Anti-Citrullinated Peptide Antibody (ACPA) Assays and their Role in the Diagnosis of Rheumatoid Arthritis

Abstract

Increasingly, assays for the detection of anti-citrullinated peptide antibodies (ACPA) are used in RA diagnosis. This review summarizes the biologic basis and development of ACPA assays, available ACPA assays and their performance characteristics, and diagnostic properties of ACPA alone and compared to rheumatoid factor (RF) in early RA. We also review correlations, precision, costs and cost-effectiveness, availability, stability and reproducibility of the available assays. Taken together, data indicate that ACPA has a higher specificity than RF for early RA, good predictive validity, high sensitivity, apparent cost-effectiveness and good stability and reproducibility. Given its superior performance characteristics and increasing availability, ACPA is emerging as the most useful single assay for the diagnosis of RA.

Introduction

The diagnosis of early rheumatoid arthritis (RA) has relied upon clinical criteria, including history and physical exam findings, laboratory and radiographic results. Irreversible damage frequently occurs early in RA (1-5). With mounting evidence supporting early diagnosis and aggressive treatment to prevent damage and disability, there is need to improve identification and diagnosis of early RA (6). Until recently, assays detecting rheumatoid factor (RF), antibodies directed against the Fc portion of the immunoglobulin G (IgG) molecule, have been the primary serological tests for RA diagnosis. Anti-citrullinated peptide antibody (ACPA) assays, developed and commercialized in the past decade, are now being employed clinically. Since ACPA are present before the onset of RA symptoms and are predictive of RA development, they are a valuable diagnostic test early in the course of the disease (4).

This review synthesizes currently available data regarding the diagnostic properties of RF and ACPA for the diagnosis of early RA. We focus on ACPA given their recent development and their potential role in the improved identification of early, undifferentiated RA. Data included in this review were obtained from medical literature searches, websites and contact with companies marketing the assays, and information and opinions obtained from experts in the field. We have included information on the biologic basis and development of ACPA assays, the available assays, and data concerning assay performance characteristics, in particular those published in peer-reviewed journals, but also those publicized by manufacturers. Diagnostic properties of these tests, including, but not limited, to sensitivity, specificity, positive and negative predictive values, are reviewed.

Rheumatoid factor (RF)

In 1940, Waaler observed that mixing serum from an RA patient with IgG-sensitized sheep erythrocytes inhibited hemolysis, but caused cell agglutination (7). Rose and colleagues later reported that RA sera agglutinated sheep erythrocytes coated with rabbit anti-sheep erythrocyte antibody more than did sera healthy individuals (8). These findings formed the basis of the earliest RA assay, the Waaler-Rose test.RF assays most commonly detect IgM antibodies directed against the Fc portion of the IgG molecule. The agglutination test measures RF IgM only and remains the most widely used assay. Agglutination assays are reported as either titers or units. Cut-offs for positivity are determined by manufacturers and based upon results from RA patients compared to healthy controls (4, 9). Agglutination assays have sensitivities for RA r from 70-85% and specificities ranging from 40-90%, as agglutination in individuals without RA may occur (10-12).

Other assays for RF have been developed, including enzyme-linked immunosorbent assays (ELISAs), radioimmune assays (RIAs), and laser or rate nephelometry techniques (13). Assays for the detection of IgA and IgG RF are also available (14-19). The sensitivity of RF for RA diagnosis by these techniques is 50-90% and specificity is 50-95%. These wide ranges reflect differences in populations tested (20-26). Studies directly comparing RF detection techniques in cohorts of established RA patients, healthy controls, and patients with non-inflammatory joint disease, have reported latex agglutination test performance to be similar to that of nephelometry and radioimmune assays (12, 27). In a meta-analysis of 50 studies of RF assays from 1998-2005, the pooled likelihood ratios (dependent upon both sensitivities and specificities) were quantitatively similar for IgM, IgA and IgG RF assays, and for using a higher versus lower RF titer for positivity (26) (Table 1).

Table 1.

Positive and Negative Likelihood Ratios for IgM Rheumatoid Factor in the Diagnosis of Rheumatoid Arthritis

	Positive LR (95% CI)	Negative LR (95% CI)	References
Pooled LR	4.86 (3.96-5.97)	0.38 (0.33-0.44)	(9, 14-17, 19, 25, 27, 39, 40, 44, 56, 94, 101, 102, 115-145)
RF Assay Type
Nephelometry	4.15 (2.95-5.84)	0.32 (0.25-0.41)	(17, 25, 46, 94, 116, 118, 119, 121, 123, 125, 127, 128, 130- 133)
Latex   Agglutination	5.05 (3.01-8.50)	0.39 (0.27-0.56)	(9, 27, 44, 102, 115, 117, 120, 122, 126, 134, 137-139, 145, 146)
ELISA	6.13 (4.6-8.17)	0.42 (0.34-0.51)	(14-16, 19, 38-40, 124, 129, 135, 136, 140, 141, 143, 144, 147)
RF Value
≥ 20 U/ml	4.42 (3.02-6.47)	0.39 (0.31-0.50)	(26)
≥ 40 U/ml	5.49 (2.25- 13.38)	0.50 (0.37-0.69)	(26)
≥ 80 U/ml	4.57 (4.60-8.17)	0.42 (0.34-0.51)	(26)

LR: Likelihood ratio; RF: Rheumatoid factor.

Adapted from Nishimura K, et al (26).

Reprinted with permission from Annals of Internal Medicine.

False positive RF results commonly occur in the setting of chronic infections, malignancy, and other rheumatic diseases (21). RF is detected in the sera of 1-4% of healthy young persons and in a higher percentage of elderly persons without RA (28, 29). The RF assay however, is widely available, relatively inexpensive, and understood by both primary care physicians and arthritis specialists (21).

Antibodies to citrullinated peptides

In 1964, Nienhuis and colleagues described an autoantibody they called anti-perinuclear factor. Detected by indirect immunofluorescence test on human buccal mucosa cells, anti-perinuclear factor recognized antigen present in keratohyaline granules surrounding the nucleus (30). Anti-perinuclear factor was present in up to 90% of established RA patients, with a 73-99% specificity (31). Young and colleagues later detected anti-keratin antibodies using indirect immunofluorescence on cryosections of rat esophagus (32). The reported sensitivity of the anti-keratin assay in RA patients of 36-59% and specificity 88-99% (31). Despite the high specificity for RA, these tests were not used widely because of difficulty in standardization of natural substrates and arbitrary interpretation of the indirect immunofluorescence pattern.

In 1995, Sebbag and colleagues demonstrated that both of these antibodies belonged to a family of autoantibodies directed against citrullinated fillagrin, an epithelial cell protein (33). Citrullination is posttranslational modification of arginine to citrulline by the enzyme peptidyl arginine deiminase (PAD). This process occurs naturally during inflammation, apoptosis and keratinization (9). While fillagrin is not present in the synovium (34), several citrullinated proteins, including fibrinogen and fibronectin, are present in RA synovium, and other citrullinated epitopes have been identified as targets of highly RA-specific autoantibodies (35-37). In 1998, Schellekens and colleagues produced synthetic linear citrullinated peptides derived from human fillagrin, easily detected by ELISA with enhanced sensitivity and no loss of specificity (35). To improve antigen composition and antibody recognition, a cyclic citrullinated peptide (CCP) was developed (38).

The first commercially available ACPA assay (1^st generation or CCP1) was developed by Eurodiagnostica, used in early studies (2000–2001). This ELISA–based test employed a single cyclic citrullinated peptide derived from fillagrin.(38). The assay detected autoantibodies in 53% of established RA patients with 96% specificity (38).

Peptide libraries were then screened for better epitopes Since 2000, 2^nd generation cyclic citrullinated peptide (CCP2) and 3^rd generation cyclic citrullinated peptide (CCP3) assays have been developed. Several companies market these assays for RA diagnosis. CCP3 assays rely upon additional epitopes not present in the CCP2 antigen sequence (39, 40). Apart from the main difference in substrate, both CCP2 and CCP3 use ELISA methods and similar dilutions (1:101), diluents, controls, conjugates, and rinterpretation. AxSYM Anti-CCP utilizes microparticle enzyme immunoassay (MEIA) for the semi-quantitative determination of the IgG class of auto-antibodies specific to CCP 2. Most studies, however, show no evident improvement of CCP3 compared to CCP2 assays (41-43). The compositions of many new CCP3 peptides are not yet publicly available as patents are pending. The anti-CCP3.1 assay marketed by INOVA detects both IgG and IgA CCP3 antibodies in an effort to increase sensitivity (41). Eurodiagnostica has developed a “point-of-care” assay, employing a finger lancet to obtain a drop of blood for rapid office-based results.

Newer assays detect non-cyclic citrullinated peptides (41); the term anti-citrullinated peptide antibody (ACPA) has thus replaced anti-cyclic citrullinated peptide (anti-CCP) antibody. Citrullinated vimentin is present in synovial fluid. Anti-Sa antibodies, directed against it, are detectable in RA synovium (44, 45). Anti-Sa antibodies have reported sensitivity of 20-25% and specificity of 95% in early RA (46). An ELISA for the detection of autoantibodies against mutated citrullinated vimentin (anti-MCV) has better sensitivity than anti-Sa antibody. The sensitivity of anti-MCV is comparable (or even higher in some studies) to that of ACPA (82% vs 72%) (47), while specificity of anti-MCV is slightly lower than ACPA in several studies (90-92% vs. 96-98%) (41, 48). Unlike ACPA assays, the anti-MCV levels may correlate with disease activity (47, 49).

Pathogenetic role of ACPA in RA

The roles of citrullinated peptides and autoantibodies to them in RA pathogenesis remain unclear. ACPA are strongly associated with an increased risk of developing RA in healthy individuals and are detectable in the blood of healthy persons prior to clinical RA (14, 50, 51). Among those with RA, their presence is associated with more severe structural damage, radiographic progression and poorer response to therapy. (26, 38, 52-60,66). Geneticists and epidemiologists hold ACPA-positive RA to be a homogeneous phenotype of severe RA. ACPA is strongly associated with the HLA-DRB1 shared epitope (61) and PTPN22 (62, 63), strong genetic risk factors for RA, and smoking (64, 65), the strongest known environmental risk factor for RA. Smoking by individuals with inherited HLA-DRB1 shared epitope genes may trigger RA-specific immune reactions to citrullinated peptides, the generation of ACPAs and, ultimately, disease (64).

ACPA reproducibility and stability over time

In stored blood bank samples, Nielen and colleagues detected ACPA antibodies present up to 14 years prior to RA onset, with gradually increasing prevalence and increased sensitivity and specificity for RA compared to RF (51). The duration of the preclinical autoantibody positive, symptom-free period prior to RA may iincrease with increasing age (60). In a 3-year study off 97 individuals with RA, ACPA status was relatively stable: three ACPA positive subjects became negative, while two ACPA negative subjects became positive (67). Decreases in ACPA may be observed with some RA therapies, but generally patients do not lose their positive results (68-72). Although in some small studies ACPA levels paralleled RA disease activity (68, 69, 73-75), this has not been corroborated in subsequent studies and ACPA assay results are not employed clinically to monitor disease activity (70-72).

Currently Available ACPA Assay Performance Characteristics

Several ACPA assays are currently approved by the U.S. Food and Drug Administration (FDA) (Table 2). The ACPA assays employed by European and Canadian early arthritis cohorts are mainly CCP2 assays (Diastat™ from Axis–Shield, Immunoscan-CCP Plus™ from Eurodiagnostica, and ELIA-CCP™ from Phadia, and Quanta Lite from Inova, etc). Most currently available assays are kits employing a substrate derived from the synthetic cyclic peptide described by Schellekens and colleagues (38, 41), but differ in incubation time, volume and dilution of serum, type of conjugate and of enzymatic substrate, and range of units reported and thresholds for positive results (41, 42, 76-78). To determine the diagnostic performance, manufacturers have tested established RA patients meeting the 1987 ACR criteria (79), and healthy individuals. Sensitivities range from 60-80% and specificities from 85-99%. CCP2 assays have slightly higher sensitivity than CCP1 assays; the newest non-cyclic ACPA assays report similar performance compared to CCP2 (42, 76-78, 80, 81).

Table 2.

Available ACPA assays on the market (1987 ACR Criteria for RA used for calculating sensitivity and specificity in most studies).

Name	Manufacturer	Type of Assay (generation)	Sensitivit y for RA *	Specificity for RA *	Within- run assay CV *
Diastat	Axis-Shield Diagnostics Scotland, UK	ELISA (2nd)	76.5%(77)	86.1%(77)	13.6%(148)
CCPoint	Euro- diagnostica Netherlands	Colloidal gold immunoassay (2nd)	56.0%(149)	98.8%(149)	N/A
CCPlus	Euro- diagnostica Netherlands	ELISA (2nd)	70%(150)	97.5%(150)	0.4- 5.1%(151)
EDIA	Euro- diagnostica Netherlands	ELISA (2nd)	66.7%(78)	97%(78)	1.9- 7.9%(151)
RA anti- CCP ELISA	Euro- diagnostica Netherlands	ELISA (2nd)	76.5%(41)	95.4%(41)	12.6- 34.3%(41)
Euroimmu n	Euroimmun Germany	ELISA (2nd)	72.5%(41)	96.4%(41)	6.4- 12.1%(41)
Quanta Lite	Inova United States	ELISA (2nd)	70%(77)	91.3%(77)	6%(148)
ELIA CCP	Phadia Sweden, Germany	Immunocap method (2nd)	77.5%(41)	95.9%(41)	7.2- 9.8%(41)
Quanta Lite CCP3	Inova United States	ELISA (3rd)	77.5%(41)	87.8%(41)	3.7- 5.1%(41)
Quanta Lite CCP3.1	Inova United States	ELISA (3rd)	74%(133)	89.6%(133)	0.5- 4.8%(133)
Org 548 anti-MCV	Orgentec Germany	ELISA mutated citrullinated vimentin	74.5%(41)	90.3%(41)	8.4- 12.3%(41)

^*Value in parenthesis represents the respective reference numbers.

As ACPA assays are based on detection of autoantibodies by ELISA or MEIA or immuno-enzymofluorimetry, reactivity is related to the quantity of antibody present in a non-linear fashion. While changes in antibody concentration are reflected in a corresponding rise or fall in results, the change is not proportional in most assays (i.e. a doubling of the antibody concentration will not double the reactivity) (41). In a head-to-head comparison of the technical performance of six different commercially available ACPA assays, Inova, Eurodiagnostica and Genesis (41) demonstrated significant deviation from linearity; the best linearity was achieved by Euroimmun.

ACPA assay precision

Studies comparing different ACPA assays have concluded that the majority of assays are precise, with within-assay (intra-assay) coefficient of variations (CVs) for most available assays ranging from 4-19% (41, 78). In a study by Coenen and colleagues comparing six ACPA assays, the greatest precision was found with Genesis (4.8-5.9% intra-assay CV) and Inova assays (3.7-5.1% intra-assay CV) and the lowest with the Eurodiagnostica assay (12.6-34.3% intra-assay CV) (41, 78).

ACPA Assay correlation

Although different antigens and methods are employed to quantitate and report ACPA, the results, expressed as positive or negative values, are highly correlated among commercially available ACPA assays, with correlation coefficients from 0.59 to 0.96 (Table 3) (41, 78). Vander Cruyssen and colleagues studied four ACPA assays, including INOVA’s CCP3 assay. They found that discrepancy between the ACPA assays was due to borderline results, inter-assay variability and inter-test variability. The lowest intertest discrepancy is observed between tests using the same substrate (82). If one false positive ACPA was found in an individual without RA, there was a high probability that ACPA would be negative in a different ACPA assay (82).

Table 3.

Spearman Rho Correlations Comparing Quantitative Anti-Citrullinated Peptide Antibody Assays

	INOVA CCP	INOVA 3 CCP	Euroimmun CCP	Euro- Diagnostica CCP	Axis- Shield CCP	Edia CCP	Pharmacia CCP	Triturus CCP
INOVA 3 CCP	0.86
Euroimmun CCP	0.84	0.76
Euro- Diagnostica CCP	0.96	0.83	0.8
Axis-Shield CCP	0.83	0.71	0.81	0.8
Edia CCP	0.86	0.82	0.93	0.62	0.93
Pharmacia CCP	0.82	0.66	0.71	0.79	0.8	0.87
Triturus CCP	0.75	0.74	0.77	0.71	0.67	0.7	0.59
IgM RF Nephelometry	0.67	0.64	0.63	0.65	0.59	0.73	0.58	0.48

RF: Rheumatoid factor

Reprinted with permission from Clinica Chimica Acta(78)

Development of an international reference range for standardized ACPA reporting

Given the variety of ACPA assays, quantitative results are not currently comparable between studies. Work is underway to develop standardized ACPA units. (83). Results were promising, but require additional confirmation in large numbers of samples and acceptance by assay manufacturers (84).

ACPA assays in other diseases

While the specificity of ACPA assays for RA compared to healthy individuals is good, the potential for lower specificity in the setting of other inflammatory disorders, such as psoriatic arthritis, scleroderma, systemic lupus erythematosus (SLE) and seronegative spondyloarthropathies is of concern (85). The presence of immune complexes or other immunoglobulin aggregates can cause increased non-specific binding and false positive results.

We identified and reviewed 63 studies that examined the cross-reactivity rate of ACPA in non-RA rheumatic diseases and common infections. The highest frequency of ACPA positivity in non-RA autoimmune conditions are found in psoriatic arthritis (9%), SLE (8%), and juvenile idiopathic arthritis (JIA) (8%), as well as scleroderma and CREST syndrome (7%), followed by Sjögren’s syndrome (6%) and vasculitis (5%) (Table 4). As many patients in these studies do not have long-term follow-up, they may have ultimately been diagnosed with ACPA-positive RA or an overlap syndrome. For example, 7 of 126 psoriatic arthritis patients with detectable ACPA had more severe, erosive disease and high prevalence of the RA-associated HLA-DRB1 shared epitope (86). A high frequency of ACPA positivity has been observed in patients with erosive arthritis and overlap syndromes with features of scleroderma and SLE (41, 43, 78, 82, 87-91). ACPA in JIA has been associated with RF-positive disease similar to RA in adults (92).

Table 4.

Detection of ACPA in other diseases

	n	ACPA+ (n, %)	Refs.
Psoriatic arthritis	1343	115 (8.6%)	(41, 78, 82, 86, 87, 152-156)
SLE	1078	84 (7.8%)	(41, 78, 82, 89-91)
Sjögren’s syndrome	609	35 (5.7%)	(41, 78, 87, 89, 91)
Spondyloarthropathy	431	10 (2.3%)	(78, 89, 104, 149, 155)
Scleroderma/CREST	380	26 (6.8%)	(41, 43, 78, 87, 88, 91)
Hepatitis C/cryoglobulinemia	285	10 (3.5%)	(87, 157, 158)
Osteoarthritis	182	4 (2.2%)	(78, 87, 89, 91, 149)
Hepatitis B	176	1 (0.6%)	(159)
Juvenile idiopathic arthritis	169	13 (7.7%)	(87, 91, 152, 160-162)
Polymyalgia rheumatica	146	0 (0%)	(89, 91, 149, 163)
Vasculitis/ Wegener’s granulomatosis	107	5 (4.7%)	(78, 87, 89, 91, 104)
Tuberculosis	96	33 (34.3%)	(93, 164)
Polymyositis/dermatomyositis	75	0 (0%)	(41, 78, 82, 87)
Fibromyalgia	74	2 (2.7%)	(78, 149)
Gout and pseudogout	58	0 (0%)	(87, 89, 149)

The surprisingly high prevalence of ACPA in active tuberculosis has been studied by Kakamanu and colleagues (93). They reported that reactivity to uncitrullinated arginine-containing residues was common in tuberculosis, but not in RA. The mechanism of induction of ACPA in active pulmonary tuberculosis known. ACPA levels decreased somewhat, but not rapidly, after treatment for tuberculosis (93).

Comparison of 1^st, 2^nd, and 3^rd Generation and newer ACPA assays

Given the rapid evolution of ACPA assays, establishing the comparative sensitivity and specificity of the three “generations” of assays is crucial if they are to be used interchangeably. CCP2 and CCP3 assays offer slightly improved sensitivity over that of CCP1 assays (85, 94), although they have similar specificity for RA (86-96%). CCP2 and CCP3 assays in most (41-43), but not all studies (91), have had similar performance characteristics with sensitivities 68-79% and specificities 86-96% (26, 43, 75, 78, 81, 95).

New anti-MCV assays also have similar performance, with sensitivities 70-82% and specificities 90-98% (47, 48, 96, 97). Higher false positives rates have been reported with Orgentec™ (anti-MCV) and Inova Quantalite™ (CCP3) assays (41). There is some lack of agreement between the results obtained from different ACPA assays on same subjects, which could be partially attributed to borderline results and inter-assay variability. One study has shown 18% discrepancy between two different ACPA assays tested on RA patients (82).

Cost and availability of RF and ACPA assays

RF assays have been widely used for years and are familiar to general practitioners. They are relatively inexpensive and easy to obtain. Since 2000, when ACPA assays were first introduced, the availability of these tests has drastically increased and costs have decreased. They are now marketed almost worldwide by a variety of companies. The Diastat™ CCP2 assay from Axis-Shield, for example, is sold globally. It has the approval of the U.S. Food and Drug Administration (FDA), the European Medicines Agency (EMEA) and the Japanese Ministry of Health and Welfare (MOHW). The price per kit varies from market to market, but is approximately $US 250-300 per 96 well kit. Immunoscan™ 2^nd generation ACPA assay 96 well kits from Euro-Diagnostica are currently marketed for 350-400€, $US 500-600, or £250-300 in Great Britain. Fully automated and point-of-care assays are beginning to be marketed by several companies.

Konnopka and colleagues performed a cost effectiveness analysis to address the incremental benefit of testing for ACPA in addition to the current ACR criteria for RA classification (98). They developed a Markov model of the 10-year progression of RA in patients presenting with undifferentiated arthritis, and estimated the effects of ACPA testing on incremental costs and quality-adjusted life years, including the impact of late diagnosis and treatment. Their analysis revealed that up-front use of ACPA testing, rather than waiting and testing after a few years of symptoms, was cost effective, and, when indirect costs were incorporated, saved in the range of 1000€ per quality-adjusted life year. While based on multiple assumptions, this study does provide evidence for changing the current approach to early inflammatory arthritis.

Diagnostic accuracy of ACPA assays

More than 300 studies have been published concerning the diagnostic accuracy of ACPA assays in RA diagnosis (26). These studies vary substantially in focus: some have addressed technical aspects, while others have compared the diagnostic accuracy in different populations of individuals (early or established RA; patients with other diseases or healthy controls). The studies are heterogeneous in their comparison of ACPA assay utility to other tests, including IgM, IgA and IgG RF (26) and their use of a gold standard for RA diagnosis (most often the existing 1987 American College of Rheumatology criteria for the classification of RA (79)).

In studies of early or undifferentiated RA, ACPA testing is generally more specific and equally sensitive to RF (Table 5). In cohorts containing both established and early RA, the performance characteristics of the two tests are comparable (Table 4). The definition of early arthritis or early RA has varied in these studies. In the majority, early arthritis has been defined as symptom duration of less than 2 years (median of approximately 2 months) and initial serologies of patients who developed RA have been compared to those who did not (14, 26, 38-40, 46, 51, 56, 59, 76, 99-102, 103, 104, 105, 106, 107-111). Most of these data are from the prospective follow-up of early arthritis cohorts in Japan, the Netherlands, and Austria.

Table 5.

Comparison of performance characteristics of RF and ACPA (CCP2) assays in early RA cohorts and cohorts containing both early and established RA

	ACPA (CCP2)	IgM RF	ACPA (CCP2) or IgM RF	ACPA (CCP2) and IgM RF
Early RA Cohorts (38, 46, 58, 76, 101, 102, 104, 106, 111, 119)
Sensitivity range, %	41-63	41-66	52-67	33-58
Specificity range, %	91-100	87-97	72-82	98-100
Early and Established RA Cohorts (9, 41, 82, 96, 102, 106, 120, 124, 133, 155, 165)
Sensitivity range, %	41-77	62-87	70-81	33-57
Specificity range, %	88-98	43-96	80-91	91-99

RF and ACPA: One, either or both in Early, Inflammatory Arthritis?

Given the substantial overlap between the diagnostic performance and utility of RF and ACPA for the diagnosis of RA, the marginal diagnostic value of adding one test to the other and the added value of performing both must be addressed. In particular, the challenge is to decide on the combination of assay or assays that offers superior performance for the identification of RA among patients presenting with early, undifferentiated inflammatory arthritis. Although correlated, RF and ACPA assays detect different underlying biological phenomena in RA, and thus agreement between assay results is not static, but likely fluctuates during disease course (103).

In our review of data from early RA cohorts, ACPA was slightly more specific than RF, but the two assays have equivalent sensitivity (Table 5). The positive predictive value for ACPA in the setting of early undifferentiated arthritis is 78-96% in t early RA cohorts, with most values in the low to mid 90% range and the negative predictive value is 62-96% (38, 46, 58, 76, 101, 102, 104, 111). The positive predictive value for RF is broader, from 36-97% with most values in 70-80%, and the negative predictive value is 69-95%. Positive ACPA results may be particularly helpful in the setting of a negative RF. The positive predictive value of a positive ACPA test was 91.7% among 260 IgM-RF negative early arthritis patients followed for one year (76). Employing both ACPA and RF positivity further increases specificity and positive predictive value to above 95%, but decreases sensitivity substantially. When either ACPA or RF positivity are required, the sensitivity is somewhat increased (52-67%), but specificity is similar to that of RF alone (72-82%) (102, 104).

In cohorts containing both established and early RA, the performance characteristics of RF and ACPA are comparable and the sensitivity of both RF and ACPA is improved,,(although the ranges of performance characteristics are large and tdata are mixed). A strategy requiring either ACPA or RF may improve sensitivity for both early and established RA. In one study, the presence of either ACPA or RF increased testing sensitivity for RA from 66% (ACPA) and 72% (RF) to 81%, with a good specificity of 91% (9). The specificity of requiring both to be present is comparable to that of ACPA alone.

The addition of ACPA testing improved the sensitivity of the 1987 ACR criteria (which rely upon the presence of RF as one of the 11 possible criteria, 4 of which must be present) for the correct classification of early RA subjects (112). Adding ACPA results to the 1987 criteria increased sensitivity for early RA (≤ 6 month disease duration) from 25 to 44% and did not change the specificity of 86%. ACPA also played an important role in a rule developed by Van der Helm-van Mil and colleagues to predict which patients with undifferentiated arthritis would progress to RA (113). Five hundred and seventy patients with undifferentiated arthritis in the Leiden Early Arthritis Center were selected and reassessed at one year for RA development. The prediction rule consisted of nine variables: sex, age, location of symptoms, morning stiffness, tender joint count, swollen joint count, C-reactive protein, and RF and ACPA positivity. ACPA was one of the strongest predictors, and if positive, a subject received 2 points (113). A modified form of this prediction rule was validated in three cohorts of patients with recent onset undifferentiated arthritis and was found to have excellent discriminative ability to assess progression to RA (114).

ACPA assays are increasingly available and affordable. The assays have good predictive validity as ACPA are associated with known genetic and epidemiologic risk factors for RA and therefore identify a population of RA patients with more severe, erosive joint disease that is at high risk for rapid joint destruction. Positive and negative results are highly correlated between current assays. International standardization of reporting units is underway and will facilitate inter-assay comparisons. Both CCP2 and CCP3 assays have improved upon CCP1 assays, and have comparable diagnostic utility, with sensitivities of 68-79% and specificities from 86-96% for RA (26, 75, 78, 81, 95)

We did not obtain all data, published and unpublished, from past comparisons of RF and ACPA assay performance or perform a formal meta-analysis. We did review published studies and presented sensitivity and specificity ranges of assays, alone and in combination in both early and established RA cohorts. Our results suggest that ACPA assays offer a slight advantage over RF (including high titer RF and combined IgM, IgA and IgG RF levels) due to higher specificity. RF and ACPA are two different autoantibody systems that do not measure or reflect the same underlying biology (103). While there is substantial correlation between ACPA and RF seropositivity within patients, the ACPA assay may be especially valuable in predicting RA in patients who are RF-negative but nevertheless have a high probability of RA (76). If the role of the assay is to aid in the identification of patients developing RA among those presenting with early undifferentiated symptoms, a high-risk population with a high prevalence of disease (rather than screening the general population), the positive predictive value of the ACPA assay is on the order of 95% (76).

ACPA assays have high specificity, high predictive validity, high specificity, apparent cost-effectiveness and good reproducibility for the diagnosis of early RA. In prior studies, accepting either ACPA or RF positive assay results for the diagnosis of RA did not improve upon testing for RF alone and requiring both assays to be positive for diagnosis is a very specific, but not extremely, sensitive approach. Ultimately, the decision to use one or both tests depends upon the population tested, the indications for the testing, and the inherent trade-off between sensitivity and specificity.

Summary and Overall Recommendations

• ACPA assays have good predictive validity in that they are associated with the known genetic and epidemiologic risk factors for RA and identify a population of RA patients with more severe, erosive joint disease, at high risk for more rapid joint destruction.

• RF and ACPA are two different autoantibody systems and do not measure or reflect the same underlying biology.

• ACPA assays are becoming increasingly available and less expensive. Cost-effectiveness analyses suggest that up-front testing of ACPA in patients presenting with undifferentiated arthritis is cost-effective, in particular in terms of the saved indirect costs of delayed diagnosis.

• ACPA offers similar sensitivity, but higher specificity for RA than RF in early RA. When used in the identification of patients potentially developing RA among those presenting with early undifferentiated symptoms, a high risk population, (rather than screening the entire population), the prevalence of disease will be high and the positive predictive value of the ACPA assay is on the order of 95% (76).

• In the setting of a relatively high clinical suspicion (pre-test probability) and a positive ACPA result, the patient has a high likelihood of having or developing RA. If ACPA is negative, further testing may be indicated depending on the level of clinical suspicion.