T

Mi

P

a

b

c

d

RA

1

SHa

p

0d

at University of L

iverpool onhttp://trstm

h.oxfordjournals.org/D

ownloaded from

ransactions of the Royal Society of Tropical Medicine and Hygiene (2009) 103, 588—593

avai lab le at www.sc iencedi rec t .com

journa l homepage: www.e lsev ierhea l th .com/ journa ls / t rs t

ortality and loss to follow-up among HAARTnitiators in rural South Africa

eter MacPhersona,∗, Mosa Moshabelaa, Neil Martinsonb,c, Paul Pronyka,d

School of Public Health, University of Witwatersrand, P.O. Box 2, Acornhoek, Limpopo 1360, South AfricaJohns Hopkins University Center for TB Research, 15 N. Wolfe Street, Baltimore, MD 21205, USAPerinatal Health Research Unit, University of Witwatersrand, P.O. Box 114, Diepkloof 1864, South AfricaLondon School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK

eceived 10 August 2008; received in revised form 3 October 2008; accepted 6 October 2008vailable online 13 November 2008

KEYWORDSHAART;HIV;Tuberculosis;Mortality;Immunereconstitutioninflammatorysyndrome;South Africa

Summary A retrospective cohort study of mortality rates and potential predictors of deathwas conducted in public-sector patients initiating highly active antiretroviral therapy (HAART)between October 2005 and September 2007 in a rural, under-resourced region of South Africa.The aims were to determine the relative contribution of death to cohort exit and the causesand predictors of mortality among HAART initiators. A community outreach programme tracednon-attenders. Patients categorised as dying at home underwent a verbal autopsy (by inter-viewing family members) and case-file review, and those dying in hospital a case-file review,to determine the probable cause of death. At 24 months 1131 (83.6%) patients were retainedon treatment in the programme, 124 (9.2%) had died, 63 (4.7%) had transferred out, and 35(2.6%) were lost to follow-up. The most common causes of death were tuberculosis (44.3%) anddiarrhoeal diseases (24.5%). Death was the major reason for cohort exit. As experience is gained

with rural HAART programmes mortality rates may decrease. These results draw attention to theneed for early HIV diagnosis, increased access to HAART services with earlier treatment initia-tion, and routine screening and aggressive management of opportunistic infections, particularly

opica

i

May 1, 2015

tuberculosis.

© 2008 Royal Society of Trreserved.

. Introduction

outh Africa has the highest number of people living withIV/AIDS globally,1 and despite recent progress in improvingccess to highly active antiretroviral therapy (HAART),2 lim-

∗ Corresponding author. Tel.: +27 78 3900310; fax: +27 13 7955082.E-mail addresses: petermacpherson@mac.com,

etermacp@gmail.com (P. MacPherson).

cnucfm

tl

035-9203/$ — see front matter © 2008 Royal Society of Tropical Medicinoi:10.1016/j.trstmh.2008.10.001

l Medicine and Hygiene. Published by Elsevier Ltd. All rights

ted uptake,3,4 poor retention5 and difficulties in accessingare6 remain serious issues. While there have been sig-ificant efforts to expand access to HAART services fromrban centres to resource-constrained rural communities,7

oncerns have been raised regarding high rates of loss to

ollow-up in a number of sub-Saharan African HAART treat-ent programmes.5

Experience from urban African settings8—17 has suggestedhat mortality post-HAART initiation is a major reason foross to follow-up. However, there is less evidence for the

e and Hygiene. Published by Elsevier Ltd. All rights reserved.

bsamasocwtdaatfdp

ottptfidn

3

Bwprdua

rl3cw((w

d1ipaC2

mc

at University of L

iverpool on May 1, 2015

http://trstmh.oxfordjournals.org/

Dow

nloaded from

HAART in rural South Africa

contribution of mortality to cohort exit from operationalprogrammes in rural areas. In addition, accurately assessingcause of death in such contexts poses numerous challengesdue to poor-quality clinical records, high rates of out-of-hospital death, the absence of strong systems for monitoringadherence and tracing defaulters, and the social and geo-graphic barriers that limit access to care in rural areas.

We employed a community tracing programme for track-ing loss to follow-up, alongside a systematic assessment ofcause of death using verbal autopsies and clinical-recordreviews among HAART initiators in a rural South AfricanHIV/AIDS treatment programme. The objectives were todocument the relative contribution of death to cohort exit,to understand better the causes and predictors of mortality,and to generate lessons that might inform the scaling up ofHAART services in resource-limited areas.

2. Methods

This study was conducted in the Bushbuckridge district ofSouth Africa’s rural north-east, a densely populated formerhomeland area of approximately 500 000 people living in 133villages.18 The region has been designated one of 13 nodalareas of extreme deprivation in the country. Antenatal HIVseroprevalence was 32.1% at the time of the study2 and theprevalence of pulmonary tuberculosis (TB) in the region was157/100 000.19

The clinical-care site where the study took place (RixileARV Clinic, Tintswalo Hospital, Acornhoek, Bushbuckridgedistrict) began initiating HAART in October 2005. A packageof comprehensive HIV/AIDS care was offered free of chargeas part of the government programme. Patients with HIVwere eligible for HAART if they met South African NationalDepartment of Health guidelines (CD4 < 200 × 106 cells/�l orWHO Stage 4 condition, and psychosocial preparedness toundertake therapy).20 After HAART initiation patients weregiven monthly review appointments at the clinic. Patientswho failed to return on their designated appointment daywere traced with the assistance of a local community-basedorganisation. Initial tracing was attempted by clinic staff bytelephone on the day following a missed appointment and,if this was unsuccessful, the community-based organisationdeployed a local field worker from the patient’s nearest pri-mary health centre to locate the patient. Care was takennot to disclose HIV status to families or other householdmembers. Patients who could not be traced and did not sub-sequently return to the clinic during the study period wereclassified as lost to follow-up. Patients who notified clinicstaff of their intention to transfer to another designatedHAART treatment site were given a written referral formand classified as transferred out.

A retrospective analysis was conducted of all those aged18 years and over who initiated HAART between 1 Octo-ber 2005 and 30 September 2007. For HAART initiators whowere identified as having died in hospital, two medicalofficers with experience of HIV in the rural setting inde-

pendently reviewed the patients’ hospital files (includinglaboratory and radiological investigations) to determine theprobable cause of death. Where the opinion of the twophysicians differed an arbitration process was carried outby a third physician. For patients dying at home a ver-

121ch

589

al autopsy was conducted by a trained verbal autopsypecialist who interviewed family members. The verbalutopsy tool had previously been validated in the local com-unity under investigation.21 For each patient who died

t home the verbal autopsy report was reviewed along-ide the clinical records to determine the probable causef death using the same process outlined above. Tuber-ulosis was suspected on clinical symptoms of cough >2eeks, night sweats, weight loss and chest pain, or symp-

oms and signs suggestive of extrapulmonary TB, and wasivided into microbiologically confirmed (which was defineds a single positive microscopy sample or culture result)nd non-microbiologically confirmed. Immune reconstitu-ion inflammatory syndrome (IRIS) was considered to be aactor contributing to death if the medical record prior toeath suggested a clinical worsening due to an opportunisticathogen despite an adequate response to HAART.22

Data were censored at the date of death or at the pointf last clinical contact as of 30 September 2007. Those losto follow-up were not assumed to have died unless confirma-ion of their death was obtained by the community tracingrogramme. The Cox proportional hazards model was usedo evaluate associations between risk factors for death andxed covariates at baseline, and a multivariate model waseveloped for variables which were identified as having sig-ificance in the univariate analysis.

. Results

etween October 2005 and September 2007, 1353 patientsere initiated onto HAART and followed up for 11 589atient-months. At 24 months 1131 (83.6%) patients wereetained on treatment in the programme, 124 (9.2%) hadied, 63 (4.7%) had been transferred out, and 35 (2.6%) werenable to be traced by the community-based organisationnd so were defined as lost to follow-up (Figure 1).

The median age of initiators was 37 years [interquartileange (IQR) 31—45] and 67.0% were female. Median CD4ymphocyte count was 93 × 106 cells/�l (IQR 37—148) and0.9% of patients presented with a CD4 count of ≤50 × 106

ells/�l at baseline. A baseline CD4 count ≤50 × 106 cells/�las seen in 36.7% of males but only in 28.0% of females

Mann-Whitney test: P = 0.001). Median weight was 55 kgIQR 48—62.5) and 53% and 10.7% of patients, respectively,ere WHO stage 3 and 4.

The median time to death from HAART initiation was 57ays (IQR 27.5—169 days) and the overall mortality rate was3.0 per 100 person-years (95% CI 10.9—15.5). From HAARTnitiation to 6 months the mortality rate was 19.5 per 100erson-years (95% CI 16.0—23.8); between 6 and 12 monthsfter HAART initiation it was 7.5 per 100 person-years (95%I 4.9—11.5); and between 12 and 24 months the rate was.9 per 100 person-years (95% CI 1.2—7.0).

Patients who died between HAART initiation and 6onths had a lower median baseline CD4 count (27 × 106

ells/�l; IQR 11—83) than patients who died between 6 and

2 months after initiation (median 44 × 106 cells/�l; IQR4—77) or after 12 months (median 158.5 × 106 cells/�l; IQR39—198); test for trend P = 0.008. Individuals with a CD4ount ≤50 × 106 cells/�l at baseline had a 3.7 or 2.9 timesigher risk of death than individuals with higher CD4 counts,

590 P. MacPherson et al.

n hi

budCdw

(1dcpctu

d(1fd

Tt

w5fa

4

ImHfo

at University of L

iverpool on May 1, 201

http://trstmh.oxfordjournals.org/

Dow

nloaded from

Figure 1 Outcomes of patients initiated o

y univariate and multivariate analysis, respectively. Bothnivariate and multivariate analysis of baseline variablesemonstrated that the presence of WHO stage 3 or 4 disease,D4 count ≤50 × 106 cells/�l, weight ≤45 kg, and male gen-er were significantly associated with death (Table 1). Ageas not predictive of death.

Among 124 patients who were known to have died, 3225.8%) died at home and 92 (74.2%) died in hospital. Of the24 deaths, 18 were not included in the analysis: 13 (10.5%)ied in other hospitals from which we were unable to accessase records, we were unable to locate the families of fouratients for interview and the family of one patient refusedonsent. The opinion on the cause of death differed betweenhe two physicians in 44/106 (41.5%) cases and so these casesnderwent the arbitration process.

Tuberculosis in all its forms was the leading cause of

eath in HAART-treated adults and accounted for 44.3%47/106) of all deaths (Table 2). Of those who died of TB,9/47 (40.4%) had been on TB treatment at the time of deathor a median of 2 months (IQR 0.25—4 months). Diarrhoealiseases accounted for a further quarter of deaths (24.5%).

fLcZe

Table 1 Cox proportional hazards model of associations between

Baseline characteristic Hazard ratio for death

Univariate analysis

Age 0.992 (0.Male gender 1.551 (1.Weight ≤45 kg 2.950 (2.CD4 count ≤50 × 106 cells/�l 3.694 (2.WHO Stage 3 or 4 2.376 (1.

ghly active antiretroviral therapy (HAART).

he remaining deaths were due mostly to AIDS-related infec-ions, carcinomas and adverse events related to HAART.

IRIS was implicated in 18/106 (17.0%) deaths — nineomen and nine men. This was found to have occurred in/14 (35.7%) deaths from extrapulmonary TB, 8/33 (24.2%)rom pulmonary TB, 4/26 (15.4%) from diarrhoeal diseasesnd 1/3 (33.3%) from hepatic failure.

. Discussion

n this South African government-run rural HIV/AIDS treat-ent programme more than 80% of patients initiated ontoAART were retained at 24 months, a figure that compares

avourably with other sites in sub-Saharan Africa.5 Ratesf mortality (9.2%) were higher than previously reported

rom urban South African settings14 and the recent ART-INC (2.3—6.4%) collaborative assessment8 of low-incomeountries, but comparable to other African sites such asambia,13 Uganda16 and Ethiopia.11 WHO stage 3 or 4 dis-ase, CD4 count ≤50 × 106 cells/�l, weight ≤45 kg, and

baseline characteristics and death.

(95% CI)

Multivariate analysis

974—1.010)087—2.214) 1.626 (1.120—2.360)038—4.271) 2.481 (1.683—3.659)580—5.290) 2.902 (1.997—4.215)522—3.711) 1.591 (1.007—2.513)

5

HAART in rural South Africa 591

Table 2 Causes of death in highly active antiretroviral therapy (HAART) initiators.

Cause of death n Percentage cause of death

Pulmonary TB 31.13Not microbiologically confirmed 23 21.70Microbiologically confirmed 10 9.43

Extrapulmonary TB 13.20TB meningitis 6 5.66TB pleura 4 3.77TB pericarditis 3 2.83Miliary TB 1 0.94

Diarrhoeal diseases 24.52Not microbiologically confirmed 24 22.64Dysentery 1 0.94Cryptosporidiosis 1 0.94

Meningitis 4.72Cryptococcal 3 2.83Bacterial 2 1.89

Hepatic failure 2.83Drug-induced 1 0.94Unknown aetiology 2 1.89

Carcinoma 6.59Kaposi’s sarcoma 3 2.83Prostate 1 0.94Cervix 1 0.94Bladder 1 0.94Endometrium 1 0.94

Obstetric 1.88Abruptio placentae 1 0.94Ectopic pregnancy 1 0.94

Cerebral space-occupying lesion 3.77Brain abscess 2 1.89Toxoplasmosis 1 0.94Other 1 0.94

Congestive cardiac failure 2 1.89 1.89Pneumonia 2 1.89 1.89Septicaemia 4 3.77 3.77HIV encephalopathy 1 0.94 0.94Diabetic ketoacidosis 1 0.94 0.94Renal failure 1 0.94 0.94

wfnis

mT

at University of L

iverpool on May 1, 2015

http://trstmh.oxfordjournals.org/

Dow

nloaded from

Upper GI bleed 1

Total 106

male gender were predictive of death, similar to patternsreported elsewhere.9,11,12

Although the incidence of TB is dramatically reducedby HAART23 it remains the leading cause of morbidityand mortality, both in untreated and in HAART-treatedadults,24 and its contribution to death may be under-estimated. Tuberculosis was implicated as the cause innearly half the deaths in this study, higher than previously

25

reported. We also note that approximately 40% of patientswho died of TB had been on TB treatment for a mediantime of 2 months, suggesting that patients with TB andHIV co-infection may be vulnerable to death, perhapsbecause of IRIS. Autopsy studies similar to those done

pHhpo

0.94 0.94

100 100

hen antiretroviral treatment was not available26 couldurther refine our findings. Diarrhoeal diseases and otheron-tuberculous AIDS-associated pathologies were alsomplicated in mortality, similar to the results of other Africantudies.25

IRIS is increasingly recognised as a major contributor toortality in HAART patients,22 especially in relation to TB.27

he incidence of IRIS in South Africa is reported to be 25.1

er 100 patient-years and the presence of a low CD4 count atAART initiation is predictive of occurrence.22 Our findingsighlight the importance of being vigilant for IRIS in the ruralatient, who may present later with a low CD4 count and anccult opportunistic infection.

5

(imbHfttaCs

matattinndosrgaittpitodefsabrfi

drtpscDmiposbiFtn

po

Apgcwto

AwiHcAwa

Fbt6Mgw

C

EoRw

R

at University of L

iverpool on May 1, 2015

http://trstmh.oxfordjournals.org/

Dow

nloaded from

92

Of note was the low median presenting CD4 count93 × 106 cells/�l). Those who died in the early post-HAARTnitiation period (0—6 months) had a significantly loweredian CD4 count at HAART initiation than those who diedetween 6—12 months or after 12 months. Presentation forAART with advanced immunosuppression is a key risk factor

or death25 and these findings suggest that late presenta-ion or barriers to accessing HAART may be contributingo mortality. Thus, earlier diagnosis of HIV, improvedccess to HAART services and HAART initiation at higherD4 counts may reduce mortality rates in rural Africanettings.

The strengths of this study included the use of a com-unity tracing system to track loss to follow-up and death,

nd the systematic assessment of cause of death, both inhe community and in the hospital. However, there werelso several limitations. Patients unable to be traced byhe community-based organisation and those transferred outo other HAART treatment sites may also have died, giv-ng an underestimation of mortality rates. Future researcheeds to be directed towards understanding the reasons foron-attendance at rural HAART programmes (e.g. poverty,ifficulty in accessing care) and determining the outcomesf patients transferred to other treatment sites. This maytimulate programmatic improvements aimed at increasingetention rates. Although the clinical site was part of aovernment-run programme, it worked in partnership withn academic institution and a local community-based organ-sation and may have had higher patient retention rateshan government programmes without such support. Fur-hermore, given the rural nature of the site and the largeroportion of deaths due to TB and diarrhoeal diseases, its possible that socio-economic factors such as living condi-ions, income and water supply may have been predictivef mortality. It was not possible to collect socio-economicata for the entire cohort and so the contribution of socio-conomic status to mortality in HAART programmes requiresurther investigation. Finally, it is important to note that thistudy was conducted when HAART was first introduced in therea and those initiated on treatment may, therefore, haveeen awaiting treatment for prolonged periods. Lower deathates may be observed when monitoring patient cohortsrom more established programmes with earlier treatmentnitiation.25

These findings have a number of implications for theelivery of comprehensive HIV/AIDS services in under-esourced settings. Foremost, our experience suggestshat high rates of patient retention in HAART treatmentrogrammes are possible, even in under-resourced ruralettings, highlighting the importance of developing robustommunity-based partnerships to aid patient retention.eath was the major reason for cohort exit and the high earlyortality draws attention to the urgent need for expand-

ng access to testing services to ensure early HIV diagnosis,rompt HAART initiation, and the recognition and treatmentf opportunistic infections. This research also suggests thattrategies to enhance the uptake of testing and treatment

y men, who present later, with a more advanced state ofmmunosuppression and die more often, remains a priority.inally, the high burden of potentially avoidable deaths dueo TB re-affirms the importance of TB/HIV integration, theeed for rapid diagnosis and TB treatment initiation, and

P. MacPherson et al.

ossibly the expansion of TB preventative therapy prior tor soon after HAART initiation.28

uthors’ contributions: PP and NM conceived the studyrotocol and implemented the clinical treatment pro-ramme; PM and MM co-ordinated and undertook the dataollection; PM undertook the statistical analysis; PM and MMrote the first draft of the paper. All authors contributed

o, read and approved the final manuscript. PM is guarantorf the paper.

cknowledgements: Dr Sibusiswe Makhanya for assistanceith the case-file review; Thembi Matokane for undertak-

ng verbal autopsies and Elizabeth Russell, London School ofygiene and Tropical Medicine, for assistance with statisti-al analysis. Prof. Steve Tollman and Dr Kathleen Kahn of thegincourt Health and Population Unit (University of the Wit-atersrand) for technical support and the use of the verbalutopsy instrument.

unding: Patient care at Tintswalo Hospital is funded byoth the President’s Emergency Plan for AIDS Relief, throughhe US Agency for International Development (USAID) (award74-A-00—05—00003—00) and the Department of Health inpumalanga. NM receives a Fogarty International Centerrant (1U2RTW007370—01A1) for TB research training; MMas a recipient of training from this grant.

onflicts of interest: None declared.

thical approval: Ethical approval for this study wasbtained from the University of the Witwatersrand Humanesearch Ethics Committee (Medical), University of the Wit-atersrand, Johannesburg, South Africa.

eferences

1. UNAIDS/WHO. AIDS epidemic update. Geneva: Joint UnitedNations Programme on HIV/AIDS and World Health Organiza-tion; 2007.

2. National Department of Health. HIV and AIDS and STI strate-gic plan for South Africa for 2007—2011. Pretoria: NationalDepartment of Health, Republic of South Africa; 2007.

3. Lawn SD, Wood R. How can earlier entry of patients intoantiretroviral programs in low-income countries be promoted?Clin Infect Dis 2006;42:431—2.

4. Lawn SD, Myer L, Harling G, Orrell C, Bekker L-G, WoodR. Determinants of mortality and nondeath losses from anantiretroviral treatment service in South Africa: implicationsfor program evaluation. Clin Infect Dis 2006;43:770—6.

5. Rosen S, Fox MP, Gill CJ. Patient retention in antiretroviral ther-apy programs in sub-Saharan Africa: a systematic review. PLoSMed 2007;4:e298.

6. Mills EJ, Nachega JB, Bangsberg DR, Singh S, Rachlis B, Wu P, etal. Adherence to HAART: a systematic review of developed anddeveloping nation patient-reported barriers and facilitators.PLoS Med 2006;3:e438.

7. Bedelu M, Ford N, Hilderbrand K, Reuter H. Implementing

antiretroviral therapy in rural communities: the Lusik-isiki model of decentralized HIV/AIDS care. J Infect Dis2007;196:S464—8.

8. The Antiretroviral Therapy in Lower Income Countries (ART-LINC) Collaboration and ART Cohort Collaboration (ART-CC)

http://trstmh.oxfordjournals

Dow

nloaded from

HAART in rural South Africa

groups. Mortality of HIV-1 infected patients in the first yearof antiretroviral therapy: comparison between low-income andhigh-income countries. Lancet 2006;367:817—24.

9. Coetzee D, Hildebrand K, Boulle A, Maartens G, Louis F,Labatala V, et al. Outcomes after two years of providingantiretroviral treatment in Khayelitsha, South Africa. AIDS2004;18:887—95.

10. Etard JF, Ndiaye I, Thierry-Mieg M, Guèye NF, Guèye PM,Lanièce I, et al. Mortality and causes of death in adults receiv-ing highly active antiretroviral therapy in Senegal: a 7-yearcohort study. AIDS 2006;20:1181—9.

11. Jerene D, Endale A, Hailu Y, Lindtjørn B. Predictors of earlydeath in a cohort of Ethiopian patients treated with HAART.BMC Infect Dis 2006;6:136.

12. Lawn SD, Myer L, Orrell C, Bekker LG, Wood R. Early mortalityamong adults accessing a community-based antiretroviral ser-vice in South Africa: implications for programme design. AIDS2005;19:2141—8.

13. Stringer JS, Zulu I, Levy J, Stringer EM, Mwango A, Chi B,et al. Rapid scale-up of antiretroviral therapy at primarycare sites in Zambia: feasibility and early outcomes. JAMA2006;296:782—93.

14. Bekker LG, Myer L, Orrell C, Lawn S, Wood R. Rapid scale-upof a community-based HIV treatment service: programme per-formance over 3 consecutive years in Guguletu, South Africa.S Afr Med J 2006;96:315—20.

15. Ivers LC, Kendrick D, Doucette K. Efficacy of antiretroviral ther-apy programs in resource-poor settings: a meta-analysis of thepublished literature. Clin Infect Dis 2005;41:217—24.

16. Weidle PJ, Malamba S, Mwebaze R, Sozi C, Rukundo G, Down-ing R, et al. Assessment of a pilot antiretroviral drug therapyprogramme in Uganda: patients’ response, survival, and drugresistance. Lancet 2002;360:34—40.

17. Wester CW, Kim S, Bussmann H, Avalos A, Ndwapi N, Peter TF,

et al. Initial response to highly active antiretroviral therapy inHIV-1C-infected adults in a public sector treatment program inBotswana. J Acquir Immune Defic Syndr 2005;40:336—43.

18. Department of Local and Provincial Government. Bushbuck-ridge Nodal Development Profile. Pretoria: Department of

593

Local and Provincial Government, Republic of South Africa;2007.

19. Pronyk PM, Kahn K, Hargreaves JR, Tollman SM, CollinsonM, Hausler HP, et al. Undiagnosed pulmonary tuberculosisdeaths in rural South Africa. Int J Tuberc Lung Dis 2004;8:796—9.

20. National Department of Health. National ARV TreatmentGuidelines. Pretoria: National Department of Health, Republicof South Africa; 2006.

21. Kahn K, Tollman S, Garenne M, Gear J. Validation and appli-cation of verbal autopsies in a rural area of South Africa. TropMed Int Health 2000;5:824—31.

22. Murdoch DM, Venter WD, Feldman C, Van Rie A. Incidence andrisk factors for the immune reconstitution inflammatory syn-drome in HIV patients in South Africa: a prospective study. AIDS2008;22:601—10.

23. Badri M, Wilson D, Wood R. Effect of highly active antiretroviraltherapy on incidence of tuberculosis in South Africa: a cohortstudy. Lancet 2002;359:2059—64.

24. Lawn S, Myer L, Bekker L, Wood R. Burden of tuberculosis inan antiretroviral treatment programme in sub-Saharan Africa:impact on treatment outcomes and implications for tuberculo-sis control. AIDS 2006;20:1605—12.

25. Lawn S, Harries A, Anglaret X, Myer L, Wood R. Early mortalityamong adults accessing antiretroviral treatment programmesin sub-Saharan Africa. AIDS 2008;22:1897—908.

26. Martinson NA, Karstaedt A, Venter WD, Omar T, King P, MbengoT, et al. Causes of death in hospitalized adults with a pre-mortem diagnosis of tuberculosis: an autopsy study. AIDS2007;21:2043—50.

27. Lawn SD, Myer L, Bekker LG, Wood R. Tuberculosis-associatedimmune reconstitution disease: incidence, risk factors andimpact in an antiretroviral treatment service in South Africa.AIDS 2007;21:335—41.

28. Grant AD, Charalambous S, Fielding KL, Day JH, Corbett EL,Chaisson RE, et al. Effect of routine isoniazid preventive ther-apy on tuberculosis incidence among HIV-infected men in SouthAfrica: a novel randomized incremental recruitment study.JAMA 2005;293:2719—25.

at University of L

iverpool on May 1, 2015

.org/