INDIA HAS MORE PEOPLE INFECTED with HIV than any other country. Of the roughly 4 million people with HIV, approximately 550,000 have AIDS and another 300,000 a year will develop AIDS over the next 15 to 20 years. Although several states have made progress in improving prevention, much more is needed to limit the impact of HIV/AIDS in the future. If India's prevention programs fail to halt the expansion of the AIDS epidemic, these numbers could double or triple over this period. This article is the product of a 2002–2003 World Bank-sponsored consultation to the government of India. Its purpose is to use the best available data and epidemiologic models to assess the costs, cost-effectiveness, and epidemic impact a range of HIV treatment policy options.
Antiretroviral (ARV) drugs were initially available only in rich countries at an annual cost of more than $20,000 per person.1 Once Indian pharmaceutical firms began to manufacture generic versions of advanced therapies and selling them for approximately $240 per person per year, national and international interest groups lobbied the government to increase access to antiretroviral therapy (ART). Proposals ranged from simply encouraging the use of ART by patients of private physicians to providing free ART to all people infected with HIV.2 Even at today's relatively low prices, only a minority of Indians are able to finance ART with their own families' money.3 Increasing access therefore requires a substantial increase in government health expenditures. The availability of donor resources relieves somewhat, but does not eliminate, the government's resource constraint. The government committed to expanding publicly financed access to free AIDS treatment in 2004. In June 2006, free treatment is available for approximately 39,000 patients at 71 national, state, and NGO health facilities in more than 10 states with plans to expand access to more patients and treatment sites. In addition, an unknown number of patients pay out of their pockets for ART obtained from the private sector.4
The objective of this study is to assess the costs, cost-effectiveness, and HIV epidemic impact of 3 ART policy options from the perspective of the Indian government in 2004. Because the actual policy of the Indian government in 2006 is a mix of the policy options analyzed, the results of the analysis will cast light on the consequences for India of emphasizing one or another dimension of current policy as they expand treatment in the future.
This article examines the full range of costs and consequences accompanying 3 policy choices regarding government funding of ART. It is intended to help policymakers make decisions by:
* Reviewing the effects and consequences of ART
* Using an epidemiologic model to predict the course of the epidemic, including its consequences and costs, under current government policies; and
* Using the same model to determine the consequences and costs of 3 alternative policies and comparing their effects with the consequences and costs of maintaining current policies.
Possible Consequences of Antiretroviral Therapy
After the launch of generic ARV drugs by Indian pharmaceutical companies in 2000 and the decline in the costs of these drugs, an increasing number of people with HIV/AIDS have been using ART. Of the estimated 550,000 people with AIDS in India in 2004, 370,000 resided in 60 major cities. In 2004, physicians in these cities were treating 90,000 of these people, 11,700 of whom (8,700 males and 3,000 females) were receiving ART.5 However, little of the ART is “structured,” that is, in conformity with the guidelines of the World Health Organization and the National AIDS Control Organisation of India (NACO).
Policymakers need to consider both the direct and indirect effects (i.e., “spillover effects” or “externalities”) of ART programs. Direct effects improve well-being for the patients receiving the therapy; whereas externalities affect others, for example, by affecting HIV transmission.
Of the 5 possible spillover effects on transmission, 2 are positive: ART may reduce infectiousness and may encourage prevention. There are 3 possible negative effects. Use of ART may promote resistance to the drugs may increase the duration of infectivity and may encourage HIV positives and negatives to engage in more risky sexual behaviors. The consequences of these externalities on a population are not well understood. We therefore make a range of assumptions about their effects in portraying the consequences of ART policies.
Antiretroviral Therapy May Reduce HIV Transmission
ART may reduce transmission by reducing the concentration of the virus in body fluids and genital secretions. ARV drugs definitely reduce the amount of virus in the blood.6 Based on the observed reduction in viral load in blood plasma, experts estimate that ART reduces infectiousness by a factor of 2 to 8.7–13 Because the effect of ART on infectiousness is unclear, this analysis uses the optimistic assumptions that structured therapy eliminates transmission for 5 years and unstructured therapy reduces it by 50% for up to 3 years.
Antiretroviral Therapy May Strengthen Prevention
Increased availability of ART could strengthen prevention by motivating people to come forward for voluntary counseling and testing, resulting in more prevention counseling, less high-risk behavior, and lower transmission. The evidence in support of such an effect is anecdotal, however, and the effect is likely to be context-specific. We include such a positive synergy by modeling the result of simultaneous increases in publicly financed treatment and condom use.
Antiretroviral Therapy Increases Drug Resistance
Users of ART sometimes react adversely to the drugs either when they start them or after months of use. These adverse reactions as well as psychosocial problems impede adherence to the prescribed regimen.14 In the absence of strict adherence to a 3-drug regimen, most people develop drug-resistant strains of the virus, which can then be spread. In Mumbai, 18% of people newly diagnosed with HIV were resistant to at least one ARV drug. This figure is higher than the 14% of new infections estimated to be resistant in North America in 2001.15 These high rates of resistance among the newly infected are likely to increase as resistant strains accumulate in the population. In our model, those for whom treatment has failed will transmit a strain of the virus, which is not susceptible to first-line therapies, and we do not consider second-line therapies in this study.
Antiretroviral Therapy Lengthens the Period of Infectiousness
High levels of adherence to recommended ART drug regimen are expected to result in 4 additional years of life and in better health during those years, including a return to usual levels of sexual activity. Increased longevity can thus increase the chance that an individual can pass on his or her infection and this effect is captured in our model.
Antiretroviral Therapy May Increase Risky Sexual Behavior
The availability of ART may induce people to engage in riskier sexual behavior. This phenomenon, sometimes called “disinhibition,” potentially affects both those being treated and the general population. North American and European studies of men who have sex with men indicate a sharp increase in the proportion of people having unprotected sex since ART became available.16–19 Evidence from Kenya on heterosexual behavior suggests that substantial disinhibition occurs if the availability of ART is announced without strengthening prevention measures.20 This effect like the possibility of beneficial synergy between ART and prevention mentioned is modeled and discussed subsequently.
The Configuration of Alternative Antiretroviral Therapy Programs
Three attributes of ART policy design, population coverage, the relative mix of public and private services, and the extent of HIV transmission-minimizing features largely determine its costs and consequences.
Population coverage refers to the number of eligible people receiving any form of ART (ranging from self-medication to model clinical practice) as a percentage of the total number of people eligible for such therapy. Coverage can be increased by expanding availability to all infected peopled or only to specific groups such as pregnant women or poor people. According to World Health Organization guidelines, people are eligible for ART if they have AIDS or are HIV-positive and have significant AIDS-related illnesses or a CD4 count of less than 200 mm2. On average, this period of eligibility includes approximately 5 of the 10 years between initial HIV infection and death. Thus, as many as half of people who are HIV-positive are eligible for ART (the proportion is lower early in the epidemic, higher later in the epidemic). This report conservatively assumes that 1.9 million or 50% of the estimated 4 million Indians who are HIV-positive (including people with AIDS) are eligible for ART. Only 12,000 are currently on ART of any kind, indicating coverage of much less than 1%.
The second attribute of ART provision is the degree to which it is financed and provided by the public sector. The Indian government subsidizes an estimated 20% of total healthcare costs, but the percentage of ART costs funded by the public sector is much smaller. Given that people with HIV/AIDS are often refused treatment at public hospitals, public spending probably represents an even smaller percent of total spending on ART.
Although the way in which ART is delivered varies from country to country, the essential elements of model clinical practice, referred to in this report as “structured treatment,” consists of the following features:
* Standardized, competency-based training of physicians in ART management
* Prescription of a standard triple-drug regimen
* Support from a multidisciplinary team, including a counselor and a nutritionist
* Regular clinical and laboratory-based monitoring of the patient's treatment status
* Counseling to prevent transmission
* Prophylaxis for opportunistic illnesses when indicated; and
* Diagnosis and treatment of opportunistic illnesses.
The third attribute of ARV programs is the extent to which they are able to reduce transmission by capitalizing on the preventive potential of ARVs. Because most ART provided in India before 2004 was “unstructured,” that is, did not follow these guidelines, structured therapy reached few people and was unconnected to prevention programs. As typically practiced, structured treatment also does little to address the risk behavior of either the patients being treated or of both the HIV-infected and -uninfected people in the vicinity who are not under treatment. Because an increase in the risk behavior of those not under treatment could easily cause new HIV transmission that offsets the beneficial effects of ART, it is imperative that delivery of ARV programs be designed to prevent increased risk behavior in both groups.
We define transmission-minimizing ART as consisting of 3 elements: 1) structured ART, 2) incentives to state and local government and community leaders to strengthen prevention programs, and 3) rigorous, impartial monitoring and evaluation of both treatment and prevention program effectiveness and of the extent to which they contribute to one another. In a well-functioning transmission-minimizing ART program, these 3 elements work together to decrease transmission as well as improve treatment.
Willingness to Pay for Antiretroviral Therapy
Because the ART policy scenarios analyzed all involve individual payments for some or all ART patients, a study was commissioned to look at the treatment-seeking behavior and willingness to pay (WTP) for ART in India.3 A survey was conducted among 269 HIV-positive individuals in 4 cities based on a purposive sample of individuals known to have been diagnosed with HIV infection and to be receiving at least some treatment for opportunistic illnesses, if not ART. Such a sample does not permit us to describe the average HIV-infected person in India. However, it tells us how some HIV-infected people in India are responding to their infection. Furthermore, the study can tell us the responses of an average HIV-infected person with certain characteristics such as what an HIV-infected person with a certain level of assets would pay for ART.
The WTP survey revealed a very high willingness to be on ART; approximately 94% of all individuals not currently on ART wanted to be on ART. However, the willingness to take treatment did not necessarily translate into WTP. Of those who are willing to be on ART, 90% were willing to pay for drugs, 74% for initial tests, and 83% for follow-up tests. Based on the patients' maximum WTP, the authors generated demand curves for each of the 3 components of ART. In the 2 panels of Figure 1, we present the predicted percent willing to pay for drugs on the horizontal axis as a function of the monthly cost of drugs on the vertical axis. The 2 curves are derived from 2 alternative questions eliciting this information from the same respondents. Both demand curves indicate that ART is a “normal good” for which demand increases with assets and price is a significant deterrent to purchase. Results for tests were qualitatively similar but WTP was significantly smaller.
The policy implications of this analysis are straightforward; to get a greater number of individuals on ART would mean finding a mechanism for lowering the price to the patient of all the 3 components of ART, but especially the initial tests. These tests form the entry point into ART and can also be linked to voluntary testing and counseling because subsidizing initial tests could be one way of stepping up voluntary testing and counseling in any system that offers ART.
The Epidemiologic Projection Model
To estimate the health consequences of alternative government policies, we modified an epidemiologic model previously used to project the course of an Indian epidemic. The model draws on the international AIDS literature to set biologic parameters relating to transmission and disease progression. When there is uncertainty about a biologic effect of ART such as whether structured ART eliminates transmission, we choose assumptions that favor ART.
The model assumes that the probability of transmission of HIV from a male to a female on a single sexual contact is 0.0052, whereas the probability from female to male is 0.0036.21 These transmission probabilities apply to people on average over the duration of their infection provided they are not treated. However, we assume that infectivity is lower by 24% during the 5-year period without symptoms and, in the absence of treatment, higher during the symptomatic period by 30%. We assume that unstructured ART treatment reduces infectivity by 50%, whereas structured ART eliminates transmission altogether. However, after treatment failure, these effects are vitiated and infection can occur. Individuals for whom treatment has failed as well as those whose primary infection is with a resistant virus will transmit resistant viruses.
Figure 2 presents the aspects of the model that account for transmission of the epidemic. The population is divided into 4 groups according to their sexual behavior, low-risk men, low-risk women, male clients of sex workers, and female sex workers. Each of these groups is further divided into a small subgroup of people who are infected with HIV and the majority who are susceptible to infection, making a total of 8 groups.
The model assumes that in 1998, at the start of the simulation run, 15% of the adult male population are clients of sex workers with each of these clients purchasing 50 sexual contacts annually. The model further assumes that 1.1% of the adult female population is sex workers so that the average sex worker has approximately 675 commercial sex transactions per year.1 Condoms are used in half of these transactions and are assumed to reduce the probability of infection to zero in those cases.
Few data are available on the impact of ART on life expectancy in developing countries. Based on data from Western countries, we assume that the “triple-drug” therapy most widely and inexpensively available in India increases the time from the onset of AIDS-related complex to the onset of AIDS from 4 to 8 years; the time between the onset of AIDS and death remains 1 year. The use of ARV drugs in accordance with strict guidelines thus increases the average time from infection to death from 10 to 14 years.
Expanded use of ART could reduce the number of AIDS orphans, for example. Children who lose their mother at age 1 will experience 14 years of under-15 orphanhood, a welfare loss that can never be reversed. Assuming that the average orphaned child is 7 years old, government policy that lengthens survival or slows HIV infection rates in the current year will have most of its effects on orphanhood 7 to 15 years later. These important long-term consequences are also included in the model.
The baseline scenario uses the assumptions of the epidemiologic model and additional assumptions about how rapidly the Indian population will adopt ART without any change in policy. It supposes that under current policies, unstructured ART will:
* Grow steadily until half of symptomatic Indians are using ART by 2012 and 80% are doing so by 2018
* Attain lower rates of patient compliance than structured care, therefore yielding lower health benefits for patients
* Facilitate the spread of resistant viral strains; and
* Decline in quality as it becomes more widely used; when adoption reaches 80%, most of the health gains will have been eroded.
Given the very low rates of health insurance in India, the model assumes that in the absence of government intervention, the growth of structured ART would have been slow (2% a year) restricted by the limited capacity to raise funds from private sources and the need to reallocate and train manpower.
Alternative Policy Scenarios
The epidemiologic consequences of 3 policy options, “ADHERE,” “Mother-to-Child-Transmission-plus” (MTCT+), and below the poverty line (BPL), are evaluated and each is compared with the same baseline model.
This policy aims to maintain the quality of unstructured care projected under the baseline scenario. ADHERE would provide government financing to train private physicians and laboratories in ART and diagnostic testing and subsidize laboratory tests. The outcome of this policy would be the net result of 2 offsetting effects on transmission: the decrease in the proportion of patients who transmit resistant strains of HIV and the increased opportunity of each patient to infect others.
An estimated 54,000 HIV-infected children are born in India each year. MTCT+ would provide government-financed ART for mothers and their husbands (the “+” portion of the program) who meet the criteria for beginning ART. Under this policy, the government would introduce a program to test all women who present themselves at government antenatal clinics and implement a nevirapine-based preventive strategy to reduce MTCT for seropositive women. Few women testing positive will yet have AIDS or be eligible for ART. The model assumes that:
* The public sector screens half of all pregnant women
* The average HIV-positive woman is not eligible for ART until 2 years 6 months after screening
* Women already in unstructured therapy begin structured ART immediately
* Women with resistant viruses are not recruited into the program; and
* The program recruits at least 25% of HIV-positive women and approximately half of their husbands.
Below the Poverty Line.
The BPL option assumes that a mechanism is in place for identifying the poorest people with symptomatic HIV infection and financing their access to ART. This policy would provide subsidized access to structured care for 40% of people with HIV. By 2033, the model projects that 7 million people would be receiving ART of whom 4.7 million would be receiving structured care financed by the government and 2.3 million would be receiving unstructured care. The effect of the policy is first assessed assuming that it induces no behavioral changes. That assumption is then relaxed to assess the effects if the policy alters condom use.
Evaluating the Impact of Antiretroviral Policy Options on the Epidemic
Figure 3 breaks out the health impacts of all 3 policies into the indirect and direct components. Contrasting the 2 panels shows the effect of discounting on the relative contributions of the 2 components, assuming a rate of 10%. Each group of 3 bars pertains to a single ART policy. Within each group, the first bar shows the impact of the policy on HIV incidence, whereas the second shows its impact on life-years. The third bar of each triple is the sum of the first 2 and measures the net impact of the policy on the selected health measure. Figure 3A presents the undiscounted results, whereas Figure 3B presents results after future life-years are discounted.
Both Figures 3A and 3B show that the benefits of the ADHERE policy accrue primarily to those receiving subsidized tests whose lives are lengthened. The indirect effects of the policy are on balance negative and serve to offset the positive direct effects in computing the net impact on healthy life-years saved.2 By contrast, for both the MTCT+ and the BPL policies, most of the benefit is the result of indirect rather than direct effects. Most of the potential health benefit of ART is the result of the reduced infectivity of those who receive high-quality structured care.
Figure 4 displays the net impacts of the ADHERE, MTCT+, and BPL policies on the 4 measures of health benefit and the 2 measures of orphanhood benefit, each of which is expressed as a percentage of the burden of the epidemic in the baseline scenario. This figure demonstrates that publicly subsidized ART has the largest impact on the epidemic when it directly supports structured treatment for the largest number of persons as in the BPL program. Depending on the measure used, the BPL program could reduce the epidemic's burden by between 6% and 14%.
Figure 4A shows that the MTCT+ program achieves a modest 1% to 2% improvement by all 4 measures. A comparison of its impact to that of the ADHERE program depends on the benefit index chosen. MTCT+ has a higher impact than the ADHERE policy about the first 3 of the indices. However, as was evident in Figure 3B, when saved life-years are discounted at 10%, the ADHERE policy's immediate health benefit to all the people receiving unstructured care dominates its long-term pernicious effect on the number of HIV infections and enables it to achieve 5 times the health gains of the MTCT+ program.
Figure 4B shows that, by both the undiscounted and the discounted calculations, all 3 programs reduce the orphanhood burden of AIDS by larger relative amounts than they improve either measure of health. The ADHERE program does particularly well in this regard, achieving almost the relative improvement of the more ambitious and expensive BPL program. In contrast, the MTCT+ program has a smaller beneficial effect on orphanhood. Despite being targeted at mothers, the MTCT+ program does not provide treatment to enough people to compete with either of the other 2 programs about the total reduction in orphanhood burden.
Evaluating Costs and Cost-Effectiveness
In the absence of detailed information on program costs, we use a small set of basic cost assumptions to estimate the costs and cost-effectiveness of alternative ART policies: ADHERE policy would cost $100 for every patient-year of structured or unstructured treatment and the MTCT+ and BPL programs would cost $500 per patient-year for patients in structured therapy. These figures include the cost of drugs, clinic visits, and laboratory tests that enable physicians to monitor treatment success and patient adherence. All costs are in 2002 U.S. dollars and are net of the offsetting reduction in costs as a result of the postponed costs of treating opportunistic illnesses for the 20% of people with HIV/AIDS who would have used public sector facilities. Although based on rough ex ante calculations of the cost of ART in India using the new drug regimens, these assumptions are consistent with a recent detailed study of the cost of India's publicly financed ART program in 7 treatment centers.4
Under these assumptions, government financing of ART would increase the net discounted cost of future health expenditures through the year 2033 by $598 million for the MTCT+ program, $1,682 for ADHERE, and $7,057 million for the BPL program, all evaluated at a 10% discount rate (see Fig. 5).
Figure 6 presents the cost per life-year saved for the 3 ART policies analyzed in the absence of behavioral change and for the BPL policy under the alternative assumption that the availability of ART financing can be used to minimize transmission. For each of these 5 options, cost-effectiveness is evaluated not only at the 10% discount rate, a rate at which governments frequently discount public investments, but also at the 3% rate, which is sometimes recommended for social investments like those in health. At the 10% rate, the first 3 policy options give costs per life-year saved in the range from $145 to $280. These figures are small in view of historical estimates 8 to 20 times larger. These numbers are also much smaller than the $600 per year cost we are assuming the government will spend for each patient-year of structured treatment. Two features of the Indian environment, as captured in this modeling exercise, contribute to the relatively favorable cost-effectiveness of these subsidized ART programs: The medications cost much less than they did even 5 years ago, and we assume that ART reduces the infectivity of treated persons. These transmission-reducing effects of ART, which operate independently of any behavioral change, account for most of the life-years saved of the structured ART policies that we call MTCT+ and BPL. These costs per life-year are still an order of magnitude higher than the estimated cost of saving life-years through HIV prevention or through tuberculosis or malaria control. Studies in Africa estimate the cost per disability life-year saved at $1 for condom distribution and $19 for prevention of mother-to-child transmission.22
When costs and benefits are compared at the lower 3% discount rate, the ADHERE policy changes its rank from the most to the least cost-effective of the 3 options that ignore behavioral change. Using the lower discount rate allows the long-run disadvantages of the ADHERE policies to offset the short-run advantages yielding the alternative ranking, which ranks the MTCT+ policy as the most cost-effective at $115 per life-year saved.3
If ART affects risk behavior, the results are very different. If ART has a sufficiently disinhibiting effect on condom use, the negative effects of riskier behavior could easily outweigh the positive effects of any treatment program. This disinhibition possibility is illustrated the by 40% condom use column in Figure 7. On the other hand, if the Indian government uses ART to enhance prevention, it may be able to design a “transmission-minimizing ART policy” such as the one outlined. In that case, it will be justifiable to attribute the success of the prevention program to the ART program, thus dramatically increasing the benefits of ART.
The fourth and fifth bars in Figure 6, which represent the transmission-minimizing ART policy, present a different picture. The costs per life-year saved of $51 or $30 are higher than, but of the same order of magnitude, as estimates for the cost of saving life-years through HIV prevention or other public health interventions. Unlike HIV prevention programs, which have been shown to be cost-effective in several settings, and disinhibition, which has been observed in several parts of the world, transmission-minimizing ART has not yet been demonstrated. Where ART programs have attempted to influence behavior, they have focused on the behavior of those under treatment, not on the behavior of untreated high-risk groups who would be subject to disinhibition and are crucial to HIV control. Thus, the low cost per life-year saved of ART policy displayed in these last 2 bars is less a result of this analysis than it is a proposal or a hypothesis.
If the government of India (GOI) proposes to implement the BPL option, it should do so with incentives to motivate state government and medical decision-makers to dramatically increase condom us. We suggest that India could design a mechanism that would increase condom use during high-risk contacts from the current rate of 50% to 70%. Assuming that the change in behavior could not have been achieved without the incentives provided in the ART program and that the costs of this improved prevention effort had already been allocated, all of the epidemiologic benefits of the change and none of its costs could be attributed to the transmission-minimizing ART policy. Under these very favorable assumptions, the cost per life-year saved of the ART program falls to just $51 at a 10% discount rate (Fig. 6). If the rate of condom use rose to 90%, the cost per life-year saved by the BPL policy would fall to approximately $30 at the 10% discount rate. A lower discount rate of 3% reduces the costs more than the benefits and therefore results in even more advantageous cost-effectiveness ratios for these synergistic scenarios of $20 and $10, respectively.
Can India design and implement a structured ART program that would effectively motivate prevention efforts? If the GOI can meet this challenge, it has an opportunity to harness the benefits of ART in a way that will benefit the country for decades to come.
Even at very low prevailing prices for generic ART medications in India, financing such therapy is very expensive and eliminates only a small portion of the burden of the AIDS epidemic. Although more cost-effective than ever before, the cost of ART still exceeds $100 per life-year saved, much more than many other life-saving interventions.
In view of the large costs and potentially dangerous externalities of financing ART, we recommend that the government proceed cautiously to:
* Implement the “ADHERE” strategy. Support improvements to the quality of unstructured ART provided by the private sector to minimize its negative externalities at the lowest possible cost to the government. Under the central assumptions we adopted, including the higher discount rate of 10% often used for government investment decisions, such a policy is the most cost-effective approach to ART. Furthermore, in the Indian context, the ADHERE policy can reduce the near-term harm done by widespread and otherwise unstructured private ART delivery
* Because exclusive dependence on the ADHERE strategy has negative long-run consequences for the development and spread of resistant strains, continue the current policy of expanding publicly financed ART at a moderate pace that is consistent with maintaining high rates of patient adherence
* Evaluate both the costs and the effects of prevention programs. Not only are they likely to save life-years less expensively than ART, but they will reduce the future need for ART
* Evaluate the costs and effects of alternative ARV regimens. It would be very useful to better understand the determinants of sustained patient adherence to a drug regimen in India
* Measure the prevalence of resistant strains of the virus among people with HIV
* Monitor the effects of improved access to ARV treatment on the risk behaviors of those not under treatment, especially high-risk groups such as sex workers. The government should be prepared to modify its ART and prevention programs in response to evidence of increased risky behavior or disinhibition; and
* In consultation with state and national stakeholders, design and implement an institutional arrangement that rewards effective prevention programs, thereby increasing the likelihood that the availability of treatment generates positive (rather than negative) externalities.