Line blot immunoassays in idiopathic inflammatory myopathies: retrospective review of diagnostic accuracy and factors predicting true positive results

Background Line blot immunoassays (LIA) for myositis-specific (MSA) and myositis-associated (MAA) autoantibodies have become commercially available. In the largest study of this kind, we evaluated the clinical performance of a widely used LIA for MSAs and MAAs. Methods Adults tested for MSA/MAA by LIA at a tertiary myositis centre (January 2016–July 2018) were identified. According to expert-defined diagnoses, true and false positive rates were calculated for strongly and weakly positive autoantibody results within three cohorts: idiopathic inflammatory myopathy (IIM), connective tissue disease (CTD) without myositis, and non-CTD/IIM. Factors associated with true positivity were determined. Results We analysed 342 cases. 67 (19.6%) had IIM, in whom 71 autoantibodies were detected (50 strong positives [70.4%], 21 weak positives [29.6%]). Of the strong positives, 48/50 (96.0%; 19 MSAs, 29 MAAs) were deemed true positives. Of the weak positives, 15/21 (71.4%; 3 MSAs, 12 MAAs) were deemed true positives. In CTD without myositis cases (n = 120), 31/61 (51.0%; 5 MSAs, 26 MAAs) autoantibodies were strongly positive, with 24/31 (77.4%; 0 MSAs, 24 MAAs) true positives. 30/61 (49.2%; 13 MSAs, 17 MAAs) were weakly positive, with 16/30 (53.3%; 0 MSAs, 16 MAAs) true positives. In non-CTD/IIM cases (n = 155), all 24 MSAs and 22 MAAs were false positives; these results included 17 (37.0%; 7 MSAs, 10 MAAs) strong positives. Individual autoantibody specificities were > 98.2 and > 97.5% for weakly and strongly positive results, respectively. True positivity was associated with high pre-test for IIM (odds ratio 50.8, 95% CI 13.7–189.2, p < 0.001) and strong positive (versus weak positive) results (4.4, 2.3–8.3, p < 0.001). Conclusions We demonstrated the high specificity of a myositis LIA in a clinical setting. However, a significant burden of false positive results was evident in those with a low pre-test likelihood of IIM and for weakly positive autoantibodies.

Line blot immunoassays (LIA) for MSA/MAA have become commercially available, increasing the availability of testing in clinical practice [19]. Literature suggests LIA is an appropriate substitute to conventional immunoprecipitation for MSA/MAA testing, but only small samples have been studied with variable accuracy demonstrated [13,19,20].
In the largest study of its kind, we evaluated the diagnostic accuracy of a commercially available LIA for MSA and MAA testing in a clinical setting and examined factors associated with true positive results.

Cases
We retrospectively identified patients tested with the EUROLINE Inflammatory Myopathies 16 Ag (IgG) commercial LIA (Euroimmun, Lubeck, Germany) from January 1st, 2016, to July 30th, 2018, at Salford Royal NHS Foundation Trust (SRFT), United Kingdom. The search was limited to patients being reviewed in tertiary IIM, systemic sclerosis and neuromuscular outpatient clinics. Patients without available clinical data were removed. Where duplicate testing occurred, only the most recent results were analysed.
This study was performed as part of a quality improvement project evaluating LIA usage at SRFT. Case notes and other data were reviewed retrospectively without alteration to patient management. Given this context, and after consultation with the Health Research Authority (via www.hra-decisiontools.org.uk) this study proceeded without further requirement for ethical authorization.

Clinical data collection methodology
Patient records were reviewed by 2 authors (FT and CR) independently. Demographics, electromyography (EMG) results, muscle biopsy results, and peak serum total creatine kinase (CK) levels were collated. Indications for ordering the LIA (pre-test diagnoses) were categorised retrospectively as "suspected IIM", "CTD without evidence of myositis", or "myopathic syndromes with low likelihood of IIM". The final diagnosis made by the expert treating clinician was recorded and categorised retrospectively as "IIM" (including overlap syndromes), "CTD without myositis", or "non-IIM/CTD". Categorisation was agreed on between the authors (FT, CR), and a third author was consulted in indeterminate cases (JBL or HC). Diagnoses were verified through review of extensive clinical information reflecting several years of care in each case. Research classification criteria were not applied as by their nature these are restrictive and would limit the real-world applicability of this study. In addition, recent studies have demonstrated classification criteria may not accurately reflect clinical diagnoses made by expert clinicians [21].

Assay performance
Ab results were reviewed and categorised as true or false positive according to the available clinical information. True positive MSAs were defined as those in patients diagnosed with IIM with the phenotype and IIM subtype expected of that MSA. True positive MAAs were defined as those in patients diagnosed with either CTD or IIM phenotypes expected of that MAA. Otherwise, results were deemed to be false positives. For patients with multiple MSAs, that which best reflected the clinical phenotype was assigned true positive, as MSAs are generally mutually exclusive [8]. Additional MSAs in such cases were false positives, except for anti-Mi2A and anti-Mi2B (isoforms of the same Ab) where simultaneous true positives were accepted [22,23]. All negative results were deemed true negatives. In cases of uncertainty, an immunologist (SE) reviewed the source data to ensure accuracy.

Statistics
Analysis was performed with STATA version 14 (College Station, USA). Descriptive statistics examined characteristics of different groups according to Ab status.
Categorical data were summarised as frequencies and proportions. Continuous data were summarised using means and standard deviations. For individual Abs, the rate of true and false positivity and the associated specificity for the presence of a consistent diagnosis or disease subtype was calculated. Logistic regression was performed to investigate factors associated with true positive results. A p-value < 0.05 represented a statistically significant difference.
Seven IIM patients had apparent dual MSA positivity. 3/7 (43.0%; DM (n = 2) and OM (n = 1)) had concurrent anti-Mi2A and anti-Mi2B. 1/7 (14.3%; ASS (n = 1)) was strongly positive for both anti-EJ and anti-PL7; anti-EJ was deemed the true positive based on significantly higher signal intensity when reviewed with the source data compared to anti-PL7 which just met the cut off for a strong positive result. Another was strongly positive for both anti-Jo1 and anti-SAE1; anti-Jo1 was felt to be the true positive as the patient clinically had ASS. 2/7 (28.6%) patients had dual weak false positives (anti-MDA5 and anti-TIF1γ in a patient with OM; anti-SAE1 and anti-SRP in a patient with PM).

Sensitivity, specificity, and factors associated with a positive result
When considering individual Abs, specificity for the presence of a consistent diagnosis or disease subtype was generally high across all Abs (98.2-100.0% for weak positives and 97.5-100.0% for strong positives; Supplementary Table 1). However, weak positive anti-SRP had the lowest specificity (97.0%). A pre-test working diagnosis of IIM and a strong positive Ab result were significantly associated with true positivity (OR 50.8, 95%CI 13.66-189.22, p < 0.001 and OR 4.38, 95%CI 2.32-8.26, p < 0.001, respectively) ( Table 3).  . This was particularly true for weakly positive anti-SRP results which were all false positives in our study. Our study is in agreement with other recent publications which have also demonstrated that weak positives are more likely to be false [24]. Whilst weak positive MSAs were more likely to be false positives, specificity was high. Weak positive anti-SRP had the lowest specificity. Our results suggest that this LIA's accuracy may be improved if the threshold for defining weak positivity was increased, although this may vary according to each antibody on the assay [25].
We also found only 4/67 (5.9%) IIM cases with multiple MSAs (excludes concurrent anti-Mi2A and anti-Mi2B, isoforms of Mi2 autoantibodies which coexist frequently [23]). Two of these cases only had 1 true MSA each and in the other 2 cases, all were false positive results. This is congruent with recent large cohort studies demonstrating mutual exclusivity of MSAs in IIM individuals [8] and highlights that when multiple MSAs are found in LIA testing, results should be treated with suspicion. Dual positivity for the MAAs anti-PM-Scl100 and anti-PM-Scl75, in contrast, improved the reliability of the results in both IIM and CTD without IIM cases.
Another notable finding was that a high proportion of IIM patients were seronegative (37.3%). This number is comparable to recent findings from a large cohort of European IIM patients which found 38.3% of their cases to be seronegative [8]. A growing number of Abs currently not available in this LIA may be useful in the correct clinical context. For example, in addition to anti-HMGCR, recent larger cohorts of IIM demonstrate that other emerging Abs such as anti-KS and anti-Zo are also useful in the diagnosis of IIM [8] .
Limitations of this study include data drawn from a single centre, although they represent a population of nearly 3 million people. Secondly, data were analysed retrospectively, and no specific additional review or tests were performed to confirm the diagnostic categorisation. Of the patients without CTD or IIM, most had at least 3 years of follow-up in their case notes, but it remains possible that positive Ab results may represent preclinical IIM or CTD. Additionally, negative Ab results were assumed to be true negatives. It is possible that some seronegative patients have detectable Abs via another method such as immunoprecipitation or have a hitherto undescribed Ab. Additionally, in cases where duplicate testing on the same patient occurred, we included only the most recent results. There is some evidence that certain MSAs might be lowered with treatment [26] so including only the most recent LIA result may have affected our results. However, clinicians seldom use the LIA for disease monitoring and the seven duplicates which occurred were more likely to be cases where the accuracy of first LIA test was in question. Finally, the final diagnoses made by the treating physicians could have been biased by the Ab results. However, most patients had several years of follow-up allowing for their diagnoses to be confirmed or reclassified over time and these final diagnoses were used in this study.

Conclusions
MSAs and MAAs are increasingly gaining importance in the diagnostic workup and management of IIMs. MSA myositis-specific autoantibody, MAA myositis-associated autoantibody, OR odds ratio, P p-value, CI confidence intervals, SD standard deviation, IIM idiopathic inflammatory myopathy, EMG electromyography, CK creatine kinase