Research Methods Measurement and Statistics

Error in histologic dating of secretory endometrium: variance component analysis

Gibson M, Badger GJ, Byrn F, Lee KR, Korson R, Trainer TD

Published August 1991 Fertility and Sterility, 56(2), 242-247
DOI 10.1016/s0015-0282(16)54479-3 PMID 2070853

RRM Academy Synopsis

Evaluator disagreement drives most error in endometrial dating

Disagreement between evaluators accounted for 65% of the variability in biopsy dating. Five evaluators dated paired biopsies from 25 women in infertility testing. The authors say the dating error could lead to many false-positive luteal phase defect diagnoses.

Key Findings

  • Interevaluator inconsistency accounted for 65% of the variability, intraevaluator inconsistency for 27%, and regional differences within the uterus for 8%.
  • Repeat readings of a slide by the same evaluator agreed exactly in 43.1% of instances and fell within one day in an additional 39.3%.
  • On first readings, two evaluators gave identical dates in 25% of comparisons, fell within 1 day in 62% and within 2 days in 78%. Another 22% differed by more than 2 days.
  • The authors calculated a 1.4% chance of a false luteal phase defect diagnosis in a patient whose true histology was exactly in phase.

Interpretation

This study measured how precisely readers date endometrial biopsies. Three pathologists and two IVF doctors (reproductive endocrinologists) from one institution each read 50 slides from 25 women twice, and the analysis used the 21 complete cases. The false diagnosis figure assumes errors follow a normal pattern around the true date. The authors write that the data bear only indirectly on accuracy, since no external gold standard exists. They suggest more readers per slide would improve precision.

RRM Context

Restorative reproductive medicine puts evaluation before treatment, so a test must be reliable before decisions rest on it. The paper measures the error in the endometrial biopsy, a long-used test of luteal function. Reader disagreement is the largest source.

Abstract

Objective

To characterize the extent and sources of imprecision in histologic dating of the endometrial biopsy.

Design

Duplicate endometrial biopsies from 25 women were dated by five evaluators on two separate occasions to evaluate the overall precision of the measure. Using variance component analysis, estimates of intrauterine, intraevaluator, and interevaluator variability were determined.

Setting

Samples were obtained during outpatient fertility testing. Evaluators were colleagues at the same institution.

PATIENTS,

Participants

Women presenting with infertility undergoing routine evaluation.

Interventions

None.

Main Outcome Measures

Variability in histologic dating of the endometrium.

Results

Inconsistencies between evaluators accounted for 65% of the observed variability, whereas 27% was because of inconsistencies in duplicate readings by the same evaluator. Regional differences in the uterus accounted for only 8% of the total variability.

Conclusions

The overall error from these sources have the potential to result in a substantial false-positive rate for diagnosis of luteal phase defect.

Topics

By this author

Related research

Research Methods › Measurement and Statistics › Research Tooling and Reproducibility · Diagnostics › Endometrial Assessment › Luteal Phase Biopsy · Menstrual Cycle › Cycle Biomarkers › Hormonal Markers
Mark Gibson, Gary J. Badger, Francis Byrn, Kenneth R. Lee, Roy Korson, Thomas D. Trainer
M Gibson, G Badger, F Byrn, Ken Lee, K Lee, R Korson, Tom Trainer, T Trainer
PMID 2070853 2070853 DOI 10.1016/s0015-0282(16)54479-3 10.1016/s0015-0282(16)54479-3 Gibson et al. 1991, Gibson 1991