Estimating birth prevalence of Down's syndrome.

Abstract

Background: Estimates of maternal age-specific prevalence of Down's syndrome are needed for the assessment of environmental factors, for counselling and monitoring screening programmes. The estimates should relate to populations of women who have not received prenatal screening. This is normally achieved by using data collected before the widespread use of screening. The problem of under-ascertainment in some data-sets has been recognised in the literature, but has not been dealt with satisfactorily in the statistical models used to estimate live-birth prevalence.

Methods: In this paper we develop a model that takes explicit account of under-ascertainment and apply this model to data from nine published studies. The primary aim of our analysis is to provide an improved model for live-birth prevalence. A secondary aim is to examine the ascertainment rates in the nine studies.

Results: The proposed model provides a good fit to all but one of the nine studies, although exclusion of this study does not affect the estimated risks. The estimate of risk weighted across the maternal age distribution is 1.41 in 1000 live-births [90% confidence interval (CI) 1.37-1.49].

Discussion: Comparing this figure with those obtained from published rate schedules suggests that the proposed model predicts rates that are some 10% higher than those obtained when ascertainment is assumed to be complete in all studies. The predicted rates are similar to those calculated when only those studies known to have high levels of acertainment are included.