Test accuracy of drug and antibody assays for predicting response to antitumour necrosis factor treatment in Crohn’s disease: a systematic review and meta-analysis

OBJECTIVE: To present meta-analytic test accuracy estimates of levels of antitumour necrosis factor (anti-TNF) and antibodies to anti-TNF to predict loss of response or lack of regaining response in patients with anti-TNF managed Crohn’s disease. METHODS: MEDLINE, Embase, the Cochrane Library and Science Citation Index were searched from inception to October/November 2014 to identify studies which reported 2×2 table data of the association between levels of anti-TNF or its antibodies and clinical status. Hierarchical/bivariate meta-analysis was undertaken with the user-written ‘metandi’ package of Harbord and Whiting using Stata V.11 software, for infliximab, adalimumab, anti-infliximab and anti-adalimumab levels as predictors of loss of response. Prevalence of Crohn’s disease in included studies was meta-analysed using a random effects model in MetaAnalyst software to calculate positive and negative predictive values. The search was updated in January 2017. RESULTS: 31 studies were included in the review. Studies were heterogeneous with respect to the type of test used, criteria for establishing response and loss of response, population examined and results. Meta-analytic summary point estimates for sensitivity and specificity were 65.7% and 80.6% for infliximab trough levels and 56% and 79% for antibodies to infliximab, respectively. Pooled results for adalimumab trough levels and antibodies to adalimumab were similar. Pooled positive and negative predictive values ranged between 70% and 80% implying that between 20% and 30% of both positive and negative test results may be incorrect in predicting loss of response. CONCLUSION: The available evidence suggests that these tests have modest predictive accuracy for clinical status; direct test accuracy comparisons in the same population are needed. More clinical trial evidence from test–treat studies is required before the clinical utility of the tests can be reliably evaluated.