Effect of study design on the association between nuchal translucency measurement and Down syndrome.

OBJECTIVE To evaluate the effect of verification bias on the accuracy of first-trimester nuchal translucency measurement for Down syndrome detection. METHODS We used MEDLINE and EMBASE to identify all papers relating the results of nuchal translucency measurement to fetal karyotype. The detected studies were scored for verification bias. Fifteen studies without and ten with verification bias were included. RESULTS Sensitivity and specificity were calculated for each study. For studies with verification bias, adjusted estimates of the sensitivity were calculated assuming a fetal loss rate for Down syndrome pregnancies of 48%. The sample size weighted sensitivity was 55% in studies without and 77% in those with verification bias, for specificities of 96% and 97%, respectively. After adjustment for verification bias, the sample size weighted sensitivity changed from 77% to 63%. CONCLUSION Studies with verification bias reported higher sensitivities, but also slightly higher specificities of nuchal translucency measurement than studies without verification bias. The difference in sensitivity is greater than could be explained by verification bias. We postulate that the experience of the sonographist might be an explanation for the differences.