In low-income settings, testing for CD4 cell counts and viral load (VL) are not routinely available . Monitoring of CD4 cell counts permits the proper identification of two important subgroups of patients: those with an adequate CD4 cell count but with symptoms that emulate advanced HIV infection, and those who are asymptomatic but whose CD4 cell count shows that HIV disease has progressed to the point where ART would be beneficial. Basing treatment on CD4 cell count results spares the former subgroups unnecessary ART treatment, and the latter patients from imminent risk of opportunistic infections and progression of AIDS. After treatment initiation, when clinicians decide whether to switch to second-line treatment, laboratory tests can improve decision-making.
World Health Organization (WHO) guidelines classifies monitoring tests for antiretroviral therapy (ART) into three categories: level 1 (for primary health centers, requiring only a rapid HIV test and hemoglobin), level 2 (for district hospitals, adding CD4 cell count assessment), and level 3 (for regional referral centers, adding viral load assessment) . Programs for ART must decide whether or not these tests can be offered locally, based on the availability of scarce resources.
Besides laboratory monitoring, another strategic decision is whether or not to offer second-line ART regimens. Early detection of non-response followed by initiation of an effective second-line regimen should reduce the duration of impaired immune function and improve survival. However, second-line regimens tend to cost four to six times more than first-line regimens .
Can investments in CD4 cell count and viral load testing capabilities be justified in terms of cost per quality-adjusted life years (QALYs) gained? Specifically, this simulation was designed to compare outcomes and costs of a reference no treatment strategy (NO ART) to antiretroviral treatment in settings with four types of disease monitoring: (a) syndromic management without laboratory tests (ART ONLY); (b) syndromic management plus total lymphocyte counts every 6 months (TLC); (c) syndromic management plus CD4 cell count assessment every 6 months (CD4); (d) syndromic management plus CD4 cell count every 6 months and viral load assessment 4 weeks after the initiation of treatment, then every 6 months (VL).
A computer-based discrete event simulation of HIV disease was developed for a cohort of 10 000 virtual patients in order to assess the incremental impact and cost effectiveness of alternative laboratory monitoring strategies. The course of each individual patient's disease status, as reflected by CD4 cell counts, viral load, WHO stage (I–IV), and likelihood of death was updated every 6 months as a succession of draws from an array of state-contingent statistical distributions. The model was calibrated to reflect the evolution of disease before and after ART initiation, using parameters from clinical studies.
A 10 year horizon was adopted and outcomes were assessed as life years, QALYs, and costs from the societal perspective, discounted at 3%. The external effect of the patients' health behavior or health condition on outsiders was not considered. One-way sensitivity analysis and multivariate stochastic simulation assessed the robustness of the conclusions.
Health systems model
Each of the four monitoring strategies could operate either with or without availability of a second-line treatment regimen, making a total of eight strategies. Outcomes were obtained for eight comparable cohorts of 10 000 simulated patients managed using one of eight treatment strategies under consideration. A comparison scenario, no ART treatment, assumed palliative treatment of end-stage AIDS, without ART or prophylaxis for opportunistic infections. Starting at time 0, all patients began to visit the clinic every 6 months for 10 years, or until death. It was assumed that all clinicians in a health system would be fully adherent to assigned guidelines, and that all patients were naive to previous ART. Patients were assumed to be as adherent to their treatment as were the cohorts that generated the clinical outcome parameters that the model is based on.
The ART ONLY strategy refers to an ART program in which clinicians have access only to HIV serology, hemoglobin, and the WHO clinical stage determined by history and physical examination. In the TLC strategy, clinicians have HIV serology, clinical staging, plus total lymphocyte count assays every 6 months. In the CD4 strategy, clinicians have access to all of the above as well as CD4 cell counts every 6 months. Finally, in the VL strategy, clinicians have access to all of the above plus viral load assays every 6 months as well as 4 weeks after treatment initiation, to assess virological failure. In each strategy except VL, clinicians were kept blinded to one or more pieces of information known to the modeler.
Treatment switching decisions were also based on laboratory monitoring. It was assumed that clinicians who have viral load information would switch patients who fail to show a fall in viral load by 1 logarithm 1 month after starting treatment. In the CD4 algorithm, clinicians would switch patients at the 6th month visit or later whose CD4 count stayed < 200 cells/μl and who had failed to gain at least 50 cells/μ. For the ART ONLY and TLC strategies, patients would be switched if they exhibit clinical stage IV symptoms at both the 6th month and 12th month visits. This implies a minimum 12 month delay in switching therapy for failing patients whose clinicians do not have access to CD4 cell or viral load assessment.
The simulation model is described in more detail in a technical appendix available from the authors online . Briefly, an array of joint normal distributions for CD4 cell count and viral load describes the natural history of disease as well as the statistical correlation between viral load decrements and CD4 cell increments during treatment. AIDS clinical stage, total lymphocyte counts, QALY, and mortality rates for a root population of 10 000 derive from each simulated patients' CD4 cell count and viral load. Parameters are drawn from published literature [5–9].
Costs for untreated patients accrued with each semiannual clinic visit. For the base case, relatively low estimates of $200 per year for pharmacy costs of first-line treatment, and $900 per year for second-line treatment were used. In sensitivity analyses, ranges from $100 to $1000 per year in treatment costs for first-line treatment, and $400 to $1400 per year for second-line treatment were tested. Baseline opportunistic infection costs for patients with stage IV disease was taken to be $150 [10,11]. A one-time cost of $325 was assigned for switching to a second-line regimen, to cover additional clinic visits, testing, and hospitalization for some patients.
Treatment initiation protocols are outlined in Table 1 and costs for patients accruing with each semiannual clinic visit are listed in Table 2.
The findings from the reference case are summarized in Fig. 1. Because of discounting, the theoretical maximum number of discounted QALYs that 10 000 people in perfect health could accrue in 10 years is 87 861. Without treatment, the baseline model of natural history predicts that a population infected on average 7.5 years ago will experience 22 676 QALYs (not shown in Fig. 1) at a cost of $1.3 million, primarily for the medical care of AIDS complications. Under these natural history conditions, 9271 of the 10 000 will die during the subsequent decade. With ART, the number of deaths could be reduced to 2311–2714 in the same population, depending on the clinical strategy. Furthermore with ART, the population could more than double their QALY expectancy to between 45 717 and 47 637 discounted QALYs. Of the lost QALYs, approximately 30% are attributable to deaths, and the remainder to reduced quality of life. With ART, the population attains roughly half of the theoretical maximum healthy life expectancy.
The least costly option, ART ONLY with no second-line therapy available, generates 45 736 discounted QALYs and 2714 deaths with discounted costs of $15.8 million over the decade. Paying for more than the least cost option would save more lives at higher costs. Without second-line treatment available, an average of 900 additional QALYs can be gained by adopting CD4 cell count testing at an additional cost of $221 000 for a mean cost of $245 (median $238) per QALY gained (Table 3). This additional QALY gain is primarily from the more timely initiation of ART treatment. The costs of the CD4 cell count tests are offset significantly by eliminating costly drug treatment for patients who meet criteria on clinical grounds, but whose CD4 cell counts remain adequate.
The availability of second-line treatment dramatically alters the effect of the various laboratory monitoring options on costs and outcomes. In this case, advancing from ART ONLY to CD4 strategy yields 1380 (540 + 840; as per Table 3) additional QALY at an additional cost of $7.9 million ($612 000 + $7 348 000; as per Table 3). Adding CD4 cell count testing capacity in this situation enables clinicians to identify more patients who would benefit from second-line treatments than they could by using only clinical judgement or total lymphocyte counts. The second-line treatments combined with either CD4 cell count and/or viral load tests rescue more patients from treatment failure, but drugs are substantially more expensive, thus increasing the cost of the overall program.
As shown in Table 3, the median incremental cost effectiveness ratio (ICER) for introducing ART ONLY compared with no ART is $628 [interquartile range (IQR), 626–630] without second-line treatment and $684 (IQR, 680–687) with second-line treatment. In the absence of second-line treatment, any decision to advance to a TLC strategy would be dominated by the CD4 strategy in the sense that advancing to TLC expresses a willingness to pay at least $776/QALY, and a decision-maker would be well-advised to advance further to the CD4 strategy where additional QALYs can be gained at a savings of $635/QALY.
To this point, the analysis assumed that the decision whether to have second-line drugs available was independent of the choice of laboratory monitoring approach. However, the simulation was also able to address the cost effectiveness of second-line therapy independent of the choice of laboratory monitoring. Based on the mean values from Table 3, having second-line drugs available in the CD4 algorithm compared with not having them available in the CD4 algorithm would produce an additional 470 QALYs at a cost of an additional $9million, and thus cost $19 147/additional QALY.
Studying the various algorithms in scenarios with no second-line treatment available helps to isolate cost effectiveness of the laboratory tests in regard to treatment initiation alone, because treatment switching is not an option. Interestingly, the model suggests that in the absence of second-line treatment it would not be cost effective to use viral load testing, even if the viral load tests were free. If second-line treatment is unavailable, resetting the price per viral load test from $25 to $0 would change the ICER for the VL algorithm from $16 139 to $1622 in the bottom row of the top panel of Table 3, leaving all other ICER values unchanged. The reason the cost of the VL algorithm is higher in programs with no second-line treatment available is that viral load detects some additional patients whose viral load is low enough to merit first-line treatment even though they do not meet treatment criteria by CD4 cell count or clinical examination. Even if the tests themselves are free, they lead to more person years of treatment, which is less efficient when initiated by viral load criteria.
If second-line treatment is available, the cost of viral load testing would have to decline to $14/test to achieve the same median ICER as the CD4 strategy, holding CD4 cell count at $5 per test. The CD4 strategy breaks even with a $25 viral load test at a price of $21/CD4 cell count test. The relative cost effectiveness of the various strategies also depends on the costs of treatment, with higher costs of first-line treatment rendering CD4 strategies more advantageous (Fig. 2). In the absence of second-line regimens, the CD4 strategy begins to save money (i.e., has a negative cost per incremental QALY) when first-line ART treatment costs more than $158 per year. When second-line treatments are available, CD4 strategy begins to save money when a year of first-line treatment costs more than $747 (Fig. 3). Higher prices for second-line treatments make both VL and CD4 strategies cost more per incremental QALY but have little impact on the cost effectiveness of the TLC and ART ONLY strategies.
Results for first-line treatments were robust to univariate sensitivity analyses except if it was assumed that ART has no side effects, leading to no reductions in quality of life, then the TLC strategy produces more QALYs than the CD4 strategy. This occurs because ART slows disease progression, but in this particular sensitivity analysis the downside of adverse drug effects from a premature ART start has been assumed away. Sensitivity tests to a discount rate of 6% did not alter the basic pattern of results shown in Fig. 1.
The cost effectiveness of second-line treatment was sensitive to the assumptions about the efficacy of both first- and second-line treatment. When first-line treatment was assumed to cause CD4 cell increments and viral load decrements that were half as large as baseline, this led to higher utilization of second-line treatment. In the baseline model, each patient could expect to spend 0.3, 0.4, 2.8, and 3.6 years on second-line medication per decade under ART ONLY, TLC, CD4, and VL algorithms respectively. When first-line treatment was assumed to be half as effective as baseline, the corresponding expected time spent on second-line treatment per patient was 0.3, 0.4, 4.8, and 6.7 years, respectively. In the CD4 algorithm, use of second-line treatment added only 0.15 additional discounted QALYs per person while costing an additional $1612 per person, yielding an ICER of $10 730 per discounted life year. If it was further assumed that second-line treatment is as effective as first line at improving CD4 cell counts, then the gain per patient rose to 0.24 life years, yielding an ICER of $7225 per discounted life year. This assumed a $900 annual cost of second-line drugs. The ICER declined linearly such that the ICER is $940 lower for every $100 reduction in the price of second-line drugs. For example, if the cost of second-line treatment declined to $400 per year, the ICER would be $3465 per discounted life year.
The estimated cost effectiveness of first-line treatment with no monitoring in our study at $626–630/QALY gained is close to a prior estimate of $547 per disability-adjusted life year averted , despite substantial differences in methodology. Another recent study estimated a cost of $620 per life year gained obtained from ART plus prophylaxis without CD4 cell count testing compared with prophylaxis alone in a model calibrated to data from a cohort in Côte d'Ivoire, where first-line treatment was only 51% effective at achieving virological suppression . For purposes of comparison, our model predicts a cost of $380 per discounted life year gained comparing the ART ONLY model with no second-line treatment with NO ART. Important differences are that in our model first-line treatment is 74% effective at achieving virological suppression at 1 year and our comparison group did not receive prophylaxis against opportunistic infections. The Côte d'Ivoire model suggested much higher benefits from introducing second-line therapy, with an average gain of 10 months of survival for every patient; this contrasts to 1.8 months per patient gained from second-line treatments in our model. This difference may be explained by both poorer effectiveness of first-line treatment in the Ivoirian model and the different assumptions about the efficiency with which failing patients are detected and switched to second-line treatment. Survival outcomes data emerging from current studies in low-resource countries should clarify the relative value of second-line treatment. Further tests of the validity of our model are discussed in detail in the technical appendix.
An important limitation of our model that is shared by some other recent models is the failure to consider disease transmission. ART could have opposing effects on transmission by lowering infectivity but prolonging the survival of potentially infective hosts. In addition, the quality of preventive counseling, laboratory monitoring, and consistency of patient adherence could alter the spread of resistance [14,15].
As expected, health improvements gained by introducing a basic ART package are large , outweighing the incremental benefits gained from refining the treatment process by introducing laboratory monitoring and second-line regimens. As shown in Table 3, introduction of a basic ART regimen for a population of 10 000 would save 6500 lives and stop the loss of 23 000 discounted QALYs over one decade. Each subsequent refinement to laboratory monitoring and to the availability of second-line drugs saves no more than 50–150 lives and saves only 250–850 QALY. However, these refinements do offer real benefits, and they are already being introduced progressively in resource-limited countries. A major potential benefit from laboratory monitoring and second-line regimens is improved capability to stem the emergence and spread of drug-resistant HIV strains in affected populations by rapidly identifying and treating patients with virological failure. Emerging data on the scope of this problem will likely indicate that control of drug resistance is essential if ART programs are to succeed.
In cost-effectiveness terms, many of the monitoring options offer similar ICER values to that obtained when first introducing ART. From the solid line in Fig. 1, one sees that ART ONLY, TLC, and CD4 strategies in the absence of second-line treatment are all of similar total cost ($15–35 million), but that of these three the CD4 strategy in the absence of second-line treatment registers the highest gain to health. By offering second-line treatment, CD4 and VL strategies in the model will magnify the health gains from this decision, but will increase costs from under $17 million to $25–32 million. The model predicts that without CD4 cell and viral load testing, the second-line treatments will seldom be used; in this case, neither their benefits to health nor their high costs would be realized. Finally, according to the model, if second-line treatments are not available, the costs of CD4 cell count tests would be offset by savings from fewer inappropriately treated patients. The cost offsets from CD4 cell count testing are notably larger in populations with less-advanced disease (see Technical appendix, Fig. 10a).
Ongoing cohort trials of ART will shed additional light on the effects and costs of various strategies. The sensitivity analyses in this model predict that the absolute costs and gains will not generalize across populations with different disease severity. However, our extensive sensitivity analysis predicts that CD4, TLC, and ART ONLY strategies will have relatively similar costs to each other in the absence of second-line treatment under a wide variety of assumptions and settings.
Changing prices of drugs and laboratory tests will alter the cost effectiveness of the strategies in the model, which can be updated easily to reflect these changes. In addition the model can be readily altered to accommodate populations with less advanced or more advanced disease.
Simulation models of HIV have been helpful in guiding policies and informing research strategies [17,18]. Ultimately, policies are directed by the objectives of the people affected and the decision-makers serving them. There are many considerations other than cost effectiveness that would bear upon the choice of an optimal strategy for scaling up an ART program in a resource-constrained setting. The existing infrastructure and availability of facilities, trained personnel, logistics, and availability of funds would be foremost considerations. Many other important consequences stem from these decisions that are not captured by the model. The occurrence and spread of resistant strains of virus would be influenced by the availability of second-line treatment.
It would be unusual if cost effectiveness became the only deciding factor in making health policy. If cost effectiveness were the primary concern, this model indicates that the ART ONLY program is sufficiently cost effective that it would be unwise to postpone ART scale-up until second-line drugs or better laboratory facilities became available. In the absence of second-line treatments, investments in CD4 cell count testing capability would offset their costs by restraining the use of therapy in patients who do not need it. In the presence of costly second-line treatments, CD4 cell count and viral load tests would enhance the appropriate selection of cases for second-line treatment, resulting in higher utilization of second-line treatment, more lives saved, and substantially higher costs unless the cost of second-line treatment falls dramatically. Ultimately, the choices between these options will rest with payers and policymakers.
DB and AC had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Helpful comments were received from John Bartlett, Richard Chaisson, Steven Chapman, Ricardo Diaz, Ronald Gray, Sabina Haberlen, Andrew Kambugu, Keith McAdam, Carolyn Mohan, David Thomas, Thomas Quinn, Steven Reynolds, Jeanne Brosnan and seminar participants at the Infectious Diseases Institute in Kampala and Rakai Health Sciences Program in Kalisizo, Uganda. We thank the MACS/WIHS study personnel and Lisa Jacobson for permitting us access to these data to compute the covariance structure of levels and changes in log viral load and the fourth root of CD4.
This study was conceived jointly by BD Biosciences and the investigators. BD Biosciences funded and participated in the study through a grant from Becton, Dickinson and Company through its BD Biosciences Segment to Johns Hopkins Bloomberg School of Public Health. David Bishai has been a consultant to BD Biosciences; David Durack is an employee of BD Biosciences.
1. World Health Organization. Scaling Up Antiretroviral Therapy in Resource-Limited Settings: Guidelines for a Public Health Approach. Geneva: World Health Organization; 2002.
2. World Health Organization. Scaling Up Antiretroviral Therapy in Resource-Limited Settings: Treatment Guidelines for a Public Health Approach, 2003 Revision. Geneva: World Health Organization; 2004.
3. Gutierrez JP, Johns B, Adam T, Bertozzi SM, Edejer TT, Greener R, et al. Achieving the WHO/UNAIDS antiretroviral treatment 3 by 5 goal: what will it cost? Lancet 2004; 364:63–64.
5. Taylor JM, Tan SJ, Detels R, Giorgi JV. Applications of a computer simulation model of the natural history of CD4 T-cell number in HIV-infected individuals. AIDS 1991; 5:159–167.
6. Taylor JM, Law N. Does the covariance structure matter in longitudinal modelling for the prediction of future CD4 counts? Stat Med 1998; 17:2381–2394.
7. Taylor JM, Cumberland WG, Sy JP. A stochastic model for analysis of longitudinal AIDS data. J Am Stat Assoc 1994; 89:727–736.
8. Lyles RH, Munoz A, Yamashita TE, Bazmi H, Detels R, Rinaldo CR, et al. Natural history of human immunodeficiency virus type 1 viremia after seroconversion and proximal to AIDS in a large cohort of homosexual men. Multicenter AIDS Cohort Study. J Infect Dis 2000; 181:872–880.
9. Fauci AS, Pantaleo G, Stanley S, Weissman D. Immunopathogenic mechanisms of HIV infection. Ann Intern Med 1996; 124:654–663.
10. Hansen K, Chapman G, Chitsike I, Kasilo O, Mwaluko G. The costs of HIV/AIDS care at government hospitals in Zimbabwe. Health Policy Plan 2000; 15:432–440.
11. Guinness L, Arthur G, Bhatt SM, Achiya G, Kariuki S, Gilks CF. Costs of hospital care for HIV-positive and HIV-negative patients at Kenyatta National Hospital, Nairobi, Kenya. AIDS 2002; 16:901–908.
12. Hogan DR, Baltussen R, Hayashi C, Lauer JA, Salomon JA. Cost effectiveness analysis of strategies to combat HIV/AIDS in developing countries. Br Med J 2005; 331:1431–1437.
13. Goldie SJ, Yazdanpanah Y, Losina E, Weinstein MC, Anglaret X, Walensky RP, et al. Cost-effectiveness of HIV treatment in resource-poor settings: the case of Côte d'Ivoire. N Engl J Med 2006; 355:1141–1153.
14. Baggaley RF, Garnett GP, Ferguson NM. Modelling the impact of antiretroviral use in resource-poor settings. PLoS Med 2006; 3:e124.
15. Bertozzi S, Bautista-Arredondo S. Modeling the impact of antiretroviral use in developing countries. PLoS Med 2006; 3:e148.
16. Koenig SP, Leandre F, Farmer PE. Scaling-up HIV treatment programmes in resource-limited settings: the rural Haiti experience. AIDS 2004; 18(Suppl 3):S21–S25.
17. Paltiel AD, Weinstein MC, Kimmel AD, Seage GR III, Losina E, Zhang H, et al. Expanded screening for HIV in the United States: an analysis of cost-effectiveness. N Engl J Med 2005; 352:586–595.
18. Goldie SJ, Paltiel AD, Weinstein MC, Losina E, Seage GR III, Kimmel AD, et al. Projecting the cost-effectiveness of adherence interventions in persons with human immunodeficiency virus infection. Am J Med 2003; 115:632–641.
19. Mulligan JL, Fox-Rushby J, Adam T, Johns B, Mills A. DCPP Working Paper 9: Unit Costs of Health Care Inputs in Low and Middle Income Regions
. Baltimore, MD: National Institutes of Health; 2003. http://www.fic.nih.gov/dcpp/wps/
20. Gidwani G. Patterns of Health Care Seeking in Uganda. Baltimore: Johns Hopkins Press; 2004.
21. Uganda Bureau of Statistics. Statistical Abstract. Entebbe: Government of Uganda; 2000.
cost effectiveness; HIV; AIDS; CD4 lymphocytes; viral load; antiretroviral treatment; resource-limited settings; laboratory testing; developing countries
© 2007 Lippincott Williams & Wilkins, Inc.