Objectives: The principal objective was to compare automation-assisted reading of cervical cytology with manual reading using the histological end point of cervical intraepithelial neoplasia grade II (CIN2) or worse (CIN2+). Secondary objectives included (i) an assessment of the slide ranking facility of the Becton Dickinson (BD) FocalPoint™ Slide Profiler (Becton Dickinson, Franklin Lakes, NJ, USA), especially 'No Further Review', (ii) a comparison of the two approved automated systems, the ThinPrep® Imaging System (Hologic, Bedford, MA, USA) and the BD FocalPoint Guided Screener Imaging System, and (iii) automated versus manual in terms of productivity and cost-effectiveness.
Design: A 1 : 2 randomised allocation of slides to either manual reading or automation-assisted paired with manual reading. Cytoscreeners were blinded to whether samples would be read only manually or manually paired with automated. Slide reading procedures followed real-life laboratory protocol to produce a final result and, for paired readings, the worse result determined the management. Costs per event were estimated and combined with productivity to produce a cost per slide, per woman and per CIN2+ and cervical intraepithelial neoplasia grade III (CIN3) or worse (CIN3+) lesion detected. Cost-effectiveness was estimated using cost per CIN2+ detected. Lifetime cost-effectiveness in terms of life-years and quality-adjusted life-years was estimated using a mathematical model.
Setting: Liquid-based cytology samples were obtained in primary care, and a small number of abnormal samples were obtained from local colposcopy clinics, from different women, in order to enrich the proportion of abnormals. All of the samples were read in a single large service laboratory. Liquid residues used for human papillomavirus (HPV) triage were tested (with Hybrid Capture 2, Qiagen, Crawley, UK) in a specialist virology laboratory in Edinburgh, UK. Histopathology was read by a specialist gynaecological pathology team blinded to HPV results and type of reading.
Participants: Samples were obtained from women aged 25-64 years undergoing primary cervical screening in Greater Manchester, UK, with small proportions from women outside this age range and from women undergoing colposcopy.
Interventions: The principal intervention was automation-assisted reading of cervical cytology slides which was paired with a manual reading of the same slide. Low-grade cytological abnormalities (borderline and mild dyskaryosis) were triaged with HPV testing to direct colposcopy referral. Women with high-grade cytology were referred for colposcopy and those with negative cytology were returned to recall.
Main outcome measures: The principal outcome measure was the sensitivity of automation-assisted reading relative to manual for the detection of CIN2+. A secondary outcome measure was cost-effectiveness of each type of reading to detect CIN2+. The study was powered to detect a relative sensitivity difference equivalent to an absolute difference of 5%.
Results: The principal finding was that automated reading was 8% less sensitive relative to manual, 6.3% in absolute terms. 'No further review' was very reliable and, if restricted to routine screening samples, < 1% of CIN2+ would have been missed. Automated and manual were very similar in terms of cost-effectiveness despite a 60%-80% increase in productivity for automation-assisted reading.
Conclusions: The significantly reduced sensitivity of automated reading, combined with uncertainty over cost-effectiveness, suggests no justification at present to recommend its introduction. The reliability of 'no further review' warrants further consideration as a means of saving staff time.
Trial registration: Current Controlled Trials ISRCTN66377374.
Funding: This project was funded by the NIHR Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 15, No. 3. See the HTA programme website for further project information.