A Methodological Review of Models Used to Estimate the Cost-Effectiveness of Antiretroviral

A Methodological Review of Models Used to Estimate the Cost-Effectiveness of Antiretroviral Regimens for the Treatment of HIV Infection

Supplementary Material

Journal: Pharmacoeconomics

Author: Josephine Mauskopf, PhD

Corresponding Author:

Josephine Mauskopf

RTI Health Solutions

E-mail:

Contents

Table1...... Markov Cohort Models in Discrete or Continuous Time: Detailed Review

Table2...... Monte Carlo Simulation Models: Detailed Review

Table3...... Discrete Event Simulation or Continuous Time Markov Models: Detailed Review

Table4...... Other Models: Detailed Review

Table5. Economic Models for HIV Treatment: Number of Articles Found in Literature Search, January 11, 2013

Table6...... PubMed Search Strategy

Table7...... Embase Search Strategy

Table8...... Cochrane Library Search Strategy

References

Table1.Markov Cohort Models in Discrete or Continuous Time: Detailed Review

Author, Year, Country / Health States/
Cycle Time / Other Model Structure Assumptions / Uncertainty Analyses / Validation
Simpson et al. (1994) [1] / SixCD4 cell count ranges (50, 50-99, 100-149, 150-199, 200-299, ≥300) and death
Cycle time 2months / Treatment changed pattern of CD4 cell count change over 1year period only; OI risk by CD4 cell count; if OI, assigned life expectancy; if no OI after 12months, assigned life expectancy / One-way sensitivity analysis for risk of OI, starting distribution of CD4 cell counts, life expectancy, future costs / None
Chancellor et al. (1997) [2] / ThreeHIV disease states: (CD4 200 and 500 and non-AIDS, CD4 200 and non-AIDS, AIDS and death)
Cycle time 1year / Treatment reduced risk of transition between health states; 1- and 2year and continuous effect tested / One-way sensitivity analysis for relative risk of progression, duration of effect, starting cohort CD4 cell counts, costs, and discount rates / None
Mauskopf et al. (1998) [3] / Five CD4 cell count ranges (100, 100-199, 200-349, 350-500, 500) and death
Cycle time 1year / Treatment stops CD4 cell count decrease for 6.5 vs. 18months, then progression at natural history rate / One-way sensitivity analysisfor discount rates, duration of efficacy, post-trial efficacy treatment duration / Compared model results for firstyear to 1-year trial results
Biddle et al. (2000)[4] / Year 1 model:six CD4 cell count ranges (50, 50-99, 100-149, 150-199, 200-299, ≥300) and death
Cycle time 2months
Long-term model: four HIV disease states (CD4 ≥500; CD4 200 and 500 and non-AIDS; CD4 200 and non-AIDS; AIDS) and death
Cycle time 1year / Year 1 model: treatment changed pattern of CD4 cell count change over 1year period only; OI risk by CD4 cell count
Long-term model: first-line treatment increased CD4 cell count with one-time transition to higher CD4 cell count health state; median time to return to baseline CD4 cell count was used to modify transition rates between health states; median time to virologic breakthrough and switch was 24months longer for triple therapy than dual therapy. Second-line therapy with PIs,increase of 100CD4 cells for 18months; third-line PI, only 90% had CD4 cell count increase for 15months / One-way sensitivity analysisfor CD4 cell count at which treatment is started, duration of effect, and discount rates / None; but some validation of input parameters from published studies for durability of regimens including either nevirapine or PIs
Simpson et al. (2004) [5] / Twelve combinations of CD4 cell count and viral load
(CD4 500 +VL400 or ≥400; CD4 351-500 +VL400 or ≥400; CD4 201-350 +VL400, 400-19,999, or ≥20,000; CD4 50-200 +VL400, 400-19,999, 20,000-100,000, or 100,000; CD4 50 + VLany level) and death
3month cycle time / First-line treatment efficacy allows progress into better or worse health states or health state remains unchanged; after 12months, assume no further divergence in survival curve; switch to secondline when CD4 500 and VL≥400; switch to third line when CD4 350 and VL≥400; health state improves on switch for 3months (CD4 cell count increase and viral load decrease) then changes at the rates observed in two large observational database studies of people from their third quarter of HAART to the end of the study; overall OI rates varied by health state but relative frequency of OIs assumed the same for all health states; for this study only a sensitivity analysis was run that allowed differential efficacy for the second-line regimen based on resistance patterns after first-line / One-way sensitivity analysis for costs, utility weights and OI event rates, and discount rates / Model validated in separate publication though validation of OI rates from second observational database
Hornberger et al. (2005) [6] / FourHIV-related stages: start treatment no virologic failure (VF), VF, immunologic failure (IF) defined as return of CD4 cell count to baseline level, AIDS, and death
Monthly cycle time / Virologic failure defined as viral load never suppressed or rebound after suppression; rates of VF for first 12months as in the trial; rates of VF after 12months based on trial 3- to 12month failure rates; CD4 cell count increases based on trial data for those without VF; between VF and IF, rate of decline of CD4 cell count depends on amount of viral suppression attained during therapy; after IF, rate of CD4 cell count decline equal to untreated rate; patients allowed to re-optimize OBR after VF but no efficacy assumed / One-way sensitivity analyses for treatment effectiveness, relevant progression probabilities, discount rates, medical care costs, time horizon and utilities / Stated that estimates for optimized background regimen for AIDS-free and survival are similar to those from observational studies in highly treatment-experienced individuals
Hubben et al. (2007) [7] / See Simpson et al. (2004) [5] / See Simpson et al. (2004) [5]; but switch to second or third line when CD4 200 and VL≥400 / One-way sensitivity analysis for costs, disease progression rates, OI event rates, and discount rates / See Simpson et al. (2004) [5]
Simpson et al. (2007) [8] / See Simpson et al. (2004) [5] but also including CHD, CHD death, and other death / See Simpson et al. (2004) [5]; but switch to second or third line when CD4 350 and VL≥400; in addition, impact of ART on cholesterol levels and CHD risk using Framingham risk equations included / One-way and multi-way sensitivity analyses and scenario analyses with different impacts on cholesterol and different relationships between cholesterol and CHD / See Simpson et al. (2004) [5]
Simpson et al. (2007) [9, 10] / See Simpson et al. (2004) [5] but also including CHD, CHD death, and other death / See Simpson et al. (2004) [5]; but risk of viral rebound different for the different treatment regimens based on proportion with virologic suppression in the trial for the first-line regimen; switch to second or third line when CD4 350 and VL≥400; in addition, impact of ART on cholesterol levels and CHD risk using Framingham risk equations included / One-way sensitivity analysis for duration of viral suppression, risk of heart disease, costs / See Simpson et al. (2004) [5]
Badia et al. (2007) [11] / Four HIV-related stages (start treatment and no virologic failure [VF], VF, IF, AIDS)and death
Monthly cycle time / See Hornberger et al. (2005)[6]
CD4 cell count constant at the value in the last week of the clinical trial for those without VF and decreasing as a function of suppression of VLin last week of trial for those with VF or AIDS until IF, then at natural history rates / One-way sensitivity analyses for rate of virologic failure, relevant progression probabilities, discount rates, medical care costs, time horizon and base CD4 cell count and viral load / None
Simpson et al. (2008) [12] / See Simpson et al. (2004) [5] / See Simpson et al. (2004) [5]; but switch to second or third line when CD4 200 and VL≥400 / One-way and probabilisticsensitivity analyses (PSA) based on cost of AIDS events, discount rate, utilities, switch drug costs, adverse event costs, and switch treatment efficacy; both Monte Carlo and bootstrap approaches for PSA / See Simpson et al. (2004) [5]
Simpson et al. (2009) [13, 14] / See Simpson et al. (2004) [5] but also including CHD, CHD death, AIDS event, and other death
3month cycle time / See Simpson et al. (2004) [5]; but switch to second or third line when CD4 350 and VL≥400; in addition, impact of ART on cholesterol levels and CHD risk using Framingham risk equations included / One-way sensitivity analyses on cost of ritonavir boosting, cost of statins, cost of AIDS event, variation in prices of PIs for first regimen, effect of difference in baseline risk of CHD, cost of drugs used for second- and third-line treatment / Some comparison with CHD rates in the DAD observational cohort
Elbasha et al. (2009) [15] / 42 health states of combined CD4cell count (500; 301-500; 201-300; 101-200; 51-100; 0-50) and viral load (100,000; 30,001-100,000; 10,001-30,000; 3,001-10,000; 401-3,000; 51-400; 0-50), also divided into those with and without history of OI; those with an OI enter temporary health states based on OI type (PCP, MAC, CMV, toxoplasmosis, fungal infections, other); continuous transitions modeled / Drug efficacy estimated using trial data for weeks 0 to 4 and 4 to 48 for movement through viral load states based on differential equations; CD4 cell count changes were estimated based on a regression equation relating viral load to CD4 cell count changes; virologic and immunologic failure rates were assumed to remain stable at the observed trial rates for 5years, then 8% failure rate assumed while still on raltegravir; when raltegravir discontinued, rate of movement between states was set equal to the control group; no switch included, just remain on same OBR after discontinuing raltegravir; OI risk adjusted for impact of HAART; mortality from OIs, HIV-related, and general were included / One-way sensitivity analyses for utility weights, risk of OIs, price of raltegravir, duration of treatment with raltegravir, mortality rates, efficacy duration and treatment failure rates; variability analyses for analytic time horizon and discount rate / Model results compared to data from trials, cohort studies, and ART intervention studies including annual rate of death and outcomes of OIs
Chaudhary et al. (2009) [16] / See Elbasha et al. (2009) [15] / See Elbasha et al. (2009) [15] / See Elbasha et al. (2009) [15] / See Elbasha et al. (2009) [15]; also life expectancy compared to that estimated in other models in treatment-experienced patients (life expectancy in this model is much longer than in other models)
Moreno et al. (2010) [17] / Six CD4 cell count health states (500, 351-500, 201-350, 101-200, 51-100, 0-50) and death
1month cycle time with half-cycle correction / Transitions in firstyear taken from clinical trial; base-case duration of maraviroc therapy was 1year only; after 1year, both arms received OBR alone with virologic failure and CD4 cell count decline at the OBR rate for 6-12months; after virologic failure, 1month lag before CD4 cell count decline; CD4 cell count decline after AIDS taken from untreated cohort; AIDS-defining event estimated using observational data / One-way and probabilistic sensitivity analyses; one-way included treatment duration, model time horizon, rate of treatment success, lag time before CD4 cell count decline, transition probabilities between CD4 cell states, utilities, AIDS-defining event mortality rates, discount rates, and cost of maraviroc, OBT, enfuvirtide, HIV-related care, and AEs; 1,000 runs for PSA / None
Mauskopf et al. (2010) [18] / Six CD4 cell count health states (500, 351-500, 201-350, 101-200, 51-100, 0-50) plus death
3month cycle time / Efficacy via a rapid rise in CD4 cell count followed by a slower increase; rate of CD4 cell count increase (rapid and slow) and duration of slow increase varied by virologic response at 24 weeks (viral suppression,1log10 drop but not suppressed, and 1log10 drop); switch to second OBR after treatment failure (defined as decline in CD4 cell count for at least 6months) and continued for remaining lifetime; enfuvirtide therapydiscontinued after 6month period of decreasing CD4 cell count; after second regimen failed, declining CD4 cell count at a lower rate than untreated based on observational studies; HIV-related mortality rates related to CD4 cell count; non–HIV-related mortality related to age and washigher than general mortality rates by a factor / One-way and probabilistic sensitivity analyses; one-way analyses included extent and duration of treatment-induced increase in CD4 cell count for both initial treatment regimen and second-line treatment regimen; drug use, utility, costs, and HIV-related mortality; 1,000 runs for PSA; variability analyses for time horizon and discount rates / None
Moeremans et al. (2010) [19, 20] / See Mauskopf et al. (2010) [18] / See Mauskopf et al. (2010) [18] / See Mauskopf et al. (2010) [18] / None
Brogan et al. (2010) [21] / See Mauskopf et al. (2010) [18] / See Mauskopf et al. (2010) [18] / See Mauskopf et al. (2010) [18] / None
Thuresson et al. (2011) [22] / Three health states for angina, myocardial infarction, stroke and death; also states for first-, second-, and third-line treatment; no health states for CD4 cell count
Cycle time ?? / Treatment efficacy based on treatment discontinuations for any reason; after 96 weeks, discontinuation rate assumed equal for both regimens; subsequent treatment depended on reason for failure; CD4 cell count not tracked; adverse event incidence based on Framingham equation; costs for CV adverse events only; disutility from CVs from published studies and overall utility rate for HIV infection not varied by disease severity; second-line regimen depends on reach for discontinuation of first-line (adverse events, virologic failure, or compliance); efficacy of second- and third-line regimens from clinical trials with discontinuation rates resulting in switch to next line / One-way and PSA sensitivity analyses varying adverse events rates, treatment discontinuationrates, and cardiovascular profile / Validation against life expectancy in disease cohorts
Colombo et al. (2011) [23] / Eight health states from combinations of CD4 cell count andVL
(CD4500, VL50;CD4500, VL50; CD4350-500, VL50; CD4350-500, VL50; CD4200-350, VL50; CD4200-350, VL50;CD4200, VL50;CD4200, VL50) plus AIDS and death
Cycle time of 1year and time horizon of 10years / Efficacy measured by response to therapy and person placed in health states with viral load50; CD4 cell count allowed to move through states but no information provided on rates; response taken from trial data and assumed to stay constant at last observed value; treatment failures assigned to CD4 cell count states with VL50 and CD4 cell count assumed to decrease at untreated rates; with CD4 cell count 200, transition to the AIDS state occurred; HIV-linked mortality related to CD4 cell count / One-way sensitivity analysis including virologic response, HIV-associated mortality rates, and initial distribution of patient CD4 cell counts and viral load / Compared average annual costs from the model with annual costs from an administrative database
Brogan et al. (2011) [24] / Six CD4 cell count health states(500, 351-500, 201-350, 101-200, 51-100, ≤50) plus death
1-year cycle period and lifetime horizon / Efficacy for first-line therapy measured using indirect comparison analysis for virologic response and CD4 cell count increase for the 144-week trial period; switch rate and CD4 cell count increase after trial period extrapolated from trial data; all patients switch to same second and subsequent therapy lines; durationon second- and third-line therapy taken from observational data; CD4 cell count increases based on clinical trials; no increase in CD4 cell count for salvage therapy; remain on salvage therapy for remaining lifetime; HIV-related mortality varied by CD4 cell count, while general mortality varied by age and sexwith a 2.5factor increase; adverse events for first-line therapy included in the model / One-way and probabilistic sensitivity analysis with realistic ranges for variables including efficacy of first-line therapy, annual changes in CD4 cell counts for later lines of therapy, and antiretroviral drug costs for later lines of therapy; variability analysis for model time horizon, discount rate, and incidence and cost of adverse events / Validated results with observational data on life expectancy for those with HIV infection initiating therapy and with percentage of deaths from HIV-related causes
Simpson et al. (2011)[25] / See Simpson et al. (2004) [5]; in addition CHD, CHD death, AIDS event, and other death and lipoatrophy submodel
3-month cycle time / See Simpson et al. (2004) [5]; but switch to second or third line when CD4 cell count 350 and VL≥400; resistance profile from first-line treatment assumed to increase the cost of third-line treatment by adding enfuvirtide; impact of ART on cholesterol levels and CHD risk using Framingham risk equations included; impact of ART on lipoatrophy with costs and disutility also included / One-way sensitivity analyses on impact of resistance on the cost of the third regimen and cost of AIDS, heart disease, or lipoatrophy and utility loss with lipoatrophy / See Simpson et al. (2004) [5]
Mauskopf et al. (2012) [26] / See Mauskopf et al.(2010) [18] / See Mauskopf et al. (2010) [18]; but switch regimen to include raltegravir plus OBR / See Mauskopf et al. (2010) [18]; variability analysis also included patient characteristics / One-year CD4 cell count distribution compared with that from the clinical trials used for efficacy estimates; proportion of HIV-related deaths compared with observed data in Canada; life expectancy compared with published estimates of life expectancy
Simpson et al. (2012) [27] / See Simpson et al. (2004) [5]; in additionAIDS event, CHD, CHD death, and other death;
3-month cycle time / See Simpson et al. (2004) [5];but switch to second or third line when CD4 cell count 350 and VL≥400; impact of ART on cholesterol levels and CHD risk using Framingham risk equations included / One-way sensitivity analysis for baseline risk of heart disease and costs of different lines of therapyonly / See Simpson et al. (2004) [5]

AE=adverse events; AIDS=acquired immunodeficiency syndrome; ART=antiretroviral therapy; CD4=CD4 T-lymphocyte or CD4 T-lymphocyte cell count; CHD=coronary heart disease; CMV=cytomegalovirus; CV=cardiovascular; DAD=Data Collection on Adverse events of Anti-HIV Drugs (observational cohort study); HAART=highly active antiretroviral therapy; HIV=human immunodeficiency virus; IF=immunologic failure; MAC=Mycobacterium avium complex; MI=myocardial infarction; OBR=optimized background regimen; OBT=optimized background therapy; OIs=opportunistic infections; PCP=Pneumocystis pneumonia; PI=protease inhibitor; PSA=probabilistic sensitivity analysis; VF=virologic failure; VL=viral load.

Table2.Monte Carlo Simulation Models: Detailed Review

Author, Year, Country / Health Events/ CycleTime / Other Model Structure Assumptions / Uncertainty Analysis / Validation
Miners et al. (2001) [28] / Three HIV health events (CD4 ≥200 no AIDS, CD4 200 no AIDS, AIDS) plus death;
1-year cycle / Cohort of 1,000; treatment effects taken from observational data with different effects for firstyear and subsequentyears of treatment; HAART effect mean of 5years but costs of dual NRTI or HAART continued until death / One-way sensitivity analyses for additional cost of third drug, probability of disease progression, duration of cost for third drug, duration of HAART effect, model time horizon and discount rate / None
Freedberg et al. (2001) [29]
CEPAC model / Chronic health events stratified by CD4 cell count (500, 301-500, 201-300, 101-200, 51-100, 0-50), viral load (30,000; 10,001-30,000; 3,001-10,000; 501-3000;≤500), acute health state (with OIs) and history of clinical events plusdeath