Guidelines on when to start treatment in resource poor settings
Nathan Geffen, TAC
The appropriate CD4 count to initiate treatment is a topic of much research and discussion. In South Africa, the threshold in the public health system remains 200 cells/mm3, based on the Department of Health guidelines published in 2004. In the private sector, medical schemes are generally using a threshold of 350 cells/mm3, based on the Southern African HIV Clinicians Society Guidelines.  Other sub-Saharan African generally use a 200 threshold as well. Meanwhile, Europe and North America also initiate at first CD4 count below 350 cells/mm3.
The 350 cells/mm3 threshold is supported with evidence from several studies:
The When to Start Consortium, in an analysis of 15 cohort studies, found a greater than two times hazard ratio for progression to AIDS or death in people with CD4 counts from 100 to 200 cells/mm3 compared to people with CD4 counts of 201 to 300 cells/mm3 (HR: 2.21; 95%CI 1.91-2.56). There was also a significant difference in progression between the CD4 count strata of 151-250 cells/mm3 and 251-350 cells/mm3 (HR:1.71; 95%CI 1.43-2.04). (See HTB South April-June 2009.)  Also see Kitahata et al., which found a 1.69 [95%CI: 1.26-2.25] greater risk of death in patients who started treatment with a CD4 count between 350 and 500 cells/mm3 versus patients who deferred. 
The CIPRA HT 001 study was started in Port-au-Prince, Haiti in 2005. The 816 volunteers with CD4 counts between 200 and 350 cells/mm3 were randomised to either start HAART immediately or defer until their CD4 counts dropped to below 200 cells/mm3, the current standard of care in Haiti. The DSMB stopped the trial in May 2009 when it found clear evidence favouring the earlier treatment group. In the immediate group six people died versus 23 in the deferred group. Twice as many people (36 vs 18) contracted TB in the deferred group. These results, according to NIAID, are statistically significant. 
The START trial will answer the question as to whether treatment should be initiated at CD4 counts > 500 cells/mm3 or when it drops to below 350 cells/mm3. The trial began this year, but its estimated completion date is March 2015. In the meanwhile guideline developers need to make decisions based on the best, albeit incomplete, evidence available.
Francois Venter’s presentation at IAS2009
Francois Venter presented six issues affecting the discussion on when to start at a symposium at IAS2009. 
First, he presented the currently available evidence and ongoing trials to determine the appropriate CD4 count initiation threshold. He contrasted this with data presented by Matthias Egger at CROI 2007 that reviewed 176 sites from 2003 to 2005 in 42 countries that included 33,000 patients and showed when people actually are starting HAART. In developed countries, the average patient initiates treatment with a CD4 count of 150 to 200 cells/mm3. In Sub-Saharan Africa, it has increased from 50 cells/mm3 to about 100 cells/mm3. Venter pointed out that this was despite a large increase in testing in South Africa over the same period, with only about 20% of tested patients eligible for HAART according to current guidelines. In Venter’s programme in the inner city of Johannesburg, the average CD4 count at initiation is 106, despite 70% coverage and a massive escalation in testing. He stressed that retaining patients who have been tested in care is the key challenge as opposed to escalating testing.
Secondly, he explained that many people die waiting for treatment. For example, of 4570 patients followed up for at least one year in a cohort in Free State Province, South Africa, 53.2% died. Of these, 87% died before receiving HAART.  Venter presented data from Braitstein and colleagues showing that expedited care reduced mortality by 60%.
Third, he considered whether treatment should start earlier to reduce non-AIDS morbidity and mortality. He presented data from Andrew Phillips and colleagues that examined the risk of non-AIDS events in four studies: Flexible Initial Retrovirus Suppressive Therapies (FIRST) trial, Data Collection on Adverse Events of Anti-HIV Drugs study (DAD), Concerted Action on SeroConversion to AIDS and Death in Europe collaboration (CASCADE) and the Strategies for Management of Antiretroviral Therapy (SMART) trial.
The adjusted relative hazards for the effect of CD4 cell count on the risk of liver disease was significant in DAD, CASCADE and SMART. For non-AIDS cancer, it was significant in FIRST, DAD and CASCADE. For renal disease it was significant in FIRST and DAD (not measured for CASCADE) and for cardiovascular disease it was significant in CASCADE. In all comparisons, even when the results were not significant for a particular study, higher CD4 counts were correlated with less disease. Venter demonstrated that these results are relevant to South Africa by pointing out that there are overlapping epidemics of obesity and HIV in South Africa. He noted the high prevalence of hypertension and diabetes and suggested this put HIV-positive people in South Africa at high risk of dying from non-AIDS illnesses at low CD4 counts.
Fourth, he considered the effect of earlier HAART on opportunistic infections. There are high rates of opportunistic infections above CD4 counts of 200 cells/mm3 in resource poor settings. For example, an analysis by Badri and colleagues compared the six-monthly risk of death according to CD4 count and WHO stage. They found that the risk of death for WHO stages 1 and 2 was 3.5% (95%CI 1.4-7.1) for people with less than 200 cells/mm3, 2.8% (95%CI 1.3-5.3) for 200 to 350 cells/mm3 and 1.2% (0.5-2.3) for greater than 350 cells/mm3. For WHO stage 3 the death rates were 10.1% (95%CI 7.5-13.2) for <200 cells/mm3, 4·3% (95%CI 2.1-7.8) for 200 to 350 cells/mm3, and 4.9% (95%CI 2.3-9.1) for > 350 cells/mm3. For WHO stage 4, 22.2% (95%CI 17.9-27.1) for < 200 cells/mm3, 10.3% (95%CI 3.5-22.4) for 200 to 350 cells/mm3, and 13.8% (95%CI 4.1-32.6) for >350 cells/mm3. Interestingly, 52% of deaths took place in patients without AIDS. 
Venter pointed out the additional concerns in resource-poor settings about opportunistic infections. This has been well characterised for tuberculosis (see HTB South July to September 2009, pp. 22-28). But there are also concerns about greater risk of other bacterial (and fungal) infections including cryptococcal meningitis, pneumococcus, salmonella, wasting and malaria, even at higher CD4 counts. Cryptococcis is a particularly important concern. For example, a study by Olivier Lortholary and colleagues found that the mortality rate per 100 person-years was 15.3 [95%CI: 12.2-18.4] in the HAART era versus 63.8 [95%CI: 53.0-74.9] in the pre-HAART era for people diagnosed with this infection.  A study in Thailand by Jongwutiwes and colleagues retrospectively analysed 149 patients with cryptococcis. The cumulative 75% survival from relapse duration was over 10 months amongst those who did not receive HAART versus 42 months in those who did (p<0.01). 
Fifth, Venter considered maternal HAART and paediatric treatment. He examined the CHER results (discussed extensively in several issues of HTB ) which led to the WHO guidelines recommendation of universal treatment for all HIV-infected infants younger than 12 months. [13,14] Although mortality on the immediate treatment arm of CHER was as low as 4%, Venter pointed out that in 2007, only 8% of HIV-exposed infants were tested in the first two months of life. Using data from Malawi, he showed that the vast majority of children entered HIV care via hospital wards, as opposed to PMTCT follow-up, VCT and child health institutions, indicating systemic problems with getting HIV-infected children onto early treatment. Venter stated that it was better to prevent paediatric infection in the first place and that it was much easier to treat mothers than babies.
Venter considered the pros and cons of universal treatment for pregnant women, summarised in Table 1. He also emphasised the need for universal treatment for pregnant women with CD4 counts < 350 cells/mm3, citing data from Louise Kuhn shown in her plenary talk at IAS2009 that in the ZEBS study 84% of maternal deaths and 82% of postnatal infections occur in women with CD4 counts < 350 cells/mm3 (cf < 200 cells/mm3 at which 55% of maternal deaths and 47% of postnatal infections occur). 
Table 1: Pros and cons of universal treatment for pregnant women
Finally Venter considered the effect of earlier treatment on prevention. He pointed out that prevention programmes have had disappointing results and asked whether reducing viral load earlier might have a public health impact. He said there could be a convenient convergence of using HAART for both treatment and prevention purposes.
Venter concluded by pointing out that so long as stavudine is part of the standard care in Sub-Saharan Africa, earlier treatment would be harder to implement. The drug is in wide use. Venter presented data from Westreich and colleagues showing that stavudine is used by over 80% of Zambian, Mocambican and Tanzanian HAART patients and over 70% of Côte d’Ivoire ones. That study analysed a cohort of over 7,000 patients in Johannesburg. It found that for ongoing TB treatment at HAART initiation, the risk (adjusted hazard ratio) of a stavudine switch (compared to not being on TB treatment) was 3.18 (95%CI: 1.82-5.56) in the first 2 months of HAART, 2.51 (95%CI: 1.77-3.54) in months 3 to 6, and 1.19 (95%CI: 0.94-1.52) thereafter. For concurrent initiation of TB and HAART treatment, the risk was 6.60 (95% CI: 3.03-14.37) in the first 2 months, 1.88 (95%CI: 0.87-4.09) in months 3-6, and 1.07 (95%CI: 0.65-1.76) thereafter. There was no effect on the risk of stavudine substitution if patients were treated for TB after they had initiated HAART. 
In a study from a programme in Kigali Rwanda published by MSF researchers in September, stavudine side-effects were responsible for substantial switching. Out of 2190 adults followed up for a median of 1.5 years, stavudine was replaced in 175 patients (8.0%) for neuropathy, 69 (3.1%) for lactic acidosis and 157 (7.2%) for lipoatrophy. Lipoatrophy was the most frequent adverse event by three years of treatment. The authors concluded that stavudine is associated with significant long-term toxicity. They suggested stavudine-dose reduction, increased access to safer HAART regimens in low-income countries and close monitoring for those at risk. 
Data from IAS2009 on drug switching due to stavudine is presented in another article in this issue titled ARV Programme Results: Reports at IAS.
Cost-effectiveness of 350 cells/mm3 initiation
A study by Walensky and colleagues, published in the Annals of Internal Medicine, has modeled the cost-effectiveness of three options: (1) no treatment, (2) treatment initiated at CD4 count less than 250 cells/mm3 and (3) treatment initiated at CD4 count less than 350 cells/mm3. The study found that initiating HAART in South Africa at 350 cells/mm3 will reduce morbidity and mortality, improve long-term survival, and be cost-effective compared to initiation at 250 cells/mm3. The authors recommend that treatment guidelines should be changed to allow initiation when a patient’s CD4 count is below 350 cells/mm3. 
The objective of the study was to inform the crucial decision of when to start now, ie in the period before clinical trial results are published. Based on their results, the editors write, “Earlier antiretroviral therapy will probably prove to be superior in South Africa.”
In all scenarios, cotrimoxazole prophylaxis at a CD4 count < 500 cells/mm3 was modelled. The strategies were analysed over short-term (5 years) and life-time periods. Only direct HIV-associated uses of medical resources were considered.
The study used the CEPAC (Cost-Effectiveness of Preventing AIDS Complications) International model. All costs and life-expectancies were discounted at a rate of 3% per year. All monetary amounts were calculated using 2006 US dollars. The model has many input parameters, whose values were derived from clinical trials and observational studies in South Africa. Baseline at treatment parameters include mean age, ratio of men to women, mean CD4 count and distribution of the population over several HIV viral load strata. Parameters for modeling the onset of disease include monthly CD4 cell change (linked to viral load), monthly risk for severe opportunistic infection; efficacy of cotrimoxazole at reducing severe bacterial diseases, toxoplasmosis and PCP; monthly risk of HIV-related death and efficacy of HAART at viral load suppression by 48 weeks (broken down by first versus second-line treatment). In accordance with the South African treatment guidelines, the model provides for clinical assessments to occur every 3 months and CD4 and viral load testing every 6 months while receiving therapy.
The WHO considers an intervention cost-effective if it costs less than three times the per capita GDP to save one quality adjusted life-year. It considers an intervention very cost-effective if it costs less than the per capita GDP to save one quality adjusted life-year. However, the study examined years of life saved without adjusting for quality. Nevertheless, this is a minor difference and for practical purposes it is acceptable to directly compare the study’s findings with the WHO measures of cost-effectiveness. The cost-effectiveness of the study’s outputs was compared with the 2006 per capita GDP of $5,400 in South Africa.
The discounted per person lifetime cost for the strategies is $3,930 (no treatment), $12,730 (250 cells/mm3) and $13,620 (350 cells/mm3). Both treatment strategies would increase life-expectancy by at least 7.9 years over the no treatment strategy. Compared with no treatment, a treatment threshold CD4 count of 250 cells/mm3 would have an incremental cost-effectiveness ratio of $1,100 per life-year saved. And compared with a threshold of 250 cells/mm3, a threshold of 350 cells/mm3 would have an incremental cost-effectiveness ratio of $1,200 per life-year saved. (The incremental cost-effectiveness ratio is the additional cost of the intervention divided by the number of additional deaths prevented by the intervention.)
Table 2 summarises the clinical outcomes and costs over the next five years in the two treatment scenarios.
Table 2: Clinical outcomes and costs over the next five years in the treatment scenarios
Adapted from Walensky et al. (b = billion, m = million).
Using the WHO cost-effectiveness guidelines, the authors calculated that initiating treatment at a CD4 count of 350 cells/mm3 over the next five years should be used if the probability that a trial shows benefit for this threshold is 17% or greater.
The authors also conducted sensitivity analysis on the model’s lifetime projections and made the following findings:
1. Fewer than 39% of second-line patients initiated using a 350 cells/mm3 threshold would have to achieve viral suppression at 48 weeks to match the projected survival rate from using a threshold of 250 cells/mm3. This is a 32% relative decrease from the base case.
2. Discontinuation of care in the 350 cells/mm3 group (eg due to treatment fatigue) would have to be more than 19% to decrease survival compared to the 250 cells/mm3 group.
3. Including a hypothetical third-line regimen increased life-expectancy in both treatment groups.
The cost-effectiveness study by Walensky and colleagues presents a strong economic argument for changing the HAART guidelines in South Africa to treat everyone with a CD4 count < 350 cells/mm3. Shortly before their study was published, NIAID announced the termination of the CIPRA HT 001, arguably settling the question that 350 is a better threshold than 200 cells/mm3 (although the cost-effectiveness study compared 350 to 250 and not 200 cells/mm3).
Venter’s analysis provides an excellent summary of the complex issues affecting when to start. His presentation shows that systemic problems are at least as important as changing guidelines. Even in developed countries, many patients start late, ie well within guideline recommendations, and the situation in resource-poor areas is much worse.
Interestingly when asked at IAS2009, one of the CIPRA HT 001 study trial researchers explained that the median CD4 count for the immediate group was 280 cells/mm3 (IQR: approximately 230-300) versus 170 cells/mm3 for the deferred group. It is notable that even in a trial, people did not start close to the stated threshold but the difference between early versus delayed treatment was still profound.
Changing the CD4 threshold will not by itself remedy this situation. It is a positive development that more people are getting tested earlier for HIV, but if there is then nothing to offer them until they are ill, they are likely to be lost to follow up, only to reappear with very low CD4 counts. Ensuring that patients who are tested early and then retained in care is the key challenge and probably requires system changes.
Ironically, if the START study demonstrates that immediate HAART is better than deferred, then this might prove to be a useful mechanism for retaining people who have tested HIV-positive in the health system.
This should not detract from the need for effective public communications programmes that encourage early testing, or provider-initiated testing. It is just that these programmes would be more effective at keeping patients in the system.
The move towards earlier treatment has become possible because of improved medicines. But the long-term side effects of stavudine make it much harder to determine an appropriate threshold in programmes that use this drug. Venter’s concern about the widespread use of stavudine in sub-Saharan Africa can be alleviated if steps are taken to gradually replace it with tenofovir. As reported in the Treatment Update section of the previous issue of HTB South, there is at least one generic manufacturer making a low-cost 3-in-1 once-daily tenofovir containing fixed-dose combination pill that is WHO prequalified and FDA tentatively approved. Other generic manufacturers are intending to do the same. Surely these should become the standard first-line regimens in sub-Saharan Africa?
Discussions are underway, and it seems likely that 350 CD4 as a starting point and tenofovir in first line to replace d4T, will be added to South African guidelines. However, both of these come with significant extra costs. So qualifications may apply, at least initially.
1. South African Department of Health. National antiretroviral treatment guidelines. 2004.
2. Southern African HIV Clinicians Society. Antiretroviral therapy in adults. 2008.
3. Sterne J et al. When should HIV-1-infected persons initiate ART? Collaborative analysis of HIV cohort studies. 16th CROI, Oral abstract 72LB.
4. Kitahata M et al. Effect of early versus deferred antiretroviral therapy for HIV on survival. NEJM. 2009 Apr 30;360(18)
5. NIAID. Questions and answers: The CIPRA HT 001 clinical trial. 2009 June 8.
6. Venter F. When to start in resource-poor areas. 5th IAS Conference on HIV Pathogenesis, Treatment and Prevention, 19-22 July 2009, Cape Town. Plenary session MOSY103.
7. Egger M. Outcomes of ART in resource-limited and industrialized countries. 14th CROI. Plenary abstract 62.
8. April MD et al. HIV testing rates and outcomes in a South African community, 2001-2006: implications for expanded screening policies. JAIDS 2009 Jul 1;51(3):310-6.
9. Fairall LR et al. Effectiveness of antiretroviral treatment in a South African program: a cohort study. Arch Intern Med. 2008 Jan 14;168(1):86-93.
10. Badri et al. Short-term risk of AIDS or death in people infected with HIV-1 before antiretroviral therapy in South Africa: a longitudinal study. Lancet. 2006 Oct 7;368(9543):1254-9.
11. Lortholary O et al. Long-term outcome of AIDS-associated cryptococcosis in the era of combination antiretroviral therapy. AIDS. 2006 Nov 14;20(17):2183-91.
12. Jongwutiwes U et al. Impact of antiretroviral therapy on the relapse of cryptococcosis and survival of HIV-infected patients with cryptococcal infection. Curr HIV Res. 2007 May;5(3):355-60.
13. Violari et al. Early antiretroviral therapy and mortality among HIV-infected infants. N Engl J Med. 2008 Nov 20;359(21):2233-44.
14. WHO. Antiretroviral therapy of HIV infection in infants and children: towards universal access. 2008.
15. Kuhn L. Prevention of mother-to-child HIV transmission (PMTCT). 5th IAS Conference on HIV Pathogenesis, Treatment and Prevention, 19-22 July 2009, Cape Town. Plenary session MOPL103.
16. Westreich DJ et al. 2009. Tuberculosis treatment and risk of stavudine substitution in first-line antiretroviral therapy, Clin Infect Dis. Jun 1;48(11):1617-23
17. Walensky et al. When to start antiretroviral therapy in resource-limited settings. Ann Intern Med. 2009 Aug 4;151(3):157-66.
18. van Griensven J et al. 2009. Stavudine- and nevirapine-related drug toxicity while on generic fixed-dose antiretroviral treatment: incidence, timing and risk factors in a three-year cohort in Kigali, Rwanda. Trans R Soc Trop Med Hyg. Sep 2.