Factors related to inter-observer reproducibility of conventional Pap smear cytology: a multilevel analysis of smear and laboratory characteristics.

Abstract

To identify factors that influence the inter-observer reproducibility of the routine, conventional Pap smear cytology (Pap smear test) in a network of certificated laboratories in a middle-income Latin American country. Twenty-six laboratories provided each an average of 26 negative for malignancy (NILM) and high-grade squamous intraepithelial lesion (HSIL) Pap smears. An external panel reviewed the slides. The kappa index and multilevel logistic regression were used to estimate the reproducibility and odds ratios (OR) of a false result with 95% confidence intervals (95% CI), respectively. Results are presented for laboratories that collect (collector laboratories) and do not collect (non-collector laboratories) samples. The agreements ranged widely (median kappa 0.51, range 0.16-0.70). The overall false-positive (FP) and false-negative (FN) rates were 31% (95% CI 27-35) and 11% (95% CI 7-17). Among collector laboratories (N = 14), a bigger sample collection volume decreased the probability of a FP (OR-adjusted 0.05, 95% CI 0.02-0.1) whereas the number of quality defects (OR-adjusted 1.67, 95% CI 1.25-2.24), high workload (OR-adjusted 5.52, 95% CI 3.85-7.92) and collection by cytotechnologists (OR-adjusted 1.28, 95% CI 1.15-1.42) or health professionals (OR-adjusted 2.26, 95% CI 2.04-2.49) instead of nursing assistants increased it. Among non-collector laboratories (N = 9), the FP rate increased with the number of quality defects (OR-adjusted 1.86, 95% CI 1.06-3.26) but decreased if the samples were collected by health professionals instead of nursing assistants (OR-adjusted 0.37, 95%CI 0.17-0.80). No significant associations were observed for FN. Staff in charge of cervical sampling significantly determined the reproducibility of the Pap smear test, but this depended on whether the laboratory collects samples or read samples collected elsewhere.