MAVARICA randomised controlled trial that compared manual and automated (machine-read) screening of liquid-based cytology samples from the uterine cervix.
The primary outcome measure (OM) was the sensitivity of automation-assisted reading relative to manual for the detection of CIN2+. A secondary OM was cost-effectiveness of each type of reading to detect CIN2+.
Automated reading was 8% less sensitive than manual reading. “No further review” by automated reading was very reliable; if restricted to routine screening samples, <1% of CIN2+ would have been missed.
The much lower sensitivity of automated reading, combined with uncertainty over its cost-effectiveness, suggests that there is no justification to recommend its use. The reliability of the “no further review” diagnosis may be of value as a way of saving staff time.