Agreement Among Multiple Observers on Crystal Identification by Synovial Fluid Microscopy.

Abstract

Despite the fact that polarized microscopy remains the gold standard for diagnosing crystal arthritis, some uncertainties hamper full implementation in clinical practice. We undertook this study to analyze the agreement among multiple observers in crystal identification using compensated polarized microscopy, as well as to assess potential outcome modifiers. This was a cross-sectional, observational study with consecutive synovial fluid sampling. Samples were immediately analyzed when possible or kept refrigerated at 4°C. Five observers analyzed them separately, blinded to clinical data, using a compensated polarized optical microscope (400×), through 3 stages (ordinary, simple polarized, and compensated polarized light) to detect and identify crystals. They recorded the presence and type of crystal (no crystals, monosodium urate [MSU], and calcium pyrophosphate dihydrate [CPP]). Interrater agreement was measured by Cohen's kappa. Subanalyses were performed on visualization delay and cumulative expertise. Discrepancies between each stage of the microscopy and the final decision were also examined. A total of 250 observations from 50 samples completed full assessment. Overall, κ=0.74 (95% confidence interval [95% CI] 0.64-0.84), indicating good agreement. Agreement for crystal detection was κ=0.71 (95% CI 0.59-0.82), for MSU was κ=0.88 (95% CI 0.75-1.0), and for CPP was κ=0.69 (95% CI 0.56-0.82). Most of the crystal identifications were already made by ordinary light. No differences between observations made before or after 24 hours (P=0.859) or in expertise on crystal analysis (P=0.989) were found. Observations performed under ordinary light matched the final diagnosis in 96.8% of cases (242 of 250). Compensated polarized microscopy remained consistent in detecting and identifying crystals in synovial fluid, even when examined among multiple observers, confirming its high utility for clinical practice.