Page 137 - 2021_07-Haematologica-web

FVIII activity and bleeding risk in hemophilia A
have a higher FVIII:C for the majority of time between doses. When we describe FVIII:C at 15% in our analysis, this reflects a momentary FVIII:C value between peaks and troughs, rather than a trough value (Figure 1). As such, bleeding risk will not be the same in a patient at a momen- tary FVIII:C level of 15%, if the trough in this patient is much lower, compared with a patient with a trough level of 15%, who would have much higher FVIII:C for most of the time.
In addition to reporting negative binomial distribution estimates of ABR, we used Kaplan-Meier analysis and log- rank statistics to assess the bleeding risk at different FVIII levels. This analysis provides a slightly different perspec- tive on the data by depicting the proportion of bleed-free patients over a given period of observation, confirming the impact of FVIII levels on the risk of bleeding. It should be noted, however, that there was no plateau reached dur- ing the analysis period of 2 months in our study, and it can be expected that the proportion of bleed-free patients would decrease further over time. The proportion of bleed-free patients depicted in Figure 3 should not, there- fore, be compared directly to proportions of bleed-free patients reported from other studies over different periods of time.
Our analysis also has some potential sources of error such as: the misinterpretation of symptoms of pain as a
AB
bleed; delayed recognition of bleeds resulting in a shift of the start of a bleed to some time after the next injection; and inaccuracy of injection records. Assuming these potential errors decreased during the guardian trials, this may provide some explanation for the decline in ABR over time. On the other hand, we cannot exclude underreport- ing of bleeds by patients during the extension phase, par- ticularly when visits changed from every 2 months during the pivotal phase, to every 6 months during the extension phase.
PK data from patients treated with FVIII products can be described using one- or two-compartmental PK models.17 Two-compartmental models may not be identifiable from PK data with less intensive sampling, such as data from children.17 Furthermore, one-compartmental models are less sensitive than two-compartmental models to the han- dling of observations below “limits of quantification”. This is because a smaller number of parameters are to be estimated from the given pool of data. While there is no well-established consensus regarding handling in popula- tion PK modeling,18 these choices can have profound effects for FVIII PK modeling.19
Overall, our data suggest that trough level targets need to be adapted according to a patient’s age and previous treatment and should be revised regularly over time. We hypothesize that higher levels may be required in patients
CD
Figure 2. Estimates of annualized bleeding rates for spontaneous bleeds for five FVIII:C categories in the two trial phases and age groups. (A-D) Negative binomial estimates of annualized bleeding rate (ABR) for five FVIII:C categories (0-1%, >1-5%, >5-20%, >20-50%, and >50%) for spontaneous bleeds during the pivotal (A) and extension (B) phases for adults/adolescents, and during the pivotal (C) and extension (D) phases for children. FVIII:C: factor VIII activity.
haematologica | 2021; 106(7)
1907