A CRITICAL-APPRAISAL OF TOXICITY INDEXES IN RHEUMATOLOGY

Our objective was to design a grading system to judge the methodologic al quality of toxicity symptoms assessment in rheumatologic randomized controlled trials, and to use this grading system to evaluate rheumat ology case report forms. For comparison, we also evaluated instruments derived in other clinical specialties, To further determine whether v ariability seen on the rheumatology case report forms would lead to va riability of side effect measurement in a clinical trial setting, we c onducted a survey of investigators and their coordinators in a multice nter randomized trial. We used Feinstein's definition of sensibility a nd a conceptual model of the process of toxicity measurement in the cl inical trial setting to develop a grading system. Thirty case report f orms were obtained from a convenience sample, and 7 other instruments from other disciplines were obtained through literature review, These forms were evaluated in a blinded manner by 3 reviewers. A survey of 3 0 trial staff from 15 centers involved in a multicenter, randomized tr ial of 2 active nonsteroidal antiinflammatory drugs (NSAID) in rheumat oid arthritis (RA) was undertaken to evaluate whether the variability identified in the review of the case report forms translated into vari ability in the measurement of side effects in clinical trial setting. The rheumatology case report forms had a mean score of 9.1 (SD = 2.6) out of 20, where 0 is the worst score and 20 is a perfect score. The s tructured instruments from other disciplines had a mean score of 14.9 (SD = 2.7). These means were significantly differ ent, with p = 0.001. The reviewers showed good levels of agreement, with kappa values of 0 .93, 0.89, and 0.80 for intrarater agreement, and 0.70 (95% CI 0.58 to 0.82) for interrater agreement. The survey demonstrated variability i n the manner of enquiry and recording of side effects in the case repo rt form. We conclude the current case report forms in rheumatology fai led to meet half the defined criteria expected of a valid instrument, while structured forms from other disciplines performed much better. T hese identified weaknesses lead to variability in side effect measurem ent in the usual clinic setting, as demonstrated by a survey in a mult icenter randomized trial.