Reliability of sparing papanicolaou test conventional reading in cases reported as No Further Review at AutoPap-assisted cytological screening

Abstract

AutoPap-assisted smear reading has been proposed prior to conventional manual reading; the latter may be unnecessary for cases reported as No Further Review (NFR) and would be required for cases reported as Review (REV). The authors evaluated comparable concurrent screening cohorts who were undergoing a conventional manual (CONV) or an AutoPap-assisted smear reading within the same screening program. The authors evaluated the prevalence of CIN2+ at repeat screening in subjects 1) with a negative report at conventional Papanicolaou test (CONV- = 9605), 2) with a REV report at AutoPap, followed by a negative conventional reading (REV- = 17,576), and 3) with a NFR report at AutoPap, followed by a negative rapid review (NFR- = 3477) at previous (baseline) screening. Crude CIN2+ detection rate was 0.176% (17 of 9605), 0.187% (33 of 17,576), or 0.02% (1 of 3477) among baseline CONV-, REV-, or NFR- subjects, respectively. No significant difference in CIN2+ detection rates was evident when comparing the whole baseline CONV- and AutoPap- cohorts (0.176% versus 0.161%; chi(2) ((df=1)) = 2.48, P = .87), or baseline CONV- with baseline REV- subjects (chi(2) ((df=1)) = 2.49, P = .96), whereas a borderline difference was evident when comparing baseline NFR- with CONV- (chi(2) ((df=1)) = 3.07, P = .079) or with REV- (chi(2) ((df=1)) = 3.61, P = .057) subjects. The authors' findings suggested that AutoPap assisted reading is comparable to conventional reading, as the frequency of CIN2+ detected at repeat screening did not differ between baseline CONV- or AutoPap- cohorts. The baseline NFR- subgroups showed a lower, borderline significant CIN2+ detection rate when compared to the CONV- cohort. The very low observed negative predictive value of an NFR report (0.02%) suggested that these subjects may safely return to periodic screening and that quality control measures such as rapid review or full manual reading of a random sample are probably not necessary.