Economic evaluation of HLA-B*15:02 screening for carbamazepine-induced severe adverse drug reactions in Thailand

Authors

  • Waranya Rattanavipapong,

    1. Health Intervention and Technology Assessment Program (HITAP), Nonthaburi, Thailand
    Search for more papers by this author
  • Tanunya Koopitakkajorn,

    1. Health Intervention and Technology Assessment Program (HITAP), Nonthaburi, Thailand
    Search for more papers by this author
  • Naiyana Praditsitthikorn,

    Corresponding author
    1. Health Intervention and Technology Assessment Program (HITAP), Nonthaburi, Thailand
    2. Bureau of AIDS TB and STIs, Department of Disease Control, Ministry of Public Health, Nonthaburi, Thailand
    • Address correspondence to Naiyana Praditsitthikorn, Department of Health, Health Intervention and Technology Assessment Program (HITAP), Ministry of Public Health, 6th Floor, 6th Building, Tiwanon Road, Muang, Nonthaburi 11000, Thailand. E-mail: naiyana.p@hitap.net

    Search for more papers by this author
  • Surakameth Mahasirimongkol,

    1. The National Institute of Health, Department of Medical Sciences, Ministry of Public Health, Nonthaburi, Thailand
    Search for more papers by this author
  • Yot Teerawattananon

    1. Health Intervention and Technology Assessment Program (HITAP), Nonthaburi, Thailand
    Search for more papers by this author

Summary

Purpose

There is strong evidence of an association between the presence of the human leukocyte antigen (HLA)-B*15:02 and two severe adverse drug reactions—Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN)—in patients taking carbamazepine (CBZ), a common treatment for patients with epilepsy and neuropathic pain. As a result, there are calls for all patients that are due to undergo CBZ therapy to be screened for this genetic marker before commencing their therapy. This study aims to determine the value for money of HLA-B*15:02 screening compared to the following: (1) administering CBZ therapy without conducting patient screening, and (2) not prescribing CBZ but alternative drugs that are less likely to result in severe reactions, but that come at a higher cost.

Method

An economic evaluation was carried out by using a decision tree and Markov models to examine the cost-utility of providing HLA-B*15:02 screening for all patients with either newly diagnosed epilepsy or neuropathic pain in the Thai setting. All transitional probabilities were derived from the national and international literature. The majority of the data on direct medical care costs were collected from 10 community, provincial, and regional hospitals throughout Thailand. Direct non-medical cost and health-related quality of life (HRQoL) data were obtained from interviews that were conducted with 33 patients, some of whom had experienced severe drug reactions.

Key Findings

The incremental cost-effectiveness ratio (ICER) of adopting a universal HLA-B*15:02 screening policy was estimated at 222,000 Thai baht, THB/quality-adjusted life year (QALY) gained for epilepsy patients and 130,000 THB/QALY gained for patients with neuropathic pain. Furthermore, we found that 343 patients need to be tested for HLA-B*15:02 allele to prevent one case of SJS/TEN.

Significance

Universal HLA-B*15:02 screening represents good value for the money in terms of preventing SJS/TEN in CBZ-treated patients with neuropathic pain at the Thai ceiling ratio of 120,000 THB/QALY gained. However, the prevalence of CBZ-induced SJS/TEN in the Thai population and the positive predictive value (PPV) are major factors that influence the cost-effectiveness of HLA-B*15:02 screening. Therefore, an active surveillance system to make a more accurate assessment of the prevalence CBZ-induced SJS/TEN in the Thai population would enhance the generalizability of the results.

According to national and global pharmacovigilance systems, the most common adverse drug reactions (ADRs) are cutaneous. The most severe life-threatening forms of cutaneous ADRs are Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN), two related acquired bullous disorders of the skin that, in the majority of cases, are caused by reactions to certain drugs, such as sulfonamide-antibiotics, antiepileptic agents—especially carbamazepine (CBZ), allopurinol, and oxicam-type nonsteroidal antiinflammatory drugs (NSAIDs) (Harr & French, 2010). Incidence rates for SJS/TEN vary according to ethnicity, and the highest rates are seen among Han Chinese, Malays, and Thais (Lim et al., 2008).

CBZ is the primary treatment choice for patients with epilepsy and neuropathic pain according to current Thai clinical practice guidelines (National Drug Committee, 2008; Thai Association for the Study of Pain, 2008; The Epilepsy Society of Thailand, Prasat Neurological Institute, 2011). As such, many doctors are forced to make a difficult choice—they can either prescribe patients CBZ treatment, with the knowledge that it brings with it high risk of SJS/TEN (Health Product Vigilance Center, 2011a), or they can prescribe a less effective and more expensive treatment. Evidence suggests that host genetics play an important role in determining the probability of developing SJS/TEN reactions to certain drugs (Chung et al., 2004; Hung et al., 2005; Chantarangsu et al., 2009; Chen et al., 2011). In 2004, HLA-B*15:02 was the first genetic marker identified for predicting CBZ-induced SJS/TEN in Taiwanese patients.

Since the initial discovery of HLA-B*15:02 and CBZ-induced SJS/TEN, this association has been confirmed in a number of other Asian populations. For instance, in South East Asian populations and Southern Chinese populations, where HLA-B*15:02 carriers are relatively common (>10% for carriers in the population and >5% for minimal allele frequency), this association was also demonstrated. However, in countries where the proportion of the population who are HLA-B*15:02 carriers is low, such as Japan and Korea, the association has not been found (McCormack et al., 2011). This is also true for European nations, where carriers of HLA-B*15:02 are also relatively rare (McCormack et al., 2011). As a result of these studies, the U.S. Food and Drug Administration (FDA) now recommends that HLA-B*15:02 screening be conducted for all patients of Asian descent before undergoing CBZ treatment. Moreover, the screening for HLA–B*15:02 is covered under the National Health Insurance system in Taiwan (Chantratita et al., 2011).

At present in Thailand, allele-specific genotyping for HLA-B*15:02 is currently available through nine regional centers of the laboratory network of the National Institute of Health (NIH) at the Department of Medical Sciences (DMSc) and at a number of medical school laboratories. Despite the availability, a universal HLA-B*15:02 screening policy for patients due to undergo CBZ therapy is not yet in place in Thailand. To decide whether screening should be conducted universally, an economic evaluation study is required to help policy makers assess the value for money of universal HLA-B*15:02 screening compared to (1) the current practice, in which patients receive no HLA-B*15:02 screening before undergoing CBZ treatment, and (2) prescribing drugs other than CBZ that have a lower risk of triggering severe reactions but are more expensive. The results of the study will be submitted to the Subcommittee for Development of the Benefit Package and Service Delivery (SCBP)—the coverage decision authority responsible for issuing recommendations on whether HLA-B*15:02 screening should be offered to individuals who are eligible under the health benefit package of the Universal Coverage Scheme (UC).

Methods

Overview of potential strategies

Current practice

In Thailand, patients with either newly diagnosed epilepsy or neuropathic pain commonly commence their treatment with CBZ because it is an effective and inexpensive drug (1.39 Thai baht [THB] per tablet) (Center of Essential Information for All Health Officers, 2011) that is widely available at all types of hospital throughout the country. However, these patients usually commence their CBZ without first undergoing screening for HLA-B*15:02. The standard World Health Organization (WHO)–recommended defined daily dose for CBZ is 1 g for both patients with epilepsy and neuropathic pain (The WHO Collaborating Centre for Drug Statistics Methodology, 2012). Although the majority of patients with epilepsy require life-long treatment, clinicians have indicated that this is likely to be unnecessary for most patients in the Thai setting (Rattanavipapong et al., 2012a), where the average treatment duration for epilepsy is about 4 years. These findings are similar to those found in Taiwan (Hu et al., 2011). However, we did assess the value for money of providing lifetime treatments for patients with epilepsy. For patients with neuropathic pain, the average treatment duration in the Thai setting was found to be 2 years (Rattanavipapong et al., 2012a).

Universal HLA-B*15:02 screening

Because CBZ-induced SJS/TEN has been found to be associated with the presence of the HLA-B*15:02 allele (Locharernkul et al., 2008; Tassaneeyakul et al., 2010), the provision of HLA-B*15:02 screening prior to starting treatment with CBZ should serve as a useful pharmacogenetic test. At present, screening is available to anyone who can afford it (at a cost of 1,000 THB, according to the DMSc price). However, there are growing calls for HLA-B*15:02 screening to be publicly funded and alternative treatment options offered for those who test positive for HLA-B*15:02.

Prescribing alternative drugs without screening for HLA-B*15:02

To prevent CBZ-induced SJS/TEN, patients can be prescribed alternative drugs at the outset of treatment, thereby removing the need for HLA-B*15:02 screening. As recommended by the Thai Clinical Practices Guidelines and an expert consultation meeting, sodium valproate (VPA) and gabapentin (GBP) are suitable alternative treatments for patients with epilepsy and neuropathic pain, respectively (Rattanavipapong et al., 2011; The Epilepsy Society of Thailand, Prasat Neurological Institute, 2011). The standard dosage regimen for VPA is 1.5 g per day at a cost of 12.66 THB per 500 mg tablet (Center of Essential Information for All Health Officers, 2011; The Epilepsy Society of Thailand, Prasat Neurological Institute, 2011; The WHO Collaborating Centre for Drug Statistics Methodology, 2012). The standard dosage regimen for GBP is 1.2 g a day at a cost of 8 THB per 300 mg tablet (Thai Association for the Study of Pain, 2008; Center of Essential Information for All Health Officers, 2011; Moore et al., 2011; Rattanavipapong et al., 2012a).

Analyses and model

Model-based economic evaluations consisting of a decision tree and a Markov model were constructed for the Thai context using the societal perspective. Two separate models were constructed—one for patients with newly diagnosed epilepsy and one for patients with newly diagnosed neuropathic pain. These models were used to evaluate the costs and consequences of each strategy, by following the same adult cohort (made up of individuals aged 20 years or older) for all strategies. The lifetime time horizon was used with a cycle length of 1 year. The primary outcomes of interest were lifetime costs, quality-adjusted life-years (QALYs) gained, and the incremental cost-effectiveness ratio (ICER) in THB per QALY gained.

The decision tree model (Fig. 1) imitated the three potential strategies for treating patients with epilepsy or neuropathic pain. In the first strategy, patients started CBZ without HLA-B*15:02 screening (reflecting current practice). In the second strategy, all patients were tested for the HLA-B*15:02 allele, and patients who tested positive for HLA-B*15:02 allele were prescribed the alternative drugs, whereas patients who tested negative received CBZ. In the third strategy, patients were given alternative drug treatments and did not undergo HLA-B*15:02 screening. For all patients—whether they received CBZ or alternative drugs—there were three possible outcomes: (1) the development of SJS/TEN, (2) the development of other ADRs without SJS/TEN, and (3) no development of ADRs.

Figure 1.

Decision tree model showing the three practical strategies for epilepsy or neuropathic pain treatment with carbamazepine.

A Markov model (Fig. 2) was used to predict the lifetime costs and outcomes of each outcome that occurred after taking CBZ or alternative drugs. As shown in Figure 2A, there are four possible health scenarios for patients who develop SJS/TEN: (1) patients develop SJS/TEN, which may or may not lead to complications during hospitalization; (2) patients recover but experience sequelae such as blindness; (3) patients recover without any complications; and (4) patients die from SJS/TEN or other causes, such as diseases and accidents. Patients who develop other ADRs could move to a state of recovery or death in the next cycle, as shown in Figure 2B, whereas patients who do not develop any ADRs could remain in this state or could die from other causes, as shown in Figure 2C. All parameters used in the analysis are presented in Table 1 and discussed further in the model parameter sections. The Monte Carlo simulation was used to model costs and outcomes over a lifetime period. All costs and outcomes occurring after 1 year were discounted at a rate of 3%, as recommended in the health technology assessment guidelines of Thailand (Permsuwan et al., 2008).

Table 1. Input parameters used in the model
ParametersDistributionMeanSE
  1. THB, Thai baht; PPV, positive predictive value; NPV, negative predictive value; N/A, not applicable; CBZ, carbamazepine; VPA, sodium valproate; GBP, gabapentin; ADRs, adverse drug reactions; DMSc, the Department of Medical Sciences, Ministry of Public Health.

  2. a

    Analysis of primary data using bootstrap method.

  3. b

    Analysis of primary data collected by the authors.

  4. c

    Because no data are available on the costs of treating SJS/TEN with sequelae, the costs of treating patients with psoriasis, which normally involves the skin, were used instead.

  5. d

    The utility of patients with complete vision impairment was used in this study because there were no available data on the utility of patients with SJS/TEN with sequelae.

  6. e

    Utility of patients with other ADRs is assumed to be 10% lower than the utility of recovered patients from SJS/TEN without complication.

Epidemiologic parameter and transitional probabilities    
Prevalence of HLA-B*15:02 allele in the Thai population (Kupatawintu et al., 2010; Romphruk et al., 2010)Beta0.1550.003
Probability of CBZ-induced SJS/TEN in patients testing positive for HLA-B*15:02 allele (PPV) (Tassaneeyakul et al., 2010)Beta0.0190.019
Probability of CBZ-induced SJS/TEN in patients testing negative for HLA-B*15:02 allele (1-NPV) (Tassaneeyakul et al., 2010)Beta0.00040.0004
Probability of patients developing sequelaeBeta0.5700.060
Probability of CBZ-induced other ADRsBeta0.0320.032
Probability of VPA-induced other ADRsBeta0.0040.004
Probability of GBP-induced other ADRsBeta0.0250.025
Probability of death due to CBZ-induced SJS/TENBeta0.0020.002
Costing parameters (THB per year)    
Epilepsy   
Direct medical cost   
Treatment with CBZaNormal4,09438
Treatment with VPAbNormal15,47720
CBZ-induced SJS/TENaNormal25,868192
VPA-induced SJS/TENbNormal25,666189
Treatment of other ADRs (Kankeawlar & Phurmsuwan, 2006)Gamma2,3192,319
Follow-up of sequelae (Thavorncharoensap et al., 2008)cGamma1,5051,505
Direct non-medical cost   
Treatment with CBZ or VPAaNormal6,43123
CBZ-induced SJS/TENaNormal20,812199
VPA-induced SJS/TENbNormal21,020202
Treatment of other ADRs (Kankeawlar & Phurmsuwan, 2006; Rattanavipapong et al., 2012b)Gamma1,8701,870
Follow-up of sequelae (Rattanavipapong et al., 2012b)cGamma794794
Neuropathic pain   
Direct medical cost   
Treatment with CBZaNormal5,38736
Treatment with GBPbNormal14,57633
CBZ-induced SJS/TENaNormal26,970322
GBP-induced SJS/TENbNormal26,885305
Treatment of other ADRs (Kankeawlar & Phurmsuwan, 2006)Gamma2,3192,319
Follow-up of sequelae (Thavorncharoensap et al., 2008)cGamma1,5051,505
Direct non-medical cost   
Treatment with CBZ or GBPaNormal3,52728
CBZ-induced SJS/TENaNormal37,230226
GBP-induced SJS/TENbNormal37,608231
Treatment of other ADRs (Kankeawlar & Phurmsuwan, 2006; Rattanavipapong et al., 2012b)Gamma1,8701,870
Follow-up of sequelae (Rattanavipapong et al., 2012b)cGamma794794
Utility parameters    
Epilepsy   
No ADRs aBeta0.680.003
Develop SJS/TENaBeta−0.080.002
Recover from SJS/TEN without complicationBeta0.520.003
Recover from SJS/TEN with sequelae (Teerawattananon et al., 2007)dBeta0.300.200
Develop other ADRseBeta0.460.003
Neuropathic pain   
No ADRsaBeta0.630.003
Develop SJS/TENaBeta−0.180.002
Recover from SJS/TEN without complicationBeta0.650.003
Recover from SJS/TEN with sequelae (Teerawattananon et al., 2007)dBeta0.300.200
Develop other ADRseBeta0.580.003
Figure 2.

(AC) Markov model representing the three events that could occur for patients after treatment by carbamazepine or alternative drugs. Each event shows all health states in each cycle. It is possible to transit to another health state with transitional probability as shown by the arrows. (A) Patients who developed SJS/TEN after receiving drugs. (B) Patients who developed other ADRs after receiving drugs. (C) Patients who did not developed any ADRs after receiving drugs.

Model parameters

Health state transitional probabilities

Transitional probabilities between health states were obtained from published studies and primary data. The prevalence of HLA-B*15:02 allele carriers in the Thai population was derived from our own meta-analysis of national studies (Kupatawintu et al., 2010; Romphruk et al., 2010). We conducted an analysis of the Health Product Vigilance Center (HPVC) database of the Thai Food and Drug Administration (TFDA) from 2006 to 2011  (Health Product Vigilance Center, 2011b), to acquire the probability of patients developing SJS/TEN, sequelae, or other ADRs from either CBZ or alternative drugs. The age-adjusted mortality rate for the general population was taken from the Thai Burden of Disease and Injury Study in Thailand (The Thai Working Group on Burden of Disease and Injuries, 2004).

Intervention effectiveness

The PG 1502 test (PharmiGene, Inc., Taipei, Taiwan) was used to detect the presence of the HLA-B*15:02 allele; this has been approved by Taiwan's Department of Health as a recommended in vitro diagnostic (IVD) test. The sensitivity of the test was 100%, and the specificity was 98.7%, as specified by the manufacturer (Tassaneeyakul et al., 2010). On the basis of previous research, the prevalence rate for CBZ-induced SJS/TEN in the Thai population was set at 0.27%; the positive predictive value (PPV) and negative predictive value (NPV) of HLA-B*15:02 screening were set at 1.92% and 99.96%, respectively (Tassaneeyakul et al., 2010). The PPV was used to indicate the probability of CBZ-induced SJS/TEN occurring in patients who tested positive for the HLA-B*15:02 allele; 1-NPV (0.04%) was used to indicate the probability of CBZ-induced SJS/TEN occurring in patients who tested negative for the HLA-B*15:02 allele.

Cost and utility parameters

Costs and health-related quality of life (HRQoL) data were collected from Thai patients who had been treated with CBZ from 10 community, provincial, and regional hospitals. The case groups comprised patients who were diagnosed by specialists as having CBZ-induced SJS/TEN and who met the inclusion criteria. To minimize recall bias, only patients who had started CBZ treatment for epilepsy or neuropathic pain treatment between 2006 and 2010, when they were 20 years old or older, were included in the dataset. However, patients with conditions that can act as underlying etiologic conditions for SJS and TEN, such as graft-versus-host disease, human immunodeficiency virus (HIV) infection, tuberculosis, and systemic lupus erythematosus (SLE), were excluded from the study. For the control group, patients who had not developed any adverse drug reactions from CBZ were individually matched with patients who had developed SJS/TEN in terms of age, gender, disease, CBZ starting time, and treatment settings. Following this, 15 patients who had developed SJS/TEN were identified as suitable for inclusion in the case (5 of whom had epilepsy) and 18 patients were identified as suitable for the control group (5 of whom had epilepsy). Ethics committee approval was given and the patients gave informed consent.

In the model, costs were defined as (1) direct medical costs—the cost of treating the disease and any ADRs, the cost of follow-up, the cost of screening, outpatient fees, and blood testing fees—all of which were collected from hospital databases, and (2) direct non-medical costs—travel and food costs for patients and their caregivers, personal facility costs, and opportunity costs incurred by patients (e.g., patient time spent on visits to healthcare facilities). These were defined based on information provided during detailed face-to-face interviews with patients. Indirect costs were excluded in order to avoid double-counting, since the effectiveness outcome or QALYs already take into account morbidity and mortality effects (Riewpaiboon, 2008). The medical records of all patients were reviewed to identify the level of health care resource utilization; costs were estimated using a standard costing menu and reference price (Center of Essential Information for All Health Officers, 2011; Riewpaiboon, 2011). All costs were converted to 2011 values using the Thai consumer price index (Bureau of Trade & Economic Indices, 2011). For international comparison, costs were converted into international dollars using the purchasing power parity (PPP) conversion rate, where a PPP 2011 dollar is worth 17.505 THB (The World Economic Outlook Database, 2012).

The mean direct medical costs per year of SJS/TEN treatment induced by CBZ and alternative drugs were 25,868 THB (SE = 192) and 25,666 THB (SE = 189), respectively, in the epilepsy model, versus 26,970 THB (SE = 322) and 26,885 THB (SE = 305), respectively, in the neuropathic pain model. The direct non-medical costs of SJS/TEN treatment induced by CBZ and alternative drugs were 20,812 THB (SE = 199) and 21,020 THB (SE = 202), respectively, in the epilepsy model, versus 37,230 THB (SE = 226) and 37,608 THB (SE = 231), respectively, in the neuropathic pain model.

The QALYs gained from the interventions were measured as health outcomes. The HRQoL results were obtained from face-to-face patient interviews using the EQ-5D-TH questionnaire (The EuroQol Group, 1990; Brooks, 1996; Tongsiri, 2009). All patients were asked about their health state preferences, and these data were then converted to utility scores, which were in turn multiplied by life expectancy to generate QALYs. Detailed information about the means and standard errors of each health state are presented in Table 1. The mean utility score of patients with epilepsy was 0.68 (SE = 0.003), whereas the mean utility score of patients with neuropathic pain was 0.63 (SE = 0.003). The mean utility scores of patients who developed SJS from CBZ decreased dramatically to −0.08 (SE = 0.002) in patients with epilepsy, and −0.18 (SE = 0.002) in patients with neuropathic pain.

The bootstrap method was used to analyze the primary data from both the dataset and control group. This method was chosen because of its suitability for studies with small sample sizes and large variability. The bootstrap technique creates a new sample by randomly sampling individuals with replacements from the original sample, and then computing bootstrap replicates of statistical quantities, such as the sample mean, standard error, and confidence intervals; in most cases, this is based on at least 1,000 bootstrap samples. As a result, the bootstrap method uses simulated data to create large sample sizes, which provides more accurate and reliable estimates (Campbell & Torgerson, 1999; Grunkemeier & Wu, 2004; Walters & Campbell, 2005; Sharma & Kim, 2012). According to the Central Limit Theorem, the sampling distribution will show normal distribution because the sample size is sufficient (Briggs et al., 2006).

Uncertainty analyses

Probabilistic sensitivity analyses (PSAs) were conducted to offset the effect of parameter uncertainty. The input variables were assigned a probability distribution to reflect the feasible range of values that each input parameter could attain (Briggs, 2000). Consequently, one value from each parameter was taken out and calculated for cost and effectiveness. This process was repeated 1,000 times, and the range of possible values was given as a cost-effectiveness acceptability curve. This analysis used the cost-effectiveness ceiling threshold of 120,000 THB per QALY gained, as recommended by the Health Economic Working Group under the Subcommittee for Development of the National List of Essential Drugs (The Health Economic Working Group, 2012).

In addition, a threshold analysis was performed for the sake of generalization. The results of different levels of selected parameters—prevalence of the presence of the HLA-B*15:02 allele in the Thai population, PPV, NPV, cost of VPA, cost of HLA-B*15:02 screening, and duration of epilepsy treatment—were assessed to evaluate their potential benefits in other settings.

Results

Costs

The lifetime costs of each option are presented in Figure 3. Overall, the average lifetime cost for the current treatment policy tends to be lower than the other two strategies, by approximately 42,100 THB for epilepsy patients, and 18,600 THB for patients with neuropathic pain. The average lifetime cost of providing universal HLA-B*15:02 screening before prescribing CBZ to patients with epilepsy and neuropathic pain was about 50,200 and 22,700 THB, respectively. The average lifetime cost of prescribing drugs other than CBZ to all patients who would currently receive CBZ treatment was nearly 83,700 for epilepsy patients and 35,800 for patients with neuropathic pain.

Figure 3.

The lifetime costs of each strategy categorized by disease.

As Figure 3 shows, the cost of treating epilepsy according to the current policy is almost the same as the cost of providing universal HLA-B*15:02 screening before prescribing CBZ. However, current practice accounts for most of the cost of SJS/TEN treatment, a cost that can be avoided with the implementation of prospective HLA-B*15:02 screening. Because the cost of prescribing VPA is nine times higher than CBZ, the cost of treating epilepsy represented a sizeable proportion of the total costs in cases where VPA rather than CBZ was prescribed. For patients with neuropathic pain, a similar pattern was observed.

An analysis was conducted on the number needed to screen (NNS), which showed that 343 patients need to be screened for HLA-B*15:02 allele to prevent one case of SJS/TEN. As a result, the cost of preventing one case of SJS/TEN was calculated as 375,200 THB (the cost providing HLA-B*15:02 screening for 343 patients).

Health outcomes

It is estimated that without HLA-B*15:02 screening, 187 patients will develop SJS/TEN annually. Universal HLA-B*15:02 screening or treatment with alternative drugs can reduce this number to approximately 23 patients per year. When the provision of universal HLA-B*15:02 screening is compared with current practice, the number of life years saved and QALYs gained is insignificant for patients with both epilepsy and neuropathic pain, reflecting the fact that SJS/TEN is a rare condition. A similar finding is observed when comparing universal HLA-B*15:02 screening and providing alternative drugs (Table 2).

Table 2. Life time costs and health outcomes of each strategy using societal perspectivea
OptionsCosts (THB)Incremental LYs (days)Incremental QALYs (days)ICER (THB per QALY gained)
  1. THB, Thai baht as of 2011 value; LYs, life years; QALYs, quality adjusted life years; ICER, incremental cost-effectiveness ratio.

  2. a

    The number presented in table was rounded to the nearest whole number.

Epilepsy model    
Current practice42,000
Universal HLA-B*15:02 screening50,0000.0713.44222,000
Alternative drugs84,0000.110.3832,522,000
Neuropathic pain model    
Current practice19,000
Universal HLA-B*15:02 screening23,0000.0511.53130,000
Alternative drugs36,0000.030.1335,877,000

Cost-utility analysis

Table 2 shows the average lifetime costs and the QALYs gained for each of the three strategies compared in this study. Compared to current practice, universal HLA-B*15:02 screening incurs higher costs but results in slightly more QALYs gained. The ICER of the universal HLA-B*15:02 screening strategy was estimated at 222,000 THB/QALY gained for epilepsy patients and 130,000 THB/QALY gained for patients with neuropathic pain. Compared to universal HLA-B*15:02 screening, the policy of providing alternative drugs without screening yielded only marginal benefits but did result in a significantly higher ICER of 32,522,000 THB/QALY gained for epilepsy patients and 35,877,000 THB/QALY gained for patients with neuropathic pain.

Uncertainty analysis

A probabilistic sensitivity analysis was undertaken, and the results are presented in Figure 4. The line graph illustrates the probabilities of the three strategies being cost-effective at different ceiling ratios. The results reveal that the probability of universal HLA-B*15:02 screening being cost-effective at the Thai ceiling ratio of 120,000 THB/QALY gained for epilepsy and neuropathic pain models, is 16% (Fig. 4A) and 32% (Fig. 4B), respectively. It can be clearly seen that the cost-effectiveness of the screening rises in correlation with the increase of the ceiling threshold.

Figure 4.

(A and B) Cost-effectiveness acceptability curves demonstrate that the probability of three strategies for epilepsy and neuropathic pain treatment would be cost-effective at the different ceiling threshold values. (A) Probability of three strategies of epilepsy model being cost-effective at different ratio. (B) Probability of three strategies of neuropathic pain model being cost-effective at different ratio.

The probability of CBZ-induced SJS/TEN occurring in a person who tested positive for HLA-B*15:02 was sensitive to the result of the cost-effectiveness of universal HLA-B*15:02 screening. Because this parameter was obtained from only one paper, which was not based on a national survey, we conducted our own threshold analysis to determine the ICER for the range of prevalence of CBZ-induced SJS/TEN in the Thai population, the positive predictive value (PPV), and the negative predictive value (NPV). The results (Fig. 5) show the effect of the prevalence, PPV, and NPV values on ICERs. In the base case, universal HLA-B*15:02 screening in patients with neuropathic pain is found to be cost-effective at a rate of nearly 0.27% based on current evidence; conducting screening for patients with epilepsy is not found to be cost-effective. An increased prevalence of CBZ-induced SJS/TEN resulted in a reduction of ICERs for both patients with neuropathic pain and those with epilepsy. This particular information is useful because data on CBZ-induced SJS/TEN, PPV, and NPV values might change in the future, and this trend may be applicable to other relevant settings.

Figure 5.

Result of threshold analyses. *Prevalence of CBZ-induced SJS/TEN in the Thai population.

In addition, the reduced cost of VPA was also found to increase the value for money of HLA-B*15:02 screening. Screening becomes cost-effective if the cost of VPA decreases from 12.66 THB to 7.07 THB. However, a reduction in the cost of HLA-B*15:02 screening does not affect the value for money of the screening itself, because the screening cost is only a minor part of the total lifetime cost. This study also examined the effect of the duration of epilepsy treatment on the value for money of the proposed regimen. The durations assessed ranged from 4 years to lifetime, and the model found that the longer the duration of treatment, the lower the value for money of the screening.

In a voluntary reporting system, it is likely that the number of drug-induced ADRs is likely to be underestimated. To account for this, we performed an analysis that showed that the change of other CBZ-induced ADRs had a moderate effect on the ICER of screening in patients with epilepsy but exerted very little effect on patients with neuropathic pain. If the rate of other CBZ-induced ADRs is as high as 11.2% (the upper bound of its 95% confidential interval), the ICER drops to 143,000 THB/QALY. In contrast, if the rate of other CBZ-induced ADRs is as low as 0.1% (the lower bound of its 95% confidential interval), the ICER increases to 275,000 THB/QALY. Lastly, the analysis found that none of the other model parameters exerted an effect on the value for money of the gene screening. These include the direct non-medical costs and utility parameters data collected from a small number of samples.

Discussion

The findings of this study reveal that a significant decrease in the number of CBZ-related SJS/TEN cases is observed if either a universal HLA-B*15:02 screening policy or a policy of prescribing drugs other than CBZ is adopted for the treatment of epilepsy and neuropathic pain. In addition, universal HLA-B*15:02 screening was found to be preferable to alternative drug treatment because, although both policies generate similar health outcomes, the former is less expensive. Furthermore, our analysis of the NNS found that 343 patients need to be tested for the HLA-B*15:02 allele to prevent one case of SJS/TEN. As a result, the additional cost of preventing one case of SJS/TEN is assessed to be 375,200 THB (the cost of providing HLA-B*15:02 screening for 343 patients).

Using the cost-effectiveness ceiling threshold in Thailand, a program of universal HLA-B*15:02 screening represents good value for the money when conducted on patients with neuropathic pain, but not when conducted on patients with epilepsy. This is because the cost of alternative drug regimens is much higher for epilepsy (83,700 THB for lifetime treatment with VPA) than it is for neuropathic pain (lifetime treatment with GBP costs 35,800 THB). Moreover, with a universal screening policy, a significant proportion of patients with epilepsy who test positive for the HLA-B*15:02 allele will not develop SJS/TEN (98%), but will still switch to the more expensive treatment, as a precaution. It is interesting to note that our findings pose a significant challenge for policy implementation, since there are significant ethical issues involved with introducing HLA-B*15:02 screening for one group of patients and not for another when their risk level is identical. At this stage, no final decision has been made by the SCBP, and we believe that this evidence needs to be considered alongside other priorities, for example, equity grounds, ethical principles, and legal issues.

In 2012, a research study examining the cost-effectiveness of HLA-B*15:02 screening in newly diagnosed patients with epilepsy in Singapore from the perspective of the health care provider was published (Dong et al., 2012). The study used the HRQoL of burn patients as a proxy measurement for the HRQoL of SJS/TEN patients. They compared three similar treatment strategies, as we did in this study, and found that prescribing VPA without HLA-B*15:02 screening is not a cost-effective choice. However, in contrast to our findings, their study found that a universal HLA-B*15:02 screening would be cost-effective. This may be explained by the fact that a higher ceiling threshold (50,000 USD) and PPV (5.96%) were used in the Singapore study.

To the best of our knowledge, there is no published literature on the cost-utility analysis of HLA-B*15:02 screening for preventing CBZ-induced SJS/TEN in patients with neuropathic pain. The results of this study can be utilized as information to support decision making in other settings regarding the adoption of this pharmacogenetic test, given the increased concern over the safety of CBZ. On the other hand, if decision makers in other settings prefer local data, information derived from this study can also be employed for other economic evaluation studies. For example, data on the effect on HRQoL in patients with SJS/TEN has yet to be reported, and the primary data on HRQoL collected in this study could be used to assess the value for money of the test in other settings.

However, the results of this study should be interpreted with caution. First, because of the rarity of cases of SJS/TEN and the rigorous inclusion criteria, only a small number of patients were used in our evaluation, representing both cost and utility limitations. A larger number of patients with a wider variability in the severity of their SJS/TEN hypersensitivity would be more likely to ensure that the results accurately reflect the current situation among all patients. Second, no active surveillance system is in place to quantify the prevalence of CBZ-induced SJS/TEN in the Thai population, and this study employed data from only one study, which was conducted in a medical school in Bangkok. Therefore, further research on this topic would enhance the generalizability of the results and provide more valid and reliable evidence for applying the data to the study of the cost-effectiveness of HLA-B*15:02 screening policies for patients set to undergo CBZ treatment.

Acknowledgments

The Health Intervention and Technology Assessment Program (HITAP) is funded by the Thailand Research Fund under the Senior Research Scholar on Health Technology Assessment (RTA5580010), the National Health Security Office, the Thai Health Promotion Foundation, the Health System Research Institute, and the Bureau of Health Policy and Strategy, Ministry of Public Health. The findings, interpretations, and conclusions expressed in this article do not necessarily reflect the views of the funding agency. The authors would like to thank the Health Product Vigilance Center of the Thai Food and Drug Administration for providing the data derived from health product surveillances, and also Assoc. Prof. Dr. Wichittra Tassaneeyakul for providing the patient information. We would also like to express our appreciation for the standardized instrument used to measure HRQoL supported by the EuroQol Group. Finally, we acknowledge the input from all of the patients and health care professionals at the 10 hospitals that participated in this study.

Disclosure

Surakameth Mahasirimongkol is employed by DMSc, public health laboratory services under the Ministry of Public Health. He has received funding for research projects related to pharmacogenomics of severe cutaneous adverse drug reactions, but unrelated to the data presented in this article. The other authors declare no conflicts of interest. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Ancillary