Validation plots in antenatal screening for Down's syndrome.

OBJECTIVE

To validate empirically the accuracy of antenatal Down's syndrome screening using the Integrated test, to compare this with other screening tests (including the Integrated test with the addition of cross trimester [CT] marker ratios) and to suggest how such validation analyses should be presented and interpreted.

METHODS

Using data from 7809 unaffected and 27 Down's syndrome pregnancies that had had an Integrated test, risk estimates for various screening tests (maternal age, Double, Triple, Quadruple, Combined, Integrated and serum Integrated tests) that use Integrated test markers were categorized according to quintile categories of risk estimates of the 27 affected pregnancies. For each screening test, the median risk estimate for each category was plotted against the observed prevalence within each category. Such validation plots were also produced for the Integrated test with CT marker ratios by measuring the level of the serum markers in the trimester of pregnancy not already measured in stored samples of all affected and a one-in-five sample of unaffected pregnancies. The robustness of the method was assessed by repeating the analysis for the Integrated test after re-classifying affected pregnancies with low risk estimates as unaffected, simulating the underascertainment of cases.

RESULTS

The validation plots (i) confirmed the accuracy of risk estimation for the different tests (by how close the points lay to the line of identity between predicted risk and observed prevalence), (ii) demonstrated the differences in screening performance of the different tests (by the range of risk spanned by the points and, in particular, the separation between the points representing the lowest risk and the next point), and (iii) are robust to underascertainment of affected pregnancies (by having little influence on the closeness of the points to the line of identity).

CONCLUSION

The validation plot is a useful, simple and robust way to assess the validity of new screening methods, to assess the accuracy of risk estimation and to audit the performance of screening programmes.