 |
|
|
 |
Dr. Taylor replies To the Editor: Classification criteria are very helpful in identifying a group of people with meaningful homogeneity with respect to a particular disease, syndrome, or health condition. The more closely the criteria concur with the true state of the patient and the more specific contexts in which this is demonstrated, the more useful the criteria. The CASPAR criteria for psoriatic arthritis (PsA) are the most robust and accurate classification criteria yet demonstrated for any rheumatic disease1. The criteria remain accurate in the context of early disease and, as shown by Dr. Chandran and colleagues, in undifferentiated patients presenting to a family medical clinic. This is possibly unprecedented in rheumatology. Yet some caveats remain. Chandran and colleagues propose that the excellent performance of the CASPAR criteria across different contexts constitutes evidence for "diagnostic criteria." This is a more challenging concept than "classification criteria." First, diagnosis is a process in which all available information is processed, interpreted, and synthesized. Data beyond those listed in classification criteria may be important in the final diagnostic analysis. For example, in some patients, magnetic resonance imaging findings may be important to aid diagnosis, yet these are not listed in current classification criteria. Second, classification criteria are supposed to be applied to groups of people, in which probabilistic terms such as sensitivity and specificity or positive and negative predictive value have meaning. In the case of an individual, who either has the disease or not, such probabilistic statements are of limited value. The usefulness of a probability statement when applied to an individual patient depends not only on the absolute value of that probability but also on the consequences of error2. Greater likelihood of misdiagnosing a person as having the disease when they do not may be acceptable when the diagnosis leads to relatively innocuous treatment, for example. It is rarely possible to be able to specify in advance the "costs" associated with making a correct or incorrect diagnosis. It may be the case that "diagnostic criteria" can only be meaningful when the criteria define the disorder, so that it is logically impossible for the disease to be present when the criteria are not met and logically impossible for the disease not to be present when the criteria are met. An example of this could be synovial uric acid crystals for the disease of gout. At this time, such a pathognomic feature for PsA is more elusive, and until it is found perhaps we should be more circumspect about claims for "diagnostic criteria." The other difficulty with the CASPAR criteria that may limit direct application to nonrheumatology settings, despite their accuracy, is the requirement of "inflammatory articular disease." The presence of this item was determined by a rheumatologist in the study reported by Chandran, et al. It is unclear whether nonrheumatologists can confidently or accurately judge the presence of inflammatory articular (joint, entheseal, spinal) symptoms and signs, although there is evidence that examinations of joints and for dactylitis are as reliable (or unreliable) among dermatologists as rheumatologists3. Nonetheless, further evidence for the broader application of CASPAR criteria might come from testing nonrheumatologist accuracy in determining whether the criteria are met or not. WILLIAM J. TAYLOR, PhD, MBChB, FAFRM, FRACP, Department of Medicine, University of Otago, Wellington, New Zealand. E-mail: will.taylor@otago.ac.nz 2. Guggenmoos-Holzmann I, van Houwelingen HC. The (in)validity of sensitivity and specificity. Stat Med 2000;19:1783-92. [MEDLINE] 3. Chandran V, Cook R, Helliwell P, et al. International multi-centre psoriasis and psoriatic arthritis reliability Trial (GRAPPA-IMPART): assessment of skin, joints, nails and dactylitis [abstract]. Arthritis Rheum 2007;56 Suppl:S798-9.
|
