Introduction

Most HIV-infected individuals on antiretroviral therapy (ART) in low/middle-income countries are treated following the WHO public health approach [1]: the public sector provides one standard first-line regimen, with alternative drug substitutions for anti-tuberculosis co-therapy/toxicity; when first-line failure occurs, the patient switches to a standard boosted-protease inhibitor (bPI)-based second-line regimen. Current WHO guidelines[2] define failure by virological (>5,000 copies/ml), immunological (CD4 below pre-therapy baseline; 50% fall from on-treatment peak; persistently <100 cells/mm³) or clinical criteria after 6 months on ART. Low-income countries differ in their ability to provide laboratory tests to identify first-line failure and support routine follow-up; if available at all, CD4 testing is most common with viral loads (VL) sometimes used to confirm clinical/immunological failure[2]. Routine virological monitoring is rarely available or feasible[3]. Such approaches differ markedly from individualised management in high-income countries, where routine VL monitoring is used to modify initial or subsequent therapy and many drugs are available.

WHO 2010 definition of clinical failure includes WHO 4 and certain WHO stage 3 conditions. Without VLs, it is strongly recommended that immunological criteria confirm clinical failure (noting moderate quality of evidence), but no CD4 threshold value is proposed. We therefore evaluated switches from first to second-line ART in the DART trial[4], [5], particularly considering the unique group randomised to clinically-driven monitoring (CDM) and managed without CD4 counts, but for whom CD4s (and some VLs) were available for retrospective analysis. The aims were to investigate the characteristics of immunological/clinical failures determined without routine VL monitoring and with or without routine real-time CD4 monitoring; and to identify optimal CD4 thresholds to confirm clinical failure and switch to second-line ART when VLs are unavailable.

Materials and Methods

Results

Patients monitored clinically without CD4 (or VL) results

1660 CDM participants initiated ART with median(IQR) CD4 86(31–139) cells/mm³ and were clinically monitored without CD4 counts for median 5 years. 314(19%) switched to bPI-containing second-line for first-line failure after median(IQR) 3.4(2.5–4.2) years on first-line (only 2 before 48 weeks when switching was discouraged, both at 46 weeks). In those who switched, median(IQR) pre-ART CD4 was 47(14–104) cells/mm³ and age at switch 39(34–44) years; 193(61%) were female.

223(13%) CDM participants had new/recurrent WHO 4 events accepted by the ERC after 44 weeks on ART: 187/223 (84%) switched to second-line, 14(6%) died on first-line before switching and 22(10%) had not switched before trial closure. The Kaplan-Meier median(IQR) time to switch after meeting failure criteria was 7(1–23) weeks. The most commonly reported reasons for delaying switch for >8 weeks/not switching were that the WHO 4 event was judged unrelated to ART failure by the clinician (45%) or because of drug-drug interactions between rifampicin and bPI (32%) (Table 1; n = 44). An additional 20 patients switched to second-line for WHO 4 events eventually judged not to meet pre-defined protocol criteria by the ERC, leading to a total 207/314 (66%) switches in CDM being for WHO 4 events.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Reasons for not switching when first met immunological/clinical criteria for switch in patients receiving and not receiving CD4 monitoring.

https://doi.org/10.1371/journal.pone.0057580.t001

In the same period, 70 multiple (within 60 days) and 392 single WHO 3 events were reported in participants not switching for WHO 4 events. Clinicians used clinical judgement to assess which of these events were likely first-line ART failure, leading to 30(10%) and 77(25%) of the 314 CDM switches being for multiple or single WHO 3 events respectively (Table 2). More switches for multiple/single WHO 3 events occurred over calendar time (2004–2008) reflecting wider promotion of WHO 3 events as switch criteria in WHO 2006 guidelines[9]; eg 85% of pre-2007 switches were due to WHO 4 events compared to 53% subsequently.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. CD4 and VL at switch to second-line for first-line failure in patients not receiving CD4 or VL count monitoring (CDM).

https://doi.org/10.1371/journal.pone.0057580.t002

CD4 counts were performed 12-weekly in all CDM participants, but, as not returned, did not influence the decision to switch. Although the median(IQR) CD4 count at switch was 56(15–196), 64(20%) participants had CD4 ≥250 cells/mm³ (Table 2). Switching for failure determined by a single WHO 3 event was significantly more likely to occur with CD4≥250 cells/mm³ (36%) compared to multiple WHO 3 (10%) or WHO 4 (16%) events (p = 0.0002).

Most WHO 4 events triggering switch had relatively similar proportions with CD4≥250 cells/mm³ and <50 cells/mm³ (Table 3), with the exception of cryptococcal meningitis, where one third (11 events) triggered switch with CD4≥250 cells/mm³. Interestingly CD4 was ≥250 cells/mm³ in three of the four switches to second-line triggered by lymphoma. Although generally considered a less severe WHO 4 event, median(IQR) CD4 at switch triggered by oesophageal candidiasis was only 30(8–68) cells/mm³. Weight loss, severe bacterial infection (SBI) and diarrhoea were the main single WHO 3 events triggering switch with CD4≥250 cells/mm³ (44%, 43% and 100% respectively), likely reflecting their frequency in adults irrespective of HIV status or CD4. However, combinations of ≥2 WHO 3 events triggered switch at lower CD4 counts, similar to WHO 4 events.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Clinical events triggering switch after 44 weeks on ART in patients monitored without CD4 cell counts (CDM).

https://doi.org/10.1371/journal.pone.0057580.t003

CDM participants switching with WHO 4 events were more likely to die within a year of switch than those switching with multiple or single WHO 3, consistent with their severity (19% versus 3% versus 5% respectively, exact p = 0.002, Table 2). Interestingly, subsequent mortality was similar in those switching at high and lower CD4 counts: 17% (11/64) CDM participants switching with CD4≥250 cells/mm³ died within a year of switch versus 13% (33/250) switching with CD4<250 cells/mm³ (p = 0.41, p = 0.51 adjusted for WHO 3/4 events); and 27% (9/33) CDM participants switching for WHO 4 events with CD4≥250 cells/mm³ died within 1 year versus 17% (30/174) with CD4≤ 250 cells/mm³ (p = 0.22).

Overall 181(58%) CDM participants had VL at switch to second-line assayed retrospectively. 49(27%) had VL<400 copies/ml, with similar proportions across reasons for triggering switch (p = 0.29) (Table 2). Thus 3.7 ‘tie-breaker’ VL tests at clinical failure would be needed to prevent one switch with suppressed VL. There was a very wide range of VLs in clinically-monitored participants failing and switching with CD4<100 cells/mm³ (Figure 1(a)). In contrast, most with CD4>250 cells/mm³ had suppressed VL.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. VL and CD4 at switch to second-line for first-line failure in patients receiving and not receiving CD4 count monitoring.

(a) in patients not receiving routine CD4 count monitoring (CDM: 20% >250 cells/mm3). (b) in patients receiving routine CD4 count monitoring (LCM: 2% >250 cells/mm3). Footnote 1 Points at >750000 and <400 have a small amount of ‘jitter’ added so that all observations are visible.

https://doi.org/10.1371/journal.pone.0057580.g001

To inform practice where VL testing is unavailable or performed off-site (when return of results may be delayed considerably), we evaluated the predictive ability of a single tie-breaker CD4 count at clinically-triggered switch to identify participants with VL<400 copies/ml, using data on VLs assayed retrospectively on stored samples and CD4 counts which had not been returned to clinicians. VL was <400 copies/ml in 38/48(79%) with CD4≥250 cells/mm³ versus only 11/133(8%) with CD4<250 cells/mm³ (p<0.0001). The area under the receiver-operating-characteristic (ROC) curve (Figure 2(a)) was 0.91 (95% CI 0.86–0.96), with an optimal threshold where most observations were correctly classified (90%) of CD4≥220 cells/mm³. This cutoff had 86% sensitivity, 92% specificity, 79% positive predictive value and 95% negative predictive value for identifying participants with VL<400 copies/ml (LR+ = 10.3, LR- = 0.16). ROC areas were similar according to reason for switch and in those joining and not joining second-line studies (Table 2). Therefore a threshold of >250 cells/mm³, a close but more likely cut-off for a CD4 point-of-care assay, would capture most individuals without virological failure, for whom switching could be premature and unnecessary.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Ability of a single CD4 at switch to second-line for first-line failure to predict VL.

(a) in patients not receiving routine CD4 count monitoring (CDM: 20% >250 cells/mm3). (b) in patients receiving routine CD4 count monitoring (LCM: 2% >250 cells/mm3). Footnote 2 Receiver operator curves (ROC) show how the sensitivity and specificity of CD4 thresholds for predicting VL<400 copies/ml varies as CD4 increases from 1 to 788 (CDM) or 505 (LCM) cells/mm³. The straight line indicates performance no better than chance. The threshold with the greatest probability of correctly classifying each CD4 count according to whether it has VL<400 copies/ml or not is indicated with sensitivity (proportion with VL<400 c/ml who have CD4≥threshold), specificity (proportion with VL≥400 c/ml who have CD4 <threshold), positive predictive value (proportion of patients with CD4 ≥threshold who have VL<400 c/ml) and negative predictive value (proportion of patients with CD4 <threshold who have VL≥400 c/ml).

https://doi.org/10.1371/journal.pone.0057580.g002

Patients monitored clinically with 12-weekly CD4 but no VL results

1656 LCM participants initiated ART with median(IQR) CD4 86(32–140) cells/mm³ and were clinically monitored together with 12-weekly CD4 counts for median 5 years. 361(22%) switched to bPI-containing second-line for first-line failure after a median(IQR) 2.8(2.1–3.8) years on first-line (shorter than in CDM, p = 0.0001) (only 1 <48 weeks). In those who switched, median(IQR) pre-ART CD4 was 42(17–85) cells/mm³ and age at switch 39(34–45) years; 201(57%) were female.

326(20%) had a new/recurrent WHO 4 event accepted by the ERC or met immunological failure criteria after the first 44 weeks on ART. 286/326 (88%) switched to second-line, 6(2%) died on first-line before switching and 34(10%) had not switched before trial closure. The Kaplan-Meier median(IQR) time to switch after meeting the criteria was 4(2–20) weeks, similar to CDM (p = 0.70). The most commonly reported reasons for delaying switch for >8 weeks or not switching was high CD4s (71%) or because of drug-drug interactions between rifampicin and bPI (24%) (Table 1; n = 42). An additional 75 participants switched to second-line ART for WHO 4 events not judged to meet pre-defined protocol criteria by the ERC, single/multiple WHO 3 events or other CD4-related reasons.

As expected, since CD4 decline generally precedes development of WHO 4 events, most (73%, 265/361) LCM participants who switched failed by the CD4<100 cells/mm³ protocol criteria. An additional 37(10%) switched for WHO 4 events; 43(12%) for other CD4 reasons (mainly rapid CD4 decline to >100 cells/mm³); and only 6(2%) and 10(3%) with multiple or single WHO 3 events respectively (Table 4). Over time, more switches occurred for CD4s which were low but ≥100 cells/mm³; for example, 2% pre-2007 switches were due to CD4 ≥100 cells/mm³ compared to 24% subsequently, reflecting changes in WHO2006 guidelines[8].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. CD4 and VL at switch to second-line for first-line failure in patients receiving 12-weekly CD4 count monitoring, but no VL monitoring (LCM).

https://doi.org/10.1371/journal.pone.0057580.t004

The median(IQR) CD4 count at switch was 63(36–95) cells/mm³. In contrast to CDM participants, but reflecting clinician reluctance to switch LCM participants with WHO events and high CD4s (Table 1), only 7(2%) CD4-monitored participants switched with CD4≥250 cells/mm³ (3 for WHO 4; 1 single WHO 3; 3 other CD4 reasons, all in participants with substantial prior CD4 variability). Mortality following switch was highest in participants who switched for WHO 4 events before meeting immunological failure criteria (32% (12/37); p<0.0001)

Overall 187(52%) participants had VL at second-line switch assayed retrospectively. 23(12%) had VL<400 copies/ml (Table 4; p<0.0001 versus CDM). Thus 8.1 ‘tie-breaker’ VL tests at immunological/clinical failure would be needed to prevent one switch with suppressed VL. Fewer participants switching for CD4<100 cells/mm³ had VL<400 copies/ml (7%) than those meeting other CD4 or clinical criteria (p = 0.03). Those with CD4<100 cells/mm³ had a very wide range of VLs (Figure 1(b)), similarly to participants monitored without CD4s.

In those receiving 12-weekly CD4 counts, CD4 at switch was a much poorer predictor of suppressed VL (Figure 2(b)); with area under the LCM ROC curve 0.70 (95% CI 0.58–0.73) (similar according to reason for switch and participation in second-line studies). This may be because only 14(4%) had CD4 above the optimal threshold of 220 cells/mm³ for identifying VL suppression found in CDM, since those LCM participants with high CD4, in whom VL was more likely suppressed, were rarely switched by clinicians (Table 1). In fact, 54% (13/24) of LCM participants who switched with VL<400 copies/ml actually had CD4<110 cells/mm³ (Figure 1(b)) (median(range) 3.4(1.1–4.9) years on first-line), highlighting that completely accurate CD4-based prediction of suppressed VL is impossible.

Interestingly, irrespective of monitoring strategy, virological-responders/immunological-non-responders with VL<400 copies/ml but CD4<110 cells/mm³ had excellent response to bPI-containing second-line ART. In 14 participants (10LCM,4CDM) the median(IQR) CD4 increase 24 weeks after switching was +129 (+42,+216) cells/mm³ from median(IQR) 70(38–94) cells/mm³ at switch (4 others (3LCM,1CDM) died before 24 weeks on second-line).

Discussion

Most public sector ART clinics in low-income settings have very limited laboratory capacity to monitor patients on therapy, so justifying and prioritising services provided to support clinical monitoring is critical. DART has already shown that routine CD4 monitoring alone from the second year of ART has a small but important impact on survival[4]. WHO recommends the use of VLs to confirm immunological/clinical failure[2]. Currently, neither are widely accessible; eg in April-June 2011, there were only 50 functional CD4 machines in 449 ART clinics in Malawi[10] and in early 2012, only 4/59 ART centres in Malawi, Zimbabwe, Uganda, including hospitals, had the possibility to monitor VL, including off-site[3]; even if theoretically functional, lack of electricity/consumables/personnel may further reduce their availability in practice. Further analysis of the unique group of DART participants failing first-line ART who were clinically-monitored without routine CD4/VL tests, but with retrospective CD4 and VLs available, now shows the utility of multiple, but not single, WHO 3 events as clinical failure criteria in the absence of any CD4 monitoring; and that, where access to single CD4 tests is available, a CD4 tiebreaker at a 250 cells/mm³ threshold could identify ∼80% of those failing clinically with VL<400 copies/ml who may be unlikely to benefit from switching to more costly second-line therapy.

Despite WHO recommendations to use a VL ‘tie-breaker’ test to confirm clinical/immunological failure[2], access to expensive HIV RNA testing is unlikely to improve soon given the current financial crisis. Point-of-care VL testing could dramatically change this, but is unlikely to become available for several years, and may still be relatively costly. Meanwhile, many public sector ART programmes will continue to monitor ART patients with negligible access to VL. In contrast, CD4 testing is more widely accessible[3], and point-of-care devices already in evaluation will soon increase coverage[11]. However, the sheer volume of testing will remain challenging, as evidenced by stockouts even of simple HIV tests[3] and given the 6.6 million adults/children receiving ART in low/middle-income settings[12]. Making routine CD4 monitoring available to all would require significant additional investments in laboratory infrastructure, personnel and consumables, which may not be possible given the current financial situation, particularly as, at current costs, it is not cost-effective for most African countries[13], [14], and a more pressing priority is to rollout ART to more who need it. Additional benefits of routine VL over CD4 monitoring are small[15], [16] or negligible[17] and even less cost-effective[18]. It is therefore essential to consider parsimonious ways to use CD4 testing without VLs to support clinical monitoring in the critical decision of when to switch to second-line.

As our analysis investigated characteristics of those patients switched to second-line, we did not (and cannot) estimate the overall accuracy of CD4 (or clinical) criteria for identifying virological failure in all individuals on treatment. However, our data clearly confirm that monitoring for clinical failure alone over-identifies immunological failure, potentially resulting in unnecessary and premature switching to more costly second-line ART[14], [19]: 20% of clinical failures/switches had CD4>250 cells/mm³. The low CD4 nadir in DART participants (10% had pre-ART CD4 <10 cells/mm³) may have contributed to this, with patients at long-term risk for events such as lymphoma, despite immune reconstitution. Furthermore, 12% of CD4-monitored and 27% of clinically-monitored participants had suppressed VL at failure/switch, as previously reported[19]-[22]. Discordance between clinical, immunological and virological failure at any single timepoint is expected, as they track different processes. Nor is failure always absolute: eg 50% of patients with virological failure and genotypic NNRTI resistance in one South African cohort re-suppressed while receiving an NNRTI[23]. In resource-limited settings with access to single confirmatory laboratory tests, the challenge of how to deal practically with discordance remains.

Whilst relatively short periods on ART leading to incomplete immune reconstitution may account for some discrepancies in previous studies[19], [20], the 18 participants switching with CD4<110 cells/mm³ and VL<400 copies/ml in DART had been on first-line ART for median 3.4 years. Some had variable CD4 responses on first-line, with periods of low CD4 without documented non-adherence; others had never responded immunologically, similarly to Kantor et al where 3/7 patients with persistent CD4<100 cells/mm³ had undetectable VL[20]. Most DART participants with VL<400 copies/ml but very low CD4 at failure benefited considerably immunologically from second-line switch; however 22% (4/18) died shortly after switching. Given increased mortality risks at CD4<100 cells/mm³ even with suppressed VL in resource-rich[24], [25] and resource-limited settings[26], there may be clinical benefits from switching this specific group of discordant responders with very low CD4 from an NRTI/NNRTI-based first-line to a bPI-based second-line regimen. Of interest, we also observed relatively high mortality in participants switching with high CD4 counts and WHO 4 events, possibly reflecting that developing such events despite apparently high absolute CD4 counts may indicate underlying functional immune deficits that may themselves impact mortality risk.

Those ‘failing ART’ clinically with only single WHO 3 events can be viewed in two ways. The fact that nearly 40% had CD4>250 cells/mm³ and 33% VL<400 copies/ml demonstrates lack of sensitivity of these events for ART failure, probably because of their frequency in the underlying population irrespective of HIV status. Whilst our results suggest single WHO 3 events should not trigger switch in isolation, conversely 64% of this group had CD4≤250 cells/mm³ and 67% VL>400 copies/ml, greater than the wider population on first-line. Targeting them for confirmatory tiebreak CD4 testing could therefore identify additional ART failures whilst avoiding premature switching. Of note, participants switching with multiple WHO 3 events had lower CD4 and higher VL at switch than single WHO 3 events. Adding WHO 3 (single or multiple) events to the WHO 4 criteria for clinical failure increased the numbers identified by about 50%.

One potential limitation of our study is that clinical monitoring was conducted by nurses and clinicians in relatively well-supported, staffed and supervised sites with no ART stock-outs. DART had excellent retention (just 7% loss to follow-up over five years) and 5-year survival (87% CDM, 90% LCM). Among participants randomised to CDM, CD4s were measured but not returned and health-workers remained blinded to CD4s throughout the trial. This enabled us to perform analyses impossible outside a trial design such as DART. Whilst several WHO 4 events triggering switch in DART were not considered to meet criteria for trial endpoints on independent review, we included them in these analyses. In ART programmes, more WHO 3/4 events might be over-diagnosed (with clinicians conservatively ascribing clinical episodes as WHO 3/4 events), but these would likely occur with high CD4 and VL<400 copies/ml supporting generalisability of our findings. DART participants were severely immune-compromised (median 86 cells/mm³) when initiating ART: generalizability to programmes initiating ART earlier is unknown. However, such patients would take longer to fail on first-line therapy, and therefore it is plausible that a greater (rather than smaller) proportion of clinical events on first-line would be single WHO 3 rather than multiple WHO 3 or 4 events, in whom we found greatest VL suppression. Our findings may therefore be more, rather than less generalizable, to such settings.

The DART protocol only included one of the three WHO immunological failure criteria (confirmed <100 cells/mm³) for LCM participants (although a small number of participants switched for other CD4 concerns before this). The other criteria require a series of CD4s (50% decline from peak) or a pre-ART CD4 (drop below pre-ART baseline) and were not included in the protocol because they were judged impractical in settings with limited access to CD4 testing and are anyway not validated; it was considered that switching should be determined only by known predictors of mortality on ART (ie overall CD4, not declines or drops below baseline). These other immunological failure criteria also generally lead to switch at higher CD4 counts and might be expected to be associated with higher rates of VL suppression at switch than demonstrated here, but we cannot assess this. In addition, as real-time VL monitoring was not performed, we cannot evaluate the VL>5000 copies/ml WHO criteria for switching without immunological or clinical failure[2], nor investigate the performance of CD4 criteria in identifying this threshold value. Figure 1 shows that most participants switching for immunological/clinical failure with VL>400 copies/ml had values around or above this threshold.

The unique opportunity afforded by the DART trial design to evaluate the value of using CD4 testing parsimoniously to support clinical monitoring when VL monitoring is unavailable, provides clear evidence that a single CD4 tie-breaker at clinical failure with a 250 cells/mm³ threshold can identify patients who have suppressed VL, providing the potential to reduce unnecessary switching to costly second-line ART. Adding a single CD4 count in those failing clinically could therefore improve the specificity (and positive predictive value) of clinically identified failure for virological failure: clearly a single count cannot improve the sensitivity for detecting virological failure in patients without clinical events. Our results further suggests that multiple (but not single) WHO 3 events should be considered as well as WHO 4 events in the definition of clinical failure when CD4 testing is unavailable. If limited CD4 testing is possible, targeting this to patients with a single WHO stage 3 event could identify additional failures with detectable VL whilst still avoiding premature, costly switching to second-line. Programmes in low-income countries that are considering how to scale-up laboratory services to rationally support regular clinical follow-up can use these results to plan how to widen access to CD4 monitoring taking advantage of new point-of-care technologies.

Acknowledgments

We thank all the patients and staff from all the centres participating in the DART trial.

MRC/UVRI Uganda Research Unit on AIDS, Entebbe, Uganda: H Grosskurth, P Munderi, G Kabuye, D Nsibambi, R Kasirye, E Zalwango, M Nakazibwe, B Kikaire, G Nassuna, R Massa, K Fadhiru, M Namyalo, A Zalwango, L Generous, P Khauka, N Rutikarayo, W Nakahima, A Mugisha, J Todd, J Levin, S Muyingo, A Ruberantwari, P Kaleebu, D Yirrell, N Ndembi, F Lyagoba, P Hughes, M Aber, A Medina Lara, S Foster, J Amurwon, B Nyanzi Wakholi, K Wangati, B Amuron, D Kajungu, J Nakiyingi, W Omony, K Fadhiru, D Nsibambi, P Khauka.

Joint Clinical Research Centre, Kampala, Uganda: P Mugyenyi, C Kityo, F Ssali, D Tumukunde, T Otim, J Kabanda, H Musana, J Akao, H Kyomugisha, A Byamukama, J Sabiiti, J Komugyena, P Wavamunno, S Mukiibi, A Drasiku, R Byaruhanga, O Labeja, P Katundu, S Tugume, P Awio, A Namazzi, GT Bakeinyaga, H Katabira, D Abaine, J Tukamushaba, W Anywar, W Ojiambo, E Angweng, S Murungi, W Haguma, S Atwiine, J Kigozi, L Namale. A Mukose, G Mulindwa, D Atwiine, A Muhwezi, E Nimwesiga, G Barungi, J Takubwa, S Murungi, D Mwebesa, G Kagina, M Mulindwa, F Ahimbisibwe, P Mwesigwa, S Akuma, C Zawedde, D Nyiraguhirwa, C Tumusiime, L Bagaya, W Namara, J Kigozi, J Karungi, R Kankunda, R Enzama.

University of Zimbabwe, Harare, Zimbabwe: A Latif, J Hakim, V Robertson, A Reid, E Chidziva, R Bulaya-Tembo, G Musoro, F Taziwa, C Chimbetete, L Chakonza, A Mawora, C Muvirimi, G Tinago, P Svovanapasis, M Simango, O Chirema, J Machingura, S Mutsai, M Phiri, T Bafana, M Chirara, L Muchabaiwa, M Muzambi, E Chigwedere, M Pascoe, C Warambwa, E Zengeza, F Mapinge, S Makota, A Jamu, N Ngorima, H Chirairo, S Chitsungo, J Chimanzi, C Maweni, R Warara, M Matongo, S Mudzingwa, M Jangano, K Moyo, L Vere, I Machingura.

Infectious Diseases Institute (formerly the Academic Alliance) Makerere University, Mulago, Uganda: E Katabira, A Ronald, A Kambungu, F Lutwama, I Mambule, A Nanfuka, J Walusimbi, E Nabankema, R Nalumenya, T Namuli, R Kulume, I Namata, L Nyachwo, A Florence, A Kusiima, E Lubwama, R Nairuba, F Oketta, E Buluma, R Waita, H Ojiambo, F Sadik, J Wanyama, P Nabongo, J Oyugi, F Sematala, A Muganzi, C Twijukye, H Byakwaga.

The AIDS Support Organisation (TASO), Uganda: R Ochai, D Muhweezi, A Coutinho, B Etukoit.

Imperial College, London, UK: C Gilks, K Boocock, C Puddephatt, C Grundy, J Bohannon, D Winogron.

MRC Clinical Trials Unit, London, UK: J Darbyshire, DM Gibb, A Burke, D Bray, A Babiker, AS Walker, H Wilkes, M Rauchenberger, S Sheehan, C Spencer-Drake, K Taylor, M Spyer, A Ferrier, B Naidoo, D Dunn, R Goodall.

Independent DART Trial Monitors: R Nanfuka, C Mufuka-Kapuya.

DART Virology Group: P Kaleebu (Co-Chair), D Pillay (Co-Chair), P Awio, M Chirara, D Dunn, DM Gibb, C Gilks, R Goodall, A Kapaata, M Katuramur, F Lyagoba, R Magala, B Magambo, K Mataruka, A McCormick, L Mugarura, T Musunga, M Nabankkema, J Nkalubo, P Nkurunziza, C Parry, V Robertson, M Spyer, D Yirrell.

Trial Steering Committee: I Weller (Chair), A Babiker (Trial Statistician), S Bahendeka, M Bassett, A Chogo Wapakhabulo, J Darbyshire, B Gazzard, C Gilks, H Grosskurth, J Hakim, A Latif, C Mapuchere, O Mugurungi, P Mugyenyi; Observers: C Burke, M Distel, S Jones, E Loeliger, P Naidoo, C Newland, G Pearce, S Rahim, J Rooney, M Smith, W Snowden, J-M Steens.

Data and Safety Monitoring Committee: A Breckenridge (Chair), A McLaren (Chair-deceased), C Hill, J Matenga, A Pozniak, D Serwadda.

Endpoint Review Committee: T Peto (Chair), A Palfreeman, M Borok, E Katabira.