Full disclosure

The authors did not blind the study, which meant that patients who tested positive for the FVL mutation, along with their physicians, were notified of the results. As might be expected, this information influenced care. Affected participants were more likely to receive short courses of anticoagulation and were more frequently sent to another physician or service for investigation of possible DVT symptoms. Presumably, this reflects only a modicum of the increased anxiety that awareness of FVL mutation produced in patients and their physicians.

It is important to remember that like much of the literature in this area, the work by Kjellberg et al is not a clinical trial of any suggested therapy or intervention. While observational trials are very helpful in outlining the natural history of a disease process or condition, the generated data generally do not support particular treatments. As clinicians, we need to keep this in mind. For example, it would not follow from this study that patients with an FVL mutation should receive prophylactic anticoagulation in order to avoid the outcomes of interest.

No surprises

A number of findings were interesting but not unexpected. The group with the FVL mutation had a higher incidence of both DVT and pulmonary embolism than the control group. In fact, all of the diagnosed thrombotic events were in the FVL group. However, the numbers were small, and the results were not statistically significant.

In addition, the number of postpartum hemorrhages was lower in the FVL group, although the average blood loss at delivery was not different in the 2 groups. Journal Club members felt that the methods used to determine blood loss at delivery were suboptimal, an opinion that diminished their confidence in the reported estimations. They were more convinced that postpartum hemorrhage had been accurately assessed.

These 2 findings reinforced important points that had been found in previous studies. First, the main problem caused by FVL mutation is thrombosis, not adverse obstetric events. At the same time, it appears to offer some protection against excessive blood loss at delivery; perhaps that explains the persistence of the mutation.

Notably absent

The principal purpose of this study was to examine adverse obstetric outcomes, and these were defined broadly. A close examination of the statistical section indicated that the 3 primary outcomes of interest in this study were the incidences of preeclampsia, premature delivery, and IUGR, which were each scrutinized individually. Careful study of other possible outcomes—placental abruption, late miscarriage, intrauterine fetal death, and neonatal asphyxia—was impossible, since there were too few cases for meaningful comparisons.

Gestational hypertension was defined as 2 diastolic blood pressure measurements of ≥ 90 mm Hg, obtained at least 6 hours apart, after 20 weeks’ gestation. An increased incidence was noted but was not statistically significant. The study was powered to find significance if relative risk (RR) exceeded 2; the RR was 1.7. No increase in preeclampsia or severe preeclampsia was recorded. Again, cases were too rare to allow conclusions about differences in incidence.

When premature delivery was defined as birth before 37 weeks’ gestation, no difference was identified between the 2 groups. However, when the analysis was limited to infants delivered before 32 weeks, the surprising result was that FVL seemed to confer an advantage. National birth data suggest the incidence among control patients was at approximately the expected level. In contrast, the FVL group had fewer cases than expected. Although the finding was not statistically significant, it was striking and should be a catalyst for further study.

Most perplexing was the lack of association between FVL and IUGR. The authors labeled as a birthweight more than 2 standard deviations below the mean or a birthweight in less than the 10th percentile. Then they looked at mean birthweight deviations. The first is their clinical standard practice and what they used as their primary outcome for statistical purposes. They assumed that a rise in incidence of any adverse outcome of less than 2-fold was not statistically significant, as it was not clinically significant. In Table 3, the odds ratio for IUGR in their study group was 2.7, and the confidence interval did not cross 1. Yet, they concluded that this rise was not statistically significant.

At first, the reasoning behind this conclusion puzzled discussants. Many metaphors could explain this confusing statistical finding but basically, if you compare too many factors when looking for an association, you might find a spurious relationship by chance. This is more likely with an increasing number of comparisons, and the authors knew this. They made 6 comparisons, 3 involving primary outcomes. Because they made multiple comparisons between outcomes, they corrected for this in their statistics, making it harder to prove any association.

Selective screening

In negative studies, like this one, we are often most concerned with type II errors. Here, this is magnified by the authors’ care in voiding type I errors. By saying we want a P value of less than .05, we are saying, “Please do not reject the null hypothesis unless you are 95% sure you are correct in doing so.” The authors used 99%. As they did not reach this degree of certitude, they have not rejected the null hypothesis. But, we are unconvinced in the case of IUGR that this means they should embrace it.

Does this same reasoning apply to their other obstetric outcomes? In looking at the data tables, it does not appear so. However, certain results, such as severe preeclampsia and fetal death, were likely underpowered.

Overall, Journal Club members were concerned about several factors in this study. First, although there is no clear way to study this, it is possible that patients with the strongest gene expression of this mutation are those who had a prior history of DVT and treatment, and, as these patients were excluded, we cannot determine their risk status from this study. They might represent the group at highest risk for adverse obstetric outcomes, as well as thrombotic outcome, and so the associations might have been understated.

In addition, Journal Club participants wondered about possible connections between FVL and IUGR, severe preeclampsia, and other severe, less common outcomes. At present, patients who have experienced these events undergo a thrombophilia workup before they embark on a subsequent pregnancy. These data do not convince us that we are incorrect in ordering the relevant testing. In other words, these data do not exclude an association between FVL and these severe outcomes. However, we agree that data such as those presented here indicate that widespread screening for FVL mutation would be, at best, of uncertain benefit and possibly of some harm if it incites undue apprehension in patients.