Cost-effectiveness and cost-benefit in the prevention of mother-to-child transmission of HIV in developing countries
Introduction HIV infection has important economic, societal and financial consequences, in terms of loss of productive labour, disruption of household units and burden of treatment costs on health-care systems [1]. Approximately 25–30% of infants born to HIV-infected mothers in developing countries are infected [2,3], and approaches to reducing the rate of mother-to-child (vertical) transmission (MCT) not only include prevention of new infections in women of childbearing age, but also specific therapeutic interventions targeted at pregnant women, modification of obstetric practices, and substitution of formula (bottle) feeding for breast-feeding where feasible and affordable [3]. To date, the only approach of proven effectiveness is the administration of zidovudine therapy during pregnancy, delivery and in the neonatal period in the absence of breastfeeding, as demonstrated in the United States and France [4], and very recently with a short regimen from 36 weeks of pregnancy and during delivery in Thailand [5]. Within the next year, the results of the ongoing antiretroviral trials for prevention of MCT of HIV infection in several African countries could confirm the benefit of zidovudine in breastfeeding populations. However, reduction of the risk of MCT is only one of many demands on the health–care systems, and before decisions are taken on implementation of vertical transmission therapeutic interventions, issues relating to the implementation into public health measures need to be addressed. In this article we review economic evaluation as an approach to assist in the development of policies regarding the prevention of MCT in developing countries, using data from studies evaluating antiretroviral therapy in sub-Saharan Africa and Thailand. Principles of economic evaluation It is generally agreed that economic evaluations of MCT interventions should conform to the widely agreed practices regarding economic evaluations in general, with transparency in analysis and presentation [6–9]. The following methodological elements are applicable to the case of MCT interventions. Costs are defined as the monetary value of resources used as a result of the intervention(s) under consideration. The intervention's outcome(s) might be measured in several ways, including clinical outcome (paediatric HIV infection), monetary benefits, and subjective preferences, for example quality of life, often combined with survival to produce expected quality-adjusted life-years (QALY) gained or productivity/disability-adjusted life-years (DALY). By convention, evaluations comparing monetary costs with monetary outcomes are termed ‘cost-benefit', whereas evaluations comparing monetary costs with commensurate non-monetary outcomes are termed ‘cost-effectiveness'. As a subset of cost-effectiveness evaluations, those using subjective preferences as outcomes are termed ‘cost-utility’ analyses. Economic evaluations typically involve a relative assessment (i.e., a comparison of consequences of two or more approaches, one of which is usually either the status quo or no intervention). This requires that the outcomes of all the interventions under consideration are measured in the same units. Costs and outcomes generally have a different impact on different individuals or groups, such as the patient, the patient's household, the health-care system or society as a whole, and the frame of reference must be clearly defined and presented. It is possible that an intervention that is cost-beneficial to the public health system is not so for certain individuals. If some parameters used in the evaluation are variable rather than constant, results must be subjected to sensitivity analysis, and be interpreted as ranges of values. Costs and outcomes occurring over time must be the subject of discounting to make time-dependent values commensurate [9]. The discount rate reflects social time preference, which under perfect markets equals the opportunity cost of capital employed. The discount rate is likely to be a function of economic affluence, with people in poorer countries placing a higher premium on present consumption than those in richer countries. By implication, the appropriate discount rate for poorer countries would be higher than that for richer countries. Debate persists over whether non-monetary outcomes should be discounted at all and, if they are, whether the appropriate rate is higher or lower than, or equal to, the rate for costs [10]. However, it is essential that a sensitivity analysis of the choice of discount rate is conducted for all evaluations, and that when several interventions are being compared, common assumptions regarding the discount rate are employed. Review of evidence Available evidence relates to two studies in sub-Saharan Africa [11,12] and two in Thailand [13,14], and refers to settings where trials are currently ongoing to evaluate specific antiretroviral therapy interventions to reduce vertical transmission. The main methods, parameters and results for the economic models in these studies are summarized in Tables 1 and 2. The potential cost-effectiveness of a short-course (4 weeks at 600 mg per day prepartum) of zidovudine therapy for a hypothetical 1-year birth cohort in a developing country was estimated using a decision model [11], with assumptions made regarding the prevalence of HIV infection, the extent of breastfeeding, HIV transmission rates, costs of zidovudine therapy, testing and counselling. For the modelled cohort of 100 000 births with base-case parameters, the incidence of infant HIV infection would be reduced by 12%. The cost of the antiretroviral programme would be about US$ 20 per birth, of which 24% represents drug costs. Health-care costs of infected infants would fall by a small amount. Sensitivity analysis indicated that the programme would be cost-saving from a societal perspective if antenatal HIV prevalence was above 18%, if the cost of testing/counselling was below US$ 12 or if medical care costs for infected infants exceeded US$ 4145. However, the estimated programme costs of around US$ 20 per capita would represent a major proportion of total health-care costs in many developing countries. It was concluded [11] that it is unlikely that, in developing countries, a zidovudine intervention programme would create direct savings from the health-care system perspective, but it may prove cost-effective from a societal perspective.Table 1: Summary of methods, parameters and results of the economic models.Table 1: (continued)Table 2: Summary of clinical and efficacy parameters applied in the economic models*.The second study relevant to the African context considered three short-term regimens of combination anti–retroviral therapy (zidovudine and lamivudine), as used in the UNAIDS-sponsored trial in Africa [12], and compared these three regimens with no intervention. In regimen A, combination therapy was started at 36 weeks of pregnancy and continued until 1 week postpartum, regimen B omitted the prepartum therapy, whereas regimen C omitted both pre- and postpartum therapy. The frame of reference was the public sector medical system and the study thus considered the costs of testing and counselling and antiretroviral treatment costs in relation to expected savings in infant health-care costs. Estimates are also provided of marginal net cost per expected life-year gained, weighted for both productivity and quality of life. Basic clinical parameters were derived from the published and unpublished evidence and allowed for imperfect adherence to the regimens. The shortest and most cost-effective option was regimen C. For a cohort of 100 women with 15% HIV prevalence, net costs were estimated at US$ 3617 for regimen A, US$ 1667 for regimen B, and US$ 351 for regimen C. Regimen C had a cost of US$ 1129 per HIV infection averted and a cost of US$ 60 per DALY. Sensitivity analysis revealed that regimen C was the most sensitive to testing costs and particularly sensitive to assumptions about maternal prevalence. Cost-effectiveness declined rapidly when prevalence was below 7% or when efficacy fell below 10%. When estimated drug prices fell to low levels (e.g., < US$ 0.5 per pill), the cost-effectiveness of the more drug-intensive regimens (A and B) improved. The base-case estimate of around US$ 60 per DALY gained from the most cost-effective combination antiretroviral regimen at current drug prices compares unfavourably with other public health initiatives in most settings within the developing world, for example, immunization and treatment of other sexually transmitted diseases. However, at 20% of current drug costs, both arms B and C attain cost-effectiveness values of under US$ 50 per DALY, and a MCT control programme based on short-term anti-retroviral drug regimens might then be a sensible option, particularly in areas where more cost-effective public health interventions have already been widely adopted. The first Thai study [13], conducted before the 1997 Asian economic crisis, considered the selection of an optimal intervention strategy, given a prior government commitment to devote resources to the prevention and cure of HIV/AIDS. It argued that affordability must be a key consideration in the selection process, in that policies will only be sustainable if either governments or households (or both) are in a position to provide the programme's resource base. The study modelled 10 possible strategies, eight of which involved antiretrovirals. Two concerned themselves with MCT (formula feeding alone and zidovudine plus formula feeding), whilst eight targeted infected adults either before or after the onset of AIDS. The unit costs of treatment for adults with AIDS range from about US$ 2000 per patient-year with zidovudine monotherapy to US$ 4000 per patient-year with zidovudine plus didanosine or zalcitabine combination therapy. The unit cost of treating pregnant women following the AIDS Clinical Trials Group (ACTG) 076 [4] protocol was about US$ 500 per mother–child pair. Assuming a pricing policy that fully subsidizes antiretrovirals whilst allowing for one-third recovery for other medical services, all antiretroviral options were shown to be unaffordable within the budget of the Thai National AIDS Programme, with the exception of the two MCT programmes. The modelled subsidized pricing policy would be affordable by around 50% of the population. Simulations suggested that the effectiveness (expected QALY gain) of preventing MCT is of a similar order to that of treating adults, but their cost-effectiveness ratios are about 20 times superior than antiretroviral programmes for adults. A further Thai study [14] explored the expected cost implications assuming that the full ACTG 076 protocol [4] was to be implemented in Thailand. It concluded that implementation of the ACTG 076 protocol in Thailand would not only reduce the MCT rate but would be cost-saving overall. This finding was insensitive to the choice of discount rate (within plausible bounds). The programme would be affordable within the government's AIDS budget, again before the 1997 economic crisis, even without cost recovery from private individuals. The study by Wilkinson and colleagues in this issue [15] provides additional information regarding the affordability of zidovudine therapy to reduce vertical transmission (Tables 1 and 2). In a modelling exercise, they evaluated both the long zidovudine regimen used in the ACTG 076 trial [4] and a combination regimen used in the UNAIDS trial [12] in the context of a rural population in KwaZulu Natal, South Africa, and considered issues relating to the implementation of these antiretroviral therapy interventions in a general rural population. Cost data were combined with estimates of effectiveness to calculation the cost per infection prevented; the cost per potential year of life saved was calculated using a life expectancy of 63 years. They concluded by saying that providing short-course combination therapy of zidovudine and lamivudine may be cost-effective, but questioned whether it would be a realistic public health intervention. The cost of this combination therapy is relatively high despite its short antepartum duration due to the need for two-drug therapy for both mother and child, and the recently demonstrated efficacy of short-course zidovudine monotherapy [5], with a reduction in the cost of zidovudine [12], would make this a more realistic option. Sensitivity of estimates to parameter variation To understand the implications of parameter variation for the MCT evaluation results, a simplified model of antiretroviral therapy in developing countries was constructed, which included the essential features of the studies reviewed above, and two carried out in the United States [16,17]. This was part of a consensus meeting held in June 1997, and full details of the model can be found in the report of the workshop [18]. Intervention with antiretroviral therapy necessitates screening of pregnant women, counselling and other additional antenatal care and the administration of the drug regimen. Although test sensitivity and specificity were assumed to be 100%, treatment compliance and adherence rate to the drug regimen were not. This means that a proportion of pregnant women will not be tested, or not treated adequately, and their infants will thus continue to be at high risk of acquiring HIV infection. The modelling exercise showed that the values of the parameters associated with decreased cost–benefit were high costs for screening, counselling, antenatal care and drugs, low efficacy, low costs for paediatric treatment and productivity losses, low prevalence of HIV infection in the antenatal population, and low compliance rates [18]. Movements of one or more of these parameter values in the opposite direction would make the intervention more cost–beneficial. The model also allows evaluation from the health-care system perspective, and, assuming positive efficacy and positive productivity losses, the net gain to society as a result of the intervention will always exceed that of the health-care system for any given set of parameter values. The cost–benefit model can be reinterpreted in a cost-effectiveness framework, and the relevant parameter movements noted above to decrease net benefits in the cost–benefit approach also make the intervention appear less cost-effective. In addition, the parameter that would make the intervention less rather than more cost-effective is a lower differential in discounted outcome between infected and non-infected infants. However, given the possibility of compensating effects, the reliability of parameter estimates used in these cost–benefit and cost-effectiveness exercises needs to be considered. Reliability of parameter estimates Model building to evaluate an as yet unimplemented intervention generally entails both measurement (when a relevant activity is currently in place) and assumption (when an activity is not yet in place), where the former are considered to be more reliable than the latter. A randomized controlled trial provides more convincing evidence and more precise estimates of the effect of an intervention than any other study method [19], and replication of the first trial improves confidence in reliability of the estimates [4,5]. However, although randomized controlled trials may offer the best prospect of yielding reliable estimates of clinical effect, they may be less reliable as broader indicators of how an intervention would perform in everyday practice. This is especially true for economic considerations. For example, trial participants are volunteers and clinical efficacy in terms of the number of subjects enrolled gives no indication about likely compliance rates in a general population. Furthermore, clinical trials are generally conducted within a supportive infrastructure. Resource usage in such a setting might not be typical of resource usage when the intervention is introduced across a health-care system more generally. The reported drug cost varied across studies [11,12,14,17], which reflects either the world price or the purchase price negotiated at the time of analysis. There was a more than fourfold variation in counselling and testing costs in the two African studies using evidence from Zambia [12] and Uganda [11]. This difference partly reflects the costs of labour and capital involved in the procedures. The estimate for United States costs is around US$ 100 [16]. Similarly, base estimates of mean discounted lifetime paediatric HIV treatment costs range from US$ 118 [12] and US$ 396 [11] for African settings, through approximately US$ 5000 for Thailand [14], to US$ 98 915 [17] and US$ 161 137 [16] for the United States. The conventional method for estimating the value of productivity losses is to assume that premature mortality has prevented the individual concerned from generating mean per capita gross domestic product (GDP), or some other macroeconomic aggregate, such as national income. In developing countries generally, GDP figures are considered to be less reliable, owing to measurement error and the large volume of activity generated by non-market production. Using the GDP figures specifically for the present purpose makes the strong assumption that production resulting from the avoidance of HIV infection can be represented as a mean contribution to output. This would clearly be misleading if infants concerned were born in, for example, regions where subsistence agriculture predominates or in urban areas of high unemployment. Generally, when ranking these costs on a per intervention basis, treatment costs savings and productivity losses avoided greatly exceed unit testing, counselling and drug costs. The margin between these two cost classes increases as the level of development of the country concerned increases. This leads to the general conclusion that the likelihood of an antiretroviral intervention to be shown to be cost–beneficial increases with the level of development of the country, considering only the cost variable and assuming that other factors remain unchanged. Despite the variety of settings for which MCT intervention models have been constructed, their assumed effectiveness parameters are based on the results of a single trial [4] conducted in the United States and France on mildly symptomatic women with no prior antiretroviral treatment. Although little information is available on adherence to the drug regimen, our descriptive model demonstrated the sensitivity of cost-effectiveness and cost–benefit results to compliance effects. Lower compliance reduces the intervention's effectiveness of MCT risk reduction. A study of zidovudine use among women in the United States suggests that diverse attitudinal and perceptual barriers to the use of zidovudine amongst HIV-infected women exist and that these may be strongest among minority women, particularly African Americans [20]. This may also be the case in many African countries, where fear of stigmatization is high. Completeness of the models Models are designed to answer clearly specified questions, and are constructed in order to make a complex process comprehensible. However, the assumptions used in the construction of these models should not be overlooked, especially when we consider using the results of a model as a criterion for effecting policy changes. Below we note a number of issues, explicitly absent from the economic models thus far considered, which a policymaker might consider relevant. Reducing the risk of MCT is likely to increase the number of orphans in society. Although in many countries the extended family system has been able to accommodate these children, the impact on the household economy is unknown. As HIV/AIDS deaths increase over time, more reliance may have to be placed on the government. If this factor was to be included in the evaluation models, the costs of care of orphans would offset to some degree the treatment cost saved by less HIV infection amongst infants. In addition, it is possible that the employment circumstances of orphans in any particular country would differ from that of the average individual; it is believed, for example, that orphans and children in one-parent families typically receive less education than children in two-parent families [1]. A relative lack of education would make such individuals less competitive in the labour market, other things remaining equal. Assuming we accept this proposition, then the estimates of productivity losses avoided by preventing the birth of an HIV-infected child, which are based on mean contribution to GDP, are likely to be overestimates. Assuming that antiretroviral therapy is adopted and proves acceptable to women, we could anticipate a change in behaviour with respect to decisions about reproduction that could impact on overall population growth. Reducing MCT transmission increases the probability of producing an uninfected child. This may result in either higher levels of reproduction (if families value higher numbers of healthy children) or lower levels (if families aim to achieve a given family size). Economic intuition supports either possibility and thus the reaction cannot be predicted reliably without further evidence. In the models discussed thus far, the health-care system frame of reference has not been disaggregated into public and private contributions. In policy terms, such a disaggregation may be significant, in view of the importance currently attached to user charges and cost recovery in the health-care systems of developing countries. Although there is a strong case in favour of cost recovery in health care generally, there are also important counter-arguments in specific cases, such as the administrative costs of operating charges, equity considerations and the likely disincentive effects to service use generated for households subject to charges [21]. It is quite possible, for example, for cost recovery to transfer a sufficient part of the public funding burden to households such that the practical effectiveness of the new intervention is nullified by poor uptake. In the MCT case, we do not yet possess information regarding household's price and income elasticity of demand with respect to risk-reduction services. Comparability and affordability Economic evaluation as a decision-making tool facilitates the ranking of alternatives: for example, intervention A is cost-saving relative to the status quo, or intervention A is more cost-effective than intervention B. With respect to MCT, evaluations in developing countries have generally concerned antiretroviral treatments, although other risk-reduction options are recognized [3,22]. To date, however, of these of MCT reduction have been proven to be of cost-effectiveness within MCT MCT policies are part of a more general public health programme at the of women and children and their cost-effectiveness should be compared with other public health programmes might include and and to the around of all the health of the under in developing countries can be by interventions less than US$ 100 per DALY saved The one study that has cost-effectiveness estimates [12] suggested that the cost-effectiveness of antiretroviral therapy could be below this but the reliability of this estimate for antiretroviral therapy would be by the of further evidence. However, there may be other paediatric interventions, such as prevention and treatment of which offer value for than antiretroviral therapy, and their could be more cost-effective than the of maternal antiretroviral therapy to be considered before implementation of interventions to reduce vertical evaluations of MCT interventions, either in randomized controlled trials or in should include economic evaluation to whether the if found to be can or should be implemented as a public health economic evaluations should be and include a sensitivity analysis of the most important to effectiveness and costs. However, in addition, the issue of affordability should also be and outcome measures and of both to comparison with other interventions and to potential between the of of decision should be employed. information regarding the cost-effectiveness of antiretroviral therapy in reducing vertical transmission a degree of particularly until the Thai results available the efficacy of short-course treatment was [5]. is and policy decisions regarding anti-retroviral interventions should be based on of the results of trials currently in In the cost-effectiveness of antiretroviral therapy in countries is more sensitive to the drug price than it is in or However, within the range of antiretroviral interventions for HIV treatment in general, it is likely that MCT prevention will prove both more cost-effective and affordable than and will part of the HIV in many developing countries. relating to affordability and remain to be addressed. The for the and the workshop for their general and their on the
Read more
3