Re: Risk of miscarriage following amniocentesis and chorionic villus sampling

I read with much interest the systematic review and meta-analysis by Dr Akolekar et al.1. The authors attempt to provide an updated figure on the risk of miscarriage associated with invasive prenatal diagnostic testing. They do an admirable job of trying to minimize the risk of bias in the series selected, by including only studies with control groups, published after the year 2000 and with more than 1000 procedures. They conclude that the pooled risk of amniocentesis-related miscarriage before 24 weeks is about 0.11% and that the risk is not significantly different between amniocentesis and control groups. Both in the discussion and in the conclusion of their review, the authors also suggest that the pooled procedure-related risk is likely to be overestimated, because pregnancy characteristics which may increase the risk of chromosomal abnormalities (and hence the uptake of invasive testing) may also be associated with increased risk of miscarriage. Surprisingly, the authors do not mention the obvious counterbiases, i.e. those in both study and control groups that are likely to lead to underestimation of the procedure-related risks. The most glaring of these biases is that, in all of the studies, fetal losses among controls included both normal and abnormal karyotypes, whereas the losses after procedures involved euploid fetuses (since aneuploid pregnancies were usually terminated). This limitation leads to an underestimation of the post-amniocentesis spontaneous fetal loss rate: it is estimated that about 25% of trisomy 21 fetuses are lost during pregnancy, and 40% of such losses occur by 24 weeks2. Loss rates are even higher for other chromosomal anomalies, such as trisomies 13 and 18 and Turner syndrome. Some of the seven studies included in the meta-analysis on amniocentesis had additional biases. In two studies, an ultrasound scan to demonstrate fetal viability and absence of major anomalies in the midtrimester was not done for fetuses in the control group3, 4; not surprisingly, in one study, eight of 39 fetal losses in the amniocentesis group were detected at ultrasonography (and were excluded from the computation of the meta-analysis), whereas there was no mention of any such losses among the 197 fetal deaths in the control group3. The same study had significant differences in patients lost to follow-up between cases and controls (6.2% vs 7.4%, respectively; P = 0.005). The other study listed the gestational ages at miscarriage among cases (allowing selection of cases below 24 weeks for the meta-analysis), but not among controls (all losses among controls were included in the meta-analysis)4. A third study had a suspiciously low number of control subjects (ratio of study to controls was 2:1) and it provided no description of the criteria for inclusion of the controls5. Similar biases were also present in the three studies included in the meta-analysis for risk of miscarriage following chorionic villus sampling. The most important conclusion of the meta-analysis by Akolekar et al. may be that there is significant heterogeneity among studies (P < 0.0001, Cochran's Q test), thus limiting greatly the robustness of any conclusion drawn from such an analysis. As we move into the era of cell-free DNA testing, with its associated decline in uptake of invasive procedures, it is unlikely we will see better-quality studies published, and so the uncertainty regarding risk of procedure-related loss is likely to persist. I applaud the authors' enormous effort to provide our community with contemporary data on the risks of invasive prenatal procedures. This work is likely to be widely disseminated. I offer these counter-arguments in the spirit of open discourse, as it is in the best interest of our patients to have accurate statistics, to enable them to make an informed decision for their family at such a time of great stress. A. Ghidini Antenatal Testing Center, Inova Alexandria Hospital, 4320 Seminary Road, Alexandria, VA 22304, USA (e-mail: [email protected])