A meta-analysis over multiple thresholds to compare two prediction rules for acute pharyngitis

Problem

Centor’s score and McIsaac’s score are two rules commonly used to diagnose Group A beta-haemolytic streptococcus (GABHS) infection. However, McIsaac’s score applied to primary care has not been meta-analysed; the scores have not been compared by meta-analysis; and a model which may handle all the data reported at multiple cut points has not been used.The aim was to compare the performance of Centor’s score and McIsaac’s score in diagnosing patients with GABHS presenting to primary care with pharyngitis

Approach

Medline, Embase, PsychoInfo, Google and the references from two published reviews were searched for relevant citations. Searches were restricted between 1980 and February 2019. Studies were included if they were: a cross-sectional study; patients with a sore throat were recruited from a primary care population; the study evaluated Centor’s or McIsaac’s score; GABHS infection was the target diagnosis; the reference standard was a culture throat swab; complete 2 x 2 tables across multiple thresholds were reported. Selection was conducted independently by two reviewers.Data were extracted according to PRISMA-DTA guidance by two independent reviewers. The risk of bias and applicability of results were assessed using the QUADAS-2 tool. A novel meta-analysis random effects model which incorporates data at multiple thresholds was used to synthesise estimates.The main outcomes were a summary receiver operating characteristic (SROC) curve, the area under the curve statistic and a calibration plot corrected for optimism.

Findings

10 studies on Centor’s criteria and 8 on McIsaac’s score were included. In both sets of studies the prevalence of GABHS ranged between 4 and 44%. Two SROC curves were synthesised and the area under curve, C statistics for McIsaac’s score and Centor’s score were 0.7052 and 0.6888 respectively. The p-value for the difference (0.0164) was 0.353, suggesting the SROC curves for the tests are equivalent. Both scores demonstrated poor calibration. Both rules provide only fair discrimination of those with and without GABHS infection. McIsaac’s score and Centor’s score appear broadly equivalent in performance. The poor calibration for a positive test result suggests other point of care tests are required to ‘rule in’ GABHS. However, with either rule a score of zero may be sufficient to ‘rule-out’ infection.

Consequences

This comparative meta-analysis demonstrates that Centor’s score and McIsaac’s score have broadly similar performance characteristics in diagnosing GABHS in primary care. Neither score is sufficiently accurate to rule in GABHS infection and if applied as recommended could lead to over 1 in 2 patients being prescribed antibiotics inappropriately. Other point of care diagnostics that augment these scores are needed if rates of inappropriate antibiotic prescriptions are to be reduced. However, a score of zero (or less) in either rule potentially has a role in ruling out GABHS infection in primary care.