Erratum to: Incidence of malignancy in adult patients with rheumatoid arthritis: a meta-analysis

It has come to our attention that, in our recently published article on the incidence of malignancy in adult patients with rheumatoid arthritis [1], the risk of melanoma requires additional clarification. Indeed, the meta-analysis of the risk of melanoma among patients with rheumatoid arthritis compared with the general population showed a standardized incidence rate ratio of 1.23 (95 % confidence interval 1.01, 1.49). This finding was erroneously included in the list of malignancies for which there was no consistent trend in risk among patients with rheumatoid arthritis versus the general population. Instead, with our current, updated meta-analysis it should be considered that there is now a suggestion of a possible increased risk for melanoma. With respect to the Discussion section of our article, we would like to note that the summary standardized incidence ratio shows a modest increased risk for melanoma.


Introduction
Rheumatoid arthritis (RA) is a polygenic, multifactorial and chronic immune-mediated disease characterized by chronic joint inflammation, a predilection for development of joint damage and deformity, and extraarticular involvement [1]. The management of RA includes the use of biologic and non-biologic disease-modifying antirheumatic drugs (DMARDs) [2,3], which act by directly modifying immunologic pathways involved in the pathogenesis of RA. The risk of malignancy among patients with RA has been of ongoing interest and research because of the autoimmune pathogenesis that underlies RA, the common etiology between rheumatic disease and malignancy, and the use of immunomodulatory therapy, such as DMARDs, that may alter normal immunosurveillance and elevate the risk of malignancy [4,5]. Understanding this potential therapeutic risk has become more relevant with the increasing use of biologic DMARDs as a routine therapeutic approach to RA management [3]. With more biologic treatment options available and initiation of biologic treatments occurring earlier, it is important to understand the baseline risk of malignancies in patients with RA. Furthermore, continuous updates on the incidence of malignancies published in the literature are crucial to allow better understanding of the background rates for malignancy in clinical trials and observational research evaluating real-world practice.
Smitten et al. [6] performed a meta-analysis on the risk of overall malignancy and several site-specific cancers, including overall lymphoma, Hodgkin disease, non-Hodgkin lymphoma, lung cancer, colorectal cancer and breast cancer, using data published between 1990 and 2007. Their meta-analysis suggested a small overall increase in risk of malignancy, which was elevated for lymphoma, Hodgkin disease, non-Hodgkin lymphoma and lung cancer, but they found a decreased risk of colorectal and breast cancer in patients with RA compared with the general population [6]. The malignancies included were prespecified and based on the most frequently reported malignancies in an RA population.
In this article, we review the data on the incidence of malignancy reported since the Smitten et al. [6] metaanalysis and additionally evaluate the risk of other important site-specific malignancies, namely, melanoma, cervical cancer and prostate cancer, which have been a topic of discussion in recent publications [7][8][9].

Literature search
We conducted a search of the MEDLINE, BIOSIS Previews, Embase, Derwent Drug File and SciSearch databases for literature published between 1 January 2008 and 30 November 2014, using the following specified search terms: cancer or tumor, tumour, malign* and rheumatoid arthritis or RA, and epidemiolog* or inciden*, population, observation, retrospective or occurren*.

Exclusion algorithm
Predefined inclusion criteria were used to identify relevant observational studies that provided estimates of relative risk of malignancy in patients with RA compared with the general population. The same criteria as detailed by Smitten et al. [6] were applied. Studies were eligible for inclusion if they fulfilled the following criteria: (1) observational study design (including prospective, retrospective, epidemiologic, database, survey, registry, cohort and case-control), (2) reported malignancy outcomes in patients with RA and a general population, (3) enrolled more than 100 patients, (4) included only patients older than 18 years of age, (5) covered any geographic region and (6) were reported in English as a full-length publication. Citations meeting the inclusion criteria were obtained and screened for the outcomes of interest, which included the observed incidence rates of total malignancy and lymphoma, lung, colorectal, breast, melanoma, prostate and cervical cancers in patients with RA compared with the expected incidence rates in the general population. Lymphoma was reported as Hodgkin or non-Hodgkin where available. The selection of studies for inclusion was made without regard to evaluation of specific RA management strategies. We attempted to avoid overlap by excluding studies for which updated publications were available. We also excluded studies that compared specific RA biologic treatments with other RA treatments, unless overall RA and general population rates were reported.

Data presentation
Citations meeting the inclusion criteria were screened for outcomes of interest. Risk data on overall malignancy and site-specific malignancies (overall lymphoma, Hodgkin disease, non-Hodgkin lymphoma, melanoma, and lung, colorectal, breast, cervical and prostate cancers) were extracted independently by two people and compared for accuracy. Study-specific estimates of relative risk were measured mostly by overall and age-and sex-adjusted standardized incidence ratios (SIRs) relative to non-RA patients or the general population. In one study using a case-control approach, researchers reported an odds ratio, which is an accurate estimator of the relative risk [10]. Thus, all measures of relative risk and their respective 95 % confidence intervals (CIs) were then compared with previously published rates. Pooled relative risks and 95 % CIs were then computed using the DerSimonian and Laird technique [11], in which a random-effects model takes into consideration both within-study and between-studies variation by incorporating the heterogeneity of the effects in the overall analysis. Some studies in which authors reported data separately for colon and rectal cancer, or for men and women, were considered separate studies when we pooled relative risks.

Overall malignancies
In general, SIRs for overall malignancy across the studies were similar. Six of seven studies reported a significant increase in overall risk of malignancy ( Fig. 2)

Colorectal cancer
In seven studies [12,[14][15][16][17]19] Fig. 3 Relative risk of malignant lymphoma in patients with rheumatoid arthritis (RA) compared with the general population. CI, confidence interval; n, number of malignancies; N, population size; OR, odds ratio; SIR, standardized incidence ratio; TNF, tumor necrosis factor. *SIRs by sex are not included in the total pooled SIR. † Reported as odds ratio  Fig. 2 Relative risk of overall malignancy in patients with rheumatoid arthritis (RA) compared with the general population. CI, confidence interval; DMARD, disease-modifying antirheumatic drug; n, number of malignancies; N, population size; RR, relative risk; SIR, standardized incidence ratio; TNF, tumor necrosis factor. *SIRs by sex are not included in the total pooled SIR. † Excluding non-melanoma skin cancer. ‡ All solid tumors. § Excluding lymphatic and hematopoietic plus those included here was found to be 0.86 (0.73-1.01) for breast cancer.

Discussion
Although individual studies can show considerable variation, the results of this updated meta-analysis of the incidence of malignancy in adult patients with RA are generally consistent with previously reported data for all specific malignancies [6]. The updated data reported here show that patients with RA have a modest increased risk  Fig. 4 Relative risk of Hodgkin disease in patients with rheumatoid arthritis (RA) compared with the general population. CI, confidence interval; n, number of malignancies; N, population size; MTX, methotrexate; SIR, standardized incidence ratio. *SIRs by sex for Parikh-Patel et al. [16] are included in the total pooled SIR, as overall SIR was not available in their study  Fig. 5 Relative risk of non-Hodgkin lymphoma in patients with rheumatoid arthritis (RA) compared with the general population. CI, confidence interval; DMARD, disease-modifying antirheumatic drug; n, number of malignancies; N, population size; SIR, standardized incidence ratio. *SIRs by sex included in total pooled SIR only if overall SIR was not available of overall malignancy as well as an increased risk of lung cancer and lymphoma when compared with the general population. Lymphoma was associated with the greatest relative risk, especially Hodgkin disease, with a 12-fold increase reported in one study [15] and an increase in women versus men in another [16]. Lung cancer generally showed a twofold increase in risk [10,[12][13][14][15][16][17]19]. As previously noted, a decreased risk overall of colorectal cancer in patients with RA compared with the general population was found. For melanoma, the pooled SIR was significantly greater than unity, although few of the individual point estimates were statistically significant. Breast and cervical cancers in general showed a decrease in risk overall that failed to reach statistical significance. Various explanations for the differences observed in the risk of certain malignancies in patients with RA have been proposed. It is thought that ongoing immunologic stimulation over time may increase the risk of malignant transformation of immune system cells and decrease the number of T-suppressor lymphocytes, thus increasing rates of lymphoma malignancy in patients with RA [4]. Chronic lung inflammation due to the disease in patients with RA may explain the increased risk of lung cancer.
Smoking increases the risk of both RA and lung cancer; however, the SIRs reported here were not adjusted for smoking, so it is not possible to rule out the increase in SIR being due to an indirect association. Thus, the risk of both lymphoma and lung cancer has been hypothesized to be dependent on the level of disease activity experienced by the patient [4,6].
This review, like that of Smitten et al. [6], is based on observational, real-world clinical data; however, these individual studies may have potential limitations. Although most of the individual estimates were derived from similar database studies, their study designs (including populations, data sources, follow-up times, data analysis and procedures of reporting, particularly when linked to a cancer registry) were a source of heterogeneity among the studies and may be associated with bias of varying magnitude. In addition, other potential sources of heterogeneity included geography and/or region, stage and severity of RA, and treatment. Such heterogeneity should be considered in any meta-analysis; the DerSimonian and Laird technique used in this study [11] assumes heterogeneity across studies and incorporates it into the estimation of the pooled SIRs and their confidence limits. Therefore, although the variation observed among the individual SIRs reported should be treated with caution when interpreting these data, it should be recognized that this variation was accounted for in the analyses resulting in the pooled SIRs.
For example, with regard to treatment, Dreyer et al. [13] evaluated the effect of tumor necrosis factor (TNF) inhibitor treatment on risk rate by examining the treated and non-treated populations separately (only non-treated data included in this analysis), whereas Mercer et al. [15] evaluated biologic-naive patients. Buchbinder et al. [19] Fig. 10 Relative risk of cervical cancer in patients with rheumatoid arthritis (RA) compared with the general population. CI, confidence interval; DMARD, disease-modifying antirheumatic drug; n, number of malignancies; N, population size; SIR, standardized incidence ratio; TNF, tumor necrosis factor Fig. 11 Relative risk of prostate cancer in patients with rheumatoid arthritis (RA) compared with the general population. CI, confidence interval; n, number of malignancies; N, population size; SIR, standardized incidence ratio. *Reported as odds ratio examined malignancy risk only in patients with RA receiving methotrexate. However, the rest of the patients in these studies are likely to have received treatments for RA, therefore making it difficult to differentiate the effects of treatment and disease. Nonetheless, to further understand which malignancies may be related to treatment versus the underlying disease, it is important, where possible, to quantify differences in malignancy rates between patients with RA treated with non-biologic DMARDs and those treated with biologic DMARDs. Whether biologic agents, predominantly TNF inhibitors, increase the risk of malignancy has been debated [9]. Data have shown that use of TNF inhibitors was not associated with a major further increase in risk in the already elevated lymphoma occurrence in patients with RA [39][40][41][42][43][44]. However, they do appear to increase the risk of skin cancer, including melanoma [39][40][41][42][43]. Furthermore, the data presented by Simon et al. suggest that the observed incidence of total malignancy and those for breast, colorectal and lung cancers and lymphoma in patients treated with abatacept was largely consistent with that expected of patients with RA treated with non-biologic DMARDs [45].

Conclusions
The additional data presented here are consistent with previously reported data [6] for all specified malignancies. For overall malignancy, the new studies analyzed here show a similar increase in risk compared with that previously reported. In addition, patients with RA continue to be at an increased risk of lung cancer and lymphoma when compared with the general population. Further studies examining specific aspects such as treatments, smoking or other lifestyle factors are needed to investigate the underlying mechanisms for the increased or decreased risk of specific cancers observed in patients with RA compared with the general population. Also, it is important to quantify any differences in malignancy risks between patients with RA who are treated with non-biologic DMARDs and those treated with biologic DMARDs, as understanding which malignancies may be related to the treatment as opposed to being related to underlying disease will allow patients to be advised and monitored accordingly.
Competing interests TAS and KKG are employees and stockholders of Bristol-Myers Squibb. AT is an intern at Bristol-Myers Squibb and a medical student at Rowan University School of Osteopathic Medicine, Stratford, NJ, USA. MCH receives grant/ research support from the National Institutes of Health and is a consultant for Bristol-Myers Squibb, Eli Lilly, EMD Serono, Genentech/Roche, Novartis Pharma AG, Pfizer and UCB. SS is a consultant for Bristol-Myers Squibb and Genentech/Roche. The authors declare that they have no non-financial competing interests.
Authors' contributions TAS participated in the design and coordination of data collection, contributed to analysis and interpretation of data and helped to draft the manuscript. AT conducted the literature search and helped to draft the manuscript. KKG contributed to the study design and data collection and interpretation and helped to draft the manuscript. MCH contributed to the study design and interpretation of the data and helped to draft the manuscript. SS contributed to the study design, performed the meta-analysis, contributed to data interpretation and helped to draft the manuscript. All authors read and approved the final manuscript.