6 McMaster-Fay RA. Br J Obstet GynaecolI992; 99: 527-528.
AUTHORS' REPIS
Sir;We thank Fay et al. for their interest in our publication. Although they direct their criticisms at our methods and results, we think that the main reason for their disagreements with our paper are dependent on their interpretation of the results. Although a test can be statistically associated with a disease, this does not prove that this test is clinically useful. For a screening test, sensitivity and positive predictive value are the two performance characteristics required to justify further surveillance or interventional procedures.Fay et al. find the definition of pre-eclampsia we used inappropriate. Yet it is in accordance with the National High Blood Pressure Education Program Working Croup Report on High Blood Pressure in Pregnancy', as cited in our article, and with the recent Report of the Canadian Hypertension Society Consensus Conference'. Using this definition, the incidence of pre-eclampsia was 3.6% in our cohort', compared with 4.7% (36/768, see Table 1 of that paper) in the study cited by Fay'. Hyperuricaemia is not part of any definition of preeclampsia, although it has once been proposed as a prognostic factor of the severity of the disease.We did not report in detail on the predictive value of uterine artery waveforms (notch) at 18 weeks. We cited A:C ratio as we have previously validated this as a substitute for the uterine artery notch5. In our cohort the performance of the uterine artery notch assessed at 18 weeks to predict pre-eclampsia was equally poor: sensitivity = 0.50, specificity = 0.64, and positive predictive value = 0.05.Contrary to the perception of Fay et al. we not only considered each complication individually, but also studied the three complications together. We reported that when pre-eclampsia, small for gestational age and spontaneous prematurity were combined as indicating a poor perinatal outcome, the predictive performance remained similar (sensitivity = 0.24, specificity = 0.89, likelihood ratio for a positive test 2.18), except for the positive predictive value (0.38). which increased, as the prevalence of the studied outcome was higher (21%). We showed that to predict induction before term for pre-eclampsia or low fetal weight (prevalence I%), the sensitivity and the specificity of an abnormal Doppler test were better (0.70 and 0.87, respectively), but the positive predictive value was still disappointing at 0.04.McMaster-Fay and colleagues assert that a test with a high sensitivity is a good predictive test and deserves a better quality assessment than ours. Yet, when using the numbers they published', we calculate a low sensitivity of 0.22, a low positive predictive value of 0.11 and a likelihood ratio of an abnormal test of 2.43 (95% CI 1.26 to 4.66). This is very similar to our findings, as reported in Table 1 of our article'. The strengths of our study were that we specified diagnostic criteria for an abnormal Doppler test and for abnormal outcomes before analysis, and that blinding w...