Cost-effectiveness analysis of ‘test and treat’ policy for antiretroviral therapy among heterosexual HIV population in India

Antiretroviral therapy (ART) to human immunodeficiency virus (HIV) patients in India is provided free of cost through a network of 319 care and support centres, 579 ART centres, 1261 link ART centres, 85 ART plus centres and 18 centres of excellence (CoE)1. However, despite this the coverage of ART treatment remains low at 43 per cent1. On the contrary, India is committed to end AIDS epidemic as a public health threat by 2030, a goal which aspires 90-90-90 targets aimed at diagnosing 90 per cent of total people living with HIV (PLHIV), putting 90 per cent of them on active ART and achieving viral suppression in at least 90 per cent of those on ART1. In order to achieve this, the National AIDS Control Organisation (NACO) has revised its ART policy as per the World Health Organisation (WHO) guidelines, to initiate ART treatment immediately upon diagnosis (test and treat)1,2.

Any shift in the treatment guidelines such as test and treat implies large scale-up of ART services. A previous analysis of implication of changing the cut-off of initiation of ART treatment from 350/mm³ CD4 cell count to 500/mm³ reported that number of eligible patients will increase from 0.92 to 1.17 million3. However, no assessment of economic implications associated with change in ART guidelines was undertaken. With the setting up of Health Technology Assessment in India (HTAIn) in India, there is an increasing emphasis on justifying new interventions based on their value for money through cost-effectiveness analyses4,5. Moreover, National Strategic Plan for HIV/AIDS and sexually transmitted infections (STI) 2017-2024’ of NACO specifically enlist the identification of cost-effective approaches through stochastic modelling of interventions among the areas of priority for evidence generation6. Furthermore, other countries have evaluated cost-effectiveness of test and treat in their settings7-9 for evidence-based decision-making and rationale resource allocation. It hence, becomes important to ascertain cost-effectiveness of this intervention in the Indian context also, for best outcomes with given funding.

In terms of evidence on cost-effectiveness of ART therapy, two Indian studies are available10,11. Maddali et al11 reported that early initiation (CD4 <500/mm³) is cost-effective as compared to late initiation (CD4 <350/mm³). However, it does not specifically include the test and treat as a scenario. While Eaton et al10 assessed the cost-effectiveness of introduction of test and treat in four countries including India; most of the parameters on the valuation of cost and consequences were regional and not India specific. Second, the clinical evidence was derived from men having sex with men (MSM) population and considered an idealistic scenario which did not factor in the lack of adherence in the real-world scenario.

Hence, this analysis was undertaken to evaluate the cost-effectiveness of the fairly recent NACO intervention, i.e. immediate initiation of ART (test and treat policy), through estimation of the cost per quality-adjusted life-year (QALY) gained.

Material & Methods

Model overview: A probabilistic Markov model (Fig. 1) was used to simulate the disease progression and estimate the costs and consequences. Scenario I composed of ‘test and treat’. The two comparators included ART initiation at CD4 count <500/mm³ (Scenario II) and CD4 count <350/mm³ (Scenario III), respectively (Fig. 1). All costs were calculated from health systems perspective because ART is delivered to all HIV patients for free through the public healthcare system2. The analysis was undertaken using the annual cycles and a life time horizon of HIV patients, i.e. we modelled the costs and health outcomes of the cohort were modelled till the average life expectancy of the cohort was reached. All future costs and consequences were discounted at three per cent annual rate to adjust for the time difference between money spent and benefits gained. The probability of acquiring opportunistic infections (OI) such as herpes, other viral infections, tuberculosis and candidiasis was also incorporated according to their CD4 cell count. Second, the development of malignancies such as non-Hodgkin’s lymphoma, Kaposi’s sarcoma and head-and-neck cancers was also modelled for all scenarios. Third, the development of adverse drug reactions such as hepatitis, lipodystrophy, anaemia, skin reactions, gastric disturbances and immune reconstitution inflammatory syndrome, etc. as a result of ART administration was also incorporated into this model.

Close

Fig 1

Markov state model. ART, antiretroviral therapy; STI, sexually transmitted infections.

Export to PPT

Taking 2017 as a base year, all adult PLHIV registered with NACO (1,141,531) entered the model and were assigned to different stages according to their CD4 cell count12. Each of the three scenarios differed in terms of ART initiation. In all three scenarios, the disease followed its natural course and patients were under continuous care through counselling, testing of CD4 levels and prophylaxis/treatment for OIs. Finally, incremental cost per QALY gained was computed from ‘test and treat’ (Scenario I) as compared to Scenario II and III, respectively. In addition, incremental cost per QALY gained was also estimated for Scenario II as compared to Scenario III as depicted in the Supplementary Figure 1.

Transition probabilities from one stage of HIV to another were calculated using the primary longitudinal follow up data obtained from a CoE situated in a large tertiary care hospital in North India. A cohort analysis was done after extracting information comprising of 1115 life-years follow up data of HIV patients registered with CoE. This included patients on ART as well as those on pre-ART care. The mix of patients included patients with the reported levels of adherence to the therapy along with those on the second or third line of ART drugs making results more realistic as compared to an RCT. The annual rate of transition from one stage to another was calculated, which was then used to compute annual transition probabilities which are summarized in Tables I and II. Detailed methodology used to derive these probabilities is provided

HIV stage wise health-related quality of life data, i.e. health state utility values corresponding to our model transition states are not available from India. Hence, these data were used from a study of health-related states of HIV patients reported for the United States13. We also modelled the risk of subsequent transmission of HIV by PLHIV through the heterosexual route for each of the different scenarios using Weinstein’s equation14. By means of Weinstein’s equation, we incorporated the effect of various behavioural factors such as number of sex partners, frequency of sex acts per partner and condom use along with other factors such as probability of transmission through different routes, prevalence of HIV and STDs in the general population and efficacy of condom in the prevention of HIV transmission. The input parameters for Weinstein equation were obtained from the national level behavioural sentinel surveys in India and other published literature7. Major model input parameters used in the analysis along with sources are given in Table III.

Costing: Overall, the cost in each scenario comprised of the cost of ART treatment, pre-ART care, and treatment of OIs and management of adverse drug effects due to ART. The difference in cost between different scenarios is typically attributable to the differences in the number of eligible patients on pre-ART and ART treatment and the duration of treatment due to differences in longevity3. The difference in the number of HIV patients requiring pre-ART and ART care was a result of change in criteria for the initiation of treatment and the number of new HIV transmissions. Differences in longevity are attributable to improved survival with early initiation of ART. Cost of delivery (COD) of ART as reported from a recent study which used the bottom-up costing approach to ascertain the annual cost at different levels of service delivery, i.e. CoE and ART centre, was used15. This analysis also incorporated cost of delivering ART to patients on second as well as third line therapy, which make the results of this analysis more realistic. Cost of treatment of OIs was taken from published literature. Cost of treatment of herpes infection, other viral infections and candidiasis was taken from an analysis done in 14 public sector STI clinics in the State of Andhra Pradesh16. Cost of treatment of tuberculosis was obtained from the study done in Tamil Nadu State covering all public health facilities in one district17. This cost included capital costs such as infrastructure, furniture, equipment, instruments, etc. and various recurrent costs such as HR cost, cost of drug regimen, sputum examination cost, cost of chest X-rays done and cost of monitoring/supervision, etc. Cost of management of malignancies was derived from an economic costing done in a large tertiary care hospital in north India18. This analysis included the capital as well as recurrent costs pertaining to service delivery including staff salaries, equipment, space rent and consumable, etc. Cost of management of adverse drug reactions or complications due to ART was used as reported from the analysis of data from a Chennai-based treatment and research institute covering about 16 cities of India19. All costs were adjusted to 2018 using year specific inflation rates for India based on consumer price index.

Sensitivity analysis: Univariate analysis was done by varying various input parameters pertaining to cost, health utility states, demography and epidemiology from their lower to upper value to ascertain their effect on overall incremental cost-effectiveness ratio (ICER) of the intervention. The results were sorted according to their impact on ICER and a tornado chart was formulated to present the effect of the most impactful input parameters on ICER of Scenario I vs. Scenario II.

Probabilistic sensitivity analysis (PSA) was also performed using MS Excel and Visual Basic (Microsoft Office 2013) to compare the effect of joint variation of all the inputs parameters on ICER. We performed 1000 simulations to ascertain the variability in ICER using the different random values for selected input parameters using beta, gamma and log-normal distribution. The results were plotted in cost-effectiveness acceptability curve and cost-effectiveness plane.

Results

Costs: Implementing test and treat in India at the national level will impose extra cost to the healthcare system due to increased number of eligible patients and overall longevity of treatment. This cost was ascertained by this study to be ₹ 348542.30 million during lifetime horizon in test and treat, in comparison to ₹ 326272.30 million in Scenario II and ₹ 274694.10 million in Scenario III, as summarised in Table IV and Supplementary Table V. Major proportion from the total expenditure made by the healthcare system on treatment and care of PLHIVs was of cost of ART delivery which constituted about 90-95 per cent of the total cost, whereas only 5-10 per cent was the share of all other expenditures combined, viz. treatment of OIs and management of complications and adverse effects of antiretroviral drugs as summarized in Table IV. If implemented with immediate effect across the PLHIV registered with NACO, the test and treat strategy will put extra budgetary impact (estimated using HTAIn Budgetary Impact Assessment Guidelines20) of ₹ 1382.10 million on the first year. Subsequently, the figure may decrease with 1047.20 million in second year and 804.10, 624.70 and 492.00 million in the third, fourth and fifth year, respectively.

Effects/consequences: As expected, there was a reduction in the number of new HIV infections through heterosexual population due to the effect of ART. Considering reported levels of adherence to ART, new HIV transmissions were estimated to be minimum, i.e. 0.23 million in the test and treat scenario as compared to Scenario II and III as depicted in Table IV. A total of 16,105 new HIV infections were averted by implementing the test and treat over Scenario II and as 74,875 new HIV infections were averted over Scenario III. Initiating ART earlier also increases the overall life expectancy of HIV patients due slow progression of disease and hence reduction in HIV-related deaths. By implementing test and treat, life expectancy increased by 0.17 years over Scenario II and by 0.60 years over Scenario III. HIV-related deaths during the given time horizon in Scenario I were 0.209 million as compared to 0.227 million in Scenario-II and in 0.270 million in Scenario III, respectively as depicted in Table IV and Supplementary Table I. Apart from the difference in years of life lived by PLHIVs in all three scenarios, a substantial difference in the quality of life of HIV patients was estimated, as highlighted in the Supplementary Table II. The number of QALYs lived in the ‘test and treat’ scenario as compared to Scenario-II and III was ascertained to be 12.91, 12.57 and 12.18 million, respectively in lifetime horizon as depicted in the Supplementary Table III.

Cost-effectiveness: To compare health outcomes and costs in all three scenarios of our analysis, ICERs were calculated. Extra cost incurred by implementing test and treat over Scenario II and III was compared with QALYs gained over these scenarios to get the results in the form of cost per QALY gained. After discounting, test and treat was estimated to have ICER of ₹ 46,599 per QALY gained as compared to Scenario II. This ICER was about one third of per capita GDP of India in financial year 2017-2018. Test and treat when compared to Scenario III, was estimated to have ICER of 80,050 per QALY gained which again was less than the then per capita GDP of India. ICER of Scenario II over Scenario III was estimated, which came around to be 0.10 million per QALY gained; again, less than per capita GDP of India. Discounted ICERs for various scenarios are also summarized in Table IV.

Sensitivity analysis: One-way sensitivity analysis run for all the input parameters found that three input parameters, viz. - cost of ART, cost of pre-ART and discount rate have the maximum impact on the cost-effectiveness of the intervention. Seven parameters sorted according to their impact on ICER in the decreasing order are used to construct a tornado chart as shown in Fig. 2.

Close

Fig. 2

Tornado chart of one-way sensitivity analysis (OWSA) using lower and upper bound of input parameters. ART, antiretroviral therapy; ADR, adverse drug reaction; t/t, test and treat; ICER, incremental cost-effectiveness ratio.

Export to PPT

PSA done using randomly selected values using beta distribution for probabilities, utility values and proportions; gamma distribution for all cost parameters. After 1000 simulations using visual basics, results plotted as cost-effectiveness acceptability curve and cost-effectiveness plane are given in Figs. 3 and 4.

Close

Fig. 3

Cost effectiveness plane cloud for 1000 iterations of incremental cost and QALYs gained. CE, cost-effectiveness; ART, antiretroviral therapy; QALYs, quality adjusted life years; PSA, probabilistic sensitivity analysis; LYs, life years.

Export to PPT

Close

Fig. 4

Cost effectiveness acceptability curve (CEAC): Probability of cost-effectiveness and willingness to pay. QALYs, quality adjusted life years.

Export to PPT

Considering all uncertainties involved with estimation of input parameters, it was found that the of test and treat strategy cost may be effective with a 100 per cent probability and at a willingness-to-pay threshold equal to per capita GDP of India.

Discussion

This study was done to assess the cost-effectiveness of the test and treat strategy as per the recommendations of WHO. As evident from the results, the test and treat intervention falls under highly cost-effective interventions (based on cost-effectiveness threshold of and WHO21 and HTAIn22) in the Indian context with ICER less than the per capita GDP of India. This cost-effectiveness is largely due to decreased burden of OIs and averted new HIV cases. Our analysis estimates that, a total of 23,472 new infections can be averted by implementing test and treat, as depicted in the supplementary Table IV. This decrease in new infections can be a significant cost saver as the health system then has to incur less costs to treat new PLHIVs in the future.

In addition, an increase of 0.26 years in life expectancy per HIV patient was estimated by our analysis which then gets translated in additional 0.246 QALYs lived per person in comparison to Scenario II. This increase in life expectancy and quality of life due to early initiation of ART can then be compared to the costs to ascertain its value for money. Positive impact of test and treat may also be due to a reduction in the load of OIs and can further increase over time based on the cumulative averted HIV infections.

Two economic evaluations of changes in strategy for the initiation of ART have been reported previous10,11. Eaton et al10 reported the test and treat to be cost saving over a time horizon of five years as compared to the initiation of ART below a CD4 count of 350/mm³, while Maddali et al11 found early ART (above CD4 count of 350/mm³) in current care continuum as compared to delayed ART (CD4 count of 350/mm³) to be cost-effective over the time horizon of 20 years with an incremental cost of US$ 442 per QALY gained.

The present analysis has several strengths. First, as compared to previous analyses which used either financial cost of ART delivery or some other proxy data for costing10,11, we used cost inputs from a local economic costing analysis of ART delivery15. This cost analysis also factors in the proportion of patients developing resistance to first line therapy, and hence, cost of second or third line therapy, which is significantly higher than the first line therapy15. Second, this study estimated the transition probabilities, based on the analysis of follow up data of Indian HIV patients. However, the data include representation from the northern regions such as Punjab, Haryana, Chandigarh in Himachal Pradesh, Rajasthan, and Jammu and Kashmir. This study incorporates the realistic estimates by including patients at the reported level of adherence to the therapy along with those on second and third line ART therapy; however, to address the variation across States, we varied the adherence rate by 20 per cent on both sides for generalizability of results. Rates of transmission of HIV were ascertained in our study by using Weinstein’s equation which factors in several behavioural parameters specific to the Indian population23. Our model also incorporated the most common OIs and their effect on health utility weights as well as cost of care due to early initiation of ART, which is a life long therapy and cannot be stopped once started. As a result, chances of development of toxicity to first line therapy and adverse effects can be a major issue, which was incorporated in our model.

Despite the strengths, the study also had some limitations. This is, however, acknowledged as a limitation that the data used to calculate transition probabilities are assumed to represent the national population as this is the first evaluation which uses local data on clinical effectiveness of ART. Furthermore, the adherence to the therapy could be different in other regions such as the southern States of the country; therefore, it would impact the cost-effectiveness of intervention. The spectrum of OIs in HIV is wide, but only five potential OIs were considered to be modelled to avoid the complexity. In the selected OIs too, co-infections are possible in real-life situation. However, this was not included into the model in view of data limitations. It was assumed that all these infections, which were being covered are mutually exclusive and are independently occurring in case of HIV infection. We modelled only for the heterosexual route of subsequent transmission of HIV. This is likely to underestimate the cost-effectiveness of test and treat by not accounting for the reduced HIV transmission for MSM and injecting drug users. Finally, while major data used in parameters were local, quality of life data for the Indian population was not available in the literature. Hence, data from similar studies elsewhere were used as input to the model, which is a limitation and can impact the accuracy of ICER due to potentially different health preferences of the Indian population. It would also be valuable if some research on number of patients not registered with NACO is undertaken in the future. This will help to give more accurate estimates about cost-effectiveness of test and treat. Furthermore, this analysis only covers public sector patients in terms of cost, clinical impact and the overall cost-effectiveness in the private sector could be different. Cost of service delivery in the private sector is comparatively more, which means, if we assume a scenario where more patients start availing services from public health facilities through NACO, the ICER would be further lowered, i.e. the intervention would become more cost effective.

The implementation of test and treat will substantially increase the cost of care and support for India’s HIV control programme in the short run, but in later years, the total cost incurred by healthcare will decrease drastically. Hence, test and treat should be considered as a long-term investment in healthcare. Our model suggests that, at the reported rates of adherence to ART, many potential benefits of this policy will remain unexplored. As cost of delivery of ART through a tertiary care centre is found more than that of other ICTCs; patients receiving ART from tertiary care settings and CoEs must be kept low. It can be further reduced by linking maximum patients to link ART or ART centres to further reduce the overall cost of delivery of ART. Major expenditure on the delivery of ART to patients is of antiretroviral drugs; so, government can make efforts to further reduce the cost by amendments in procurement. As evident form the transition matrix developed and used in this study, the mortality was high in the advanced stages of HIV with low CD4 count; efforts must hence be made to diagnose and link the HIV patient to the treatment at the early stages of infection. This will reduce the HIV-related mortality and OI and hence the overall cost incurred by the health system.

Supplementary Material