Risk factors for pre-eclampsia at antenatal booking: prediction or causality?
We congratulate Duckitt and Harrington  for their inspiring review on risk factors for pre-eclampsia at antenatal booking. Yet we have difficulty with some aspects of their approach. Most importantly, predicting illness is an issue of (statistical) association, not necessarily one of causality. Hence confounding, matching, and statistical adjustment are irrelevant . The authors appear to have missed this distinction. Inability to discriminate between prediction and causation at the outset has downstream implications for the rest of the review process and its validity and inferences. Our critical observations concern literature searching, quality assessment and data interpretation.
First, the authors do not describe their search strategy in the detail expected for a rigorous systematic review. The number of references retrieved depends on how terms are combined and truncated. Did the authors use MeSH terms, keyword terms, or both and were search terms combined with the Boolean operators “AND” or “OR”? Furthermore, independent duplicate selection of papers should minimise the number of relevant articles missed. For example, at least one relevant paper has been missed .
Second, the authors excluded all studies that did not report on comparability at baseline or those that showed incomparability, although these causality-related issues are irrelevant to prediction . Adoption of Taggart et al.’s quality checklist was inappropriate, since Taggart et al.’s research question (‘Effect of arterial revascularization on survival’) was indeed one of causality, not prediction .
Third, regarding interpretation it is confusing where the terms risk versus relative risk have been used. The objective of the review is stated to be “To determine the risk of pre-eclampsia …at antenatal booking” (Abstract). The Abstract concludes with the claim that the factors identified could be used to assess risk at booking. However, the review determined relative risks, not absolute risks. If clinicians know a woman’s background (a priori, pre-‘test’) risk given all the information on that patient prior to learning the results of a factor reviewed by the authors, the absolute, or post-‘test’, risk at booking can be attained by multiplying the pre-test risk with the relative risks from the review.
Given the (potential) problems that we have highlighted, patients, obstetricians, and midwifes may be advised to postpone adoption of recommendations from this review until more definitive data are available.
1. Duckitt K, Harrington D: Risk factors for pre-eclampsia at antenatal booking: systematic review of controlled studies. BMJ 2005;330:565.
2. Grobbee DE: Epidemiology in the right direction: the importance of descriptive research. Eur J Epidemiol 2004;19:741-744.
3. Konijnenberg A, van der Post JA, Mol BW, Schaap MC, Lazarov R, Bleker OP, Boer K, Sturk A: Can flow cytometric detection of platelet activation early in pregnancy predict the occurrence of preeclampsia? A prospective study. Am J Obstet Gynecol 1997;177:434-442.
4. Taggart DP, D'Amico R, Altman DG: Effect of arterial revascularisation on survival: a systematic review of studies comparing bilateral and single internal mammary arteries. Lancet 2001;358:870-875.
Competing interests: ongoing research project on prediction of pre-eclampsia
Competing interests: No competing interests