Diagnostic strategies for the histological examination of muscle biopsy specimens for the assessment of vasculitis in rheumatoid arthritis.

For the diagnosis of rheumatoid vasculitis (RV), histological examination of a blindly-taken muscle biopsy is advocated. If fibrinoid necrosis (FN) is observed in one or more tissue sections of the biopsy the diagnosis of RV is confirmed. The diagnostic value of such histological investigation depends on the prevalence of FN in biopsies of RV patients, the number of tissue sections that are investigated, and the sampling design with which the tissue sections are obtained from the biopsy. In this paper we determine the mathematical relation between the sensitivity of the RV diagnosis and these three factors for four different models for the FN distribution in RV biopsies. The goodness-of-fit of these models was assessed by analysing 18 829 tissue sections of the biopsies of 56 patients, among which were 24 (otherwise histologically proven) RV patients. It appeared that the sensitivity was moderate: FN was observed in only 14 out of 24 (58 per cent) muscle biopsies of the RV patients. The prevalence of FN in the tissue sections of those 14 biopsies was low, about 17 per cent according to the best fitting model. We proved that the sampling design, maximizing the minimum distance between tissue sections (equidistant sampling) was optimal, and that with such optimal sampling, examination of about 20 tissue sections was sufficient. With practical sampling designs, however, considerably more tissue sections had to be inspected. FN appeared to cluster in the RV biopsies, and a first-order Markov model satisfactorily described the (auto)correlation between adjacent tissue sections in RV biopsies.