As developing countries launched large-scale HIV/AIDS treatment programs in the first half of this decade, the cost of antiretroviral medications, as well as their efficacy and ease of storage and use, was a major consideration in selecting first-line ARV regimens. As a result, many countries chose a first-line regimen that contained the drug stavudine (d4T). A survey conducted in March 2006 of 23 resource-limited countries, of which 17 were in sub-Saharan Africa, found that nearly 70% of adult patients were on first-line regimens that included d4T, with d4T-lamivudine (3TC)-nevirapine (NVP) accounting for 61% of patients and d4T-3TC-efavirenz (EFV) for 8%. For most other patients, zidovudine (AZT) was substituted for d4T.1
Despite its widespread use, d4T is known to be the cause of a large share of the side effects suffered by patients on antiretroviral therapy (ART).2 Peripheral neuropathy, lactic acidosis and symptomatic hyperlactatemia, and lipodystrophy and lipoatrophy are all attributed to d4T. Several studies have reported high rates of peripheral neuropathy and hyperlactatemia in African cohorts. Almost a third of South African patients in a public sector program experienced peripheral neuropathy within 2 years of starting treatment3; in Kenya4 and Cote d'Ivoire,5 a fifth of observed patients did. In South Africa and Botswana, high rates of symptomatic hyperlactatemia and lactic acidosis have been reported from multiple cohorts; all suffered high levels of mortality among patients diagnosed with lactic acidosis.6-8 More than a third (34%) of Rwandan patients on d4T for at least 6 months developed lipodystrophy.9 The rates of d4T-related toxicities reported in African cohorts are somewhat higher than those found in clinical trials in industrialized countries.10
All these toxicities reduce the quality of life of patients experiencing them. In a study in Uganda, where d4T is included in first-line treatment, patients cited toxicities or fear of toxicities as the most common reason for modifying ART regimens and, importantly, the second-most common reason offered for discontinuing ART entirely, cited by 21% of the 14% of patients who discontinued treatment.11 It is thus possible that d4T-related toxicities contribute to the relatively high levels of loss to follow-up (LTFU) that frustrate many African ART programs.12 Fear of toxicities may also deter some patients from ever starting ART.
One drug that is often considered as a potential replacement for d4T is tenofovir disoproxil fumarate (TDF). TDF has been associated with renal toxicities, but the incidence of such events and of other side effects is extremely low.10 It is widely regarded as a superior drug to d4T, but it is also more expensive. In South Africa, for example, the private sector price for a 1-month supply of Viread by the generic manufacturer Aspen Pharmacare is $30.66,13 whereas d4T costs just $3.39 at the public sector tender price. (An exchange rate of R6.5/$1 is used throughout this paper). The lowest confirmed price for TDF is that negotiated for public sector procurement in Zambia, $17.00/month (J. Wilson, BA, BS, MS, MPH, personal communication, October 2007).
Although TDF is substantially more expensive to purchase, management of d4T-related toxicities consumes a good deal of ART providers' time and resources, and patients' survival and quality of life are diminished. Switching to TDF may thus result in significant savings in health care resource utilization that should be taken into account in estimating the budgetary impact of TDF. It may also generate quality-of-life improvements that would justify some or all of its higher cost.
In 2007, Zambia announced that it would shortly replace d4T with TDF in its first-line regimen. Namibia, in contrast, stated its intention to replace d4T with AZT.14 AZT is also considerably more expensive than d4T, however, with a public sector price in South Africa of $12.47/month. The World Health Organization has recommended reducing the dose of d4T from 40 mg twice daily to 30 mg twice daily with the hope of delaying side effects without incurring additional costs.15
In South Africa, where the first-line regimen in the national treatment program contains d4T and TDF is not currently available through public sector procurement, an active debate is under way on whether to replace d4T with TDF in the first-line regimen. To help inform this debate, we estimated the costs in terms of medical care expenditures and quality-adjusted life years (QALYs) loss incurred by a large treatment facility in South Africa for d4T-associated toxicities and compared these with the estimated costs of switching to a first-line regimen containing tenofovir.
Using a model developed for this analysis, we compared the costs of care for patients initiated on d4T under South African regimen 1a (3TC-d4T-EFV) (d4T scenario) with the expected costs of care for a regimen that substitutes TDF for d4T (TDF scenario) during the first 24 months after ART initiation. To estimate the costs of each scenario, we modeled a hypothetical cohort of 1000 patients under the 2 scenarios using primary data from the study site and parameters drawn from the literature. Outcome measures were medical care expenditure for ARVs and toxicities (budgetary cost) and incremental cost per QALY gained (cost-effectiveness).
Primary data for the d4T scenario were drawn from the medical record database maintained by the Themba Lethu Clinic (TLC) of Helen Joseph Hospital in Johannesburg, South Africa. TLC is one of the largest ART providers in the country, with more than 6000 adults receiving ART on an outpatient basis as of September 2007. The clinic's database reports the starting and stopping dates for all ARV medications, patient conditions and laboratory results related to starting and stopping decisions, and, in most cases, specific reasons for stopping or switching individual drugs. The database also records all resources used for patient care at the clinic, including outpatient visits, drugs, and laboratory tests. It does not contain complete data on inpatient admissions, and patient conditions are recorded inconsistently, using multiple ICD-10 codes. Reasons for patient LTFU and causes of death are also entered inconsistently or not at all.
Starting with the full roster of patients who ever initiated ART at TLC, we identified all patients who were at least 15 years old and started an ART regimen containing d4T on or after January 1, 2005. Patients who initiated treatment in 2004 were excluded due to concerns about data quality during the early months of the program. We then identified the subsample of patients who had stopped d4T in the first 24 months after ART initiation and collected starting and stopping dates for d4T, reasons for stopping, and clinical conditions reported before the d4T stop date. We also recorded the new drug regimen and the patient's status at the time of data collection (still on new regimen, stopped all ARVs, lost to follow-up, died, or transferred).
For the TDF scenario, no data were available from TLC, as TDF has not yet been prescribed to public sector patients. We relied on published reports to estimate parameters for the TDF scenario.
Definition of Events
Based on published reports of d4T- and TDF-related toxicities and the experience of TLC clinicians, we defined 10 discrete toxicity-related events that could be attributed to d4T and 2 that could be attributed to TDF, as shown in Table 1. Some of these events led to an immediate drug change, others a drug change after a 3-month interruption of all ARVs, and the remainder no change in drugs. The severity levels assigned to each type of toxicity were intended to represent homogeneous resource use patterns. For example, event 2c, severe symptomatic hyperlactatemia with drug change, represents greater resource use than event 2d, nonsevere symptomatic hyperlactatemia without drug change. Events that did not result in drug changes are included because they consumed resources such as outpatient visits. Patients who transferred to another treatment site, died (except deaths due to d4T-related lactic acidosis), or were lost to follow-up (except patients known to have discontinued due to d4T-related side effects) were excluded from the analysis. Although an increase in pancreatitis has previously been attributed to d4T, recent research calls this into question,16 as does the very low incidence of pancreatitis at the study site. We have therefore conservatively excluded pancreatitis as a toxicity attributable to d4T.
In the model, all switches from d4T or TDF were assumed to be to AZT. At the study site, 90% of patients who stopped d4T and remained on ART had a single drug substitution to AZT. All other substitutions were to second-line drugs that are more expensive than AZT. It is likely that AZT will also be the most common drug selected for patients who cannot start or cannot tolerate TDF. Assigning all switches in the model to AZT is thus a conservative assumption.
Model Structure and Transition Probabilities
We developed a state-transition model to estimate the costs and utility gains of the 2 scenarios.17,18 In the model, a hypothetical cohort of patients ages in a sequence of time periods, called cycles. In each cycle, each patient in the cohort occupies one of a set of states that has a fixed cost and utility value associated with it. Transition probabilities determine how many patients in the cohort occupy each state in each cycle. By summing the number of patients in each state and multiplying by the cost of that state, a total cost for treating the cohort can be estimated.
The structure of the model is shown in Figure 1. We defined 5 possible states that patients could occupy, as shown in the Figure. All patients in the study cohort begin in state 1 at time 0. In each cycle, patients can either remain in the same state (dotted arrows) or transition from state 1 to one of states 2-5 (solid arrows) based on the transition probabilities presented below. Some patients who remain in state 1 have an event without a drug change (hatched arrow) during the cycle, whereas others do not have any event.
Each state represents a specific drug regimen that is costed in the model. In addition, transitions between states are assigned costs related to the specific event associated with the transition. For example, a patient who transitions from state 1 in the first cycle to state 3 in the second cycle incurs the cost of one of the events in Table 1 that leads to a drug change with no interruption (ie, event 1a, 1b, 3a, or 4b) and the cost of the new regimen in all future cycles.
Each cycle in the model is 3 months in duration. The model was estimated over a 24-month period covering the first 2 years after ART initiation and thus incorporates 8 cycles. Each patient was allowed only one change to a different state, as states 2-5 all represent end states within the model. Patients who had an event without a drug change and remained in state 1 were exposed in future cycles to the same transition probabilities as all other patients in state 1. Patients who changed regimens (substituted AZT for either d4T or TDF) were assumed to remain on the new regimen for all remaining cycles and to suffer no further events or drug changes. All events were assumed to take place at the midpoint of the cycle.
Transition probabilities from state 1 to all other states in the d4T scenario were calculated using the primary data set and standard survival analysis methods. Probabilities of having an event that did not lead to a drug switch (ie, event 1c, 2d, or 2e) were estimated from the primary data set and/or published reports; where timing of such events could not be determined from the data set, the timing was assumed to parallel that of events that did require a drug change. For example, cases of event 1c, mild peripheral neuropathy not requiring a drug change, were assumed to occur in proportion to the incidence of events 1a and 1b.
Almost no information is available about the rates and timing of TDF-related events and drug switches in Africa, and reports of clinical trials, while offering estimates of event rates, rarely provide detailed descriptions of timing or resource use. We used published information whenever it was available and filled in the gaps with assumptions based on the experience of clinical colleagues.
Estimates of Resource and Event Costs
Unit costs for drugs, laboratory tests, outpatient visits and infrastructure, and other fixed costs at TLC were taken from an earlier analysis of the costs of ART at the site (Rosen et al, in press, 2008). Because no recent estimate has been made of the cost of inpatient care at the site, we took this value from a recently published study conducted in Cape Town, South Africa.19
To determine the average resource utilization for the d4T-related events defined in Table 1, we selected a random sample of TLC patients who had experienced the event and inventoried the clinical resources used from the first indication of the toxicity until d4T was switched to a substitute drug. All outpatient visits, ARVs, and laboratory tests were included; one additional outpatient visit was added for each event that required a drug change, to capture monitoring of AZT tolerance. We then estimated the average resource utilization and cost per event, using the unit cost estimates described above. Because inpatient admission days were not consistently reported in the data set, TLC clinicians estimated the average duration of inpatient care for events requiring hospitalization. For TDF-related events, we estimated resource utilization based on clinicians' experience. In the TDF scenario, 4 urea and electrolytes test groups (sodium, potassium, chlorides, urea, and creatinine) were included for all patients in the cohort and 1 urea and electrolyte was included for the additional 5% of ART patients who are estimated not to be able to start TDF at all. In both scenarios, all patients who switched to AZT were allocated 3 hemoglobin tests. The only non-ARV drug routinely prescribed for toxicity management and not captured in the estimates of inpatient care costs was amitriptyline for peripheral neuropathy; the cost of this drug is negligible and was therefore excluded from the analysis.
Total Cost per Scenario and Cost-Effectiveness Analysis
The total cost of each scenario was the sum of all drug costs and all event costs for the 1000-patient cohort. Second-year costs were not discounted. We then compared the total cost of each scenario and estimated the threshold price of TDF that would make the change to TDF in the first-line regimen cost neutral for the South African Department of Health. It is important to note that the costs included in this analysis are for ARVs and toxicity event management only; costs for routine monitoring and care not associated with side effects of d4T and TDF are excluded. The costs estimated here should thus not be taken as estimates of the total cost of treatment, which can be found in an earlier analysis (Rosen et al, in press, 2008).
In addition to a cost, each event was assigned a quality adjustment weight and duration to determine the number of QALYs lost per event. Weights and durations were taken from the literature or estimated based on clinical experience. Finally, an incremental cost-effectiveness ratio (ICER)-the additional cost of the TDF scenario divided by the additional QALYs gained under the TDF scenario-was estimated and compared with South Africa's per capita gross domestic product (GDP). It is a common practice to define an intervention with an ICER less than 3 times per capita GDP as “cost effective” and an intervention with an ICER less than per capita GDP as “very cost effective.”20
As mentioned above, it is possible that some LTFUs from the treatment program are attributable to d4T toxicities. The database used for this analysis does not contain complete information about reasons for LTFU, however, making it impossible to use primary data to estimate d4T-attributable LTFU rates. We have therefore omitted LTFU from the basic model and instead assumed that all patients, except those who die from lactic acidosis, remain on ART until the end of the study period. We then explore the effects of varying LTFU rates on the results in the sensitivity analysis. Other causes of LTFU and death are also excluded from the model.
Data Set, Parameters
The primary data set contained records for 5766 adult patients of TLC who started on first-line ART in 2005 or later. Of these, 3534 remained on d4T 2 years later; of the 2232 who had discontinued d4T, 1435 were no longer in care at the clinic, whereas 797 had changed to a new drug. The data set used to estimate switch rates and resource utilization is illustrated in Figure 2. Of the 5766 eligible patients, 62% were female; the mean age was 37 years, mean body mass index 22.2, and mean CD4 count 100 cells/mm3.
As noted above, values for model parameters were drawn from both the primary data set and from the literature. Parameter values are shown in Table 2.
Transition Probabilities, Event Numbers, Final States, and Scenario Costs
Transition probabilities for the model are presented in Table 3. As noted above, transition probabilities for the d4T scenario came from the primary data set, whereas transition probabilities for the TDF scenario were estimated from published sources.10,21,22
After eight 3-month cycles, 82.5% of the d4T scenario cohort remained on d4T, 16.6% had switched to AZT, and 0.8% had died. The cohort had also experienced 414 events that did not lead to a drug change, as shown in Table 4. Table 4 also shows that in the TDF scenario 2.5% of the cohort switched from TDF to AZT over the study period, whereas 97.5% remained on TDF.
The total cost for ARVs and toxicity management in the d4T scenario was $1,067,408. ARVs comprised 83% of the total and events 17%. The total cost of the TDF scenario was $1,323,445 of which ARVs comprised 96% and events 4%. On an annual basis, the TDF scenario costs $128 more per patient than the d4T scenario. The purchase price for a year's supply of TDF, $204, is $161 more than the annual purchase price of d4T. The annual cost of the TDF scenario in this analysis is thus $33 less than the difference in drug costs alone, indicating that, at a TDF price of $17/month, savings on toxicity management costs offset roughly 20% of the higher cost of TDF.
Had our cohort experienced no adverse events, each scenario would have generated a total of 2000 QALYs (2 years per patient × 1000 patients). When toxicity-related events were taken into account, the d4T scenario generated 1970.85 QALYs, indicating a loss of approximately 30 QALYs, as shown in Table 5. The TDF scenario generated 1999.39 QALYs, reflecting the small number of toxicity-related events ascribed to this scenario. The ICER for the TDF scenario was $9007 per QALY gained. This is far less than 3 times South Africa's annual per capita GDP of $5632,23 the rule of thumb for defining cost-effectiveness. Switching to TDF can thus be regarded as a cost-effective intervention. The ICER is about 60% more than per capita GDP, however, and so cannot be labeled very cost effective at a TDF price of $17.
Threshold Prices for TDF
We estimated 2 threshold prices for TDF: the price at which the change would be cost neutral for the Department of Health and the price at which the change would be very cost effective based on an ICER equal to the country's per capita GDP. The cost-neutral price was $6.17/month, approximately 36% of the price used in the model. The price at which the ICER would equal per capita GDP was $12.94, more than three fourths of the modeled price. It would thus take only a modest reduction in the price of TDF to make the switch qualify as very cost effective, even though a larger reduction would be required to achieve budget neutrality.
In sensitivity analysis, we explored the effect of d4T-related LTFU on overall costs and cost-effectiveness. Overall LTFU of TLC patients who initiate a d4T regimen in the first 2 years on treatment is 19.1%. In the baseline case presented above, transition probabilities to state 4 were assumed to be 0%. In Table 6, results that attribute to d4T-related toxicities 5%, 10%, and 20% of overall LTFU are presented. LTFU is assumed to take place in each cycle in proportion to the total number of other events in that cycle.
As might be expected, both the total cost of the d4T scenario and the cost-neutral price of TDF decrease with the d4T-attributable LTFU rate. This is because the cost of state 4 is $0-patients who are lost to follow-up cease to utilize health care provider resources and thus incur no further costs. In contrast, the cost-effective price of TDF increases with the proportion of patients in state 4 because patients who discontinue treatment also lose an average of nearly half a QALY each in the first 2 years after ART initiation.
Using data on the frequency and severity of d4T-related toxicity events from one of South Africa's largest AIDS treatment providers, we found that the reduction in toxicity management costs stemming from a switch from d4T to TDF in first-line regimens would offset about 20% of the higher price of TDF. Even under the conservative assumptions of no d4T-related LTFU and a TDF price of $17.00/month, the switch to TDF would be cost effective on the basis of the incremental cost per QALY gained, using the standard of 3 times GDP per capita per QALY. If we assume that 10%-20% of observed LTFU results from d4T side effects, then changing to TDF would become very cost effective (GDP per capita per QALY) at a price of $14-$16/month, very close to the baseline price used in the model. For the switch to be cost neutral for the South African government, however, a considerably lower price for TDF would be required in the neighborhood of $6/month.
Despite the high rate of d4T-related events and drug changes observed in the study cohort, the cost of the d4T scenario continued to be driven largely by the price of ARVs and not by the cost of managing toxicities. The predominance of drug costs is even greater in the TDF scenario because event rates are so low. These proportions explain why the price of TDF must fall so far for the new regimen to become cost neutral: Even if all costs of managing d4T-related toxicities were eliminated, the difference in drug prices would continue to dominate the comparison.
Several other responses to the high rate of d4T-related toxicities can also be considered by policy makers and should be investigated in future cost-effectiveness analyses. Most simply, d4T could be replaced by AZT rather than TDF. The unit cost estimates reported in Table 2c, however, argue against this strategy because the price of AZT is not substantially less than that of TDF, and AZT entails its own set of side effects, particularly during a patient's first 6 months. Alternatively, patients could be initiated on d4T and then routinely switched to AZT after 6 months. This strategy has been proposed by clinicians in South Africa, but to our knowledge, it has not yet been evaluated in the literature. The recent recommendation by the World Health Organization that the dosage of d4T be lowered from 40 to 30 mg should also be considered. Future sensitivity analysis could also examine how differences in laboratory testing capacity, for example between urban and rural areas, might affect both the costs of managing serious toxicities, such as lactic acidosis, and the outcomes for patients who suffer them.
The study has a number of limitations stemming from data availability and quality. The major data limitations in the d4T scenario are as follows: (1) the absence of information about the severity of toxicity events; (2) the inconsistent recording of reasons for changing from d4T to a new drug; (3) the lack of information about inpatient care in the outpatient clinic's medical records; (4) the incomplete information about why some patients were lost to follow-up; (5) the lack of information about side effects suffered by patients who did not seek care for them; and (6) the relatively short average duration of most patients on ART, which necessitated terminating the model after 24 months. In the first 3 cases, we either used parameters from the literature to make up for the gap in the primary data set or relied on the judgment of clinicians working at the study site. It is difficult to know in what direction and to what extent each of these limitations biases our findings. We addressed the fourth limitation in sensitivity analysis; the fifth was not addressed but would tend to bias the results in favor of d4T by excluding events that may have imposed quality-of-life costs on patients but were not captured in the analysis. The sixth limitation can only be addressed in coming years as the average duration of patients on ART rises. For the TDF scenario, we had no primary data available at all and relied entirely on the limited information available in the published literature.
There are also several limitations related to the model itself. First, as implied above, the relatively short duration of the observation period, 24 months, means that the longer term costs and effects of the 2 regimens were missed. Over time, the loss of QALYs caused by death and LTFU would increase, causing the estimated cost-effective price of TDF to rise. Both d4T- and TDF-related toxicity patterns could also change substantially. Second, the model excluded “second-generation” effects of toxicities, drug changes, and dosing, such as the impact of d4T-related toxicities on adherence and thus the development of resistance; how an early change to AZT might affect the durability of second-line regimens; or the effect of once-a-day dosing of TDF on long-term adherence. Third, the model also excluded side effects of the new drug, AZT, on the assumption that these would be identical regardless of whether the original drug was d4T or TDF. Fourth, the model conservatively assumed that transition probabilities among patients who had already experienced an event not leading to a drug change are the same as among patients who experienced no events. It is likely that the transition probabilities for this group of patients are in fact somewhat higher.
Another source of uncertainty in the analysis was the unit costs of ARVs. We used the drug prices paid by the study site in November 2007. Information obtained from other ART sites in South Africa, however, makes clear that there is wide variation in drug prices, both among sites and over time. Drug prices available to different countries at different points in time also vary widely24 and may be influenced by the actions of international organizations such as the World Health Organization, advocacy groups such as the Clinton Foundation HIV/AIDS Initiative, and historical pricing experience. The price of d4T, a relatively “old” ARV, for example, has already fallen significantly and may not decline further, but the price of TDF, a newer drug, may continue to decrease. For obvious reasons, the results presented here are sensitive to changes in the prices of d4T, AZT, and TDF and should thus be interpreted with caution.
Finally, many of the estimates of QALY losses for specific toxicity-related events were obtained from unrelated events with similar symptoms, transferred from research in other countries, or loosely estimated from clinical judgment. As with some of the data-related limitations mentioned above, we have no way of knowing whether we over- or underestimated QALY losses or in what direction the results are biased.
Despite these limitations, the analysis provides policy makers and program managers in resource-constrained settings a framework for considering the change from d4T to TDF and some initial estimates of the potential costs and cost-effectiveness of making this change. We conclude that replacing d4T with TDF at current prices is cost effective (cost/QALY gained < 3 × per capita GDP) and would become very cost effective (cost/QALY gained <1 × per capita GDP) with only a modest reduction in the price of TDF. The change would not be cost neutral for the South African Department of Health, however, without a more substantial decrease in the price of TDF. These conclusions should be interpreted with caution due to the large number of data gaps and model limitations described above. Better data about the severity of toxicities, reasons for LTFU, and long-term consequences of both d4T and TDF would improve the accuracy of the findings.
The funder had no involvement in study design, collection analysis or interpretation of data, writing of this paper, or decision to submit it for publication. We thank the several colleagues in South Africa and the United States who reviewed the analysis, helped determine model parameters, and provided other input, and in particular, Francesca Conradie of the Clinical HIV Research Unit, Pappie Majuba of Themba Lethu Clinic, David Hamer of Boston University, and Daniel Westreich of The University of North Carolina. We also thank Themba Lethu Clinic, Helen Joseph Hospital, and the Gauteng Province Department of Health for allowing us to conduct the study at their site and for their ongoing support of research efforts. Finally, we thank Right to Care for its efforts in facilitating this study.
1. Renaud-Thery F, Nguimfack BD, Vitoria M, et al. Use of antiretroviral therapy in resource-limited countries in 2006: distribution and uptake of first- and second-line regimens. AIDS
. 2007;21(Suppl 4):S89-S95.
2. Subbaraman R, Chaguturu SK, Mayer KH, et al. Adverse effects of highly active antiretroviral therapy in developing countries. Clin Infect Dis
3. Wong EB, Murdoch D, Wing J, et al. High Rates of Non-Fatal Toxicities in a 24-Month Cohort Receiving Publicly Funded HAART in South Africa: Proceedings of the 14th Conference on Retroviruses and Opportunistic Infections (CROI), Los Angeles, CA, 25-28 February 2007
. Abstract 795.
4. Hawkins C, Achenbach C, Fryda W, et al. Antiretroviral durability and tolerability in HIV-infected adults living in urban Kenya. J Acquir Immune Defic Syndr
5. Toure S, Seyler C, Messou E, et al. Main reasons of modification of the first-line antiretroviral regimen in adult patients who initiated HAART in the International Family Health Initiative (ACONDA/ISPED/EGPAF) in Abidjan, Cote d'Ivoire. Paper presented at: 3rd HIV/AIDS Implementers' Meeting; June 16-19, 2007; Kigali, Rwanda. Abstract 1215.
6. Bolhaar MG, Karstaedt AS. A high incidence of lactic acidosis and symptomatic hyperlactatemia in women receiving highly active antiretroviral therapy in Soweto, South Africa. Clin Infect Dis
7. Osler M, Stead D, Rebe K, et al. Severe Hyperlactatemia Complicating ART With Stavudine First-Line Therapy in South Africa: Incidence, Risk Factors, and Outcomes: Proceedings of the 14th Conference on Retroviruses and Opportunistic Infections (CROI), Los Angeles, CA, 25-28 February 2007
. Abstract 792.
8. Wester CW, Okezie OA, Thomas AM, et al. Higher-than-expected rates of lactic acidosis among highly active antiretroviral therapy treated women in Botswana: preliminary results from a large randomized clinical trial. J Acquir Immune Defic Syndr
9. Mutimura E, Stewart A, Rheeder P, et al. Metabolic function and the prevalence of lipodystrophy in a population of HIV-infected African subjects receiving highly active antiretroviral therapy. J Acquir Immune Defic Syndr
10. Gallant JE, Staszewski S, Pozniak AL, et al. Efficacy and safety of tenofovir DF vs stavudine in combination therapy in antiretroviral-naive patients: a 3-year randomized trial. JAMA
11. Kiguba R, Byakika-Tusiime J, Karamagi C, et al. Discontinuation and modification of highly active antiretroviral therapy in HIV-infected Ugandans: prevalence and associated factors. J Acquir Immune Defic Syndr
12. Rosen S, Fox MP, Gill CJ. Patient retention in antiretroviral therapy programs in sub-Saharan Africa: a systematic review. PLoS Med
13. Aspen Pharmacare. Launch of Viread and Truvada. Press release. April 2007.
14. Alcorn K. First-line treatment choices proving challenging for African ART programmes. Aidsmap News
. June 27, 2007.
15. Alcorn K. CROI: high burden of side-effects from d4T leads to WHO dose reduction recommendation. Aidsmap News
. March 21, 2007.
16. Smith CJ, Olsen CH, Mocroft A, et al. The role of antiretroviral therapy in the incidence of pancreatitis in HIV-positive individuals in the EuroSIDA study. AIDS
17. Briggs A, Schulpher M, Claxton K. Decision Modelling for Health Economic Evaluation
. Oxford, England:Oxford University Press; 2006.
18. Drummond MF, Sculpher M, Torrance GW, et al. Methods for the Economic Evaluation of Health Care Programme
. 3rd ed. Oxford, England: Oxford University Press; 2005.
19. Harling G, Wood R. The evolving cost of HIV in South Africa: changes in health care cost with duration on antiretroviral therapy for public sector patients. J Acquir Immune Defic Syndr
20. Wolf LL, Ricketts P, Freedberg K, et al. The cost-effectiveness of antiretroviral therapy for treating HIV disease in the Caribbean. J Acquir Immune Defic Syndr
21. Amoroso A, Sheneberger R, Edozien A, et al. ART-Associated Toxicities Leading to a Switch in Medication: Experience in Uganda, Kenya, and Zambia: Proceedings of the 14th Conference on Retroviruses and Opportunistic Infections (CROI), Los Angeles, CA, 25-28 February 2007
. Abstract 789.
23. World Bank. South Africa Data Profile, World Development Indicators Database, April 2007
. Washington, DC: The World Bank; 2007.
24. World Health Organisation. A Summary Report From the Global Price Reporting Mechanism on Antiretroviral Drugs, October 2007
. Geneva, Switzerland: World Health Organization; 2007. Available at: http://www.who.int/hiv/amds/GPRMsummaryReportOct07.pdf
. Accessed March 28, 2008.
25. Forna F, Liechty CA, Solberg P, et al. Clinical toxicity of highly active antiretroviral therapy in a home-based AIDS care program in rural Uganda. J Acquir Immune Defic Syndr
26. Rapoport J, Teres D, Zhao Y, et al. Length of stay data as a guide to hospital economic performance for ICU patients. Med Care
27. Ortegon MM, Redekop WK, Niessen LW. Cost-effectiveness of prevention and treatment of the diabetic foot: a Markov analysis. Diabetes Care. 2004;27:901-907.
28. Schackman BR, Scott CA, Sax PE, et al. Potential risks and benefits of HIV treatment simplification: a simulation model of a proposed clinical trial. Clin Infect Dis
29. Nyman JA, Barleen NA, Dowd BE, et al. Quality of life weights for the US population: self-reported health status and priority health conditions, by demographic characteristics. Med Care