SEARCH

SEARCH BY CITATION

There is no doubt that recurrent fetal loss and pregnancy complications such as preeclampsia, intrauterine growth restriction (IUGR) and placental abruption are important and emotionally charged medical events. When offered therapy to treat or prevent these events many women are prepared to take personal risk if there is the hope of an increased probability of a successful pregnancy [1]. In the western world, marriage and child-bearing are occurring later in life and hence the years that are available for conception are fewer. All this tends to lead to decisions by patients and physicians that are often driven by emotion rather then facts or logic. However, as physicians we have an obligation to try to see above the emotional influences on treatment decisions and counsel appropriately. Primum non nocere, first do no harm. Whenever possible we should offer prophylaxis or treatment supported by evidence from studies with the best experimental design possible, and that, of course, is provided by randomized controlled trials. In the case of previously unproven therapies the ideal design is to compare with the current standard. In the case of complications of pregnancy and miscarriage there is no proven therapy and as such the comparator should be placebo or a no-intervention arm.

Brenner believes the information currently available is adequate to justify anticoagulant prophylaxis in women with thrombophilia and pregnancy complications [2]. Although the success rates described by Brenner, 75% live birth rates in women who previously had multiple pregnancy losses, are compelling and exciting, they do not come from an adequate study design [3]. The error in not having a placebo comparison group has been highlighted many times before. Many severe diseases that are not self-limited, such as lupus, may nevertheless undergo spontaneous remission. This remission can be interpreted as a treatment effect if we simply observe change in the course of disease after a treatment with no comparator.

The natural history of many diseases is unpredictable. The phenomenon of regression to the mean often guarantees that the use of historical patient data as a comparator will suggest an intervention has improved outcomes. Sacks et al. reviewed 50 randomized trials and 56 studies with historical controls [4]. Seventy-nine percent of trials with historical controls found the experimental treatment to be better but only 20% of trials with a concurrent randomized control group found the treatment to be better. The differences between the two kinds of trials occurred mainly because the control patients in the historical trials did considerably worse. Adjustments for prognostic factors when possible did not change the results of this review. Although it is probably true that short periods of time between the control period (preintervention) and the treated period will lower the risk that other aspects of medical care have changed during the interval, and result in a safer comparison, this type of comparison is still risky because of the placebo and Hawthorne effects. In the studies quoted by Brenner prior fetal loss was the main inclusion criterion and hence one would expect a low prior live birth rates in study participants [3,5,6]. As highlighted by Middledorp [7], in other trials in patients with recurrent fetal loss subsequent live birth rates have varied widely, which makes comparison with historical controls risky. Studies examining the use of heparin in patients with recurrent fetal loss and antiphosholipid antibody syndrome observed widely varying live birth rates in the control arms (44–72%) [8–10]. Further, in a similar study population comparing ASA to placebo live birth rates in both arms were over 80%[11].

Primum non nocere. Low-molecular-weight heparin (LMWH) is associated with serious side-effects such as hemorrhage, osteoporosis/fractures, heparin-induced thrombocytopenia, and LMWH may limit peripartum analgesic options because of the risk of epidural hematoma [12–16]. One must not forget that we are in the infancy of LMWH use in pregnancy (fewer than 2000 published cases) and that given the rarity of these serious drug-related events it is perhaps only a question of time before they appear.

The cost arguments put forth by Brenner need formal analysis to prove cost-effectiveness. Proper ‘number needed to treat’ data can only come from a randomized trial with a proper comparator group. We believe that Middeldorp is right to worry that physicians will gradually prescribe therapy to a healthier, better prognosis population if LMWH becomes the standard therapy without adequate data. There are over four million pregnancies per year in North America and as many as 400 000 women may have a thrombophilia. Given that miscarriage occurs in up to 30% of women widespread adoption of an attitude that LMWH will improve pregnancy outcomes could lead to enormous numbers of tests for thrombophilia in pregnant women and enormous numbers of women receiving LMWH without strong indication. It behooves us to define exactly who will benefit from LMWH, what are the complications associated with this therapy, and what is the cost-effectiveness or cost-benefit of this intervention before widespread use occurs in women who have suffered recurrent miscarriage or adverse complications in pregnancy. Given the scarcity of evidence for benefit and potential for serious harm, LMWH should be considered an experimental drug in pregnancy for these indications until data from controlled trials are published. Thrombophilic women, even those with previous pregnancy complications, should be closely followed in high-risk pregnancy units without prophylactic heparin if they are not in a clinical trial.

Acknowledgements

  1. Top of page
  2. Acknowledgements
  3. References

Dedicated to the memory of Noah Joseph Rodger, stillborn 12 February 2004.

References

  1. Top of page
  2. Acknowledgements
  3. References