Introduction

Acquired immunodeficiency syndrome (AIDS) continues to be a major health epidemic with an estimated 39.5 million people living with human immunodeficiency virus (HIV) worldwide and 2.9 million people dying from AIDS in 2006 1. Based on data from the Mexican National Center for the Prevention and the Control of the HIV/AIDS, it was estimated in 2006 that 182,000 people in Mexico were living with the disease 2, 3. In 2007, the World Health Organization (WHO) estimated HIV prevalence in Mexico, among adults aged 15 years and older, to be 244 per 100,000 population 4. From 1992 to 1997, only social security beneficiaries had free access to antiretroviral therapies (ARTs) and HIV care in Mexico 5. However, by 2006, accessibility to ART had increased to 78% of the infected Mexican population 6.

Currently, highly active antiretroviral therapy (HAART) represents the mainstay among treatments for HIV-infected individuals, which suppress viral load, improve immune function, and delay the progression of the disease. Since its introduction in 1996, HAART has dramatically improved survival rates for individuals with HIV infection. However, new classes of antiretroviral agents with novel mechanisms of action are now receiving greater consideration due to side effects and poor tolerability with existing agents and the emergence of drug-resistant viral strains after long exposure to ART therapies, resulting in treatment failure.

The newly developed drug maraviroc is the first of a class known as CCR5 antagonists. Maraviroc selectively and reversibly binds to CCR5 receptor on CD4+ cells, preventing interaction with the gp120 protein of CCR5-tropic virions and subsequent entry. In double-blind, placebo-controlled clinical trials, maraviroc in addition to optimized background therapy (OBT) were compared to OBT alone in treatment-experienced adults infected with CCR5-tropic HIV-1 7. Patients treated with maraviroc achieved a substantially greater viral load reduction than did patients treated with OBT alone (−1.84 and −0.78 log10 copies/mL, respectively), with over twice the proportion of patients achieving undetectable levels of HIV-1 (56.1% and 22.5% with <400 copies/mL at 48 weeks) 8, 9.

To predict the long-term clinical and economic impact of maraviroc as add-on therapy to OBT, we used the Anti Retroviral Analysis by Monte Carlo Individual Simulation (ARAMIS) model 10 to analyze the cost-effectiveness of maraviroc for the management of individuals infected with only CCR5-tropic HIV-1 detectable, and who have evidence of viral replication and HIV-1 strains resistant to multiple antiretroviral agents (The ARAMIS cost-effectiveness model is the property of Pfizer Inc., and is protected by US and international copyright laws. Unauthorized reproduction, modification, republication, distribution, or display of all, or any portion, of the ARAMIS cost-effectiveness model is prohibited). The analysis was conducted from the perspective of the health care payer in Mexico to a lifetime horizon.

Methods Analytic Overview

The analysis was conducted using ARAMIS, a previously validated and published computer-based microsimulation model of HIV disease progression and treatment 10. The ARAMIS model was designed to be applicable to countries with well-developed systems for the care of individuals with HIV/AIDS, and was adapted to include Mexico-specific inputs. The model was used to extrapolate 48-week results from the MOTIVATE-1 and -2 studies and predict clinical and economic outcomes over a lifetime, using disease progression data derived from epidemiological cohort studies. Individuals screened for entry to the MOTIVATE trials (the “MOTIVATE screening cohort”) included adults with an HIV-1 RNA of greater than 5000 copies/mL despite at least 6 months of prior therapy with at least one agent from three ART classes or documented resistance or intolerance to at least one member of each class 8, 9. All individuals in the MOTIVATE screening cohort were tested for viral tropism status, and only those in whom only CCR5 tropic virus was detectable were eligible for entry into the trial.

The primary economic analysis was a comparison of two treatment strategies: 1) a “contingent maraviroc” strategy in which individuals would be treated with OBT, with or without the addition of maraviroc according to viral tropism test results (“OBT ± maraviroc”) versus 2) an “OBT alone” strategy of treating the entire population with OBT alone, without testing for tropism. Hence, the economic comparison was specified slightly differently from the design of the MOTIVATE trials, in order to represent the ex ante policy decision on the use of maraviroc in a treatment-experienced population, without prior knowledge of individuals' tropism status. The primary outcomes of the analysis were lifetime costs, life expectancy and quality-adjusted life years (QALYs) resulting from each treatment strategy. Incremental cost effectiveness ratios (ICERs), including cost per life year and cost per QALY gained, were calculated and reported in 2008 Mexican pesos (with values converted to 2008 US$ reported in parentheses). The analysis was conducted using a lifetime time horizon and a Mexican payer perspective.

The ARAMIS Model Structure and Global Parameters

The ARAMIS model has previously been described within the context of an analysis of maraviroc conducted from a US payer perspective 10. The microsimulation model simulates the progression of a cohort, one individual at a time, through a set of health states representing the natural history of HIV/AIDS from initial infection to death. Six mutually exclusive health states are defined by CD4+ cell count categories (>500, 351–500, 201–350, 101–200, 51–100, and <50 cells/µL), and mortality is modeled as a seventh, absorbing state (Fig. 1). In a given 1-month cycle, an individual may remain in the same CD4+ state, progress to an adjacent CD4+ state, enter an opportunistic infection transition state, or die.

Markov state transition diagram for the ARAMIS model. OI, opportunistic infection. ↓ Denotes allowable transitions. ↑ Denotes transitions only allowable with treatment.

Disease progression is modeled as CD4+ cell decline, the occurrence of an acute opportunistic infection, or death. CD4+ cell count is the marker of current disease state (which is tracked at an individual patient-level using a tracker variable) 11. Depending upon treatment status and history of opportunistic infections, CD4+ cell count determines the subsequent risk of opportunistic infections and HIV-related mortality that does not occur within the acute phase of an opportunistic infection. Mortality within the month immediately after a new OI (referred to as “acute HIV mortality”) is captured separately in the model.

Data on disease progression were based on published cohort studies 12-16. The natural rate of decline of CD4+ cells in untreated patients is known to be strongly related to the HIV viral load set point (i.e., the stabilized viral load in the postacute phase) (Table 1) 15. As set point data were unavailable for the study population, baseline viral load was deemed to be a reasonable proxy for modeling the rate of decline of CD4+ cells in patients failing treatment. The loss of CD4+ cells in HIV-infected individuals who are untreated or failing treatment drives transition through successively lower CD4+ states, while successful antiretroviral treatment results in reconstitution of CD4+ cells and transitions from lower to higher CD4+ states. The CD4+ cell count trajectory and other aspects of each individual's simulated lifetime clinical history are recorded within the model.

Table 1. Mean monthly CD4+ cell count decline by HIV RNA stratum HIV RNA (copies/mL) CD4+ count (cells/mm3) >30,000 6.375 10,001–30,000 5.400 3,001–10,000 4.600 501–3,000 3.733 ≤500 3.025

Independently of CD4+ cell count, the risk of an opportunistic infection and chronic HIV mortality is modified by history of an opportunistic infection and current exposure to antiretroviral treatment 13, 14. In the model, causes of death include chronic HIV disease, acute fatal opportunistic infection, and all-cause mortality.

The probabilities of progression between health states each cycle depend on each individual's baseline characteristics, including current CD4+ status, viral load set point and history of opportunistic infections. The risk of events does not remain fixed through the model, but depends on each individual's accumulated history of events, particularly the occurrence of opportunistic infections and antiretroviral treatment status. Concurrently, individuals progress through successive lines of ART contingent on the occurrence of virologic failure or adverse events. With successful treatment, viral load is assumed to drop below detectable levels (<400 copies/mL) at 48 weeks and is described as “suppressed.” Although tests with a lower detection limit of 50 copies/mL are now in common use, the 400 copies/mL criterion is used in ARAMIS because many of the clinical trials used to calibrate the model reported only suppression rates to <400 copies/mL 16, 17. Forty-eight weeks has become established as a standard length of follow-up for clinical trials of ART. In ARAMIS, when HIV is suppressed, the decline in CD4+ cell count is arrested. If suppression is sustained, treatment specific CD4+ count recovery is modeled with time.

Each cycle, ARAMIS updates its account of each individual's accumulated survival time, along with the associated costs and quality-adjusted survival associated with his or her pathway through the modeled health states and treatment. Specific monthly costs are assigned to treatment status and to each CD4+ state, the latter representing the costs of routine health care in HIV/AIDS, as described later. Additional cost tolls are imposed when individuals experience acute opportunistic infections, which are additive to the monthly routine care cost. Each CD4+ state and opportunistic infection state is associated with a utility value on a 0 to 1 scale, by which survival is quality-adjusted to allow calculation of QALYs. Opportunistic infections are modeled as transitions that may be encountered at each cycle. In this manner, individuals incur transition tolls (cost of treatment and quality of life decrements) when opportunistic infections occur.

Study cohort. The distributions of the baseline characteristics of the cohort simulated in this study were directly derived from individual level data for the MOTIVATE screening cohort, as previously described (mean values are shown in Table 2, n = 3244 patients screened) 7-9. A secondary analysis was performed simulating the subset of individuals with only CCR5-tropic HIV-1 detectable after tropism testing, and who were thus eligible to enroll in MOTIVATE (the “MOTIVATE trial cohort”).

Table 2. Baseline characteristics of cohorts from the MOTIVATE trials Characteristic Screening population Trial population (CCR5 monotropic) Mean age at baseline (years) 45.3 45.6 Gender (% male) 86 87 Mean CD4+ count (cells/µL) 183.94 217.15 Viral load set point (log10 copies/mL HIV-1 RNA) 4.72 4.76 Proportion receiving enfuvirtide with optimized background therapy (%) 58 56 Proportion exclusively CCR5 tropic (%) 51 100 Proportion entering with history of opportunistic infection (%)  Bacterial 7 8  Fungal 43 41  Viral 20 19  Protozoal 3 3  Other illness 30 28  Any opportunistic infection 67 63

ART intervention. The treatment effects for individuals in the OBT alone strategy and for individuals in the contingent maraviroc strategy who were not CCR5-tropic were derived from pooled 48-week data from the placebo arms of the MOTIVATE -1 and -2. The treatment effects for CCR5 positive individuals in the maraviroc arm were derived from 48-week pooled data from the 300 mg twice-daily maraviroc arms from the same studies (Table 3) 8.] Within the model, CD4+ cell count increased in each cycle that viral suppression was sustained. Consistent with the normal pattern observed in trials of ARTs 8, 18-22, a rapid CD4+ increase was modeled during the first two months of effective antiretroviral therapy, and a slower rate thereafter. The modeled rate of CD4+ cell recovery to 48 weeks was treatment-specific and based on observed data from the MOTIVATE trials.

Table 3. Efficacy parameters from the MOTIVATE trials First regimen MVC + OBT (CCR5+) OBT alone Mean Mean First year virologic suppression rate (<400 copies/mL)  Pooled over both trials 8* 56.1% 22.5% Monthly late failure rate†  Week 48 to week 96 0.75% 1.06%  Thereafter 1.06% 1.06% Monthly CD4+ increase when successful (cells/µL)‡  First 2 months 54.1 39.3  Months 2 to 12§ 5.4 5.3  Thereafter§ 5.6 5.6 Monthly adverse event risk (grade 3 and 4)* 1.2% 1.0% Subsequent regimen, given failure of first regimen (reoptimized background therapy; rOBT) 42, 43 ¶ 20%#OBT + T20 80%OBT 20%#OBT + T20 80%OBT First year virologic suppression rate (<400 copies/mL) 30.4% 12% 30.4% 12% Monthly late failure rate 1.87% 1.87% 1.87% 1.87% Monthly CD4+ cell increase when successful (cells/µL)‡  First 2 months 33.64 16.63 33.64 16.63  Months 2 to 12 3.39 1.68 3.39 1.68  Thereafter 3.50 1.73 3.50 1.73 * Pooled 48-week data from trials A4001027 and A4001028 (MOTIVATE-1 and -2). Maraviroc efficacy taken from 300 mg twice-daily maraviroc arms of trials. Pfizer Global Research and Development, data on file. † The late failure rates in fully suppressed patients are estimated by projecting the rate of viral rebound between week 16 and week 48. ‡ Success is defined as viral load dropping below detectable levels (<400 copies/mL) at 48 weeks. § Estimated by linear regression on pooled (MVC and placebo arms) CD4+ count data from week 8 to 48. ¶ Failure on either of the compared regimens results in a switch to an unspecified reoptimized background therapy regimen; a four-month time lag was assumed between failure and switching, to mimic the likely delay in discovering and confirming virologic failure in clinical practice. # 20% of patients were assumed to receive T20 in their subsequent OBT regimen based on unpublished information for the MOTIVATE open-label continuation phase. MVC, maraviroc; OBT, optimized background therapy.

Beyond 48 weeks, CD4+ increases were conservatively assumed not to differ between treatments and were estimated by linear extrapolation from the observed trial data. Individuals who experienced virologic failure after 48 weeks were considered “late failures.” The late failure rate was treatment-specific and derived by linear regression from the observed failure rate from the peak of suppression at week 8 through week 48 in the MOTIVATE trials. As no long-term studies provide information on sustainability of suppression beyond 10 years, all patients modeled as suppressed at 10 years were conservatively assumed to fail at this point in time.

After an individual failed either of the compared regimens, they were assumed to switch to an unspecified reoptimized background therapy regimen (rOBT), based on treatment history and the results of resistance testing. A 4-month time lag was assumed between failure and switching, to mimic the likely delay in discovering and confirming virologic failure in clinical practice. Based on unpublished data from the open-label continuation phase of the MOTIVATE studies, it was assumed that for 20% of these switched patients, the rOBT regimen would include enfuvirtide. After virologic failure, the CD4+ cell count was modeled to remain static for 12 months 23 and thereafter to decline at a rate determined by the viral load set point 23. For initial failures (i.e., those whose CD4+ count never reaches <400 copies/mL during their lifetime), CD4+ cell count was modeled to remain static for 12 months after the start of the model, consistent with the assumption of virologic failure to the treatment received prior to entering the analysis. Failing patients were assumed to remain on ART, and, consistent with evidence from the literature 13, 14, they were assumed to benefit from a reduced risk of opportunistic infection incidence conferred by treatment in the absence of suppression.

ART-related adverse events were modeled based on pooled rates of drug-related grade 3 and 4 adverse events from the MOTIVATE studies, adjusted for exposure time. Monthly risks of an adverse event for individuals receiving OBT plus maraviroc and OBT alone were 1.2% and 1.0%, respectively.

Routine care and testing. Prophylaxis against Pneumocystis (PCP) and Mycobacterium avium complex (MAC) was assumed to be initiated and ended when the CD4+ cell count passed thresholds of 200 and 50 cells per µl, respectively, and consisted of trimethoprim-sulfamethoxazole 800 mg/160 mg (TMP-SMX) 24. Routine testing was assumed to include: CD4+ cell count, HIV-1 RNA level, and, after treatment failure, genotypic resistance testing, and was assumed to be performed according to US clinical guidelines 25. The cost of one viral tropism test was applied to all individuals in the maraviroc treatment strategy arm. Although the tropism test, as it was available and used in MOTIVATE trials, is known to be imperfect, sensitivity and specificity were set to 100% in the analysis to avoid double-counting, because individuals with false-positive test results contributed to the reported results for the OBT+ maraviroc arms of the trials. The frequency of the remaining tests followed clinical guidelines.

Quality of life. With the exception of the impact of adverse events, utility values used in the model for calculating quality-adjusted life expectancy were based on published estimates (Table 4) 26, 27. In the absence of published utility estimates of the impact of adverse events in HIV infection, a decrement of 0.14 was applied for one cycle, based on expert opinion, for all patients experiencing a grade 3 or 4 adverse event. This is somewhat less than the utility decrement of 0.20 reported for gastrointestinal adverse events in chemotherapy 28.

Table 4. Utility values by model health state Health state Utility value Duration* Chronic HIV by CD4+ category 26  CD4+ >500 0.870 Present month  CD4+ 351–500 0.860 Present month  CD4+ 201–350 0.860 Present month  CD4+ 101–200 0.850 Present month  CD4+ 51–100 0.850 Present month  CD4+ <50 0.832 Present month Acute OI 27  Bacterial 0.561 3 months  Protozoal 0.561 3 months  Viral 0.652 3 months  Fungal 0.652 3 months  Other illness 0.561 3 months Post-acute OI 26  Bacterial 0.735 6 months  Protozoal 0.731 Rest of life  Viral 0.760 Rest of life  Fungal 0.743 6 months  Other illness 0.770 Rest of life Disutility of AEs 0.140 Present month * When more than one criterion applies, the lowest utility value is applied. AEs, adverse events; OI, opportunistic infection. Adaptation of ARAMIS to Mexico

Adaptation of ARAMIS to the Mexican setting involved use of Mexico-specific costs, age- and gender-specific population mortality data for risk of deaths due to non-HIV-related causes 29, and a discount rate for costs and outcomes of 5% 30.

The scope of Mexico-specific costs included direct monthly health care costs for routine HIV care (for both inpatients and outpatients) including ART, costs of tropism testing, adverse events associated with ART, acute and prophylactic treatment of opportunistic infections, CD4+ cell count tests, HIV RNA tests, treatments, and HIV- or opportunistic infection-related palliative care preceding deaths. The daily cost of maraviroc was $181.50 (US$16.96) based on 300 mg twice daily dosing. The complete set of cost inputs is presented in Table 5. Mexican pesos were inflated 31 and converted to US$ using a published rate of $1 Mexican Peso = US$0.09346 32.

Table 5. Summary of cost inputs 2008 Mexican pesos (US$) Daily ART drug costs 33  MVC (300 mg twice daily) 181.50 (16.96)  OBT 144.80 (13.53)  Enfuvirtide (90 mg twice daily subcutaneous injection) 796.41 (74.43) Routine care cost (per month)*†  No OI history   CD4+ >500 1,577 (147)   CD4+ 351–500 1,406 (131)   CD4+ 201–350 1,512 (141)   CD4+ 101–200 1,737 (162)   CD4+ 51–100 1,940 (181)   CD4+ <50 2,094 (196)  OI history 1,687 (158) OI prophylaxis (per month) 33  M. avium complex 223 (21)  P. carinii 9.3 (1) Acute OIs (per episode)†  Bacterial 82,285 (7,960)  Fungal 53,928 (5,040)  Other 41,703 (3,898)  Protozoal 64,732 (6,050)  Viral 78,811 (7,366) Last month of life costs  HIV death (assumed to be 69% of acute OI death cost 44) 56,784 (5,307)  Acute OI death (assumed same cost as bacterial OI episode) 82,285 (7,690) Test costs (per test)  Tropism test 7,700 (720)  CD4+ cell count 1,249 (117)  Viral load test 2,458 (230) Treatment-related AE‡ 4,463 (417) * Includes inpatient and outpatient care cost during one month. † Resource patterns and unit costs collected through review of patient records (n = 637) at nine hospitals in Mexico. ‡ Represents cost of a 1-day hospital stay. US$ = 2008 US$. AE, adverse event; MVC, maraviroc; OI, opportunistic infections; OBT, optimized background therapy.

Patterns of resource use were estimated based on a review of patient records from nine hospitals in Mexico City belonging to Social Security Mexican Institute (IMSS). The data collection study gathered information from 637 patients treated for HIV/AIDS at one of the participating hospitals during 2004 to 2005. The study sample had an average age of 49 years (standard deviation [SD] 9 years) and was 87% male. At the time of data collection, patients had been infected with HIV for an average of 8.8 years (SD 2.3), with 46% of the sample being diagnosed with AIDS and 78% having failed at least one line of therapy.

Through review of clinical records, the number of physician visits, laboratory tests, and utilization of retroviral, non-retroviral, and opportunistic infection-related therapies were identified to define patterns of routine care for individuals with HIV/AIDS. Data related to hospitalizations for HIV/AIDS-related treatment were also collected, including length of stay and frequency of inpatient visits, tests, surgeries, and medical procedures. Monthly routine care costs by CD4+ cell count for individuals with and without a history of opportunistic infection were determined by dividing the sum of all resources used within a year by 12 months. To determine costs associated with acute opportunistic infection treatment, HIV/AIDS-related resources utilized 30 days prior and 60 days after the diagnosis were considered. Because the metropolitan area from where the participating hospitals are located shows the highest HIV prevalence in Mexico, it is likely that the patterns of resource utilization observed are representative of the average treatments used in Mexico. Drug costs were obtained from the IMSS 33, while mean unit costs for other HIV/AIDS-related health care resources were based on costs collected from the hospital databases of the nine participating sites. It was assumed that patients experiencing grade 3 or 4 adverse events due to ART incur the additional cost of one hospital inpatient day per event; for simplicity, efforts were not made to model individual types of adverse events.

Results Base-Case Analysis

Applying the base-case assumptions to the MOTIVATE screening cohort, the model predicted that the contingent maraviroc strategy would lead to an increase in mean undiscounted life expectancy of 14 months (15.5% increase; 8.7 vs. 7.5 years), a 10-month increase in AIDS-free survival time (38%; 35.3 vs. 25.5 months), a decrease in the proportion of individuals experiencing AIDS of 2.4% (87.4% vs. 89.8%), and a decrease in the proportion of deaths caused by HIV (87.9% vs. 90.2%) relative to the OBT alone strategy. The contingent maraviroc strategy produced a decrease in the incidence of all types of opportunistic infections (19.2% vs. 20.0%) and a corresponding increase in the mean time to the first opportunistic infection event among incident cases (10.2 vs. 8.1 years). The average time to treatment failure was 28.5 months for the contingent maraviroc strategy compared to 18.7 months for the OBT alone strategy (Table 6). Corresponding outcomes for the comparison between treating with OBT + maraviroc versus treating with OBT alone in CCR5 monotropic individuals (i.e., based on the MOTIVATE trial cohort) are shown in Table 6.

Table 6. Undiscounted clinical outcomes predicted for the base-case scenario Clinical outcomes MOTIVATE screening cohort MOTIVATE trial cohort (CCR5 monotropic) Treat with OBT alone Test and treat with OBT ± MVC Treat with OBT alone Treat with OBT+MVC Life expectancy (years) 7.54 8.71 8.12 10.46 Proportion experiencing AIDS 89.8% 87.4% 88.0% 83.2% AIDS-free survival (months) 25.49 35.25 31.13 50.30 Causes of death