Rheumatoid arthritis (RA) is a common autoimmune disease characterized by chronic inflammation of the joints, which can ultimately lead to cartilage and bone destruction. In the past decade it has become apparent that citrullinated proteins/peptides, and in particular autoantibodies directed to them (anti-citrullinated protein antibodies (ACPA)), are likely to be involved in the development of this disease in at least 70% of the patients (reviewed in [1]). In the clinical setting, ACPA have mostly been detected using the anti-cyclic citrullinated peptide (anti-CCP) test, although more recently other tests using various citrullinated proteins have also been employed.

Recently it became clear that RA patients can be classified into two major subsets; namely, those who have ACPA (anti-CCP(+)) and those who do not (anti-CCP(-)) [2]. Whilst in the early phase of the disease these two groups of patients show a very similar clinical presentation, the picture changes considerably as the disease develops further. The presence of ACPA at early diagnosis predicts more pronounced radiographic progression, as demonstrated by many studies showing a strong association between anti-CCP positivity and the development of bone erosions. Importantly, environmental risk factors (for exam ple, tobacco smoking) differ to a large extent between these two populations [3], and the risk of developing ischemic heart disease is clearly higher in anti-CCP(+) patients compared with anti- CCP(-) RA patients [4]. Furthermore, treatment response to, for example, synthetic disease-modifying antirheumatic drugs such as methotrexate may differ between these groups of patients [5]. It is therefore important for a clinician to be able to accurately separate anti-CCP(+) patients from anti-CCP(-) patients. During such a decision-making process it is important that both clinicians and laboratory specialists are fully aware of the advantages and disadvantages of the various ACPA tests that are commercially available.

The present review intends to critically review the literature on comparisons between the most frequently applied commercial tests in terms of specificity and sensitivity of ACPA detection.

Anti-CCP antibodies of these patients can be eluted from the CCP2 ELISA plate (by low pH or high salt) and the eluate can subsequently be used to stain western blots containing different citrullinated proteins, such as fibrinogen, histones or vimentin. The eluted antibodies react with all of these citrullinated proteins, indicating broad cross-reactivity between anti-CCP and these various antigens (R Toes, personal communication). These data have been complemented by studies of synovial exosomes from RA patients, which were shown to contain ACPA as well as a number of citrullinated constituents - for example, citrullinated fibrinogen peptides and citrullinated Spα (a CD5 antigen-like protein) [7]. Moreover, the number of anti-CCP(-) sera that show reactivity with other citrullinated antigens is very small [8, 9]. Taken together, these data indicate that the vast majority of ACPA can be detected by the CCP2 test.

Early diagnosis of RA coupled with rational use of disease-modifying antirheumatic drugs has been shown to have a favorable effect on the course of the disease. Early and accurate diagnosis has therefore become increasingly important. For several reasons, the presence of anti-CCP antibodies is a great help to clinicians in deciding which patient needs early treatment.

First, anti-CCP antibodies are very specific for RA, and they are produced at significant levels very early in disease. The specificity of anti-CCP antibodies for the diagnosis of RA is high, as shown in Table 1. In addition, it is known that anti-CCP antibodies can be present many years before the first visit to the clinic (up to 18 years) [10–12].

Second, studies of early arthritis cohorts have shown that a large number of these early arthritis patients cannot be accurately diagnosed at their first visit, and hence are often referred as undifferentiated arthritis patients. If patients are found to be anti-CCP(+) when referred to the clinician, however, more than 90% develop RA within 3 years - in contrast to only 30% of the anti- CCP(-) patients. The presence of anti-CCP antibodies in undifferentiated arthritis therefore accurately predicts development of RA [13, 14].

Third, the presence of anti-CCP antibodies at the first visit to the clinician predicts radiographic progression, as demonstrated by many studies that have shown a strong association of anti-CCP positivity with the development of bone erosions [1, 15, 16]. In the past, IgM rheumatoid factor (RF) positivity was assumed to predict radiographic progression, but a recent report clearly indicates that the radiographic progression seen is actually associated with ACPA(+)/RF(+) and ACPA(+)/RF(-) RA, but not with ACPA(-)/RF(+) and ACPA(-)/RF(-) RA [17]. The conclusion therefore seems to be that the presence of ACPA as such is associated with an erosive course and the presence of IgM RF is just a co-expressed autoantibody, as has been known for a long time.

Fourth, more germinal centers in synovial tissue infiltrates are found in anti-CCP(+) RA patients [18]. It is known that germinal centers contribute to RA pathogenesis by supporting autoantibody production [19]. In the same report, distinct synovial features such as increased fibrosis in the synovial tissue and a thicker synovial lining layer were found in anti-CCP(-) RA patients [18].

In conclusion, although at baseline the clinical features of both RA subsets are very similar, anti-CCP(+) RA is more strongly associated with poor outcome than anti- CCP(-) RA (reviewed in [20]).

This conclusion may also have important implications for treatment. Synthetic disease-modifying antirheumatic drugs, such as methotrexate, are often used in treating RA, frequently in combination with TNFα blockers to enhance the treatment response. In a large Dutch study - the so-called PROMPT study (methotrexate versus placebo treatment) - Van Dongen and coworkers found that anti-CCP(+) patients responded well to methotrexate treatment, while a parallel anti-CCP(-) patient group did not [5]. In a subsequent study it was shown that methotrexate treatment resulted in a more favorable response in patients with a low or intermediate pretreatment level of ACPA [21]. These data not only suggest that the effectiveness of a drug can be different in anti-CCP(+) as compared with anti-CCP(-) arthritis, but also that very early treatment, even in patients not yet fulfilling American College of Rheumatology (ACR) criteria for RA, can be beneficial if used in a selective way [20].

In the past 3 years about 25 articles have appeared in which performances of different ACPA tests were compared. Only nine of these studies, however, followed the essential principle that any such comparisons can only be made when the test results are properly stratified. Stratification means that sensitivity values have to be calculated at a predefined specificity (mostly 98% or more) using the same cohort of RA patients and disease control sera. The chosen specificity should be as high as possible without compromising essential sensitivity, and thus the overall diagnostic efficiency. Another important point is the cohort of patients selected for such studies.

The main advantage of ACPA lies in their proven ability to predict the development of RA and their potential use as a criterion for RA even at baseline [22]. The ideal cohort to evaluate the clinical value of ACPA tests is therefore the mixed population of patients visiting an early arthritis clinic where some patients eventually will develop RA [6]. The use of a cohort of established RA patients with longstanding disease is clearly less useful. In the studies reviewed in the present article, all sorts of RA patient cohorts and control groups have been used, and this, at least in part, explains some of the differences in sensitivity and specificity of ACPA tests between different studies.

Several studies have addressed the diagnostic performance of the MCV assay. In stratified studies this test shows a lower sensitivity (see Table 2), and in nonstratified studies a lower specificity, than the CCP2 test [27–30]. A similar conclusion was reached in a recent review on the diagnostic and prognostic properties of the MCV assay [31]. There are a few reports indicating that anti-MCV is present in a significant number of anti-CCP(-) sera [21, 32–34], and this subgroup of patients appears to have a higher rate of radiographic destruction than seronegative patients. Anti-MCV positivity therefore seems to indicate poor radiographic prognosis in a larger group of RA patients than anti-CCP positivity does [34]. Data from the study of Van der Linden and collaborators, however, indicate that the presence of anti-MCV antibody with negative anti-CCP does not strongly affect the level of joint damage in RA [17]. It is clear that further evaluation of the MCV antibody in clinically wellcharacterized cohorts is needed. It would also be useful to introduce a control mutated vimentin antigen (not citrullinated) in order to test whether this CCP-negative population of anti-MCV antibody is directed to the vimentin protein or to the citrulline moiety of mutated vimentin [9, 35].

Only the CCP3 test appears in some studies to be comparable with the CCP2 test [36–38], although in the majority of published studies the CCP3 test shows a somewhat lower specificity and/or sensitivity [23, 24, 39–43]. Based upon the combined results of these studies it can be concluded that the CCP3 test has no apparent diagnostic advantage compared with the CCP2 test. Inova has also recently introduced the CCP 3.1 test, which additionally includes measurement of IgA antibodies. In general this test does not appear to be better than the CCP3 test [23, 41], and the CP 3.1 test would appear to have a limited usage in a routine laboratory setup.

Citrullinated peptide antigens have also been derived from proteins like Epstein-Barr virus nuclear antigen (Astra) and IgG (Aesku). The tests with these peptides were included in the stratified study of Bizzaro and collaborators [23], but scored rather low in sensitivity (Aesku 44%, Astra 47%). In a single study, the Genesis cFil test shows a positive predictive value that is lower than that of the CCP2 test, but higher than the positive predictive value of the CCP3 and MCV tests [44]. Finally, Lutteri and coworkers also measured the diagnostic abilities of a new test using synthetic citrullinated peptides (RA/CP; Triturus, Bad Kreuznach, Germany). The specificity, sensitivity, positive predictive value and negative predictive value of this test were all much lower than those of the CCP2 and CCP3 tests [24].

As outlined above, ACPA are not only important predictors of RA development but are also among the most potent predictors of the outcome of RA, as measured by the rate of radiographic joint destruction. There are at least four clinically important reasons to perform and compare ACPA tests: the ability to confirm or predict the development of RA with the highest reliability; the ability to predict radiographic progression; the ability to predict remission; and the ability to predict response to anti- TNFα treatment.

First, from Tables 1 to 3 it is clear that although most ACPA tests are perfectly able to predict or confirm a diagnosis of RA, none of the tests has a better diagnostic record than the CCP2 test. In addition, it was reported recently that the positive predictive value for predicting progression from undifferentiated arthritis to RA was highest for CCP2 (67.1%). Combinations of two or more ACPA tests appear to give no additive value [17]. A somewhat different situation may exist for the additional testing of RF, since RF testing may carry additional clinical value beyond testing for anti-CCP alone [45]. It should be noted that RF positivity still is a criterion for RA, and that RF-positive/anti-CCP(-) RA patients display different environmental risk factors than those that are only anti-CCP(+) [3].

Second, a positive test for anti-CCP2, anti-CCP3, or anti-MCV was associated with a higher Sharp/van der Heijde score at all time points except baseline and was also associated with a higher rate of joint destruction over a period of 7 years [17]. There was no difference between these tests with regard to their ability to predict radiographic progression. The use of a second or third autoantibody test did not increase the predictive accuracy for the rate of joint destruction [17]. Similar results were reported by Dejaco and colleagues [36], Majka and colleagues [46], and Syversen and colleagues [47]. It is also interesting to note that not only in the presence of ACPA, but also in the absence of these antibodies, RF did not significantly correlate with increased rates of joint destruction. This indicates that RF, in contrast to ACPA, does not by itself contribute to disease progression. In some studies MCV was reported to have a somewhat higher sensitivity (mostly accompanied by a lower specificity) than CCP2 (for example [27, 32–34]). These studies, however, also showed that the rate of joint destruction in MCV-positive/CCP(-) patients was comparable with that in patients lacking ACPA, indicating that the presence of anti-MCV antibody alone does not affect the level of joint damage in RA [17, 33, 48].

Third, the test's ability to predict the likelihood of not achieving sustained disease-modifying antirheumatic drug-free remission was highest (11.6%) for anti-CCP2 and varied between 4.7 and 6.0% for anti-CCP3, anti- MCV and RF [17]. Again, performing two ACPA tests had no additional value compared with the anti-CCP2 test alone. It is clear, however, that we need more data on this aspect of ACPA testing.

Finally, treatment of RA is mostly assumed to combat important disease mechanisms and thus lower the inflammation. As a consequence one might also expect a reduction of the activation of autoreactive B cells followed by reduced ACPA levels, which would allow monitoring of the effect of treatment. In the initial studies no significant effect of infliximab on anti-CCP levels was observed (reviewed by Zendman and collaborators [49]). Also in later studies neither anti-CCP3 nor anti-CCP2 levels were found to be influenced by TNFα blocking agents, and the test results failed to predict responses to anti-TNFα treatment [36]. A possible reason for these observations may be that although the inflammation may decrease, the citrullinated antigens (and consequently the production of autoantibodies and immune complexes) are still there, and this is reflected in the antibody levels (see [1]). In other studies, however, significant decreases of anti-CCP2 and anti-MCV titers at 18 months and/or 24 months of infliximab treatment have been reported [27]. At the moment we have to conclude that none of the available ACPA tests unequivocally shows the ability to predict response to treatment.