Use of sorafenib in patients with hepatocellular carcinoma before liver transplantation: A cost-benefit analysis while awaiting data on sorafenib safety
Hepatology Jan 2010, Pages 165-173
Alessandro Vitale 1 *, Michael L. Volk 2, Davide Pastorelli 3, Sara Lonardi 3, Fabio Farinati 4, Patrizia Burra 4, Paolo Angeli 5, Umberto Cillo 6
1Unita di Chirurgia Oncologica, Istituto Oncologico Veneto, IRCCS, Padova, Italy
2Division of Gastroenterology, University of Michigan Health System, Ann Arbor, MI
3Oncologia Medica, Istituto Oncologico Veneto, IRCCS, Padova, Italy
4Divisione di Gastroenterologia, Azienda-Universita di Padova, Italy
5Clinica Medica V, Azienda-Universita di Padova, Italy
6Unita di Chirurgia Epatobiliare e Trapianto Epatico, Azienda Universita di Padova, Italy
email: Alessandro Vitale (email@example.com)
*Correspondence to Alessandro Vitale, Istituto Oncologico Veneto, IRCCS, Unita di Chirurgia Oncologica, Padova, Italy, Via Gattamelata 44, 35128 Padova, Italy
Potential conflict of interest: Dr. Volk advises Onyx.
The role of bridging therapies for patients with hepatocellular carcinoma (HCC) on the waiting list for liver transplantation (LT) remains controversial. There is strong evidence to support the effectiveness of sorafenib in extending the time to progression of HCC. Using a Markov model, we compared two strategies: one using sorafenib as neoadjuvant therapy before LT (Strategy A), and the other using no bridging therapy in the first 6 months (Strategy B). Reference case: T2 HCC patient with compensated cirrhosis. The benefit of sorafenib in delaying time to HCC progression was expressed as the hazard ratio (HR) and taken from recently published randomized trials. The endpoints considered were: survival benefit measured in quality-adjusted life days (QALDs), transplant probability, costs (C) in E, willingness to pay (WTP), and net health benefit (NHB), where NHB = survival benefit - C/WTP. The calculated WTP of sorafenib in Italy was 346 E per QALD. Probabilistic sensitivity analysis showed a median survival benefit of 94 QALDs (10% percentile = 38, 90% percentile = 210). In the base-case scenario (HR = 0.47, monthly dropout probability = 5%, median time to LT = 3 months), the gain in LT probability due to sorafenib was 5% and it increased proportionally with increasing median times to LT and decreasing HR. In the cost-benefit analysis, the incremental NHB of Strategy A versus Strategy B was 37 QALDs; it increased as sorafenib HR decreased and when median times to LT were shorter than 6 months, whereas for longer times it gradually dropped, particularly when Strategy B included effective locoregional treatments. Conclusion: Sorafenib neoadjuvant therapy is cost-effective by comparison with no therapy for T2-HCC patients waiting for LT, particularly for median times to LT under 6 months. (HEPATOLOGY 2009.)
Due to the disparity between organ availability and demand, dropout while on the waiting list (WL) has become the main predictor of survival for adult patients awaiting liver transplantation (LT) for malignant or nonmalignant (NM) chronic liver disease when an intention-to-treat analysis is applied.[1-3] Tumor progression before LT was the main reason for removing patients with hepatocellular carcinoma (HCC) meeting the Milan criteria (MC) from the WL, whereas for NM patients the main reason was the patient's death due to complicated cirrhosis.
A first way to contain this considerable dropout risk is to proportionally increase the probability of transplantation for patients with more severe liver disease by adopting specific prioritization policies. This is the primary strategy used by the US liver allocation system, which has adopted the model for endstage liver disease (MELD) in recent years to establish which HCC and non-HCC patients take priority for transplantation.
For HCC patients, the dropout risk might also be reduced by treating the tumor in order to slow its progression. Locoregional treatments, such as resection, ablation (percutaneous or laparoscopic), and transarterial chemoembolization (TACE) have been proposed as neoadjuvant therapies before LT.[7-9]
Although these procedures have a well-established efficacy in prolonging the survival of HCC patients, no studies strongly support and exactly measure their effectiveness in reducing the risk of dropout among HCC patient candidates for LT. This is the main reason why recent guidelines have prudently suggested that locoregional bridging therapies can be considered only if the median time on the WL exceeds 6 months.
A new systemic, molecularly targeted therapy, sorafenib, was recently tested in two large Phase 3 randomized clinical trials (RCTs), showing a significant efficacy in delaying tumor progression in patients with intermediate-advanced HCC. This effect was maintained in demographically different study populations, as demonstrated by the similar hazard ratios (HRs) in the two RCTs. Unlike the case of locoregional therapies, therefore, the efficacy of sorafenib in slowing tumor progression has been demonstrated and quantified with the highest level of scientific evidence. On the other hand, such a powerful antiangiogenic effect as that of sorafenib may interfere with vessel repair and thus give rise to a potentially higher risk of postsurgical complications, especially in the case of unscheduled measures such as transplantation. There are no data, however, to demonstrate and measure this potential toxicity of sorafenib in surgical patients.
In the present study, we hypothesized that by delaying tumor progression sorafenib could decrease dropout from the transplant WL and thus increase the number of patients able to be transplanted. We developed a Markov model to represent and quantify the potential cost-benefit ratio of sorafenib as a neoadjuvant therapy for HCC patients meeting the MC and awaiting LT. It has to be emphasized, however, that this model was designed while awaiting robust data on the safety of sorafenib.
To the best of our knowledge, this is the first study to analyze the neoadjuvant role of sorafenib in the context of LT for HCC patients. Monte Carlo probabilistic sensitivity analysis showed with a high level of confidence (Fig. 2) that neoadjuvant therapy with sorafenib before LT had a beneficial effect on survival with respect to a strategy without therapy. This central result of our study may be essentially explained by the positive impact of sorafenib on the transplant probability of HCC patients listed for LT (Fig. 2A). Our data confirmed previous findings concerning other Markov models of pre-LT bridging therapies. The results of the present study are very strong, however, because they are the first to be based on the findings of two RCTs. In fact, whereas locoregional therapies such as TACE, percutaneous ablation, or resection[7-9] have been recommended to reduce the dropout risk for HCC candidates awaiting LT, the scientific evidence to support and quantify their efficacy against tumor progression remains weak, especially as concerns the first 6 months on the WL. For the same reason, however, it is extremely important to emphasize that the results of this study cannot be used to promote sorafenib as a first-line neoadjuvant strategy for HCC patients awaiting LT. In fact, locoregional therapies have a well-known relevant impact on the survival of early HCC patients, so they are probably more powerful bridging strategies (when properly indicated). The basic assumption of this study is that we know the effect of sorafenib (HR) on time to progression, but the same cannot be said of conventional bridging therapies. On these grounds it is structurally impossible to use this model to draw any direct comparisons between sorafenib and locoregional therapies (resection, percutaneous ablation, or TACE) as bridging strategies. On the other hand, the relationship between sorafenib and standard locoregional therapies that emerges from this study may have two facets:
1 While awaiting more robust data on the efficacy of locoregional therapies in delaying tumor progression, our model may be used to assess the utility of sorafenib as a neoadjuvant therapy before LT, and also as a more general tool for studying the role of bridging therapies, taking sorafenib as a paradigm due to its well-known, evidence-based beneficial effect on tumor progression.
2 The no bridging therapy strategy before LT is now accepted by many centers around the world that assign HCC patients high priority for LT. In this light, the results of this study can easily be transposed to clinical situations where HCC patients have low median waiting times before they receive a transplant. In real life, however, when HCC patients have prospects of longer waiting times (generally longer than 6 months) they are given locoregional bridging therapies almost everywhere. That is why our model also included a specific sensitivity analysis considering the potential introduction of locoregional therapies in Strategy B patients when their median time on the WL exceeded 6 months (Fig. 7). We took the HCC patient with a T2 tumor for reference in this study because, among the candidates meeting the Milan criteria, these characteristics identify a higher dropout risk. On these grounds, many centers around the world attribute to T2 HCC patients 22 arbitrary points on the MELD score to contain the risk of tumor progression by reducing their waiting times. An excessive prioritization carries the risk of underestimating the biological aggressiveness of liver tumors, however, or of harming the competing NM transplant candidates at high risk of early WL mortality (MELD score >20). Similarly, the higher transplant probability due to sorafenib (Fig. 3A) risks being beneficial for treated HCC patients but harmful for patients denied the chance of a transplant. From a population utility perspective, therefore, the individual gain in life expectancy thanks to sorafenib neoadjuvant therapy could seem irrelevant. Nonetheless, the availability of predictive models that exactly correlate treatment efficacy with gain in transplant probability and survival benefit represents a potentially relevant tool for adapting the transplant priority of individual patients to a given local WL (relative proportions of HCC patients with T2 tumors and NM patients with high MELD scores).
We found a linear correlation between sorafenib HR and the transplant priority of HCC patients (Fig. 4); in this context, a second potential benefit of sorafenib neoadjuvant therapy might be its use as a corrective factor of HCC patient transplant priority in favor of NM patients with high MELD scores (Fig. 4). A minimum WL period maintaining a low risk of tumor progression (using sorafenib as bridging therapy), therefore, could be beneficial both for HCC and NM transplant candidates.
This favorable scenario changes dramatically when we consider the results of the cost-utility analysis, however. In fact, marginal cost-utility ratios increased in our model for longer times to LT or higher HR values (Fig. 4), which means that the costs of sorafenib therapy increased more rapidly than its utility according to these variables. Our study confirmed that traditional cost-effectiveness analyses cannot answer the underlying moral and policy issues raised by expensive treatments, such as molecular targeted therapies. To evaluate the threshold cost-utility of our strategy, therefore, we referred to the accepted cost-utility of sorafenib therapy in Italy, based on the results of the Sharp trial. This seemed reasonable in terms of an ethical concept of equity between patients with the same cancer, and given that the proportion of early-stage HCC candidates for LT is far lower than that of patients with intermediate or advanced tumors.
The WTP obtained was used to calculate the incremental NHB of Strategy A versus Strategy B. When this cost-utility reference was introduced, the sorafenib neoadjuvant strategy became cost-effective in almost all clinical scenarios tested in our Markov model (Figs. 4, 5). Moreover, the incremental NHB was mainly concentrated in the first 6 months on the WL, i.e., the period during which bridging therapies are currently not recommended. Our findings, thus, suggest that different or combined bridging therapies might be adopted, depending on the median time to LT in a given area. In fact, the incremental NHB of Strategy A dropped faster when locoregional therapies were included in Strategy B (Fig. 7) according to current guidelines. In such a clinical scenario, our specific sensitivity analysis supports the use of sorafenib mainly for HCC patients with median waiting times up to 12 months (Fig. 7).
There are some issues that might make our results under- or overestimate the actual benefit of sorafenib before LT. The actual benefit may be underestimated for two main reasons: (1) because a declining trend in the sorafenib HR on time to progression has been demonstrated from advanced to intermediate stage disease, the actual HR range of sorafenib for T2 tumors may plausibly be lower than was assumed in our model; and (2) sorafenib acts on the molecular pathways promoting tumor dedifferentiation and microscopic vascular invasion, but in this study we did not consider the potential benefit of sorafenib due to its effect on the tumor's biological aggressiveness before LT and thus on the post-LT risk of tumor recurrence.
The actual benefit of sorafenib before LT might be overestimated, on the other hand, because the antiangiogenic effect of sorafenib might have a negative effect on the outcome of surgery, although such a negative effect has never been demonstrated in the literature. This potentially toxic effect may also be more relevant in transplant candidates due to the unscheduled nature of LT (making it impossible to prudently suspend sorafenib some days before surgery) and to the presence of arterial, venous, and biliary anastomoses at risk of leakage or thrombosis. Only specifically designed clinical trials will provide definitive data on these issues. While awaiting such data, all the findings of this study must be considered with great caution and cannot be transferred to daily clinical practice.
In conclusion, sorafenib neoadjuvant therapy is cost-effective by comparison with no therapy for T2-HCC patients waiting for LT, particularly for median times to LT under 6 months. This Markov decision analysis, therefore, strongly supports the need for designing clinical trials in this complex field to comprehensively study the safety profile of sorafenib used before LT.
Survival Benefit and Transplant Probability.
To take into account the impact of variable uncertainties on the model results we performed a Monte Carlo probabilistic sensitivity analysis. According to this analysis, the median utility of Strategy A was 1,350 QALDs (10% percentile = 1,151, 90% percentile 1,434), whereas the median utility of Strategy B was 1,244 QALDs (10% percentile = 978, 90% percentile = 1,368). In Fig. 2 the distribution of incremental QALD gains of Strategy A versus Strategy B are represented: Strategy A showed a median survival benefit versus Strategy B of 94 QALDs (10% percentile = 38, 90% percentile = 210).
In the base-case analysis (Table 1), the strategy involving sorafenib treatment for HCC patients with a T2 tumor and compensated cirrhosis increased the probability of having a transplant by 5% with respect to no treatment (from 47% to 52%) if a time horizon of 10 years was considered. As a consequence, the same strategy reduced the individual risk of death by 5%, from 53% (for Strategy B) to 48% (for Strategy A). This lower mortality risk coincided with a gain of 89 QALDs for each patient treated.
In our utility-gain model, we performed one-way sensitivity analysis for all variables (Table 1). The variables most affecting the gain in LT probability and survival benefit were the HR (expressing the ability of sorafenib to delay tumor progression) and the median time to LT, as shown in Fig. 3.
As Fig. 3A clearly shows, higher median times to LT corresponded to a greater gain in transplant probability of Strategy A versus Strategy B, and this prognostic relationship had a clearly linear behavior. The angular coefficient of this relationship, on the other hand, was strongly influenced by the particular sorafenib HR. The median time to LT and sorafenib HR also had a considerable influence on survival benefit (Fig. 3B), but this effect was almost logarithmic rather than linear.
In Fig. 4 we evaluated the impact of the sorafenib HR on the transplant prioritization (expressed as the transplant probability ratio) of HCC patients on the WL. We found an almost linear relationship between the sorafenib HR on time to tumor progression and the ratio applied to transplant probability. According to this relationship, therefore, our model found that the effect of sorafenib on tumor progression can be used to proportionally reduce the priority of HCC patients without impairing their intention-to-treat survival rate.
Cost-Utility and Cost-Benefit Analyses.
In the base-case analysis, the marginal cost-utility ratio of Strategy A versus Strategy B was E 197 per QALD, whereas the incremental NHB (assuming WTP = E 346 per QALD) was 37 QALDs. Figure 5 shows that the marginal cost-utility ratios of Strategy A / Strategy B correlated strongly with the median times to LT and the sorafenib HR, but these ratios were below the calculated WTP value in the majority of cases. In particular, we found an inverse relationship between these two variables, i.e., the longer the median time to LT, the lower the HR threshold had to be in order to balance the utility against the costs.
One-way sensitivity analyses (Fig. 6) confirmed that, using the calculated WTP value, the incremental NHB of Strategy A versus Strategy B increased as the sorafenib HR decreased (Fig. 6A) and the threshold value of HR where Strategy A became harmful was 0.75.
The incremental NHB tended to rise for median times to LT below 6 months (Fig. 6B), whereas it dropped for longer waiting times and only became negative more than 24 months after starting the neoadjuvant therapy.
As expected, the incremental NHB of sorafenib dropped more rapidly when locoregional therapies were introduced after the first 6 months on the WL (Fig. 7). For example, sorafenib maintained a positive NHB up to 12 months on the WL only when the impact (HR) of conventional therapies on the dropout rate was higher than 0.5 (Fig. 7).