The Authors Respond

DISCLOSURES
No funding was received for the writing of this letter. Simpson reports consulting honoraria for participation in advisory boards for Bayer, CSL Behring, HEMA Biologics, Novo Nordisk, Octapharma and Takeda, and speakers bureau for Bayer and Novo Nordisk; Yan and Desai are employees of CSL Behring; Maro is an employee of Adivo Associates.


■■ The Authors Respond
We thank Preblick et al. for their comments on our article. rVIII-SingleChain has been developed with an improved pharmacokinetic and pharmacodynamic profile compared with standard-acting recombinant FVIII (rFVIII) products 1 ; this allows for a less frequent dosing regimen compared with standard-acting rFVIII, similar to regimens achievable with other long-acting products. Long-acting products are generally considered to include all the newer rFVIII products with

■■ Limited Generalizability of a Retrospective Chart Review Comparing 3 Recombinant FVIII Products for Hemophilia A Prophylaxis
We have read with great interest the April issue of JMCP, which featured an article by Simpson et al. that was a retrospective chart review comparing bleeding rates and factor consumption of 3 recombinant factor VIII products for hemophilia A prophylaxis. 1 We commend the authors for conducting this timely analysis, given the increased number of new therapies that have become available to the hemophilia community.
In the introduction, the authors discuss the concepts and definitions of long-acting, standard half-life (SHL), and extended half-life (EHL) factor therapy. To provide a framework for future assessments of rFVIII products, Mahlangu et al. (2018) published an evidence-based definition for EHL rFVIII. 2 Based on 3 criteria defined to classify an rFVIII as an EHL, it should be noted that rFVIII-SingleChain does not meet the criteria as an EHL, by definition and design, 2 as defined by Lambert et al. (2018). 3 Also, we highlight additional limitations, beyond those listed in the Simpson et al. article, that potentially skews study results and interpretation.
First, since the matching methodology is not documented in the article (e.g., 1:1 match, propensity score matching), it is challenging to understand the relevance of the findings. For example, while the study attempted to match cohorts on age and severity, the rVIII-SingleChain cohort has a much lower mean weight versus the rFVIIIFc (−12kg) and PEG-rFVIII (−22kg) cohorts. While there was an attempt to account for weight differences as part of the ANCOVA and normalization of factor utilization to IU per kg per week, given its significant effect on consumption, the analysis would be strengthened if the matching variables included weight or additional sensitivity analysis conducted.
Second, other known confounders of treatment outcomes of hemophilia were not accounted for in the analysis. Previous treatment regimen, previous bleeding events, existence of arthropathy, inhibitor history, and treatment center are examples of potential confounders that should be addressed. 4,5 Third, 17 (42.5%) of rFVIIIFc patients had a weekly dosing frequency other than 2 or 3 times per week, with 14 (35%) dosing < 2 times per week (n = 8 [20%] on weekly dosing). Based on Table 3 in the article, it can be calculated that the mean weekly factor consumption for these 17 patients was 77.6 IU per kg. It should be noted that 10% of the rFVIII-SingleChain group and none of the PEG-rFVIII group were on a weekly dosing schedule. Therefore, the comparison in Table 3 would be subject to bias; it would be of interest to assess outcomes in comparable dosing strata.
Fourth, the article limits the pre-vs. post-switch subgroup analysis to the rVIII-SingleChain cohort only; inclusion of all 3 cohorts in the switch analysis would have strengthened the study as a comprehensive comparative analysis.
improved pharmacokinetics, compared with standard-acting products, and rVIII-SingleChain has been described as such. 2,3 Given the very limited eligible/available patient pool on each product and the practicality of comparing multiple products across multiple centers, 1:1 and propensity score matching are not possible. We adopted a practical, sequential-like sampling method by completing data collection for the product with the smallest patient pool first and then aiming to obtain charts for patients with similar age and disease severity distributions treated with the other products, in order to achieve comparability across products. As Preblick et al. pointed out, we used the consumption measure normalized by weight (IU/kg/week) and further included weight in the ANCOVA models (in case of any potential effect not known on the normalized consumption); therefore, the effect of weight is already adequately adjusted. In fact, there were studies that suggested a negative association between body mass index and dosing, or using ideal body weight instead of actual body weight in dosing. 4,5 While it is not clear whether this translates to a negative correlation between weight and normalized consumption, we have not found literature indicating the contrary.
While we recognize that additional variables may affect treatment outcomes, there are many challenges to obtain them from patient charts because of frequently missing or incomplete information. Patients may have been using their product for years, and getting accurate previous treatment data may not be feasible. History of inhibitor status is sometimes unknown. Treatment patterns can vary by center, and in practice, it is not possible to account for center effects due to greatly varied product usage across centers. Finally, the existence of arthropathy could be a previous condition or an outcome of current treatment; therefore, it is difficult to standardize. Because of such data quality issues, including these variables can be problematic.
Dosing frequencies vary greatly across products, so the only unbiased estimate for consumption of a product is the overall across all dosing frequencies, as presented in Table 3. We also compared consumption for the 2 times per week and 3 times per week groups because they had the highest patient numbers and were also the only dosing frequencies used across all 3 products. In ANCOVAs for annualized bleeding rates, we included consumption as a covariate, which is statistically more efficient than comparing arbitrarily defined consumption strata.
Because rVIII-SingleChain became available in 2016, we were able to collect previous treatment information for all 40 patients on the product. However, previous treatment data are typically more challenging to collect for products that have a longer history (6 products were included in the study, and a