Factor replacement therapy is the standard treatment for hemophilia. However, some patients with hemophilia develop inhibitors (alloantibodies) against the clotting factor administered. The development of these clotting inhibitors may render replacement therapy ineffective and, consequently, increase morbidity because of the inability to control or prevent hemorrhages. Inhibitor development has been reported to occur in as many as 33% of patients with hemophilia A, and as many as 7.5% of patients with hemophilia B.1-3
Currently, 2 bypassing agents are available for use in patients who develop clotting factor inhibitors:
- Factor VIII inhibitor bypassing activity, an activated prothrombin complex concentrate (aPCC), also known as anti-inhibitor coagulant complex4
- A recombinant activated factor VII (rFVIIa).5 .
The ability of these agents to control bleeding episodes has been documented, with success rates ranging from 80% to 93%.6-9 However, responses can vary between patients,10-12 as well as within the same patient during the course of a single bleeding episode.13 Therefore, making appropriate adjustments to therapeutic regimens is critical for controlling problem bleedings in this patient population.14,15
Treatment Algorithm, Definitions
Systematic approaches to treating hemorrhagic episodes can potentially guide the most appropriate pharmacologic decisions for patients with hemophilia who develop clotting inhibitors. To date, there are no approved treatment guidelines for such patients who are experiencing problem bleeding episodes.
In the present study, problem bleeding is based on the definition by Teitel and colleagues, who have published a consensus algorithm addressing the treatment of problem bleedings (specifically iliopsoas muscle hematoma as the archetypal limb-threatening bleeding and intracerebral hemorrhage as the archetypal lifethreatening bleeding) in this population (Figure 1).14
Teitel and colleagues have defined problem bleeding as bleeding unresponsive to initial therapy with a single agent within a reasonable time (8-12 hours for non– life-threatening bleedings and 2-4 hours for life-threatening bleedings).14
The goal of the treatment algorithm was to aid physicians in determining the appropriate timing for treatment approaches to optimize patient outcomes. This algorithm includes the designation of time intervals between assessments, to allow for more effective treatment changes or dosage adjustments based on individual responses, in an effort to facilitate faster responses and improved outcomes.14
Although the algorithm specifies intervals and general recommendations for the treatment of problem bleedings (eg, increasing a dose or increasing the frequency of doses), it does not give specific dose recommendations or provide insight into the likelihood of improvement for patients at each interval. The lack of readily available clinical data precludes a direct comparison of outcomes and costs associated with adherence versus nonadherence to the consensus treatment algorithm.
Methods: Developing a Simulation Model
To circumvent the absence of data for this patient population, a simulation model was developed, informed by expert opinion, to estimate the likely outcomes and costs in relation to the treatment algorithm. Outcomes and costs were modeled for following the treatment algorithm (ie, adherence) and for not following it (ie, nonadherence) for each of these scenarios and for each of the 2 bypassing agents (ie, aPCC and rFVIIa). The present article describes the development of this simulation model and its benefits in evaluating outcomes and costs in this patient population.
The structure of the model was constructed from a consensus algorithm previously developed for controlling limb-threatening muscle bleedings in patients who have severe hemophilia A and high-titer inhibitors.14Using the algorithm as a foundation, the new model allowed the investigators to assess the impact of guideline adherence on outcomes and costs, assuming therapy would be initiated with either an aPCC or an rFVIIa.
Treatment patterns (ie, dosing and frequency) and the likelihood of a successful clinical outcome (defined as improvement in bleeding status at specific time points) associated with guideline adherence versus nonadherence were analyzed in this study based on the consensus opinion of an expert panel.
A decision-analytic model was used to evaluate the clinical and economic impact of adherence to the algorithm. Panel members developed the model to assess improvement at 12, 24, 48, and 72 hours after bleeding onset.
According to the treatment algorithm, after initial treatment, the dose or frequency of administration of both bypassing agents should be increased in patients who do not improve after 12 hours. If no improvement is observed after the specified adjustment at 24 hours, the algorithm suggests switching to the other bypassing agent. If no improvement has occurred at 48 hours, the dose or frequency of administration of the new agent should be increased.
Finally, the algorithm specifies that patients not improving after 72 hours should receive combined sequential therapy with aPCC and rFVIIa for 3 days. Combined sequential therapy consists of aPCC 50 units/kg administered every 12 hours (for a total of 7 doses in 3 days), with rFVIIa 90 mcg/kg (12 doses in 3 days) given between the aPCC doses. It is further assumed that after 3 days of sequential therapy, all bleeding episodes would be resolved and treatment stopped.
A corresponding scenario assuming nonadherence to the treatment algorithm has also been constructed in the present model. In that scenario, patients who did not improve at 12 hours would continue at the same dose and frequency for another 12 hours, thereby delaying the increase in dose or frequency of administration to 24 hours and the switch in agents to 48 hours.
The model assumed 2 possibilities: one in which treatment was initiated with aPCC and one in which treatment was initiated with rFVIIa. Within each treatment possibility, clinical outcomes and costs were mapped for following (ie, adherence) and not following (ie, nonadherence) the algorithm. Outcomes evaluated included resolution of bleedings at 72 hours, the need for combined sequential therapy, and treatment cost, which were compared for adherence and nonadherence to the algorithm.
Total drug costs were calculated according to the 2008 Medicare Part B allowance payment limits for aPCC and rFVIIa, considered to be $1.48/U for aPCC and $1.23/mcg for rFVIIa. One-way sensitivity analyses were performed to determine if the results were affected by changes in dosing or in drug pricing.
Expert Panel and Delphi Methodology
No evidence-based guidelines have been validated for the treatment of problem bleedings, and clinical studies have not been conducted to establish guidelines. Therefore, we used a modified Delphi approach16 in conjunction with the consensus algorithm to populate the parameters in the decision-analytic model. The Delphi method is a systematic interactive process led by a facilitator who relies on a panel of experts to anonymously answer questions during 2 or more rounds of discussion.16
After each round, a summary of responses or forecasts is presented, and panelists reevaluate their answers in light of the replies. During this process, the range of answers becomes smaller as the group converges on consensus. The process is repeated until a predefined stop criterion is reached. The average scores of the final round determine the end result.16,17
Consensus groups have been utilized previously to address the treatment of hemophilia.12,14 In addition, the Delphi method has been used in other complex disease states, such as asthma and epilepsy.18,19
To obtain model parameters for input, a modified Delphi methodology was used. Given the specialized nature of hemophilia, 7 international physician experts with experience in treating hemophilia in patients with clotting inhibitors convened in person to form the panel. An audience response system allowed panelists to respond anonymously to questions using a numeric keypad and displayed the results immediately. A moderator facilitated the questions and subsequent discussion for each question addressed to the panel.
A consensus was considered to be reached when the standard deviation of the responses divided by the mean response was less than 0.25. If a consensus was not achieved, a panel discussion would ensue, followed by a revote. Because this was a modified approach, as many as 3 revotes were allowed per question. Input was provided on the number of doses, average dose in U/kg or mcg/kg, and probability of improvement for each agent at specified intervals (12, 24, and 48 hours) separately for aPCC and for rFVIIa.
For example, panelists were initially asked how many doses they would administer during the interval from 0 hours to 12 hours. After panelists responded to the question the first time, discussion ensued, and after the question was asked again, a consensus was reached on the number of doses initially administered. Questioning continued to further delineate the dosing strength until all dosing parameters were established.
Because clinical data are not readily available, an approach similar to what was used to get dosing information from the physicians was used to determine the likelihood of improvement in bleedings based on the panelists’ own experience. No specific parameters were established for determining clinical improvement, because there is currently no standardized routine laboratory assay available,14 and improvement status would ultimately be up to each physician’s evaluation of a patient’s response.
The general assumption provided to the panelists regarding improvement was based on the criteria established for the timing algorithm by Teitel and colleagues: treatment should continue until the bleeding resolves, as indicated by the functional recovery of the muscle.14 Questioning ensued for both aPCC and rFVIIa for the adherence and nonadherence algorithm situations. Specific questions posed to the panelists are listed in Appendix A and Appendix B.
Figures 2, 3, 4, and 5 (pages 222-225) show the treatment algorithm chosen by the panel. Results of the modified Delphi process regarding the dosing, switching patterns, and likelihood of bleeding improvement are provided in Appendix A and Appendix B. Consensus was achieved for approximately 70% of the questions. Data obtained from questions for which consensus was not reached were still used in the model; however, these are indicated in Figures 2, 3, 4, and 5 by a superscript “a.” None of the 4 pathways modeled had 100% consensus for dosing/frequency or likelihood of improvement.
The total dosing for each interval, as well as the likelihood for improvement, are provided for each time interval. Of note, the dosing displayed in the figures is reflective of the total dose for each time interval. For example, in the algorithm with aPCC as the initial agent, a patient who improves at 12 hours and at 24 hours should receive a decreased dose. Initially, this patient would have received 81 units in the first 12 hours and 81 units in the next 12 hours but would receive 130 units total during the next 24 hours (assuming 2 doses of 65 units).
In the evaluation of clinical outcomes, patients who demonstrated consistent improvement from initiation of therapy were excluded from the model, because they would not be expected to require changes deviating from the treatment algorithm. The model showed that regardless of the initial bypassing agent administered, when the treatment algorithm was followed, an average of 74.4% of patients would improve during 72 hours (73.7% for aPCC; 75.1% for rFVIIa) compared with an average of 56.7% (51.7% for aPCC; 61.7% for rFVIIa) when the algorithm was not followed—a 31.2% relative increase in clinical improvement after 72 hours, which was attributed to algorithm adherence (Table).
Similarly, the proportion of patients not improving after 72 hours and requiring combined sequential therapy was lower when the algorithm was followed. If the subset of patients who consistently improved from initiation of therapy is removed, an average of 25.6% (26.3% for aPCC; 24.9% for rFVIIa) of patients would require combined sequential therapy when the algorithm is followed; if it is not followed, an average of 43.3% (48.3% for aPCC; 38.3% for rFVIIa) would require this regimen (Table).
Algorithm nonadherence increased the requirement for combined sequential therapy by 69.1%.
Based on a decision-analysis approach, costs associated with bypassing agent therapy are lower for adhering to the algorithm, regardless of the agent initiated. If 50% of patients used aPCC initially and 50% used rFVIIa, adhering to the treatment algorithm would reduce the overall costs by 5.6%.
Additional cost-savings are noted from algorithm adherence as a result of a reduction in the percentage of patients requiring combined sequential therapy (Table). Using combined sequential therapy for 3 days is estimated to be $92,604 per patient in drug costs alone. Avoidance of combined sequential therapy in 17.6% of patients, based on an absolute reduction from 43.3% to 25.6% of patients requiring sequential therapy, would generate savings as high as $16,305 per patient.
Sensitivity analyses indicated that the model is robust to changes in drug prices and clinical efficacy. When modeled prices were increased or decreased by 20% around base-case pricing, treatment algorithm application still resulted in cost-savings from 3.3% to 7.7% for aPCC use and from 2.3% to 9.7% for rFVIIa use.
Cost-savings were also demonstrated when initial efficacy (improving or worsening after the first dose) varied by as much as 20%. The range of savings for patients requiring combined sequential therapy was between 17.4% and 17.6% among those receiving aPCC initially and between 16.6% and 18.4% among those receiving rFVIIa initially.
The model’s purpose was to quantify the clinical impact of adherence to a treatment algorithm in addressing problem bleedings and to assign a monetary value associated with adjusting therapy in a timely manner, based on a systematic approach to therapy. The efficacy of bypassing agents in controlling bleeding episodes in patients with hemophilia who develop inhibitors has been documented, and the reported success rates range from 80% to 93%.6-9 Variability in response to therapy does occur11 and is attributed by some investigators to the different mechanisms of action of aPCC and rFVIIa.10,20,21
Responsiveness to treatment has also been documented to change during the course of a single bleeding episode.13 A clinical trial comparing aPCC and rFVIIa determined that both agents exhibit a similar effect on joint bleedings, but that the efficacy of the 2 agents is rated differently by a substantial proportion of patients with clotting inhibitors.10
Further complicating matters is the lack of agreement on dosing.22 Some protocols suggest the use of high daily doses, whereas others propose lower doses, and the success rates with both approaches are similar.23 These factors underscore the need to assess the adequacy of the response at regular intervals and make appropriate therapy adjustments in suboptimal situations.14,15
If patients who are not improving received treatment according to the algorithm, the dose or frequency of administration would be increased at 12 hours; therapy would be switched to the alternate bypassing agent at 24 hours for those still not improving; followed by an in crease in dose or in frequency of administration at 48 hours for those whose bleeding remains refractory. As a final step, patients would be given combined sequential therapy if they demonstrated no improvement at 72 hours.
If the treatment algorithm was not followed, the model assumed that no treatment evaluation would be made at 12 hours, and all subsequent decisions would be delayed. Based on the model simulation, adhering to the algorithm could result in a relative improvement of more than 30% at 72 hours. This increase in improvement demonstrates the effect of timely therapy adjustments in this population.
Regardless of the bypassing agent used at the initiation of therapy, algorithm adherence also results in fewer patients requiring combined sequential therapy. Based on the present model results, reducing the percentage of patients requiring sequential therapy by 17.5% has the potential for average cost-savings of $16,035 per patient.
Combined sequential therapy has been used for severe refractory bleedings,24 but its use is controversial.12 Although there may be situations when this approach is justifiable, minimizing the number of patients who require this regimen clearly has economic advantages and averts theoretical patient safety concerns. However, the agents examined in this analysis are not approved for sequential therapy, and adverse events, such as an increased risk of thrombosis, are a concern for some physicians.12
The costs of managing problem bleedings in patients with inhibitors can vary tremendously, depending on the bleeding episode itself and on the medication used.12 As a result, payers, providers, and patients face a significant challenge in trying to contain costs while achieving the best possible patient care. In view of the complexity of the condition, and the lack of consensus on optimal treatment, models such as the one described in this article can help providers with the decision-making process. Because this model is based on expert opinion, the information and recommendations set forth may assist with the construction of clinical guidelines to guide future therapies. Utilizing the assumption that the inputs derived from the consensus panel reflect real-world costs and outcomes, the information contained in this model may provide insight to clinicians when assessing patients with problem bleeding. In addition, guideline implementation may assist with payment and reimbursement strategies for the use of these agents.
No evidence-based treatment guidelines are currently available for this patient population, and the results of this study are based on a simulated model and must be interpreted as such. As with any model, there are limitations associated with interpreting the results. The model considers only limited direct costs associated with bypassing agent therapy (ie, drug costs). Additional direct costs, including those for hospitalization, physician visits, and rehabilitation care, as well as indirect costs associated with decreased productivity, were not considered.
In addition, costs were assessed according to Medicare Part B limits, which may not be reflecting an individual institution’s costs. The model aimed to apply the consensus algorithm to a real-world setting by using a panel of experts to characterize treatment patterns and effectiveness in an assessment of the clinical and economic benefits and consequences of algorithm adherence.
For several of the questions posed to the panel, no consensus was reached. In both of the consensus algorithm adherence and nonadherence scenarios presented here, the panel did not reach a consensus on multiple occasions regarding the dosing of rFVIIa. The panel was mostly in agreement with the likelihood of clinical improvement, with the noted exception regarding bleeding worsening at 12 hours and again at 24 hours.
The data presented in this article are based solely on expert opinions, recommendations, and assumptions; therefore, generalizing this information to real-world settings should be done cautiously. However, in the absence of studies directly assessing the impact of guideline adherence on clinical and economic outcomes associated with the use of bypassing agents to control spontaneous bleeding episodes in patients with hemophilia who develop inhibitors, these data provide the best available estimate for those outcomes.
Future studies that assess treatment algorithms such as this one may be warranted to obtain further evidence for the outcomes of bypassing agent therapy in patients with hemophilia. Being able to incorporate real-world data may further help to clarify areas of hemophilia treatment where currently no consensus exists.
Adherence to an algorithm in which treatment is altered at regular intervals based on clinical response has the potential to result in an increase in clinical improvement and reduction in the number of nonresponsive patients requiring sequential therapy. Moreover, algorithm adherence would result in a modest cost-savings in patients with hemophilia who develop clotting factor inhibitors and are experiencing problem bleedings.Appendices (click below to download PDFs)
- Appendix A - Panel Discussion on Dosing and Frequency Questions and Consensus Results
- Appendix B - Panel Discussion on Likelihood of Improvement Questions and Consensus Results
This study was supported by funding from Baxter Healthcare.
Author Disclosure Statement
Dr Gringeri has received research grants from Baxter Healthcare, Biotest, CSL Behring, and Grifols; he is a Consultant to Baxter Healthcare and Wyeth and is on the Speakers’ Bureau of Baxter Healthcare, Biotest, CSL Behring, Grifols, Kedrion, Novo Nordisk, Octapharma, and Wyeth. Dr Gomperts is a Consultant to Baxter Healthcare. Dr Leissinger has received research support from and is on the Speakers’ Bureau of Baxter Healthcare and Novo Nordisk. Dr d’Oiron is a Consultant to and on the Speakers’ Bureau of Baxter Healthcare, and Novo Nordisk. Dr Teitel is a Consultant to Baxter Healthcare and Pfizer and has received grant support from CSL Behring. Dr Young is a Consultant to Baxter Healthcare and Novo Nordisk. Dr Berntorp has received honoraria and research support from Baxter Healthcare. Dr Bonnet, Dr Ewenstein, and Dr Franklin have nothing to disclose.
- Ehrenforth S, Kreutz W, Scharrer I, et al. Incidence of development of factor VIII and factor IX inhibitors in haemophiliacs. Lancet. 1992;339:504-598.
- Ljung R, Petrini P, Tengborn L, Sjorin E. Haemophilia B mutations in Sweden: a population-based study of mutational heterogeneity. Br J Haematol. 2001;113:81-86.
- DiMichele D. Inhibitors: resolving diagnostic and therapeutic dilemmas. Haemophilia. 2002;8:280-287.
- FEIBA VH. Anti-Inhibitor Coagulant Complex (vapor heated) [package insert]. Westlake Village, CA: Baxter Healthcare Corporation; 2005.
- NovoSeven [package insert]. Princeton, NJ: Novo Nordisk; 2006.
- Negrier C, Goudemand J, Sultan Y, et al. Multicenter retrospective study on the utilization of FEIBA in France in patients with factor VIII and factor IX inhibitors. Thromb Haemost. 1997;77:1113-1119.
- Parameswaran R, Shapiro AD, Gill JC, Kessler CM. Dose effect and efficacy of rFVIIa in the treatment of hemophilia patients with inhibitors: analysis from the Hemophilia and Thrombosis Research Society Registry. Haemophilia. 2005;11:100-106.
- Key NS, Aledort LM, Beardsley D, et al. Home treatment of mild to moderate bleeding episodes using recombinant factor VIIa (Novoseven) in haemophiliacs with inhibitors. Thromb Haemost. 1998;80:912-918.
- Hilgartner MW, Knatterud GL. The use of factor eight inhibitor bypassing activity (FEIBA Immuno) product for treatment of bleeding episodes in hemophiliacs with inhibitors. Blood. 1983;61:36-40.
- Astermark J, Donfield SM, DiMichele D, et al; for the FENOC Study Group. A randomization comparison of bypassing agents in hemophilia complicated by an inhibitor: the FEIBA NovoSeven Comparative (FENOC) Study. Blood. 2007;109: 546-551.
- Quintana-Molina M, Martínez-Bahamonde F, González-García E, et al. Surgery in haemophilic patients with inhibitor: 20 years of experience. Haemophilia. 2004;10 (suppl 2):30-40.
- Allen G, Aledort L. Therapeutic decision-making in inhibitor patients. Am J Hematol. 2006;81:71-72.
- Hayashi T, Tanaka I, Shima M, et al. Unresponsiveness to factor VIII inhibitor bypassing agents during haemostatic treatment for life-threatening massive bleeding in a patient with haemophilia A and a high responding inhibitor. Haemophilia. 2004;10:397-400.
- Teitel J, Berntorp E, Collins P, et al. A systematic approach to controlling problem bleeds in patients with severe congenital haemophilia A and high-titre inhibitors. Haemophilia. 2007;13:256-263.
- Gomperts ED, Astermark J, Gringeri A, Teitel J. From theory to practice: applying current clinical knowledge and treatment strategies to the care of hemophilia A patients with inhibitors. Blood Rev. 2008;22(suppl 1):S1-S11.
- Linstone HA, Turoff M, eds. The Delphi Method: Techniques and Applications. Reading, MA: Addison-Wesley Publishing Company; 1975.
- Rowe G, Wright G. The Delphi technique as a forecasting tool: issues and analysis. Int J Forecast. 1999;15:353-375.
- Kaliner M, Amin A, Gehling R, et al. Impact of inhaled corticosteroid–induced oropharyngeal adverse effects on treatment patterns and costs in asthmatic patients: results from a Delphi panel. P&T. 2005;30:573-601.
- St Louis EK, Gidal BE, Henry TR, et al. Conversions between monotherapy in epilepsy: expert consensus. Epilepsy Behav. 2007;11:222-234.
- Turecek PL, Varadi K, Gritsch H, Schwarz HP. FEIBA: mode of action. Haemophilia. 2004;10(suppl 2):3-9.
- Hoffman M, Monroe DM III. The action of high-dose factor VIIa (FVIIa) in a cell-based model of hemostasis. Dis Mon. 2003;49:14-21.
- Lloyd Jones M, Wight J, Paisley S, Knight C. Control of bleeding in patients with haemophilia A with inhibitors: a systematic review. Haemophilia. 2003;9:464-520.
- Ingerslev J. Hemophilia. Strategies for the treatment of inhibitor patients. Haematologica. 2000;85(10 suppl):15-20.
- Schneiderman J, Nugent DJ, Young G. Sequential therapy with activated prothrombin complex concentrate and recombinant factor VIIa in patients with severe hemophilia and inhibitors. Haemophilia. 2004;10:347-351.