Antineutrophil cytoplasmic antibody (ANCA), first described in 1982 in patients with systemic vasculitis and glomerulonephritis, has emerged as a new diagnostic tool and marker of disease activity for vasculitis [4, 5]. Originally thought to be a response to arboviral infection, ANCA was later identified as a seromarker for Wegener granulomatosis [4, 6]. It is currently believed that the presence of ANCA is an important factor in the pathogenesis of this disease [5]. Two ANCA patterns may be seen with indirect immunofluorescence of ethanol-fixed neutrophils: a cytoplasmic pattern (c-ANCA) and the artifactual perinuclear pattern (p-ANCA) [7, 8]. The major antigen for c-ANCA is a 29-kd serine protease, known as proteinase 3, which is found within the azurophilic granules of the neutrophil [9]. The major antigen for p-ANCA is myeloperoxidase, a lysosomal enzyme found in neutrophils [7, 10]. Some clinical overlap has been seen, but the two patterns have different disease associations. The c-ANCA pattern has predominantly been associated with Wegener granulomatosis, and p-ANCA has been associated with microscopic polyarteritis, other vasculitides, idiopathic necrotizing and crescentic glomerulonephritis, and other diseases [7, 11-13].

Early studies of c-ANCA in patients with Wegener granulomatosis, especially those with active disease, have been promising. Reported sensitivities and specificities have exceeded 90% [10-12, 14-18]. Because of the morbidity and expense associated with biopsy, c-ANCA has attracted interest as a rapid and noninvasive way to diagnose Wegener granulomatosis. In fact, some investigators advocate immunosuppressive therapy for patients with positive c-ANCA test results and symptoms compatible with Wegener granulomatosis, even in the absence of biopsy results [19-21].

The history of c-ANCA testing appears to mirror the natural history of many new diagnostic tests. Important and early pioneering studies of new tests are typically done in highly selected patients with confirmed disease to determine the test's operating characteristics. However, when the new test is subsequently applied to unselected populations, clinical experience and additional studies often provide a less optimistic picture of its diagnostic power. Several recent reports [22-27] document false-positive c-ANCA test results in patients with tuberculosis, Hodgkin lymphoma, human immunodeficiency virus infection, nasal septal perforation, monoclonal gammopathies, and drug-induced Wegener-like disease. Similarly, additional reports [28, 29] describe negative c-ANCA test results in patients with Wegener granulomatosis, even those with active disease. Given the current enthusiasm for a serologic diagnosis of Wegener granulomatosis, underrecognition of the rate of false-positive c-ANCA test results may lead to inappropriate immunosuppressive therapy in some patients who do not have Wegener granulomatosis. Conversely, underappreciation of the frequency of false-negative c-ANCA test results may delay therapy for patients who have clinical features compatible with Wegener granulomatosis. We did a literature review and meta-analysis to assess the utility of c-ANCA as a diagnostic marker for Wegener granulomatosis overall and in relation to disease activity.

Methods

Top Methods Results Discussion Author & Article Info References

As noted above, two ANCA patterns have been described in the literature, and each has separate disease associations. Patients with Wegener granulomatosis often have c-ANCA, whereas p-ANCA is found in patients with various syndromes with and without vasculitis [7, 30, 31]. Because c-ANCA is thought to have a higher test sensitivity than p-ANCA for the diagnosis of Wegener granulomatosis, we focused only on c-ANCA.

Search Strategy

We could not determine a systematic approach for identifying unpublished data on c-ANCA, and thus we selected only published articles for our literature review. We did a MEDLINE search for English-language studies done in humans and published between 1966 and June 1993. Three searches were done: 1) a textword and registry search for antineutrophil cytoplasmic antibody [ANCA], anticytoplasmic antibody, and antineutrophil antibody; 2) a combination of search terms "antibodies, neutrophils" and keywords such as "pulmonary-renal syndrome," or "Wegener granulomatosis," or "vasculitis," or "glomerulonephritis"; and 3) a search using the keyword "pulmonary-renal syndrome" alone. Using this initial search strategy, we identified 720 citations. Manual examination of the reference lists of included articles and the bibliographies of review articles and standard texts yielded an additional 27 studies, for a total of 747.

Article Selection

To be included in the analysis, all articles had to pass through a four-stage review (Table 1). First, the titles and abstracts of the 747 articles were independently examined by two reviewers. One reviewer was a rheumatologist, and the other was a health services researcher with a PhD who had experience in the critical appraisal of medical literature. Articles were excluded at this stage if they were not pertinent to the topic (for example, if they addressed antineutrophil antibodies and malignancy) or if they did not present patient-level data (for example, if they were reports of tissue culture experiments). If a cited article did not have a published abstract but appeared to be relevant, or if either reviewer included it, it was retained for further review. Of the original 747 articles, 407 were excluded and 340 were retained for second-level review.

In stage 2 of the review process, the complete articles were photocopied and reviewed by a rheumatologist. Articles were excluded if they were case reports (n = 64), reviews (n = 75), or letters to the editor (n = 58), or if they were irrelevant to our topic (n = 81). Sixty-two of the 340 articles remained for potential inclusion.

In stage 3, a rheumatologist (JKR) and an investigator with a PhD reviewed the methods section of each article in detail. The methodologic criteria required for inclusion at this stage were specification of the patient selection method and the use of standard reference criteria to define Wegener granulomatosis. For case–control and cohort studies, the authors had to have described a systematic method of patient selection (for example, consecutive patients or a random sample based on prespecified criteria) [32]. Thus, case-series studies (such as those defining persons with a common feature nonsystematically without the presence of a control group) were excluded. Authors also had to specify standard reference criteria for establishing Wegener granulomatosis exclusive of the c-ANCA test result. Acceptable reference criteria included the Ear, Nose, Throat, Lung, and Kidney staging system [2], the Fauci criteria [1], and the American College of Rheumatology criteria [33]. The Ear, Nose, Throat, Lung, and Kidney system and the American College of Rheumatology criteria require biopsy confirmation of Wegener granulomatosis, and the Fauci criteria are clinicopathologic in nature. Thirty-nine articles were excluded at this stage for the following reasons: case-series design (n = 24); the method of patient selection was not clearly stated (n = 4); a reference standard was not specified (n = 7); or the article was irrelevant to our review (n = 4). If either reader included an article, it was retained for further review.

In stage 4, three physicians each independently read the remaining 23 articles. Articles were excluded at this stage if a 2 x 2 contingency table, needed to calculate the operating characteristics, could not be constructed from the results (n = 8). Thus, 15 articles remained for detailed review [10, 11, 14, 15, 28, 34-43]. For these 15 articles, 2 x 2 contingency tables were constructed for patients with and without Wegener granulomatosis compared with patients with and without positive c-ANCA test results. Disagreements between reviewers were resolved by consensus.

Definition of Wegener Granulomatosis

We included only those studies that specified standard reference criteria for defining Wegener granulomatosis independent of the patient's c-ANCA test result [1, 2, 33]. Patients with Wegener granulomatosis can be further divided into two subtypes: those with classic and those with limited disease [2]. In classic Wegener granulomatosis, glomerulonephritis and inflammation of the upper or lower respiratory tract, or both, are present. In limited disease, the kidneys are not involved, but inflammation of the upper or lower respiratory tract, or both, is present. Only 6 of the 15 studies separated patients by subtype. Furthermore, 2 of these 6 studies did not separate the c-ANCA test results by Wegener granulomatosis subtype. Because of these inconsistencies, we considered a single classification (overall Wegener granulomatosis) to represent any patient with confirmed disease (either classic or limited).

We also assessed the relation of c-ANCA to disease activity in patients with Wegener granulomatosis. For an article to be included in this subset of the review, its authors had to have specified objective activity criteria exclusive of the c-ANCA test result. These criteria included, but were not limited to, the mention of histologic evidence of active inflammation, fever, elevated erythrocyte sedimentation rate, active glomerulonephritis, or new pulmonary infiltrates. Four of the 15 articles met this requirement [11, 15, 34, 37].

Definition of a Positive c-ANCA Test Result

Three methods are currently available for measuring ANCA: indirect immunofluorescence, enzyme-linked immunosorbent assay (ELISA), and radioimmunoassay. Indirect immunofluorescence is considered the gold standard for measuring c-ANCA [44], in part because of its widespread availability. We did not select articles on the basis of the methods the investigators used to do the ANCA tests, but indirect immunofluorescence was used to measure c-ANCA in all of the selected studies, sometimes in conjunction with ELISA. We considered only the indirect immunofluorescence results in our analyses.

As with any other serologic test results, c-ANCA test results may be reported as positive or negative or as a specific titer based on serum dilution. Five of the 15 articles reported c-ANCA results as titers, and 10 reported the results as positive or negative. Because of the lack of a standard titer dilution across all studies, we considered any positive titer to be a positive c-ANCA test result.

Quality Score

We constructed a quality-assessment score for each article. This score was based on five areas of methodologic importance for studies evaluating diagnostic tests [45, 46]. We defined a high-quality study a priori as one that fulfilled the following criteria Table 4: 1) it was either a case–control or a cohort study; 2) data collection was prospective or retrospective-prospective [for example, retrospective selection of patients with prospective follow-up of all patients with Wegener granulomatosis and patients positive for c-ANCA]; 3) the method of patient selection was consecutive or random; 4) more than 50% of the controls had diagnoses [such as other vasculitides, pulmonary-renal syndromes, malignancy, or fungal infections reasonable for ruling out Wegener granulomatosis]; and 5) the investigators reading the biopsy results were blinded to the diagnosis being considered. Thus, a high-quality study would score at least 10 of a possible 13 quality points; however, we included all articles that passed the initial quality screening. Table 2 shows study characteristics and overall quality scores for all included articles.

Statistical Analysis

The sensitivity and specificity of c-ANCA testing were calculated from the data reported in the articles for Wegener granulomatosis overall and, where applicable, for active and inactive Wegener granulomatosis. Summary estimates of sensitivity and specificity were derived by pooling averages weighted by the number of patients in each study [47]. Before determining the summary estimates, we evaluated a chi-square test of homogeneity for the 15 studies. We applied a fixed-effects model for the pooled estimates unless there was heterogeneity across studies, in which case we applied a random-effects model [47, 48]. In general, the random-effects model provided wider CIs to account for study heterogeneity. We derived 95% CIs for all estimates, including the summary measures, using exact methods for proportions. In a secondary analysis, we repeated the above analyses for studies with a quality score of more than 7 compared with those with a quality score of less than 7.

Results

Top Methods Results Discussion Author & Article Info References

The characteristics of the studies selected for review are summarized in Table 2. The 15 studies reported c-ANCA test results in a total of 13 562 patients (range, 33 to 7500 patients). There were 736 patients (5.4%) with Wegener granulomatosis defined according to our inclusion criteria (range, 3 to 277 patients), 1186 normal controls (8.8%), and 11 641 patients with other diseases (85.8%). The total number of positive c-ANCA test results was 616 (range, 6 to 213 positive results); the total number of negative results was 12 946 (range, 26 to 7483 negative results).

Thirteen studies were done in academic institutions and 1 was done at the National Institutes of Health (NIH); the location of 1 study was unspecified. The NIH contributed patient serum specimens for the controls in 1 additional study [15]. Three studies included patients from community hospitals, but none of the 15 studies was done solely in a community hospital setting. Twelve studies were retrospective, and 6 provided prospective follow-up of the patients with Wegener granulomatosis and other patients positive for c-ANCA. None of the authors gave information about the duration of symptoms for patients or controls. Only 1 study specified the temporal relation of c-ANCA testing to immunosuppressive therapy [28].

The sensitivity and specificity estimates with 95% CIs for Wegener granulomatosis overall are shown in Figure 1. The range of sensitivities was wide (34% to 92%), and the range of specificities was narrow (88% to 100%) among the reported results. The pooled sensitivity of c-ANCA for Wegener granulomatosis overall was 66% (95% CI, 57% to 74%), and the pooled specificity was 98% (CI, 96% to 99.5%).

View larger version (29K): [in this window] [in a new window]   Figure 1. Operating characteristics of c-ANCA testing for confirmed Wegener granulomatosis.Top. Overall disease. Middle. Active disease. Bottom. Inactive disease.

Four of the 15 articles gave disease activity information; this subset reported on 468 patients with Wegener granulomatosis (64% of all patients with Wegener granulomatosis). Only 1 study reported on the disease activity of controls [37]. In 2 of the 4 studies, tissue confirmation was included in the criteria used to assess disease activity [15, 37]. In all 4 studies, the sensitivity of c-ANCA was higher for patients with active disease than for patients with inactive disease. For active Wegener granulomatosis, the summary sensitivity was 91% (CI, 87% to 95%) and the summary specificity was 98.6% (CI, 97% to 99.9%). For inactive Wegener granulomatosis, the summary sensitivity and specificity were 63% (CI, 57% to 69%) and 99.5% (CI, 99.1% to 99.7%), respectively.

The quality scores of the studies ranged from 4 to 10 of a possible 13 points. Only three studies satisfied our criteria for a highest-quality study ( 10 points) [10, 42, 43]. Two of the three highest-quality studies reported sensitivities of 39% and 46% [42, 43]; the sensitivity of the third study was 92% [10]. The number of articles that met the highest-quality standards was insufficient to stratify the summary calculation of sensitivity and specificity. When the articles were stratified on the basis of a quality score of 7, the results did not differ significantly. For overall disease, the pooled sensitivity and specificity for the 8 articles that had quality scores of less than 7 were 66% (CI, 57% to 73%) and 99% (CI, 97% to 100%), respectively. For the 7 articles with quality scores of 7 or more, the pooled sensitivity was 71% (CI, 49% to 89%) and the pooled specificity was 96% (CI, 92% to 99%). In 7 of the 15 studies, more than 50% of the controls tested were either disease-free persons or patients in whom Wegener granulomatosis would not have been clinically suspected [10, 11, 34, 38-40].

Discussion

Top Methods Results Discussion Author & Article Info References

Fifteen studies published between 1987 and 1993 met the minimum criteria for inclusion in our literature review and meta-analysis. Excluding normal controls, the overall prevalence of Wegener granulomatosis in the group of patients tested was 5.9%. For Wegener granulomatosis overall, the 15 studies reported sensitivities for c-ANCA ranging from 34% to 92%; the specificities were between 88% and 100%. When these results were pooled, the summary estimates for sensitivity and specificity were 66% and 98%, respectively. Four of the studies provided data on disease activity. In patients with active, histologically confirmed Wegener granulomatosis, the reported sensitivities were more homogeneous and higher than those reported for patients with inactive disease and overall disease. For active disease, the pooled sensitivity was 91%; for inactive disease, it was 63%, similar to that obtained for disease overall.

When applying our findings to clinical practice, both the anticipated prevalence of Wegener granulomatosis in the population tested and the patient's disease activity are key variables determining the usefulness of c-ANCA testing. Not surprisingly, this testing works better in patients with known, active Wegener granulomatosis. Using the summary estimates of sensitivity (66%) and specificity (98%) for overall disease in a hypothetical sample of 1000 patients and a prevalence of Wegener granulomatosis of 5%, as found in our study, the positive predictive value of a c-ANCA test result would be 63%. Thus, 37% of patients with a positive c-ANCA test result could be falsely classified as having Wegener granulomatosis on the basis of the test alone. However, if c-ANCA testing is done only in patients with signs of active Wegener granulomatosis, the false-positive rate would decrease from 37% to 16%. The improved test performance of c-ANCA among patients with active Wegener granulomatosis raises the question of its utility as a disease-monitoring tool in some patients with Wegener granulomatosis, analogous to the use of anti-double-stranded DNA to monitor patients with lupus nephritis. Indeed, two of the studies we reviewed came to different conclusions about this: Cohen Tervaert and coworkers [15] concluded that an increasing c-ANCA titer was a sensitive marker for active Wegener granulomatosis. Kerr and colleagues [37] found that changes in serial c-ANCA titers temporally correlated with changes in disease activity in only 24% of patients with disease in remission or smoldering activity. We were unable to address this issue quantitatively in our analysis because of the paucity of articles fulfilling our selection criteria.

Although the relation of c-ANCA's test performance to disease activity may appear intuitive, its relation to the prevalence of disease is less obvious. Nephrologists, pulmonologists, and rheumatologists probably see more patients with this disease than do general internists and family practitioners. Generalists probably encounter a lower prevalence of Wegener granulomatosis than do subspecialists. The prevalence, or prior probability, of Wegener granulomatosis thus varies among clinical practice settings, and, as is true for diagnostic tests in general, this variation profoundly affects the predictive value of a positive c-ANCA test result (Table 3).

At the extremes of prevalence, c-ANCA has the least diagnostic utility. If the prevalence, or probability, of Wegener granulomatosis is low (perhaps 1% or less), a positive c-ANCA test result is very likely to be a false-positive result, leading to much unnecessary, costly, and even risky additional diagnostic testing. Therefore, given that Wegener granulomatosis is a rare condition, there is little value in using a c-ANCA to screen for disease in low-prevalence situations. On the other hand, if the pretest probability of Wegener granulomatosis is high because of a classic clinical presentation, a positive c-ANCA test result has marginal clinical value because other evidence suggests that the diagnosis is likely. Thus, a clinician might choose to proceed to biopsy, regardless of the test result, because cyclophosphamide therapy is not innocuous.

The c-ANCA test may be most valuable in settings in which the prevalence of Wegener granulomatosis is about 5% to 10% and the clinician is considering several diagnoses, including Wegener granulomatosis. In such a setting, a negative test result may not significantly reduce the likelihood of Wegener granulomatosis, but a positive result might boost that likelihood high enough to warrant further testing. In this prevalence range, clinicians might do an ANCA test to help identify patients who might benefit from further evaluation, such as biopsy.

We found significant heterogeneity among the sensitivities reported in those articles that did not define the clinical activity of Wegener granulomatosis. This variability forced us to use a model that resulted in wider CIs for sensitivities and specificities. This variation can be explained in several ways. First, the laboratories involved may have differed in their ability to do c-ANCA testing; significant measurement error has been previously described with other rheumatologic tests, such as that for antinuclear antibodies [49]. Although clinical laboratories strive for the most precise measurements possible, variation inevitably occurs, and some serum samples are classified as positive by one laboratory and as negative by another [49, 50]. All studies measured c-ANCA by using indirect immunofluorescence, the internationally accepted gold standard for c-ANCA testing, and yet few reported the screening dilution used to define a positive result. This factor may have affected the reported sensitivities. For example, more positive c-ANCA results could have been obtained if a positive test was defined at a lower dilution cutoff, thus potentially overestimating sensitivity. Although most studies provided for two independent readings of the indirect immunofluorescence slides, few confirmed the positive results by using ELISA [11, 39, 40, 43] or measured interassay variability in the samples positive for c-ANCA [11, 37].

Second, the patients with Wegener granulomatosis may have differed in the extent or severity of their disease. These differences arise partly from true biological variation and partly from the lack of a universally accepted reference standard for the diagnosis of Wegener granulomatosis. Renal tissue and specimens from the upper airways of patients with Wegener granulomatosis typically show suggestive but nonspecific findings [1, 51-53]. Although lung tissue is considered most likely to yield the characteristic findings of granulomatous inflammation and vasculitis, not all patients have lung involvement, particularly at the time of presentation [1]. Thus, most patients have their disease diagnosed on the basis of compatible clinical symptoms and histopathologic findings from involved organs. All of the selected studies were done in academic institutions and used tissue confirmation and clinical symptoms to establish the presence or absence of Wegener granulomatosis, but few provided details on the extent or severity of disease in patients with Wegener granulomatosis. Patients with extensive, severe disease might be expected to have a higher rate of positivity than those with limited disease. Third, patients may have differed among the institutions in terms of disease activity at the time c-ANCA was measured. The studies we reviewed provide supporting evidence for these conclusions. The sensitivities and specificities were more homogeneous in the 4 studies that reported disease activity than in the 11 studies that did not.

Fourth, immunosuppressive therapy done at the time of testing may have influenced the number of positive and negative results. Only one study provided information about immunosuppressive therapy at the time of testing [28]; in this study, five of the patients tested, including three who had Wegener granulomatosis, were receiving corticosteroid or cyclophosphamide therapy, or both, at the time of testing. Four of these patients were negative for c-ANCA.

Finally, the composition of the control groups used in these studies may have influenced the reported operating characteristics of the c-ANCA test. In almost 50% of the articles reviewed, the test would not have been clinically warranted for most of the controls. For example, MacIsaac and colleagues [38] screened for c-ANCA in 7500 consecutive samples. This group of samples probably included patients with a minimal likelihood of autoimmune disease or vasculitis. In the study by Ludemann and coworkers [34], almost one third of the control group, or 300 persons, were normal controls. Several of the studies done before 1990 included many normal controls, because the prevalence of c-ANCA positivity was unknown at the time. However, the resulting specificity estimates pertain to a group of persons for whom vasculitis was not an active possibility.

Several methodologic flaws were noted among the studies included in this review. Few studies provided the method used to select patients and controls; the number of patients excluded from the study and the reasons for exclusion; and information on whether physicians were blinded to c-ANCA test results when referring patients to the study. This lack of information may diminish the generalizability of the results to other patients with Wegener granulomatosis. Despite our extensive study selection criteria, the presence of these methodologic flaws is reflected in the overall low-quality scores of most of the included studies.

Our literature review and meta-analysis has several important limitations. First, we used only published data. Negative and weaker findings may not have been published, and this "publication bias" may have led to the overestimation of our calculated test sensitivity and specificity. Second, because only a few articles satisfied our inclusion criteria, our ability to compare different reference standards or assay techniques was limited. Further-more, none of the 15 articles provided information on the long-term follow-up of patients negative for c-ANCA who had clinical syndromes similar to Wegener granulomatosis. Thus, we could not determine how many persons developed the disease over a specified period of time, and we could not calculate the extent to which verification bias may have affected the estimated operating characteristics. This error in study design may have led to an overestimation of test sensitivity. Lastly, we were unable to restrict our analysis to the highest-quality studies. Using the most stringent quality score would have excluded all but 3 studies. Interestingly, 2 of the 3 high-quality studies included in this review reported widely divergent sensitivity estimates [10, 42].

Since it was first described in 1982, c-ANCA has continued to attract the interest of clinicians and investigators. For investigators, the discovery of c-ANCA and its antigens provides exciting new insights into the pathogenesis of Wegener granulomatosis. For clinicians, c-ANCA serves as a useful clinical marker of Wegener granulomatosis in some patients and of active disease in others. On the basis of the results of our meta-analysis, however, we urge clinicians to use c-ANCA testing cautiously when pursuing the diagnosis of Wegener granulomatosis. Although the test sensitivity of c-ANCA appears to be higher for active disease than for overall or inactive disease, a substantial number of false-positive results occur in low-prevalence situations. The false-positive rate may be too high to allow the test to serve as the principal basis for immunosuppressive therapy without obtaining tissue to confirm the diagnosis and exclude other causes. Because no prospective study has tested of a cohort of patients with suspected vasculitis for c-ANCA and then followed them for outcome determination, serious questions remain about the operating characteristics of c-ANCA testing for the diagnosis of Wegener granulomatosis. Moreover, until these concerns are addressed, we recommend judicious ordering of c-ANCA testing and cautious interpretation of the test results, tailored to the patient's clinical presentation, disease activity, and previous probability of Wegener granulomatosis at the time of testing.