The new preliminary ACR classification criteria for SS

A new approach to classification criteria for Sjögren's syndrome (SS) has been recently proposed.1 This new set of criteria was the result of an expert consensus based on the analysis of data collected from a large number of patients enrolled during a multicentre study, that is, the Sjögren's International Collaborative Clinical Alliance (SICCA). On the basis of these new criteria, a patient can be classified as having SS, when she (or he) meets at least two of the following three criteria: (i) a positive serum anti-Ro and/or anti-La antibodies, or positive rheumatoid factor and antinuclear antibody (titre >1 : 320); (ii) presence of keratoconjunctivitis sicca (KCS) defined by an ocular staining score >3; and (iii) presence of focal lymphocytic sialadenitis defined by a focus score >1 focus/4 mm² in labial salivary gland biopsy samples.

As the authors stated, this criteria set shows a very high specificity in selecting true patients with SS, a quality that they believe to be really important when classification criteria are used to select patients to be enrolled in clinical trials where new therapies, namely, biological agents with potentially serious side effects, are going to be tested. In addition, the authors affirm that their criteria specificity is strongly reinforced by the fact that symptoms of dry mouth and dry eye are excluded from the criteria set, since subjective complaints lack the specificity associated with objective tests.1

The method employed to derive these new criteria is unusual for two reasons: (i) no preliminary definition of the disease was provided, as usually happens in studies aimed at developing classification criteria for rheumatic diseases and (ii) very few examples exist of classification criteria for rheumatic diseases derived by applying this methodology.2 After an expert consensus selection and the restriction of candidate tests for entry into the criteria set, a specific type of cluster analysis, that is, latent class analysis, was applied to define a number of disease classes.3 Ten variables were selected by expert consensus as the best predictors for SS. A model formed by two classes (which separates disease cases and control cases) was chosen on the basis of the statistical analysis. At the end of the analysis three variables were selected and the combined presence of two out of three of these variables was shown to have the best performance in terms of classification accuracy.1

Critical points of the new classification criteria

Some points can be raised on the assumptions made, and on the results obtained by the whole procedure. Two disease classes seem to be too few in view of the clinical spectrum of SS and sicca complaints, which include at least three disease classes, that is, patients with primary SS, patients with SS associated with other systemic autoimmune disorders (SADs) and patients with sicca complaints alone which may simulate SS. The SICCA clinical cohort included only a few patients with SS associated with other SADs, and these patients were excluded by the analysis. Nevertheless, the preliminary criteria were proposed as a valid tool even to classify these kinds of patients who are well known to present different patterns of presentation according to the associated SADs.4 Furthermore, the absolute exclusion of symptom items may induce a misclassification of patients presenting with ocular and oral sicca complaints, anti-Ro/La autoantibodies, and a clear reduction of salivary or lachrymal flow, who can certainly be classified as patients with SS. Moreover, these patients who have not yet developed a KCS (or have an ocular staining score less than 3), and where the lip biopsy is (falsely) negative, may have earlier disease and consequently a higher possibility to respond to effective treatment, including the new target therapies. This seems to demonstrate that, besides their low specificity, a precise assessment of symptoms, associated with that of more specific tests, often makes the classification of some patients easier.

Another important point is represented by the inclusion in the SICCA criteria, as an alternative serological item, of a positive antinuclear antibody and rheumatoid factor instead of anti-Ro/La antibodies. This statement was derived by a consensus decision of the SICCA experts, and it was adopted although both sensitivity and specificity of this alternate serological item were lower.

These items certainly require further assessment when a third disease class including patients with SADs and associated SS is considered in the statistical model and used as an additional control group.

The SICCA group have developed a new Ocular Staining Score;5 it would be useful to know if this can be substituted by and substitute for the American European Consensus Group (AECG) objective ocular component (reduced Schirmer-I-test or positive van Bijsterveld score) in either criteria set. This should be possible to evaluate using the existing SICCA data.

Comparison of the new ACR criteria with AECG classification criteria

The new criteria were tentatively validated by comparing their accuracy with that of the AECG classification criteria,6 which was considered to be the gold standard comparison. This was done since, as the authors underlined, the AECG criteria have better specificity1 with respect to all of the criteria previously proposed.7–11 On the other hand, this statement is also supported by the fact that the AECG criteria have been widely adopted by the scientific community since their publication, as demonstrated by the large number of citations they received (more than 1300 from the articles released from 2002 until now).6 ,12 The statistical comparison between the AECG criteria and the newly described criteria showed a strong agreement between the two criteria sets. This is particularly true when the version of the AECG criteria that includes all the items1 ,6 was taken into account. Measure of agreement (k) with SICCA criteria was 0.88 with a sensitivity of 92.1% (CI 89.8% to 94.0%) and a specificity of 95.5% (CI 93.8% to 96.9%).

The AECG classification criteria: criticisms and unsolved problems

The AECG classification criteria were the results of an international collaborative study that was begun in the 1990s and completed at the beginning of the new millennium. The multicentre study was developed and carried out in the first phase at a European level only,13 but a number of American experts (including some members of the SICCA group) took part in the final consensus definition and validation of the AECG criteria.

The statistical method used to derive the AECG criteria was the same as that adopted by the American College of Rheumatology (ACR) for the definition of the 1987 criteria for rheumatoid arthritis14 and the 1982 criteria for systemic lupus.15 Following this methodology, patients had to be preliminarily classified as having or not having a specific disease simply on the basis of the physician's clinical judgment. At a later time, statistical analysis made it possible to find the best combination of items able to more accurately distinguish between case patients, previously judged as having the disease, and case controls with similar and misleading clinical symptoms or diseases, but initially classified as not having the disease.

Criticisms of this statistical procedure can certainly be raised. This method may introduce some biases in selecting the criteria items, due to a circularity process, since the preliminary classification of patient cases is certainly influenced, in the physician's mind, by the positivity of specific tests which, at the end of the process, have a high probability of inclusion among the items selected by statistical tools.

However, since no specific gold standard is available to correctly classify a complex disease like SS, the process applied in the development of the AECG criteria can be simply defined as the way to statistically validate the common meaning of the experts on what is a sufficient requirement to define a patient as having SS. In the case of the AECG criteria the statistics and the expert consensus lead to defining a patient as having SS by the mandatory evidence of autoimmune activation (presence of specific autoantibodies or focal lymphocytic infiltration in target tissue), plus the presence of a number of other items defining the following domains: (i) dry eye symptomatology, (ii) dry mouth symptomatology, (iii) evidence of functional or structural derangement of salivary glands and (iv) lachrymal/conjunctive involvement defined by reduction of lachrymal flow and/or evidence of KCS.

The presence of six domains for a total of 13 items in the AECG criteria set may induce one to argue that these criteria are less useful in practice than the new ACR criteria that include only three domains. This is not completely true since the greater part of the items are constituted in the AECG criteria set by simple questionnaires and by objective tests, such as Schirmer's test and unstimulated whole saliva collection, which are very easy to do, even at the patient's bedside.

The fact that non-equivalent tests were considered to be equivalent in some domains and invasive and expensive tests were introduced in the criteria set is another important point that is still being debated. However, the wide application of the AECG criteria surely supports their validity and feasibility. Some of these criticisms, and the fact that some tests like sialography are presently rather obsolete, may undoubtedly induce one to think that the need for some revision is now emerging. Mainly, specifically designed studies aimed at evaluating sensitivity and specificity of the ultrasound examination of major salivary glands in SS are needed.

The future scenario of classification criteria for SS

Once the new criteria are definitely validated, two different criteria sets will be available to classify patients with SS. This fact may induce some confusion in the scientific community, since cohorts of patients selected for clinical studies by using different methods could not be equivalent in terms of disease duration and clinical expression, and the results obtained in the epidemiological studies and therapeutic trials will not be entirely comparable. This is obviously a critical point, mainly at the present time in which new biological agents that may potentially contrast some of the key mechanisms in the disease pathological cascade have become available to be tested in this (until now) orphan disorder. Consequently, to reach a final agreement on classification criteria for SS overcoming the differences among the various criteria proposed and validated seems to be an objective that we should all be striving for. This goal can be certainly obtained by a joint initiative that should involve experts on SS coming from different countries. A validation of provisional ACR classification criteria, using an external cohort of case patients and controls, a comparison between ACR and AECG criteria performance, and finally an expert panel discussion on the cases with discordant diagnoses could be the proper methodology for this purpose.