A meta-analysis of melanoma risk in idiopathic inflammatory myopathy patients

Two independent authors (Luo and Xia) searched three English databases (PubMed, Embase, Web of Science) and three Chinese databases (CNKI, VIP, and Wanfang) for studies on IIM related to tumors published up to October 2023. A combination of MeSH terms and free words was used. The search strategy combined terms ‘idiopathic inflammatory myopathy,’ ‘dermatomyositis,’ ‘polymyositis,’ or ‘myositis’ with ‘cancer,’ ‘malignancy,’ ‘tumor,’ ‘tumour,’ ‘neoplasm,’ or ‘carcinoma.’ The term ‘melanoma’ was then searched in all the full texts. This meta-analysis was performed according to the guidelines specified by PRISMA [12] and registered on INPLASY under registration https://​doi.​org/​10.​37766/​inplasy2023.​11.​0031.

The following inclusion criteria were applied: 1) written in English or Chinese; 2) type of study: cohort study reporting estimates with corresponding confidence intervals (CIs); 3) study population: patients with IIM combined with tumors; 4) outcome indicators: occurrence of tumors as one of the observed outcomes. Studies were excluded if 1) the study population was a repetitive population; 2) specific raw data indicators were unavailable; 3) the type of literature was a conference, case report, review, lecture, abstract, and so on.

Two authors (Luo and Xia) independently read all literature sources and full texts. A pre-designed extraction Microsoft Excel file was used for the following information: first author, year of publication, type of study, region, period of study, number of people followed, diagnostic criteria, risk estimate such as relative risk ratio (RR), hazard ratio (HR), or standardized incidence ratio (SIR), with corresponding 95% confidence interval (95% CI). In case of any disputes, mutual discussion was organized with a third researcher.

The Newcastle–Ottawa Scale (NOS) was used for the quality assessment of non-randomized controlled studies. Eight items were categorized into three dimensions, which include selection, comparability, and outcome.

The extracted data were processed using Stata software (version 16.0; Stata Corp LLC, College Station, TX, USA). The measure of interest was regarded as a continuous variable. The confidence intervals for each effect size were used with a 95% CI. We plotted a funnel plot to show the publication bias of studies: symmetry indicated no publication bias, and asymmetry publication bias requiring further examination by Begg’s correlation test and Egger’s linear regression.

The heterogeneity test was done by means of Q statistic and I² test. Regarding the Q test, if P > 0.05, there was homogeneity among the studies; conversely, if P ≤ 0.05, there was heterogeneity. The size of heterogeneity was then quantitatively analyzed according to I² test: if I² < 50%, it indicated less heterogeneity in the included studies within the acceptable range; conversely, if I² ≥ 50%, there was more heterogeneity. The corresponding effect model was selected according to the result. If P > 0.05 and I² < 50%, the fixed effects model was used; on the contrary, if the heterogeneity was large (P ≤ 0.05 or I² ≥ 50%), the random effects model was used for meta-analysis. The following subgroup analyses were performed if the original articles provided enough data: 1) sex, 2) age, 3) subtypes of IIM.

Sensitivity analysis was performed to assess the robustness and reliability of the combined results of meta-analysis. In this study, each single study was excluded one by one. If the effect on the combined effect size is small after eliminating a single study, it means that the results are stable and reliable; if the effect on the combined effect size is large after eliminating a single study, it means that the results are unreliable and further analysis is needed.

A total of 1660 papers were examined, including 59 papers in PubMed, 569 in Web of Science, 523 in Embase, 297 in CNKI, 88 in VIP, and 124 in Wanfang, with 528 removed as duplicates. By reading the title and abstract, 1104 documents were removed, and 24 documents were removed by reading the full text. Finally, 4 studies were included in the study. The literature screening flow diagram is shown in Fig. 1.

The four included papers were all cohort studies written in English, with a total sample size of 4239 IIM patients in the studies. The general information of the included studies is listed in Table 1. As shown in the table, Sigurgeirsson’s study was the earliest one, published in 1992. It followed hospitalized individuals from the Swedish National Board of Health and Welfare from 1964 through to 1983. A total of 58 cancers were diagnosed in 396 PM patients and 94 in 392 DM patients. Only one melanoma was observed after the first dermatomyositis diagnosis. Later, Stockton et al. conducted a retrospective population-based cohort study with 286 DM and 419 PM from 1982 to 1996 in Scotland. The observed number of malignant melanomas was also one, but in PM, with a standardized incidence ratio (SIR) of 3.7 (95% CI 0.1, 20.4). Chen et al. led a large sample size study of 1012 DM and 643 PM patients from 1997 to 2007 in Taiwan, China. In total, three melanoma cases were observed, one in DM with SIR 4.33 (95% CI 3.54 to 5.29) and two in PM with SIR 9.76 (95% CI 8.66 to 11.49). Dani in Sweden performed a nationwide study between 2002 and 2016. A total of 1181 patients with IIM and 6194 non-IIM comparators were followed: 14 cases of melanoma were found after IIM diagnosis, with crude incidence rate (1000 person-years) of 1.8/0.6 (IIM/non-IIM) and adjusted hazard ratio (AHR) of 3.2 (1.6–6.4).

The quality of the included studies was evaluated according to the Newcastle–Ottawa Quality Rating Scale. The results of the quality evaluation for each study are shown in Table 2. All four studies were of high quality, with a score of more than 5.

Four cohort studies involving 4239 IIM patients were analyzed for risk. Two studies providing RR and AHR, and another two studies providing SIR were pooled separately. The pooled overall RR/HR was 3.08 (95% CI 0.79–5.37) and the pooled overall SIR was 6.30 (95% CI 1.59–11.02), with low heterogeneity (I² = 0%, P > 0.001) using a fixed effects model (See Fig. 2a, b). A total of 19 cases of melanoma were observed without definite age and sex information or IIM subtype, so we could not perform further corresponding subgroup analyses.

The funnel plot drawn was symmetric in appearance, indicating a lack of publication bias.

Melanoma has not been frequently investigated with IIM. To the best of our knowledge, this is the first meta-analysis that takes into account this most dangerous type of skin cancer. The included cohort studies are more reliable compared to case–control studies. The four studies were reported in different decades and regions, making results more representative. Table 1 showed the diagnosed criteria in each study. Moreover, the time of development of cancer was also considered, which means that all the melanoma cases were observed after the diagnosis of IIM, not before.

The first limitation of the study is that the number of included studies is not large. Due to the small number of 19 melanoma cases observed, there was no definite age and gender information, so we could not perform further subgroup analyses. At our initial plan, Chinese studies were searched due to the large population; however, there seemed to be a lack of high-quality cohort studies in China.