Can a “center effect” explain the higher frequency of inhibitors for a second-generation recombinant factor VIII product?

To the editor:

The Research of Determinants of Inhibitor Development (RODIN) study published a higher risk for inhibitors by a second-generation full-length recombinant factor VIII (rFVIII) (adjusted for several potential confounders) (hazard ratio [HR], 1.60; 95% confidence interval [CI], 1.08-2.37) as compared with third-generation full-length rFVIII, based on 547 previously untreated patients (PUPs) with severe hemophilia A.¹  Recently, these findings were confirmed by 2 independent cohort studies from France and the United Kingdom.^2,3  The French group reported on 303 PUPs with severe hemophilia and found adjusted HR of 1.55 (CI, 0.97-2.49), and in the United Kingdom report, the adjusted HR of inhibitor development for the second-generation full-length rFVIII product compared with the third-generation full-length rFVIII product was 1.75 (95% CI, 1.11-2.76). Recently, it was suggested that the preference for certain products in a hemophilia treatment center might lead to an additional confounder, a so-called “center effect.”⁴  A “center effect” might arise when centers differentially choose to prescribe certain products for predefined patients. In a recent letter in Blood by Iorio and Berntorp, our group was requested to share data on a potential “center effect.”⁴  The leading hypothesis was that larger centers might include more patients with a higher initial risk. The PedNet Study Group is a collaboration of 29 centers and for this analysis, the most recent update of January 2015 was used.

We identified 10 large centers (≥25 patients) that included a median of 35 patients (range, 25-49) over a 10-year period. The 19 smaller centers included a median of 13 patients (range, 2-23) (Table 1).¹  Large centers showed a higher incidence of inhibitor development as compared with small centers: 34.5% vs 24.0% (P = .006). Overall, large centers had significantly more patients with a high-risk gene mutation. Dosing during the first peak treatment moment was higher in large centers (P = .001) (Table 1).

However, the higher total inhibitor incidence of large as compared with small centers was mainly due to the detection of more low-titer inhibitors: 11.3% for the large centers vs 6% for the small centers (P = .026). The incidence of high-titer inhibitors was not significantly different: 23.1% (large centers) vs 18% (small centers) (P = .125). It is likely that the higher number of low-titer inhibitors found in the large centers is at least partly due to the observed higher testing rate for inhibitors. Large centers tested a median of 5 times in 50 EDs, whereas small centers tested only 3 times (P = .002).

The difference in inhibitor incidence between large and small centers did not affect the higher risk for inhibitors found with a second-generation rFVIII product. The percentage of patients using second-generation full-length rFVIII products was similar: 29.9% in large centers and 32.4% in small centers (P = .520). This contradicts the hypothesis of an additional “center effect” caused by a preferential use or avoidance of second-generation products. A factor that has been suggested to influence treatment choice is a positive family history for inhibitors. As is clear from Table 1, however, a positive family history for inhibitors was present in only a small percentage of our patients. It is highly unlikely, therefore, that a preferential choice of product could have had an impact on our results. In conclusion, no “center effect” was found comparing small and large centers that contributed to the higher inhibitor incidence with a second-generation full-length rFVIII product.