included in meta-analyses) were retrospectively collected by record linkage among various healthcare databases, and therefore some important factors such as smoking status were not completely accounted for. In this study, baseline clinical data were collected prospectively at CSRG cohort entry, and therefore adjusting for confounding factors, including smoking status, was possible. We confirmed the well-known risk that smoking confers on lung cancer.

A link between idiopathic pulmonary fibrosis and lung cancer has been relatively well recognized for years. Autopsy series have confirmed the high frequency of lung cancer in patients with usual interstitial pneumonia (UIP), the well recognized histologic pattern that defines idiopathic pulmonary fibrosis.^14,15 The association of idiopathic pulmonary fibrosis with lung cancer was found to be independent of smoking. ¹⁶ Several autoimmune rheumatic diseases have been linked to an increased risk of lung malignancy, including rheumatoid arthritis,^17,18 systemic lupus erythematous,^19–21 and idiopathic inflammatory myopathy. ²² Only scant data are available to suggest a possible relationship between rheumatoid arthritis-associated UIP and lung cancer risk ¹⁸ and, interestingly, one recent study of idiopathic inflammatory myopathy suggested that there was a negative correlation between cancer risk and ILD. ²³

In SSc, reports on the association between ILD and lung cancer in SSc have to date been conflicting.^4–8 This study provides additional support that confirms that ILD is independently associated with increased lung cancer risk in SSc.

Mechanisms underlying the increased risk of lung cancer development in SSc remain speculative. Some theories have invoked persistent inflammation, defects in immune surveillance and clearance of carcinogens, cytotoxic immunosuppressive treatments, genetic predisposition to both carcinogenesis and autoimmune dysregulation, and/or exposure to common environmental triggers that could play a role in cancer development.^24–28 A role of chronic lung inflammation (independent of prior smoking) in lung cancer has been suggested in several studies involving patients with idiopathic pulmonary fibrosis and chronic obstructive pulmonary diseases.^29,30 Several markers of systemic inflammation have been linked to an enhanced lung cancer risk, namely, acute phase reactants such as c-reactive protein ³¹ and pro-inflammatory cytokines like IL-6, IL-7, and IL-8.^32,33 Persistent inflammation triggers repeated events of injury and repair to the respiratory epithelium, leading to cellular atypia, epithelial metaplasia, and genetic alterations eventually causing tumorigenesis.^14,34 On the molecular level, chronic pulmonary inflammation causes upregulation of tumor suppressor proteins, such as p53, in patients with idiopathic pulmonary fibrosis and ILD related to connective tissue diseases, triggering apoptosis of DNA damaged cells. With persistent inflammation, induction of mutational changes in p53 gene is observed, thereby altering its tumor suppressor activity and potentially leading to uncontrolled cellular proliferation and carcinogenesis.^35–38

We were unable to document a clear association between exposure to immunosuppressive therapy in our SSc patients and risk of lung cancer, possibly because few subjects were exposed to these drugs. Previous studies have linked cyclophosphamide to an increased risk of bladder, non-melanoma skin and hematological malignancies in patients with rheumatoid arthritis,^39–41 systemic lupus erythematous, and other systemic vascultis.^42–45

Serum autoantibodies have been reported to be associated with the risk of malignancy in idiopathic inflammatory myopathies.^46,47 In SSc, some data, although quite scarce, suggest a link between SSc-specific antibodies, namely, anticentromere, anti-topoisomerase I antibodies, and cancer,^9,10 and more specifically lung cancer.^6,48 An association between anti-RNA polymerase III antibodies and cancer diagnosed within 2–3 years of the clinical onset of SSc has been reported, suggesting a possible paraneoplastic phenomenon in this subset of SSc patients.^9,49,50 In this study, we did not find any associations between serum antibodies and lung cancer risk, although this may be due to lack of power given the small numbers of lung cancer cases.

The histologic subtype of lung cancer associated with SSc has been quite heterogeneous across studies. One meta-analysis reported an equal proportion of adenocarcinoma and squamous cell carcinomas, ³ while other case series showed a predominance of bronchioloalveolar carcinoma, now referred to as adenocarcinoma in situ.^51,52 In this study, adenocarcinoma was the most commonly encountered lung cancer subtype. In fact, adenocarcinoma is the most common lung cancer subtype in contemporary studies, particularly among women and nonsmokers.^45,53

This study has many strengths. Data were collected prospectively and in a standardized manner. Extensive baseline demographic, clinical, and serological data were available. Smoking status, a major confounding factor relative to the association of SSc and SSc-related ILD with lung cancer, was well documented. Although lung cancer case ascertainment was based on physician reports, pathology reports were successfully obtained in the majority of cases. Results were adjusted for key baseline variables, including age, sex, race/ethnicity, smoking status, and immunosuppressant drug exposure. The time interval between the first clinical manifestations of SSc, defined as onset of non-Raynaud’s symptoms, and lung cancer diagnosis was well estimated, in contrast to previous studies and meta-analyses, where such crucial information was not always readily available.

Our study also has some limitations. As in other studies investigating risk of lung cancer in SSc, it may be that some cancers were diagnosed earlier than in the general population due to the closer medical surveillance of these patients. In addition, lung cancer case ascertainment was based on physician and pathology reports. As well, the CSRG registry cohort consists mainly of referral patients assessed at 15 academic centers across Canada. Our findings could therefore be subject to a selection bias because of overrepresentation of severely ill individuals, who have higher risk of medical comorbidities and, possibly, higher risk estimates of malignancy. Nevertheless, the characteristics of our SSc cohort are highly similar to those of other large SSc cohorts around the world and our results are thus representative of this sampling frame. Finally, our study population was not age and sex-matched to the general Canadian population, although our SSc cohort was fairly young and the majority were female.

In conclusion, in this study, male sex and smoking history were important predictors of lung cancer in SSc. Most interestingly, ILD was also associated with a threefold increase in lung cancer risk, independent of male sex and smoking history. These clinical risk factors may facilitate the identification of high-risk SSc subjects who require screening for lung cancer.