Response: Sample size and statistical comparisons of GVHD rates in pediatric Niemann-Pick disease patients

We appreciate the thoughtful approach of Jakub Tolar and colleagues to the clinical implications of our studies involving a potential role for acid sphingomyelinase (ASMase)–derived ceramide in initiation and/or progression of graft-versus-host disease (GVHD). We believe they have asked the right question, that is, whether Niemann-Pick (NP) disease patients with minimal ASMase activity when treated with bone marrow transplantation might be protected from GVHD. Furthermore, we recognize that they have used standard criteria for GVHD severity. However, we do have some reservations regarding the statistical basis for their conclusion.

The authors report, in a retrospective survey, that upon marrow or cord blood transplantation, pediatric patients with NP disease had an incidence and severity of GVHD comparable with patients with inborn errors of metabolism (IEM) similarly treated. Examination of the data in Table 1 of their letter suggests that this is a reasonable conclusion for patients receiving a marrow transplant from a matched sibling donor. For patients who received either cord blood or marrow from an unrelated donor, a group with presumably a greater degree of mismatch, however, this conclusion is questionable.

We examined this inference based on data in Table 1 that 2 of 11 patients (18%) with NP disease experienced acute GVHD, grade 2-4, by day 100 versus 212 of 543 patients (39%) with IEM. Assuming patients without acute GVHD were followed for at least 100 days (a simplifying assumption since we do not have original data), an application of Fisher exact test would produce a P value equal to .22. Using this method of analysis or something similar, the authors conclude that acute GVHD rates in the 2 disease groups are comparable. An alternative explanation is that the sample size in the NP group is too small to detect a difference. The ability to detect a difference in the acute GVHD rates is driven by the minimum sample size between the 2 disease groups. In this case, the strength of the test comparison is diminished due to only 11 patients with NP disease. To demonstrate this, if we maintain the same acute GVHD rates as shown in Table 1 (18% vs 39%), but increase patient number in the NP group to 28 patients, so that 5 of 28 patients (18%) had an acute GVHD event, the P value for this comparison would equal .03 based on Fisher exact test. Under this scenario, the conclusion would be that there is a difference in acute GVHD rates between groups. Increasing sample size in the NP group to 55, resulting in 10 of 55 (18%) with acute GVHD, yields a P value equal to .002, indicating strong evidence that the acute GVHD rates are different in the 2 groups. A similar demonstration can be performed for the chronic GVHD endpoint. In conclusion, we believe the limiting factor in the authors' comparison of GVHD rates is the small number of patients (11) in the NP group. The inability to reject the (null) hypothesis that the disease groups have equal acute GVHD rates is not the same as the acceptance of this hypothesis.

Our published data are also consistent with the notion that the greater the mismatch, as would likely occur when comparing sibling-matched to unmatched donors, the larger the protection afforded by inactivating host ASMase.¹  In this regard, we reported enhanced protection from acute GVHD pathology by ASMase deficiency in a major histocompatibility complex (MHC)–disparate compared with mHA-disparate model. While we observed a 37.8% reduction in keratinocyte apoptosis in ASMase-deficient mice in the major histocompatibility antigen (mHA)–disparate model, a 71.4% reduction occurred in the MHC-disparate model. Similar differences were observed in small intestines and liver histopathologic scoring for established organ-specific acute GVHD parameters with ASMase-deficient mice scoring 35.0% and 34.6% better than wild-type littermates in liver and small intestinal scoring in the mHA-disparate model, and 54.6% and 65.0% better in the MHC-disparate model, respectively. Finally, while we observed equal protection from serum interferon-γ expression, significantly greater protection from serum TNF-α expression was measured in ASMase-deficient recipients in the MHC-disparate model (53.4% reduction compared with 28.8% in the mHA-disparate model). Although we have no current explanation for the apparent greater protection in the more mismatched model, we believe these consistent differences warrant additional research in the hopes of defining those parameters in preclinical models that might predict successful clinical intervention in acute GVHD by pharmacologic inactivation of sphingolipid signaling.

Lastly, although NP disease patients manifest up to 10% to 15% residual lysosomal ASMase activity, to our knowledge there is no information regarding residual nonlysosomal secretory ASMase activity^2-4  in this population. Secretory ASMase is the form used by cytolytic T cells for induction of GVHD, and requires added zinc, which is not routinely added to conventional ASMase assays. Hence, it currently is not possible to assess whether there might be sufficient residual secretory ASMase in NP patients for GVHD induction.