Comment on: “The Supply Side of Coagulation: Fibrinogen and Bleeding in Liver Disease and Transplantation.” The Known, the Unknown, the Possibly Known, and the Impossible

We read with great interest the commentary written by Dr Wagener regarding our recent article on preoperative fibrinogen level and bleeding in liver transplant recipients (LTRs) with end-stage liver disease.^1,2 The author pointed out multiple concerns regarding the generalizability of our results and the consequently limited scope of our findings. Although we agree on many points raised, we believe some of them must be brought into perspective.

Generalizability is an epidemiological concept that refers to how findings from a sample, the study population, may be applied to a wider population, the target population. It depends on the extent of the differences in the distribution of the variables that may modify the observed effect between the study and the target population. Hence, generalizability is not related to the study power. Power depends on the number of observations in relation to the magnitude of the effect researchers want to estimate.

Our data on LTRs were collected from one center with peculiar transfusion requirements and practices (no use of viscoelastic tests, use of phlebotomies, and restrictive transfusions strategies).³ We acknowledged that these unique characteristics may limit the generalizability of our results, but we strongly believe there is a low probability that the preoperative fibrinogen level would be no more associated with intraoperative bleeding and transfusions in different clinical settings. To cancel out the observed association, the viscoelastic tests (VETs) should be a perfect test. Indeed, VETs must individualize the use of hemostatic blood products at a point where a preoperative fibrinogen level of 1 g/L would not be associated with a higher bleeding risk than a preoperative fibrinogen level of 3 g/L. We believe this is highly improbable, even if the use of VET has been associated with fewer transfusions requirements in LTRs.⁴ In contrast, in a setting with higher mean blood loss, it would be surprising that any variable explaining this higher blood loss would at the same time cancel out the effect of preoperative fibrinogen level. Based on basic hemostatic physiopathology, it is more likely that such association would be stronger in those clinical settings.

Finally, we agree that our findings might not suggest the use of preemptive fibrinogen in LTRs with a preoperative fibrinogen level of <3 g/L. Our findings rather suggest that these patients are at a higher risk of bleeding and blood product transfusions through different clinical pathways. However, we must bear in mind that current guidelines suggesting any transfusion threshold are based on in vitro studies or epidemiological studies using empirical categorization (Figure 1).⁴ To our knowledge, we are the first group to explore an unconfounded threshold effect of preoperative fibrinogen in this population using nonlinear data-driven modeling. Our results are consistent with those found in other populations using similar analytical strategies.⁵ Thereby, our study provides better evidence than the studies used to build the current guidelines on the fibrinogen level threshold effect in LTRs. Nevertheless, there is still a lack of high-quality evidence on the efficacy of fibrinogen administration at different timing and thresholds in this population.

FIGURE 1.: Transfusion risk according to fibrinogen level using 2 simulated modeling strategies. The two strategies modeled the effect of the independent variable using the same data set. A flexible nonlinear modeling of the fibrinogen level as a continuous variable and an empirical categorization of the fibrinogen level at 2 g/L will suggest different threshold effects in this situation.