Article Text
Abstract
Objectives Recent studies demonstrate autoantibodies are powerful tools to interrogate molecular events linking cancer and the development of autoimmunity in scleroderma. Investigating cancer risk in these biologically relevant subsets may provide an opportunity to develop personalised cancer screening guidelines. In this study, we examined cancer risk in distinct serologic and phenotypic scleroderma subsets and compared estimates with the general population.
Methods Patients in the Johns Hopkins Scleroderma Center observational cohort were studied. Overall and site-specific cancer incidence was calculated in distinct autoantibody and scleroderma phenotypic subsets, and compared with the Surveillance, Epidemiology and End Results registry, a representative sample of the US population.
Results 2383 patients with scleroderma contributing 37 686 person-years were studied. 205 patients (8.6%) had a diagnosis of cancer. Within 3 years of scleroderma onset, cancer risk was increased in patients with RNA polymerase III autoantibodies (antipol; standardised incidence ratio (SIR) 2.84, 95% CI 1.89 to 4.10) and those lacking centromere, topoisomerase-1 and pol antibodies (SIR 1.83, 95% CI 1.10 to 2.86). Among antipol-positive patients, cancer-specific risk may vary by scleroderma subtype; those with diffuse scleroderma had an increased breast cancer risk, whereas those with limited scleroderma had high lung cancer risk. In contrast, patients with anticentromere antibodies had a lower risk of cancer during follow-up (SIR 0.59, 95% CI 0.44 to 0.76).
Conclusions Autoantibody specificity and disease subtype are biologically meaningful filters that may inform cancer risk stratification in patients with scleroderma. Future research testing the value of targeted cancer screening strategies in patients with scleroderma is needed.
- systemic sclerosis
- autoantibodies
- epidemiology
Statistics from Altmetric.com
Introduction
Prior investigations have demonstrated an increase in cancer risk in patients with systemic sclerosis (scleroderma) compared with the general population.1–11 In a study of patients with scleroderma and cancer, our group showed that patients with RNA polymerase III autoantibodies (antipol) had cancer occur within a short interval of scleroderma onset.12 Subsequent studies demonstrated that these patients have genetic alterations (somatic mutations and/or loss of heterozygosity) at the POLR3A locus that encodes for RNA polymerase III in their cancers, with both mutation-specific and cross-reactive immune responses seen.13 These data strongly suggest that alterations of autoantigen sequence in cancers may trigger antitumour immune responses that spread to the wild-type molecule, resulting in autoimmunity.14
Many international scleroderma cohorts have similarly observed that patients with scleroderma and antipol have a significantly increased risk of cancer at the time of scleroderma onset compared with scleroderma patients without these antibodies.15–19 In addition, patients lacking antibodies against centromere, topoisomerase-1 and RNA polymerase III (hereafter referred to as ‘CTP-negative’) also have more cancer diagnosed within a short interval of scleroderma onset, suggesting there may be other serologic subsets of cancer-associated scleroderma.17 20 21 Case reports suggest that therapy of coincident cancer may induce scleroderma remission,22–24 raising the possibility that early cancer detection and therapy in patients with new-onset scleroderma might improve scleroderma outcomes.
Our prior work suggests that investigating cancer risk in scleroderma as a group, without differentiating between serologically relevant subsets or using the cancer-scleroderma interval as a filter, may mask important differences in the relationship between cancer and autoimmunity. In the current study, we examined overall and site-specific cancer risk at scleroderma onset in distinct serologic and phenotypic subsets and for the first time compared these estimates with the general population.
Methods
Study population
Patients seen at the Johns Hopkins Scleroderma Center for their first visit between 1 January 2000 and 31 December 2015 were eligible for the study if they consented to participate in our IRB-approved cohort database and had a diagnosis of scleroderma. Scleroderma was defined by 1980 or 2013 American College of Rheumatology/European League Against Rheumatism classification criteria,25 26 at least three of five CREST (calcinosis, Raynaud’s phenomenon, esophageal dysmotility, sclerodactyly, telangiectasia) syndrome criteria or having definite Raynaud’s, abnormal nailfold capillaries and a scleroderma-specific autoantibody. Clinical and serological data are collected prospectively at baseline and at 6 months interval. Patients were classified as having limited or diffuse scleroderma by established criteria.27 Four autoantibody categories were assessed: anticentromere A/B (cenp), antitopoisomerase-1 (topo), antipol and CTP-negative. Patients were considered positive for an autoantibody if they were ever positive based on clinically obtained assays. Double autoantibody positivity was infrequent: 14 patients were positive for both anticenp and antipol, 6 for anticenp and antitopo and 14 for antipol and antitopo. Patients who could not be classified into an autoantibody subset because of missing autoantibody data were only included in the overall scleroderma cohort analyses. For all analyses, the timing of scleroderma onset was defined by the first scleroderma symptom, either Raynaud’s or non-Raynaud’s. Patient-reported cancer diagnoses and dates of diagnosis, obtained at the enrolment visit and during follow-up, were confirmed by medical record review and pathology reports if available.17 Electronic medical records were comprehensively reviewed to ensure that all cancer cases were captured during follow-up.
Examination of cancer risk in scleroderma compared with the general population
Cancer risk was determined by comparing cancer incidence in our cohort with the Surveillance, Epidemiology and End Results (SEER) registry, a nationally representative sample of the US population. Cancer incidence was examined in the overall scleroderma cohort, and in autoantibody and cutaneous subsets. We computed standardised incidence ratios (SIR) for cancer overall and individual cancer types. Our cancer subtype analyses focused on breast and lung cancers as these are the most prevalent cancers in scleroderma, but other cancer sites were also examined (see online supplementary appendix 1). The observed number of cancers in our cohort was compared with the expected number of cancer cases for the US population by identifying the crude rate of incident cancers corresponding to each patient’s age (within 5-year intervals), gender, race, ethnicity and the calendar year of exposure in SEER.28 Person time prior to 1973 was not examined as SEER data began in 1973. At the time of analysis, SEER data were complete through 2014. SEER crude rates for 2014 were used as a surrogate for person time after 2014. The sum of the crude rates for all years of exposure for all patients yielded the expected number of cancer cases. To find the 95% confidence limits, we followed standard procedure.28 29
Supplemental material
Because we are interested in cancer diagnosed close to the time of scleroderma onset (defined as time zero) that may be suggestive of cancer-induced autoimmunity, we examined two time windows for our primary analyses: (i) 3 years before scleroderma onset until cancer diagnosis date or the last visit date (termed ‘overall cancer risk’ during follow-up) and (ii) 3 years before scleroderma onset until 3 years after scleroderma onset (±3 years, ‘cancer-associated scleroderma’). Patients with cancers preceding these time windows were excluded from our analysis. Administrative censoring occurred at the cancer diagnosis date or last visit date, whichever came first. The study population for our primary analyses comprised 2383 patients with scleroderma.
Since including individuals with cancers diagnosed a few years before joining the cohort may introduce a form of immortal person time bias, we performed two additional analyses restricting our study population to patients who presented to our centre within 5 years of their first scleroderma symptom (‘recent-onset scleroderma’). In the first analysis, we only included cancer diagnoses that occurred after the first visit to our centre. As referral to a tertiary centre is often delayed, we also performed an analysis involving patients with recent-onset scleroderma and examined cancer diagnoses after scleroderma symptom onset. This time point better reflects presentation to a community rheumatology practice.
Finally, we graphically examined cancer risk over time (starting 3 years before scleroderma onset) in patients with scleroderma compared with the general population. The expected cancer incidence was computed using SEER data for each patient-year of exposure. Observed and expected numbers of cancer cases, and the corresponding SIR, were plotted in 6-year time windows (ie, ±3-year increments with time zero denoting scleroderma onset). For each patient, cancer risk exposure ended on the date of cancer diagnosis, last visit or at the end of the 6-year window. The cumulative incidence of cancer was also plotted over time for patients with scleroderma overall and in each autoantibody subgroup.
Analyses were performed using MATLAB R2016b (MathWorks, Natick, Massachusetts, USA) and R V.3.4.0 (R Foundation, Vienna, Austria). Bonferroni adjustment for multiple (10) comparisons was performed, as each time window and tumour type had 10 autoantibody-subtype comparisons. Therefore, P≤0.05/10 or P≤0.005 was considered to be statistically significant.
Results
The study population for our primary analyses consisted of 2383 patients with scleroderma contributing 37 686 person-years (table 1). The mean age at scleroderma onset was 42.4±15.1 years. Sixty per cent of patients had limited scleroderma, 83% were female and 76% self-identified as white race. Among the 1712 patients with autoantibody data, 608 (35.5%) were positive for anticenp, 481 (28.1%) for antitopo and 278 (16.2%) for antipol; 379 patients (22.1%) were CTP-negative. An additional 671 patients could not be classified into an antibody subset because of missing data. Approximately 9% of patients (205/2383) had a history of cancer (see online supplementary appendix figure 1 for tumour sites). Additional scleroderma characteristics of this population are detailed in online supplementary appendix table 1.
Determination of cancer risk relative to the general population: all cancers
Patients with diffuse scleroderma did not have an increased risk of cancer (SIR 1.12, 95% CI 0.88 to 1.39; table 1, overall cancer risk). In contrast, an increased risk of cancer was observed among antipol patients with diffuse disease (SIR 2.05, 95% CI 1.44 to 2.84). Patients with limited scleroderma had a 30% lower risk of cancer (SIR 0.70, 95% CI 0.59 to 0.84), and this was notable in anticenp-positive patients (SIR 0.59, 95% CI 0.44 to 0.77).
Next, we sought to determine the risk of cancer within 3 years of scleroderma onset (‘cancer-associated scleroderma’) compared with individuals in the general population. While patients with limited scleroderma did not have an increased risk of cancer-associated scleroderma (table 1, ±3 years), patients with diffuse scleroderma had a 56% increased risk compared with the general population (SIR 1.56, 95% CI 1.13 to 2.10). This risk increase was notable among antipol patients with diffuse disease (SIR 3.13, 95% CI 2.03 to 4.62). Additionally, CTP-negative patients with limited scleroderma had an increased risk of cancer-associated scleroderma (SIR 2.43, 95% CI 1.36 to 4.00).
The increased risk of cancer at scleroderma onset among antipol-positive and CTP-negative patients is illustrated in figure 1. The number of cancer cases observed around the time of scleroderma onset (top row, blue curve) is greater than the number of expected cancer cases based on SEER data (red curve) in these two autoantibody subsets. The relative risk of cancer compared with the general population is presented in time-dependent SIRs (middle row) and was increased for antipol-positive and CTP-negative groups close to scleroderma onset. The cumulative incidence of cancer was significantly higher among antipol patients (blue lines, blue dashed lines 95% CI) compared with that expected in the general population (red line) (bottom row, figure 1). In contrast, the cumulative incidence of cancer was lower than expected in the anticenp group.
Cancer risk in patients with recent-onset scleroderma
We performed two additional analyses restricting our study population to patients who presented to our scleroderma centre within 5 years of their first scleroderma symptom and examined cancer diagnoses (i) after first visit to our tertiary referral centre or (ii) after the first scleroderma symptom. Our findings of an increased risk of cancer among antipol-positive patients with diffuse scleroderma remained unchanged in both analyses, although in these restricted analyses this was statistically significant only after first symptom when adjusting for multiple comparisons (see online supplementary appendix table 2).
Breast cancer
Antipol-positive patients with diffuse scleroderma had an increased risk of breast cancer overall (SIR 3.06, 95% CI 1.75 to 4.98) and within 3 years of scleroderma onset (SIR 5.14, 95% CI 2.66 to 8.98; table 2). Within 3 years of scleroderma onset, CTP-negative patients with limited disease also have an increased risk of breast cancer (SIR 4.44, 95% CI 1.92 to 8.74). The marked increased risk of breast cancer at scleroderma onset in these two autoantibody subsets is illustrated in figure 2 (top and middle rows). The cumulative incidence of breast cancer is significantly higher among antipol patients compared with the general population (bottom row, figure 2).
Conclusions
In this investigation, we used autoantibodies, cutaneous subtype and temporal clustering as biologically relevant filters to investigate cancer risk and type in patients with scleroderma compared with the general population. We made several novel findings that, if confirmed by others, will inform our approach to early cancer detection in scleroderma, and also provide additional insights into mechanistic connections between cancer and scleroderma. First, while patients with scleroderma did not have an increased overall risk of cancer compared with the general population, antipol-positive patients with diffuse scleroderma and CTP-negative patients with limited scleroderma are at increased risk for cancer at scleroderma onset. Second, scleroderma patients with antipol antibodies may have increased risk of different types of cancers depending on whether they have limited or diffuse cutaneous disease. Third, patients with anticenp antibodies may have a decreased risk of cancer. Overall, these data suggest that autoantibodies could be useful tools for cancer risk stratification to maximise detection of cancer through enhanced screening of high-risk groups, while minimising the harms and costs from overscreening.
Prior studies investigating cancer incidence in patients with scleroderma relative to the general population have used study populations from national centralised registries or cohorts similar to ours.1–6 8–10 30 However, most of these studies excluded patients with cancer diagnoses shortly before scleroderma onset, and many lacked data on autoantibody status or phenotypic subtype. In contrast, we studied cancer risk within (±) 3 years of scleroderma onset, and investigated whether this varied by autoantibody specificity, cutaneous subtype and cancer type. We included the time before scleroderma diagnosis because our recent data demonstrated that POLR3A is genetically altered in short-interval cancers associated with an immune response to that protein, where both mutation-specific and cross-reactive immune responses were seen.13 These data strongly support a biological model in which cancer precedes scleroderma, and initiates a scleroderma immune response and clinical disease.13 14 The fact that 27% of patients with scleroderma with antipol antibodies and cancer have cancer shortly preceding scleroderma onset highlights the frequency of this subgroup, and the importance to include them. Our current study demonstrates that the risk of cancer around the time of scleroderma onset in antipol-positive patients is manyfold higher than that expected in the general population, supporting the idea that these patients may require more aggressive cancer screening at disease onset.
Interestingly, our data suggest that cancer risk may differ among antipol patients depending on their cutaneous subtype, as those with diffuse scleroderma had a higher risk of breast cancer and those with limited scleroderma may have an increased risk of lung cancer. These findings, particularly for lung cancer, require validation in other scleroderma cohorts given the small numbers of lung cancer cases in each autoantibody-subtype stratum. While prior studies have identified an increased risk of breast cancer concomitant with scleroderma onset in antipol-positive patients, these studies included patients with scleroderma without antipol as comparator groups, limiting the ability to determine excess risk compared with the general population.15 31 Our data suggest that enhanced breast cancer screening, incorporating sensitive measures such as MRI, may be warranted in antipol-positive women with diffuse scleroderma, but this needs further evaluation. Our exploratory analyses, if confirmed in other cohorts, suggest that antipol-positive patients may also require increased vigilance in monitoring for lung, tongue and prostate malignancies (see online supplementary appendix 1).
Our prior work demonstrated that CTP-negative patients may also be at risk of cancer-associated scleroderma.17 20 21 In this study, CTP-negative patients with limited scleroderma had an increased risk of breast cancer and melanoma (see online supplementary appendix 1) at scleroderma onset, suggesting that vigilance for breast cancer and comprehensive skin examination is most important. Of note, the CTP-negative group is likely heterogeneous, with several novel unrecognised immune responses.17 20 21 Identifying distinct autoantibodies in this subgroup (eg, anti-RNPC3)20 21 associated with an increased risk of cancer could facilitate development of a cancer risk prediction model in scleroderma.
Our study showed for the first time that patients with scleroderma with anticenp antibodies may have a substantially decreased risk of cancer compared with the general population. This unexpected finding possibly explains the different cancer risks observed in scleroderma cohorts internationally because the ratios of anticenp-positive to antipol-positive patients in cohorts dramatically impact the blended cancer risk. The finding that distinct serologic subgroups have different cancer risks suggests that cancer immunity may be a common principle across the scleroderma spectrum, with cancer emergence influenced by the different immune responses such that for anticenp, cancer emergence may be inhibited, while inhibition is only partial for antipol. Prior studies in small cohorts of patients with breast cancer have demonstrated that anticenp antibodies may be present and associate with improved disease-free and overall survival.32–34 Intriguing recent data also suggest that anti-DNA antibodies can have direct anticancer effects in cells with DNA repair defects,35 possibly explaining the decreased risk of breast and other cancers among patients with systemic lupus erythematosus.36 While it is possible that anticenp immune responses exert a similar anticancer effect in scleroderma, other possibilities exist, and mechanistic studies are needed.
This was a prospective study using a large, well-defined scleroderma cohort to investigate whether scleroderma-specific immune responses and clinical phenotypes associate with a higher risk of certain cancer types. These findings require validation in other scleroderma cohorts given the observational study design and smaller sample sizes in each subgroup as patients are divided into finer classification schemes. Our primary analyses focused on cancers that were detected up to 3 years before the clinical onset of scleroderma, as we were interested in cancer-induced autoimmunity. We recognise that including person time prior to the first visit to our centre raises concerns about immortal person time biases due to mortality from cancer diagnosis prior to presentation. To address this, we performed sensitivity analyses only including patients with recent onset scleroderma and examined cancer diagnoses after first visit to our centre. Our primary findings for antipol were similar. Our findings in the other autoantibody subsets were attenuated, likely due to decreased statistical power. Several patients were missing sufficient autoantibody data to be classified into a serologic subset; on average, these patients presented for their first visit 3 years before those who could be classified into an autoantibody category, suggesting a period effect due to limited availability of certain commercial autoantibody assays in earlier years. These differences may affect the generalisability of our findings. We do not think surveillance bias plays a major role in our findings, as historically all clinical cancer screening in our centre has been based on age and gender and was not influenced by scleroderma diagnosis or features. However, we recognise that incidental malignancies may be detected during testing performed for scleroderma; conversely, patients with early cancer or scleroderma may face a competing risk of death from either process before diagnosis of the other disease, resulting in an underestimation of cancer cases at the time of scleroderma onset. Stratified analyses suggested that smoking and interstitial lung disease were effect modifiers for lung cancer risk (data not shown). Unfortunately, smoking information was unavailable in the SEER registry, limiting our ability to fully adjust for this risk factor. Lastly, although our prior biologic studies suggest that certain subsets of patients with scleroderma may have cancer-induced autoimmunity, we acknowledge that these data do not prove causality. The relationship between cancer and autoimmunity in scleroderma is likely complex and bidirectional, with many potential links between the two diseases including immunosuppressive therapies or damage from the disease triggering malignancy, or a shared genetic or environmental exposure.
These data suggest that segregation by clinical features and autoantibody response identify scleroderma subgroups with distinct risks of both overall cancer, and specific types of cancer. Application of these simple filters may be useful in designing studies that define guidelines for cancer detection in patients with scleroderma. Investigating the mechanistic basis for differences in cancer risk across scleroderma subgroups is likely to enhance our understanding of scleroderma, autoimmunity and cancer immunity.
Acknowledgments
The authors would like to thank Adrianne Woods and Margaret Sampedro for their excellent support in database management and quality control.
References
Footnotes
29 Ar and AAS are joint senior authors
30 TI, LKH and KV are joint first authors
Handling editor Josef S Smolen
Funding This study was supported by the NIH (K23-AR061439, P30-AR053503, P30-AR070254, R01- DE12354-15A1, T32-AR048522), the Donald B and Dorothy L Stabler Foundation, the Jerome L Greene Foundation, the Chresanthe Stauralakis Memorial Discovery Fund, the Martha McCrory Professorship and the Scleroderma Research Foundation.
Competing interests TI, LC-R, AR and AAS have recently submitted a patent application entitled ’Materials and Methods for Assessing Cancer Risk and Treating Cancer'.
Patient consent Not required.
Ethics approval Johns Hopkins Medical Institutions IRB.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement There are no additional unpublished data from this study at this time.