Only studies, in which the absolute numbers of true positives, false positives, false negatives, and true negatives were available or could be calculated, were included in the analysis. The pooled specificity and sensitivity and their associated 95% confidence intervals were calculated using a random-effects (DerSimonian and Laird) model. Statistical heterogeneity was assessed using the Cochran Q and the I2 tests. Substantial statistical heterogeneity was defined as an I2 value of over 50%.
A summary receiver operating characteristic (sROC) curve was produced. Meta-regression was used to explore the possible sources of heterogeneity, such as the magnetic resonance unit, scanning technique, repetition time, echo time, maximum b factor, number of b factors used for ADC calculation, mean size of tumour (mm), and mean age of patients. The correlation between sensitivity and specificity was assessed using the Spearman correlation coefficient.
Publication bias was assessed using a funnel plot.