Background Syndromic management of vaginal discharge is common in both developed and developing countries. No systematic review has evaluated the performance of the WHO syndromic management algorithms as a case-management tool for symptomatic women. Our evaluation will inform health care professionals whether these algorithms have sufficient discriminatory power in identifying patients who would benefit from treatment for cervical infections.

Methods We conducted a systematic review and meta-analysis of diagnostic studies on syndromic management of vaginal discharge. We searched Medline, Embase and Global Health databases; we also searched relevant lists of references. We included the studies that reported their findings in sufficient detail to allow data abstraction for symptomatic women. Sensitivity, specificity and diagnostic OR (DOR) were used to evaluate the algorithms. ROC curves were pooled using the HSROC random effects model.

Findings We screened 896 abstracts for eligibility and included 101 articles for a full text review; 17 studies were included in the analysis. HSROC overall summary statistics were: sensitivity 0.58 (0.42–0.73), specificity 0.70 (0.61–0.78), DOR 3.25 (2.22–4.76) and sensitivity 0.81 (0.71–0.87), specificity 0.46 (0.40–0.52), DOR 3.71 (2.42–5.67) for the WHO algorithms without and with speculum examination respectively. For the algorithms that were modified to include local risk factors the overall sensitivity was 0.75 (0.51–0.90), specificity 0.58 (0.35–0.77) and DOR 4.18 (2.84–6.15) for the algorithms without speculum exam; sensitivity 0.70 (0.41–0.89), specificity 0.64 (0.42–0.82), DOR 4.24 (2.55–7.54) for the algorithms with speculum exam; and sensitivity 0.91 (0.68–0.98), specificity 0.54 (0.31–0.75), DOR 11.97 (4.93–30.0) for the algorithms with microscopy.

Interpretation Overall, diagnostic algorithms for the management of vaginal discharge had poor to moderate performance in symptomatic women. Incorporating locally-identified risk factors resulted in small improvement in the performance of the algorithms, however, considerable number of women were overtreated due to low specificity.