The pathophysiology of FNAIT cannot be deduced from highly selected retrospective data

To the editor:

Today there is no consensus regarding the optimal method for determining the fetal status when the mother has alloantibodies against fetal platelets. Therefore, it was with great interest we read the article by Bertrand et al¹  recently published in Blood. However, we feel that a few points need to be clarified by the authors.

The study¹  included 75 women who underwent 239 pregnancies complicated by fetal and neonatal alloimmune thrombocytopenia (FNAIT). Given the same frequency of FNAIT in France as in Norway,²  the yearly expected number of newborns with FNAIT in France would be between 400 and 700. It is surprising that the authors report only the result of 155 pregnancies in Table 2. This represents 0.8%-1.4% of the total number of FNAIT cases occurring in France³  and 4.7%-8.5% of all FNAIT cases in Paris.⁴  Furthermore, of the 75 women who were studied, we were puzzled to discover the authors present antibody values at delivery from only 30 of these women. The very low number of pregnancies included clearly indicates that the cohort of patients is not a representative selection of FNAIT cases, and in our view it is very unlikely that the authors' findings can be generalized.

Ensuring homogeneity between groups is very challenging when data collection has taken place over 3 decades. The authors state that “… homogeneity of the subgroups was carefully controlled (cases collected between 1981 and 2009),” ^1p3209 but they do not explain how they controlled for homogeneity.

Without giving any data, the authors state that “no significant difference emerged between the concentrations of the untreated mothers of blood groups A and O” and they “did not observe any significant correlation between HLA-DRB3 allele and the maternal alloantibody concentration at delivery or the neonatal platelet counts.” ^1p3210 The reader cannot evaluate whether the lack of significance is because of a type II statistical error, and it is therefore impossible to assess the validity of the authors' claims.

To predict severe thrombocytopenia from maternal alloantibody levels the authors have calculated sensitivity, specificity, and positive and negative predictive value (see Bertrand Table 4¹ ). The authors should have reported the uncertainty of their estimates considering that their calculations were based on only 28 and 37 cases. This could have been done by calculating AUC values and the 95% confidence interval for the ROC curves used for selection of their cutoff values. We are also surprised that the authors did not discuss why their cutoff value was 10 times higher than the cutoff value from a large prospective study.⁵ 

The role of IVIg in the treatment of FNAIT is still a matter of debate, and hence it is surprising that the authors do not make reference to the recent studies by Giers et al^6,7 ; these authors could not find any effect of IVIg when given to either the mother or the fetus. The results of the treatment with IVIg and steroids are surprisingly good and may have been because of selection bias.