Impact of bias in crown–rump length measurement at first‐trimester screening for trisomy 21

Abstract

To assess the repeatability of crown-rump length (CRL) measurement and examine the effect of its over- and underestimation on first-trimester combined screening. Intra- and interoperator repeatability of CRL measurement at 11-13 weeks of gestation was assessed in 124 cases by two operators. Raw data were transformed into gestational age and intra- and interoperator repeatability was evaluated by within-operator standard deviation (SD) and the SD of differences in measurements between both operators. Modeling techniques were used to assess the impact of CRL measurement error on general population screening and on the operator-specific screening performance. The impact of errors in CRL measurement were investigated by simulating fetal nuchal translucency (NT) measurements and multiple of the median (MoM) values for pregnancy-associated plasma protein A (PAPP-A) and free β-human chorionic gonadotropin (β-hCG) for 500 000 euploid and 500 000 trisomy 21 pregnancies at 12 weeks and 9 weeks of gestation, and adding to or subtracting from each CRL value up to 10 mm and recalculating patient-specific risks. Within-operator SD of the CRL measurement was 1.27 days of gestation. The SD of the differences in CRL measurement between operators was 1.37 days of gestation. Both intra- and interoperator 95% limits of agreement were around ± 5 mm. In general population-based screening, a CRL measurement error SD of 5 mm accounts for an estimated 5% of the SD of log MoM PAPP-A and less than 1% of the SD of log MoM free β-hCG. Modeling the effect of removing this measurement error on overall screening performance showed a minimal impact. For a risk cut-off of 1 in 100, the benefit in terms of overall screening performance would be an increase in detection rate of about 1% and a reduction in false-positive rate of less than 0.1%. With regard to the operator-specific screening performance, a consistent 5-mm underestimation of CRL reduces the detection rate from 84% to 79% and the false-positive rate from 2.4% to 1.2%. With a consistent 5-mm overestimation the rates would be 88% and 5.6%, respectively. The impact of the interoperator variability in CRL measurement on patient-specific risk needs to be taken into account when interpreting first-trimester screening results. A systematic under- or overestimation of CRL should be avoided.