METHODS AND RESULTS

Criteria for inclusion were the presence of diffuse musculoskeletal pain; a scan for diagnostic uncertainty; and availability of whole body scans, including a close-up scan of the peripheral joints. Three hour images were studied. Earlier “blood pool” images were not available. Patients were excluded if scans had been carried out for a regional disorder, isolated spinal disease, Paget’s disease, suspected malignancy, known arthritis or connective tissue disease, and previous disease modifying antirheumatic drug use. One researcher examined the medical records and applied the inclusion and exclusion criteria. Ethics committee approval was granted.

Two consultant radiologists read all the scans independently without reference to clinical data or the original scan report, to ensure consistency. Scans were classified according to the scheme shown in table 1. Radiologists then met to agree a final classification, which was used against our reference standard. General practitioner records were reviewed if follow up data were inadequate in the hospital records (10 cases). Two researchers agreed the final clinical diagnosis by discussion and after independent review of the records.

Three hundred and ten ^99mTc MDP bone scans were ordered between May 1995 and August 1996, but only 135 films were available. One hundred and twenty isotope scans were found, 54 patients met inclusion criteria, but four were excluded because of incomplete records (table 1). Radiologists agreed on scan classification in 20/50 (40%) cases (κ for inflammatory arthritis versus other categories 0.31). However, after conferring, the radiologists readily agreed and discrepancies arose because of difficulties in classifying osteoarthritis versus non-specific isotope uptake.

The sensitivity of an isotope bone scan for a clinical diagnosis of ” inflammatory arthritis” was 13% (agreement in 2/15; 95% confidence interval (CI) 4% to 23%; χ² =0.57; p>0.1) and the specificity 80% (agreement in 28/35; 95% CI 69% to 91%). The likelihood ratio of a positive scan was 0.65 and 1.09 for a negative scan. The positive predictive value of a bone scan was 22% and the negative predictive value 68%. The latter suggests that scans may be useful in excluding inflammatory arthritis, a finding consistent with an earlier report.¹

DISCUSSION

Our study fell short of the ideal design for assessing the accuracy of a diagnostic test in that it was not a prospective blind comparison of a test and a reference standard in consecutive patients.² A practical difficulty in meeting this requirement was that clinical follow up could not be applied in parallel with the test. Also, we studied relatively few patients and many films were not found. Possibly, the final clinical diagnosis was influenced by the bone scan report. Such “test review bias” is difficult to remove in evaluations of routine diagnostic tests and efforts to do so may ignore clinical realities.^3,4 We hoped that bias was minimised by relying on prolonged follow up and using two assessors.

Isotope bone scans seem to have limited value in confirming or refuting a clinical diagnosis of inflammatory arthritis and, in common with other radiological examinations, are prone to substantial interobserver variation.⁵ Many studies of bone scans, and of other diagnostic tests, compare patients known to have disease with those free of disease, or with another disease.^1,6 Such studies are known to overestimate test diagnostic accuracy.³ Watchful waiting may be a more appropriate clinical strategy than relying on bone scans, with appropriate intervention when diagnostic confidence increases. However, we recognise that clinicians use diagnostic tests in sequence and that test results frequently have an incremental impact on diagnostic confidence.