Reflections on the FranceCoag report on inhibitory antibodies to factor VIII in patients with severe hemophilia A

To the editor:

Calvez et al¹  exhaustively reported their excellent analysis of the FranceCoag Database, showing a higher risk of inhibitors for Kogenate over Advate. Because they acknowledge the absence of a definitive biological rationale for their findings, we wish to suggest a possible explanation for their findings.

Calvez et al¹  found that Kogenate was associated with an inhibitor rate of ∼50% (1 in 2 patients). They also showed that this rate has not changed over time, which implies that this evidence has been available from their countrywide surveillance system for some time before their recent publication, which, as discussed in the paper, has been prompted by the unexpected results from the PedNet Registry. Although it is possible that the high inhibitor rate with Kogenate had gone unnoticed, a plausible alternative explanation is that such a high event rate was somehow anticipated, as a consequence of a selective use of Kogenate in patients at higher risk of inhibitor development. Indeed, imbalance in the baseline risk of events is the most common flaw in observational studies. To test this hypothesis, we provide an additional sensitivity analysis of the FranceCoag data and an explanation.

In supplemental Table 19 from Calvez et al,¹  a breakdown by center of their database is provided. Only 3 centers contributed >10 patients on Kogenate or Advate and observed an inhibitor rate for Kogenate of >50%. In Table 1, we contrast pooled data from these 3 centers (ie, 1, 3, and 13 in their supplemental Table 19) with pooled data from the remaining 34 centers.²  The absolute difference in inhibitor rate between Kogenate and Advate was a clinically and statistically significant 41% (95% confidence interval [CI], 17% to 47%) for the 3 selected centers and a completely irrelevant 3% (95% CI, −11% to 18%) for all other centers (with a large sample size of 161 previously untreated patients [PUPs] and 56 inhibitors; overall rate, 34%).

From the participant list provided in Calvez et al,¹  centers 1, 3, and 13 are Paris Necker, Lille, and Strasbourg. During the years between 1997 and 2001, Paris Necker and Lille provided data on Kogenate for a study published in 2005,³  claiming a very low inhibitor rate. We consider it reasonable to imagine that the 3 centers, based on their experience,³  had been keen to use Kogenate in high-risk patients with the hope of reducing their inhibitor risk and had not been surprised by observing a 60% inhibitor rate in 93 higher-risk patients. Correspondingly, the inhibitor rate in the patients treated with Advate in those 3 centers was 17%, which was significantly lower than the overall rate and very close to the 15% inhibitor rate reported by Kreuz et al³  for Kogenate. Thus, a possible explanation is that with selection of Kogenate for higher-risk patients, the patients who remained for Advate were necessarily at lower risk of inhibitor development. In contrast, it is possible that all the other centers considered Kogenate or Advate as equivalent choices for their patients and thus did not introduce confounding by indication. Unfortunately, even the state-of-the-art multivariable analysis performed by Calvez et al¹  cannot adjust for all the subtle nuances (including, but not limited to, low economic and education level, difficulty in vein access, expected low compliance, and their association with known risk factors for inhibitors) that constitute the expertise of those treating hemophilia in assessing the risk of inhibitors of their patients.

The most efficient option to address residual confounding would be to randomize 120 patients to Kogenate or Advate, ruling out a hazard ratio of ≥1.6 with a power of 80% and a 1-sided α of 0.05 and adjusting the analysis for known covariates. Is this feasible, ethical, and worthy? We invite the readers’ opinion via e-letters, and we prompt manufacturers to consider sponsoring such a trial.

Kessler and Iorio⁴  suggested that a center effect was not tested in Research of Determinants of Inhibitor Development (RODIN), but the authors ignored our criticism.⁵  We could test the center effect with the FranceCoag report, because of the more detailed reporting. We again urge the RODIN authors to report center-level data for their study. These data belong to the community and cannot be kept secret, leaving patients and those treating the patients with doubts as to whether they are receiving suboptimal treatment. Any rational clinical decision has to be made after considering the entire body of available evidence.