Pioglitazone and vitamin E for nonalcoholic steatohepatitis: A cost utility analysis


  • Suzanne E. Mahady,

    Corresponding author
    1. School of Public Health, University of Sydney, Sydney, Australia
    2. Westmead Millennium Institute, Storr Liver Unit, Westmead, NSW, Australia
    • Storr Liver Unit, Westmead Hospital, NSW, Australia

    Search for more papers by this author
    • fax: 61 2 9635 7582

  • Germaine Wong,

    1. School of Public Health, University of Sydney, Sydney, Australia
    2. Centre for Transplant and Renal Research, Westmead, NSW, Australia
    3. Centre for Kidney Research, Kids Research Institute, Children's Hospital at Westmead, Westmead, NSW, Australia
    Search for more papers by this author
  • Jonathan C. Craig,

    1. School of Public Health, University of Sydney, Sydney, Australia
    2. Centre for Kidney Research, Kids Research Institute, Children's Hospital at Westmead, Westmead, NSW, Australia
    Search for more papers by this author
  • Jacob George

    1. Westmead Millennium Institute, Storr Liver Unit, Westmead, NSW, Australia
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.


Nonalcoholic steatohepatitis (NASH) is the commonest liver disease in developed countries. However, there are no current data on the cost-effectiveness of therapeutic options such as lifestyle modification, pioglitazone, or vitamin E. We undertook a cost utility analysis to compare these strategies. Using a third-party payer perspective, a deterministic Markov model was developed to compare costs and health benefits of lifestyle modification alone or with pioglitazone or vitamin E in a cohort of patients aged 50 years with biopsy-proven NASH and fibrosis level 3 or greater. We assumed an annual cycle length over a lifetime horizon. Probability and utility estimates were derived from a systematic literature review, and uncertainties in parameter estimates were tested using one- and two-way sensitivity analyses. Our outcome measure was the incremental cost-effectiveness ratio (ICER), with $A50,000 or less considered cost-effective. In comparison with lifestyle modification alone, treatment with either pioglitazone or vitamin E in addition to lifestyle modification was cost-effective, with incremental cost-effectiveness ratios of $A2748 and $A8475 per quality-adjusted life year (QALY) gained, respectively. In a direct comparison, pioglitazone was more cost-effective than vitamin E (ICER $A2,056/QALY gained). Sensitivity analyses indicated that pioglitazone was not cost-effective if either the total drug cost was greater than $A16,000 per annum, or the annual probability of developing cirrhosis in advanced fibrosis was less than 2%. Conclusion: Our modeled analyses suggest that in patients with advanced fibrosis due to NASH, pharmacological treatment in addition to standard lifestyle modification is likely to be cost-effective. (HEPATOLOGY 2012;56:2172–2179)

Nonalcoholic fatty liver disease (NAFLD) is the commonest cause of abnormal liver tests in developed countries, accounting for 20% of primary care presentations and displacing traditional causes such as viral hepatitis, which now account for less than 1%.1 NAFLD and its progressive form, nonalcoholic steatohepatitis (NASH), are strongly associated with the global obesity epidemic.2 Although the annual cost of obesity-related care is estimated at $147 billion in the United States3 and $21 billion in Australia,4 the healthcare costs associated with NAFLD and NASH are unknown but likely to be substantial, as NASH may progress to cirrhosis, decompensated liver disease, and hepatocellular carcinoma (HCC)5-9; furthermore, NASH is predicted to be the leading cause of liver transplantation in the U.S. by 2020.10

Despite these data, there remains no widely accepted therapy. Lifestyle modification remains the standard of care but there is little evidence that this improves liver fibrosis,11 the recommended endpoint for trials in NASH.12 In contrast, trials and meta-analyses of pharmacological therapy using thiazolidinediones or vitamin E as add-on therapy indicate reversal of steatohepatitis13-17 and improvement in fibrosis.17, 18 Currently, these drugs are recommended for patients with advanced disease who fail lifestyle modification19 but the incremental costs and benefits have not been studied in a formal economic evaluation.

Of the different types of economic evaluations, cost utility analysis is the methodological approach preferred by health economists20, 21 due to its outcome measure, the quality-adjusted life year (QALY). The QALY takes into account the duration and quality of life gained and, because it is not disease-specific, is a universal currency that allows policy makers to compare the value of interventions across different health conditions.

The aim of our study was to estimate the incremental healthcare costs and benefits of using either pioglitazone or vitamin E in addition to lifestyle modification in patients with recently diagnosed NASH and advanced fibrosis (F3 or greater). Second, we aimed to identify the key factors that drive cost-effectiveness and therefore prioritize areas for future research.

Materials and Methods

Markov Model

Using a third-party payer perspective, a deterministic decision analytical Markov model (TreeAge Software, Williamstown, MA) was developed using a lifetime horizon. In this model, the average patient was age 50 with elevated aminotransferases, biopsy-proven NASH with fibrosis level 3/4, and no prior treatment. We first structured the model to simulate the natural history of disease progression in patients with NASH (Fig. 1). In each cycle, patients may remain well or develop compensated cirrhosis or decompensated cirrhosis, with a proportion eligible for liver transplantation. A proportion may also develop HCC and will enter cycles governing various treatment strategies for HCC. We assumed an annual cycle length and the model terminated when all patients died. This lifetime horizon was chosen to reflect the often slowly progressive nature of liver disease due to NASH. Half-cycle corrections were included for all parameters. Reporting was performed according to peer-reviewed guidelines for economic evaluations.21, 22

Figure 1.

Simplified Markov model.

Competing Strategies in the Model and Analyses

The model had three arms that were compared in the base case analyses: lifestyle modification, pioglitazone in addition to lifestyle modification, and vitamin E in addition to lifestyle modification. These two drugs were chosen as they are considered the principle pharmacologic options for patients with NASH. Patients in the lifestyle modification arm received management consistent with international guidelines,23 including hepatologist review with diet and exercise recommendations twice per year and annual dietitian consultation. Dietary recommendations were consistent with the principles of healthy eating for patients with the metabolic syndrome including reduction in saturated fats and refined carbohydrate and increased intake of lean protein, complex carbohydrates, and adequate dietary fiber. All patients with cirrhosis underwent 6-monthly HCC screening. Patients in the pioglitazone arm received advice on lifestyle modification in addition to a daily oral dose of pioglitazone (30 mg). Pioglitazone was chosen as the thiazolidinedione of choice, as this is likely to be preferred over rosiglitazone given the latter's association with adverse cardiovascular effects24 and withdrawal from the market in some European countries. These adverse cardiovascular effects have not been reported for pioglitazone. In the vitamin E arm, patients received a daily oral dose of 800 IU, the dose used in the largest published trial of vitamin E therapy (13), in addition to lifestyle advice. Both drugs were stopped if patients developed decompensated liver disease, as they have not been tested in this stage.

Input Parameter Estimates for the Model

Clinical Data.

Our base case model incorporated a wide range of probability estimates, as shown in Table 1. These estimates were derived from a recently published systematic review, other published literature, and supplemented with data from the largest international database of NAFLD patients with biopsy-proven F3 or 4 disease.25 Individual patient data from this database was used to calculate time-specific probabilities for outcomes such as decompensation, which are nonlinear, and this method is therefore more likely to reflect clinical scenarios than extrapolated, linear estimates from short-term follow-up studies.

Table 1. Annual Transition Probabilities
Health StateBase Estimate (Range)Reference
Advanced fibrosis (F3/4)  
Probability of remaining well0.91 (0.83-0.92)25,28,29
Probability of developing cirrhosis0.04 (0.02-0.06)25,30
Probability of decompensating0.013 (0.013-0.11)25, 30
Probability of developing hepatoma0.00425
Probability of all cause mortality0.004 (0.004-0.02)25
Compensated cirrhosis  
Probability of remaining compensated0.82 (0.67-0.98)29
Probability of decompensating0.06 (0.04-0.16)29, 31
Probability of developing hepatoma0.03 (0.007-0.05)7,8,31
Probability of all cause mortality0.04 (0.02-0.04)29, 31
Decompensated liver disease  
Probability of stable decompensation0.7625
Probability of developing hepatoma0.03 (0.007-0.05)7,8,31,32
Probability of liver transplant0.05 (0.05-0.25)35
Probability of all cause mortality0.16 (0.15-0.38)8
Post liver transplant  
Probability of survival at 1 year0.88 (0.78-0.99)33,34
Probability of resection0.18 (0.1-0.2)35
Probability of liver transplant0.2 (0.1-0.3)33,35
Probability of locoregional therapy0.2336
Probability of sorafenib0.22Author assumption
Probability of supportive care0.17 (0.15-0.24)35,36
Pharmacological treatments  
Histological improvement (pioglitazone)1.38 (1.01-1.89)13,18
Histological improvement (vitamin E)1.35 (0.87-2.09)13

Probability estimates for fibrosis progression were calculated using the rate from the largest published cohort of NAFLD patients with serial biopsies26 and then applying a relative risk for histological improvement with pioglitazone derived from a meta-analysis of randomized trials,18 where pioglitazone was used as add-on therapy to standard lifestyle advice. A relative risk for histological improvement for vitamin E was determined from the largest randomized trial of vitamin E therapy,13 which was considered the highest level of evidence for vitamin E efficacy due to the rigorous methodology employed in this trial. Sensitivity analyses were performed ranging from no improvement with drug therapy to the best-case scenario as suggested by the upper limit of confidence intervals from the above studies. We also included an increased relative risk of mortality with use of high-dose vitamin E.27

Costs Data.

Our cost data are reported in 2010 Australian dollars ($A) (Table 2). We included the direct healthcare costs of caring for patients with NASH, including initial visits, screening to exclude other causes of chronic liver disease, and coexistent features of the metabolic syndrome including Type 2 diabetes and dyslipidemia. We included costs of HCC screening (6-monthly alpha-fetoprotein and liver ultrasound). Costs of inpatient and outpatient care for liver decompensation and liver transplantation were based on funding as described in the Australian Medicare Benefits Schedule,37 Pharmaceutical Benefits Scheme,38 and the National Hospital Cost Data Collection.39 Costs of palliative care for terminal HCC were based on published literature.40, 41 Where required, costs were inflated to 2010 using a national inflation index.42 All foreign currencies were converted to the 2010 Australian dollar using the Purchasing Power Parity conversion factors. The annual cost of pioglitazone treatment was based on the cost when prescribed under the Australian Pharmaceutical Benefits Scheme (around $A1,000), which is similar to the cost in the U.S. (around $US600, equivalent to $A600) and the United Kingdom (£437, equivalent to $A700). The cost of vitamin E was based on the formulation used in the largest trial (Nature Made Pharmavite, Mission Hill, CA).

Table 2. Health Care Costs ($A, Australian Dollars)
Annual Clinical Care CostsBase Case Estimate (Range)Reference
Routine care and pioglitazone1696 (1435-1865)37,38
Routine care and vitamin E675 (609-743)37,38
Compensated cirrhosis and pioglitazone1751 (1699-1803)37,38
Compensated cirrhosis and vitamin E731 (658-804)37,38
Decompensated cirrhosis13,990 (11475-16512)37,40
Hepatoma18715 (13971-18715)37,40
Liver transplant (1st year)125678 (103056-148300)37
Resection of hepatoma27970 (27691-28249)43
Locoregional therapy1155 (1145-1868)43
Palliative care7497 (5763-9732)40,41
Itemized costsCostReference
Consultation costs  
Specialist consultation (initial)145.2037
Specialist consultation (follow up)72.6537
Primary care doctor consultation34.9037
Pathology costs  
Full blood count17.0537
Liver function tests17.8037
Oral glucose tolerance test19.1037
Hepatitis C antibody17.2037
Hepatitis B surface antigen17.2037
Anti nuclear antibody24.6037
Drugs (annual cost)  
Vitamin E (800IU)7544
Radiology and Procedures  
Computed tomography scan360.0037
Magnetic Resonance Imaging (MRI) scan403.0037
Upper gastrointestinal endoscopy170.4037
Radiofrequency ablation786.1537
Transarterial chemoembolization323.4037
Liver biopsy167.8537

Health-related Quality of Life Data.

No prior health-related quality of life studies have been performed in patients with NASH-associated chronic liver disease, cirrhosis, or hepatic decompensation. We therefore used utilities from primary studies45–48 and a systematic review49 derived from other causes of chronic liver disease and tested in sensitivity analysis over a wide range. Given that cirrhosis, decompensated liver disease, and HCC represent a common pathway of chronic liver disease, we assumed that the decrement in quality of life associated with these conditions is similar regardless of the initial cause. It was considered important to include a decrement in quality of life for weight gain associated with pioglitazone treatment, as this is a clinically significant side effect. However, there are no published utility data for this in people with NASH. Therefore, utility values derived from weight gain in patients treated with antidiabetic agents50 were included and the effect tested in sensitivity analysis. Utilities for health states are shown in Table 3.

Table 3. Health-Related Quality Of Life
Health StateQuality Of Life Estimate (Range)Reference
  • *

    No health-related quality of life studies performed.

Well with F3/4 fibrosis0.92 (0.65-0.95)45,48,49
Compensated cirrhosis0.82 (0.65-0.89)45, 46, 48, 49
Decompensated cirrhosis0.6 (0.46-0.81)45, 46, 48, 49
Hepatoma0.73 (0.5-0.8)45
Liver transplant (1st year)0.69 (0.62-0.86)46-48
Surgical resection (after 1st month)0.73 (0.62-0.84)51
Locoregional treatment0.79 (0.5-0.8)52
Sorafenib0.7 (0.4-0.8)Authors assumption*
Palliative care0.4 (0.2-0.6)Authors assumption*


The outcomes of the model's three strategies were measured in costs ($A), benefits (life years saved [LYS], a measure of the number of deaths averted), and quality-adjusted life years gained (QALYs). The cost-effectiveness of each strategy was assessed by calculating its incremental cost-effectiveness ratio (ICER) according to the following formula:

equation image

An ICER of less than $A50,000 is considered cost-effective in an Australian healthcare setting.53 Discounting was performed in accordance with standard Australian guidelines at 5% with a range of 3%-8%54 to incorporate the range used internationally.

Sensitivity Analyses

To assess the effect of variation in individual probabilities and costs on the ICER, one- and two-way sensitivity analyses were performed across published ranges or 95% confidence intervals. For costs data, a clinically relevant range was tested, and where unavailable, a 10% variation for upper and lower limits was used. Probabilistic sensitivity analysis was not performed due to the absolute paucity of published data on the distributions for relevant parameters.

Model Assumptions

Our model included the following assumptions: that people with NASH who developed decompensated liver disease would cease pharmacological treatment, that health-related quality of life was similar in endstage liver disease irrespective of the initial cause, and that histological improvement is a valid surrogate for longer-term clinical outcomes. Surrogate markers have well-documented limitations55; however, this is the most commonly employed endpoint in NASH trials.

Funding Source

There was no external funding source for this study.


HCC, hepatocellular carcinoma; ICER, incremental cost-effectiveness ratio; LYS, life years saved; NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis; QALY, quality-adjusted life years.


The results of the base case analysis for the three competing strategies showing total costs, benefits (LYS) and cost per QALY gained are shown in Table 4. Lifestyle modification as a baseline strategy cost $46,000 for the cohort with a total average benefit of 6.2 LYS. Pioglitazone in addition to lifestyle modification was more costly than lifestyle modification alone, but delivered greater health benefits and was cost-effective, with an incremental benefit of an additional 4.7 LYS and an ICER of $2748/QALY gained. Vitamin E in addition to lifestyle modification was also cost-effective, with additional benefit of 0.6 LYS, resulting in an ICER of $8475/QALY gained. A direct comparison of the two pharmacological strategies indicated that pioglitazone was more cost-effective, with an ICER of $2056/QALY gained compared with vitamin E.

Table 4. Base Case Analysis of Costs and Benefits of Pioglitazone or Vitamin E and Lifestyle Modification, Compared with Lifestyle Modification Alone
StrategyCostIncremental CostBenefits (LYS)Incremental Benefits (LYS)ICER
Vitamin E$51,000$5,0006.850.59$8475/QALY

One-Way Sensitivity Analyses

The results of a one-way sensitivity analysis that tested for influential variables in the pioglitazone strategy are shown in Fig. 2. The vertical line represents the ICER for the base case estimate. The arrows show the direction of movement of the ICER across the range that the variable was tested. There were four key variables (represented as horizontal bars) that had a meaningful effect on the ICER. For example, if the annual probability of death in decompensated NASH was 15%, the ICER was more than $7000/QALY gained; however, if the probability was 38%, the ICER was less than $1000, indicating pioglitazone was more cost-effective when the risk of death in decompensated disease increased. Similarly, as the benefit of pioglitazone in preventing progression to cirrhosis increased, the cost-benefit ratio improved. A one-way sensitivity analysis testing variables in the vitamin E strategy indicated that the ability of vitamin E to prevent decompensation, and the probability of death due to decompensated disease, were the most influential variables. Nevertheless, the ICER remained cost-effective across the ranges tested for these probabilities, likely reflecting the cheap cost of vitamin E. The model was tested over a discounting rate that varied from 3%-8%. At the highest rate of discounting (8%), the ICER for both strategies became more cost-effective (ICERs of $945/QALY gained for pioglitazone and $5475/QALY gained for vitamin E).

Figure 2.

One-way sensitivity analysis for pioglitazone strategy.

Two-Way Sensitivity Analyses

Two-way sensitivity analyses were performed to assess the change in the ICER when two variables were varied simultaneously, in order to find thresholds at which the drugs were no longer cost-effective. Two-way sensitivity analyses in the pioglitazone strategy indicated that if the likelihood of developing cirrhosis for people with advanced fibrosis was less than 2% per year, then lifestyle modification was the more cost-effective option. At probabilities equal or greater than 2%, pioglitazone was more cost-effective. As expected, when the effectiveness of pioglitazone reduced, the ICER became less favorable (Fig. 3). Two-way sensitivity analyses in the vitamin E model indicated that it remained cost-effective irrespective of starting age (50-70) and until extreme limits of cost. In a direct comparison of the two agents, across a range of probability of 2%-6% per annum for the development of cirrhosis, pioglitazone remained cost-effective until its annual cost was greater than $16,000; beyond this, vitamin E was more cost-effective.

Figure 3.

Two-way sensitivity analysis: cost of pioglitazone and probability of remaining well for those with advanced fibrosis.


Our cost utility analysis indicates that for people with NASH and advanced fibrosis, treatment with a pharmacological agent in addition to lifestyle modification is likely to be cost-effective. The model identifies pioglitazone as the most cost-effective strategy, at a cost of $A2748 per additional QALY gained compared with lifestyle modification. Key factors driving this result include reduced progression to fibrosis with pioglitazone use, its relatively inexpensive cost, and the limited effectiveness of lifestyle modification at a population level. Currently, glitazones are recommended for use in people at high risk of progression to cirrhosis who fail lifestyle modification.19 Our modeled analyses add further information by indicating that their use is also likely to be cost-effective. In addition, vitamin E was cost-effective at $A8475 per QALY gained. Pioglitazone appears the more cost-effective option despite the fact that vitamin E is cheaper, although the differences in average costs and benefits between the two are not great. This is likely to reflect the slightly more favorable estimate used for reduced fibrosis progression with pioglitazone, and over the lifetime horizon of the model, even small differences in efficacy may translate to large cost savings at a population level when expensive outcomes such as liver failure and liver cancer are avoided.

Given the uncertainties inherent in the base case analysis, the robustness of the results was tested in sensitivity analyses. These suggest that the most influential variables in our model are the efficacy of pioglitazone, the probability of developing decompensated liver disease and subsequent death, and the costs of drug therapy. We assumed a base cost of pioglitazone of around $A1,000 and tested to an upper limit of around $A16,000 and found that the ICER remained cost-effective across this range but, above this, vitamin E appeared more cost-effective. The sensitivity analyses also showed that drug therapy may not be cost-effective when the likelihood of disease progression is low, and considerably greater when the likelihood of adverse outcomes is higher, as would be expected.

There are a number of caveats to our study. In particular, the ICER for both drugs appears highly favorable and, taken at face value, would represent excellent value for each healthcare dollar spent. In certain clinical situations, however, the ICER is likely to be less favorable. This is because it is not feasible to include all reported or potential side effects of drug therapies in economic models and therefore the quality of life gained may be less than our results indicate. For example, while weight gain with glitazones was included, reduced bone mineral density56 and bladder cancer57 were not. For clinical decision making, absolute risks of these events should be weighed against the likelihood of disease progression with no effective therapy. People with NASH and advanced fibrosis may progress to cirrhosis at a rate of more than 4% per annum, whereas the absolute increase, for example, in the risk of bladder cancer (assuming a causal relationship) is an extra 13 per 100,000 (or number needed to harm of 7,692). For vitamin E, meta-analytic data from observational studies suggests an increase in mortality with a high dose and this was included in the model; however, recent data on a possible increase in prostate cancer58 was not; providing these data may be part of the decision-making process for high-risk individuals.

These results cannot readily be extrapolated to patients with less advanced disease. We included patients with advanced fibrosis (F3 or greater) but the cost-effectiveness may be less for those with lower levels of fibrosis and/or reduced risk of progression. Similarly, for individuals who are very successful at adopting lifestyle change that results in weight loss and improved insulin sensitivity, the benefits of drug therapies are likely to be less. There are currently no trial data showing improvement in fibrosis with lifestyle modification, even with highly intensive and state-of-the-art programs59 and, therefore, a reduction in fibrosis due to lifestyle modification was not modeled.

Our model highlights the paucity of data in many areas required for comprehensive economic modeling in NASH, and therefore our study has a number of limitations. First, there are inherent inaccuracies and potential bias when using a surrogate marker instead of true clinical outcomes. There are currently no randomized trials of pioglitazone and vitamin E with long duration and liver-related outcomes, thus uncertainty about efficacy of one over the other remains. Such a trial is unlikely in the near future, and in this situation modeling represents a useful tool to explore potential outcomes and provide clinicians and decision makers the most reliable information in the setting of uncertainty. Future trials should aim to assess hard clinical endpoints, as previously recommended.12 Second, the lack of health-related quality of life data specifically derived from people with NASH may introduce bias. Although we felt it reasonable to assume that quality of life in endstage liver disease is similar regardless of the cause, the validity of this assumption has not been tested. To overcome this, we included a wide range for utility estimates derived from meta-analyses and other literature; however, there is a need for preference-based quality of life studies in the NASH population. We also did not include the costs of potential side effects of pioglitazone therapy such as incident fractures, as these data are not available. However, even in the largest trial of pioglitazone therapy, there were no significant differences in bone fractures, cardiac side effects, and anemia in the treatment and placebo groups, suggesting that the cost-effectiveness will not be significantly impacted upon. Equally, we did not include potential benefits of pioglitazone such as reduced progression to diabetes60 and reduction in adverse cardiovascular outcomes,61 which will be a benefit for some individuals.

Our study has a number of strengths. To our knowledge, this is the first economic evaluation of pharmacological therapies in NASH. Our clinical scenario envisaged treating only those with advanced disease (F3 fibrosis or cirrhosis), for whom prospective cohort studies on disease progression are available and for whom therapy is most required. In addition, we included a comprehensive literature search to identify data for probabilities, costs, and utilities, such that the model's estimates have incorporated the majority of data currently available for NASH. Further strengths include the level of evidence for key factors driving the model, from meta-analyses and randomized trials where available, and from long-term cohort studies.

In conclusion, current treatment recommendations for people with NASH and advanced fibrosis advise initial lifestyle modification followed by pharmacological therapies where lifestyle change alone fails. Such recommendations are not based on cost utility analysis. Our modeled analyses indicate that pharmacological therapies in addition to lifestyle modification are likely to be cost-effective. For patients, an individualized decision should be made taking into account their ability to modify their lifestyle, an evidence-based risk of fibrosis progression, and preferences regarding known side effects. For clinicians and policy makers, the decision to use pioglitazone or vitamin E where lifestyle changes fail is likely to be effective and cost-effective. Future therapeutic trials should include a prospective, parallel cost-efficacy arm according to best practice guidelines62 to permit more detailed scenarios to be modeled in the future.