Evaluation of Cost-Utility of Varenicline Compared with Existing Smoking Cessation Therapies in South Korea
Eui Kyung Lee, Graduate School of Clinical Pharmacy, SookMyung Women's University, Hyochangwongil 52, Yongsan-Ku, Seoul 140-742, South Korea. E-mail: email@example.com
Objective: This study aimed to evaluate the cost-effectiveness of varenicline compared with the other smoking cessation interventions, bupropion, nicotine replacement therapy (NRT), and willpower.
Methods: The Benefits of Smoking Cessation on Outcomes model was modified to reflect major smoking-related diseases in Korea. Transitional probabilities, resource utilization, and costs were obtained from Korean public data. The analysis was carried out from a societal perspective for the lifetime period. Also, series of sensitivity analyses, including probabilistic sensitivity analysis, were performed.
Results: With the exclusion of bupropion, which is subject to extended dominance, the incremental cost-effectiveness ratio (ICER) for varenicline versus NRT was analyzed as $US4809 per quality-adjusted life-year (QALY) during the lifetime. The results of sensitivity analysis are quite stable across most of the included parameters. The acceptability curves showed that the probability of varenicline being cost-effective was 83.3% at the willingness to pay of $US15,000.
Conclusions: Even though the maximum willingness to pay for a QALY has not officially been defined, varenicline can be regarded as cost-effective because the ICER is at the 24.0% level of per capital gross domestic product, which is an implicit reference for decision-making in Korea.
Smoking is not simply a preferential activity but a detrimental one, causing various physical diseases. Many smokers often fail to quit smoking because smoking is related to nicotine dependency, which leads to chronic diseases that require repeated therapies. There are some smoking cessation drugs that are currently available, such as nicotine replacement therapy (NRT), bupropion SR, nortriptyline, and clonidine. A new smoking cessation drug, varenicline is also available as a prescription drug in Korea. Nevertheless, it is not reimbursed by insurance coverage yet. For a new drug to be reimbursed, an economic assessment based on Korean pharmacoeconomic (PE) guidelines  is mandatory with the introduction of positive list system from 2007 (for more information on all the following references, see Evaluation of Cost-Utility of Varenicline Compared with Existing Smoking Cessation Therapies in South Korea Value in Health Supporting Information, part II at: http://www.ispor.org/Publications/value/ViHsupplementary/ViH12s3_EKLee.asp). A cost-effectiveness analysis of varenicline has already been performed in other countries [2–4], but the results cannot be applied to a Korean clinical setting for the following reasons.
First, smoking-related morbidities can be different in Korea. Unlike Western countries, stomach cancer and liver cancer rank higher than other smoking-related cancers in Korean males. Among Korean females, lung cancer, esophageal cancer, and stomach cancer are the three most prevalent disease burdens .
Second, in Korea, per capita health-care expenditures are relatively low compared with other Organization for Economic Cooperation and Development (OECD) countries. New drugs are valuable when they prevent smoking-related morbidities. In Korea, however, the treatment cost is relatively low and the trade-off between the drug and treatment cost can be different from other countries.
Third, the smoking rate by sex in Korea is different from that of other countries. There are similarities in the smoking rate between the sexes in Europe and the United States. In Asian countries including Korea, however, the female smoking rate is much lower than the male smoking rate. According to the 2005 Korean National Health and Nutritional Survey , the overall smoking rate was 27.3% (male: 52.2%, female: 3.9%). These results showed that most smokers are males. If the cost-effectiveness of smoking cessation treatment is different between the sexes, sex difference in smoking rate among countries can have some impact on the study results.
Given the above background, we conducted this study to assess the cost-effectiveness of varenicline compared with the other currently available smoking cessation drugs (NRT and bupropion) and willpower. Then, we attempted to propose the most cost-effective alternatives by estimating the long-term benefits of smoking cessation from a societal perspective based on Korean PE guidelines .
A Markov model, the Benefits of Smoking Cessation on Outcomes model, was adopted to conduct a long-term cost-utility analysis in four smoking cessation alternatives (for more information on all the following tables and figures, see Evaluation of Cost-Utility of Varenicline Compared with Existing Smoking Cessation Therapies in South Korea Value in Health Supporting Information, part I at: http://www.ispor.org/Publications/value/ViHsupplementary/ViH12s3_EKLee.asp). The time horizon was set to the lifetime years reaching the time point when the cohort became 100 years old. Based on Korea's PE guidelines , a discount rate of 5% was applied to both cost and effectiveness.
Rationales for our selection of six diseases are based on the following reasons: First, the three cancers (lung, stomach, and liver cancer) are the top three cancers from the result of multiplying the disability-adjusted life-year by the population attributable risk (PAR) due to smoking . Second, three smoking-related chronic diseases (ischemic heart disease [IHD], stroke, and chronic obstructive pulmonary disease [COPD]), known to have high incidence rates among the Korean population and to be associated with smoking greatly, were the most prevalent diseases with high PAR .
The prevalence, incidence, and mortality rate of the six diseases in this study were subdivided based on smoking state, sex, and age band using the existing epidemiological data [6,8,9–11] and the relative risk (RR) of smoking [12,13] (Table S2).
To compare the efficacy of each smoking cessation intervention, we found head-to-head trials about varenicline and bupropion, and those about varenicline and NRT. Nevertheless, there are only two clinical trials [14,15] that compared varenicline and bupropion, and no clinical studies have been conducted on varenicline and NRT. For the quitting rates for the interventions, we conducted meta-analysis of the related trials [14–18] (Table S3).
The relapse rate for the recent quitters was obtained as 6.8% from the previous study  (References 19–36 found at Evaluation of Cost-Utility of Varenicline Compared with Existing Smoking Cessation Therapies in South Korea Value in Health Supporting Information at: http://www.ispor.org/Publications/value/ViHsupplementary/ViH12s3_EKLee.asp). By contrast, the long-term quitters are those who have quitted for more than 6 years (6+ years), in whom the relapse rate was 2% during a period ranging from 6 to 10 years and 1% after a 10-year period .
We obtained utility weights based on published studies [21–25]. On the one hand, utility weights for the cohort with no comorbidity were specified depending on the age, sex, and smoking status . On the other hand, the utility weight for the cohorts with a smoking-related morbidity was applied to the model regardless of the age, sex, and smoking status based on various data resources [21–25] (Table S4).
Table S5 shows the direct annual per capita medical costs of six smoking-related diseases [26–28]. The direct cost for therapies includes drug , pharmacy dispensing, physician visit, monitoring, and traveling cost . In order to calculate the direct medical cost, we used National Health Insurance reimbursement fee schedule . As NRT is an over-the-counter drug, only drug cost was included (Table S5).
One-way sensitivity analyses were conducted on discount rate, efficacy of varenicline, cost, and utility weight. Also, the probabilistic sensitivity analysis (PSA) was conducted for major parameters used in the model, including effectiveness of smoking cessation strategies and treatment cost for morbidities and utilities. For the distribution followed by each parameter, the beta distribution was used for the quit rate and utilities, whereas the log normal distribution was used for costs based on related study .
Varenicline had the lowest absolute number of cumulative lifetime incidence cases of smoking-related morbidities (Table S6), and also have the lowest number of deaths from the included smoking-related morbidities. Drug cost was the highest for varenicline, whereas direct medical cost to treat the included smoking-related morbidities was the lowest compared with the other comparators.
The results showed that the incremental cost-effectiveness ratio (ICER) of bupropion to NRT is $US53,241 per quality-adjusted life-year (QALY), which is higher than $US859 per QALY, the ICER of varenicline to bupropion (Table S7). For the remaining alternatives, NRT versus willpower is $US1956 per QALY, and varenicline versus NRT is $US4809 per QALY. Willpower, NRT, and varenicline are placed on the kinked line of lower slopes, and this line represents cost-effectiveness frontier (Fig. S1). But bupropion is placed on the left side of the line, indicating that it is dominated (by extended dominance) and excluded from consideration.
Subgroup analysis by sex was performed, and varenicline was a consistently dominant strategy at lifetime. The ICER of varenicline versus NRT was higher in males as $US4913 per QALY, whereas it is $US3861 per QALY in females.
As shown in Table S8, results of the one-way sensitivity analysis of key model parameters such as discount rate, efficacy of varenicline, costs, and utility weights were summarized. For discount rate, 0%, 3%, and 7.5% were used based on the Korean PE guidelines. The result shows that the higher the discount rate, the higher the ICER value. For efficacy, 95% confidence level upper limit and lower limit of the quit rate of varenicline was applied to the model. As a result, the lower limit was more sensitive. For the costs, the ICER for the cases in which NRT was prescribed by doctors was relatively lower. Nevertheless, other parameters, increase of cancer cost and use of different data source, were robust. Of those without comorbidities, when the utility weights of smokers were applied as one of the nonsmokers, the cost was the same as the baseline analysis. But both QALYs and ICERs increased for all alternatives. The upper limit and lower limit of all utility weights by 0.05 did not have a big impact on the result.
The result of the PSA can be explained with the cost-effectiveness acceptability curve (Fig. S2). As shown in the figure, this study shows that the cost-effectiveness probability of NRT versus willpower is 80.9% with the critical value of $US5,000, 87.9% with $US10,000, and 90% with $US15,000. Meanwhile, the cost-effectiveness probability of varenicline versus NRT is 75.2% with the critical value of $US10,000 and 83.3% with $US15,000.
Our results showed that varenicline dominated bupropion. For the remaining alternatives, the ICER for NRT versus willpower was $US1956 per QALY, whereas varenicline versus NRT was $US4809 per QALY.
Still, in Korea, no official threshold of ICER has been announced for decision-making. At the Drug Reimbursement Evaluation Committee, however, a certain range of per capita gross domestic product (GDP) is referenced as a threshold of ICER for decision-making (Korean per capita GDP is $US20,045 in 2007). Orlewska  suggested the reference range as a ratio of 0.7∼2.3 times per capita GDP, which is $US14,000∼$US46,000 in Korea, but the actual criteria for decision-making in Korea seems to be lower than the suggestions. Even though there is no explicit range for decision-making, however, as ICERs of varenicline and NRT are at the 24.0% and 9.8% level of per capita GDP, respectively, we think both of them can be regarded as cost-effective alternatives.
A one-way sensitivity analysis showed that our results were generally robust for discount rate, utility weight, and cost. Varenicline dominated bupropion consistently, whereas ICERs for varenicline versus NRT ranged from $US1575 per QALY to $US8284 per QALY. This corresponds to less than the per capita GDP. The parameter that had the biggest impact on the result of the sensitivity analysis was efficacy of varenicline. Even though the assumption that efficacy of varenicline is at the 95% confidence interval lower limit is extremely conservative, ICER of $US16,544 is still lower than GDP. Also, the PSA suggests that the model results are consistent across multiple stochastic runs. For the comparison of varenicline with NRT, 75.2% of the simulations showed cost-effectiveness at $US10,000 per QALY, whereas 83.3% at $US15,000 per QALY. The results showed that the uncertainty around the input parameters did not influence the conclusions.
Previous studies [2–4,33] have demonstrated the cost-effectiveness of varenicline, although their detailed results are different from ours. In the current Korean study, varenicline did not dominate and its ICER was higher than the results of other countries.
The major reason is that the share of treatment cost for smoking-related morbidities is lower in Korea, although the share of drug cost is higher as compared with other countries. For example, in Korea, the cost of varenicline accounts for 16.4% of cumulative direct cost at lifetime, while the treatment cost for smoking-related morbidities accounted for 83.6%. In the United States, the varenicline cost is 1.3% and the treatment cost for smoking-related morbidities is 98.7% . The value of varenicline is to reduce the treatment cost for smoking-related morbidities. In Korea, however, the treatment cost for smoking-related morbidities takes a relatively small share. Accordingly, varenicline showed less favorable cost-effective results.
Another reason is that the number of male smokers was greater than female smokers in Korea. The share of male smokers is 92.1% in Korea, although it is 58.4% in the United States. Because the ICER of females is $US3862 per QALY, which is lower than males ($US4913), the fact that the share of male smokers is large increases the overall ICER value.
As for limitations, first, this study used outcome data from foreign clinical studies because there are only limited references on efficacies and utility weights domestically. To circumvent this difficulty, PE guideline  recommends comparing foreign data with local clinical and epidemiological data to cross-check for data transferability, and then to solicit expert opinions when published clinical or epidemiological data are not available. Although it is not enough at all, we referred to the expert opinions on the transferability of foreign outcomes data based on the guideline, and got a positive response.
Second, modeling the benefits of smoking cessation is challenging because multiple morbidities and related mortalities were obtained by transformation based on the formula of the previous studies. A lack of observable data on a population basis prevents availability of direct evidence for the model inputs. But previous studies showed that the use of the RR of mortality as a proxy for RR of morbidity is valid [34–36].
Third, we did not include productivity costs and the medical costs of morbidity during the additional life-years gained, even though it is well established nowadays that the costs for nonsmoking-related diseases in additional life-years are substantial. Regarding future increases in health-care costs originating from increased survival, however, there seems to be no consensus yet among economists on the correct way of doing it. For these debatable issues, Korean guidelines recommend that both productivity costs and unrelated health-care cost be not included in the future for the baseline analysis .
For the following reasons, our results are conservative. First, although smoking is associated with many forms of cancer, cardiovascular diseases, respiratory problems, and reproductive problems, only the six types of diseases were included. Second, we assumed that NRTs are purchased by consumers directly without visiting doctors' offices as it is a nonprescription drug. But bupropion and varenicline, both of which are prescription drugs, must be purchased after the visit to both doctor's office and pharmacy. Therefore, doctor's fee, pharmacist's fee, and medical examination fee were applied for bupropion and varenicline, whereas the medical costs for NRT and willpower were applied to a lesser extent.
Varenicline costs $US4809 more to gain an additional QALY compared with NRT. To date, no maximum willingness to pay for a QALY has officially been defined in Korea. Nevertheless, varenicline can be regarded as a cost-effective alternative because the ICER is at the 24.0% level of per capital GDP, which is an implicit reference for decision-making in Korea.
Source of financial support: This study was sponsored by Pfizer, Ltd., Seoul, South Korea. Jung Youn Bae, Cheol Hwan Kim, and Eui Kyung Lee have no conflicts to declare.