The target population of this study was selected from a systematic review. It is composed of male and female adult patients with no history of resistance to antiviral drugs, and without coinfection with HIV, hepatitis C or hepatitis D virus. Patients were also tested for cirrhosis or hepatocellular carcinoma at the beginning of treatment.

The technologies evaluated in this study were the drugs for chronic hepatitis B in the usual doses used in Brazil, according to the clinical guideline:

• -

  Adefovir dipivoxil: 10mg per day orally.

• -

  Entecavir: 0.5mg per day orally.

• -

  Interferon-α: 5MUI daily.

• -

  Pegylated interferon-α: 1.5mcg/kg of pegylated interferon alpha 2b subcutaneously administered weekly (used as reference patients weighing 60kg).

• -

  Lamivudine: 100mg per day orally.

• -

  Telbivudine: 600mg per day orally.

• -

  Tenofovir disoproxil fumarate: 300mg per day orally.

Patients receiving these drugs were compared to those receiving no treatment.

We considered patients with chronic hepatitis B who were followed up clinically, but did not receive treatment.

A systematic review was performed for published articles that evaluate the efficacy and safety of treatments for chronic hepatitis B. It was conducted in five baselines: Cochrane Library, MedLine/PubMed, LILACS, SCIELO, and IPA (International Pharmaceutical Abstracts). We considered studies published in Portuguese, English, German or Spanish, published between 2000 and July 2011.

Two independent reviewers extracted data and assessed the trial's quality using the Jadad scale.4 Only studies of medium and high quality (at least 3 points on the Jadad scale) were included. Inclusion criteria were randomized controlled trials (RCTs) lasting for at least three months. We excluded review papers and editorials, case studies or studies with fewer than 10 patients.

Eligible studies were those conducted on adult patients older than 18 years, who had been diagnosed with chronic hepatitis B, received one of the drugs evaluated in the usual doses, and were compared with patients receiving other drugs, the association of two drugs, or a placebo. Studies, where the control group was given the same drug but in different doses or for a different duration of therapy, were excluded. In case of discrepancy in any of the enumerated steps, the discordant results were resolved by consensus between the reviewers and, when necessary, by a third reviewer.

From the selected articles, extracted data included: viral response, lack of response to treatment, treatment withdrawals, viral resistance, HBeAg seroconversion, viral breakthrough, and adverse events. Data were extracted on pre-prepared tables and separated into three groups: HBeAg positive patients, HBeAg negative patients, and a third group comprising all patients. This third group was created because some articles presented the aggregate results of HBeAg positive and negative patients.

Another systematic review was conducted looking for studies that assessed the utility of patients with chronic hepatitis B in each condition. The transition probabilities of patients between health states were also extracted from these studies. This search was made in PubMed/MedLine. We also looked at studies that evaluated quality of life of patients at different stages of chronic hepatitis B, evaluated by a generic quality of life questionnaire.

The TreeAge Pro® software was used to construct a decision-analytical Markov model with estimates of the natural history of chronic hepatitis B infection. Scenarios were created assuming health conditions and situations arising from the use of drugs and also from clinical pathology. The scenarios considered that the sequence of events that might occur would be the same for all interventions, carrying only the probability of events occurring in different treatments and also their costs and utilities.

The model was first applied to patients already diagnosed with chronic hepatitis B infection. The decision node represents the patient's treatment options (ADF, ETV, IFN-α, LAM, TDF, LdT and PEG-IFN-α) or no treatment. The treatment was considered discontinued in the states, “serious adverse events” or “death”. Furthermore, it was considered that in the states, “no response to treatment” and “viral resistance,” the treatment was changed to TDF, as recommended by the Brazilian Guideline (for patients who received TDF, this was changed to ETV).

For the state, “death,” mortality data from the Brazilian general population was used, considering patients aged 40 years in whom mean disease onset was found in the systematic review.

After building the model, the yearly costs of each stage were added. The data were extrapolated to a time horizon of 40 years, considering the chronic nature of the disease. We performed a cost-utility analysis, which measures the cost of treatment related to QALY.5 The incremental cost-effectiveness ratio (ICER) was evaluated to compare the alternatives that were more expensive and more effective than others.

To evaluate the cost of each treatment, the parameters of the Brazilian public health system were used; only direct costs were included. The cost of the drugs were obtained by averaging the amounts paid by the Ministry of Health in the bids made during the year 2011, with the exception of the LdT, which is not funded by the Brazilian public health system. For this product, the ABC Pharma table was used, which standardizes the selling price of medicines in the private network in the country. For the cost of other procedures (consultations, examinations, hospitalizations and surgeries), we used the cost table produced by the Brazilian health system hospitals.

The discount rate used in the study was 5%, both for cost and effectiveness values.

The uncertainties were minimized with sensitivity analyses that were performed using univariate analysis, varying the parameters of cost and effectiveness for more and less. To evaluate the robustness of the method and reduce uncertainties, the following types of analyses were performed:

• -

  Change the weight of the patients using PEG-IFN-α 2b from 60 to 70kg; ranged the utilities of treatment as reported in the literature.

• -

  Variation of the cost of each treatment according to the maximum and minimum prices paid for each drug in the tenders held during 2011.

• -

  Discount rates varied between 0 and 10%.

• -

  Transition probabilities between health states were varied from minimum and maximum values found in literature.

In the first stage of the systematic review, 1174 articles published between 2000 and July 2011 were found in the databases. Of these, 100 randomized clinical trials were selected based on their abstracts. Of these, 246–29 studies met the inclusion criteria and were included.

We included data only from studies that evaluated the patients during one year. The extracted probabilities were described in Table 1 for HBeAg positive, negative, and all patients. For HBeAg negative patients, we found no RCTs that evaluated the IFN-α alone, without association, and met the inclusion criteria. Therefore, for this group of patients, the drug was not evaluated.

31 articles were found in the systematic review performed to search for the utilities of patients with chronic hepatitis B in different health states. Of these, only three articles met the inclusion criteria and were original, while in the other 28 papers the utilities were taken from another study. Of the three resulting articles, the study by Levy et al.30 was chosen for the extraction of utility values, because it was a multi-center study conducted on 1134 patients, on whom the Health Related Quality of Life Questionnaire (HRQOL) was conducted. The utility values found for each state of health were (range):

• -

  Uncomplicated chronic hepatitis B: 0.68 (from 0.66 to 0.70).

• -

  Compensated cirrhosis: 0.69 (0.66–0.71).

• -

  Decompensated cirrhosis: 0.35 (0.32–0.37).

• -

  Hepatocellular carcinoma: 0.38 (0.36–0.41).

• -

  Liver transplantation (first year): 0.57 (0.54–0.60).

• -

  Liver transplantation (other years): 0.67 (0.64–0.69).

Some utility values were not found in the literature and therefore were adapted from the selected study. For the outcomes “viral response” and “HBeAg seroconversion”, values were considered equivalent to the value “health”. The outcomes “no response to treatment,” “withdrawal due to adverse events,” and “viral resistance” were considered equivalent to the value “uncomplicated chronic hepatitis B”.

The transition probabilities between the health states of chronic hepatitis B were also extracted from the studies selected in the systematic review,2,31–37 obtaining the average, maximum and minimum values. The values are described in Table 2.

In untreated patients the following probabilities were considered:33,38

• •

  Compensated cirrhosis: 23.89%.

• •

  Clinical remission: 1.6%.

• •

  Death: 2.9%.

• •

  Hepatocellular carcinoma: 1.5%.

• •

  Hepatocellular carcinoma to death: 50%.

Costs of drugs per patient over a one year treatment period are shown in Table 3. We calculated the average cost of each drug in the year 2011, considering that the price can vary between the bids. The cost of IFN-α and PEG-IFN-α was added in the Markov model only for the first year because Brazilian guidelines prohibit therapy with interferon beyond one year.

For the cost of health states defined by decompensated cirrhosis, hepatocellular carcinoma and liver transplantation, values were taken from another study conducted in Brazil in 2005, for which the price list has not been updated since then:39

• -

  Decompensated cirrhosis: U$ 12,803.

• -

  Hepatocellular carcinoma: U$ 2770. This cost does not take into consideration chemotherapy.

• -

  Liver transplantation: U$ 50,798 in the first year and subsequent years we used the cost of clinical care.

Three Markov models were developed: for HBeAg positive, negative, and for all patients (Fig. 1). The main difference in the model of HBeAg negative patients is that there is no seroconversion of HBeAg.

Fig. 1.

Markov model used for cost-utility analysis.

(0.2MB).

It was considered that all patients started in the state “treatment” and then transited to one of the health states: viral response, death, no response to treatment, withdrawal due to adverse events, seroconversion of HBeAg (in HBeAg positive patients), viral resistance, compensated cirrhosis, or hepatocellular carcinoma.

The results of cost-utility analyses are shown in Fig. 2, for HBeAg positive, negative and all patients, respectively. The costs, effects, and cost-utility are described in Table 4. The results are in increasing order of cost-utility values, and compared to the item “no treatment”.

Fig. 2.

Cost-utility of treatments in HBeAg positive, HBeAg negative and all patients. PEG-IFN, pegylated interferon-α; IFN, interferon-α; QALY, quality adjusted life years.

(0.17MB).

By varying the initial utility or utility of the state viral response in HBeAg negative patients, in some cases, PEG-IFN becomes less cost-effective than the option “no treatment”. Varying the cost of LdT in HBeAg negative patients, LdT shows better cost-utility values than ADF. In the three groups of patients, varying the discount rate to 10%, LAM was more cost-utility than ETV.

Because TDF had the best cost-utility of the evaluated drugs, the worst scenario was created for TDF based on data from systematic reviews. The probability of developing hepatocellular carcinoma, compensated cirrhosis, and decompensated cirrhosis was ranged to the highest value found in the literature. Still, TDF was the alternative with the better cost-effectiveness ratio among the three groups of evaluated patients; all other strategies were dominated.

The modeling for economic evaluations can bring limitations because values are extrapolated based on findings in the literature and created scenarios. Sensitivity analyses are performed to try to minimize possible variations in the model. However, some limitations should be considered.

First, the results of clinical outcomes of this study were derived from clinical studies from the last ten years. Clinical trials evaluate the efficacy of therapies, and not effectiveness, which reflect the clinical practice in use. The course of chronic HBV infection is dynamic, which could be modified significantly with anti-HBV agents but it is still dynamic. We used data extracted from systematic review that only had limited therapy interval and limited follow-up duration to estimate the long-term cost-effectiveness of drug therapy. Moreover, there might be significant differences in the HBeAg seroconversion rate and viral resistant rate in the 1st and 2nd year of treatment. Thus, one year data could lead to significant bias. However, these data were used because there are no clinical studies available for all these drugs with more than one year duration.

The population involved in clinical trials is heterogeneous, and the responses to treatments may be different in our population. Also, different genotypes were found in the included studies, which are not necessarily the same genotypes found in the Brazilian population.

Another important limitation of the study is that in the Markov model, it was assumed that the patient cannot be in two health states at the same time. This assumption differs from reality because patients could, for example, have decompensated cirrhosis and hepatocellular carcinoma at the same time, and still experience a serious adverse event. However, the model precludes the representation of this type of a situation because the clinical trials report the data separately.

TDF was found to be the drug with the best cost-utility ratio for the three groups of patients. However, this information is ascertained only from two studies, as described in the article by Marcellin et al.11 A larger number of studies may provide a more reliable result. New clinical studies of TDF would, therefore, increase the robustness of this result.

The results of this study establish the cost-utility of drugs available in Brazil used to treat chronic hepatitis B. Understanding this relationship is crucial for health professionals; it bolsters the decision-making process. It is essential for doctors to be aware of the cost-utility ratios of each drug. However, each patient responds to therapy in different ways and should be evaluated individually.