Nathan Geffen, TAC

IAS2009 was distinguished by the large number of reports from operational HAART programmes across the world. A search of the conference website using the phrase “ART cohort” returns nearly 170 abstracts, the vast majority of them from resource-poor settings. At least 25 of these report cohorts with more than a 1,000 people in poor countries. And five of these include more than 10,000 people. Many report on children and several now include five-year data. Most report good or reasonable retention rates. Here is a summary of some of the research presented:

• Two Cambodian facilities treating over 1,160 children reported 91% survival at 36 months. Of 4,150 people receiving HAART in these facilities, 86%, 84% and 81% were alive and in care at 2, 3 and 5 years respectively. [1-2]
• In Khayeltisha South Africa, an analysis of 3,595 adults enrolled on HAART over five years showed that the cumulative proportion of patients remaining in care at 54 months was 78%. Early loss to follow-up was higher in women who had been exposed to PMTCT, in patients enrolled later in the programme and in patients initiated on efavirenz. Interestingly, diagnosis of AIDS and lower weight were protective. Increased risk for late loss to follow-up was associated with younger age, exposure to PMTCT and initiation on efavirenz. In a similar study, 87% of children on this programme remained in care after five years. [3-4]
• The Lesotho government’s programme, with extensive support from Medecins Sans Frontieres (MSF), reported a 24 month retention rate of 80%. This programme comprises 14 clinics and one hospital. The number of people enrolled is 11,367 (5% children). Of these 4,347 initiated HAART (6.5% children). Interestingly, the programme also uses a three-in-one once-daily tenofovir based generic antiretroviral. [5]
• In a Malawian MSF-assisted programme, over 2,300 adults were followed up for over 1,500 person-years. There were 277 (12%) deaths, 206 (74%) of which occurred during the first three months of HAART. The cumulative mortality rate was 13 per 100 person years. A common theme through many of the IAS treatment cohort reports was early mortality associated with patients presenting with low CD4 counts and advanced disease. The extent to which patients are enrolling late was shown by a study of PEPFAR programmes. This study included PEPFAR cohort-level aggregate indicator data on the median CD4 count at HAART initiation from 248 sites in eight sub-Saharan African countries –Ethiopia, Kenya, Lesotho, Mozambique, Nigeria, Rwanda, South Africa and Tanzania– where a total of 1,438 cohorts (defined as comprised of patients who initiated HAART at a given site each quarter) representing 103,124 patients initiated HAART. Median baseline CD4 count for all of the cohorts was 135 cells/mm³ (IQR: 109-175). This had increased steadily from 115 to 140 cells/mm³ over the study period. [6-7]
• Solidarmed, a church group running HAART facilities in several countries, reported over 65% retention after 24 months at its facilities in Tanzania, Zimbabwe, Lesotho and Mozambique, comprising over 4,300 patients. [8]
• Absolute Return for Kids (ARK) reported the results of 2,332 children enrolled on HAART at seven rural and 37 urban sites. Patients were divided into (1) urban, (2) rural and (3) rural attending urban facility. There were 1,727; 228 and 377 people in groups 1, 2 and 3 respectively. After 18 months the probability of death was 3.0% (CI: 2.2-4.1%), 7.7% (CI:4.5-13.0%) and 3.1% (CI:1.7-5.6%) in the groups respectively. Loss to follow-up probabilities were 9.9% (CI: 8.3-11.8%), 5.8% (CI:2.7-12.2%) and 15.0% (CI:11.2-20.0%) respectively. [9] • In Gugulethu, South Africa of 3,546 people eligible for HAART, 11 (6.5%) children and 442 (13.1%) adults died while waiting for treatment. Death while awaiting HAART initiation is another common theme dealt with in more detail below. An analysis of 2,906 adults and children who started HAART found that virological failure averaged 6.0% per year for children and 3.5% per year for adults. Adult on-treatment mortality was 12.9% at 5 years with 209 of 249 deaths (84%) within the first year. Kaplan-Meier estimates for loss-to-follow-up were 8.5% for children and 16.5% for adults at 3 years. Retention in care at 3 years was 89.3% for children and 74.1% for adults. [10]
• A total of 9,460 HIV-infected children from 11 different cohorts in South Africa, Malawi, Mozambique and Zimbabwe were included in one analysis. Overall, 5,228 (55.3%) were younger than five years. Mortality at one year after initiating HAART ranged from 0.7% to 17.7% in children aged < 5 years and from 0% to 8.9% in children ≥5 years. Overall mortality, including estimating mortality in children lost to follow-up, was calculated to be 8.8% [95%CI: 5.5-12.1. Loss to follow-up ranged from 1.9% to 33.7% (overall 13.3%). [11]
• Paediatric data from the Thai government’s National Access to HAART for People Living with HIV/AIDS (NAPHA) programme were presented. This cohort consists of 3,409 treatment-naive children enrolled before January 2006. Median age at baseline was 7 years (IQR: 5-9). Median baseline CD4 was 5% (IQR: 2-13). Weight-for-age z-score was -2.0 (IQR: -2.6 to -1.4). At the end of the follow-up period (median of 1.7 years; IQR: 1-2.5), 9% of children had died, 10% were lost to follow-up (> 9 months since last appointment), 1% had stopped HAART and 80% were still on treatment and in the programme. Of the deaths, 90% were due to AIDS. The overall death rate was 5.2 per 100 person years (95%CI: 4.6-5.8) and the AIDS death rate in the first 90 days was 16.7 per 100 person years (95%CI: 14.0-19.7). The probability of a child surviving to one year was 92.7% (95% CI: 91.7, 93.5) and to five years it was 88% (95%CI: 85.9-89.8). Age, weight-for-age z score, HIV stage, baseline CD4 and attending a community as opposed to a district hospital were significantly associated with mortality. [12]
• In perhaps the oldest programme in the developing world, the São Paulo State STD/ AIDS Program, Brazil, patient records are available from 1985 and treatment, albeit with fewer than three drugs, began in 1991. An analysis of 4,191 people including 14,690 HAART prescriptions showed that by 2005, 70% of patients were still in the programme. There were 518 deaths and 726 people were lost to follow-up. [13] • A study from Zambia compared a (1) rural government, a (2) rural mission facility and an (3) urban government facility, the latter two sponsored by PEPFAR. The first has more than 300, the second more than 1,100 and the third more than 1,700 patients. Medical records from each facility of the last 150 patients to initiate HAART in 2006 were compared. At 12 months, 67%, 84% and 75% of patients were still in care in facilities 1, 2 and 3 respectively. Despite having the best results, baseline CD4 count was lower in facility 2 (133 vs 117 vs 125 respectively). But the cost of care per patient of facility 2 was also substantially higher than the other two facilities. Cost per 150 patients in the sample were $24,940, $55,721, $35,877 respectively. [14]
• A study of an MSF project in Chiradzulu district Malawi showed the effectiveness of decentralised treatment. From March 2001 to December 2008, over 17,000 people initiated HAART, of whom 70% were followed in decentralised care. In 2003/4 6-month survival was about 84% in both centralised and decentralised care. This rose to about 94% in 2007/8. The percentage of patients with CD4 counts available rose from 53% in 2003/4 to 86% in 2007/8 (rising from 49% to 86% in decentralised and from 56% to 83% in centralised care). In decentralised care, the proportion of patients starting HAART with advanced disease declined over time, from 77% of adults in stage 3 or 4 in 2003/4 to 44% in 2007/8. Similarly for children this declined from 61% in 2003/4 to 42% in 2007/8. [15]
• Osler and colleagues presented consolidated results of the Western Cape HAART programme. The number of sites has risen from 35 in 2004/5 to 78 in 2008/9. The number of patients newly enrolled on HAART were 5,929 in 2004/5, 10,169 in 2005/6, 12,424 in 2006/7, 13,820 in 2007/8 and 20,035 in 2008/9. The number of patients remaining in care were 7,661 in 2004/5, 16,343 in 2005/6, 26,111 in 2006/7, 37,435 in 2007/8 and 54,703 in 2008/9. The proportion of adults starting HAART with a CD4 count < 50 cells/mm³ decreased from 52.9% in 2001 to 16.2% in 2008. Using the Actuarial Society of South Africa’s model, this study estimated that coverage had risen from 47% in 2004/5 to 85% in 2008/9

By 12 months, 9.6% of adults were lost to care. After six years on HAART, 72.4% of ARV-naive adults remained in care. With each year, retention in care continued to improve. By 12 months of treatment on HAART, 5.8% of children were lost to care. After five years, 86.0% of children remain in care. Just under 90% of adults and nearly 80% of children from 6 months to over four years in care were virally suppressed. By four years of HAART just under 10% of both adults and children are on second-line treatment.

Cost per patient dropped from an average of R7,504 in 2005/06 to R5,635 in 2007/08 . In the HAART budget, drugs comprised 40%, personnel 35% and laboratory services 11% of cost. Efavirenz, prescribed to 61% of first-line patients, was the most significant cost driver of HAART expenditure. [16]

Several studies showed problems with some aspects of HAART implementation. They are described here in more depth.

Retention in South Africa: a mixed picture

While the Western Cape data presented by Osler reports high retention, this was not the case for all large South African databases.

Morna Cornell and her colleagues presented data from the IeDEA-SA database, which contained about 10% of South African public sector adult patients from seven large sites by end of 2007. Loss-to-follow-up was defined as last contact greater than six months before the database was closed for the study. The cohort was followed up for another six months following database closure to get an accurate estimate of loss-to-follow-up. Four of the sites are in the Western Cape so many of the patients in the Osler study above must be in this cohort. [17]

The cohort consists of over 27,000 people, of whom over 18,000 (67%) are female. Median age is 34 years (IQR: 29-40). The number of patients initiated per year were: 1,152 (4% of the cohort) in 2002 and 2003, 4,614 (17%) in 2004, 6,838 (25%) in 2005, 7,669 (28%) in 2006 and 6,936 (25%) in 2007 (for a total of 27,209 patients). The median CD4 count at HAART initiation was 96 (IQR: 39-162). Over 7,000 patients, of the nearly 24,000 who had baseline CD4 data, initiated with a CD4 count less than 50 cells/mm³ Median HIV RNA level at HAART initiation was 4.84 log (IQR: 4.3-5.4) for the approximately 10,500 patients for whom results were available. Over 4,000 (40%) had viral loads > 5 log. WHO staging data was available for over 12,500 patients at HAART initiation. Of these 1,132 (9%) were in stage I, 1,508 (12%) were in stage II, 6,060 (48%) were in stage III and 3,936 (31%) in stage IV.

In each successive year of the programme, median CD4 increased (from 68 in 2002/03 to 104 in 2007), the percentage of stage IV patients decreased (to 27% of all patients by 2007), 12-month mortality improved (until 2006) but 12-month loss to follow-up increased over the same period and 12-month overall retention decreased with time on the programme (see Table 1), so that by four years, only 56% of patients were still in care. In multivariate analysis the decreasing 12-month mortality per calendar year was partially explained by increasing baseline CD4 count. The authors conclude that the increasing loss to follow-up reflects the major challenge of patient retention in large HAART programmes.

ADD TABLE

Table 1: Mortality, loss to follow-up and overall retention in Cornell et al.

Long waiting periods

 

Two disturbing studies from South Africa show the long time that patients have to wait for treatment in the Free State and Kwazulu-Natal provinces and the consequent high mortality.

Margaret May and her colleagues followed, until December 2007, 22,000 patients aged 15 years and older enrolled between May 2004 and December 2006 in the Free State HAART programme. They used the province’s HAART data supplemented with National Health Laboratory System CD4 data to determine that over 13,400 (61%) were eligible for HAART (CD4 count <200 cells/mm³. Of these over 3,450 (26%) died while waiting for treatment. The median waiting time to death was 3.1 months (IQR: 1.1-6.9). For those who started treatment, the median waiting time was 3.6 months (IQR: 2.0-6.4). Waiting times varied by a factor of two across districts and were halved in combined assessment and treatment clinics as opposed to when these services were separated. Waiting times were shorter in 2006 than in 2004, also shorter for women (possibly due to HAART referral from PMTCT facilities) and decreased with decreasing CD4 count (p for trend <0.001), but as the authors explain many of the patients with the lowest CD4 counts will drop out of survival analyses because they have died before they have started treatment and this has the potential to bias results. [18]

This data is broadly consistent with a study by Fairall and colleagues published in January 2008 which found that of 4,570 patients followed up for at least one year in the Free State, 53% died. Of these, 87% died before receiving HAART. [19]

Similarly concerning was a study by Ingrid Bassett and colleagues. Their study had two objectives: to evaluate rates of HAART initiation within 12 months of a positive HIV test in Durban, South Africa and to identify baseline factors that predict failure to be on HAART at one year. They examined two pay-for-care semi-private, but government subsidised, facilities in Durban (McCords and Marianhill Hospitals). Their cohort included patients ≥ 18 years in outpatient departments who were English or Zulu speaking. They were enrolled prior to their HIV test. Pregnant women, people on stretchers and people with known HIV-positive status were excluded. Patients were enrolled from November 2006 to October 2008 and followed up through June 2009. [20]

After screening 3,401 patients, 1,467 HIV-positive people met the inclusion criteria for the study (HIV prevalence was 54%). In the final cohort, 54% were female, the median age was 34 years and median follow-up time was 11.4 months. Only 607 (41%) had CD4 counts within 90 days of testing positive. Of these, 368 (61%) were eligible for HAART (CD4 count < 200 cells/mm³). An additional 19 were clinically eligible for HAART. Only 213 (55%) were known to have initiated HAART within 12 months. Median time from CD4 count to ART initiation was 3.5 months (IQR: 2.0-5.5). There were 68 known deaths (18% of the HAART-eligible cohort) of which 52 (76%) occurred prior to starting treatment. Adjusting for age, CD4 count, having work outside home and distance from clinic, patients who did not start treatment were more likely to be male (RR: 1.5; 95%CI: 1.1-2.1) and to not have an HIV-positive family member or friend (RR: 5.1 95%CI: 1.8-14.9).

The authors point out several limitations of their work: the sites may not be representative of public sector hospitals in South Africa, 30% of pre-HAART patients were unreachable and they likely underestimated mortality and HAART initiation that occurred at non-study sites.

Stavudine

Some studies considered the stavudine side-effects on regimen switching and retention. Barbara Castelnuovo and colleagues prospectively followed 559 people who initiated HAART from April 2004 to April 2005 in Kampala, Uganda. At baseline the cohort was 70% female, median age was 38 years (IQR: 33-44), median CD4 count was 98 cells/mm³ (IQR: 21-163) and viral load was 5.4 log (IQR: 5.0-5.8). The primary endpoint was the substitution of at least one drug included in theinitial combination for any reason.

Most of the cohort (413, 74%) were prescribed stavudine, lamivudine and nevirapine as a fixed combination, while the remainder received zidovudine, lamivudine (fixed combination) and efavirenz. [21]

There were 148 (27%) patients with at least one treatment change (incidence rate 14.3 per 100 person years; 95%CI: 12.2-16.9), of whom 91 changed because of toxicity. Stavudine accounted for 76 (84%) of the treatment changes followed by AZT (10, 11%).

In a multivariate analysis, being male, and treatment with the zidovudine/lamivudine/efavirenz regimen were associated with protection from any drug switches (RR 0.47 [95%CI: 0.30-0.74] and 0.38 [95%CI: 0.23-0.64]), respectively. The analysis was repeated to investigate risk factors for drug switches due to toxicities. Being male and on efavirenz were then found to be the only protective factors.

For the first 18 months, there was no difference between men and women, but drug switching in women increased in the second half of the observation period. There were significantly more women with lactic acidosis (25 versus 10, p<0.05) and lipodystrophy (26 versus 0 , p<0.05) due to stavudine that led to drug switching.

The authors concluded that the majority of the treatment changes were due to stavudine-related toxicity and that long-term stavudine use is less well tolerated in women. On the positive side, patients that are switched due to toxicity have a similar virologic outcome compared to patients on a stable regimen.

However, for children in a Malawian study, stavudine-containing regimens were relatively well tolerated. Kabue and colleagues from the Baylor College of Medicine reviewed patient records of 1,434 children who initiated HAART using fixed dose stavudine, lamivudine and nevirapine from November 2004 to October 2008. The mean and median ages at HAART initiation were 5 and 3 years respectively. They were followed up for a mean of 1.3 years. The first-line regimen was discontinued by 28 (2%) children due to adverse events (17 nevirapine-related of which 10 rash, 5 Steven Johnson Syndrome and two hepatitis; 11 stavudine-related of which 5 pancreatitis, 4 lactic acidosis and 2 peripheral neuropathy). The median time to an adverse event was two months (IQR: 10 days to 28 months). [22]

COMMENT

These studies show the substantial benefit of HAART in reducing AIDS mortality in developing countries. The Cambodian five-year results are particularly impressive. A number of these studies report improving CD4 counts and earlier stage of disease at initiation. This is possibly related to the decentralisation of HAART services.

However, even better results can be achieved if a number of problems are resolved. The most problematic is the long waiting time between enrollment and initiating HAART. This accounts for far more lives lost than any other shortcoming in programmes, including loss to follow-up after treatment begins, inferior drug regimens, lack of second and third-line regimens and lack of diagnostics. While the data on waiting time is from three South African areas [10, 18. 20], it is implausible that these are unusual either in South Africa or beyond. While most of the studies presented in this article present a positive view of HAART programmes, it is worth investigating if intention-to-treat analyses (ie including all patients who have enrolled in HAART programmes, not only those who have started treatment) would have much worse results.

Long-term patient retention is another challenge. While Morna Cornell’s group’s data is worrying, it appears to be inconsistent with the high retention of patients in Osler’s Western Cape data, even though more than half of Cornell’s sites are in the Western Cape. There are several plausible possibilities: the Western Cape programme is probably much better run than other provinces, patients that Cornell’s group has counted as lost to follow-up might have transferred to other facilities or –least likely– the Western Cape provincial data might be inaccurate. More research is needed to answer this.

But more critically, the data that Osler and Cornell have reported should be reported regularly as part of the National Department of Health’s monitoring and evaluation of the HAART programme. May and Bassett also both report missing data. This points to the urgent need to improve the monitoring and evaluation of the South African public sector programme, which does not include patients who have died prior to commencing treatment. While the Western Cape programme appears to have sound data, this is not the case in the other provinces. Unfortunately the monitoring and evaluation of the state’s programme is very poor currently, although the Department has committed, in a meeting with the newly formed Budget and Expenditure Monitoring Forum, to rectifying this soon.

The stavudine data suggests that while it is the least tolerated of the currently used ARVs, at least in adults, it still has an important role because of its pervasive use and low price. It should be phased out, but gradually and not at the expense of putting new patients onto treatment. Where substitution with tenofovir is possible, it appears that women should be prioritised. The use of 30mg instead of 40mg dosing and lactic acid testing of patients with symptoms of hyperlactatemia in some facilities appear to have made stavudine more manageable, but more data is needed on this. However, as developing country guidelines slowly change to earlier treatment, it will become increasingly important to phase out stavudine.

References

All references are for the 5th IAS Conference on HIV Pathogenesis, Treatment and Prevention, 19-22 July, Cape Town. 1. Isaakidis P et al. 2009. High survival and treatment success sustained after up to three years of ART for children in Cambodia. Oral abstract MOAB102.
http://www.ias2009.org/pag/Abstracts.aspx?SID=2420&AID=850
2. Raguenaud ME et al. 2009. Good ART patient outcomes and survival achieved in a six-year HIV/AIDS program in Cambodia. Poster abstract WEPED186.
http://www.ias2009.org/pag/Abstracts.aspx?AID=842
3. Van Cutsem G et al. 2009. Loss to follow-up and associated factors at different durations of antiretroviral therapy in Khayelitsha, South Africa. Poster abstract WEPEB284.
http://www.ias2009.org/pag/Abstracts.aspx?AID=2882
4. Colvin CJ et al. 2009. Paediatric outcomes after five years on ART in Khayelitsha Township, South Africa. Abstract poster CDB109.
http://www.ias2009.org/pag/Abstracts.aspx?AID=3035
5. Cohen R et al. 2009. Nurse-driven, community-supported HIV/AIDS care and treatment: 2 year antiretroviral treatment outcomes from a primary care level programme in rural Lesotho. Oral abstract MOAD102.
http://www.ias2009.org/pag/Abstracts.aspx?AID=3133
6. Zachariah R et al. 2009. Very early mortality in patients starting antiretroviral treatment at primary health centres in rural Malawi. Poster abstract WEPED185.
http://www.ias2009.org/pag/Abstracts.aspx?AID=805
7. Nash D et al 2009. Program-level determinants of low CD4 count ART initiation in cohorts of persons aged ≥ 6 years initiating ART in 8 sub-Saharan African countries. Oral abstract WEAD103.
http://www.ias2009.org/pag/Abstracts.aspx?AID=1882
8. Ehmer J et al. 2009. Characteristics and outcomes of HIV-infected patients starting cART in rural treatment programmes in four sub-Saharan African countries. Poster abstract MOPED045.
http://www.ias2009.org/pag/Abstracts.aspx?AID=2615
9. Fatti G et al. 2009. Antiretroviral therapy outcomes in rural and urban children attending public health facilities in South Africa. Poster abstract MOPEB077.
http://www.ias2009.org/pag/Abstracts.aspx?AID=1170
10. Kaplan R et al. 2009. Long term programmatic outcomes for adults and children at a primary health care antiretroviral clinic in South Africa. Oral abstract MOAD105.
http://www.ias2009.org/pag/Abstracts.aspx?SID=2432&AID=2901
11. Fenner L et al 2009. Mortality, loss to follow-up and transfer-out in paediatric ART programmes in Southern Africa. Poster abstract WEPEB276.
http://www.ias2009.org/pag/Abstracts.aspx?AID=2492
12. Carvalho Monteiro AL et al. 2009. Initial antiretroviral therapy (ART) in a 20-year observational cohort of patients followed at a reference center (São Paulo STD/AIDS reference center-CRT-DST/AIDS), in the city of São Paulo, Brazil, according to race/color. Abstract poster CDD142.
13. Rosen S et al. 2009. Outcomes and outpatient costs of different models of AIDS treatment delivery in Zambia. Poster abstract MOPED014.
http://www.ias2009.org/pag/Abstracts.aspx?AID=1374
14. McConnell M et al. 2009. Survival rates following expansion of the national pediatric antiretroviral treatment program, Thailand, 2000-2005. Oral abstract MOAB101.
http://www.ias2009.org/pag/Abstracts.aspx?SID=2420&AID=1715
15. Le Paih M et al. 2009. Scale up impact on care access and early program attrition: seven years of providing ART in rural Malawi. Oral abstract WEAD104.
http://www.ias2009.org/pag/Abstracts.aspx?SID=2430&AID=1754
16. Osler M et al. 2009. Despite increased burden due to scale-up, ART programme retention remains high in the Western Cape. Poster abstract MOPED048.
http://www.ias2009.org/pag/Abstracts.aspx?AID=3069
17. Cornell M et al. Temporal changes in patient mortality and programme retention among adult patients initiating antiretroviral therapy across South Africa, 2003-2007: the International Epidemiologic Databases to Evaluate AIDS (IeDEA) Southern Africa collaboration. Oral abstract MOPED041.
http://www.ias2009.org/pag/Abstracts.aspx?AID=879
18. May M et al. 2009. Determinants of waiting times for ART in the Free State Province, South Africa: prospective cohort study with retrospective database linkage. Late breaker poster LBPED05.
http://www.ias2009.org/pag/Abstracts.aspx?AID=3708
19. Fairall et al. 2008. Effectiveness of antiretroviral treatment in a South African program: a cohort study. Arch Intern Med. 168(1):86-93.
http://www.ncbi.nlm.nih.gov/pubmed/18195200
20. Bassett et al. 2009. Who starts ART in Durban, South Africa?…Not everyone who should! Oral abstract WEAD102.
http://www.ias2009.org/pag/Abstracts.aspx?SID=2430&AID=1921
21. Castelnuovo B et al. 2009. Stavudine toxicity in women is the main reason for treatment change in a 3 year prospective cohort of adult patients started on first line antiretroviral treatment (ART) in Uganda. Poster abstract CDB079.
http://www.ias2009.org/pag/Abstracts.aspx?AID=874
22. Kabue MM et al. 2009. Discontinuation of standard first-line antiretroviral therapy in a cohort of 1434 Malawian children. Poster abstract WEPED175.
http://www.ias2009.org/pag/Abstracts.aspx?AID=1064