For this systematic review and meta-analysis, MEDLINE, EMBASE and COCHRANE were searched by two investigators (AA, MH) for 1) articles on ANCA measurements predicting relapse and 2) diagnosing disease activity in ANCA associated vasculitis (AAV). No restrictions were selected for publication date (inception to September 2021) and/or language. Articles were filtered on studies performed in humans. A combination of MeSH terms (Antibodies, Antineutrophil Cytoplasmic, Recurrence) and title + abstract (tiab) terms (ANCA, relapse, biomarker, disease activity) was used to identify relevant articles (Supplementary Table S1). Cross-references from reference lists of found studies and similar articles from PubMed were reviewed as well.

Found articles (last search on 01-09-2021) were screened on title and abstract by AA, MH and PLK independently. After title and abstract screening, selected studies were reviewed for final inclusion or exclusion (Figure 1). Inclusion criteria consisted of the use of antigen-specific immunoassays, a timeframe from measurement to relapse of a maximum of 12 months and a definition for a rise as either negative to positive conversion or a ANCA level increase. When patient cohort(s) were used in multiple studies, only one study was included. Since the majority of studies are performed using ELISA, we selected ELISA studies over alternative solid-phase immunoassays (such as FEIA) to increase the comparability between studies. No disagreements on study inclusion were found. In case of missing crucial data elements to calculate sensitivity, specificity and subsequent measures such as odds ratios, the study was excluded.

Data extraction was performed by AA and YV independently. Name of first author, year of publication, study type, number of included patients, percentage of renal involvement, used immunoassay, definition of ANCA rise, and sampling interval were extracted. The slope of ANCA increase (%/month) was calculated by dividing the percentage of ANCA increase by the longest sampling interval. Studies were further separated on type of ANCA (PR3, MPO or pooled), time to relapse (rise during relapse, 1–6 months or 6–12 months), and the definition of a rise. From several studies we obtained data regarding ANCA rises during a relapse and prior to a relapse. From all selected studies the total number of relapses and total number of patients were extracted and four groups were created (rise+relapse+, rise-relapse+, rise+relapse-, rise-relapse-).

All included studies were assessed independently by AA and YV using the Quality Assessment Tool for Diagnostic Accuracy (QUADAS-2), in accordance with systematic review recommendations (11). According to QUADAS-2 guidelines, the content of the tool was first tailored to the two main research questions; to investigate the simultaneous presence of ANCA (rises) and relapses, and to investigate relapses that would follow ANCA rises within 6 or 12 months. This was followed by an independent pilot to reach agreement on the rating method by YV and MH. The QUADAS-2 supports a critical quality assessment of studies by providing four domains to assess: patient selection, index test, reference standard and timing; with subsequent questions in each domain. A study is regarded as having a high potential risk of bias when 2 or more negative points are found in each section. In case of discordance between quality assessments of AA and YV, the study was discussed with the other authors.

To study ANCA positivity or rise during a relapse, odds ratios were calculated by dividing the odds of having a clinical flare accompanied by a positive or rising ANCA test, divided by the odds of having a clinical flare accompanied by a negative or stable ANCA test. To study relapses following ANCA rises, odds ratios were calculated by dividing the odds of having an ANCA rise in 6 or 12 months before a clinical flare is diagnosed, divided by the odds of having an ANCA rise without the presence of a clinical flare in the following months. Finally, a meta-analysis based on the odds ratio was performed to estimate the odds of having a relapse when ANCA rises as opposed to a non-ANCA-rise. A random effects model was used and 95% confidence intervals (CI) were calculated. Four meta-analyses were performed: ANCA positivity during a relapse, having a relapse when ANCA rises, having a relapse within 6 months of an ANCA rise and having a relapse within 12 months of an ANCA rise. Statistical heterogeneity was assessed by Chi-squared test (Chi²) and between-study inconsistency was quantified by the I² statistic. In subsequent analyses, studies were pooled based on whether they were PR3-ANCA-only, MPO-ANCA-only or pooled ANCA. Forest plots were generated using Review Manager 5.4 software.

The literature search provided 806 results. These included all articles found in the last known meta-analysis on this subject from 2012(6). After title and abstract screening, 50 studies were selected for full-text review (Figure 1) (12). Inclusion and exclusion criteria were reviewed and 30 studies were excluded due to the use of IIF only (N = 10), missing information to complete calculations or to subdivide studies in 6 or 12 month prediction analyses (N = 16) or due to reuse of patient cohorts (N = 4) (Supplementary Table S2). 20 studies were included (Table 1). In total, 13 studies were used for simultaneous ANCA rise and disease relapse analysis, with eight studies investigating whether ANCA is positive during disease relapse and six studies investigating whether a relapse is simultaneously present with an ANCA rise. A total of 16 studies were used for the meta-analyses of ANCA rises preceding subsequent relapses within 6 or 12 months. Some studies provided information on both simultaneous ANCA rise and disease relapse, as well as prediction of disease relapses. Selected studies were published between 1993 and 2019. Nine of 20 studies were prospectively conducted. Detailed characteristics of included studies are provided (Table 1).

A risk of bias inventory was made and schematically overviewed for each of the conditions (Supplementary Tables S3–5). Reasons for scoring risk of bias are annotated (Supplementary Tables S6, S7).

Eight studies investigated the moment of relapse and tested for the positivity of ANCA after patients had been in remission prior (13, 16, 17, 21, 25, 28, 31, 32). None of these studies had included individuals with newly diagnosed AAV. Only one of the seven studies, Miloslavsky et al. (25), demonstrated an insignificant OR of 2.07 (CI 0.92–4.69) (25). All other studies favor the presence of a positive ANCA when a relapse is present, with a grand total OR of 10.80 (3.82–30.55) (Figure 2). Dolman et al. (17) and Segelmark et al. (31) included only PR3-ANCA positive individuals, whereas Ara et al. (13) and Terrier et al. (28) only included MPO-ANCA positive individuals. Ara et al. (13) and Dolman et al. (17) demonstrated the highest OR with 71.67 (2.26–2276.12) and 44.20 (1.80–1088.14), respectively, but had the lowest number of patients. The results significantly favor the detectability of ANCA at the moment of disease relapse (Figure 2).

ANCA is often used to monitor AAV patients, even though fluctuations are common during follow-up. We reviewed the literature to understand whether a concurrent ANCA rise indicates a disease relapse or might associate with disease relapses in the future. Six studies investigated whether an ANCA rise is simultaneously associated with a disease relapse (8, 18–21, 26). In these studies, an ANCA rise was defined as 1.3–4-fold elevation in ANCA level. Jayne et al. (21), Han et al. (20), Finkielman et al. (18), Kemna et al. (8), and Fussner et al. (19) strongly favor an absence of a disease relapse, as opposed to the presence of a disease relapse, simultaneously with an ANCA increase. Nowack et al. (26) slightly trends toward favoring a relapse instead, with an OR of 1.62 (0.75–3.48). Overall there is an OR of 0.13 (0.03–0.53) that significantly favors not having a relapse, as opposed to the presence of a relapse, simultaneously with an ANCA increase (Figure 3).

ANCA increases could be indicative of an ongoing or starting inflammatory process that could indicate future relapses. We set out to investigate, within clinically relevant time frames, if this was the case. Eight studies investigating ANCA rises with subsequent relapse within 6 months of the rise are included in the meta-analysis (8, 14, 16, 17, 23, 26, 29, 30) (Figure 4).

De'Oliveira et al. (16), Boomsma et al. (14), Lurati-Ruiz et al. (23), Yamaguchi et al. (30) and Watanabe et al. (29) all significantly favor a relapse within 6 months, as opposed to no relapse, after a measured ANCA increase. Additionally, the study by Dolman et al. (17) trends toward a relapse within 6 months with an OR of 10.82 (0.46–252.79). The studies by Nowack et al. (26) and Kemna et al. (8) remain inconclusive. The grand total OR of having a relapse within 6 months of an ANCA increase compared with not having a relapse is 3.65 (1.66–8.03).

Ten studies investigated ANCA rises with subsequent relapses within 12 months of the rise (8, 14, 15, 18–22, 24, 28) (Figure 5). Part of these studies overlap with the 6 months analysis, but have looked separately at relapses within 12 months. Boomsma et al. (14), Terrier et al. (28), Damoiseaux et al. (15), and McClure et al. (24) significantly favor the presence of a relapse as opposed to no relapse, within 12 months of an ANCA increase. Additionally, Jayne et al. (21), Han et al. (20), Kemna et al. (8), and Jones et al. (22) trend toward favoring a relapse within 12 months of an ANCA increase. However, Finkielman et al. (18) and Fussner et al. (19) favor no relapse instead, making the 12-month data slightly more variable compared with the 6-month data. This also resulted in a slightly lower, yet significant grand total OR of 2.88 (1.21–6.88), favoring a relapse as opposed to no relapse within 12 months of an ANCA increase.

In previous analyses PR3-ANCA and MPO-ANCA patients were pooled. To extrapolate our findings to patient groups, it would be of interest to understand whether disease relapses are predicted more accurately by changes in one of the two ANCA types. Within 6 months of a ANCA rise, Yamaguchi et al. (30) and Watanabe et al. (29) investigated MPO-ANCA only. A grand total OR of 9.29 (3.59–24.07) was reached in favor of having a relapse, as opposed to no relapse, within 6 months after MPO-ANCA increase (Supplementary Figure S1). In contrast, Dolman et al. (17), Boomsma et al. (14) and Nowack et al. (26), all investigating PR3-ANCA positive patients, only showed a trend toward favoring having a relapse within 6 months of a PR3-ANCA increase at an OR of 2.46 (0.57–10.62) (Supplementary Figure S2).

At 12 months, Boomsma et al. (14) and Terrier et al. (28) both favor a relapse within 12 months of an MPO-ANCA increase with an OR of 27.11 (4.14–177.41) (14, 28) (Supplementary Figure S3). However, both studies have studied a low number of patients. For PR3-ANCA positive patients only, the results are heterogeneous. Finkielman et al. (18) and Fussner et al. (19) favor no relapse within 12 months of a PR3-ANCA increase with OR of 0.45 (0.18–1.10) and 0.34 (0.14–0.82) respectively. However, Boomsma et al. (14), Damoiseaux et al. (15) and McClure et al. (24) significantly favor a relapse within 12 months of a PR3-ANCA increase, shifting the grand total of the five studies toward favoring a relapse within 12 months at an OR of 2.53 (0.52–12.44) (Supplementary Figure S4). Although the results are widespread, clinical characteristics of the five studies are similar for ANCA increase definition, but the two studies that trend toward no association are prospective cohort studies (18, 19) (Table 1).