Systemic Lupus Erythematous and Malignancy Risk: A Meta-Analysis

Background Pilot studies have estimated cancer incidence in patients with systemic lupus erythematous (SLE). However, the results have been inconclusive. To ascertain the correlation between SLE and malignancy more comprehensively and precisely, we conducted a meta-analysis. Methods PubMed, the Cochrane Library and Embase databases through June 2014, were searched to identify observational studies evaluating the association between SLE and malignancy. The outcomes from these studies were measured as relative risks (RRs). A random or fixed effects model was chosen to calculate the pooled RR according to heterogeneity test. Between-study heterogeneity was assessed by estimating I2 index. Publication bias was assessed by Egger’s test. Results A total of 16 papers, including 59,662 SLE patients, were suitable for the meta-analysis. Of these papers, 15 reported RRs for overall malignancy, 12 for non-Hodgkin lymphoma (NHL) and lung cancer, 7 for bladder cancer, 6 for Hodgkin lymphoma (HL) and leukemia, 5 for skin melanoma, and liver and thyroid cancers, 4 for multiple myeloma (MM), and esophageal and vaginal/vulvar cancers and 3 for laryngeal and non-melanoma skin cancers. The pooled RRs were 1.28 (95% CI, 1.17–1.41) for overall cancer, 5.40 (95% CI, 3.75–7.77) for NHL, 3.26(95% CI, 2.17–4.88) for HL, 2.01(95% CI, 1.61–2.52) for leukemia, 1.45(95% CI, 1.04–2.03) for MM, 4.19(95% CI, 1.98–8.87) for laryngeal cancer, 1.59 (95% CI, 1.44–1.76) for lung cancer, 1.86(95% CI, 1.21–2.88) for esophageal cancer, 3.21(95% CI, 1.70–6.05) for liver cancer, 3.67(95% CI, 2.80–4.81) for vaginal/vulvar cancer, 2.11(95% CI, 1.12–3.99) for bladder cancer, 1.51(95% CI, 1.12–2.03) for non-melanoma skin cancer, 1.78(95% CI, 1.35–2.33) for thyroid cancer, and 0.65(95% CI, 0.50–0.85) for skin melanoma. Only the meta-analyses of overall malignancy, NHL, and liver and bladder cancers produced substantial heterogeneity (I2, 57.6% vs 74.3% vs 67.7% vs 82.3%). No apparent publication bias was detected except for NHL studies. Conclusions Our data support an association between SLE and malignancy, not only demonstrating an increased risk for NHL, HL, leukemia, and some non-hematologic malignancies, including laryngeal, lung, liver, vaginal/vulvar, and thyroid malignancies, but also a reduced risk for skin melanoma. Although an increased risk of MM, and esophageal, bladder and non-melanoma skin cancers was identified from the accumulated data in these studies, this observation requires confirmation.


Introduction
Systemic lupus erythematous (SLE) is one of the most common systemic autoimmune rheumatic diseases. It often affects multiple organ systems and has a broad range of clinical and laboratory manifestations [1][2][3] and occasionally co-exists with Sjogren's syndrome or other overlapping syndromes [4]. Systemic glucocorticoids are used alone or combined with other immunosuppressive or cytotoxic agents including methotrexate, cyclophosphamide, and azathioprine to treat patients with significant organ involvement or refractory symptoms [5,6]. Although survival in SLE has improved since the introduction of new biological drugs such as rituximab [5], there continues to be significant morbidity like cancer adversely affecting longterm outcome.

Search Strategy for Identification of Studies
We searched PubMed, the Cochrane Library and Embase databases through June 2014 for English articles using the following words (all fields), which were retrieved from all eligible articles and the references were reviewed to identify additional relevant studies: ['autoimmune diseases' OR 'systemic lupus erythematous' OR 'SLE'] and ['lymphoma' OR 'malignancy' OR 'cancer' OR 'neoplasm' OR 'tumor'] AND ['case-control' OR 'cohort']. control group; (5) relative risk (RR), standardized incidence rate (SIR), or standardized morbidity rate (SMR) with 95% confidence interval (CI) available (or data available for calculations). Reviews [3,7,[25][26][27] or editorials, letters to the editor without original data, and case reports were excluded. In the event of multiple publications from the same study or overlapping study populations, only the most relevant one was selected.

Data Extraction
Data extraction was conducted by one investigator (L-HC) and checked by two investigators (G-XX and PL). Discrepancies were resolved through team consensus. The collected data included general information (design type, author, published year, country, study period, and follow-up), cohort of SLE characteristics (gender, mean age at the time of SLE diagnosis, diagnostic criteria, exclusion criteria, use of immunosuppressive drugs, and adjusted variables), and results (number of participants, reference population, and RR, SIR, or SMR with corresponding 95% CI). The quality of each study was evaluated independently by the above three authors using the Newcastle-Ottawa scale [28](S1 Table).

Data Synthesis
The preferred method of data presentation for cohort studies are the calculated RRs compared with the general population, generally estimating the age-and gender-adjusted SIRs or SMRs. SIRs or SMRs are estimated as the ratio of the observed over the expected number of cases for an exposed population. The expected number is calculated as the product of gender-, age-, and calendar period-specific person-years of follow-up, and the corresponding incidence of malignancy in the general population. The 95% CI is calculated assuming a Poisson distribution for the observed cases.
All analyses were performed with Stata 12.0 software. Fixed (Mantel-Haenszel) or random effects (DerSimonian-Laird) models were used to calculate the pooled effect estimates [29]. Between-study heterogeneity was assessed by estimating Ι 2 (significance level at P <0.1) [52]. In the case of significant heterogeneity, irrespective of the Ι 2 estimation, random effects models were used. The Galbraith plot was used to detect potential source of heterogeneity [30]. We also performed subgroup analyses with data available. In addition, sensitivity analyses were carried out by excluding specific studies. Egger's tests were used to evaluate publication bias for all meta-analyses [31]. The presence of publication bias was ascertained if the P value was <0.05.

SLE and Risk of Overall Malignancy
Of all included studies, fifteen, involving 58,077 patients with SLE, estimated RRs for overall malignancy. As shown in Fig 2, RR of each study ranged from 0.89-2.60 and the pooled RR by random effects analysis was 1.28 (95% CI, 1.17-1.41), with substantial heterogeneity (Ι 2 = 57.6%). The Galbraith plot showed that three studies might be the major source of the heterogeneity (Fig 3), two conducted in Europe [9,23] and one in North America [13]. The regional  There wasn't heterogeneity between the two studies when it was less than 5 years since the diagnosis of SLE. The heterogeneity was moderate when it was 5-19 years. Considerable heterogeneity emerged when it was more than 20 years (Ι 2 = 87.7%). Egger's test was used to evaluate publication bias for the metaanalysis of overall malignancy. There was no apparent asymmetric distribution occurring in it (P = 0.055; Fig 4). When sensitivity analysis for the meta-analysis was executed, no significant change in pooled RR was found by sequential omission of individual studies, indicating that the result was stable and reliable.  [20]. Of them, one study [9] revealed a dramatically elevated RR (44.40; 95% CI, 11.90-111.00). The pooled RR of all studies was 5.40 (95% CI, 3.75-7.77), but with substantial heterogeneity (Ι 2 = 74.3%). The possible source of heterogeneity might be attributed to two studies [9,21], one performed in Finland and the other in the USA, according to the Galbraith plot (Fig 6). The regional subgroup analysis confirmed the European studies produced considerable heterogeneity (Ι 2 = 78.5%) while the North American studies didn't (  Fig 7). Sensitivity analysis confirmed that the risk estimate for NHL was stable and reliable.

SLE and the Risk for Other Site-specific Malignancies
As shown in Table 4   As for non-hematologic malignancies, an increased risk was discovered in lupus patients for laryngeal, lung, esophageal, hepatic, vaginal/vulvar, bladder, non-melanoma skin and thyroid malignancies while a decreased risk was uncovered for melanoma skin cancer (Table 4). Twelve studies estimated RRs for lung cancer, 7 for bladder cancer, 5 for hepatic, melanoma skin and thyroid cancers, 4 for esophageal and vaginal/vulvar cancers and 3 for laryngeal and non-melanoma skin cancers. The sample size of these meta-analyses ranged from 7,538 to 57,890.  Figs). The meta-analyses of esophageal, melanoma skin and thyroid cancers didn't produce heterogeneity. The heterogeneity in the meta-analyses of laryngeal, lung and non-melanoma skin cancers might not be important. There was moderate heterogeneity among the vaginal/vulvar cancer studies (Ι 2 = 39.6%), substantial heterogeneity among the liver cancer ones (Ι 2 = 67.7%), considerable heterogeneity among the bladder cancer ones (Ι 2 = 82.3%). The Galbraith plot showed that no study contributed to the heterogeneity among the liver cancer studies and one study [22] might be the major source of the heterogeneity among the bladder cancer ones. No publication bias was found in any non-hematologic malignancy. If the study [21] was excluded, the summary risk estimate of esophageal cancer would be 1.89(95%, 0.93-3.83). If the study [23] was omitted, the pooled RRs for bladder cancer (1.93; 95% CI, 0.97-3.84) and skin non-melanoma (1.35; 95% CI, 0.95-1.91) would lose their statistical significance, which suggested that these risk estimates were potentially unstable and unreliable. However, the risk estimates of the other non-hematologic malignancies were confirmed stable and reliable by sensitivity analyses. From Table 4, we observed not only a higher pooled RR for malignancies involving the oropharynx, stomach, colon, rectum, anus, pancreas, cervix, uterus, kidney, prostate and brain, but also a lower pooled RR for ovarian and breast cancers. However, they are statistically insignificant, indicating that these malignancies have no clear link with SLE, compared with the general population.

Discussion
The present study is by far an ever-comprehensive meta-analysis of SLE and cancer risk, providing more precise risk estimates of various malignancies including some rarer cancers such as laryngeal and esophageal cancer in SLE. It demonstrates that NHL, HL, leukemia, and laryngeal, lung, liver, vaginal/vulvar, and thyroid malignancies are more frequently observed in lupus patients than in the general population. In addition, there is a potentially increased risk of MM, and esophageal, bladder and non-melanoma skin cancers. Conversely, there is a reduced risk of melanoma skin cancer. Hence, the overall risk of malignancy in SLE is slightly elevated, compared to the general population.
Nevertheless, these findings may be influenced by some confounding factors. The subgroup analyses for overall cancer and NHL have shown that the risk for hospital-based cohorts is higher than population-based ones. A potential persuasive explanation is that hospitalized lupus patients have a more severe condition so that a greater number develop into a cancer. European and North American lupus populations are at almost the same risk for the development of a cancer as Asian populations but at a much lower risk to develop NHL. The SLE patients with ACR criteria confer a lower risk to develop a cancer, especially NHL, than those with ARA criteria, revealing that preliminary criteria(ARA, 1971) probably overestimates the true risk owing to lack of specificity [33]. From the subgroup analyses stratified by whether cancers diagnosed in the first year of SLE being excluded, it is suggested that the selection of early-occurring malignancies, especially NHL, with SLE-like symptoms potentially overestimate the true risk, too. Males have a higher risk to develop a cancer than females. However, the conclusion is unreliable due to a small sample of men, requiring more studies for confirmation. Regarding trends over SLE duration or follow-up, an increased risk is initially apparent, followed by trends for somewhat lower RRs, and sequentially a rebound. The highest risk estimate occurs during the first year since diagnosis of SLE or the beginning of follow-up probably because some SLE patients with symptoms of cancer are included in the original cohorts. The more elevated risk estimate occurs again during >20 years, which potentially contributes to more cases of SLE-associated malignancy developing later in the disease process [23]. Therefore, it is essential to undertake a longer follow-up period to evaluate the association more precisely. The cancer incidence in the age-specific groups is quite similar while the NHL incidence at the age of forty to fifty-nine is higher than the other age-specific groups. Nonetheless, additional studies are needed to provide stronger evidence due to limited studies. In all, whatever subgroup doesn't change the risk trends of overall malignancy and NHL at all.
There have been some hypothesized explanations for the differences in the risks of certain malignancies in lupus patients versus the general population. Significant concern is that the  baseline immune system defects in this disease, combined with exposures to cytotoxic and immunosuppressive medications, may increase malignancy susceptibility [25]. According to our study, lymphoma, especially NHL, is a common malignancy in lupus patients. Possible mechanism may be composed of several aspects. First, diffuse large B cell lymphoma, originating from activated lymphocytes, has been shown to be the most common NHL subtype arising in SLE, suggesting that chronic inflammation might heighten lymphoma risk in autoimmune diseases like SLE [34]. Second, translocations involving the juxtaposition of an oncogene adjacent to an important gene for immune cell function may favor the emergence of a lymphoma [35]. The probability of a translocation is proportional to the rate of lymphocyte proliferation. The autoimmunity of SLE may up-regulate lymphocyte proliferation [24]. This might explain some of the excess lymphoma risk in SLE. Third, the use of immunosuppressive agents may lead to lymphoma by direct mutagenesis or by disturbing immune surveillance [36]. Fourth, the persistence of EBV infection may be a common etiological pathway for both the SLE and the B-cell malignancy [20,37]. As to other hematologic malignancies, there is a 2.01-fold increase for leukemia risk. Persistent cytopenia, prominent dysplasia on bone marrow smears, and azathioprine treatment may be considered as possible triggers for the development of this disease [38]. The etiology of increased incidence for MM is uncertain, needing further investigation.
With respect to non-hematologic malignancies, lung cancer risk is increased by 59% in lupus patients compared to the general population. Two potential issues might explain the association. Firstly, it is confirmed that there is a case history of more pack-years in lupus patients developing lung cancer than the general population with lung cancer [39]. Moreover, tobacco smoking has been demonstrated not only to be an independent risk factor for lung cancer but to affect disease severity of SLE [40]. This might provides additional motivation for smoking cessation in patients with SLE. Secondly, systemic chronic inflammation derived from  SLE could lead to the presence of interstitial lung disease, which could cause lung cancer [41]. Regarding liver cancer risk, SLE patients experiences a 3.21-fold increase compared to the general population. It has been suggested that there is a higher prevalence of hepatitis virus infections in lupus patients possibly due to immunologic dysfunction and immunosuppressant therapy [42,43] and hepatitis B or hepatitis C infection has a potential association with primary liver cancer [44]. The elevated risk for vaginal/vulvar cancer in women with SLE may be ascribed to susceptibility of human papillomavirus (HPV) infection [45]. However, an increased risk for cervical cancer, another HPV-associated cancer hasn't been suggested in our study. It is probably because vagina/vulva cancer may also share a common inflammatory mechanism with SLE [46]. As for the increased risk for bladder cancer, cyclophosphamide is considered to be a traditional factor [22]. Viral infection is another potential cause [23]. In regard to the higher incidence of non-melanoma skin cancer in SLE, sun sensitivity of the skin [47,48] is an established reason [49]. The following mechanism might explain the excess occurrence of thyroid cancer in SLE. There is a direct association between SLE and thyroid autoimmunity [50]  and thyroid autoimmunity itself could increase thyroid cancer risk [51]. The mechanism underlying the higher risk for laryngeal and esophageal cancers and the lower risk for skin melanoma in SLE are not apparent, further studies required. There are two main limitations which should be considered when interpreting the data from the present study. First, the primary limitation is the heterogeneity especially in the metaanalyses of overall malignancy, NHL, and liver and bladder cancers. Although subgroup analyses have indicated that some confounding factors don't change the risk trends for overall malignancy and NHL, the other factors like SLE management strategies, especially cytotoxic drugs use, coexistence of overlap syndrome, and lifestyle factors such as smoking have been unavailable from some studies so that these subgroup analyses couldn't be conducted, which may have had an impact on the estimated accuracy. Second, even though no significant evidence is found by egger's test in all the meta-analyses with the exception of NHL, we are not able to completely exclude the publication bias because we didn't unveil unpublished data and only collected the published articles in English.

Conclusions
Despite the undeniable limitations, owing to a vast sample size, our study support an association between SLE and malignancy, not only demonstrating an increased risk for NHL, HL, leukemia, and some non-hematologic malignancies including laryngeal, lung, liver, vaginal/ vulvar, and thyroid malignancies, but also a reduced risk for skin melanoma. Although an increased risk of MM, and esophageal, bladder and non-melanoma skin cancers was identified from the accumulated data in these studies, this observation requires confirmation. The findings highlight the importance of follow-up and cancer screening in lupus patients. Future studies should focus on the underlying mechanisms related to the development of site-specific cancers, especially MM, and laryngeal, esophageal and melanoma skin cancers, in patients with SLE and warrant correlation with our epidemiologic data.