Author:

Keywords:

HPV testing, cytological abnormality, cervix

Abstract:

1.Background. Management of equivocal and low-grade lesions, observed in the context of cervical cancer screening, remains until today controversial. Triage with HPV DNA testing and repeat Pap smears are two possible strategies to select women who need further follow-up. 2. Methods. Systematic literature research targeted studies published over the last ten years starting from 1992. Data were extracted from 30 selected references containing results of concomitant cytological and virological testing followed by colposcopically directed biopsy in cases with an index smear showing atypia or low-grade abnormality. Random effect models were used for pooling of accuracy parameters in case of significant inter-study heterogeneity. We distinguished studies by cytological result of the index case: a) ASCUS (atypical squamous cells of uncertified significance); b) LSIL (low-grade squamous intraepithelial lesion) and c) ASCUS/LSIL (index cases where the distinction between ASCUS and LSIL could not be separated. All retrieved studies are assessed individually by checking a standard list of study design issues: description of the study population with inclusion and exclusion criteria, the determination of the age distribution of included subjects; the particular design of the study; description of the index cytological test; description of the enrolment triage test (HPV DNA testing system; possibly concomitant repeat cytology); description of the golden standard; timing of index, enrolment and golden standard testing; the blinding of interpreters for other test results; presentation of the absolute figures and relative accuracy parameters and finally a short discussion. The accuracy parameters (sensitivity, specificity, predictive values, proportion of test positives and prevalence of disease are subsequently pooled using fixed or random effect models according to the level of inter-study heterogeneity. The variation in accuracy measures in the individual studies and the pooled measure are displayed graphically using forest plots. The contrast in accuracy between repeat cytology and HPV DNA testing is studied by the ratio of sensitivity and specificity of both tests applied in parallel on the same women and relative to the golden standard. The change in accuracy by combining two triage tests in comparison with one single test is assessed by the difference in accuracy. The influence of study characteristics on study outcomes was explored by metaregression. Existence of verification bias was assessed graphically using funnel plots and statistically using the rank correlation test ad the asymmetry regression test. Finally the trade-off between sensitivity and specificity is studied graphically by the way of summary ROC (Receiver Operating Characteristics) curve analysis.