When implemented appropriately, cervical cancer screening has the power to save lives. Its effectiveness, however, rests on the sensitivity and specificity of the diagnostics themselves—namely, pap cytology and HPV DNA testing. The introduction of p16/Ki-67 dual-stain cytology as a triage test for abnormal pap results offers the ability to improve detection of precancer in equivocal and abnormal cervical cells. This study seeks to evaluate the performance of dual-stain “diagnostic cytology” in Belgium (women age 25-65) through a systematic literature review; this includes a component of meta-analysis in order to draw conclusions about the increasingly complex, quantitative body of existing knowledge available.
A literature search was conducted of published studies from 1 May 2016 according to CRD, PRIMSA, and NICE guidelines. Studies needed to report diagnostic performance with key metrics: sensitivity, specificity, detection rate, odds ratio, PPV, NPV, true and false positives, and true and false negatives. These outcomes were chosen since they are the most significant and widely used outcomes to compare performance, and can be calculated from one another using established formulas. Sensitivity and specificity were chosen as outcome measures in the final meta-analysis based on guidance from the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy. The meta-analysis of dual-staining performance focused on sensitivity and specificity. Results of multiple studies collected during the literature review were combined into a single, standardized metric for comparison across different tests.
Unlike meta-analyses of treatment intervention effects, meta-analyses of diagnostic test accuracy must allow for the trade-off between sensitivity and specificity that occurs between studies whose threshold values for test positives and negatives vary. Hierarchical models (in contrast to many classical regression models) allow for correlation between sensitivity and specificity, in addition to the aforementioned trade-off. Therefore, the bivariate random effects model was used in this meta-analysis because of its ability to jointly evaluate sensitivity and specificity, thus providing an estimate of the diagnostic accuracy of dual-staining.
Dual-stain “diagnostic cytology” with p16/Ki-67 is an attractive biomarker for triage in cervical cancer screening. In a Belgian screening population (age 25-65 years), dual-stain cytology offered significant gains in sensitivity with minimal reduction in specificity. This could lower the number of cases of missed disease and reduce morbidity from additional interventions such as colposcopy and biopsy.