- © 2008 Canadian Medical Association or its licensors
[Six of the authors respond:]
We thank Agustín Conde-Agudelo and Marshall Lindheimer for giving us the opportunity to clarify the interpretation of our findings. We regret that they interpreted our words as a strong recommendation for routine use of Doppler ultrasonography in clinical practice. In the abstract we stated that “a pulsatility index, alone or combined with notching, is the most predictive Doppler index. These indices should be used in clinical practice.”1 Our intention was not to recommend the routine use of Doppler ultrasonography but rather to emphasize that if it is used then the pulsatility index, alone or combined with notching, is the best choice.
More generally, we do not think that firm clinical recommendations should be made on the basis of what might be called early-phase diagnostic studies or meta-analyses thereof.2 A more formal economic modelling analysis on this topic, although still hampered by the use of early-phase diagnostic studies only, showed that the routine use of Doppler ultrasonography cannot currently be considered cost-effective.3
Conde-Agudelo and Lindheimer raise 3 methodologic concerns. First, the statistical test for heterogeneity has bad statistical properties, making such tests virtually superfluous. Although the I2 statistic is an improvement,4 we agree with its inventors that “quantification of heterogeneity is only one component of a wider investigation of variability across studies, the most important being diversity in clinical and methodological aspects.”5 We carefully dealt with methodologic diversity using predefined stratified analyses. Second, funnel-plot asymmetry may be caused by at least 6 different mechanisms, of which publication bias is just 1. This is why experts in the field now prefer the term small-study bias. Without firm criteria to distinguish the sources for the asymmetry, interpretation of such plots remains speculative.6,7 Finally, the nonindependence of sensitivity and specificity is a phenomenon for which the bivariate method explicitly accounts.8 In conclusion, we concur with Conde-Agudelo and Lindheimer that it is still too early to recommend routine use of Doppler ultrasonography to predict a pregnant woman's risk of developing pre-eclampsia and intrauterine growth restriction.
Footnotes
-
Competing interests: None declared.