Background/Purpose: We had previously proposed a selection bias in studies estimating cancer risk in patients with rheumatoid arthritis (RA) stemming from the comparison of a selected subpopulation to the general population (Arthritis Rheum. 2011;63;2543-4). Biased estimates would be the result of a) death of a part of the subpopulation due to cancer before being selected, b) prevention of an autoimmune/inflammatory condition by cancer treatment and c) a lower probability of being selected/detected merely because of having cancer. These would cause the rate of accumulation of cancer cases in the selected subpopulation to be lower than that in the general population  and hence a decrease in the comparative incidence if tracked over time. We had also demonstrated such a decrease in the comparative incidence over time in studies reporting cancer risk in RA patients (Ann Rheum Dis 2012;71(Suppl3):456). We conducted a systematic literature search to see whether such a bias exists in studies reporting the cancer risk in other autoimmune/inflammatory disorders.

Methods: We conducted multiple PubMed searches using the search terms “polymyositis”, “dermatomyositis”, “sarcoidosis”, “psoriasis”, “Crohn’s”, “ulcerative”, “lupus”, “SLE”, “Sjogren”, “ANCA”, “polyarteritis”, “Wegener”, “polyangiitis”, “vasculitis”, and “scleroderma” between June 2001 and February 2012. We combined these terms with “cancer” and “standardized” to capture studies that reported standardized incidence ratios or standardized morbidity ratios. We retrieved full-text manuscripts of studies that reported comparative incidence. Studies that reported multiple incidence ratios with respect to follow-up time were included.

Results: Our search identified 192 articles and we retrieved 36 articles of relevance in full-text. Among the 36 articles retrieved, 10 reported incidence comparison of overall cancer at multiple timepoints. There were 3 studies on inflammatory myopathies, 1 study each with scleroderma, Wegener’s granulomatosis, ulcerative colitis, Crohn’s disease, polymyalgia rheumatica, SLE and sarcoidosis. Number of timepoints ranged from 2 to 10, sampling period ranged from 10 to 48 years. The comparative incidence at the latest timepoint was lower in 9 of the 10 articles by 56 to 98% as compared to the earliest.

Conclusion: We provide further evidence that current methods of comparision of cancer incidence in a selected subpopulation to the general population results in biased estimates.

---

« Back to 2012 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/further-evidence-on-biased-cancer-risk-estimation-in-studies-comparing-a-subpopulation-to-the-general-population/