Cancer in an SLE Inception Cohort: Smoking May out-Perform Tumor Markers As a Risk Predictor

Abstract

Background/Purpose: We assessed incident cancers in a large inception SLE cohort, and examined demographic and clinical factors, including tumor-related autoantibodies against proliferating cell nuclear antigen (PNCA) and centromere protein B (CENP-B).

Methods: Patients with new-onset SLE were enrolled at 32 centres. Incident cancers were recorded by physicians at annual visits, confirmed by chart review. Of 1,848 patients enrolled from 2000-2011, we studied 1,676 who had at least one follow-up visit. Multivariate survival regression was performed, with baseline demographics (age, sex, race), drugs (corticosteroids, anti-malarial drugs, immunosuppressives), smoking, SLEDAI-2K, and calendar year. Serology was available in 29 cancer cases and 1114 cancer-free patients. Subjects were followed until death, last visit, or end of the interval for this analysis (August 2015).

Results: Of the 1,676 patients evaluated, the majority (88.7%) were female, and 828 (49.4%) were Caucasian. Mean age at SLE diagnosis was 34.6 (standard deviation 13.3) years. At baseline, 1085 (64.7%) patients were never-smokers, with the remainder being current (n=248) or ex-smokers (n=342). Mean follow up was 7 years,

We observed 45 cancers in 45 subjects, occurring at a mean SLE duration of 4.6 (SD 3.1) years. Cancers included breast (9), non-melanoma skin (8), lung (5), prostate (5), head and neck (4), hematologic (3) cervical (3), thyroid (2), melanoma (2) and one each of meduloblastoma, renal carcinoma, carcinoid, thymoma, and dermatofibrosarcoma. Baseline anti-PNCA and anti-CENP-B antibodies were respectively found in 80 (7.2%, 95% CI 5.8, 8.9) and 26 (2.2%, 95% CI 1.6,3.4) of cancer-free patients, and none of the 29 cancer cases where serology was tested (including all 5 lung cancers).

Univariate hazard regression suggested that across all cancers, events were more common in whites and older patients; similar trends were evident in the multivariate analyses, but 95% CIs around the adjusted hazard ratio, HR, estimates were relatively wide. For lung cancer specifically, the adjusted hazard regression showed age and smoking were factors clearly associated with risk (adjusted HR for smoking 31.4, 95% CI 3.0, 327.1).

Conclusion: This was a first look at potentially predictive factors, including baseline levels of two tumor-associated antibodies, for cancer risk in a large inception SLE cohort. At a mean follow up time of 7 years, baseline smoking was more helpful in identifying future cancer events (specifically lung cancer), than baseline anti-tumor antibodies. However, additional work will assess a broader range of tumor-associated antibodies and follow-up over a longer period.