Meta-analysis of the association of dermatomyositis and polymyositis with cancer


  • Funding sources None.
  • Conflicts of interest None declared.


Guimei Guo.


Liming Bao.




Although some features of dermatomyositis (DM) and polymyositis (PM) have been reported as possible prognostic indicators for cancer development, previous studies were small in size and were unable to establish a definitive relationship between neoplasms and DM and PM.


To evaluate risk factors for developing malignancies in patients with DM and PM.


Meta-analysis of the studies reported in the literature was performed to unveil risk factors for developing cancer among patients with DM and PM. The included studies were either cohort or retrospective case–control studies with information on cancer status. Risk for malignancy was determined as the odds ratio (OR) or weighted mean difference (WMD) with a 95% confidence interval (CI), determined by fixed and random effects models. Stata 10.0 software was used to identify possible publication bias.


Twenty studies with 380 patients and 1575 controls were included in the analysis. The factors that may increase the risk of cancer in patients with DM and PM were older age (WMD 11·41, 95% CI 9·84–12·98), male sex (OR 1·92, 95% CI 1·49–2·48), cutaneous necrosis (OR 5·52, 95% CI 3·49–8·74) and dysphagia (OR 2·41, 95% CI 1·50–3·86), whereas those that may provide protection against cancer included arthritis (OR 0·38, 95% CI 0·24–0·61) and interstitial lung disease (OR 0·32, 95% CI 0·20–0·51).


Our data suggest that age, sex, cutaneous necrosis, dysphagia, arthritis and lung complications may influence susceptibility to cancer in patients with DM and PM.

Dermatomyositis (DM) and polymyositis (PM) are rare autoimmune disorders with inflammatory myopathies, the main clinical features of which include progressive weakness of the proximal muscles and, in the former, the development of cutaneous lesions. An association of both DM and PM with cancer has been suggested, with malignancies being more frequent in DM than in PM.[1, 2] Approximately 13–42% of patients with DM and 3–18% of patients with PM may develop cancer,[3] and malignancy is therefore considered to be a major complication in both conditions. Many tumours have been reported in patients with myositis, but ovarian and breast cancer, and lung cancer are the leading malignancies among female and male patients with DM, respectively.[4] Neoplasms may be present in patients either before, after or during diagnosis of myositis.

Although the increased incidence of cancer in patients with DM and PM is well established, the underlining pathogenesis remains unknown, partly because factors that might influence cancer susceptibility remain unidentified.[5] The following reasons have been considered probable for the association of myositis and malignant diseases:[6-8] (i) paraneoplastic conditions producing bioactive mediators that cause immune reactions against muscle fibres and skin; (ii) compromised immune systems leading to development of both tumours and myositis; (iii) neoplasms that are secondary to the cytotoxic effect of agents used in therapeutic regimens to treat myositis; and (iv) exposure to environmental hazards that are both carcinogenic and trigger immune reactions resulting in myositis.

Because of the poor outcome associated with malignancy, it becomes important clinically to identify patients who are at high risk of developing cancer.[9] Several features have been reported as possible prognostic indicators for cancer development, namely age, serological autoantibodies, human leucocyte antigen polymorphism, erythrocyte sedimentation rate (ESR), skin lesions and neoplastic markers.[10-14] Some studies have shown that the presence of necrotic skin ulcerations and pruritus, elevated ESR, periungual erythema and dysphagia increase the risk of developing cancer among patients with DM.[13, 15] However, previous studies to characterize such risk factors for malignancy in patients with DM and PM were small in size and the findings were not conclusive.[9, 10] Additional studies are needed to uncover factors that may underline cancer-associated myositis (CAM), including DM and PM. We undertook a meta-analysis in an attempt to identify these factors on the basis of previously published observational studies.

Materials and methods

Data sources and searches

We used the following terms – idiopathic inflammatory myopathy, dermatomyositis, polymyositis, tumour, cancer, neoplasm, risk factor and predictor – to search the databases, namely PubMed, Medline, Embase, Ovid, Springer, Cochrane Library, Institute of Scientific Information, China Bio-Medical Literature Database, China Wan Fang Database and the Chinese Science and Technology Database, up to January 2012. We reviewed relevant references cited in the original articles. Attempts were made to get additional information on a study reported in a conference abstract. The DM and PM studies included in the meta-analysis were either cohort or retrospective case–control studies with information available on cancer status.

Study selection and data extraction

The inclusion criteria for the study were (i) sufficient information to calculate odds ratios (ORs) and 95% confidence intervals (CIs) between a characteristic of interest and malignancy; (ii) patients aged ≥ 18 years with a probable or definitive diagnosis of DM or PM according to the Bohan and Peter criteria;[16, 17] (iii) all types of cancer; and (iv) at least one of the characteristics analysed in our meta-analysis. Two authors (J.W. and G.C.) conducted the database search independently. Discrepancies between any two search results were reconciled through reviews by a third author (G.G.). Information collected from selected studies included article title, author, publication date, regions in which the study was conducted, number of patients, age, sex, cutaneous necrosis, dysphagia, arthritis, interstitial lung disease (ILD), Gottron sign and antinuclear antibodies (ANA). Furthermore, the data quality of every selected study was evaluated by the Newcastle–Ottawa Scale[18] for subject groups, comparability and ascertainment of exposure (or outcome for case–control analysis). The selected studies were assigned as high, moderate or low methodological quality with scores > 6, 4–6 and < 4, respectively.

Statistical analysis

Meta-analysis was performed using RevMan software from the Cochrane Collaboration ( We present the pooled effects as a point estimation with the OR or WMD, 95% CI and the corresponding P-value (P < 0·05 considered significant) using binomial methods with a forest plot for primary outcome measure. Inter-study heterogeneity was assessed using the Cochrane Q-test; those with P < 0·05 were considered as having significant heterogeneity. The I2 statistic, which defines the percentage of total variation across studies due to heterogeneity rather than chance, was also calculated, with I2 > 50% suggesting significant heterogeneity. If the Cochrane Q-test showed significant heterogeneity, a random effects model was employed to calculate the OR and 95% CI. Otherwise, a fixed effects model was applied. Begg's funnel plot and Egger's test were employed to estimate the impact of possible publication bias, with P > 0·05 being considered as insignificant. Publication bias was examined using Stata 10.0 software (StataCorp, College Station, TX, U.S.A.).


Database search

Figure 1 shows the process for selecting the studies to be included in the meta-analysis. Of a total of 1100 articles identified in the databases, 1055 did not meet the study criteria and were excluded from the meta-analysis. Of the remaining 45 studies, 20 met the selection criteria,[5, 9, 10, 13, 19-34] and the remaining 25 were excluded due to not being original research, being case reports, or having inadequate study design.

Figure 1.

Flow diagram of the study selection procedure. CAM, cancer-associated myositis.

Study characteristics and quality assessment

The 20 studies selected for meta-analysis included a total of 1955 patients with DM or PM: 380 were reported as having CAM or an accompanying cancer, and 1575 patients with no features associated with CAM were considered as controls. Most patients with CAM had DM. The studies analysed the following characteristics: age (mean ± SD; n = 13),[5, 9, 10, 19-22, 26-29, 31, 32] sex ratio (n = 18),[5, 9, 13, 19-25, 27-34] cutaneous necrosis (n = 10),[5, 9, 10, 13, 21-23, 26, 33, 34] dysphagia (n = 12),[5, 20-22, 24, 26, 27, 29-33] arthritis (n = 7),[5, 20-22, 24, 30, 31] ILD (n = 10),[5, 9, 19-21, 24, 27, 30, 32, 33] Gottron sign (n = 8)[5, 20, 21, 23, 26, 30-32] and ANA (n = 9).[5, 19-22, 24, 26, 27, 31] The quality scores ranged from 5 to 8: 14 were high quality, six moderate quality and none low quality (Table 1).

Table 1. Studies included in the meta-analysis
StudyRegionAge (years), mean ± SD (n)Matched or adjusted variables analysedMyositis typeQuality scoreStudy design
  1. ILD, interstitial lung disease; ANA, antinuclear antibodies; DM, dermatomyositis; PM, polymyositis; NA, not available.

András 2008[5]Hungary57·11 ± 10·06 (28)43·25 ± 12·6 (79)Age of onset, sex, cutaneous necrosis, dysphagia, arthritis, ILD, Gottron sign, ANADM and PM7Cohort
Antiochos 2009[19]U.S.A.60·1 ± 11·4 (24)54·6 ± 16·7 (37)Age of onset, sex, ANA, ILDDM and PM7Cohort
So 2011[20]Korea62·4 ± 11·5 (25)46·9 ± 15·2 (126)Age of onset, sex, arthritis, dysphagia, ILD, ANA, Gottron signDM and PM7Cohort
Basset-Seguin 1990[10]France60 ± 14 (13)44 ± 17 (19)Age of onset, cutaneous necrosisDM6Cohort
Ponyi 2005[21]Hungary56 ± 11 (13)46·2 ± 12·3 (68)Age of onset, sex, Gottron sign, cutaneous necrosis, ILD, dysphagia, ANA, arthritisDM and PM7Cohort
Prohic 2009[22]Bosnia and Herzegovina63·1 ± 4·69 (8)52·8 ± 6·87 (24)Age of onset, sex, dysphagia, cutaneous necrosis, ANA, arthritisDM7Cohort
Sparsa 2002[9]France61·5 ± 16·3 (16)54·4 ± 16·2 (24)Age of onset, sex, ILD, cutaneous necrosisDM and PM7Cohort
Fardet 2009[23]FranceNA (29)NA (92)Sex, Gottron sign, cutaneous necrosisDM6Cohort
Rose 1994[33]FranceNA (10)NA (19)Sex, cutaneous necrosis, dysphagia, ILDDM5Cohort
Gallais 1996[13]FranceNA (9)NA (23)Sex, cutaneous necrosisDM7Cohort
Burnouf 2003[34]FranceNA (8)NA (18)Sex, cutaneous necrosisDM7Cohort
Azuma 2011[24]JapanNA (23)NA (113)Sex, arthritis, ILD, dysphagia, ANADM and PM7Cohort
Amerio 2002[25]ItalyNA (14)NA (45)SexDM5Case–control
Wu 2010[26]China50·86 ± 11·93 (37)44·07 ± 13·48 (110)Age of onset, Gottron sign, cutaneous necrosis, dysphagia, ANADM8Cohort
Lian 2005[27]China51·84 ± 13·93 (20)42·63 ± 17·73 (130)Age of onset, sex, ILD, dysphagia, ANADM and PM7Case–control
Zou 2009[28]China50·89 ± 7·25 (22)36·02 ± 16·19 (157)Age of onset, sexDM and PM6Cohort
Tang 2000[29]China47·5 ± 12·4 (18)38·2 ± 15·8 (125)Age of onset, sex, ILD, dysphagiaDM and PM6Cohort
Yu 2004[30]ChinaNA (16)NA (60)Sex, Gottron sign, dysphagia, arthritis, ILDDM7Cohort
Zhou 2012[31]China59·2 ± 12·47 (25)47·2 ± 14·4 (186)Age of onset, sex, Gottron sign, dysphagia, ANA, arthritisDM7Cohort
Chen 2008[32]China56·6 ± 15·5 (10)41·6 ± 15·1 (133)Age of onset, sex, Gottron sign, dysphagia, ILDDM and PM7Cohort

Heterogeneity test

No heterogeneity was observed for age (P = 0·18, I2 = 26%), male sex (P = 0·3, I2 = 13%), cutaneous necrosis (P = 0·04, I2 = 48·5%), Gottron sign (P = 0·06, I2 = 48·2%), arthritis (P = 0·27, I2 = 21%) or ILD (P = 0·24, I2 = 22·1%), thus the fixed effects model was applied to analyse those features. Where heterogeneity was observed – for dysphagia (P = 0·009, I2 = 56·1%) and ANA (P = 0·03, I2 = 53·2%) – the random effects model was employed.


The fixed or random effects model was applied depending on the outcome of the heterogeneity testing. The characteristics that were evaluated for risk of developing neoplasm were age (WMD 11·41, 95% CI 9·84–12·98; Z = 14·25, < 0·01), male sex (OR 1·92, 95% CI 1·49–2·48; Z = 5·08, < 0·01), dysphagia (OR 2·41, 95% CI 1·50–3·86; Z = 3·65, < 0·01), cutaneous necrosis (OR 5·52, 95% CI 3·49–8·74; Z = 7·29, < 0·01), arthritis (OR 0·38, 95% CI 0·24–0·61; Z = 4·06, < 0·01) and ILD (OR 0·32, 95% CI 0·20–0·51; Z = 4·80, < 0·01; Figs 2 and 3). The results indicate that being male or of older age are risk factors for cancer. No positive association between cancer and ANA (OR 0·75, 95% CI 0·44–1·28; Z = 1·05, P = 0·29) or Gottron sign (OR 0·98, 95% CI 0·70–1·36; Z = 0·13, P = 0·89) was noted (Fig. 4). There were 11 reports on the relationship between ILD and CAM; one by Tang et al.[29] was excluded from the meta-analysis due to heterogeneity and low median score.

Figure 2.

Forest plots for meta-analysis of the association of cancer with (a) age, (b) male sex, (c) dysphagia and (d) cutaneous necrosis. The sizes of the squares reflect the study-specific statistical weights, horizontal lines represent 95% confidence intervals (CIs) and the diamonds represent summary estimates. OR, odds ratio; CAM, cancer-associated myositis; DM, dermatomyositis; PM, polymyositis; df, degrees of freedom.

Figure 3.

Forest plots for meta-analysis of the association of cancer with (a) arthritis and (b) interstitial lung disease. The sizes of the squares reflect the study-specific statistical weights, horizontal lines represent 95% confidence intervals (CIs) and diamonds represent summary estimates. OR, odds ratio; CAM, cancer-associated myositis; DM, dermatomyositis; PM, polymyositis; df, degrees of freedom.

Figure 4.

Forest plots for meta-analysis of the association of cancer with (a) antinuclear antibodies (ANA) and (b) Gottron sign. The sizes of the squares reflect the study-specific statistical weights, horizontal lines represent 95% confidence intervals (CIs) and diamonds represent summary estimates. OR, odds ratio; CAM, cancer-associated myositis; DM, dermatomyositis; PM, polymyositis; df, degrees of freedom.

Sensitivity analysis

In our study, heterogeneity was present in the analysis of the relationship between dysphagia, ANA and risk of cancer. To identify possible sources of heterogeneity, the analyses were repeated by removing one study per iteration. For dysphagia, the OR derived from 11 studies was 2·11 (95% CI 1·38–3·23) after the study by So et al.[20] was removed from the analysis, and 2·62 (95% CI 1·60–4·30) when the study of Wu[26] was excluded. For the cases of ANA, the OR estimated from the data on eight studies was 0·64 (95% CI 0·39–1·06) when the study by So et al.[20] was removed, and 0·84 (95% CI 0·48–1·45) with exclusion of the study by András et al.[5]

Publication bias

Publication bias associated with eight possible risk factors for CAM was examined; none showed significant publication bias with the Begg's and Egger's tests, respectively, for age (P = 0·502, P = 0·298), sex (P = 0·705, P = 0·690), cutaneous necrosis (P = 1·0, P = 0·932), dysphagia (P = 0·837, P = 0·941), arthritis (P = 0·548, P = 0·456), ILD (P = 0·707, P = 0·388), Gottron sign (P = 0·711, P = 0·141) and ANA (P = 0·602, P = 0·528).


We report a meta-analysis to assess characteristics that may influence cancer development among patients with DM and PM. Our results show that six characteristics are likely to be associated with risk of neoplasm in DM or PM: four increase the risk (old age at diagnosis, male sex, cutaneous necrosis and dysphagia) and two conform protection against malignancy (arthritis and ILD). Neither ANA nor Gottron sign showed any impact on susceptibility to cancer, based on our meta-analysis. We did not examine creatine kinase, lactate dehydrogenase, ESR and aldolase because of significant heterogeneity among different studies. Diagnosis at older age has been suggested by others as a risk factor for cancer in DM and PM.[10, 32] Our analysis revealed that elderly patients are more likely to develop cancer. Furthermore, male patients with DM and PM are more prone to develop cancer than their female counterparts. Skin lesions such as cutaneous necrosis are also positively associated with cancer, which is in agreement with the findings from several previous reports.[10, 13, 35] Our analysis also demonstrated that dysphagia, one of the major complications in DM and PM, is a major risk factor for malignancy.

Trallero-Araguás et al.[36] recently reported that anti-p155 autoantibody may be linked with cancer in patients with DM. Their analysis included 312 adult patients with DM pooled from six studies. The pooled sensitivity and specificity of anti-p155 for diagnosing cancer-associated DM were 78% (95% CI 45–94%) and 89% (95% CI 82–93%), respectively. Anti-p155 had positive and negative predictive values of 58% and 95%, respectively. Targoff et al.[37] demonstrated that patients with DM and PM who are positive for anti-p155 autoantibodies are more likely to present with heliotrope rash, Gottron papules and V-sign rash. Although patients who are positive for anti-p155 autoantibodies may also be positive for ANA autoantibodies, ANA is less specific and may present in healthy individuals as well. Our meta-analysis did not show correlation between ANA autoantibodies and cancer development.

Our study concurs with other reports that have shown that patients with DM and PM with ILD are less likely to develop malignancy.[5, 19-21] Moreover, it has been suggested that anti-tRNA synthetase autoantibodies may be a risk factor for developing ILD in DM and PM.[38] Melanoma differentiation-associated protein (MDA)5 is an RNA-specific helicase that recognizes single-stranded RNA viruses.[39] Although anti-MDA5 autoantibodies have been positively associated with both ILD[40-42] and cancer,[43] such correlation was observed mostly among patients with myositis who presented with either absent or mild muscle phenotypes, suggesting that such correlation might be restricted to a subset(s) of patients with idiopathic inflammatory myopathies (IIMs) such as amyopathic DM. Therefore it is possible that the effect of anti-MDA5 autoantibodies on increased risk of ILD might also be confined to these IIM subsets, and susceptibility to cancer might be influenced by other factors in patients with DM with skin lesions, who represent the majority of the patients included in our meta-analysis. Finally, our study demonstrates that arthritis is a protective factor against cancer in DM and PM, which is consistent with other reports in the literature.[9, 21, 33]

There are several limitations in our meta-analysis. Firstly, because some studies included in our analyses were retrospective case–control studies, possible recall and selection errors cannot be excluded. Secondly, various studies adopted different criteria for diagnosis of DM, PM and CAM. Only six studies provided a clear definition of CAM.[5, 10, 19, 28, 29, 31] András et al.[5] and Antiochos et al.[19] defined neoplasm associated with DM or PM as a cancer detected either 2 years prior to or within 3 years after a diagnosis of DM or PM. Fardet et al.[23] considered a diagnosis of CAM when it was seen < 1 year prior to a diagnosis of DM or PM or within 5 years afterwards. Others used < 6 months after the diagnosis as the cut-off.[28, 31] Thirdly, the number of reported patients with heliotrope rash, Raynaud phenomenon and calcinosis was too small for meta-analysis to yield reliable results, and they were not included in the study. Fourthly, patients with juvenile DM, body myositis and overlap syndrome were not excluded from our study; however, they represented only a very small proportion of the total patients analysed, and potential micro-bias can be omitted. Fifthly, it was not possible for us to perform stratification analyses or subgroup analyses to limit potential compounding effects of the multiple characteristics analysed because the studies included in our meta-analysis did not provide the necessary data to conduct such analyses. Future meta-analysis with stratification analyses would be helpful to confirm our findings. Finally, differences in pathogeneses between DM and PM have been noted. DM is thought to be caused by humorally mediated involvement of the microcirculation, including early capillary deposition of the complement C5b-9 membranolytic attack complex and secondary ischaemic changes, while PM is likely to develop from CD8 T-cell-mediated and major histocompatibility complex 1-restricted autoimmune attack of myofibres.[44] Therefore it is possible that DM and PM may have some different risk factors for cancer. As the number of studies on malignancy in PM is limited, and many studies on the association of cancer with DM and PM have not separated DM from PM, we were not able to analyse these individually.

Nevertheless, overall our results are statistically robust. The one-study-removed test method showed that the sensitivity of the analysis was stable. Our analysis demonstrates that older age, male sex, cutaneous necrosis and dysphagia are risk factors for developing cancer in DM and PM, whereas arthritis and lung complication render protection against malignancy. These findings may not only shed some light on the pathogenesis of cancer developed with DM and PM, but may also provide new prognostic indicators for management of patients with myopathies. Additional large studies would be needed to confirm the results.