| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From Lund University Centre for Health Economics (Drs. Bolin and Lindgren), Vårdal Institute, and Department of Health Sciences, Lund University; and Unit of Preventive Medicine (Dr. Willers), Heart and Lung Centre and Occupational and Environmental Medicine, Lund University Hospital, Lund, Sweden.
Correspondence to: Kristian Bolin, PhD, Lund University Centre for Health Economics, PO Box 705, SE-220 07 Lund, Sweden; e-mail: Kristian.Bolin{at}luche.lu.se
Abstract
Study objectives: To calculate incremental cost-utility ratios (cost per quality-adjusted life-year [QALY] gained) for bupropion (Zyban; GlaxoSmithKline; Gothenburg, Sweden), as compared to nicotine replacement therapy (NRT) in smoking cessation programs for a follow-up period of 20 years.
Design: The Global Health Outcomes simulation model was used for a male cohort and for a female cohort as a point of departure but was further extended in order to include the following: (1) the indirect effects of smoking cessation on production and consumption in the economy, and (2) morbidity-specific QALYs gained.
Setting: Sweden in 2001.
Patients or participants: Model cohort consisting of 612,851 male and 780,970 female smokers, distributed by age, 
35 years old, as in the Swedish population of 2001.
Interventions: Bupropion, as compared to NRT (nicotine patches and nicotine gums), in smoking cessation programs for a follow-up period of 20 years.
Measurements and results: When the indirect effects on production and consumption were taken into account, bupropion was cost saving in comparison to both NRTs. When only the direct costs were included, bupropion was still cost saving in comparison to nicotine gum. The incremental costs per QALY gained were relatively low for bupropion in comparison to nicotine patches, 6,600 Swedish kronas (SEK) (approximately 
725) per QALY gained for men and 4,900 SEK (approximately 535) for women, all calculations in 2001 Swedish prices. The comprehensive sensitivity analysis showed robust results; results were, however, more sensitive to quit rates and intervention costs than to other variables.
Conclusions: Bupropion is a cost-effective therapy in smoking cessation programs. Furthermore, recent studies report even higher effectiveness in terms of quit rates than was assumed here, indicating that our estimated cost-utility ratio should be even more favorable to bupropion.
Key Words: bupropion • economic evaluation • nicotine-replacement therapy • smoking cessation intervention • Sweden
Tobacco smoking is one of the leading causes of mortality. On a global level, almost 5 million premature deaths per year may be attributed to smoking.1 Smoking has been estimated to cause 440,000 deaths annually in the United States, 120,000 in the United Kingdom, 137,000 in Germany, and 7,000 in Sweden.2345 Smoking has been found to be the single most important cause of lost disability-adjusted life-years (DALYs) in the developed regions of the world, 12% of the total loss.6 It has been estimated that smoking accounts for 8.0% of total annual health-care costs in the United States, 6.1% in Great Britain, 3.8% in Canada, 3.7% in Germany, and 1.5% in Sweden245789; differences among countries reflect both differences in smoking prevalence rates and the health-care cost per smoker but also differences in estimation methods and data.5 In addition, lost productivity in the economy due to smoking adds an indirect cost, which may be at least as large as the direct health-care cost.2459
Thus, smoking cessation health programs are potentially beneficial. Whether or not additional resources spent on a smoking cessation program are worthwhile depends, however, on the extent to which there will be subsequent reductions in the costs imposed by smoking on the economy and/or sufficiently large increases in population health and welfare.
Bupropion is a non-nicotine-based substance that is used as an aid to smoking cessation (Zyban; GlaxoSmithKline; Gothenburg, Sweden).10 There are still only a few published studies on the cost-effectiveness of bupropion. Two of these studies1112 examine the cost-effectiveness of smoking cessation using bupropion compared to nicotine-replacement therapy (NRT); both conclude that smoking cessation interventions using bupropion are cost saving compared to NRT.
In a systematic review13 of the clinical effectiveness and the cost-effectiveness of bupropion and NRT for smoking cessation, the British National Coordination Centre for Health Technology Assessment concluded that bupropion generally showed better cost-effectiveness than NRT; the incremental cost per quality-adjusted life-year (QALY) gained was estimated to be £473 to £1,106 for adding bupropion to physician counseling only and £741 to £1,777 for adding NRT to counseling. However, none of the studies applied a comprehensive societal perspective.
The Health Economic Consequences of Smoking Model is a simulation model prepared for and reviewed by the World Health Organization European Partnership Project to Reduce Tobacco Dependence.14 It has been used in order to provide country-specific measures of the burden of smoking-related disease. The model has also been extended in order to incorporate the possibility of making comparisons among different smoking cessation interventions: the Global Health Outcomes (GHO) Smoking Control Model.15 This is a discrete and deterministic simulation model that predicts smoking-related morbidity and mortality, and the corresponding health-care costs.
In this study, incremental cost-utility ratios (cost per QALY gained) for bupropion as compared to NRT (nicotine patches or nicotine gums) were calculated for a follow-up period of 20 years, starting in 2001, using the GHO simulation model and Swedish data. Our calculations also included the impact of smoking on the economy at large. In the following, we first describe the simulation model and the data. Second, the baseline cost-effectiveness results are presented. Third, the robustness of the results is tested in a comprehensive sensitivity analysis. Fourth, results are discussed and the study is concluded.
Methods and Materials
Simulation Model
The GHO simulation model was used as a point of departure but further extended in order to include the following: (1) the indirect effects of smoking cessation on production and consumption in the economy, and (2) morbidity-specific QALYs gained. The model simulates the development of morbidity and mortality for a cohort of Swedish smokers, reflecting the prevailing age and morbidity structure in the year 2001 (612,851 men and 780,970 women), during 20 years after intervention. Five diseases were considered: (1) COPD, (2) asthma, (3) coronary heart disease [CHD], (4) stroke, and (5) lung cancer. Together, they cover most of the health problems associated with smoking, according to present epidemiologic knowledge, even though there are other smoking-related diseases.25 Smoking cessation affects the relative risks of being attacked by the respective disease. The relative risk reductions resulting from smoking cessation were calculated as the ratio between the relative mortality risk of smokers vs neversmokers and the relative mortality risk of former smokers vs neversmokers, using US estimates.16 Estimates specific for Sweden were not available.
The model distinguishes between men and women in three age groups: (1) < 35 years; (2) 35 to 69 years; and (3) 70 years. In order to simplify calculations, we assumed that there is no smoking-related morbidity at < 35 years of age and that the prevalence of smoking is zero < 16 years of age, assumptions that do not substantially affect the results. The model also distinguishes between three kinds of smokers (current smokers; recent quitters, who have the same relative risks as smokers; and long-term quitters, who have enjoyed the reduction in relative risks) and between three health states: no morbidity, morbidity, and dead. Thus, since it does not matter for the analysis whether a dead person happened to be a current smoker or a recent or long-term quitter during his or her lifetime, there are seven different states in the model.
The model employs discrete difference equations in order to calculate the distribution of the population over the states of the model in each year. The transitions from one state to another are represented by rates of change. The proportions, which transit from one state to another, were calculated given population-specific smoking habits and morbidity and mortality rates. Each year, 1 in 19 smokers < 35 years old (16 to 34 years) moves to the 35- to 69-year age group and hence acquires the same morbidity and mortality risks as those in the 35- to 69-year age group. Similarly, at the end of each year, 1 in 35 smokers in the 35- to 69-year age group advances to the 70-year age group. We assumed that those who die in a particular year survived half that year. In order not to complicate matters further, the model does not permit shifts from morbidity to no morbidity or from no morbidity to dead. The model also excludes comorbidity.
The number of smokers in each state at time t + 1, Nit+1 (i = 1,2,3,... 7), can be expressed by the following discrete difference equation:
 |
it+1 is the number of individuals who move to state I, and Nj
it+1 is the number of individuals who leave state i.17 This formulation is more general than a strict Markov process, since transition from one state to another is not restricted to be a function of time and the state to which the individual belongs.18 The model performs simultaneous but separate calculations for the intervention strategy and a chosen comparator strategy. The proportions of smokers who move between different smoking states were calculated using the following: (1) estimates of the proportion of smokers who attempt to quit each year, (2) estimates of the effectiveness of the different treatment strategies, (3) estimates of morbidity rates, (4) information regarding mortality rates, and (5) the relative risk reduction from smoking cessation. Each year, the proportion of individuals in the recent-quitter state, who remained abstinent from smoking for 12 months, moves to the long-term quitter state. Approximately one third (35%) of former smokers who have maintained abstinence for at least 1 year may eventually relapse.19 This relapse rate has been transformed into our 20-year time horizon; we assume that 35% of those who are long-term quitters will relapse over the 20-year interval. This means a yearly rate of approximately 2.1% (the annual long-term relapse rate used in our simulations).
Data
The simulation model was provided with the following input data for men, women, and age groups: (1) prevalence, incidence, and mortality rates of each of the diseases considered; (2) average annual direct health-care costs by disease; (3) QALY weights; (4) intervention cost; (5) smoking prevalence and quit rates; (6) treatment effectiveness; and (7) average net value of the indirect effects (consumption – production). Using this information, the simulation model calculated the incremental net cost per QALY gained over a 20-year period. Tables 123
report on the input data.
|
|
|
Estimated Average Annual Health-Care Costs by Disease:
The total (diagnosis-specific) cost of inpatient care was estimated as the number of hospital admissions (for those 35 years old)22 multiplied by diagnosis-specific average costs per admission.23 Data on physician visits and diagnosis-related information on drug prescriptions was obtained from a representative sample of 6.25% of all physicians in Sweden.24 Data are routinely collected twice a year in stratified samples. In total, 35 strata are being used, defined by region and specialty of the particular physician. At each time of observation, the participating physicians were asked to report information concerning diagnosis and prescriptions, for a period of 7 consecutive days. This data were then extrapolated to the entire population. Finally, the annual average health-care cost for each disease was calculated as the ratio between total direct health-care costs for that disease divided by the number of individuals 35 old with the disease (Table 1).
QALY Weights:
Calculations of differences in health-related quality of life among the treatment options were based on published DALY weights.25 Since DALY measures the difference between a perfect state of health (score of 1) and the actual health state, and since QALY measures the difference between dead (score of 0) and the actual health state, we defined QALY weights as 1 – DALY weight for each illness. However, DALY weights for COPD and lung cancer were not reported in the study. So, we set the QALY weight for COPD to 0.45, the argument being that it should be lower than the QALY weights for both asthma and CHD. For lung cancer, we chose the published DALY weight that reflects the level of disability in the final year of illness. The QALY weights used in the baseline case are reported in Table 1. The effects of the chosen weights are reported in the sensitivity analysis presented in the "Results" section.
Intervention Cost:
Intervention cost is the cost induced by the smoking cessation program, and it is comprised of the utilization of health-care personnel and the consumption of drugs. In Sweden, there are no established clinical guidelines regarding how to combine health-care contacts and bupropion in a smoking cessation intervention. However, the Swedish Council on Technology Assessment in Health Care employed examples of clinical regimen for smoking-cessation interventions using NRT.26 These routines are consistent with the guidelines provided by the Agency for Health Care Policy and Research in the United States and—more importantly—the clinical practice employed in the clinical trials that provided the effectiveness rates for the smoking cessation interventions under study.13 Thus, applying Swedish unit costs to the physical resource requirements, following the regimen employed by the Swedish Council on Technology Assessment in Health Care, can generate relevant intervention costs for the different smoking cessation programs.
Clinical practice varies regarding the utilization of health-care personnel, since smoking cessation interventions may be provided both in primary and in specialized health care, implying great variety both in personnel characteristics and in the mix of health-care contacts and drugs. It should be noticed that bupropion differs principally from NRTs in that it is only available on a prescription-only basis, and the need for motivational support is mandatory in the instructions for prescription. However, even though motivational treatment/support is not mandatory in NRT therapy as it is in bupropion therapy, the two therapies do not differ on this point: current smoking cessation treatment provided by health-care providers in Sweden always includes motivational treatment/support. Further, comprehensive estimates of smoking cessation intervention costs corroborate the picture that smoking cessation interventions using bupropion should be considered more health-care intensive than interventions using NRT. Thus, as our baseline case, we considered the case in which smoking cessation interventions using bupropion comprise two additional motivational support visits to a nurse as compared to a smoking-cessation program using NRT.
Men and women were treated identically with respect to the smoking cessation intervention. Base-case calculations for bupropion were performed for heavy smokers only, while we made separate calculations of the NRT intervention costs for light and heavy smokers. The calculations and the estimated intervention costs are reported in Table 2.
Smoking Prevalence and Attempted Spontaneous Quit Rates:
The age-specific prevalence of smoking was estimated from Swedish Survey of Living Conditions20 data for the years 1996/1997. Further, the proportion of smokers who make an attempt to quit each year was assumed to be 30%, based on a recent Swedish study.27 Data are reported in Table 3.
Treatment Effectiveness:
A metaanalysis13 provided the treatment-effectiveness rates. These effectiveness rates are usually lower than those reported by single studies.2829 The reason for this is that the metaanalysis comprised a wider range of smoking populations (for instance, smokers who have attempted to quit using a smoking cessation intervention previously). Further, the quit rates reported in the metaanalysis13 are continuous quit rates at 12 months, which are lower than the corresponding point estimates. The importance of the quit rate for the incremental cost-utility ratios is reported in the sensitivity analyses presented in the "Results" section. Quit rates used in the baseline case are summarized in Table 3.
Average Net Value of the Indirect Effects:
When a life is saved, both total production and consumption in the economy will be affected. If production increases more than consumption, total material wealth will be larger; if, however, production increases less than consumption, total material wealth will be smaller. Figures on the net value of the indirect effects for different age-groups were taken from a recent Swedish study.30 Age groups certainly did not match perfectly, but we calculated the value for the age group 35 to 69 years by weighting the values reported for the age groups 35 to 49 years and 50 to 64 years by their respective share of the population in the same age groups. The difference between the values of production and consumption for those > 69 years of age was estimated analogously. Using this method, we calculated the differences between the value of production and the value of consumption (value of production – value of consumption) for the two age groups to be 85,000 Swedish kronas (SEK) and 159,000 SEK, respectively (Table 3). It should be noticed that the value of consumption includes the average value of health-care consumption. This implies that to some extent changes in the value of consumption were double counted. However, the health care included in the value of consumption is here the average value, while the health-care consumption included in the direct cost component comprises disease-specific consumption.
The above figures were incorporated into the simulation model in order to take the indirect effects of reduced mortality into account. Since some of the individuals who die would not have been active in the labor market, had they not died, this method will overestimate the true indirect effects of mortality for the 35- to 69-year age group. However, individuals who would not have been active on the labor market had they not died would instead have evoked larger than average direct health-care costs. Analogously, our method will underestimate the true indirect effect of mortality for those > 65 years of age, since their overconsumption of health care, had they survived, is only included to a certain extent.
Discounting:
In the baseline analysis, both costs and effects were discounted at a 3% rate. The sensitivity analysis provides cost-effectiveness ratios for a number of different discounting rates.
Sensitivity Analysis
Univariate and bivariate nonstochastic and stochastic sensitivity analyses were performed in order to illuminate the importance of the assumptions made for the baseline case. Nonstochastic sensitivity analyses are useful for examining the range of a variable for which the incremental cost-utility ratio (ICUR) falls below a certain threshold value. However, the likelihood of achieving a cost-utility ratio that falls below that threshold value cannot be inferred from such analysis. Available information concerning the distribution and its characteristics for the parameters of the ICUR may be used for performing a Monte Carlo simulation and, hence, forecast the likelihood distribution of the incremental cost-utility ratio. Such simulations were performed using software (Crystal Ball; Decisioneering; Denver, CO), which computed year 2000 estimates of the incremental cost per QALY over 20 years.31 Each estimate was achieved by applying a sample from the probability distributions, which have been defined for the stochastic variables.
The discount rate might be regarded as part of the method employed and, if so, should not be regarded as stochastic.32 Thus, in order to incorporate the effect on the simulation of different discount rates, without applying more specific assumptions concerning the parametric form of the probability distributions, we included uniformly distributed discount rates. For the same reason, we included uniformly distributed QALY weights.
Results
Baseline Case
Table 4
shows the baseline-case results of our 20-year simulations for the male and female populations. Results are reported regarding the following: (1) total incremental intervention cost (the additional cost imposed by using bupropion instead of the comparison treatment); (2) total health-care cost averted (the cost saved in the health-care sector from using bupropion instead of the comparison treatment; the negative sign indicates that this amount was subtracted from the total incremental cost); (3) total indirect cost averted (reduction in the difference between the value of production lost and the value of consumption lost induced by saved life years, as a result of using bupropion instead of the comparison treatment; the negative sign indicates that this amount was subtracted from the total incremental cost); (4) total QALYs gained (the number of life years saved adjusted for the loss of quality implied by the disease); and (5) the incremental cost per QALY gained, ie, costs in (1) to (3), above, divided by (4). Results are reported separately for bupropion vs nicotine patches and bupropion vs nicotine gums, respectively. They are all based on the population of Sweden in 2001; all costs are in 2001 prices.
|
Nonstochastic Sensitivity Analysis
In this section, we will examine the effects on the incremental cost per QALY gained of variations in the following: (1) the follow-up period; (2) the discount rate; (3) QALY weights; (4) intervention cost; (5) effectiveness; and (6) relative risk reduction. The results of the sensitivity analyses are presented in Table 5
for the length of the follow-up period, and, for the remaining input data, in Table 6
for bupropion vs nicotine patches and in Table 7
for bupropion vs nicotine gum. All reported incremental costs per QALY gained in Tables 56789
include indirect costs.
|
|
|
|
|
Discounting:
A lower discount rate than the baseline 3% for both costs and QALYs would decrease (ie, improve) the ICUR; a higher rate would increase the ratio. A lower discount rate for QALYs but at the same time higher discount rate for costs would increase the incremental cost per QALY gained. Varying the discount rate between zero and 5% would still leave bupropion as the dominant option (Tables 6, 7).
QALY Weights:
Lowering the average QALY weight would increase the ICUR; increasing the average weight would decrease the ratio. Varying the QALY weight between 0.1 and 0.75 does not change the baseline case result that bupropion is the dominant option.
Intervention Cost:
A lower intervention cost for the bupropion alternative than the estimated 3,135 SEK would obviously decrease (ie, improve) the ICUR; a higher cost would increase the ratio. The ICUR would turn positive if the intervention cost for bupropion were roughly 20% (Table 6) or 25% higher (Table 7). If the intervention cost were still higher, the ICUR would increase even more.
Effectiveness:
A higher quit rate for bupropion therapy than the baseline 18.9% would obviously improve the ICUR; a lower quit rate would increase the incremental cost per QALY gained. The ICUR would increase gradually as the estimated quit rate for bupropion approaches the estimated quit rates for nicotine patches (15.6%) and nicotine gum (15.0%). Since the intervention cost for bupropion is higher than for the NRT, the ICUR would turn positive even before the compared therapies had the same quit rates.
Relative Risk Reduction:
Varying the relative risks would affect the direct costs of health care and the indirect effects on the economy at large in opposite directions. Assuming a higher value for the relative risk for smokers vs neversmokers would result in a lower relative risk reduction and, hence, larger benefits from smoking cessation, which means that fewer health-care costs would be inflicted on society. This per se would decrease the ICUR. However, it would also mean that a larger number of individuals would live to be old, which would increase the ICUR (since the expected length of life increases for each life gained). When using the lower and upper limits of the 95% confidence interval reported by Thun et al,16 the ICURs would still be negative.
Bivariate Sensitivity Analysis:
Both Effectiveness of Bupropion and Intervention Cost Varied Simultaneously: According to the univariate sensitivity analyses above, our results seemed to be relatively more sensitive to the data on effectiveness and intervention cost. Thus, we performed a bivariate sensitivity analysis in which these two parameters were varied simultaneously. Since we were only interested in less favorable outcomes for smoking cessation interventions using bupropion, effectiveness was varied between 18.5% and 16% and incremental cost was varied between baseline plus 100, 500, 1,000, and 2,000, respectively. As Tables 8 and 9 show, both a lower effectiveness rate and a higher intervention cost for the bupropion alternative would rapidly increase (ie, worsen) the 20-year ICURs.
Stochastic Sensitivity Analysis
Another sensitivity analysis, in which the discount rate, the QALY weights, the intervention cost, the effectiveness, and the relative risk reduction were regarded as stochastic, was also performed. The following specifications of respective probability distribution were made.
Discount Rate:
Uniform distributions, defined in the range 0 to 5%, were employed for the discount rate.
QALY Weights:
Uniform distributions, defined in the range 0.1 and 0.9, were employed for the QALY weights.
Intervention Cost:
Resource use differs between clinical settings; hence, for our purposes intervention cost may be regarded as a stochastic variable. For lack of information, we simply assume that the bupropion intervention cost has a lognormal distribution, centered on the incremental costs implied in Table 2. This recognizes that our interest focus on cases in which the incremental cost of the bupropion alternative is positive.
Effectiveness:
A ß-distribution was chosen to model the quit rate. The ß-distribution is a commonly used distribution when modeling uncertainty in the probability of an outcome.32 It is characterized by two parameters that shape the distribution: first, it is possible to model variability over a fixed range, that is 0 to 1; second, the distribution can be shaped in order to take into account the number of included participants and the number of quitters. We shaped the distribution in order to capture the baseline effectiveness rate for bupropion (18.9%) as the mean value of the distribution.
Relative Risk Reduction:
We assumed that the relative risks were normally distributed and that each distribution followed the information provided by Thun et al.16 Caution was observed so that the relative risk reduction did not exceed 1. This was achieved by truncating the normal distribution at critical levels.
Results:
The stochastic simulations regarding the comparison between smoking cessation using bupropion therapy and smoking cessation using nicotine patches showed that there is an 80% chance of observing a reduction in the cost (a negative incremental cost) for both men and women. Similarly, there is a 90% chance of observing a positive incremental cost per QALY gained, which is < 5,000 SEK for men and < 14,000 SEK for women. Thus, even after taking into account the various uncertainties regarding the variables used in the simulation analysis, there is still, for both men and women, at least an 80% chance of observing a cost reduction, while there is as high a probability as 90% of observing a positive incremental cost per QALY, which is as low as (or lower than) 5,000 SEK for men and 14,000 SEK for women. The corresponding results regarding the comparison between smoking cessation using bupropion therapy and smoking cessation using nicotine gums were even more favorable for the bupropion alternative: for both men and women, there was a 90% chance of observing a negative incremental cost per QALY gained.
Discussion
In this study, we performed a simulation-model analysis of Swedish cohorts of male and female smokers over a 20-year period. The GHO model was extended in order to incorporate indirect effects on production and consumption and morbidity-specific adjustments for quality of life. Swedish data were used concerning the following: (1) diagnosis-specific morbidity and mortality; (2) smoking prevalence and quit rates; (3) health care costs; (4) production and consumption impact on the economy at large; and (5) smoking cessation intervention costs.
In the baseline case, the incremental cost-utility ratios for the comparison between bupropion and nicotine patches and between bupropion and gums were negative. We performed several sensitivity analyses. The conclusion that smoking cessation using bupropion dominates both the NRT options seems robust though. Exceptions are for extremely high incremental intervention costs and low bupropion effectiveness rates.
When only the direct costs were included, bupropion still had lower costs and improved quality of life in comparison with nicotine gum, ie, it is a dominating alternative. In comparison to nicotine patches, the incremental direct cost per QALY gained for bupropion was relatively low: 6,600 SEK (approximately 725) for men and 4,900 SEK (535) for women. These figures are far lower than those reported in cost-utility analyses of relevant chronic pulmonary disease (asthma and COPD) treatments, ranging from 72,000 SEK (7,900) to 378,000 SEK (41,400) per QALY gained,3334353637 and far below what people report that they would be willing to pay for a QALY gained, ranging from 129,000 SEK (11,000) to 333,000 SEK (28,500), according to a recent study.38
The results of the sensitivity analysis certainly tell us something about the validity of our results. However, a sensitivity analysis per se does not contain any information concerning what range of values to study or the likelihood of the occurrence of specific values in a given range; such information would always be provided, for instance, by specific knowledge generated in clinical trials and in clinical practice and from more or less informed guesses. The baseline case of this article contains the most plausible combination of inputs judged from such knowledge.
Although our sensitivity analysis is likely to contain the relevant ranges and combinations of input values, clinical practice in smoking cessation therapies does differ between different parts of the health-care sector and, hence, the cost-effectiveness of smoking-cessation intervention differs among clinical settings. Smoking cessation interventions are provided by occupational health-care, primary health-care, and intensive health-care units. The total intervention costs differ, depending on where in the health-care system the smoking-cessation intervention takes place. However, comparisons are difficult, depending on several factors, for instance different patient populations. Also, it can be assumed that bupropion is used in more complicated cases. Further, treatment by a nurse who specializes in smoking cessation interventions results in fewer doctor visits. Generally, there is a need for a closer examination when using bupropion (for different reasons such as indication, possible negative side effects). However, the motivational treatment/support does not differ in practice between nicotine replacement and bupropion treatment. Thus, the costs associated with counseling should not differ much between bupropion and NRT.
The simulation analysis, which was performed in this article, comprised a 20-year time horizon; hence, gains from smoking cessation that might occur in the future were not taken into account. The direct costs, which would be avoided beyond the 20-year horizon, are positive and, hence, would contribute to making the cost-effectiveness ratio even more negative, ie, more favorable to bupropion. The effect on the cost-effectiveness ratio of including indirect effects beyond the 20-year horizon is more complex. For those < 40 years of age, there is a 10-year period during which the value of production exceeds that of the consumption.30 Thus, including these years in the analysis would further decrease the cost-effectiveness ratio. However, the decrease in lost years of life among those > 70 years of age would increase the cost-effectiveness ratio, since the value of consumption exceeds the value of production for this age group. However, if the effect of smoking cessation on survival probabilities diminishes with age, this effect would be small. To summarize, extending the time horizon of the simulation analysis is likely to result in further reductions of the considered cost-effectiveness ratios, ie, to improve the case for bupropion.
The results obtained in this study are certainly consistent with the results reached by the National Institute for Clinical Excellence in the United Kingdom, after taking into account the fact that the studies considered in their survey39 did not incorporate the indirect effects of smoking cessation on the economy at large. The quit rates employed in this study might, however, underestimate the true quit rates resulting from bupropion smoking cessation intervention. Aubin et al40 reported significantly higher quit rates than the rates used in our baseline case. It did not, however, report quit rates for the same population as ours. Our sensitivity analysis (Tables 56789) imply that the estimated incremental cost-utility ratios are sensitive to the true effectiveness rates. Thus, in as much as these recent estimations are valid, the incremental cost-utility ratios estimated in this article would be even lower.
A recently published model analysis41 of five face-to-face smoking cessation interventions in the Netherlands confirms the favorable cost-utility ratio of intense general practitioner (GP) counseling with bupropion compared to intensive GP counseling with nicotine-replacement therapy.
Footnotes
Abbreviations: CHD = coronary heart disease; DALY = disability-adjusted life-year; GHO = Global Health Outcomes; GP = general practitioner; ICUR = incremental cost-utility ratio; NRT = nicotine replacement therapy; QALY = quality-adjusted life-year; SEK = Swedish krona
Financial support was provided by GlaxoSmithKline, Sweden.
Received for publication June 22, 2005. Accepted for publication January 2, 2006.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
