Prediction of fetal status in fetal/neonatal alloimmune thrombocytopenia (FNAIT)?

To the editor:

We read with interest the recent paper from Bertrand et al,¹  which outlines an interesting topic regarding the use of antibody concentration as predictive parameter for FNAIT.

Unfortunately, the number of subjects studied is very low. When the authors conclude that FNAIT is a severe disease and the first offspring is already at high risk, they base their analysis on 66 “index cases.” A significant difference in antibody concentration between primagravidae and multigravidae, however, is concluded from the analysis of 54% (20/37) and 34% (10/29) of enrolled cases only. Selection criteria are not given. In consequence, complete data were only available from 30 subjects over a period of 27 years. As the incidence of FNAIT is 1:1000 live births, and 600 FNAIT cases can be expected in France per year, we fear that the number of subjects may be too small to draw general conclusions.

More specifically, the following aspects require special attention:

We believe that the design of the study does not allow concluding that no significant correlation exists between the HLA-DRB3 allele and the maternal alloantibody concentration at delivery. Primary data are missing. However, the frequency of this allele among mothers of children with anti–HPA-1a–mediated FNAIT is as high as 98%.²  Accordingly, a significant difference in the frequency of this allele between any selected groups of mothers is practically excluded a priori and should always prove statistically insignificant.

We also believe that the statistical methods applied to conclude that the most efficacious treatment is maternal therapy with IVIG and steroids are insufficient. The conclusion is drawn from 155 treated pregnancies (not 239 pregnancies as stated in the “Patient cohort” section). Statistical comparisons between platelet counts of newborns assigned to different treatment groups were made with the Mann-Whitney U test, which requires independent samples. This requirement is violated by the fact that the study compares unrelated newborns plus (groups of) siblings born to the same mothers. In addition, the “no treatment” group consists of index cases, that is, (almost exclusively) of first-born babies. This bias may have a systematic influence on platelet counts independent from therapy.

Finally, the authors state that the maternal alloantibody concentration during pregnancy is predictive of fetal thrombocytopenia. Readers should note that the negative predictive value in this study was 75%; accordingly, every fourth unborn will go untreated although it does require treatment. The authors state that their recent study confirms their own data from a previous report.³  Such a cross-reference is of value if the 2 study groups are independent. We noticed that in the former study, 27 cases were enrolled between 1984 and 2004; and in the recent study, 28 cases were collected from 1981 to 2009. Furthermore, the threshold of the antibody concentration was lowered by a factor of 10 and the scientific and statistical justification for the definition of these different cut-off values is missing.

Taken together, several limitations apply when interpreting this study. Data are not yet sufficient to allow for guiding the management of pregnancies in FNAIT and we should still consider whether antibody concentration will prove helpful in future an open question.