Article Text

Download PDFPDF

The 2010 ACR/EULAR criteria for rheumatoid arthritis: do they affect the classification or diagnosis of rheumatoid arthritis?
  1. Annette H M van der Helm-van Mil,
  2. T W J Huizinga
  1. Department of Rheumatology, Leiden University Medical Center, Leiden, The Netherlands
  1. Correspondence to Annette H M van der Helm-van Mil, Department of Rheumatology, Leiden University Medical Center, PO Box 9600, Leiden 2300RC, The Netherlands; A.H.M.van_der_Helm{at}lumc.nl

Abstract

Rheumatoid arthritis (RA) is diagnosed based on phenotypic characteristics. The 2010 ACR/EULAR criteria were derived with the aim of classifying RA earlier in the disease course than the 1987 ACR criteria. When reviewing thus far published validation studies, it is clear that the 2010 criteria can be fulfilled earlier in time than the 1987 criteria. Therefore, the taskforce that derived the 2010 criteria has succeeded in their main objective. Furthermore, it has been repeatedly shown that the sensitivity of the 2010 criteria is increased compared with the 1987 criteria, but the specificity decreased. As classification criteria aim to arrive at homogeneous groups of patients in order to compare the results of clinical or experimental studies, this decrease in specificity is of concern, as patients with diagnoses other than RA can test positive on the criteria. With regard to diagnosing RA, the overall trend in the data is that early arthritis patients in the rheumatological setting who fulfil the 2010 criteria have a high probability of the disease. Not fulfilling the criteria, in contrast, does not rule out RA in these individuals.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Rheumatoid arthritis (RA) is diagnosed based on phenotypic characteristics and not on specific pathogenetic mechanisms that are underlying to the phenotype. Most likely several different pathophysiological processes end up in the phenotype RA. These different pathways are incompletely identified. Intervening with specific pathways in explicit subphenotypes or subgroups of RA, although presumably very effective, is at present far from being possible. Initiation of disease-modifying treatments early is a treatment strategy that, independent of the underlying pathway, is relevant for the outcome of RA. Pivotal to this strategy is the ability to diagnose RA early and obtain knowledge on the effectiveness of specific drugs early in the disease.

The large majority of treatment trials published thus far included patients who fulfil the 1987 American College of Rheumatology (ACR) criteria for RA. As the 1987 ACR criteria perform well in advanced RA but not in early RA,1 the efficacy and safety of specific treatments early in the disease havee not been extensively studied. The ability to classify RA in earlier phases of the disease is therefore very valuable. This need has led to the development of the 2010 ACR/European League Against Rheumatism (EULAR) criteria and the taskforce that derived these criteria stated explicitly that the specific charge was to facilitate the study of persons at earlier stages of the disease and that the developed 2010 criteria may be less suitable for basic studies because of the risk of increased heterogeneity.2 ,3

In clinical practice the temptation to use classification criteria for diagnosis is irresistible, thereby criteria shape the clinical picture of a disease. This phenomenon is also seen in other clinical situations. Fulfilling the updated Jones criteria is frequently used to diagnose acute rheumatic fever, having four American Rheumatology Association (ARA) criteria is often used to diagnose systemic lupus erythematosus, and the New York criteria have been used to diagnose ankylosing spondylitis.4,,6 The absence of diagnostic criteria mean that classification criteria are sometimes used to this end. The taskforce that derived the 2010 criteria referred to this by stating that the new criteria will probably also be used as a diagnostic aid and that in order to assign the probability of developing RA the range of scores between 0 and 10 can be used instead of the cut-off of 6 points.2 ,3

At present, several cohort studies have evaluated the test characteristics and performance of the 2010 ACR/EULAR classification criteria in their datasets. Therefore, an important question now is how well the 2010 criteria perform for the purpose of classification and diagnosis.

Classification versus diagnosis

Although classification and diagnostic criteria may consist of similar clinical, laboratory or imaging parameters, they are intended for different aims and as such have to meet different requirements (table 1). Classification criteria serve to enable communication and scientific research. Therefore, they serve as a tool to arrive at homogeneous groups of patients with comparable features to make data obtained by different researchers at different places comparable. Classification criteria preferably have a high specificity, so that patients without the disease are not included. Diagnostic criteria are not intended to form groups of patients; diagnoses are made to help the individual patient and serve as a basis for treatment decisions. In this situation high positive predictive values and high likelihood ratios are preferable (table 1).

Table 1

Differences between classification and diagnostic criteria

Another difference is the prevalence of the disease in the population that is being evaluated; this is of little interest when using classification criteria, because these are generally applied to patients in whom the diagnosis has been made in order to ascertain the validity of their classification. However, when making a diagnosis the value of a diagnostic test is highly dependent on the pretest probability, and therefore on the prevalence of the disease in the population. It has been shown that classification criteria perform poorly to diagnose individual patients in populations with a low pretest probability of the disease.7 ,8 Although the sensitivity remains the same in such a population, the positive predictive value decreases because of an increase of false positives. In this situation, the use of classification criteria for diagnosis runs the risk of misdiagnosis.

The 2010 criteria for classification

The 2010 criteria aimed to be fulfilled earlier in the disease course. Three studies evaluated patients with early arthritis that were undifferentiated according to the 1987 ACR criteria at first presentation but fulfilled the 1987 criteria during the first 12 or 18 months. These retrospective evaluations revealed that the majority of these patients did fulfil the 2010 criteria at first presentation.9,,11 This is highly relevant as it indicates that the taskforce that derived the new criteria succeeded in their main objective.

The test characteristics of the 2010 criteria as observed in the different validation studies are summarised in table 2. The majority of these studies compared these test characteristics with those of the 1987 criteria. A difficult issue is the outcome measure, as no gold standard for RA exists. Consequently, different outcome measures were used in these studies. The population of patients studied was also variable. The majority of studies excluded patients with clear other diagnoses and only evaluated patients with arthritis that was suspected to be undifferentiated or RA. One study also included patients without clinical synovitis.12 In addition, the effect of treatment on the outcomes of arthritis persistency or fulfilling the 1987 criteria was mostly disregarded.9 ,11 ,12 Such differences may affect the results. In addition, the number of patients included differed per study and measures of precision such as standard errors or 95% CI were not provided, prohibiting a formal meta-analysis. In an attempt to summarise the data, we weighted the test characteristics in relation to the number of patients included in the particular study. With this simplified method, we estimated the overall sensitivity at 74% and the specificity at 71%. These numbers should be taken with a ‘grain of salt’, because the standard error was ignored, as well as heterogeneity in the data due to different study designs and outcome measures.

Table 2

Results of studies testing the performance of the 2010 ACR/EULAR criteria for RA

Almost all studies consistently observed that the sensitivity of the 2010 criteria is increased compared with that of the 1987 criteria (table 2) and the specificity is decreased. This latter finding suggests that the group of patients classified as having RA by the 2010 criteria is less homogenous than that classified by the 1987 criteria. These data are in line with studies reporting that up to 18% of the 2010 criteria-positive patients have diagnoses other than RA later on.9 ,13,,16 Using the correct entry requirements (excluding other diagnoses before using the criteria for RA) reduced the problem of false-positive classification. A problem here is that there is no agreement as to what extent other diagnoses should be excluded before RA can be classified. Furthermore, also when patients with other diagnoses were excluded, a proportion of 2010 criteria-positive patients received other diagnoses later on.9 In conclusion, an increased sensitivity and earlier identification of RA is obtained to some extent at the cost of a precise classification. This is imaginable as early in the disease phenotypic features specific for certain diagnoses may still be absent. This is a point of concern as classification serves to discern homogeneous groups of patients in order to facilitate communication among researchers.

From a historical perspective these results are of interest. The 1958 ARA classification criteria for RA included a list of 20 features related to other diseases that should be absent before the criteria can be fulfilled.17 The criteria were revised because the spectrum of diseases with different underlying pathogenetic mechanisms indentified by the 1958 ARA criteria was considered too broad and it was intended to derive criteria with a higher specificity.18 Indeed, the 1987 ACR criteria had an increased specificity. Intriguingly, the present data indicate that the 2010 ACR/EULAR criteria have a reduced specificity and homogeneity.

The 2010 criteria for diagnosis

Clinical diagnosis is a highly individualised process, which makes it inherently difficult to obtain adequate diagnostic criteria. Diagnostic criteria preferably have high positive predictive values, indicating that patients fulfilling the criteria have a high chance of having RA. The data presented in table 2 show reasonable positive predictive values, in the range of 80%, indicating that RA is likely to be present in patients fulfilling the 2010 criteria. It is of utmost importance to realise that the positive predictive values presented here were derived from early arthritis cohort studies, in which the previous chance of RA is relatively high. In settings in which the previous risk of RA is lower, for example, at general practitioner practices, the positive predictive value will be lower. Consequently, the use of the 2010 criteria for diagnosing RA here may more often lead to misdiagnoses.

The observed negative predictive values were variable between studies but overall were not high (table 2). This indicates that RA is not ruled out in patients who do not fulfil the 2010 criteria. This finding is underlined by the observations that some of the patients who were negative for the 2010 criteria at disease onset, were already positive for the 1987 criteria.9 ,16 ,19 More importantly, almost a quarter of the 2010 criteria-negative patients fulfilled the 1987 criteria for RA later on.10 ,20 These findings imply that 2010 criteria-negative patients with unclassified arthritis should be followed carefully.

A final question is whether the clinical picture of RA has changed when using the 2010 criteria. Data are consistent in showing that 2010 RA patients have on average a lower number of tender and swollen joints than 1987 RA patients.9 ,11 ,16 ,21 The prevalence of anticitrullinated peptide antibodies in 2010 RA patients and 1987 RA patients were not comparable, but data are inconsistent here.10 ,11 ,16 Reasoning from the concept that the problem of RA is related to the chronicity of inflammation and to structural damage, it is important to compare the outcome of 1987 RA and 2010 RA. Currently available data indicate that 2010 RA is less erosive in nature9 ,21 and is more often self-limiting over time.10 ,13 Based on these results, we believe that if 2010 ACR/EULAR criteria positivity presents the new paradigm for the syndrome ‘RA’, the clinical picture of the disease has changed.

References

Footnotes

  • Funding The work of AHM is supported by the Dutch Organisation for Health Research and Development. This work is also supported by the European Community Seventh Framework Program FP7 Health-F2-2008-223404 (Masterswitch) and by the IMI EU funded project BeTheCure, contract no 115142-2.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.