Cost effectiveness of atazanavir-ritonavir versus lopinavir-ritonavir in treatment-naïve human immunodeficiency virus-infected patients in the United States

Figures & data

Figure 1.  Structure of Markov microsimulation model of ATV + r compared with LTV/r among antiretroviral therapy–naïve HIV-infected patients. Patients enter model in LPV/r or ATV + r first-line treatment, distributed among the low-viral-load states (2, 4, 6, 8) based on CASTLE trial population. HIV states are defined by CD4 count and mRNA levels, event risks are based on HIV state, lipid levels, and treatment (LPV/r, ATV + r, or second-line), and patients advance to second-line treatment after severe diarrhea, severe hyperbilirubinemia, or treatment failure.

Table 1.  HIV states in microsimulation model of LPV/r and ATV + r for management of antiretroviral-naïve HIV-1-infected patients: definitions and transition probabilities*.

HIV state	CD4 count (cells/mL)	HIV RNA (copies/mL)	Probability of opportunistic infection	Probability of transitioning from prior state
				LPV/r	ATV + r†
1	>350	<50	0.017	0.852‡	0.864‡
2		≥50	0.022	0.067	0.054
3	201–350	<50	0.028	0.074	0.086
4		≥50	0.038	0.054	0.044
5	50–200	<50	0.051	0.029	0.033
6		≥50	0.099	0.047	0.038
7	<50	<50	0.179	0.002	0.002
8		≥50	0.179	0§	0§

*Transitions per patient, per 3-month cycle.

†ATV + r transitions based on a 19% lesser likelihood of transitioning to state with greater viral load (≥50 copies/mL) than when receiving LPV/r.

‡Probability of remaining in state 1.

§Transitions occurred from states other than 7.

ATV + r, atazanavir-ritonavir; LPV/r, lopinavir-ritonavir.

Table 2.  Model input parameter estimates and data sources for microsimulation of LPV/r and ATV + r for management of antiretroviral-naïve HIV-1-infected patients.

Parameter	Base-case estimate				Range for sensitivity analyses*				Reference
Baseline population characteristics
Age in years, mean (SD)	36.4 (9.6)				–				3
Female, %	31.6
CD4 cells/mL, mean (SD)	214 (133)
HIV RNA copies/mL, mean (SD)	181,484 (204,601)
DM, %	1.8
CHD, %	0.7
CHD risk (event/1000 person-years)					–				7
No prior CHD/DM	3.4
Prior CHD	76.4
Prior DM	16.4
Prior DM and CHD	101.9
Risk of CHD event being fatal	35.4%				Uniform (±10%)				8
OIs (event/patient/cycle)	0.017 to 0.179 (states 1 to 8)				Uniform (±50% of base-case rate)				9
Treatment-specific estimates
	LPV/r		ATV + r		LPV/r		ATV + r
Effect of treatment on transition to state  with RNA ≥50 copies/mL (greater viral load)	N/A (Baseline)		–19%		–		Uniform (±10%)		3
Effect of treatment on TC:HDL ratio	−0.17		−0.40		Normal (−0.17, 1.56);		Normal (−0.40, 2.82)
Adherence:
Rate	71%		82%		Normal (71.3%, 18.1%)		Normal (81.9%, 18.1%)		10,11
Effect of nonadherence on transition to state with  RNA ≥50 copies/mL (greater viral load)	+10%				Uniform (±10%)
Diarrhea, risk (%) at varying weeks on regimen†	0–48	49–96	0–48	49–96	0–48	49–96	0–48	49–96
Mild/moderate	2.74	0.06	0.57	0.06	1,198; 42,501	25; 43,675	252; 43,848	25; 44,075	3
Severe	0.23	0.17	0.00		1; 436	75; 43,625	N/A
Hyperbilirubinemia risk (%)†
Mild/moderate	0.46		12.24		2; 435		5; 387		3
Severe	0.00		0.68		N/A		3; 438

*Ranges shown are: means and standard deviations for normal distributions, which use 95% confidence intervals for minimum and maximum values; minimum and maximum values for uniform distributions; and number of events (alpha) and nonevents (beta) from CASTLE for beta distributions. Percentages are limited to zero-one bounds, and distribution parameters are based on assumptions that consider parameter type, original data source, and the inclusion of a broad range of values.

†Risks of diarrhea and hyperbilirubinemia are beta-distributed variables.

ATV + r, atazanavir-ritonavir; CHD, coronary heart disease; DM, diabetes mellitus; HDL, high-density lipoprotein; HIV, human immunodeficiency virus; LPV/r, lopinavir-ritonavir; N/A, not applicable; OIs, opportunistic infections; SD, standard deviation; TC, total cholesterol.

Table 3.  Net quality-adjusted life-years lost and costs incurred associated with HIV infection, diarrhea, hyperbilirubinemia, coronary heart disease, and opportunistic infections.

Disease outcome		Base-case estimate			Ranges for sensitivity analyses*		Reference
QALY
HIV state†
1		0.944			Uniform (0.888, 1)
2		0.935			Uniform (0.869, 1)
3		0.929			Uniform (0.858, 1)
4		0.932			Uniform (0.863, 1)		9
5		0.863			Uniform (0.726, 1)
6		0.849			Uniform (0.698, 1)
7		0.781			Uniform (0.562, 1)
8		0.781			Uniform (0.562, 1)
CHD		Ongoing		Event	Ongoing	Event
		0.90		0.88	Uniform (0.800, 1)	Uniform (0.760, 1)	16
Diarrhea		Mild/moderate		Severe	Mild/moderate	Severe
		0.900		0.641	Uniform (0.800, 1)	Uniform (0.283, 1)	17–19
Hyperbilirubinemia		Mild/moderate		Severe	Mild/moderate	Severe
		0.999		0.950	Uniform (0.998, 1)	Uniform (0.900, 1)	Expert opinion‡
Costs (2008 US dollars)
First-line		LPV/r		ATV + r
		4145		4604			WLP
Second-line§			4049
Effect of nonadherence on  first-line drug costs			–20%		Uniform (±10%)
Nondrug costs by HIV state**
	1, 3, 5		461		γ (9,51)
	2, 4, 6		720		γ (9,80)		1, 20, expert opinion‡
	7		479		γ (9,53)
	8†		758		γ (9,84)
CHD		Ongoing		Event
		806		12,885	γ (9,1432)		16
Diarrhea††			28		–		21
Hyperbilirubinemia			0		–		22,23, expert opinion‡
OI			4442		γ (9494)		24, MACS

*Ranges for QALY uniform distributions use as upper bounds the minimum of either the value shown above or the value of the next-least valued health state for that condition: QALY weights for severe diarrhea or hyperbilirubinemia must be ≤ those for the respective mild/moderate conditions; that for HIV state 2 must be ≤ that of state 1; that for HIV state 3 must be ≤ that of state 4; that for HIV state 4 must be ≤ that of state 2; that for HIV state 5 must be ≤ that of state 3; that for HIV state 6 must be ≤ that of state 5; those for HIV state 7 and 8 must be ≤ that of state 6. Values shown for gamma (γ) distributions are the alpha and beta parameters.

†State-specific HIV utility weights were based on published estimates from a previous model and used the EuroQol quality of life instrument; utilities for states that were defined more broadly here than in the reference model were calculated as averages of multiple states. Utility weight for state 8 includes that for states beyond first-line treatment.

‡Estimates incorporated input from infectious disease clinician for visit duration for nondrug costs as well as for hyperbilirubinemia QALY and cost estimates.

§40% abacavir/lamivudine + darunavir + ritonavir and 60% efavirenz/emtricitabine/tenofovir.

**Direct medical costs of: 1 chemistry panel, 1 complete blood count, and 1 CD4 count; a 10-minute visit to a physician for states 1–6, a 15-minute visit for state 7, and a 25-minute visit for state 8; 1 blood draw and 1 viral load assessment (ultrasensitive quantitation) in states 1, 3, 5, and 7; 2 blood draws and 2 viral load assessments in states 2, 4, 6, and 8.

††Loperamide 2 md bid.

ATV + r, atazanavir-ritonavir; CHD, coronary heart disease; LPV/r, lopinavir-ritonavir; MACS, Multicenter AIDS Cohort Study; QALY, quality-adjusted life year, WLP, wholesale list price.

Table 4.  Projected health-related, cost, and quality-adjusted life-year outcomes associated with LPV/r compared with ATV + r treatment for antiretroviral therapy--naïve HIV-infected patients: base-case results.

	Health-related outcomes
	Months in first-line treatment	AIDS	OI		CHD	Diarrhea	Hyperbilirubinemia
LPV/r	70.7	20.054	0.519		5.511	6.262	0.247
ATV + r	97.3	19.081	0.443		5.437	1.272	6.986
	Cost and QALY outcomes
	Cost	Incremental cost		QALY		Incremental QALY	Incremental cost-effectiveness ratio
LPV/r	$269,160	–		10.761		–	–
ATV + r	$275,986	$6826		11.020		0.258	$26,421

ATV + r, atazanavir-ritonavir; CHD, coronary heart disease; LPV/r, lopinavir-ritonavir; OI, opportunistic infection; QALY, quality-adjusted life year; WLP, wholesale list price.

Figure 2.  Predicted years spent in first-line treatment (panel a) and incidence (per 1000 person-years) of AIDS diagnoses (panel b), opportunistic infections (panel c), CHD events (panel d), diarrhea (panel e), and hyperbilirubinemia (panel f), by treatment type and year of follow-up.

Figure 3.  Results of one-way (panel a) and probabilistic (panel b) sensitivity analyses: incremental cost-effectiveness ratios (ICERs) of ATV + r compared with LPV/r with varying estimates of model input parameter values. Tornado diagram results (panel a) are displayed in order of each parameter’s magnitude of effect on the ICER, and lower, base case, and upper values used for each one-way sensitivity analysis are shown. In panel b, each of the 125 points represents one microsimulation analysis of 100,000 individual-level trials; 95% confidence ellipse is shown.