AN ECONOMIC EVALUATION OF HIV-ASSOCIATED FACIAL LIPOATROPHY TREATMENTS: A COST-UTILITY ANALYSIS

Transcription

1 AN ECONOMIC EVALUATION OF HIV-ASSOCIATED FACIAL LIPOATROPHY TREATMENTS: A COST-UTILITY ANALYSIS by Sirianong Peyasantiwong A thesis submitted in conformity with the requirements for the degree of Master of Science in Health Services Research Graduate Department of Health Policy, Management, and Evaluation University of Toronto Copyright by Sirianong Peyasantiwong 2009

2 An Economic Evaluation of HIV-associated Facial Lipoatrophy Treatments: A Cost-utility Analysis Abstract Sirianong Peyasantiwong Master of Science in Health Services Research Graduate Department of Health Policy, Management and Evaluation University of Toronto 2009 Introduction: Facial lipoatrophy is a stigmatizing hallmark for HIV-positive status, and can lead to poor social functioning. Information gleaned from an economic evaluation of facial lipoatrophy treatments would inform policy decision making concerning potential public insurance coverage. Methods: A decision-analytic model was used to estimate the lifetime costs and Quality Adjusted Life Years (QALYs) gained from treatments using either poly-l-lactic or and polyalkylimide gel for HIV positive patients. Disease progression probabilities and utilities were derived from the literature. Costs were obtained from interviews with physicians and product distributors. Findings: Incremental costs per QALY were $66,608 CAD/$57,352 CAD for poly-l-lactic acid, and $48,583 CAD/$45,457 CAD for polyalkylimide gel (Societal perspective/ministry of Health perspective). Sensitivity analysis did not have a significant effect on the lower incremental costs per QALY reported for polyalkylimide gel. Conclusion: Our base-case analysis revealed that treatments using polyalkylimide gel offers lower ICUR than treatments using poly-l-lactic acid. ii

3 Acknowledgments Special thanks go to my supervisor, Prof. Peter Coyte, for his tremendous support and guidance. I appreciate all his valuable suggestions toward the completion of this thesis. I also wish to express my appreciation to the advisory committee, Prof. Mona Loutfy, for her support and her guidance in clinical issues. My thanks also go to Prof. Audrey Laporte, for providing insightful advice. I would like to thank Profs. Ahmed Bayoumi and Murray Krahn for their helpful suggestions on utility calculation and modeling techniques. Finally, I am indebted to my colleagues, Kednapa Thavorn, Amy Hsu, Debbie Samek, and Lisa Masucci, for their contribution in the preliminary stages of this thesis. iii

4 Table of Contents Abstract... ii Acknowledgement... iii Table of Contents... iv List of Tables... vi List of Figures... viii Chapter 1: Introduction... 1 Chapter 2: HIV-associated facial lipoatrophy Introduction Overview of HIV-associated facial lipoatrophy Treatment options A systematic search of clinical trials Clinical trials of treatments using poly-l-lactic acid Clinical trials of treatment using polyalkylimide gel Chapter 3: Method Introduction Overview of the study design Decision Analytic Model Input Parameters of the Model Sensitivity Analysis Chapter 4: Study Results Introduction Costs and Health Outcomes Analysis Sensitivity Analyses Conclusion iv

5 Chapter 5: Discussion Introduction Results Limitations Policy Implications References v

6 List of Tables Table 1: Summary of published results from clinical trials for Poly-L-Lactic Acid Table 2: Summary of published results from clinical trials for Polyalkylimide Gel Table 3: The probability of effectiveness of treatments using Poly-l-lactic Acid Table 4: The 1-year Probability of Polyalkylimide Gel Infection Table 5: The Probability of Poly-l-lactic Reinjection Between 6 th and 24 th Months Table 6: The 1-year Probability of Reinjection of Poly-l-lactic Acid for Each Cycle Table 7: Probability Transitions in Markov Model Table 8: The CD4 level at the start of Markov process from clinical trials Table 9: The severity distribution of HIV patients with facial lipoatrophy Table 10: Quantity of fillers used for facial lipoatrophy treatments Table 11: Physicians Quotes for Treatment Using Poly-L-Lactic Acid and Polyalkylimide Gel Table 12: Cost components of Poly-L-Lactic Acid and Polyalkylimide Gel Table 13: Costs of antibiotic medication for Polyalkylimide Gel injection Table 14: Estimated Taxi fares from patients' residences to clinics Table 15: Estimated hourly range of HIV-positive patients based on annual income Table 16: Utility gained from either poly-l-lactic acid or polyalkylimide gel injection Table 17: Cost components of Poly-L-Lactic Acid in 2008 Canadian Dollars Table 18: Cost components of polyalkylimide gel in 2008 Canadian Dollars Table 19: Present value of total costs of poly-l-lactic acid and polyalkylimide gel with varying severity of facial lipoatrophy in 2008 Canadian Dollars Table 20: Parameters tested in sensitivity analyses Table 21: Results of one-way sensitivity analyses that adopt Ministry of Health perspective for treatments using poly-l-lactic acid vi

7 Table 22: Results of one-way sensitivity analyses that adopt societal perspective for treatments using poly-l-lactic acid Table 23: Results of one-way sensitivity analyses that adopt Ministry of Health perspective for treatments using polyalkylimide gel Table 24: Results of one-way sensitivity analyses that adopt societal perspective for treatments using polyalkylimide gel vii

8 List of Figures Figure 1: Flow Chart Indentifying Studies... 6 Figure 2: Simplified decision tree Figure 3: The decision tree of HIV-associated facial lipoatrophy treatments Figure 4: State Transition Diagram of the Markov Model Figure 5: Tornado diagram presenting results of one-way sensitivity analyses that adopt Ministry of Health perspective for treatments using poly-l-lactic acid Figure 6: Tornado diagram presenting results of one-way sensitivity analyses that adopt societal perspective for treatments using poly-l-lactic acid Figure 7: Tornado diagram presenting results of one-way sensitivity analyses that adopt Ministry of Health perspective for treatments using polyalkylimide gel Figure 8: Tornado diagram presenting results of one-way sensitivity analyses that adopt societal perspective for treatments using polyalkylimide gel Figure 9: Two-way sensitivity analysis of utility gained from poly-l-lactic acid and physician quotes for polyalkylimide gel for analysis from the Ministry of Health perspective Figure 10: Two-way sensitivity analysis of utility gained from poly-l-lactic acid and physician quotes for polyalkylimide gel for analysis from the societal perspective Figure 11: Two-way sensitivity analysis of utility gained from poly-l-lactic acid and quantity of poly-l-lactic acid required for reinjection for analysis from Ministry of Health perspective..66 Figure 12: Two-way sensitivity analysis of utility gained from poly-l-lactic acid and quantity of poly-l-lactic acid required for reinjection for analysis from societal perspective viii

9 Figure 13: Two-way sensitivity analysis of physician quotes for polyalkylimide gel and a 1 year probability of polyalkylimide gel infection for analysis from the Ministry of Health perspective Figure 14: Two-way sensitivity analysis of physician quotes for polyalkylimide gel and a 1 year probability of polyalkylimide gel infection for analysis from the societal perspective Figure 15: incremental cost-utility isocontour of treatments using poly-l-lactic acid, compared to treatments using polyalkylimide gel for analysis from the Ministry of Health perspective Figure 16: incremental cost-utility isocontour of treatments using poly-l-lactic acid, compared to treatments using polyalkylimide gel for analysis from the societal perspective Figure 17: cost-effectiveness acceptability curve of treatments using poly-l-lactic acid and treatments using polyalkylimide gel for analysis from the Ministry of Health perspective Figure 18: cost-effectiveness acceptability curve of treatments using poly-l-lactic acid and treatments using polyalkylimide gel for analysis from the societal perspective ix

10 1 Chapter 1 Introduction HIV-associated facial lipoatrophy is characterized by volume depletion caused by fat loss in the cheeks, temples, and orbits. Numerous studies have shown that facial lipoatrophy is one of the most distressing and socially stigmatizing side effects of treatment with highly active antiretroviral therapy (HAART). These effects occur because the visibility of this condition forces involuntary disclosure of one s HIV status which may affect an individual s ability to maintain social relationships without fear of being stigmatized. In fact, people with facial lipoatrophy often experience a reduction in self esteem and self confidence, resulting in poor social functioning and social isolation (Power et al. 2003). These psychosocial effects have been linked to non adherence to antiretroviral therapies (ARTs), and fear of developing facial lipoatrophy can discourage individuals who are HIV-positive from initiating antiretroviral therapy (Montessori 2004; Phillips and Hay 2008). Since facial lipoatrophy tends to be an irreversible condition, clinicians and patients with HIV/AIDS have experimented with cosmetic surgery. Clinical trials of pharmacological interventions demonstrate mixed results with only temporary benefits (Phillips and Hay 2008) (Carr et al. 2004). Reconstructive cosmetic techniques, such as injection of permanent fillers or semi-permanent agents that can stimulate the body to produce its own collagen, are effective and preferred by patients with facial lipoatrophy (Nelson and Stewart 2007; Orlando et al. 2007; Phillips and Hay 2008). From several clinical trials, both poly-l-lactic acid and polyalkylimide gel are fillers that offer effectiveness and safety to people with HIV/AIDS (Protopapa et al. 2003; Valantin et al. 2003; Moyle and Sutinen 2004; Borelli et al. 2005; Burgress and Quiroga 2005; Lafaurie et al. 2005; Mest and Humble 2006; Moyle et al. 2006; Treacy and Goldberg 2006; Carey et al. 2007; Hanke and Redbord 2007; Lahiri and Waters 2007; Loutfy et al. 2007; Ramon et al. 2007; Honig 2008; Karim et al. 2008; Kavouni et al. 2008). A number of studies have shown significant improvement in quality of life reported by HIV patients after receiving facial lipoatrophy treatments using poly-l-lactic acid and polyalkylimide gel. However, the cost of these treatments can be prohibitively high for some patients. Despite being approved and indicated for the treatments of HIV-associated facial lipoatrophy, the use of

11 2 these products is still considered elective cosmetic surgery, and therefore, not publicly insured in Ontario, Canada. The aim of this thesis is to conduct an economic evaluation of HIV-associated facial lipoatrophy treatments using poly-l-lactic acid and polyalkylimide gel. A cost-utility analysis is employed to estimate the cost per quality adjusted life year (QALY) gained attributable to HIV-associated facial lipoatrophy treatments. Information gleaned from an economic evaluation of treatments for HIV-associated facial lipoatrophy may inform policy decision making concerning potential public insurance coverage.

12 3 Chapter 2 HIV-associated Facial Lipoatrophy 2.0 Introduction This chapter provides a comprehensive literature review of HIV-associated facial lipoatrophy. In particular, a systematic review of clinical trials using poly-l-lactic acid and polyalkylimide gel to correct HIV-associated facial lipoatrophy is presented because these two fillers are commonly used in the province of Ontario, Canada and are the subject of the cost-utility analysis reported in this thesis. This chapter is divided into two sections. The first section presents an overview of HIV-associated facial lipoatrophy. In that Section, facial lipoatrophy is defined and its impact on the lives of HIV patients is described. In the subsequent section, clinical trials of HIV-associated facial lipoatrophy treatments using poly-l-lactic acid and polyalkylimide gel are reviewed. 2.1 Overview of HIV-associated facial lipoatrophy Since 1998, researchers in the field of treatments for HIV have identified a link between antiretroviral regimens and lipodystrophy syndrome (James et al. 2002). HIV-associated lipodystrophy syndrome is broadly divided into three subcategory of lipohypertrophy (regionalised fat accumulation), peripheral lipoatrophy (localised subcutaneous fat loss including facial lipoatrophy), and patients with a mixed pattern of fat accumulation (hypertrophy) and fat wasting (lipoatrophy). The aetiology and pathogenesis of HIV lipodystrophy are complex and multifactoral (Nelson and Stewart 2007). Many studies have shown that the nucleoside reverse transcriptase inhibitors (NRTIs), particularly Stavudine (D4T), are most strongly associated with subcutaneous fat wasting (John 2002). A meta-analysis of nine studies by Lichtenstein revealed the following risk factors in descending order of importance: treatment with D4T, low CD4/high viral load, old age, longer duration of treatment, white ethnic, and treatment with protease inhibitors (Waters and Nelson 2007). Facial lipoatrophy was cited as an important stigmatizing factor for HIV positive patients, rendering them more vulnerable to the identification of HIV status (Fernandes et al. 2007). The visibility of this condition forces involuntary disclosure of one s HIV status, affecting the

13 4 individual s ability to maintain social relationships without fear of being stigmatized (Power et al. 2003). Therefore, HIV positive patients with facial lipoatrophy significantly made more attempts to solve problems due to body change than did patients without facial lipoatrophy (Blanch et al. 2004). In addition, facial lipoatrophy was the body region associated with the most severe depression scores compare to lipohypertrophy and lipodystrophy with other body region such as neck, chest, waist, belly, arms, legs, and buttocks (Crane et al. 2008). 2.2 Treatment options There are many options for treating HIV-associated facial lipoatrophy such as autologous fat transfer, switching from lipoatrophy-associated antiretroviral, soft tissue augmentation, and cosmetic surgery (Engelhard 2006). Clinical trials of pharmacological interventions demonstrate mixed results with only temporary benefits (Carr et al. 2004; Waters and Nelson 2007). Although autologous fat injection would be the safest and most biocompatible option for treating facial lipoatrophy, many patients with HIV-associated lipoatrophy have limited fat reserves, rendering this option unsuitable in the long-term (Nelson and Stewart 2007). Since clinical trials suggested that HIV-associated facial lipoatrophy tends to be an irreversible condition, patients with HIV and clinicians have experimented with cosmetic surgery (Waters and Nelson 2007). Reconstructive cosmetic techniques, such as injection of permanent fillers or semi-permanent agents that can stimulate the body to produce its own collagen, are more effective and preferred by patients with facial lipoatrophy (Nelson and Stewart 2007; Orlando et al. 2007; Phillips and Hay 2008). Various types of fillers are available for HIV-associated facial lipoatrophy treatments (Engelhard 2006). To illustrate, silicone oil, Artecoll, Radiesse, and polyalkylimide gel offer permanent effects lasting more than 5 years while poly-l-lactic acid and calcium hydroxylapatite are semi-permanent fillers whose effect may potentially last up to 40 months (Engelhard 2006). In addition, bovine collagen, Hyaluronic acid, autogenic human collagen, and Isogenic human collagen offer temporary effects of less than one year. As the main fillers used for HIVassociated facial lipoatrophy correction in Ontario, Canada are poly-l-lactic acid and polyalkylimide gel (Loutfy 2008), the following section will focus on only these two kinds of cosmetic surgery to treat HIV-associated facial lipoatrophy.

14 5 2.3 A systematic search of clinical trials using poly-l-lactic acid and polyalkylimide gel A systematic search for previously published studies associated with treatments for HIVassociated facial lipoatrophy by poly-l-lactic acid and polyalkylimide gel was conducted using the EMBASE (1980 to 2009), PubMed MEDLINE (1996 to 2009), IBSS (1951 to 2008), and CINAHL (1981 to 2008) databases. Search terms and their affiliated truncations, used singularly and in conjunction, included HIV-associated facial lipoatrophy, lipoatrophy, lipodystrophy, facial correction, facial wasting, HIV-related lipodystrophy syndrome, sculptra, bio-alcamid, poly-l-lactic acid and polyalkylimide gel. The inclusion criteria were: 1) treatments that were administered to treat HIV-associated facial lipoatrophy. Exclusion criteria, included: 1) literature reviews; 2) trials not published in English; and 3) trials where treatments for lipoatrophy were due to aging or for aesthetic purposes unrelated to HIV. We included trials that consisted of both HIV positive and negative subgroup if the treatment effects and complications were reported separately for each subgroup. We screened 33 abstracts and excluded 16 studies because 13 studies were literature reviews, 2 clinical trials were not administered to treat HIV-positive patients, and 1 study was published in French. After screening 17 full papers, we accepted all papers as they met our inclusion criterion. Categorizing by study designs, these trials consisted of 4 randomised controlled trial, and 13 uncontrolled trials. According to the hierarchy of evidence, randomised control trials are deemed to be good since they minimize bias and error (Evans 2003). Un-controlled trials are ranked at a lower level because of greater risk of bias and error. As only 4 randomised control trials were available, we also included all un-controlled trials using poly-l-lactic acid and polyalkylimide gel for HIV-associated facial lipoatrophy correction. Classifying by type of study, we have 16 perspective cohorts and 1 retrospective cohort. The retrospective cohort study is un-controlled trial. Figure 1 illustrated the process of identifying studies eligible for inclusion criteria. In the subsequent section, HIV-associated facial lipoatrophy using poly-l-lactic acid and polyalkylimide gel are described.

15 6 Figure 1: Flow Chart Indentifying Studies Potentially relevant publications identified and abstracts screened (n = 33) Excluded because Literature review (n = 13) Not English (n = 1) Trial population was not HIV patients (n = 2) Full paper screened (n = 17) Poly-l-lactic acid (n = 10) - Randomized Control Trials (n = 3) - Un-controlled Trials (n = 7) - Retrospective cohort (n=1) Polyalkylimide gel (n = 7) - Randomized Control Trials (n = 1) - Un-controlled Trials (n = 6)

16 7 2.4 Clinical trials of treatments using poly-l-lactic acid Poly-l-lactic acid is biodegradable, bioabsorbable aliphatic polyester produced by carbohydrate fermentation of corn dextrose (Burgress and Quiroga 2005). The lyophilizate is composed of poly-lactic acid microshperes suspended in an apryrogenic mannitol and sodium cellulose preparation. Poly-l-lactic acid has been used for facial aesthetic enhancements over a decade by more than 150,000 patients in 30 countries and its safety is well-established (Moyle and Sutinen 2004). However, poly-l-lactic acid was primarily used for aesthetic enhancements not related to HIV (e.g. wrinkles, creases and folds to reduce visible signs of aging), and its efficacy in treating HIV-related facial lipoatrophy has not been studied for more than three years. Health Canada s Health Protection Branch approved Sculptra (or poly-l-lactic acid) marketed by Dermik Laboratories Canada Inc. for treating HIV-associated facial lipoatrophy in June, 2006 (Sanofiaventis Canada Inc 2007). Since poly-l-lactic acid is hypoallergenic, bioresorbable, biodegradable, and immunologically inert, it cannot cause an inflammatory reaction (Moyle and Sutinen 2004). Poly-l-lactic acid does not restore lost fat, but expands dermal thickness through an increase in fibroblasts and then a subsequent increase in the deposition of collagen fibers (Moyle and Sutinen 2004). Once injected, poly-l-lactic acid has two stages of action (Kavouni et al. 2008). The initial stage is a temporary volumising effect in the skin of the injected area. The latter stage is to stimulate collagen production overtime Clinical Trials: Poly-l-lactic acid There are 10 studies that met the inclusion and exclusion criterion of examining the use of polyl-lactic acid in the treatment of HIV-associated facial lipoatrophy. These 10 clinical trials consisted of 3 randomised controlled trials, 7 un-controlled trials, and 1 retrospective study. Specifically, Moyle et al. (2004) and Carey et al. (2007) conducted a randomised control trial to evaluate the temporal association between treatments and improvement in the objective and subjective outcomes by randomising subjects into two groups: intermediate group and delayed group. The intermediate group received a poly-l-lactic acid injection at week 0 while the delayed group received it at week 12 in Moyle et al.(2004) or at week 24 in Carey et al.(2007). Moyle et al. (2006) followed HIV-patients for 2 years and published the long-term efficacy and safety of poly-l-lactic acid in Of 10 studies, 6 studies are un-controlled trials conducted by

17 8 recruiting HIV patients with facial lipoatrophy and injecting poly-l-lactic acid to all patients at week 0. They investigated treatment effectiveness and adverse reactions at the follow up period, which was typically at week 12, 24, 48, 72, and 96 (Valantin et al. 2003; Borelli et al. 2005; Burgress and Quiroga 2005; Lafaurie et al. 2005; Mest and Humble 2006; Hanke and Redbord 2007). In addition, Kavouni et al. (2008) conducted a retrospective cohort study including 441 subjects. Overall, studies had a limited number of subjects ranging from 28 to 100 adult HIV patients. Men accounted for the overwhelming proportion of study subjects. Follow-up time after the last treatment ranged from 24 weeks to 3 years. The details of each study are summarized in Table 1. To obtain data for estimating our model parameters in this thesis, we paid particular attention to quantity of poly-l-lactic acid used to address facial wasting, treatment effects, and complication Quantity of poly-l-lactic acid used These clinical trials reported the amount of poly-l-lactic acid into two different units: vial and millilitres (ml). According to the process used to prepare poly-l-lactic acid, 1 vial of poly-llactic acid (15 gram) can reconstitute to between 1.5 ml and 5 ml depending on the addition of sterile water and plain lidocaine (local anesthetic), which varied between studies. For example, Moyle et al. (2006) reconstituted 1 vial of poly-l-lactic acid by 2 ml of sterile water and 1 ml of 2% lidocaine to give a total volume of 3 ml, while Hanke and Redbord (2007) added 3 ml of sterile water and 2 ml of 1% lidocaine to give a total volume of 5 ml. By adding sterile water and plain lidocaine more than other studies, Hanke and Redbord (2007) claimed that this-5 ml dilution decreased complication rates- the risk of subcutaneous papule formation and contributed to ease of injection. Most studies did not stratify the quantity of poly-l-lactic acid used by the severity of facial lipoatrophy except Burgress and Quiroga (2005), Mest and Humble (2006) and Kavouni et al (2008). However, they did not employ a consistent grading scale for the severity of the underlying condition. Burgress and Quiroga (2005) used their proposed 4-point grading scale, while Kavouni et al (2008) used a 3-point scale. In contrast, Mest and Humble (2006) stratified the amount of polyalkylimide gel used by the severity of HIV-associated facial lipoatrophy, using a 4-point facial lipoatrophy severity scale developed by Dr. Carruthers (James et al. 2002). According to Mest and Humble (2006), the median of 6 vials is required for subjects with grade

18 9 1 (mild and localized facial lipoatrophy) while those with grade 2 (deeper and longer lipoatrophy with the facial muscles beginning to show through) required the median of 8 vials. The amount required for patients with grade 3 (deeper and wider atrophic area with the muscles clearly showing) and grade 4 (lipoatrophy covering a wide area, extending up towards eye sockets, and facial skin lying on the muscles) are 10 vials, and 12 vials, respectively. Overall, the number of treatment sessions ranged from three to six depending on severity. Typically, HIV patients received 1-2 vials per session. For example, Valantin et al. (2003) and Moyle et al. (2004) gave 3 treatment sessions with 1 vial (3-4 ml) per session. Mest and Humbles offered 1-6 treatments with the maximum of 2 vials per Session. Carey et al. (2007) recruited only HIV patients with severe facial lipoatrophy and gave 4 treatments with 2 vials of poly-l-lactic acid per session. The other studies neither reported the number of treatment sessions nor the quantity of poly-l-lactic acid used per session. The treatment session are usually spaced at intervals of two weeks (Valantin et al. 2003; Moyle and Sutinen 2004; Lafaurie et al. 2005; Carey et al. 2007). However, three studies offered treatments every 4-6 weeks (Borelli et al. 2005; Burgress and Quiroga 2005; Hanke and Redbord 2007). In addition, Mest and Humble (2006) spaced treatment sessions every 3 weeks. Although Borelli et al. (2005) suggested 2-3 weeks interval between treatments to obtain the optimal result, their study offered treatment in 4-6 weeks interval Treatment effects using poly-l-lactic acid Poly-l-lactic acid treatments in patients with HIV have been favourable yielding decreased severity of facial lipoatrophy measured by both objective and subjective instruments. By objective measurement, Moyle et al. (2004) used ultrasound and observed a mean increase of 4-5 mm in dermal thickness in the cheek and nasolabial regions at week 12 after treatment. Valantin et al. (2003) also used ultrasound to measure skin thickness and discovered a median increase in skin thickness at the 6 th week (5.1 mm), the 24 th week (6.4 mm), the 48 th week (7.2 mm), the 72 nd week (7.2 mm), and the 96 th week (6.8 mm). Similarly, Lafaurie et al (2005) found an increase in dermal thickness at 1.9 mm after the end of treatment and up to 2.3 mm at the last follow-up visit (12 th month) by using three-dimensions photographs. Also, all patients experienced an increase in skin thickness measured by using skin calipers in the study of Mest and Humble (2006), and using 10 MHz sonography in the trial of Borelli et al. (2005). In

19 10 contrast, Carey et al. (2007) did not detect an increase in facial soft tissue volume assessed by spiral computed tomography. However, most patients perceived an improvement in facial lipoatrophy severity after treatment. By subjective measurement, Burgress and Quiroga (2005) assessed the effectiveness of poly-llactic injection by a 4-point visual evaluation scale and found that both patients and physicians agreed on the effectiveness of dermal enhancement treatments in achieving a healthy appearance at the 2 nd month and 8 th month after the final treatment. Likewise, Kavouni et al. (2007) observed a gradual volume restoration in the 2 nd week after the first injection and was viewedas optimal by 90 per cent of patients by 3 rd months after the last treatments session. By using self-assessed patient visual analogue scale for lipoatrophy, Lafaurie et al (2005) noticed an increase by 3.4/10 from the baseline. Interestingly, Hanke and Pagliai (2007) injected poly-l-lactic acid to correct facial lipoatrophy due to both HIV and aging. They discovered that poly-l-lactic acid was effective for both subgroups as all patients significantly improved on the facial lipoatrophy grading scale and all patients reported high satisfaction. In the randomized control trials, Moyle et al. (2004) found that an increase in dermal thickness in cheek and nasolabial areas by ultrasound in week 12 th was significantly higher for the immediate group compared to those in the delayed group measured. Similarly, Carey et al, 2007 reported a significantly higher mean change in tissue depth at the maxilla and the base of the nasal septum between the immediate and delayed groups at week 24measured by CT scan and 3-dimensional post-processing software. Only 4 clinical trials reported the success rate of using poly-l-lactic acid (Burgress and Quiroga 2005; Lafaurie et al. 2005; Carey et al. 2007; Hanke and Redbord 2007). Lafaurie et al. (2005) detected the success rate of 82 per cent of patients (77/94) where the success rate was defined as the proportion of patients whose self-perceived visual analogue scale scores increased at the end of the treatment procedure compared to baseline, with patients lost to follow up and prematurely discontinuing the procedure considered as failure. As this study included patients with loss to follow-up as failure, the actual success rate may be higher than 82 per cent. In contrast, the success rate in Carey et al (2007), was 90 percent (45/50) at week 12 and 84 per cent (42/50) at week 24 - for patients in the immediate group compared to 18 per cent of those in the delayed group. When measured by physicians, there was a perceived reduction in facial lipoatrophy at

20 11 week 12 -, but when measured by both physicians and patients, these reductions occurred at week 24. The 100 per cent success rates were found in 2 studies (Burgress and Quiroga 2005; Hanke and Redbord 2007). According to Burgress and Quiroga (2005), all patients and physicians agreed that the response was excellent at the 6 - month of follow-up. Similarly, Hanke and Redbord (2007) discovered that all patients experienced facial lipoatrophy improved significantly on the grading scale judged by photograph. The other studies reported the mean or median increase in skin thickness without reporting the proportion of patients with success. The factors that determine the success rate of correction in HIV-associated facial lipoatrophy differed across studies. According to Lafaurie et al. (2005), patients with success did not differ from patients with failure in their baseline characteristics such as age, sex, ethnicity, CD4 cell count, and severity of the lipoatrophy. The only difference between those two groups was in the higher median number of injections associated with success (5 versus 4). In contrast, Borelli et al (2005) measured the treatment effect of poly-l-lactic acid using photography and claimed that patients with severe facial lipoatrophy responded to poly-l-lactic acid injection less well. Moreover, all patients with failure to correct facial lipoatrophy by using poly-l-lactic acid were judged as having high degree of facial lipoatrophy severity at the beginning of study. According to these trials, HIV patients quality of life was improved after receiving poly-l-lactic acid injection. Specifically, both Lafaurie et al (2005) and Carey et al (2007) used the SF-36 questionnaire to assess this improvement. While Lafaurie et al (2005) found no improvement in both mental and physical quality of life scores at the end of the study or at the last follow-up visit (1 year) compared to the baseline, Carey et al (2007) reported that the mean change from baseline in the Mental Health scores in the immediate groups significant differed from the mean changes in the deferred group at week 12 and 24. Also, the mean change in the social functioning score was also significant differed at week 24. Similarly, Valantin et al. (2003) reported a progressive increase from baseline to week 48 - measured by the Visual Analogue Scale. Poly-l-lactic acid injection for HIV patients also leads to a reduction in anxiety and depression (Moyle et al. 2006; Kavouni et al. 2008). Moyle et al. (2004) noticed a decline in the mean of anxiety and depression scores as assessed by the Hospital Anxiety and Depression (HAD) score after receiving poly-l-lactic acid injection. Also, there was a difference in HAD scores between immediate and delayed groups at week Furthermore, Kavouni et al (2008) also discovered a

21 12 significant improvement from baseline in depression and anxiety scores at the 3 rd month after treatment in their retrospective cohort study while with 441 patients. As poly-l-lactic acid is a semi-permanent filler, the treatment effect (i.e. an increase in dermal thickness) ranges from 6 months to 2 years. Moyle (2006) showed that the positive results, both a reduction in HAD and an increase in self-perceived dermal thickness as assessed by the visual analogue scale persisted for up to 2 years. Likewise, Valantin et al (2003) noted that a significant improvement in dermal thickness was sustained until the last follow up (96 th week). In contrast, Burgress and Quiroga (2005) reported that only 8 per cent of patients retained dermal thickness for a duration of 2 years. From their experience, an average significant improvement and dermal enhancement was maintained for about 18 months. In addition, studies with only 1 year followup found that an increase in skin thickness was sustained until the last follow-up visit (Lafaurie et al. 2005; Mest and Humble 2006) Complications of poly-l-lactic acid Despite the effectiveness of poly-l-lactic acid injection, these treatments for HIV-associated facial lipoatrophy are known to have several complications. A majority of patients developed palpable, inflammatory but generally invisible subcutaneous nodules after injection of poly-llactic acid in most clinical trials (Valantin et al. 2003; Burgress and Quiroga 2005; Lafaurie et al. 2005; Mest and Humble 2006; Hanke and Redbord 2007; Kavouni et al. 2008). Although nodules are generally invisible, three patients reported occasionally visible nodules (Moyle et al. 2006; Carey et al. 2007). Also, papules occurred in 4 trials (Burgress and Quiroga 2005; Mest and Humble 2006; Carey et al. 2007; Hanke and Redbord 2007). In general, the onset of these papules occurred between three to four months after initial treatment (Mest and Humble 2006). These subcutaneous nodules may resolve completely over the next 6 months without intervention, remain the same size, or slightly decrease over the course of (1 year) (Mest and Humble 2006). In addition to nodule formation, a common complication of poly-l-lactic acid injection reported in four studies is minimal and localized oedema, which spontaneously resolved within hours without treatment (Valantin et al. 2003; Lafaurie et al. 2005; Carey et al. 2007; Kavouni et al. 2008). Bruising occurred in three studies (Moyle and Sutinen 2004; Lafaurie et al. 2005; Mest and Humble 2006). The adverse events reported in one study were minimal ecchymosis which spontaneously resolved within 2-3 days (Valantin et al. 2003),

22 13 erythema tenderness (Kavouni et al. 2008), swelling (Mest and Humble 2006), limited superficial local cellulitis which did not require antibiotic therapy (Moyle and Sutinen 2004), and pain (Carey et al. 2007). These adverse events may not have been reported in other studies either because they either they did not occur or they may have been considered to be clinically insignificant. In the trial of Moyle et al (2006), in which long-term follow up was conducted, thirty six adverse events were considered to be related to poly-l-lactic acid injection, including invisible nodules, bruising, discomfort, erythema, haemorrhage, inflammation, oedema, infection, tenderness, and skin lesions. None of these events were severe and all were localized to the injection site and cheek area. Interestingly, Borelli et al. (2005) is the only study that neither found nodule formation nor other complications related to poly-l-lactic acid injection. The possible reasons were either small sample size (n = 14) or their injection technique: increasing the amount of sterile water and injecting more deeply into subcutaneous tissue.

23 14 Study Moyle et al, 2004 Moyle et al, 2006 (longterm follow up) Table 1: Summary of published results from clinical trials for Poly-L-Lactic Acid General Information London, UK 28 males, 2 females Mean Age 41 Follow up 24 months Same as above Follow up 3 years Quantity of polyl-lactic acid used 3 treatments 2 weeks apart No quantity used reported Same as above Main findings Mean increase in dermal thickness of 4-5 mm observed at week 12 by ultrasound Selfassessment visual analogue scores increased after 12 weeks Patient s selfperceived facial thinness was significantly more positive at the recall visit than at baseline. Adverse Events & reinjection Brusing (0.03%) Limited superficial local cellulitis (0.03%) (do not required treatments) Over 2 years, no serious side effects were reported. Some patients had palpable but invisible nodules. The positive results of PLA noted at 24 weeks persisted up to 2 years post treatment. Quality of life and psychosocial Anxiety and depression scores by Hospital Anxiety Depression Scores significantly declined after the treatment. The decline in Anxiety and depression scores was sustained at the recall visit. Valantin et al, 2003 Paris, France 29 males, 1female Median age 46 Follow up 96 months (all subject had severe facial lipoatrophy) 52% required 4 treatments 48% required 5 treatments Maximum of 4 ml in each cheek per one treatment 2 week interval The median increase in total cutaneous thickness from baseline measured by ultrasound was significant in all weeks. It was highest at weeks 72 and slightly decreased at week 96. No serious adverse events were observed. Some patients developed edema, minimal ecchymosis, non-visible micronodules. The effect persisted until week 96 Quality of life assessed by Visual Analogue Scale were increased from baseline to week 48 (+.3 at week 12, +.8 at week 24 and 48,+.4 at week 72 and 96)

24 15 Burgress and Quiroga (2005) Lafaurie et al (2005) Mest and Humble (2006) Washington DC, USA 61 males Mean age 45.5 Mean years on HAART 7.18 Follow up 2 years Paris, France 88 males 6 females Median age 44 Lipoatrophy mild 34 moderate 40 severe 20 Follow up 1 year US 97 males 2 females Mean age 45 Mean years on ARTs 9.1 Lipoatrophy by James scale 1=10%, 2=23%, 3=43%, 4=22% (severe) Follow up 1year Treatments every 3 to 6 weeks over 5 months. Grade1 used 1-2 vials, Grade 2 used 2 vials, Grade 3 used 3-4 vials, Grade 4 used 4 vials per cheek Median 5 injections per cheek per patient Median volume of PLA 2.5 ml and 2.4 ml for the right and left cheek. 2% of patients required 2 treatments, 16% required 3 treatments, 17% required 4 treatments, 20% required 5 treatments, 43% required 6 treatments. The median of PLA used 6 ml. (2 vials) per each treatment 3 week interval At the 6-month follow up evaluation, 100% of patients and physicians agreed to excellent responses. Self-assessed patient median VAS score and dermal thickness assessed by 3D pictures are significantly improved at the end of the treatment. All patients experienced an increase in skin thickness mean of 69% at 6-month period and maintained at 73% at the 12- month period. No incidents of swelling, edema, allergic reactions, infection, erythema, or clinical evidence of adverse reactions. 21% of patients required reinjection at an average of 1 year 1 patients developed a peripheral facial palsy for 10 days for only the third injection, the event was considered not related to the injection of PLA The probability of reinjection 15 months after the first injection is 45 per cent No serious adverse events. Most patients experienced swelling, 30 % of patients experience bruising (only 42 of 402 total treatments are clinically significant), 13.1% experienced small papule formation Not reported. Mental and Physical QoL scores using the SF-36 questionnaire showed no significant changes from baseline and the end of the treatment or the last followup Visit No report

25 16 Carey et al (2007) Hanke et al (2007) Borelli et al (2005) Australia 92 males 8 females Mean age 49 Lipoatrophy moderate 49 Severe 51 Mean year on ARTs 10.5 Follow up 24 weeks United State 27 HIV patients Mean age 48.9 Follow up not report Germany 10 male 2 female Mean age 51.5 Follow up not report 4 treatments two weeks apart The amount of PLA used 1 vial (150 mg) per cheek for each treatment 2 week interval Mean 4.6 treatments with 4-6 weeks interval Mean 8.44 vials Median 7 treatments 4-6 weeks interval FLA severity was perceived as improved at week 12 and week 24 after injection 23.9 weeks to full correction Mean patient satisfaction 4.92/5 The average FLA scores reduced 2.33/5 All patients observed an increase in skin thickness measured by sonography 96% PLA recipients experienced at least 1 productrelated adverse event, with pain (76%), localized edema (64%), erythema(53%). Most events were of short duration and not severe % of HIV patients developed subcutaneous nodule No complication related to poly-llactic acid No nodule formation The mental health scores of SF 36 increased at week 12, the mental health and social functioning scores improved at week 24 Not reported Not reported Kavouni et al, (2008) United Kingdom 411 patients Male 96% Mean age 42 Follow up 32 months (A retrospective study) Patients with mild FLA received 3 treatments, those with moderate FLA received 4 treatments, those with severe FLA received 5 treatments A minimum 5-ml dilution was used and a minimum total of 10-ml was injected. Volume restoration was observed gradually after the second session and was accepted as optimal by 90% of patients at three months post last treatment session. Reported adverse effects during the treatment period were minimal and included injection site erythema, tenderness, oederma, and ecchymosis. The incidence of nodule formation was 2.9% Patients showed significant improvement in depression and anxiety scores assessed by HAD and Appearance Satisfactions Questionnaires at three months post last treatment compare with baseline.

26 Clinical trials of treatments using polyalkylimide gel While poly-l-lactic acid is a semi-permanent filler, polyalkylimide gel (Bio-Alcamid ) is a permanent method for treating facial lipoatrophy. Polyalkylimide gel is a non-biodegradable filler composed of 96% water and 4% polyalkylimide (Protopapa et al. 2003). When injected, polyalkylimide gel becomes covered by a thin layer of collagen which surrounds the gel and isolates it from the host tissues, making it an endoprosthesis: an artificial replacement of a body part placed internally (Lahiri and Waters 2007). Since polyalkylimide gel is a relatively new filler, the prevalence of long term complications have yet to be established. It has been suggested that complications may occur whenever the filler is within the host tissue (Goldan et al. 2007). Polyalkylimide gel has been widely used in Europe since However, prior to being granted approval by Health Canada in April 2006, more than 1,000 Bio-Alcamid treatments were performed in Canada for treating patients with HIV-associated facial lipoatrophy since 2004 under Health Canada s Special Access Program (SAP) (Maple Leaf Medical Clinic 2008). The SAP grants practitioners special access to non-marketed drugs in order to treat patients with serious conditions when conventional therapies have failed, are unsuitable, or unavailable (Canada 2002) Clinical Trials: polyalkylimide gel There were only 7 trials that met the inclusion criteria for polyalkylimide gel as this filler has been available for a shorter period than poly-l-lactic acid (Protopapa et al. 2003; Treacy and Goldberg 2006; Lahiri and Waters 2007; Loutfy et al. 2007; Ramon et al. 2007; Honig 2008; Karim et al. 2008). These studies had sample sizes of between 11 and 73 HIV-infected patients. The majority of subjects were male except in the study by Protopapa et at (2002). In terms of study design, Loutfy et al (2007) randomized between an immediate and delayed group while other studies were un-controlled trials. Even though polyalkylimide gel was injected into 57 anatomical sites, including HIV-associated facial lipoatrophy, the study of Lahiri and Waters (2007) reports separate results to correct facial lipoatrophy. The follow-up period ranged from 1 month to 3 years. The study details are summarized in Table 2. Again, in order to obtain data to estimate our study parameters, we will pay particular attention to the quantity of polyalkylimide used to treat facial wasting, treatment effects, and complications.

27 Quantity of polyalkylimide gel used As polyalkylimide gel does not require reconstitution by the addition of sterile water like poly-llactic acid, the quantity of polyalkylimide gel used is reported in millilitres (ml). The amount of polyalkylimide gel used ranged from 5 to 35 ml, depending on the severity of the underlying condition (Protopapa et al. 2003; Lahiri and Waters 2007; Loutfy et al. 2007; Ramon et al. 2007; Honig 2008; Karim et al. 2008). Specifically, Loufty et al (2007) stratified the amount of polyalkylimide gel used by the severity of HIV-associated facial lipoatrophy, using a 4-point FLSS grading scale criteria developed by Carruther (James et al. 2002). Subjects with grade 1 to 4 required the mean of 11.8, 15.7, 20.8, 29.5 ml, of polyalkylimide gel, respectively. The number of treatment sessions also differed among studies. While Protopapa et al (2008), Karim et al (2008), Treacy and Goldberg (2006), and Ramon et al (2007) offered only one treatment session, Loutfy et al (2007) reported treatment sessions that ranged from one to five depending on the degree of severity Treatment effects of polyalkylimide gel All studies reportedthat the majority of patients were satisfied with the aesthetic results of the treatment. In their randomised controlled trial, Loutfy et al. (2007) found that the immediate group experienced a reduction in facial lipoatrophy severity by two grades out of four grades of physician-perceived facial lipoatrophy scores using the scale developed by Dr. Carruthers while the severity of the delayed group remained the same at week 12. Karim et al (2008) also reported a reduction in facial lipoatrophy severity using six-point scores ranging from very severe to absence of facial lipoatrophy. This reduction in severity of facial lipoatrophy was significant from baseline over 48 weeks. Similarly, Treacy et al (2006), Ramon et al (2007) and Protopapa et al, 2008 reported an excellent result assessed by both patients and physicians at the end of treatment. Apart from uniformly high satisfaction with the correction of HIV-associated facial lipoatrophy by polyalkylimide gel, patients also reported some functional improvement in terms of ease of chewing and shaving (Lahiri and Waters 2007). None of the clinical trials using polyalkylimide gel explicitly reported the proportion of patients with success. However, the success rate can be extracted from some small trials that reported the reduction in facial lipoatrophy individually. From Karim et al. (2008), there were 9 patients with

28 19 severe and very severe lipoatrophy at the baseline, but only 1 very severe patient at the end of the study (week 48). This patient was infected at one injection site necessitating surgical drainage between week 24 - and Without reporting the success rate, Protopapa et al. (2008) reported that the aesthetic results of polyalkylimide gel were deemed excellent by both physicians and patients. Similarly, Treacy et al. (2006) stated that every treated subject noted that their quality of life improved dramatically after treatment, and none of them expressed regret at having had the procedure. After receiving the injection of polyalkylimide gel, Loutfy et al. (2007) and Karim et al. (2008) assessed the improvement in HIV patients quality of life by MOS-HIV. Specifically, Mental Health Summary score was significantly increased in both studies. Also, the social functioning domain improved in both studies but was significant in only Karim et al (2008). The depression and anxiety of HIV patients was also reduced after receiving polyalkylimide injection measured by using various instruments. Loutfy et al (2007) found a significant reduction in both the depression and anxiety from baseline to week 48 assessed using the Hospital Anxiety and Depression scale (HAD). In like manner, Karim et al. (2008) reported a reduction in depression over 48 weeks measured by the Center for Epidemiological Studies Depression scale (CES-D). Finally, Treacy et al. (2006) discovered a significant mean decreased in overall score of the modified Beck depression scores after the procedure Complication of polyalkylimide gel Despite its effectiveness, polyalkylimide gel injection may lead to several potential adverse effects. The clinically important complication associated with polyalkylimide gel injection is infection with organisms such as staphylococcus aureus which may lead to surgical drainage. These infections occurred as a result of an incomplete encapsulation process of polyalkylimide gel before isolating itself from the host tissues and become endoprosthesis (Protopapa 2009). The causes of an incomplete encapsulation process are: improper injection technique which leads to gel contamination, high exposure to UVB, and bacteria from the facial skin during infection. Of 7 clinical trials, infection occurred in 2 trials (Lahiri and Waters 2007; Karim et al. 2008). In the study of Lahiri and Waters (2007) with 15 subjects, one subject was infected in the right temporal area resulting in a small area of skin breakdown and spontaneous discharge of the injected material. According to Karim et al (2008), 1 of 17 HIV patients developed this infection.

29 20 They assumed that infection was caused by either the higher volumes injected at a single stage or their application of a single prophylactic dose of Clarithromicin instead of 3-7 days antibiotic course. However, Protopapa et al (2003) also offered a single stage of polyalkylimide gel injection and gave a week of oral antibiotic course, there was no infection reported in their study. Therefore, these infections might be attributable to a single prophylactic dose of antibiotics. In addition, patients initially experience side effects after injection, such as swelling, pain, bruising and local oedema; however, these side effects normally subsided within three days (Protopapa et al. 2003; Treacy and Goldberg 2006; Lahiri and Waters 2007; Loutfy et al. 2007; Ramon et al. 2007; Honig 2008; Karim et al. 2008).

30 21 Study Loutfy et al, (2007) Protop apa et al, (2003) Table 2: Summary of published results from clinical trials for Polyalkylimide gel General Information Toronto, Canada 30 males 1 female Median age 48 FLA severity: Carruthers scale Grade 1 = 13% Grade 2 = 50% Grade 3 = 31% Grade 4 = 6% Follow up 48 weeks Italy 40 females 33 males Follow up 3 years Quantity of polyalkylimide gel used 58 % of patients required 1 injection or 1 injection with 1 touch up, 29% required 1 injections and 2 touch up, the rest required 4-5 injections. Mean PAIG used: grade1; 11.8 ml grade2; 15.7 ml grade3; 20.8 ml grade 4: 29.5 ml (for both cheek) Up to 35 ml of the PAIG were used for facial deformity Main findings Adverse Events Quality of life and psychosocial The median change in the physician FLSS (1mild to 4 severe) score from baseline to week 12 was - 2 for intervention group and 0 for the without PIAG injection. The aesthetic results were deemed excellent by both physicians and patients. In over 90%, a second injection was carried out 4 weeks after the first injection. Adverse event were mostly mild and transient, resolving after a median of 3 days. The most common adverse events are swelling, pain, bruising, and erythema. Some patients developed edema which slowly diminished until vanishing after 3-4 days. No implant dislocation, implant migration, granuloma, allergic reaction, or intolerance were recorded. The mental health summary scores were statistically significant improved from baseline to week 48. Statistically significant change improvements were noted in both depression and anxiety portions of the HAD scale from baseline to week 48. Not reported

31 22 Karim et al, (2008) Treacy et al, 2006 Ramon,Fodor and Ulman n, (2007) Netherlands 16 males 1 female Mean age 48 Lipoatrophy Grade 2-3 assessed according to Hillebrand et al. Folllow up 48 weeks Ireland 10 male 1 female Mean age 48 Follow up 18 month Israel 10 male 3 female Age FLA severity: Carruthers scale Grade 2 = 23% Grade 3 = 46% Grade 4 = 31% Follow up 1-2 years The number of treatment sessions was single stage. The total amount of PAIG ranged from ml. The median PAIG used was 14 ml. The number of treatment sessions was single stage. The total amount of PAIG ranged from ml The number of treatment sessions was single stage. 1 touch-up was required by 2 patients. The total amount of PAIG ranged from 5-13 ml At week 48, 97% of patients indicated the extent to which FLA had disappeared. The reported severity of FLA decreased significantly from baseline over 48 weeks. The end point of the study was achieved when the physicians, nursing staff, patients agreed the changes in the facial contour reach the optimum. Both physicians and patients agreed on mild to excellent improvement results. 4 patients (23.7%) reported complication. 3 patients developed capsule formation or gel migration which did not need intervention. 1 patient had an infection at one injection site necessitating surgical drainage. Only swelling and bruising occured Local edema was present for up to 1 week in all patients Quality of mental health and social functioning assessed by MOS-HIV and SF-36 improved statistically significant at week 48. Depression tended to decline from baseline over 48 weeks. The modified Beck Depression scale showed a mean overall score decrease from to (p =.1) N/A

32 23 Honig, 2006 Lahiri and Waters, 2006 Germany 9 patients Mean age 43 Follow up 2 years. United Kingdom 15 FLA patients Mean aged 43.6 Follow up 1 18 months No treatment sessions reported. Mean PAIG used 12 ml ranging from 7 ml to 25 ml ml for FLA The level of satisfaction was high (63%), excellent (32%), and poor (5%). PAIG has remained unchanged throughout the period of follow up. High satisfaction scores by a postal questionnaire ( / 5) Edema was present for 2-3 days after surgery, an immediate modest inflammatory reaction in all patients which subsided within hours 1 patient infected with staphylococcus aureus N/A N/A

33 24 Chapter 3 Research Method 3.0 Overview This chapter describes methods used for the economic evaluation of alternative treatments for HIV-associated facial lipoatrophy. It is divided into three sections. The first section presents an overview of the study design, including comparative interventions, study population, time horizon, perspectives taken in the study, and data sources. In section two, decision analytic modeling techniques are described. Finally, input parameters for the model including transition probabilities, costs, and utilities are addressed. 3.1 Overview of the study design A cost-utility analysis is a form of health outcome evaluation that focuses on the quality and duration of health outcomes produced or forgone by health care interventions. Within such analyses, findings are generally reported in terms of costs per quality adjusted life year gained. Even though some researchers treat a cost-utility analysis as part of cost-effectiveness analysis, similarity exists in only the cost component. This economic evaluation employs a cost-utility analysis for two reasons. First, it provides a generic measure for a health care intervention which enables comparisons to be made with other interventions. In contrast, a cost-effectiveness analysis yields intervention-specific health outcomes which limit opportunities to compare alternative interventions. Second, a cost-utility analysis is suitable where interventions yield differential effects on multiple dimensions that affect a patient s quality of life. Generally, a cost-effectiveness analysis is performed when comparative interventions have a common intermediate or final effect. The examples of intermediate outcomes are percentage serum cholesterol reduction, cases detected, episode-free days, etc. Final outcomes typically refer to life prolongation: life years gained, survival rate, mortality rate, etc.. For the interventions that do not extend life, a cost-effectiveness analysis can be performed only when the outcomes of interest are intermediate effects, in other words, not life years gained. However, according to Drummond et al. (2005), final outcomes are preferred to intermediate outcomes. As the final outcome of a cost-effectiveness analysis for facial lipoatrophy treatments is not applicable, we employed a cost-utility analysis.

34 Comparators Based on the literature review and expert clinical opinion from Dr. Mona Loutfy, Faculty of Medicine, University of Toronto, the most appropriate treatments to correct HIV-associated facial lipoatrophy for this economic evaluation are treatments using either poly-l-lactic acid or treatment using polyalkylimide gel because they are administrated for HIV-positive patients in Ontario Population The model population is a HIV positive male patient with facial lipoatrophy at 50 years of age because the mean age of HIV positive patients undergoing facial lipoatrophy treatments reported by most clinical trials are 50 years and men accounted for the overwhelming proportion of study subjects Time horizon The time horizontal of the decision analytic model is assumed to be the lifetime of 50 years-old male patients with HIV Perspective of the analysis The perspective taken in this study is that of society. The societal perspective is the most comprehensive perspective for an economic evaluation (i.e. a cost-utility analysis) because it considers all costs and benefits regardless of who bears the burden or receives the benefit from the intervention of interest. Also, taking the perspective of society is recommended by many guidelines in a cost-utility analysis (Stone 2000). In addition, the model that adopts a Ministry of Health (i.e. public payer) perspective was developed. For this model, the costs borne by patients, such as medication, travel costs, and indirect costs were excluded from consideration Data Sources There are three major data sources for this thesis. First, data were obtained from the Canadian Immunodeficiency Research Collaborative (CIRC) (Loutfy et al. 2007). These data were collected as part of the research entitled intermediate versus delayed polyalkylimide gel injections to correct facial lipoatrophy in HIV-positive patients and performed at the Maple

35 26 Leaf Clinic, Toronto, from 2005 to 2007 (Loutfy et al.). Thirty-one HIV-positive patients who self-identified as having facial lipoatrophy were recruited and injected with polyalkylimide gel to correct their facial lipoatrophy. Data collected included demographic data, facial lipoatrophy severity, anti-retroviral medications used, treatment history, psychosocial assessments, including the MOS-HIV. Second, data from a systematic literature review (i.e. 17 clinical trials) were used to estimate model parameters such as utility, probability of treatment success, and probability of complications. Third, the costs for HIV-associated facial lipoatrophy treatments using poly-llactic acid and polyalkylimide gel in Ontario were obtained through interviews with four physicians who perform such treatments. 3.2 Decision Analytic Model We employed a decision tree and a Markov model to evaluate two comparative interventions: poly-l-lactic acid and polyalkylimide gel. A decision tree was constructed to show the consequences of both treatments while a Markov model was used to estimate the qualityadjusted life year (QALY) gained and the lifetime costs associated for HIV-associated facial lipoatrophy treatments. The model was created using Treeage Software (TreeAge Software Inc 2009) Decision Model A decision tree was used to estimate the cost and utility of HIV-associated facial lipoatrophy treatments because it is a widely used analytical approach to identifying a preferred course of action to take from competing alternatives (Spilker 1996). Figure 2 depicts a stylized decision tree. The tree begins with a choice node indicating the option to choose among the branches originating from it (Spilker 1996). The tree structure continues to the right as a consequence of each decision undertaken. These consequences are structured chronologically by forming an if and then question. For instance, if the decision maker selects choice A, what will happen? After that, chance nodes are inserted after the choice A branch. Chance nodes indicate a probability corresponding with the next events (i.e. the probability of effectiveness). A chance node then follows by the other chance nodes identifying possible events, which may occur and a corresponding probability of those events. For example, if the decision maker selects facial lipoatrophy treatment using polyalkylimide gel and it is effectiveness, will any adverse events occur? At a minimum, a decision tree should include both costs and outcomes of the decision.

36 27 At the end of each branch, the sum of costs and associated outcomes with that branch are estimated and used to calculate the cost per (QALY) gained. Figure 2: Simplified decision tree C - 1 C - 2 C - 3 C-4 C-5 C-6 C-7 C-8 U-1 C-2 C-3 C-4 C-5 C-6 C-7 C-8 Source: (Spilker 1996) In this study, we have constructed a decision tree with two treatment options in the choice nodes: one using poly-l-lactic acid and one using polyalkylimide gel. After the choice nodes, the chance nodes for poly-l-lactic acid and polyalkylimide gel indicate the probability of the filler s effectiveness. This refers to the proportion of patients who show a reduction in the severity of facial lipoatrophy after treatment. The effectiveness nodes are then followed by complication nodes. The major complication associated with these interventions is infection, which leads to filler drainage; therefore, the next chance node indicates the probability of infection corresponding to both interventions in the first year. As poly-l-lactic acid treatment is a semipermanent filler, the subsequent branch, entitled reinjection, was built to indicate the probability of injection in the first year, as the effectiveness of poly-l-lactic may only be retained for some patients for a period of less than one year. Only patients who experienced a successful

37 28 surgery would require further reinjection. This chance node was not created for polyalkylimide gel because it is a permanent filler, which does not require further injections. After the first year, a Markov model is built to estimate the long-term consequences. Figure 3: The decision tree of HIV-associated facial lipoatrophy treatments 3.2 Markov Model A Markov model is commonly used to conduct complex models that have multiple outcomes (Briggs et al. 2006). This model differs from a decision tree in that, instead of possible outcomes being modeled at chance nodes creating a number of possible pathways, a more complex prognosis is reflected as a set of possible transitions between disease states over a number of cycles (i.e. series of discrete time periods) (Sonnenberg and Robert 1993). The analysis is divided into equal increments of time, which are referred to as Markov cycles. We employ a cycle length of one-year with utilities and costs corresponding to its time period. The Markov model to estimate life expectancy of HIV patients is shown in Figure 4. It is called a state

38 29 transition diagram, in which each state is represented by a circle. Arrows connecting two different states indicate allowed transitions. Arrows leading from a state to itself indicate that the patient may remain in that state in consecutive cycles. Only certain transitions are allowed. For example, HIV patients with CD4 > 500 state can transit to any other state, while those in Death state cannot transit to any other states. HIV patients in the pre-aids state could remain in the same state or transit to AIDS (1993 definition) state, or to AIDS (1987 definition) state, or go to death state. A Markov process will terminate when all patients are in the death state. Figure 4: State Transition Diagram of the Markov Model Source: (Kahn et al. 2001). A Markov-cohort model refers to a type of Markov model that focuses on the average patient experience (Briggs et al. 2006). As the focus of economic evaluation is on costs and outcomes, and uncertainty in those expected values, a Markov-cohort model is suitable for a cost-utility analysis (Briggs et al. 2006). The Markov-cohort simulation considers a hypothetical cohort of patients and begins the process with some distribution among the starting states (Sonnenberg and

39 30 Robert 1993). For each cycle, the fraction of the cohort initially in each state is partitioned among all states corresponding to the transition probabilities. This results in a new distribution of the cohort among the various states for the subsequent cycle. This simulation is run for many cycles to build up a profile of the number patients in each state of the model over time. After that, utilities and costs assigned to each state are summed up with this profile. A Markov model is used to estimate the life expectancy of the model population for this thesis, due to its ability to relax assumptions. Instead of assuming that poly-l-lactic acid must be reinjected at all times, we conducted a sensitivity analysis which assumed that poly-l-lactic acid may not be reinjected in the AIDS (1987) state due to a higher risk of infection (Loutfy 2009). In addition, we conducted a series of one-way sensitivity analyses on life expectancy derived from the Declining Exponential Approximation of Life Expectancy (DEALE) method (Beck et al. 1982). Patients age and sex-adjusted life expectancy was obtained from the Ontario life Table and the disease-specific mortality rates were obtained from literature (Keiser et al. 2004; Bhaskaran et al. 2008; Lloyd-Smith et al. 2009). Since two of these studies reported a compound value of morality rates in Switzerland and British Columbia (Canada), we adjusted the population base mortality rate using either the Swiss life table or British Columbia life table (Statistics Canada 2006; Federal Statistical Office 2009). 3.3 Input Parameters of the Model Input parameters in our model are categorized into three groups: probability of event, cost parameter, and utility parameter. Section presents the probability of events for the decision tree and the Markov model. In the subsequent sections, methods used to estimate cost and the utility of facial lipoatrophy treatments are described Probability of events We seek to estimate the probabilities of three events; treatment effectiveness, complications, and reinjection of only poly-l-lactic acid. As the probabilities of treatment effectiveness of both polyl-lactic acid and polyalkylimide gel are determined within the first year, it is represented in the decision tree. However, both the probabilities of complications and reinjection occur every year until all patients are in the death state. Therefore, the first year probabilities of these two events

40 31 are represented in the decision tree while the probabilities of the following year are modeled using Markovian methods (Sonnenberg and Robert 1993). Probability of treatment effectiveness Treatment effectiveness is defined as the reduction in facial lipoatrophy after receiving an injection. This is measured through use of patient or physician perceptions of reduction in facial lipoatrophy at the end of the treatment, which varied across studies depending on the number of treatment- sessions and intervals between each treatment session. Although all published trials have reported the effectiveness of poly-l-lactic acid, only four clinical trials have reported the proportion of patients with success (Burgress and Quiroga 2005; Lafaurie et al. 2005; Carey et al. 2007; Hanke and Redbord 2007). The probability of effectiveness was based on the weighted effectiveness reported in the four trials as shown in Table 3. The probability of treatment effectiveness of poly-l-lactic acid for the base-case analysis is In addition, the highest and lowest probability values (0.82 and 1.00) are used for sensitivity analysis. Table 3: The probability of effectiveness of treatments using Poly-l-lactic Acid Clinical Trial Total no. of subjects No. of subjects who showed success Probability of Success Lafaurie et al Carey et al Burgress and Quiroga Hanke and Redbord Total All clinical trials reported improvement in skin thickness of all patients after receiving polyalkylimide gel injection (Protopapa et al. 2003; Treacy and Goldberg 2006; Lahiri and Waters 2007; Loutfy et al. 2007; Ramon et al. 2007; Honig 2008; Karim et al. 2008). Therefore, the probability of treatment effectiveness of polyalkylimide gel is equal to one.

41 32 Probability of complications Despite their effectiveness, both poly-l-lactic acid and polyalkylimide gel injections lead to several potential adverse effects. The common adverse effects associated with poly-l-lactic acid are palpable, inflammatory but generally invisible subcutaneous nodules, papules, bruising, and minimal and localized oedema. As these complications did not incur cost and also did not affect the patients utilities by assumption, we did not calculate the probability of these adverse events related to poly-l-lactic acid. The major complication associated with polyalkylimide gel is infection with organisms such as staphylococcus aureus, an infection leads to polyalkylimide gel drainage (Protopapa 2009).. As these complications are not specific to Canada, infection rates from all clinical trials conducted irrespective of location have been used. Consultations with experts in this field suggest that it is appropriate to assume a fixed rate of infection with respect to time (Loutfy and Halpenny 2009; Protopapa 2009). As shown in Table 4, the 1-year probability of infection is 0.03 for the basecase analysis. In addition, the highest and lowest values (0.00 and 0.097) are used for the sensitivity analyses.

42 33 Clinical Trial Table 4: The 1-year Probability of Polyalkylimide Gel Infection Total no. of subjects No. of subjects with infections Follow-up period ( in years) A 1-year constant instantaneous rate of infection (r)* 1-year probability of infection (p)** (Loutfy and Halpenny 2009) (Karim et al weeks ) (Treacy and months 0 0 Goldberg 2006) (Honig 2008) (Ramon et al *** ) (Lahiri and Waters 2007) months**** (Protopapa et al. 2003) Total * calculated by r = - ln[(1-p)]/ t; r is rate of infection; p is probability of events; t is a time period (Briggs et al. 2006) ** calculated by p = 1 exp (-rt) ***follow up period ranges between 1 and 2 years ****1-18 months with an average of 7.5 months Probability of reinjection of poly-l-lactic acid (reinjection rate) As poly-l-lactic acid is a semi-permanent filler, the effect of treatment ranges from six months to more than two years (Burgress and Quiroga 2005). Although Moyle et al. (2006) and Valantin et al. (2003) noted that a reduction in facial lipoatrophy was sustained up to two years, the proportion of patients with additional poly-l-lactic acid reinjection at various intervals (i.e. month 6, month 12,) were not reported in their trials. In contrast, Lafaurie et al. (2005) calculated the probability of reinjection at each time point as assessed by the Kaplan-Meier method. Likewise, Burgress and Quiroga (2005) reported the proportion of patients for whom the effect of treatment of poly-l-lactic acid was retained at similar time points. In order to acquire estimates of the reinjection rate at various time points, we used weighted probabilities based on the Kaplan-Meier method as shown in Table 5.

43 34 Duration Table 5: The Probability of Poly-l-lactic Reinjection Between 6 th and 24 th Months Probability of reinjection Lafaurie et al (2005) (94 samples) Probability of re-injection Burgress and Quiroga (2005) (61 Samples) Expected probability of reinjection 6 th month th month * th month th month th month 0.72** * linear extrapolation of 6 th month and 12 th month ** linear extrapolation by ordinary least square from data at month 0 th, 5 th, 10 th, 15 th,20 th As the probability of reinjection varies from year to year, we calculated the probability of reinjection for each Markov cycle (i.e. 1 st year, 2 nd year). The probability of reinjection in the first year was assumed to be equal to the expected probability of the reinjection at 10 th month because the poly-l-lactic acid treatment takes between one-and-a-half to three months depending on the number of treatment sessions required. As reported in Table 5, the expected probability of reinjection in the first year is So, the probability of no reinjection in the first year is After the first year, the probability of reinjection of poly-l-lactic acid was calculated by summing two groups of HIV patients 1. The first group comprised HIV patients who received poly-l-lactic acid injection in the previous year. So, the probability of reinjection of this group is equal to the probability of reinjection at within the first 12 months 2, namely The second group is comprised of those who did not receive poly-l-lactic acid injection in the previous year. Thus, their probability of reinjection is equal to the probability of reinjection between the 12 th and the 24 th month 3. Table 4 demonstrates the 1-year probability of reinjection in each time cycle. The probability of reinjection after the 14 th year remains constant at

44 35 Table 6: The 1-year Probability of Reinjection of Poly-l-lactic Acid for Each Cycle Cycle The probability of reinjection in this year for those who did not reinject in the previous year The probability of reinjection in this year for those who reinjected in the previous year The probability of reinjection in this year Transition probabilities used in the Markov model Time-dependent Markov was built to reflect higher mortality of older people (Briggs et al. 2006). The mortality rate of men in Ontario who are aged between 50 and 105 years was obtained from the Ontario life tables (Canada 2007). All input parameters corresponding to Figure 4 are derived from the study entitled Health and Federal Budgetary Effects of Increasing Access to Antiretroviral Medications for HIV by Expanding Medicaid (Kahn et al. 2001). Kahn et al (2001) used a Markov model to estimate the costs of HIV medication throughout an HIV positive patients lifetime. For this thesis, we assume that all patients are receiving Highly Active Anti-Retroviral Therapy (HAART) because evidence-suggested that HAART, especially Stavudine, is the cause of facial lipoatrophy. The probability transitions used in the Markov model are reported in Table 5. Table 6 shows the distribution of HIV positive patients at the first cycle deriving from clinical trials that reported the CD4 level at the baseline visit. At the first cycle, HIV patients with facial lipoatrophy have the probability to be in the CD4>500 state

45 36 (0.39) and 201<CD4<500 state(0.61). In addition, we conducted a sensitivity analysis that all patients are in either the CD4>500 state or the 201<CD4<500 state. This model uses a cycle length of one year because the frequency of events (i.e. infection, reinjection of poly-l-lactic acid) occurs on a yearly basis. Also, utilities and costs from either poly-l-lactic acid or polyalkylimide gel injection are not sensitive to the cycle length, since these utilities and costs in the second year onward remain constant and do not vary by health states for the base-case analysis. In addition, we assume that each individual is in only one health state at a time. Table 7: Probability Transitions in Markov Model Input Parameters Transition probability per year without HAART Reduction in Transition With HAART Transition probabilities from CD4>500 (state 1) To 201<CD4<500 (State 2) To pre-aids (State 3) To AIDS, 1993 definition (State 4) To AIDS,1987 definition (State 5) To Death (State 6) Transition probabilities from 201<CD4<500 (State 2) To pre-aids (State 3) To AIDS, 1993 definition (State 4) To AIDS,1987 definition (State 5) To Death (State 6) Transition probabilities from pre-aids (State 3) To AIDS, 1993 definition (State 4) To AIDS,1987 definition (State 5) To Death (State 6) Transition probabilities from AIDS, 1993 definition (State 4) To AIDS,1987 definition (State 5) To Death (State 6) Transition probabilities from AIDS,1987 definition (State 5) To Death (State 6) Source: (Kahn et al. 2001)

46 37 Table 8: The CD4 level at the start of Markov process from clinical trials Study (CD4>500) Sample s Mean CD4 Study (201<CD4<500) Samples Mean CD4 Carey et al Moyle and Sutinen Karim et al Valantin et al Lafaurie et al Loutfy et al Total Cost parameters This thesis estimated both the direct and indirect costs of HIV-associated facial lipoatrophy treatments. Direct costs refer to costs related to the intervention itself and associated with net downstream costs (Stone 2000). These resources could be resources in the health care sector if the perspective of the Ministry of Health were taken, and may include out-of-pocket expenses from a societal perspective. Therefore, costs of transportation to receive treatments and indirect costs are included when a societal perspective is taken. The indirect cost of a specific health care intervention refers to the productivity loss attributable to that intervention (Liljas 1998). Inclusion of indirect costs is important when comparative interventions result- in differential time losses. The following sections estimate the direct and indirect cost of HIV-associated facial lipoatrophy treatments. The discount rate of 3 percent is used for a base-case analysis since the US Public Health Service Panel on Cost-effectiveness in Health and Medicine estimated that 3 percent would be the suggest this is a discount rate which is consistent with the shadow price of capital approach (Gold 1996). In addition, we undertook a sensitivity analysis including the rates of 0 percent, and 5 per cent. Severity Distribution As the cost of facial lipoatrophy treatments depends on quantities of poly-l-lactic acid or polyalkylimide gel used to address the severity of patients facial lipoatrophy, it is crucial to determine the severity distribution of this disorder in order to conduct a cost-utility analysis. Table 9 reports the severity distribution of HIV positive patients facial lipoatrophy based on data from two trials that use the same scale developed by Dr. Carruthers (Mest and Humble 2006; Loutfy et al. 2007). According to Table 9, the fraction of the target population with grade

47 38 1 facial lipoatrophy is 0.14, while the fraction of those with grade 2 is The fraction with grade 3 and grade 4 is 0.4 and 0.19, respectively. Table 9: The severity distribution of HIV patients with facial lipoatrophy Clinical Trials Subjects Grade 1 Grade 2 Grade 3 Grade 4 Loutfy et al. (2007) Mest and Humble (2006) Total Fraction of model population in each grade Grade 1: Mild and localized facial lipoatrophy Grade 2: Deeper and longer lipoatrophy with the muscles beginning to show through Grade 3: Deeper and wider atrophic area with the muscles clearly showing Grade 4: Lipoatrophy covering a wide area, extending up towards the eye sockets, and facial skin lying on the muscles Direct Costs Estimation The direct costs of HIV-associated facial lipoatrophy treatments are comprised of three categories: initial treatment costs, re-injection of poly-l-lactic acid costs, and complication costs. As the costs of treatment- are stratified by severity of facial lipoatrophy, the quantities required for the treatment of patients with different grades of facial lipoatrophy were obtained from the clinical trials that use the same severity grading scales developed by Dr. Carruther. The quantities of poly-l-lactic acid used were obtained from the study of Mest and Humble (2006) while those which deal with polyalkylimide gel are derived from Loufty et al (2007). The quantities of both fillers, stratified by severity, are shown in Table 10. Table 10: Quantity of fillers used for facial lipoatrophy treatments Stage Poly-l-lactic acid Polyalkylimide gel Source: (Loutfy et al.) and (Mest and Humble 2006)

48 39 Patients generally need 1-2 vials of poly-l-lactic acid to maintain the effects of the collagen approximately every 2 years (Krajden 2009; Sharpe 2009). Therefore, the costs of reinjection are based on 1.5 vials of poly-l-lactic acid for a base-case analysis, ranging from 1 to 2 vials for sensitivity analyses. The costs of treatment using either poly-l-lactic acid or polyalkylimide gel product were obtained through consultation with surgeons who have performed these treatments in Toronto (Canada), and representatives from the Canadian product distributors. Table 11 presents a list of physician quotes regarding the treatments. Fee quotes from cosmetic surgeons for poly-l-lactic acid per vial ranged from $700 CAD to $800 CAD, with a mean of $750 CAD. For polyalkylimide gel, the fee quote from Dr. McCahill is at a reduced price, because his patients received support from Pur Medical Inc (Berninger 2008). Therefore, only the fee quoted by Dr. Krajden represents the market price of CAD $500 per ml. However, Dr. Krajden has never injected patients with severe facial lipoatrophy, which requires more than 8 ml and he charges $2500 CAD for the first 5 ml. As the injection of polyalkylimide gel involves only a single stage, overhead costs and the material stage are included in the first 5 ml. Therefore, the charge could range between $160 CAD (the entire cost of the product) and $500 CAD per ml for the additional unit. We assumed a median cost of $330 CAD for the additional unit (per ml) in the base-case analysis. Also, we conducted sensitivity analyses of his quotes ranging from $160 CAD to $500 CAD per ml for more than 5 ml. These fee quotes of either product consist of three components: the product (per volume), materials used to inject the fillers (e.g., antibacterial solution, injection syringe, etc.), and a consultation fee, including overhead cost, etc.

49 40 Table 11: Physicians Quotes for Treatment Using Poly-L-Lactic Acid (Sculptra ) and Polyalkylimide Gel (Bio-Alcamid ). Poly-l-lactic acid Polyalkylimide gel Dr. Nowell Solish, Cosmetic Dermatologist Dr. Zel Krajden, Plastic & Reconstructive Surgeon Dr. Mitchell H. Brown Toronto Plastic Surgery Dr. Zel Krajden, Plastic & Reconstructive Surgeon Dr. John McCahill*, President Ascente Medical Corporation $700 per vial $750 per vial $800 per vial $500 per ml $200 per ml *The cost that Dr. John McCahill charged does not reflect the market price, because he was supported by Pur medical (polyalkylimide gel distributor) Table 12 presents the cost components of poly-l-lactic acid and polyalkylimide gel. We obtained the cost of poly-l-lactic from the Sculptra distributor (Sharpe 2009) and the cost of polyalkylimide gel from Dr. Krajden. Materials used to inject the fillers while following the guidelines were obtained from Shoppers Drug Mart, which is one of the largest pharmaceutical retail chains in Canada (sanofi-aventis 2009; SkinRx Distribution Inc 2009). For poly-l-lactic acid, the physician fee and product accounted for the overwhelming proportion of the costs as other materials costs less than $2 CAD in total. Similarly, physician fees accounted for almost 70 per cent for polyalkylimide gel. Since such treatments are performed in private clinics, the physician quotes include overhead cost. Overhead costs refer to resources that have served many different departments and programs, such as clinic administration, medical records, cleaning, and porters (Drummond 2005). To estimate the cost of individual interventions, the shared costs need to be included to the extent in which such cost variation is attributable to the intervention. Such information on overhead costs, however, was not available. Therefore, we were unable to estimate the cost of physicians fee and overhead costs separately.

50 41 Table 12: Cost components of Poly-L-Lactic Acid and Polyalkylimide Gel Poly-l-lactic acid Polyalkylimide gel Cost component low High Base-case Base-case range range Physician quote product cost material cost* ** Physician fee & overhead cost *Including syringes 5 ml (1), and 3 ml (1),needle 26 G (3),and 18 G (2), Lidocaine 1-2% (1 ml), sterile water (3 ml), and anti septic cleanser (20 ml) for poly-l-lactic acid **Including syringes 5 ml (1), and 1 ml (2),needle 16 G (3),and 20 G (3), Lidocaine 1% (1mL),and anti septic cleanser (20 ml) for polyalkylimide gel Anti-biotic medication is prescribed after polyalkylimide gel injection to prevent infection. According to Loutfy et al (2007), only one clinical trial of polyalkylimide gel has been conducted in Toronto (Canada). In that trial, a 5-day course of antibiotic prophylaxis and oral Cephalexin (500 mg) four times per day were given. In addition, Levofloxacin (500 mg) a day was used in cases of penicillin allergy. We obtained the costs of these medications from Shoppers Drug Mart. Table 13 shows that the costs of antibiotics for polyalkylimide gel injection are either $61.11 CAD or $84.14 CAD for patients who are allergic to Penicillin. As the cost of medication is not covered by the Ontario Health Insurance Plan, these costs are borne by patients and are only included for the analysis which adopts a societal perspective. Table 13: Costs of antibiotic medication for polyalkylimide gel injection Drug Name Quantity Duration Estimated Total Drug Cost Clindamycin 300mg 4 times a day, for 5 days $37.42 Cephalexin 500mg 4 times a day, for 5 days $23.69 Total $61.11 Levofoxacin 500mg 1 times a day, for 5 days $46.72 Clindamycin 300mg 4 times a day, for 5 days $37.42 Total $84.14 Source: Shopper Drug Mart

51 42 In term of costs associated with complications of poly-l-lactic acid, patients may develop micronodules under the skin in the treated area. While the subcutaneous micronodules are often not visible, they are evident when the treated skin is touched. Physician recommended treatment for micronodules is through various massage techniques. Therefore, there is no cost associated with micronodules. Occasionally, patients have reported incidences of papule, bruising, rash, redness and swelling. These symptoms are expected reactions resulting from fibroblast and collagen production. Also, these symptoms do not require any interventions incurring costs. The complication costs associated with polyalkylimide gel are derived from post-surgical infections, most commonly with organisms such as staphylococcus aureus. In light of possible surgical interventions required to treat patients who had suffered severe infections, the costs for product removal and reinjection are billed to the Ontario Health Insurance Plan (OHIP). Therefore, we estimated these costs based on OHIP codes, including a consultation fee of $56.1 CAD (OHIP Code: A005), an incision fee of $40.20 CAD (OHIP Code: Z101), and a repeat consultation fee of $42.35 CAD (OHIP code: A006) (Ontario Ministry of Health and Long-Term Care 2009). After polyalkylimide gel drainage, Keflex (500 mg) is prescribed for two weeks (Loutfy and Halpenny 2009), the cost of which is $44.76 CAD from Shopper Drug Mart. Travel costs The costs of traveling to receive either poly-l-lactic acid or polyalkylimide gel treatments is an out-of-pocket payment borne by patients and is only considered for the analyses which adapts societal perspective. We estimate the round trip costs of consultation visits, and treatment visits based on the distances between the patients residence (as assessed by their residence postal code) and clinics offering treatments using poly-l-lactic acid or polyalkylimide gel. The postal codes of the residence of patients with facial lipoatrophy were obtained from the - CIIRC database while the postal codes of four clinics that offer poly-l-lactic acid or polyalkylimide gel (See Table 11) were obtained from physicians clinic- websites (Dr. Nowell Solish Cosmetic Dermatologist 2009; Maple Leaf Medical Clinic 2009; Mitchell H. Brown 2009; Spamedica 2009).

52 43 The average costs of round trip for consultation visits without treatments are estimated based on the assumption that patients use public transportation from Toronto Public Transit System (TTC). Cash fare of a one-way ticket is $2.75 CAD for adults (Toronto Transit Commission 2009). For patients outside Toronto, costs of transportation also include either Go Train or Viarail with fares that vary by distance (Go Transit 2009; VIA Rail Canada 2009). The average transportation cost of a one-way ticket for a consultation is $ 5.63 CAD. Therefore, the average cost of a round trip for consultation visit is $ CAD. Unlike the day of consultation without treatment, the average costs of transportation in the day of injection are calculated based on the assumption that patients commute by public transportation to receive a treatment but return home by taxi (Loutfy 2009). The costs of taxi are determined by the average driving distance between patients residences and clinics (Beck Taxi 2009). Table 14 shows the driving distance between patients' residences and clinics where such treatments are offered. The average driving distance between patients' residence and clinics offering poly-llactic acid and polyalkylimide gel injection is and kilometers, respectively (Zip- Codes.com 2009). The taxi fares are $23.54 CAD for poly-l-lactic acid injection and $ CAD for polyalkylimide gel injection. We assumed one visit for polyalkylimide gel injection for all severity grades. In contrast, we assumed 3-6 visits for poly-l-lactic acid injection according to severity grade (patients with grade1 requires 3 visits, grade 2 requires 4 visits, grade 3 requires 5 visits, and grade 4 requires 6 visits). In addition, HIV patients are assumed to visits these physicians two times for consultation: one visit prior to fillers injection and one visit for followup after injection. Table 14: Estimated Taxi fares from patients' residences to clinics Average distance to Dr. Nowell Solish Cosmetic Dermatologist kilometers Average distance to Spamedica (Dr. Zel Krajden) kilometers Average distance to Toronto Plastic Surgery (Dr. Mitchell H. Brown) kilometers Average distance to treatments using poly-l-lactic acid kilometers Average Taxi fare for poly-l-lactic acid $23.54 CAD Average distance to Spamedica (Dr. Zel Krajden) kilometers Average distance to Maple Leaf Clinic km. Average distance to treatments using polyalkylimide gel kilometers Average Taxi fare for polyalkylimide gel $ CAD

53 44 Indirect Costs The indirect costs of receiving facial lipoatrophy treatments refers to costs associated with loss of income due to time taken off paid-work to receive the treatments, measured by Human Capital Method (Liljas 1998). The Human Capital Approach is one of the commonly used methods to calculate indirect costs in economic evaluations, and can be defined as the lost gross value of income during the time absent from usual activities (Liljas 1998). In our analysis, the time cost is calculated by assuming that all HIV patients pay the same time cost whether they take time off from paid work or leisure to receive treatments. The parameter for evaluation was based on the wage of people living with HIV in Ontario. This value was estimated at $24,000 CAD based on a weighted calculation of the annual incomes from a sample of people with HIV/AIDS in Ontario (Williams 2005). This study was a one-time cross sectional survey of 297 people living with HIV between October 2001 and August This estimate is similar to the other study conducted in Ontario with a sample of 104 adults, where the average 37% had annual household incomes lower than $20,000 CAD (Furler 2006). The value of a working day was calculated based on a 48-week year and assuming full-time employment consisting of 35 paid-work hours per week, which results in an hourly wage of $13.88 CAD after converting to 2009 dollars (economica 2009). The detailed calculations are shown in Table 15. Table 15: Estimated hourly range of HIV-positive patients based on annual income Expected income per year = ( =$10,000)*$10,000 + ( =$20,000)*$20,000 + ( =$40,000)*$40,000 + ( =$50,000)*$50,000 Expected income per hour Converting to 2009 value = (0.39)*10,000 + (0.43)*20,000 + (0.11)*40,000+ (0.07)*50,000 = $20,400 = $20,400*(1/48 weeks)*(1/35 hours) = $ = $12.14*1.143 = $ Source: Table 1 from (Williams 2005)

54 45 The indirect costs of receiving HIV-associated facial lipoatrophy treatments are calculated based on time loss due to receiving injections with the assumption that a patient visits the clinic for three purposes: consultation prior to injection, product injection, and follow-up visit. Time loss due to consultation prior to injection visit, and follow up visit are divided into two components: time loss due to travelling to the clinics; and time loss due to consultation, which is assumed to be one hour for consultation and half an hour for the follow-up visit (Krajden 2009). These latter two time frames include waiting time and administrator time. In contrast, time loss on the day of injection was estimated based on the assumption that patients would be absent from work for one day (Loutfy 2009). In terms of time loss due to travel, a single-trip from HIV patients residence to/from clinics is assumed to be 45 minutes. This assumption is based on travel time of approximately 12 kilometres in Ontario, taking traffic into account because all clinics are located in downtown Toronto. The average speed of a street car at Carlton Road, the location of Maple leaf clinics, is 17.2 kilometers per hour (Toronto Transit Commission 2009). So, it would take minutes to commute to that clinic by street car. Furthermore, the average subway speed is 30 kilometers per hour to the other clinics. So, it would take 24 minutes by subway without changing lines. Although the average travel time of 12 kilometers by bus in the greater Toronto areas is minutes with the speed limit kilometers per hour (Alvarado 1999), commuting by bus- in downtown Toronto would be longer than this estimation due to traffic. Therefore, 45 minutes is a reasonable estimation of income loss due to travel since we did take into account waiting time. In addition, the sensitivity analysis consider time loss ranging between 10 minutes for HIV positive patients living near a particular clinic to one day off work for those who live further away.

55 Health Outcome parameters Preference describes the overall concept of desirability of different health outcomes (Drummond 2005). Preference comprises utilities and values. Values are measured with certainty through use of methods such as visual analogue scale while utilities are measured under uncertainty through use of standard gamble. Therefore, attitudes towards risk (i.e. risk aversion, risk neutrality, risk lover) for different health states will affect utilities but not values. For this reason, standard gamble is considered to be the gold standard amongst preference assessment approaches. Valantin et al. (2003) is the only study which has measured the preference of HIV positive patients attributable to treatments using poly-l-lactic acid. Using the Visual Analog Scale, the median increase in health-related quality of life of 44 HIV positive patients after receiving treatment were 0.03 at week 12, 0.08 at week 24 and 48, and 0.04 at week 72 and 96. Since the Visual Analogue Scale is a value measurement which does not capture risk preference, a power curve function has been proposed to estimate the relationship between visual analogue scale (value) and standard gamble (utility) (Brazier et al. 2007). In the following equation proposed by Torrance et al (1996) 4, u represents utility measured by standard gamble and v represents value measured by visual analogue scale. u = 1 In addition to Torrance et al (1996), other studies proposed a power curve function to estimate the relationship between visual analogue scale and utility with different power (. (Torrance 2001). The highest and lowest value from these power curve functions will be used for sensitivity analysis. In contrast, the health-related quality of life attributable to polyalkylimide gel injection is calculated based on the data obtained from the Canadian Immunodeficiency Research Collaborative (CIRC) (Loutfy et al. 2007). According to these data, thirty-one HIV positive 4 Torrance et al (1996) estimated weights for the Health Utility Index 2 using the relationship between mean standard gamble and visual analogue scale values for four health states by fitting a power function to the data, which explains 97 percent of the variation.

56 47 patients were required to complete Medical Outcome Study HIV health surveys (MOS-HIV) at both treatment visits and follow-up visits. The MOS-HIV is a brief, but comprehensive measure of health-related quality of life, which has been used extensively on patients with HIV/AIDS (Albert 1996). Mental Health Summary Scores (MHS) and Physical Health Summary Scores (PHS) deriving from MOS-HIV and other baseline characteristics are used to calculate utility through use of the standard gamble approach (Bayoumi 2009). Bayoumi (2009) estimates the coefficient of utility predictors of HIV patients using the Ordinary Least Squares (OLS) technique. The full model includes all utility predictors 5 while the short model includes only some utility predictors 6. The full model was used for a base-case analysis and the short models were used for sensitivity analysis. Utility gained from polyalkylimide gel injection was reflected by the increase of MHS and PHS scores in MOS-HIV. Table 16 shows the amount of utility gained from the injection of either poly-l-lactic acid or polyalkylimide gel. Table 16: Utility gained from either poly-l-lactic acid or polyalkylimide gel injection. Baseline 12 th Week 24 th Week 48 th Week 72 th Week 96 th Week Poly-l-lactic acid - measured by VAS Poly-l-lactic acid - transform to SG Poly-l-lactic acid - increase Polyalkylimide Gel - SG utility score Polyalkylimide Gel - increase The baseline for polyalkylimide gel is derived from week 0 (immediate group) and week 12 (delayed group); Week 12 is derived from week 12 (immediate group) and 24 (delayed group); week 24 is derived from week 24 (immediate group) and 48 (delayed group); week 48 and 72 are derived from week 48 (immediate group) and 72 (delayed group); week 96 is derived from week 72 (immediate group) and 96 (delayed group). 5 Number of Severe Symptoms, Protease inhibitor use, Non-Nucleoside Reverse Transcriptase Inhibitors use Nucleoside Reverse Transcriptase Inhibitor use, ethnicity, Having sex with men, Injection drug use, Current CD4, Nadir CD4, Current Viral load, Peak Viral load, age 6 Number of Severe Symptoms, Protease inhibitor use, Non-Nucleoside Reverse Transcriptase Inhibitors use Nucleoside Reverse Transcriptase Inhibitor use

57 48 In the first year after injection, utility gained from either poly-l-lactic acid or polyalkylimide gel injection is calculated by by the average utility gained during the first year (See Table 16). From the second year onwards, utility gained is calculated by the average of how much the utility changed during week 72 and week 96. We assume that the utility changed from these injections are constant over HIV positive patients lifetime and do not vary by health states in the Markov model. Although poly-l-lactic acid requires reinjection approximately every two years, we assume that reinjection does not affect utility gained from poly-l-lactic acid injection, because facial lipoatrophy was corrected from the first injection with large volumes deposits in several sessions (i.e vials). After the first injection, only 1-2 vials of poly-l-lactic acid reinjection is required when its effect starts to vanish. Therefore, facial lipoatrophy is fixed at all times, including the reinjection periods. However, utility gained from polyalkylimide gel may be affected by infection of organisms such as staphylococcus aureus, an infection which leads to product drainage. Typically, polyalkylimide gel is removed as soon as this infection occurs. After waiting for 6 months for the infection to heal, polyalkylimide gel is injected again (Berninger 2008; Protopapa 2009). As a consequence, utility gained from polyalkylimide gel is assumed to be absent during that 6-month period. A discount rate of 3 per cent is also applied to QALYs gained attributable to poly-l-lactic acid and polyalkylimide gel. Since individuals generally prefer health gains in the present rather than the future, the stream of health gains should be discounted to account for this time preference (Drummond 2005).

58 Sensitivity Analysis A series of one-way and two-way sensitivity analyses are conducted to test the robustness of the model to a variation in a single, and subset of variable at a time (Meltzer 2001). Uncertain variables (e.g. the probability of treatment effectiveness, the probability of infection) were altered and the model is re-run while maintaining all other variables constant. We also conducted two-way sensitivity analysis allowing for two variables to change simultaneously while the other variables remain constant. In addition, a probabilistic sensitivity analysis was performed using Monte Carlo simulations, where each model input is assigned an independent probability distribution. The decision analytic model was simulated using 1,000 repetitions to assess overall robustness of the model.

59 50 Chapter 4 Study Results 4.0 Introduction This study assessed the incremental cost-utility ratio of HIV-associated facial lipoatrophy treatments using poly-l-lactic acid and polyalkylimide gel. In this chapter, our findings are reported in two sections. Section 4.1 presents the base-case cost and health outcome analysis, while Section 4.2 describes three types of sensitivity analyses: one-way; two-way; and probabilistic sensitivity analyses. 4.1 Costs and Health Outcomes analysis Cost Analysis Tables 17 and 18 present component costs for HIV-associated facial lipoatrophy treatments using poly-l-lactic acid and polyalkylimide gel, respectively. Costs presented in these tables are stratified by a 4-point grading scale developed by Dr. Carruthers to measure the severity of facial lipoatrophy. The 6 th column of Table 17 and 18, entitled Weighted Average, is calculated by weighted average of the fraction of HIV positive patients in each grade corresponding to Table 9. The cost of facial lipoatrophy treatments are categorized into 4 groups: physician quotes, medication costs, travel costs, and indirect costs. Tables 17 and 18 present cost components for HIV-associated facial lipoatrophy. According to these tables, travel costs, medication costs, and indirect costs account for approximately 10 per cent of the total cost of treatments using poly-llactic acid irrespective of patient severity. In contrast, treatment costs are constant across severity grades when polyalkylimide gel is used because only a single stage injection is required for polyalkylimide gel regardless of the severity of facial lipoatrophy severity. Table 19 shows the present value of total costs discounted at the rate of three per cent over a patient s lifetime. Only Physician Quotes are considered in analyses that adopt a Ministry of Health perspective. Cost differences attributable to a shift in perspective from the Ministry of Health to that of Society are based on inclusion of travel, medication, and indirect costs. The costs of poly-l-lactic acid include initial treatment costs and additional treatment costs that occur approximately every two years, while the costs of polyalkylimide gel include the initial costs and

60 51 the recurrent costs associated with infections, including potential polyalkylimide gel reinjection. Since experts in this field have noted that most patients experience infection on one side of their face and require reinjection (Loutfy and Halpenny 2009), the total costs of polyalkylimide gel infection comprise both the costs to treat the infection and the costs of re-injection. Table 17: Cost components of Poly-L-Lactic Acid in 2008 Canadian Dollars Cost component Grade1 Grade 2 Grade 3 Grade 4 Weighted Average Additional Treatments Physicians Quotes $4,500 $6,000 $7,500 $9,000 $6,780 $1,125 Travel costs $ $139.2 $ $ $ $40.43 Indirect Costs $ $ $ $ $ $ Total Costs $4,964 $6,590 $8,216 $9,843 $ $1,325 Table 18: Cost components of polyalkylimide gel in 2008 Canadian Dollars Cost component Grade1 Grade 2 Grade 3 Grade 4 Weighted Average Infectious Treatments Physicians Quotes $4,744 $6,031 $7,714 $10,585 $7,389 $3,742.83* Medication Costs $61.11 $61.11 $61.11 $61.11 $61.11 $105.87** Travel costs $51.29 $51.29 $51.29 $51.29 $51.29 $102.58*** Indirect Costs $ $ $ $ $ $319.15*** Total Costs $5,016 $6,303 $7,986 $10,857 $7,661 $4, *including costs of infectious treatments and reinjection of polyalkylimide gel **including costs of medication (Keflex) for treating infection ***including costs for infectious treatments and then reinjection treatments

61 52 Table 19: Present value of total costs of poly-l-lactic acid and polyalkylimide gel for treating HIV positive patients with varying severity of facial lipoatrophy in 2008 Canadian Dollars Grade1 Grade 2 Grade 3 Grade 4 Severity Distribution Ministry of Health perspective Poly-l-lactic acid $11, $13, $14, $16, $14, Polyalkylimide gel $6, $7, $9, $12, $8, Societal perspective Poly-l-lactic acid $13, $15, $17, $18, $16,388 Polyalkylimide gel $6, $7, $9, $12, $9,242 Health Outcomes Analysis The incremental quality adjusted life years (QALYs) gained are and 0.19 for poly-l-lactic acid and polyalkylimide gel, respectively. Base-case Analysis For the analysis taken from the Ministry of Health perspective, the incremental cost-utility ratios are $45,457 CAD per QALY for polyalkylimide gel and $57,352 CAD per QALY for poly-llactic acid when compared to do nothing baseline. The incremental cost-utility ratio (ICUR) for poly-l-lactic acid when compared to polyalkylimide gel is $97,907 per QALY. For the analysis taken from the societal perspective, the cost-utility ratios were $48,583 CAD per QALY for polyalkylimide gel and $66,608 CAD per QALY for poly-l-lactic acid when compared to do nothing baseline. The incremental cost utility ratio (ICUR) for poly-l-lactic acid when compared to polyalkylimide gel is $129,734 per QALY.

62 Sensitivity analyses This section reports the input parameters and results of one-way sensitivity analyses, two-way sensitivity analyses, and probabilistic sensitivity analyses, respectively One-way sensitivity analyses Table 20 provides input parameters for one-way sensitivity analyses. The discount rate ranges from 0 per cent to 5 per cent. The probability of effectiveness of poly-l-lactic acid injection ranges from 0.82 to 1, while the probability of effectiveness of polyalkylimide gel injection is equal to one. In contrast, polyalkylimide gel injection may lead to an infection, which requires drainage, while poly-l-lactic acid injection does not lead to such infections. The probability of this infection for polyalkylimide gel is between 0 and Costs parameters differ by various physician quotes for both poly-l-lactic acid and polyalkylimide gel. Three physicians in Ontario offer poly-l-lactic acid injection between $700 CAD per vial and $800 CAD per vial. One physician offers polyalkylimide gel at price without support from the polyalkylimide gel distributor in Ontario, and he charges $2500 CAD for the first 5 ml. We, therefore, conducted sensitivity analyses using his quotes ranging from $160 CAD to $500 CAD per ml for more than 5 ml. In addition to various physician quotes, we also conducted sensitivity analyses for other uncertain costs. The quantity of poly-l-lactic acid reinjection every approximately two years ranges from 1 to 2 vials. The cost of antibiotics for polyalkylimide gel injection is either $61.11 CAD or $84.14 CAD for patients who are allergic to Penicillin. Income loss due to travel from patients residence to the clinics for a consultation visit ranges from $ CAD (10 minutes) to $ CAD (6 hours) for those who live further away and take a day off from work. Utilities gained from both treatment options (poly-l-lactic acid and polyalkylimide gel) are subject to uncertainty. Since utility gained from poly-l-lactic acid was measured through use of a visual analogue scale, we conducted sensitivity analyses on several power curve functions that estimate the relationship between the visual analogue scale and standard gamble (Torrance 2001). In contrast, the utilities gained from polyalkylimide gel were derived from two Ordinary Least Square regressions. The full regression includes all utility predictors, while a more

63 54 parsimonious model includes only a subset of those predictors (Bayoumi 2009). Both models were used for one-way sensitivity analysis. We conducted one-way sensitivity analyses on model structures that affected the life expectancy of HIV patients; CD4 level at the first cycle of the Markov model, the age of HIV patients when they initiate facial lipoatrophy treatments, and the rate of disease progression. CD 4 level at the first cycle of the Markov model includes either patients who received initial treatments with a CD4 between 250 and 500 or those who may have started treatment with a CD4 of more than 500. Since the mean age of all clinical trials fall between 40 and 50, we undertook sensitivity analyses in the cases of all HIV positive patients initiating treatment when they were 40 years of age. The rate of disease progression was derived from the original study, the value of which (i.e ) indicates proportion of base-case (Kahn et al. 2001). In addition, we conducted a sensitivity analysis where poly-l-lactic acid was not reinjected in the AIDS state (1987 definition) of the Markov model, because patients would be at a high risk of infection (Loutfy 2009). The utility gained from poly-l-lactic acid injection was assumed to be retained in the first two years after a HIV positive patient entered this state. Similarly, polyalkylimide gel would not be reinjected to recorrect facial lipoatrophy where infection occurred in this state (Loutfy 2009). Therefore, we conducted a sensitivity analysis in the case where polyalkylimide gel infection occurred in this state without reinjection of polyalkylimide gel to recorrect facial lipoatrophy. No Utility was assumed to be gained when infection occurred following polyalkylimide gel injection. To assess the influence of using the Markov model in our findings, we built three analytic models using the Declining Exponential Approximation of Life Expectancy (DEALE) method to estimate the life expectancy of our model population instead of using the Markov model (Beck et al. 1982). First, we obtained the mortality rate of HIV positive patients in British Columbia from the HAART Observational Medical Evaluation and Research (HOMER) and adjusted it by age and sex-specific British Columbia life table to estimate the life expectancy of our model population (Statistics Canada 2006; Lloyd-Smith et al. 2009). Second, Mest and Humble (2006) and Valantin et al. (2003) reported the average time since seroconversion at 13.4 and 14.6 years. As an increase in seroconversion period is associated with a shorter life expectancy, Bhaskaran et al (2008) used data from Concerted Action on Seroconversion to AIDS and Death in Europe (CASCADE), which is currently a collaboration of 23 cohorts of individuals with well-estimated seroconversion date in developed countries

64 55 including Canada. They reported the excess mortality rate of HIV positive patients corresponding to this seroconversion compared to the general population was 0.03 (Bhaskaran et al. 2008).We therefore calculated the life expectancy of HIV positive patients based on this excess mortality rate following the DEALE method. Third, the mortality rate from the life table of the perspective HIV cohort study in Switzerland adjusted by the age and sex-specific mortality rate of the Swiss life table was used to calculate life expectancy (Keiser et al. 2004; Federal Statistical Office 2009). Table 20: Parameters tested in sensitivity analyses Parameters Base-case Range source Discount rate 3% 0%-5% Probability The probability of poly-l-lactic acid effectiveness The probability of polyalkylimide gel infection Physician Quotes for poly-l-lactic acid Physician Quotes for polyalkylimide gel per ml for more than 5 ml Quantity of poly-l-lactic acid reinjection Income loss due to travel to consultation visit (Burgress and Quiroga 2005; Lafaurie et al. 2005; Hanke and Redbord 2007) (Protopapa et al. 2003; Treacy and Goldberg 2006; Loutfy et al. 2007; Ramon et al. 2007; Honig 2008) Costs (Solish 2008) (Krajden 2009) (Brown 2009) (Krajden 2009) 1.5 vials 1-2 vials (Krajden 2009) (Sharpe 2009) 45 minutes 10 minutes (Toronto Transit 6 hours Commission 2009) Antibiotic for polyalkylimide gel (Shopper Drug Mart 2009)

65 56 Utilities gained from poly-l-lactic acid Utilities (Torrance 2001) Utility gained from polyalkylimide gel Full Model Short Model (Bayoumi 2009) Model Structure The proportion of patients in the All clinical trials CD4 > 500 state in the first cycle of the Markov Model Age All clinical trials Rate of disease progression (Kahn et al. 2001) AIDS (1987) definition without Reinjection No (Loutfy 2009) poly-l-lactic acid reinjection reinjection AIDS (1987) definition without polyalkylimide gel reinjection Reinjection No reinjection (Loutfy 2009) Using Life expectancy by DEALE HOMER, British Columbia n/a LE = (Lloyd-Smith et al. 2009) CASCADE n/a LE = (Bhaskaran et al. 2008) Swiss HIV Cohort Study n/a LE = (Keiser et al. 2004) The results of the one-way sensitivity analysis for poly-l-lactic acid from the Ministry of Health perspective are presented in Table 21, and analyses that take a Societal perspective are presented in Table 22. Corresponding tornado diagrams (TreeAge Software Inc 2009) are provided in Figures 5 and 6, respectively. The use of different utility transformations causes the ICUR to vary from $39,493 CAD per QALY to $105,952 CAD per QALY. In addition, both models are sensitive to the amount of poly-l-lactic acid used in reinjection, and the discount rate, respectively. The models are not sensitive to income loss as a result of consultation visits, age, the CD4 level distribution at the first cycle of the Markov model, and the rate of disease progression. Although initiating treatments at a younger age increased costs and QALYs, the ICUR was lower than the base case. In contrast, if poly-l-lactic acid is not reinjected in the AIDS

66 57 (1987 definition) state, the ICUR was higher than the base case. Finally, the results of using the DEALE method to calculate the life expectancy of our model population did not significantly alter our findings from the base-case model that used the Markov model. All ICUA ratios from these models moderately exceeded the base-case model. Table 21: Results of one-way sensitivity analyses that adopt Ministry of Health perspective for treatments using poly-l-lactic acid Value Incremental Incremental ICUR** Parameters Cost* QALY* Discount rate 0% 16, ,259 3% 14, ,352 5% 12, ,551 Probability of effectiveness , , , , , ,596 Physician Quotes , , , , , ,176 Quantity of reinjection 1 vial 11, , vial 14, ,352 2 vials 16, ,284 Utilities gained 14, ,107 14, ,713 14, ,276 14, ,352 14, ,503 CD4 level distribution CD4>500 14, ,778 Base-case 14, , <CD4<500 13, ,481 Age 40 14, , , ,352 Rate of disease progression , , , , , ,449 AIDS (1987) definition no reinjection 12, ,655 reinjection 14, ,352

67 58 DEALE LE = , ,220 LE = , ,817 LE = , ,977 *compared to do nothing strategy **Incremental Cost-Utility Ratio (ICUR) Table 22: Results of one-way sensitivity analyses that adopt societal perspective for treatments using poly-l-lactic acid Value Incremental Incremental ICUR** Parameters Cost* QALY* Discount rate 0% 19, ,673 3% 16, ,608 5% 14, ,386 Probability of effectiveness , , , , , ,456 Physician Quotes , , , , , ,505 Quantity of reinjection 1 vial 13, , vial 16, ,608 2 vials 18, ,906 Income loss due to Travel for a 10 minutes 16, ,028 consultation visit 45 minutes 16, ,608 Day off 16, ,569 Utilities gained 16, ,585 16, ,485 16, ,993 16, ,608 16, ,416 CD4 level distribution CD4>500 17, ,817 Base-case 16, , <CD4<500 15, ,893

68 59 Age 40 17, , , ,608 Rate of disease progression , , , , , ,994 AIDS (1987) definition no reinjection 14, ,534 reinjection 16, ,608 DEALE LE = , ,809 LE = , ,211 LE = , ,112 *compared to do nothing strategy **Incremental Cost-Utility Ratio (ICUR) Figure 5: Tornado diagram presenting results of one-way sensitivity analyses that adopt Ministry of Health perspective for treatments using poly-l-lactic acid

69 60 Figure 6: Tornado diagram presenting results of one-way sensitivity analyses that adopt societal perspective for treatments using poly-l-lactic acid One-way sensitivity analysis results for polyalkylimide gel are provided in Table 23 based on a societal perspective and Table 24 when the perspective of the Ministry of Health is taken. Corresponding tornado diagrams are shown in Figures 7 and 8. Both models are sensitive to physician quotes per ml (more than 6 ml) and the probability of infection, since the highest possible ICUR differs from the lowest one by approximately $30,000 CAD. The discount rate also influences the findings by about $15,000 CAD. In contrast, the models are not sensitive to CD4 level distribution at the first cycle of the Markov model, income loss due to travel for the consultation visit, antibiotics used in case of an allergy to penicillin, and AIDS (1987) state without polyalkylimide gel reinjection. Similar to the treatments using poly-l-lactic acid, initiating treatments at a younger age lowers the ICUR. Furthermore, the results of using the DEALE method instead of using the Markov model did not greatly change our results. The ICUR from the two models moderately exceed the base-case, while the ICUR using the diseasespecific mortality rate of Swiss HIV cohort was mildly lower than the base-case analysis.

70 61 Table 23: Results of one-way sensitivity analyses that adopt Ministry of Health perspective for treatments using polyalkylimide gel Value Incremental Incremental ICUR** Parameters Cost* QALY* Discount rate 0% 9, ,561 3% 8, ,457 5% 8, ,839 Probability of infection 0 7, , , , , ,548 Physician Quotes 160 5, , , , , ,900 Utilities gained Full Model 8, ,457 Short Model 8, ,339 CD4 level at the start CD4>500 8, ,879 Base-case 8, , <CD4<500 8, ,315 Age 40 8, , , ,457 Rate of disease 0.8 8, ,267 progression overall 1 8, , , ,913 AIDS (1987) definition No reinjection 8, ,300 reinjection 8, ,457 DEALE LE = , ,421 LE = , ,384 LE = , ,724 *compared to do nothing baseline **incremental cost-utility ratio

71 62 Table 24: Results of one-way sensitivity analyses that adopt societal perspective for treatments using polyalkylimide gel Parameters Value Incremental Incremental ICUR** Cost* QALY* Discount rate 0% 9, ,163 3% 9, ,583 5% 8, ,896 Probability of infection 0 7, , , , , ,807 Physician Quotes per ml 160 6, ,946 (more than 6 ml) 330 9, , , ,221 Travel for consultation day 10 minutes 9, , minutes 9, ,583 Day off 9, ,174 Costs of Anti-biotics $61.11 CAD 9, ,583 $84.14 CAD 9, ,751 Utilities gained Full Model 9, ,583 Short Model 9, ,836 CD4 level at the start All CD4>500 9, ,869 Base-case 9, ,583 All 9, , <CD4<500 Age 40 9, , , ,583 Rate of disease 0.8 9, ,225 progression overall 1 9, , , ,223 AIDS (1987) definition No reinjection 8, ,338 Reinjection 9, ,583 DEALE LE = , ,768 LE = , ,623 LE = , ,822 *compared to do nothing baseline **incremental cost-utility ratio

72 63 Figure 7: Tornado diagram presenting results of one-way sensitivity analyses that adopt Ministry of Health perspective for treatments using polyalkylimide gel Figure 8: Tornado diagram presenting results of one-way sensitivity analyses that adopt societal perspective for treatments using polyalkylimide gel Two-way sensitivity analyses According to the results of one-way sensitivity analyses, the two most influential factors that affect the ICURs associated with the treatments for facial lipoatrophy are utility gained deriving from poly-l-lactic acid, and the amount of poly-l-lactic acid required for reinjection biennially. In contrast, the ICUR of treatments using polyalkylimide gel are sensitive to physician quotes per ml for more than 5 ml, and the probability of infection within the first year following treatment. We therefore conducted two-way sensitivity analyses to examine the impact of simultaneously

73 64 altering two of these variables; utility gained from poly-l-lactic acid and physician quotes for polyalkylimide gel; utility gained from poly-l-lactic acid and the amount of poly-l-lactic acid required for reinjection; physician quotes for polyalkylimide gel and the 1-year probability of polyalkylimide gel infection. The results of the two-way sensitivity analysis are presented in a region graph with a threshold line (TreeAge Software Inc 2009). A region graph was created in two-dimensional variable space with regions assigned to alternative treatments based on their optimality, calculated by the Net Monetary Benefit approach 7 (TreeAge Software Inc 2009), where willingness-to-pay (WTP) for an additional QALY was set at $50,000 in 2008 Canadian dollars as the implicit additional costs per one additional QALYs is $ 50,000 CAD in Canada (Anell). The threshold line (i.e. border between two regions) provides the point of indifference between treatments using poly-l-lactic acid and polyalkylimide gel. Due to the limitations of the Treeage software, the two-way sensitivity analysis on utility gained from poly-l-lactic acid is impossible to create We, then, manually conducted it using Treeage, Microsoft Excel and Microsoft Word. Green areas demonstrate that polyalkylimide gel treatments offer higher Net Monetary Benefit with the willingness to pay at $50,000 per QALY, whereas blue areas show a higher Net Monetary Benefit from poly-l-lactic acid, given the same willingness to pay. Figures 9 and 10 depict the results of a two-way sensitivity analysis of utility gained from poly-llactic acid and physician quotes for polyalkylimide gel by analyses taken from the Ministry of Health and Societal perspectives, respectively. As revealed by Figures 9 and 10, treatments using polyalkylimide gel yield higher Net Monetary Benefits if physician quotes for polyalkylimide gel are at $160 CAD per ml for more than 5 ml. The higher physician quotes lead to the lower Net Monetary Benefit of polyalkylimide gel against poly-l-lactic acid. In addition, treatments using polyalkylimide gel with a physician quote at $ 415 CAD and the utility transformation function with power equal to 0.47 provides higher Net Monetary Benefit for the analysis taken from the societal perspective, while treatments using poly-l-lactic acid with the same combination yield a higher Net Monetary Benefit for the analysis taken from the Ministry of Health perspective. 7

74 65 Figure 9: Two-way sensitivity analysis of utility gained from poly-l-lactic acid and physician quotes for polyalkylimide gel compared treatment using poly-l-lactic acid with treatment using polyalkylimide gel by analyses taken from the Ministry of Health perspective $160/mL $145/mL $330/mL $415/mL $500/mL The rows represent physician quotes for polyalkylimide gel per ml for more than 5 ml. The columns represent utility gained from poly-l-lactic acid deriving from various transformation functions. The green area (left) shows that treatment using polyalkylimide gel is optimal. The blue area (right) shows that treatment using poly-l-lactic acid is optimal. Figure 10: Two-way sensitivity analysis of utility gained from poly-l-lactic acid and physician quotes for polyalkylimide gel compared treatment using poly-l-lactic acid with treatment using polyalkylimide gel by analyses taken from the societal perspective $160/mL $145/mL $330/mL $415/mL $500/mL The rows represent physician quotes for polyalkylimide gel per ml for more than 5 ml. The columns represent utility gained from poly-l-lactic acid deriving from various transformation functions. The green area (left) shows that treatment using polyalkylimide gel is optimal. The blue area (right) shows that treatment using poly-l-lactic acid is optimal.

75 66 Figures 11 and 12 provide the results of a two-way sensitivity analysis of utility gained from poly-l-lactic acid and the quantity of poly-l-lactic acid required for reinjection. As shown by Figures 11 and 12, utility gained from poly-l-lactic acid from different transformation functions greatly impact on the Net Monetary Benefit, since the value of the power primarily determines which treatments lead to a higher Net Monetary Benefit. For example, if the values of the power are 2.3 and 2.9, poly-l-lactic would offer lower Net Monetary Benefit regardless of the quantity of poly-l-lactic acid required for reinjection. In contrast, treatments using poly-l-lactic acid provide a higher Net Monetary Benefit over treatments using polyalkylimide gel with a value of 1.6 as the power of transformation function. In addition, treatments using polyalkylimide gel provide a higher Net Monetary Benefit for analyses taken from the societal perspective if the power of transformation function of poly-l-lactic acid is either 0.47 or 0.56, with 1.25 vials of poly-l-lactic acid reinjection, while treatments using poly-l-lactic acid yield a higher Net Monetary Benefit for analyses taken from the Ministry of Health perspective. Figure 11: Two-way sensitivity analysis of utility gained from poly-l-lactic acid and quantity of poly-l-lactic acid required for reinjection compared treatment using poly-l-lactic acid with treatment using polyalkylimide gel by analysis taken Ministry of Health perspective 1 vial 1.25 vials 1.5 vials 1.75 vials 2 vials The rows represent quantity of poly-l-lactic acid required for each time of additional treatments (per vial). The columns represent utility gained from poly-l-lactic acid deriving from various transformation functions. The green area (left) shows that treatment using polyalkylimide gel is optimal. The blue area (right) shows that treatment using poly-l-lactic acid is optimal.

76 67 Figure 12: Two-way sensitivity analysis of utility gained from poly-l-lactic acid and quantity of poly-l-lactic acid required for reinjection compared treatment using poly-l-lactic acid with treatment using polyalkylimide gel by analysis taken societal 1 vial 1.25 vials 1.5 vials 1.75 vials 2 vials The rows represent quantity of poly-l-lactic acid required for each time of additional treatments (per vial). The columns represent utility gained from poly-l-lactic acid deriving from various transformation functions. The green area (left) shows that treatment using polyalkylimide gel is optimal. The blue area (right) shows that treatment using poly-l-lactic acid is optimal. Figures 13 and 14 display the results of a two-way sensitivity analysis of physician quotes for polyalkylimide gel and a 1 year probability of polyalkylimide gel infection. As revealed by Figures 13 and 14, treatments using polyalkylimide gel always yield higher Net Monetary Benefits if physician quotes for polyalkylimide gel are at $350 CAD per ml for more than 5 ml for analyses taken from the Ministry of Health perspective and $ 300 CAD for analyses taken from the Societal perspective. Similarly, treatments using polyalkylimide gel always lead to higher Net Monetary Benefits if a 1- year probability of infection is below 0.03 for analyses taken from the Ministry of Health perspective and 0.08 for analysis taken from the societal perspective.

77 68 Figure 13: Two-way sensitivity analysis of physician quotes for polyalkylimide gel and a 1 year probability of polyalkylimide gel infection compared treatment using poly-l-lactic acid with treatment using polyalkylimide gel by analysis taken from the Ministry of Health perspective The rows represent physician quotes for polyalkylimide gel per ml for more than 5 ml. The columns represent the 1-year probability of polyalkylimide gel infection. The green area (left) shows that treatment using polyalkylimide gel is optimal. The blue area (right) shows that treatment using poly-l-lactic acid is optimal. Figure 14: Two-way sensitivity analysis of physician quotes for polyalkylimide gel and a 1 year probability of polyalkylimide gel infection compared treatment using poly-l-lactic acid with treatment using polyalkylimide gel by analysis taken from the societal perspective. The rows represent physician quotes for polyalkylimide gel per ml for more than 5 ml. The columns represent the 1-year probability of polyalkylimide gel infection. The green area (left) shows that treatment using polyalkylimide gel is optimal. The blue area (right) shows that treatment using poly-l-lactic acid is optimal.