Article type
Year
Abstract
Background: In a Cochrane Diagnosis Test Accuracy review of maternal screening for Down’s Syndrome, we observed that the estimated sensitivity of the 2nd trimester serology based triple test appeared to be 20% lower in studies of women aged >35 compared to studies unrestricted by age.
Objectives: To investigate whether the study design, and associated risk of loss to follow up, could explain the observed relationship.
Methods: Methodological characteristics of the included studies were evaluated and considered. Results were plotted in summary ROC space identifying studies according to their methodological characteristics as well as maternal age. Bivariate random effects meta-regression models were used to estimate the effect of study characteristics on test performance. Sensitivity analyses were used to investigate whether differential loss to follow-up could explain the observed relationship with age.
Results: Two distinct groups of studies were evident: those in which the reference standard diagnosis involved follow-up to birth (risking loss to follow-up) and those where all patients received invasive confirmatory testing (minimising loss to follow-up). Only studies of older mothers used the latter design. Comparison of test accuracy between these groups for the largest dataset showed, for a false positive rate of 5%, significantly higher sensitivity (68.6% 95% CI 62.3-74.3%, n = 11 versus 48.4%, 95% CI 40.7-56.2%, n = 13, p<0.0001) for follow up compared to confirmation studies. Sensitivity analyses showed that the observed relationship with maternal age could partially (but not totally) be explained by differential loss-to follow-up in studies undertaken using follow-up to birth.
Conclusions: The apparent increase in sensitivity in follow up compared to confirmation studies may partially be an artefact of differential loss to follow-up, where test negative women who had Down’s syndrome fetuses (false negatives) are more likely to miscarry. In confirmation studies, all false negatives were likely to be identified, giving more reliable estimates of test accuracy.
Objectives: To investigate whether the study design, and associated risk of loss to follow up, could explain the observed relationship.
Methods: Methodological characteristics of the included studies were evaluated and considered. Results were plotted in summary ROC space identifying studies according to their methodological characteristics as well as maternal age. Bivariate random effects meta-regression models were used to estimate the effect of study characteristics on test performance. Sensitivity analyses were used to investigate whether differential loss to follow-up could explain the observed relationship with age.
Results: Two distinct groups of studies were evident: those in which the reference standard diagnosis involved follow-up to birth (risking loss to follow-up) and those where all patients received invasive confirmatory testing (minimising loss to follow-up). Only studies of older mothers used the latter design. Comparison of test accuracy between these groups for the largest dataset showed, for a false positive rate of 5%, significantly higher sensitivity (68.6% 95% CI 62.3-74.3%, n = 11 versus 48.4%, 95% CI 40.7-56.2%, n = 13, p<0.0001) for follow up compared to confirmation studies. Sensitivity analyses showed that the observed relationship with maternal age could partially (but not totally) be explained by differential loss-to follow-up in studies undertaken using follow-up to birth.
Conclusions: The apparent increase in sensitivity in follow up compared to confirmation studies may partially be an artefact of differential loss to follow-up, where test negative women who had Down’s syndrome fetuses (false negatives) are more likely to miscarry. In confirmation studies, all false negatives were likely to be identified, giving more reliable estimates of test accuracy.