Background: Survival after liver transplantation for early hepatocellular carcinoma (HCC) is worsened by the increasing dropout rate while waiting for a donor.
Aims: To assess the cost effectiveness of adjuvant therapy while waiting for liver transplantation in HCC patients.
Method: Using a Markov model, a hypothetical cohort of cirrhotic patients with early HCC was considered for: (1) adjuvant treatment—resection was limited to Child-Pugh's A patients with single tumours, and percutaneous treatment was considered for Child-Pugh's A and B patients with single tumours unsuitable for resection or with up to three nodules < 3 cm; and (2) standard management. Length of waiting time ranged from six to 24 months.
Results: Surgical resection increased the transplantation rate (>10%) and provided gains in life expectancy of 4.8–6.1 months with an acceptable cost ($40 000/ year of life gained) for waiting lists ≥1 year whereas it was not cost effective ($74 000/life of year gained) for shorter waiting times or high dropout rate scenarios. Percutaneous treatment increased life expectancy by 5.2–6.7 months with a marginal cost of approximately $20 000/year of life gained in all cases, remaining cost effective for all waiting times.
Conclusions: Adjuvant therapies for HCC while waiting for liver transplantation provide moderate gains in life expectancy and are cost effective for waiting lists of one year or more. For shorter waiting times, only percutaneous treatment confers a relevant survival advantage.
Statistics from Altmetric.com
- cost effectiveness analysis
- hepatocellular carcinoma
- liver transplantation
- adjuvant treatment
- surgical resection
- percutaneous treatment
- CEA, cost effectiveness analysis
- HCC, hepatocellular carcinoma
- OLT, orthotopic liver transplantation
- RCT, randomised controlled trial
- PEI, percutaneous ethanol injection
- RF, radiofrequency
- LE, life expectancy
- MCYLS, marginal cost per year of life saved
Surgical resection and orthotopic liver transplantation (OLT) are considered the first treatment options for hepatocellular carcinoma (HCC) although the best treatment strategy has not been established.1 Resection achieves good results in only a minority of patients but the long term outcome is worsened by a high recurrence rate (three year recurrence rate of 50%).2–5 In contrast, liver transplantation offers excellent results (70% survival with less than 20% recurrence at five years) if the indication is restricted to patients with single tumours ≤5 cm or three nodules <3 cm.4,6–8 These values have prompted most hepatologists to favour OLT as firstline treatment. However, the shortage of donors and the increasing demand of organs have lengthened waiting times to more than six or 12 months in Europe and USA, respectively.9,10 This delay can allow the tumour to grow to stages that contraindicate OLT. We have recently shown that a waiting time longer than six months is associated with a 23% rate of dropout from the waiting list,4 and values of up to 30–40% have been reported both in Europe and USA,9,11 which may reach 50% when expanded selection criteria are applied.12 This sharply worsens the outcome when assessing OLT results on an intention to treat basis.
Several adjuvant antitumoral therapies have been administered to patients on the waiting list for liver transplantation to reduce tumour growth. The most common treatment is chemoembolisation, a palliative option that may achieve extensive tumour necrosis.13–18 However, there are no prospective studies showing that it is effective in reducing the dropout rate or in modifying the outcome of these patients. Surprisingly, the usefulness of more radical options, such as resection or percutaneous ablation, that may provide the majority of patients with complete tumoral responses, are seldom used as adjuvant treatments prior to OLT. Randomised controlled trials (RCTs) assessing the benefits of any of these therapeutic approaches are unlikely to be performed because they would be costly and require a very large sample size. Therefore, we designed a decision analysis to assess the impact of treating HCC patients on the waiting list for OLT with the probability of being transplanted, overall survival, and cost effectiveness. Adjuvant therapies included surgical resection and percutaneous ablation as they provide a more extensive tumour load reduction than chemoembolisation. The model takes into account the heterogeneity of HCC candidates according to their tumour stage/liver functional impairment and the variable duration of the waiting times throughout the world. This stratification allows the reader to identify into which scenario their patient fits and thus use the obtained outputs for rational clinical decision making and also to correctly allocate health care resources in the management of this complex disease.
PATIENTS AND METHODS
Design of the study and Markov multistate transition model
We analysed the cost effectiveness of applying resection or percutaneous ethanol injection (PEI) versus standard management in cirrhotic patients with early HCC on their inclusion on the waiting list for OLT. A Markov multistate transition model was developed to estimate the impact of treatment on the probability of being transplanted, overall outcome, and cost effectiveness. A commercially available software product (Data 3.5;Tree Age Software, 1998, Williamstown, Massachusetts, USA) was used to generate the model and to tabulate all costs accrued in each group. The study was performed according to the recommendations of the Panel of Cost-effectiveness in Health and Medicine of the US, and following their indications in reporting cost effectiveness analysis (CEA).19,20
Target population and adjuvant treatments
The study focused on HCC candidates for OLT as the primary treatment option. This target population included patients with early HCC (single tumour≤5 cm or three nodules ≤3 cm) on an otherwise non-advanced liver cirrhosis (Child-Pugh's A and B classes) without associated diseases. We did not consider Child-Pugh's C class cirrhotic patients with coincidental tumours because their dropout rate is mostly related to life threatening decompensation of cirrhosis rather than tumour progression.
Our model considered two reference cases of HCC candidates for OLT which were stratified according to tumour stage and liver function impairment.
Reference case No 2 (group 2). Patients with either a single HCC ≤5 cm and Child-Pugh's B or with three tumours ≤3 cm. As these patients are poor candidates for resection due to their high risk of postoperative decompensation with associated death, they were considered for PEI.
PEI is a highly effective option that provides complete necrosis in 70–80% of cases in tumours smaller than 3 cm.21–24 This treatment has been widely assessed throughout the world over the last decade, with a safety profile and a low risk of tumour seeding along the needle tract (0.6% cases per patient; 0.08% per session).25 Percutaneous radiofrequency (RF) ablation was not considered in the model because reliable data on antitumoral usefulness and cost have not been completely established. In addition, we have recently reported a noteworthy rate of needle track seeding related to this procedure, thus increasing the concerns regarding its application prior to OLT.26 Chemoembolisation was also not considered because of its lower antineoplastic effect if compared with surgery or percutaneous ablation, and because of the lack of prospective data showing a beneficial effect. Patients in the control arm were considered to receive no specific antineoplastic treatment, which represents the standard management of patients on the waiting list of our transplantation programme to date.4,7
Decision tree and states of health
A diagram of the event pathway is depicted in fig 1. Analysis of the state of health was performed in three month cycles starting from inclusion of patients onto the waiting list until a time period of 10 years after OLT. In summary, patients with early HCC were allocated to receive adjuvant treatment according to their risk group or standard management on inclusion onto the waiting list. Patients of reference case No 1 underwent resection and entered the waiting list state. Three transition states were considered thereafter: (a) receiving OLT and entering the post-transplantation state; (b) developing any contraindication for OLT and entering the dropout state; or (c) dying of tumour progression or complications of cirrhosis. Similarly, patients of reference case No 2 received PEI treatment and entered the waiting list state where the same three transition states were considered. Finally, patients in the non-treatment arm directly entered the waiting list state without antitumoral treatment where the three transitional states were again considered. Patients effectively transplanted entered the post-transplantation state until death.
Summary of data and assumptions
Survival and recurrence on the waiting list
Data and assumptions used in the decision analysis for the reference cases are depicted in table 1. In Child-Pugh's A patients with a single HCC, the probability of survival after resection was considered to be 93% and 75% at one and two years, respectively. Recurrence rate at two years was 30%.3–5 Complete response after PEI was related to tumour size, and ranged from 80% for single tumours <3 cm, 50% for tumours of 3–5 cm, to 40% for patients with up to three nodules smaller than 3 cm.21–24 Survival outcome after PEI varied according to achievement of a complete and maintained response, and ranged from 62% to 72% at two years. The two year recurrence rate—new tumour development—was considered to be 25%.21–24
Survival data for the non-treatment arm were derived from prospective studies on the prognosis of patients diagnosed with asymptomatic HCC without vascular invasion or extrahepatic spread: one and two year survival rates of 80% and 65%, respectively.27 Mortality of these patients is due to HCC itself, and eventually to complications of cirrhosis. Age related mortality in the periods of time considered was negligible.
Data on dropout rate of HCC candidates for OLT are rarely reported. We reported a 23% dropout rate on a six month median waiting list4 whereas other authors have described even higher incidences of 30–40% while waiting for more than one year.9,11 In our model, exclusions from the waiting list included both death or OLT, contraindications derived from tumour progression (vascular invasion, lymph node involvement, or metastases), or liver functional impairment (that is, progressive hepatorenal syndrome).
The probabilities for dropout rate assumed in the model are summarised in table 1. For the treated groups, values were derived from probabilities of survival and recurrence after resection and PEI. Patients who presented recurrences were considered for exclusion when they developed major contraindications. Data for the non-treatment arm were derived from the reported probability of tumour progression, vascular invasion, and extrahepatic spread, taking into account that HCC progression does not always prompt exclusion from the list. Once exclusion of the list had occurred, median survival was modelled to be less than one year.27
Outcomes of patients effectively transplanted were obtained from the best results of groups applying restrictive selection criteria. In the reference case, five year survival after OLT and recurrence rate were 70% and 10%, respectively.4,6–8
As previously described, for the two reference cases the probabilities of survival during the waiting list (with or without treatment), dropout rate, and survival after transplantation are shown in table 1. The sensitivity analysis was applied to test different degrees of estimates of the various probabilities used.19,20 Establishment of two strata represents a type of sensitivity analysis of the scenarios that are faced in the clinical setting. None the less, additional sensitivity analyses were performed varying the main probabilities of outcomes over clinically relevant ranges.
Variables modelled in the sensitivity analysis included the dropout rate and outcome after OLT.
(1) Variations in the dropout rate. (a) Best scenario: treatment achieves a relative reduction of 20% in the dropout rate from the reference case of the treatment arm. (b) Worst scenario: treatment achieves a relative increase of 20% in the dropout rate from the reference case of the treatment arm.
(2)Variations in outcome of transplantation. Survival after OLT was modelled to include a five year survival rate as low as 50% as the worst scenario.
Economic costs were assessed from the payer's perspective, and included direct costs of procedures and treatments, expressed in 1999 US dollars, and were obtained from the current payments within the Spanish Health Care System (table 2). These cover the cost of salaries for physicians, surgeons, and support personnel, equipment, supplies, and organisational costs. Direct non-medical costs incurred by patients and their families, as well as indirect costs, were not assumed in the analysis. Future costs and benefits were discounted at a baseline rate of 3%.19,20
Estimates of effectiveness and cost effectiveness
We considered a 10% increase in the probability of being transplanted as clinically relevant. Effectiveness was measured as net gain in life expectancy (LE), and according to recent proposals we defined an increase of at least three months in LE as clinically relevant.28 Although quality of life is an important outcome in calculating cost effectiveness, reliable data on HCC patients waiting for a liver donor are not available and thus this analysis was not addressed.29 The marginal cost of the treatment strategy was divided by its incremental benefit, as measured by gain in LE. This results in a marginal cost per years of life saved (MCYLS) and we have applied the conventional threshold of $50 000 per year of life saved as the accepted cut off value to consider an intervention “cost effective”.30–34
Probability of being transplanted and overall survival
The probability of being transplanted and the overall survival are depicted in figs 2 and 3. Surgical resection increased the transplantation rate from 3.7% to 10.7% for waiting lists of six and 24 months, respectively. The seven year probability of survival increased for patients undergoing resection compared with standard management, and was higher than 10% for waiting lists exceeding one year. Percutaneous treatments increased the probability of being transplanted and the seven year survival in all waiting times. These benefits became clinically relevant when the waiting time exceeded one year.
The results of the CEA are shown in figs 4 and 5, and table 3. Resection provided a clinically relevant net gain in LE ranging from 4.8 to 6.1 months if waiting times exceeded six months. The CEA shows that the MCYLS was less than $40 000 for lists of 12–24 months. Ethanol injection achieved a clear gain in LE, ranging from 5.2 to 6.7 months, increasing according to the length of the waiting list. Its cost effectiveness ratio was less than $23 000/year of life saved (fig 5). Undiscounted results did not statistically differ from those discounted at 3%.
The results of the sensitivity analysis are shown in table 4. Our model was most sensitive to variations in the dropout rate and survival after OLT. When assuming the highest benefit as a result of treatment (best scenario: relative decrease of 20% in dropout rate), gains in LE were relevant (resection 3.3–8.3 months; PEI 6.8–9.1 months), with a cost effectiveness ratio always below $60 000/year of life saved. When considering scenarios of poor outcome after resection (worst scenario: relative increase of 20% in dropout rate), this treatment provided a minimal benefit in LE, with a marginal cost ranging from $135 600/year of life saved (six month waiting list) to $43 650/year of life saved (24 month waiting list). Conversely, even in this worst scenario, PEI offered gains in LE always exceeding three months, with a marginal cost per year of life of less than $32 000.
Assuming a 50% five year survival rate after OLT (worst scenario), the estimates of LE decreased in all cases. For this scenario, PEI retained a cost effectiveness ratio for all waiting times but resection was cost effective only for waiting times exceeding two years ($43 074/MCLYS).
The excellent results of liver transplantation for early HCC are curtailed by the increasing dropout rate while waiting for a donor, thus worsening outcome when analysed on an intention to treat basis.1,4 Several strategies have been postulated to decrease the impact of exclusions. Living donor liver transplantation, domino and split liver transplantation, as well as use of marginal livers are policies currently applied in some transplant units. A decision analysis has recently shown that living donor liver transplantation compared with OLT is cost effective for early HCC for waiting times exceeding seven months.35 However, implementation of these strategies is complex, and will probably be restricted to leading transplant units. This has led most centres to administer antitumoral treatments on entering HCC patients onto waiting lists.13–18 The benefits of this policy are unknown as prospective RCTs in the field are lacking and they are seen as almost unfeasible due to the cost, heterogeneity, and complexity of the medical interventions. This uncertainty prompted us to conduct a decision analysis to address the clinical benefits and cost effectiveness of adjuvant treatment.
The model applies the best curative therapies available for the selected strata of HCC candidates for OLT and considers a waiting time between six and 24 months. This strategy predicts moderate gains in LE in almost all cases, and the gain remains below the accepted cost effectiveness ratio ($50 000/per year of life saved).30–34 In fact, it compares favourably with accepted medical interventions, such as implantable cardioverter-defibrillator for coronary heart disease ($26 000–40 000 per year of life) or haemodialysis ($42 000/quality adjusted life year saved).32–34 Surgical resection was cost effective while waiting for at least one year. After this time, surgery increased the transplantation rate, with moderate gains in LE and an acceptable cost effectiveness ratio. In contrast, LE gains and cost effectiveness ratio were less favourable ($74 000 per year life gained) with shorter waiting times. There may be some concerns on the applicability of this invasive therapy while on the waiting list. As an alternative, primary liver resection and “salvage OLT” (performed when recurrence or decompensation after resection occurs) have been postulated to save organs.36 We have applied this policy over the last decade, where resection was considered as the firstline option, but unfortunately less than 10% of candidates benefited.1
Surgery is feasible for patients with preserved liver function and single tumours but it is too risky in subjects with impaired hepatic function and/or multiple HCC sites.2–5 Their high morbidity and mortality impede any benefit of surgery. Accordingly, these patients were modelled to receive percutaneous treatment, which was identified as cost effective in all scenarios. It could be argued that PEI is not the sole effective therapy for non-surgical HCC and that other widely used alternatives such as RF thermal ablation,26,37 chemoembolisation,13–18 or even chemotherapy38 could be modelled. Ethanol injection is a safe, cheap, and effective treatment that achieves a 40–80% response for patients with small HCC21–24 and the available studies provide robust information for decision analysis. RF thermal ablation as a primary treatment for HCC may have similar efficacy.37 Compared with PEI, its increased cost and the impact of complications may balance the benefits of reduced hospital stay. However, we have recently reported a 10% rate of tumour seeding after RF thermal ablation associated with subcapsular location or an aggressive tumoral pattern. Therefore, although there are no studies on its benefit when applied during the waiting list, we discourage this procedure prior to OLT.26 Finally, chemoembolisation is the standard antitumoral treatment for HCC prior to OLT in most transplant programs. Its antitumoral effect, even when using high dose chemotherapy, is less than that of PEI, and despite the fact that some authors have suggested benefits in patients with a favourable response to therapy (downstaging),13 there are no RCTs showing the benefit of this strategy. In addition, studies assessing this point have not identified a difference in tumour recurrence and survival attributable to therapy.
The sensitivity analysis disclosed that ethanol injection was beneficial and cost effective in all ranges used but the benefits of resection decreased when varying the assumptions of the model. In the worst scenarios for dropout rate and survival after OLT (five year survival 50%), resection provided poor gains in LE regardless of the length of the waiting list, with an expensive cost effectiveness ratio. These controversial benefits may be representative of transplantation centres achieving poor results after surgery as a consequence of both limited technical skills and expertise of the group. In support of this, it has recently been reported that OLT centres in the USA that perform less than 20 transplantations per year have mortality rates higher than those at larger centres.39 Thus decision makers should consider these data to warrant implementation of economic resources for the most efficient groups. As done for most decision analyses, we used local costs for cost effectiveness estimations. However, it is clear that health costs in Southern Europe are lower than in Northern countries or even in the USA. Accordingly, in areas where local economics are largely higher, the cost effectiveness of some scenarios may be lost.
The current analysis was limited by the scarcity of data on exclusions during the waiting list. For untreated patients we have assumed a dropout rate ranging from 15% to 36% at one year of waiting time, for the two reference cases. In the sensitivity analysis, these values varied by 20% up/down in the best/worst scenarios. These assumptions were derived from our reported data on exclusions, and are in accordance with recent 1999 data from the UNOS, where patients removed from the list due to death or other reasons in the USA ranged from 26% to 41%.9
In summary, our study indicates that adjuvant treatment on entering HCC patients onto a waiting list for OLT is cost effective and recommended in almost all scenarios. Patients with well preserved liver function and a single HCC, waiting for at least one year, may benefit from surgical resection but for shorter waiting times the economic investment is controversial. In non-surgical patients, percutaneous treatments are cost effective in all waiting times and thus their application is warranted.
Supported by grant SAF 004-98 from the Comisión Interdepartamental de Ciencia y Tecnología, and a grant from the Fundació La Marató TV3, 1999.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.