Introduction: In the field of diagnostic and screening tests, trials are only part of relevant research. Other diagnostic studies are estimating the diagnostic value of signs, symptoms and tests by comparison to a gold standard.

Objective: We intended to test methods for analysing reviews of diagnostic studies.

Methods: Different methods and software were applied in six meta-analyses of different types of diagnostic studies: trial (influence of sampling methods on the diagnostic yield in cervical screening), only correlation coefficients available when test as well as gold standard results are continuous (diagnostic value of renal function tests), continuous test results with a dichotomous gold standard (diagnostic value of the CAGE questionnaire in diagnosing alcohol problems) and both test and gold standard result are dichotomous (diagnostic value of macroscopic haematuria when diagnosing urological cancer, of extended coughing when diagnosing lung cancer and of abnormal digital rectal examination when diagnosing prostate cancer).

Results/Discussion Test as well as gold standard results can have dichotomous as well as continuous values. Outcome measures used are sensitivity, specificity, predictive values and likelihood ratios, but also ROC curves, correlation coefficients and regression coefficients. Statistical pooling can be based on random as well as fixed effect models. The same principles are used as when pooling trial-data, but formulas and software have to be adapted to the specific outcome measures. Gradually, adapted software is becoming available.