The human immunodeficiency
The Human immunodeficiency virus (HIV) is an infectious retrovirus that targets and infects special cells of the immune system responsible for fighting against infections, thus impairing or destroying their functional ability. Though the body will attempt to produce more CD4 cells, their numbers eventually decline with disease progression, thus weakening the immune system of infected individuals gradually, thus making them susceptible to other infectious diseases. The most advanced stage of infection-Acquired-immunodeficiency syndrome AIDS, though it could take 10-15 years for HIV infected individual to develop AIDS; but antiretroviral drugs have been shown to slow progression1.
South Africa (SA) is a Republic in the most Southern part of Africa with a total population of 48,282 000 people and life expectancy of 50/53 years at birth (male/female). Total health expenditure( THE) of $869 accounted for 8.6% of the country's GDP2. An estimated 5.2 million adult (15-49) were living with HIV and AIDS 3 and over 250,000 deaths as a result of AIDS as at 2008 in South Africa, thus an epidemic of enormous social and economic consequences 1, 3-5.
HAART is a combination of three drugs (2 nucleoside analogues and either a protease inhibitor or a non nucleoside analogue-NNRTI. It is a treatment used for people with HIV on an ongoing basis, affording a potent way of suppressing viral cell replication in the blood while trying to prevent the HIV cells from rapidly developing resistance to the individual ARV's; thus delaying disease progression to AIDS, and prolonging life. HAART has changed the landscape of HIV-related care in the developed countries with marked reduction in morbidity and mortality6-7.
HAART has been shown to be cost-effective for HIV infected people in developed countries where it is used as a routine care8, though HAART has been introduced in S/A since the mid-1990s, there is has been no economic evaluation to inform the decision to adopt the regimen into routine care9. Though some studies have been conducted,8-11, Sculptor et-al has suggested that economic evaluations be conducted continuously even after new technologies have been adopted so as to give a concise information about the cost and effect of all options available as it will enhance allocative efficiency and transparent deployment of limited health care resources12-14.
Economic evaluations are tools designed to guide explicit health resource allocation decisions by comparing the marginal cost and consequences of alternative health care interventions13, 15 to enable transparent health care coverage decisions16.
This cost-utility study examines the lifetime cost-effectiveness of maintenance therapy with HAART compared with a regimen which treats only AIDS-related opportunistic infections under different scenarios.
Study design and decision context
Cost-effectiveness' analysis was performed from a South African health services perspective using data on cost of providing HAART, cost of treating AIDS-related opportunistic infections in HIV infected individuals and other relevant health care cost provided by LSHTM. It compares the cost and benefit of providing ‘HAART' and ‘no HAART' with the aim of aiding South African government's decision as to whether or not providing HAART is cost-effective. The CD4 cells/µL count at which treatment was started i.e. 200<=CD4<350 is an important indicator for both cost of HAART and its clinical benefits8.
This study being a rapid assessment, study employed secondary data which were local, however where not available, data was imputed from other settings. All data was provided by LSHTM.
Study Population and Treatment option
Hypothetical cohort of 1000 HIV infected individuals receiving either HAART or treatment of HIV-related opportunistic in health facilities in South Africa. Intervention is former (HAART) while comparator later.
Main Outcome Measure
Cost in US $ per quality adjusted life year (QALY).
A Markov model (which allows for modelling of chronic and repetitive events) 6, 14, 17-20 21,was created using Microsoft Excel spreadsheet software, was used to simulate the progression of HIV infection22; which also allows for the full exploration of uncertainty around our cost-effectiveness estimates23. Disease progression was modelled as a Markov process, were all cases are ‘born' into the pre-symptomatic (200<=CD4<350 cell/µL) health state, with a given probability of moving into lower disease state, then into severe irreversible AIDS state and finally death with different probabilities of moving into the last two states.
There were four mutually exclusive health states based on CD4 counts (fig1). Modelling is an important tool- as economic evaluations depends on evidence on cost, health effects etc which are more often than not collated from various sources, thus model enables synthesis, by making explicit assumptions about incidence of disease, magnitude and duration of risk and relative risk of treatment options, determinants of utilization of health-care resources and of health-related quality of life 21. Models allow for exploring of how the incremental cost effectiveness ratio (ICER) might change with changes in key parameter in the model which are not readily observable in primary data24. If properly constructed, models can be used to assess the likely effective of the two treatment option, thus an important tool for policy formulation.
Figure 1: A simple representation of the markov model. The rectangular boxes represent health states, which vary with CD4 cell counts, and arrows represent the allowed transitions between the health states. All patients are born into the model in asymptomatic (200<=CD4<350) stage.
Similar model has been used in studies on SA8 and England13 with 4 health states known as the ‘'markov states'' defined by differing CD4 cell counts. This state's represent disease progression from early to late stages are: ‘200<=CD4<350', ‘50<CD4<200', ‘CD4<50 (AIDS)' and ‘Dead' which is consistent with WHO's staging of HIV infection25.
Model assumed each cycle to last a year; and at the end of each cycle, patients transited to a worst health state or remained within the same state based on annual average probabilities of disease progression at different CD4 cell count stages; that disease progression can only be slowed but not reversed with ‘dead' being an ‘absorbing state'26, as once within, it can't be exited. Other includes; baseline probabilities of individuals moving from one state to the other is not dependant on the CD4 cell counts of the individual before entering the asymptomatic state ‘'Makovian assumption''22; that the probability of dying from all cause mortality from stage 1-2 are not affected by HAART; that all individuals in stage 1&2 have same probability of dying from all other non HIV-related mortality without adjustment for sex and age; initial age of entry is relaxed for all.
Average annual transition probabilities for patients in the ‘no HAART' is given in table 2(as the deterministic values) while that of the HAAT group was calculated based on a relative risk of 0.5(95% CI, 0.4-0.6), estimated from an indirect comparison of RCTs comparing various regimens as no RCT had directly compared ‘HAART' with ‘treatment of HIV-related opportunistic infections (no HAART)' previously in SA.
Other task performed includes the calculation of ICER of both options, extended analysis by comparing the cost-effectiveness of commencing HAART at different CD4 stages compared with no HAART. Both deterministic and probabilistic sensitivity analysis was performed to assess the robustness of ICER to its key parameters with assumptions13. Also duration of treatment effect is known only for 10 years but the model was used to project progression of disease over a lifetime horizon for both alternatives22.
Cost and decision context
Cost data used for this analysis includes cost of providing HAART, cost of treating opportunistic infections when they occur, other related health care cost assumed to be same for each stage in both options. All cost data used were provided by LSHTM and measured in US$, and discounted at 3% as suggested by WHO and NICE and to allow for comparison with other studies8 27. However, because there was no information regarding cost associated with progressing to ‘dead stage', social services cost, productivity lost etc, made this study wholly a health services perspective28.Cost was discounted as there is often a difference in the timing of investment of health resources used and outcomes. Discounting is a way in which these future cost and outcomes are adjusted to account for the preference26, 29.
Health outcome (utility)
The Primary outcome measure of this study- QALY, was derived from the results a Euroqol (EQ-5D) questioner completed by HIV patients in SA, but valuation of scores were derived from UK population due to absence of local reference values typical to SA29.
Calculation of cost-effectiveness
All data were managed and analysed on Microsoft excel to arrive at aggregate values for the key variables in the cost-effectiveness analysis. QALY's were calculated by multiplying the utility associated with each health state by the number of life years spent in that state29. It was also assumed that for each year in a health state, a life year was accrued. Therefore, total life years in each model were the sum of the number of years spent in each markov state except the ‘dead' state.
The ICER was calculated by dividing the difference between the cost of HAART' and ‘no treatment' i.e. the incremental cost by the difference in effectiveness (incremental effectiveness) of both regimens and presented as cost per life year and cost per QALY gained. Sensitivity of the ICER to uncertainties in the parameters was tested with deterministic and probabilistic sensitivity analysis27, 30.
As Markov models allow for the projection of outcomes, same was applied on the data available for cost of ‘HAART' and ‘no HAART' to estimate lifetime cost and effect as the base case. Both options were run for a cycle of 60 years for the HAART group-to capture the long term relative effect of the drug. This resulted in an incremental life years of 620 and incremental QALYs of 560 and incremental cost of US$ 5,899,016. ICERS arrived at are US$9,511/life years and US$ 10,532/QALY. Table below shows are values of ICER at 0%, 3% and 8% discount rates respectively.
Table 1: Cost and effect of ‘HAART' and no ‘HAART' therapy regimen over a lifetime at 0%, 3% and 8% discount rates respectively
A one-way sensitivity analysis was performed on all modelled variables to assess the impact of uncertainty on the ICER by varying individual parameter by using the lower and upper limits of each parameter. Results presented in table 2 showed that cost-effectiveness ratio of HAART ranged between US$7, 027- US$ 14,038/QALY which is as a result of its sensitivity to the cost of HAART drugs. Base-case ICER was also calculated at 3% discount rate as suggested by WHO27 resulted in ICER of US$9,511/life years and US$10,532/QALY respectively. For sensitivity testing, as recommended by WHO, it was discounted at 6% discount rate applied for cost and 0% for health effect, the outcome of which resulted in significant reduction in ICER- US$ 6,248- US$7,312/QALY.
Table 2: A uni-variate sensitivity analysis of individual parameters in the model
ICER= Incremental cost effectiveness ratio *QALY= Quality adjusted life years *CI= Confidence interval
A multi-way analysis which combined the parameter values that gave the lower ICERS than the baseline to model (the best case scenario) and observed change in the baseline was conducted. Same was repeated that gave higher value than the baseline i.e. worst case scenario30. The product was an ICER range of US$5,418-US$16,893-(best-worst).
Cost-effectiveness probability curve, obtained from probabilistic sensitivity analysis shows a 64% probability of being cost effective from a willingness to pay threshold of US$11,640, (which was arrived at by multiplying SA Gross National Income by two) and 100% effective at US$20,000.
Figure 2: Cost effectiveness acceptability curve show the probability that HAART is both cost-effective at varying threshold cost-effectiveness values
Cost-effectiveness of stating at HAART treatment at different stage of disease compared to ‘no HAART' showed variation in the value for ICER. ICER comparing starting treatment at 50<CD4<200 compared to ‘no HAART' was US$17,670/QALY which is more costly than baseline value but also more cost-effective (98%), and consistent with findings of Badri et al8. This was done as the current standard for starting HIV treatment in SA followed the WHO guidelines for developing world i.e. ‘that a patient is eligible for treatment when his or her CD4-positive T-cells falls below 200/mm3 of plasma or when there is an AIDS-defining illness'.
Above is affordability curve showing how ICER decreases over time. For this study, it is only cost-effective if the cycle is run for at least 8 years with the choice of threshold adopted.
Study found out that treatment with HAART may help in reducing the burden of HIV/AIDS and help in attaining the targets for MDGs and 68% cost effective at the threshold. Findings also suggest that the program has the potential to be affordable even with a relatively low annual budget.
It also shows how CEAC provides a valuable heuristic for summarizing the distribution of the expected health and economic consequences in a setting of multivariate uncertainties, but doesn't differentiate between the joint distribution of cost and effects that share same correlations between the two dimensions, but differ in scale31-32. CEAC only addresses part of the limitation; however when the joint distribution for cost and consequences are positive, issues like budget constraint and cost-effectiveness threshold in analysis of multivariate uncertainty should burst the information available for guide in reality. Results of economic evaluations are used aid decision making for resource allocations to competing programs. To do that, a threshold has to be decided upon33. Thresholds in economic analysis are the maximum point at which a service is said to be cost-effective, and considers accommodating values below it as cost-effective which is vital for normative budgeting decisions and debates about equity34. Threshold varies across countries as there is no general harmony on what the threshold for CEA should be. E.g. in the UK, NICE has a suggested threshold range of £20,000- £30,000/QALY while the USA $50,00011, 27, 33, 35. In situations where there isn't any threshold decided upon, value twice the GDP or GNI/capita(both are normative values) has been suggested for developing countries36, while WHO commission on microeconomics for health suggested 1-3 times GDP globally37. A review by Shillcutt et al on cost-effectiveness in low- and middle income countries suggests using two times the GNI/capita of the country as a threshold for analysis34.Wolf et al also reported in38 ‘'that it is a common practice to define an intervention with an ICER less than 3 times per capita GDP as ‘'cost effective'' and an intervention with an ICER less than per capita GDP as ‘'very cost-effective''''39. With a GNI/capita of US$ 5,820 as at 20082, 40, twice this gives a threshold of US$ 11,640 or 1-3 as US$ 5,820- US$ 17,460. With a threshold of US$ 11,640QALY, this study is said to be cost-effective as its ICER falls within the WHO range of cost-effective interventions.
Several limitations exist in this analysis which includes; The modelling does not reflect the indirect cost such as adverse event/drug resistance over time, and that the estimated benefit may be higher. The analysis did not evaluate fully the cost-effectiveness of initiating HAART at different CD4 cell counts except starting at 50<CD4<200, which showed to be cost-effective but more costly, though similar to finding of Badri et al8.others include;
- Lack of utility values typical to the South Africa
- Lack of specific willingness to pay threshold (WTP)
- It is important to note that other studies in other countries quoted to be cost-effective in this study, and which may have implemented these policies successfully, have entirely different way of running their health care services.
- The discount rates used in the model were arrived at based on other studies choice for that and recommendation by WHO, which may vary (higher/lower) than what would have been experiences in reality.
- There was no information on the specific type of treatment used for the treatment of AIDS-related opportunistic infections, so we had no information on the effect of these drugs on the CD4 cell counts on the cohort.
- The model also assumes that all states are independent of each other, which may not be the case for this study.
- Memory-lessness nature of the model.
- Societal perspective would have been a better option for cot estimate as decisions if made with the result obtained will impact the whole society. This was not possible due to limitation of data.
Despite favourable cost-effectiveness of HAART to no HAART in SA, lack of tolerable long term commitment to preventive health programs, reduction in poverty rates and also unfailing external funding may lead to SA facing hitches in decisions over how to set priorities and allocate resources. In this circumstance, emphasis should be placed on the importance of considering issues of affordability and continuous cost-effectiveness of new technologies. Careful considerations' should be made about the trade-offs of competing choice which have their opportunity cost too. When making decisions, equity issues should be considered as regards who pays for the service; healthy, active and productive individuals who pay proportionately higher tax or should their be user-charges associated with access to HAART for those who need it?
Nonetheless, given the fact that this research if conducted in reality may require enormous amount of effort, and that HAART' for HIV might reasonably satisfy the assumptions necessary for the existence of the fixed budget based on HAARTS's favourable cost-effectiveness profile, an approach which considers cost-effectiveness, uncertainty, affordability can serve as a practical tool to provide valuable information to decision makers in low/middle income countries who might face resource-constraints for the provision of HAART or other similar interventions.