A Systematic Review and Meta-Analysis to Inform Cancer Screening Guidelines in Idiopathic Inflammatory Myopathies

Alexander G. S. Oldroyd; Andrew B. Allard; Jeffrey P. Callen; Hector Chinoy; Lorinda Chung; David Fiorentino; Michael D. George; Patrick Gordon; Kate Kolstad; Drew J. B. Kurtzman; Pedro M. Machado; Neil J. McHugh; Anna Postolova; Albert Selva-O'Callaghan; Jens Schmidt; Sarah Tansley; Ruth Ann Vleugels; Victoria P. Werth; Rohit Aggarwal


Rheumatology. 2021;60(6):2615-2628. 

In This Article


A total of 7030 articles were initially identified via the literature search and 141 were reviewed for eligibility following removal of ineligible papers, duplicates, case reports and reviews (Figure 1). Sixty-seven studies were included in the risk factor meta-analysis (Figure 1) and nine in the screening narrative review (two studies were included in both the meta-analysis and systematic review). Table 1 displays the summary RRs and WMDs calculated for each risk factor. See Supplementary Material for forest plots for each risk factor and further details of each study (Supplementary Table 1, available at Rheumatology online).

Figure 1.

PRISMA flow diagram

IIM Subtypes

DM was significantly associated with a higher risk of cancer compared with other IIM subtypes.[17–40] PM[17–28,30,31,33–40] and CADM[28,35,41] were found to be associated with significantly lower risk for cancer compared with remaining IIM subtypes. ASS subtype was a non-significant factor; however, data from only two eligible studies were available.[21,42] Insufficient data were available to perform meta-analysis on data relating to IMNM.

Clinical Factors Including Demographics and Laboratory Calues

Demographics. Older age at time of IIM onset was found to be significantly associated with increased risk of cancer.[21,22,28,29,32,33,36,39,42–54] The mean age of IIM onset in cancer cases was 59 years, compared with 49 years in the non-cancer cases. Male sex was found to be significantly associated with higher risk of cancer, compared with female sex.[17,18,21,22,25–30,32–34,36,38,39,41–59]

Clinical Risk Factors. Sufficient data were available to quantify the cancer risks associated with dysphagia, cutaneous ulceration, Raynaud's phenomenon and ILD. Dysphagia, which was typically not objectively defined across the majority of studies, was significantly associated with higher risk of cancer.[22,25,26,29,32,33,39,43,46,47,49,56,59,60] Cutaneous ulceration was also significantly associated with higher risk of cancer.[45,46,49,50,53] Analysis revealed that the presence of Raynaud's phenomenon was associated with a significantly lower risk of cancer.[22,25,26,29,30,39,41,45,46,50,59] The presence of ILD, which was typically diagnosed via CT scanning, was also associated with a significantly lower risk of cancer.[22,26,28–30,32,33,35,39,43,45,46,56,59]

Laboratory Values. Lower creatine kinase (CK)[22,27,29,32,33,38,39,43,46,53] values were significantly associated with increased cancer risk, therefore, conversely, very high values were associated with lower risk of cancer. It is important to note, however, that the mean CK level in cancer cases (2402 IU/l) was still raised compared with normal values, but lower than the non-cancer group (3557 IU/l). Similarly, lower lactate dehydrogenase (LDH)[22,32,33,38,39,46,53] values (mean LDH 766 U/l) were found to be associated with increased cancer risk as compared with higher LDH values (mean LDH 1078 U/l). Both alanine transaminase (ALT)[29,38,53] and ESR[22,29,33,38,43,45] levels were found to be non-significant factors, and insufficient data were available for aspartate aminotransferase and aldolase.

Autoantibodies. Anti-transcriptional intermediary factor-1 gamma (anti-TIF1γ) positivity was significantly associated with increased cancer risk.[21,31,37,41,44,48,58,60–71] Anti-nuclear matrix protein 2 (anti-NXP2) positivity was a non-significant factor.[21,37,61,62,68,71–76] Large proportions of the control cohorts in studies of anti-NXP2-positive cohorts were comprised of anti-TIF1γ-positive cases. We repeated meta-analysis after removing anti-TIF1γ-positive cases. The RR of anti-NXP2 using data from six studies with anti-TIF1γ cases excluded was 1.47 [95% confidence interval (CI): 0.57, 3.80; I2 0.00%], again indicating that positivity for anti-NXP2 is a non-significant factor for cancer relative to other autoantibody subtypes.[21,37,62,68,71,73]

Analysis was carried out for each individual ASS-related autoantibody. Anti-Jo-1[21,22,25,26,29,33,46,59,62,71,77] and anti-EJ[21,62,71,78,79] were significantly associated with reduced cancer risk. Positivity for anti-PL7,[21,62,71,78,79] anti-PL12,[21,62,71,78,79] anti-OJ[21,71,78,79] and anti-KS[71,79] were non-significant factors, although limited by small number of studies. Analysis revealed that the presence of any ASS-related autoantibody was significantly associated with lower risk of cancer.[21,22,25,26,29,33,59,62,70,71]

Positivity for other autoantibodies, including anti-3-hydroxy 3-methylutaryl coenzyme A reductase (anti-HMGCR),[21,62,71,80,81] anti-signal recognition particle (anti-SRP),[62,71,80] anti-small ubiquitin-like modifier-1 activating enzyme (anti-SAE1),[21,62,71] anti-melanoma differentiation-associated gene 5 (anti-MDA5)[21,44,48,62,71] or anti-Mi2,[21,41,58,62,64,71] were identified as non-significant factors for cancer. Both myositis specific autoantibody (MSA) negativity[21,30,45,62,71] and ANA positivity[22,26,28,30,39,41,43,46,50,53,54,56,82] were non-significant factors.

Heterogeneity and Publication Bias

Table 1 displays the standard chi-squared test results and I2 statistic for heterogeneity of each analysed factor. Influence analysis aimed to identify outlier studies for risk factors with very high (>75%) heterogeneity. One study each fulfilled the outlier criteria for CK,[53] ESR[45] and ALT.[53] Adjusted WMD after removal of data from outlier publication was calculated and did not change overall relationships (Supplementary Table 2, available at Rheumatology online).

Significant publication bias was observed with 'any ASS-antibody'. Adjusted RR following 'trim and fill' analysis with six added studies was 0.46 (95% CI: 0.23, 0.93).

Cancer Screening Utility Review

Nine studies[40,45,83–89] relating to utility of cancer screening approaches in IIM populations were identified. Table 2 displays the details of each study. A total of 90 cancers were identified via screening across 1033 patients. Studies were carried out across a number of countries, including the USA, Canada, Taiwan, China, France and Spain, and widely ranging intervals between IIM onset/diagnosis and screening were reported. All but one study was retrospective. Study population sizes ranged between 14 and 400. A wide variety of cancers were diagnosed, including but not limited to breast cancer, squamous cell carcinoma, multiple myeloma, ovarian cancer, lymphoma, lung cancer and oesophageal cancer.

The utility of 'blind screening' (i.e. investigations carried out in the absence of target symptoms) was reported by Leatham et al.[85] and Sparsa et al..[45] Leatham et al. identified 17 out of 48 cancer patients diagnosed with cancer via blind screening modalities after DM onset. CT scanning of the thorax, abdomen or pelvis detected the most cancer diagnoses (6/17, 38%), followed by mammography (3/17, 18%). Sparsa et al. reported the identification of 30 cancers via 122 investigations. Thirty-five investigations were 'directed' (i.e. initiated due to the presence of target symptoms) and resulted in the identification of 19 (54%) cancers. In contrast, 87 investigations were blind and identified 11 (13%) cancers. Again, CT scanning of the thorax, abdomen and pelvis was the single investigation that detected the most cancers (5/18, 28%).

The utility of 18F-FDG PET/CT was reported by Maliha et al.[84] and Selva-O'Callaghan et al..[88] Maliha et al. reported that fluorodeoxyglucose (18F-FDG) PET/CT scans revealed no further cancer diagnoses and actually lead to more biopsies, compared with 'conventional' screening (see Table 2 for details). Similarly, Selva-O'Callaghan reported that single 18F-FDG PET/CT scans were comparable to a large number of conventional screening investigations, which included complete physical examination, laboratory tests (complete blood count and serum chemistry panel), thoraco-abdominal CT scan, tumour markers [carbohydrate antigen-125 (CA125), CA19-9, carcinoembryonic antigen (CEA), prostate-specific antigen], gynaecological examination, ovarian ultrasonography and mammography.

The screening utility of CA125 was demonstrated by Amoura et al.[89] and Whitmore et al..[87] Amoura et al. demonstrated that increased levels were significantly associated with subsequent cancer diagnoses (OR 29.7; 95% CI: 8.2, 106.6; P-value <0.0001). Whitmore et al. also demonstrated the utility of normal values—no study participant with normal CA125 levels was subsequently diagnosed with cancer during the study period. In contrast, Lim et al. concluded that CA125 testing was not useful for detection of cancer.[40] Eighteen participants had raised CA125 levels and only one (6%) was subsequently diagnosed with cancer. Additionally, 53 participants had normal CA125 levels and two (4%) were diagnosed with cancer.

Both Amoura et al.[89] and Lim et al.[40] reported the screening utility of CEA, CA15-3 and CA19-9 (Table 2). Raised CEA or CA15-3 levels were not associated with cancer in each study. Raised CA19-9 levels were significantly associated with cancer in the study by Amoura et al.; 11 cases had raised levels and three subsequently developed cancer (OR 4.5; 95% CI: 1.00, 18.7; P-value 0.018). Raised CA19-9 levels were not found to be associated with cancer in the study by Lim et al. however. Of note, Amoura et al. reported that three cases had raised levels of both CA19-9 and CA125 and all of these were subsequently diagnosed with cancer (OR 86.3; 95% CI: 4, 1832; P-value <0.0001). Lim et al. also reported no association between raised AFP levels and cancer. Interestingly, Lim et al. reported an association between CA15-3 levels and the development of ILD; eight (89%) of the nine patients with increased CA15-3 levels were diagnosed with ILD.