Initiation of antiretroviral therapy in HIV-infected tuberculosis patients in rural Kenya: an observational study


Corresponding Author Alexander J. Stockdale, Regional Infectious Diseases Unit, Western General Hospital, Crewe Road South, Edinburgh EH4 2XU, UK. E-mail:



To provide information on the effect of timing of antiretroviral therapy (ART) initiation on outcomes of TB infection in real-life, non-clinical trial, rural settings in sub-Saharan Africa.


We conducted an observational cohort study of all HIV-infected TB patients presenting to a rural hospital in Kenya between 2005 and 2009. We analysed the association between timing of initiation of ART and mortality, using a Cox regression survival analysis, adjusted for measured confounders.


A total of 404 antiretroviral-naïve HIV/TB coinfected patients were included in the study. Initiation of ART during the first 8 weeks of TB treatment (early group) was not associated with changes in mortality at 1 year compared with initiation of ART after 8 weeks (late group) [Hazard Ratio (HR) = 0.74 (Confidence Interval (CI), 0.33–1.64, P = 0.46]. In patients with baseline CD4 counts ≤50 cells/μl, there was a significant reduction in mortality in the early group compared with the late group (HR = 0.20; 95% CI, 0.042–0.99; P = 0.049). In patients with a CD4 count >50 cells/μl, there was no significant difference between early and late groups (HR 1.79; 95% CI, 0.64–5.03; P = 0.27).


We found that in HIV/TB coinfected patients in rural Kenya, early ART initiation (within 8 weeks) was associated with reduced mortality in those with CD4 counts ≤50 cells/μl. In patients with CD4 counts >50 cells/μl, there was no association seen between timing of ART and mortality.



Fournir des informations sur l'effet du moment de l'initiation de la thérapie antirétrovirale (ART) sur les résultats de l'infection tuberculeuse dans la vie réelle, essai non-clinique en zones rurales d'Afrique subsaharienne.


Nous avons mené une étude de cohorte observationnelle de tous les patients TB infectés par le VIH, se présentant dans un hôpital rural au Kenya entre 2005 et 2009. Nous avons analysé l'association entre le moment de l'initiation de l’ART et la mortalité, en utilisant une analyse de survie selon la régression de Cox, ajustée pour les variables confusionnelles mesurées.


404 patients coinfectés par le VIH/TB et naïfs pour le traitement antirétroviral ont été inclus dans l’étude. L'initiation de l’ART au cours des 8 premières semaines de traitement de la TB (groupe précoce) n'a pas été associée avec des changements dans la mortalité à un an comparée à l'initiation de l’ART après 8 semaines (groupe tardif) [rapport de risque (HR) = 0,74 (intervalle de confiance (IC) de 0,33 à 1,64; = 0,46]. Chez les patients ayant des taux de base de CD4 ≤ 50 cellules/μl, il y avait une réduction significative de la mortalité dans le groupe précoce par rapport au groupe tardif (HR = 0,20; IC 95%: 0,042–0,99; = 0,049). Chez les patients ayant un taux de CD4 > 50 cellules/μl, il n'y avait aucune différence significative entre les groupes ‘précoce’ et ‘tardif’ (HR = 1,79; IC95% : 0,64 à 5,03; = 0,27).


Nous avons constaté que chez les patients coinfectés par le VIH/TB dans les zones rurales du Kenya, l'initiation précoce de l’ART (dans les 8 semaines) était associée à une diminution de la mortalité chez ceux avec des taux de CD4 ≤ 50 cellules/μl. Chez ceux avec une numération des CD4 > 50 cellules/μl, il n'y avait aucune association entre le moment de l'initiation de l’ART et la mortalité.



Proveer información sobre el efecto que tiene el momento escogido para iniciar la terapia antirretroviral (TAR) sobre la infección de TB, en emplazamientos rurales de África subsahariana, bajo condiciones reales y sin asociación a un ensayo.


Hemos realizado un estudio observacional de cohortes de todos los pacientes con TB infectados con VIH, atendidos en un hospital rural de Kenia entre el 2005 y el 2009. Hemos analizado la asociación entre el momento de iniciar el TAR y la mortalidad, haciendo un análisis de la supervivencia mediante una regresión de Cox, ajustada para posibles factores de confusión.


Se incluyeron 404 pacientes con coinfección VIH/TB y que no habían recibido previamente terapia antirretroviral (TAR). El comienzo del TAR durante las 8 primeras semanas del tratamiento para la TB (grupo temprano), no estaba asociado a cambios en la mortalidad tras un año, comparado con el comienzo del TAR después de 8 semanas (grupo tardío) [razón de riesgo (RR) = 0.74 (IC 95% 0.33–1.64, = 0.46]. En pacientes con conteos de CD4 ≤ 50 células/μl al comienzo del estudio, había una reducción significativa de la mortalidad en el grupo temprano comparado con el grupo tardío (RR = 0.20, 95% IC 0.042–0.99, = 0.049). En pacientes con un conteo de CD4 > 50 células/μl no había una diferencia significativa entre los grupos temprano y tardío (RR 1.79 IC 95% 0.64–5.03, = 0.27)


Hemos encontrado que en pacientes coinfectados con VIH /TB en zonas rurales de Kenia, un comienzo temprano del TAR (dentro de las 8 primeras semanas) estaba asociado con una reducción en la mortalidad en pacientes con conteos de CD4 ≤ 50 células/μl. En pacientes con conteos de CD4 > 50 células/μl no se observaba una asociación entre el momento de comenzar el TAR y la mortalidad.


In sub-Saharan Africa, human immunodeficiency virus (HIV) infection has had a devastating effect on the tuberculosis (TB) epidemic, fuelling its transmission and causing higher rates of morbidity and mortality (Lawn & Churchyard 2009; WHO 2010b). TB is the most common opportunistic infection and cause of death in people infected with HIV in this setting (Division of Leprosy, Tuberculosis & Lung Disease 2006; Cohen & Meintjes 2010; Kyeyune et al. 2010). Kenya is one of the 22 high TB-burden countries in the world, with an annual incidence of 305 per 100 000 population per year (WHO 2010a). Forty-four percent of patients with TB in Kenya were estimated to be infected with HIV in 2010 (WHO 2010a).

The optimal time to initiate antiretroviral therapy (ART) in patients with HIV-associated tuberculosis has been the subject of intense debate. Concerns about early initiation of ART include a high pill burden, pharmacological interactions, overlapping toxicities and the immune reconstitution inflammatory syndrome (IRIS). Conversely, delayed initiation of ART may be associated with HIV disease progression and death (Lawn & Churchyard 2009). Previous retrospective studies (Manosuthi et al. 2006; Velasco et al. 2009) and a randomised trial showed that delaying ART initiation until after completion of TB therapy was associated with increased mortality (Abdool Karim et al. 2010). More recently, three randomised controlled trials (RCT) have demonstrated a survival benefit of early initiation of ART (2–4 weeks into TB therapy) in patients with a baseline CD4 count <50 cells/μl, over delayed initiation (Abdool Karim et al. 2011; Blanc et al. 2011; Havlir et al. 2011). A fourth RCT in patients with HIV-associated tuberculous meningitis showed no evidence of benefit of early ART initiation (Török et al. 2011).

Both of the trials that included patients from sub-Saharan Africa were based in well-resourced teaching hospitals in urban centres. The majority of HIV-infected TB patients reside in rural areas of sub-Saharan Africa where laboratory diagnostic facilities are limited, and TB diagnosis is clinical. The aim of this study was therefore to explore the relationship between the timing of ART initiation and outcome in HIV-infected TB patients in a typical rural sub-Saharan Africa setting.


Study setting and participants

We performed an observational cohort study based on data from the registers of the Tuberculosis Clinic and the HIV clinic managed by AIDS Relief at PCEA Chogoria Hospital, a rural mission hospital in Eastern Province, Kenya. TB treatment clinic data were recorded in a standardised Kenyan Ministry of Health TB treatment register, and HIV clinic data were recorded electronically in a computerised database at each patient visit. Data were collected for the period between 1 January 2005 and 31 October 2009. Patients were included if they had completed tuberculosis treatment by the end date of the study (31 October 2009) or if an alternative outcome, such as death or loss to follow-up, had been recorded.

Clinical procedures

Patients with TB were diagnosed and treated free of charge through the programme of the Kenya Ministry of Public Health and Sanitation Division of Leprosy, Tuberculosis and Other Lung Diseases (Division of Leprosy, Tuberculosis & Lung Disease 2009). At least two sputum samples from spontaneous expectoration were tested for acid-fast bacilli using Ziehl-Nielsen staining and microscopy. Mycobacterial culture and drug susceptibility testing were not available, apart from retreatment cases whose sputum samples were sent to the National Tuberculosis Reference Laboratory. The majority of patients underwent chest radiography and the decision to start TB treatment was at the discretion of the hospital doctors, based on national guidelines (Division of Leprosy, Tuberculosis & Lung Disease 2009). The diagnosis of sputum smear-negative pulmonary TB was based on a history of cough of 2 weeks or longer, fever or weight loss and compatible radiological features. Extrapulmonary tuberculosis was diagnosed on the basis of a history of weight loss, night sweats or persistent fever with pleural or pericardial effusion, miliary appearance on radiographs, lymphadenopathy, chronic lymphocytic meningitis, ascites or radiographic bone abnormalities. Access to mycobacterial culture was extremely limited and did not influence the decision to treat. In most cases, treatment was empirical and based upon clinical suspicion without diagnostic confirmation.

The treatment regimen was rifampicin, isoniazid, pyrazinamide and ethambutol for 2 months (intensive phase) followed by rifampicin and isoniazid for 4 months (continuation phase) in accordance with national TB treatment guidelines.

HIV infection status was determined using a rapid diagnostic test, and positive results were confirmed using a second test, in accordance with national HIV guidelines. From 2006, the Division of Leprosy, Tuberculosis and Other Lung Diseases in Kenya advised that all patients attending clinics which administer TB chemotherapy should be tested for HIV infection using an opt-out approach (Division of Leprosy, Tuberculosis & Lung Disease 2006).

The HIV clinic provided cotrimoxazole prophylaxis and multivitamins free of charge to all HIV-infected patients enrolled into the programme, regardless of CD4 count. ART was also provided free of charge to all patients who met the immunological and psychological criteria for initiation. Immunological criteria were WHO HIV stage 3 or 4 disease or stage 1 or 2 disease if the CD4 count was <250 cells/μl. Psychological criteria were demonstration of sufficient understanding, psychological preparedness and capacity for adherence. The ART regimen was stavudine or zidovudine plus lamivudine and efavirenz. CD4 counts were monitored 6-monthly, and plasma HIV viral load testing was not available during the course of the study.

Statistical analysis

Statistical analysis was performed using SPSS software version 17 (2008). Categorical variables were tested using the Fisher's exact test, and continuous variables using the independent samples Student's t-test or Mann–Whitney U-test, as appropriate, to determine differences in patient demographics based on HIV testing status, outcomes, the decision to start ART or timing of ART initiation.

Survival times were evaluated using a proportional hazards regression model adjusted for measured covariates using the Cox regression survival function. Outcomes were restricted to a period of 1 year from commencement of TB treatment. Patients were divided into three groups according to timing of initiation of ART in relation to TB treatment: ‘early’ meaning starting ART within 8 weeks of commencing TB treatment, ‘late’ meaning starting ART later than 8 weeks but within 1 year and ‘not started’ meaning ART was not started within 1 year of starting TB treatment. The Cox proportional hazards regression model was used to adjust for seven variables: date of initiation of TB treatment, age, gender, baseline CD4 count; type of TB (smear positive or smear negative, pulmonary or extrapulmonary), type of patient (new, retreatment or transfer in) and ART timing treatment group (as described above). Patients were analysed in two subgroups according to their baseline CD4 count at the start of TB treatment: ≤50 or >50 cells/μl.

Ethics statement

The study was approved by the Hospital Administration Team of Presbyterian Church of East Africa Chogoria Hospital, which is the ethical review board for the hospital. The study was conducted as service evaluation of the TB clinic and individual patient consent for participation in the study was therefore not required.


Between 1 January 2005 and 31 October 2009, 1729 patients were treated for tuberculosis at the hospital clinic. HIV testing was performed in 1247 patients (72.1%), of whom 849 (49.1% of total, 68.1% of those tested) were HIV infected.

The proportion of patients tested for HIV infection rose significantly during the study period from 362/688 (52.6%, 95% CI, 48.8–56.4%) in 2005 to 192/204 (94.1%, 95% CI, 90.0–96.9%, P < 0.001) in 2009, while the proportion of those tested who were HIV positive fell from 307/362 (84.8%, 95% CI, 80.7–88.3%) to 91/192 (47.4%, 95% CI, 40.2–54.7%, P < 0.001). Females were more likely to be tested for HIV infection (618/788, 78.4%) compared with males (627/939, 66.8%) relative risk (RR) = 1.18 (95% CI, 1.11–1.25, P < 0.001).

The flow of patients through the study (Figure 1) shows that 849 patients were identified as HIV/TB coinfected. Four hundred and four antiretroviral-naïve HIV/TB coinfected patients were followed up at the local HIV clinic and included in the study. The baseline characteristics of the 404 participants and association with ART are shown in Table 1. During the study follow-up period, 314 (77.7%) patients were treated with ART and 90 patients did not receive ART. ART was significantly associated with CD4 count at the start of TB treatment: median 111 cells/μl (IQR 42–261 cells/μl) in those treated with ART vs. 203 cells/μl (IQR 30–538 cells/μl) in those not treated (P = 0.023). There was no significant association between ART and age, gender, type of TB or type of patient (Table 1).

Table 1. Baseline characteristics and association with antiretroviral therapy in included patients
 n (%)Treated with ART
Early (<8 weeks)Late (>8 weeks)Not started (at 1 year)P-value
  1. a

    Not recorded for 42 patients.

  2. b

    Not recorded for two patients.

Age (years)
Median age (Interquartile range)32.0 (24.0–39.8)34.5 (28.0–41.3)31.0 (23.0–40.0)30.0 (20.0–36.0)0.003
<1877 (19.1)17 (13.9)31 (19.9)29 (23.0)0.395
18–30110 (27.2)24 (19.7)46 (29.5)40 (31.7)
31–50181 (44.8)70 (57.4)67 (42.9)54 (42.9)
>5036 (8.9)11 (9.0)12 (7.7)3 (2.4)
CD4 count at start of TB treatmenta
Median CD4 (Interquartile range)122 (41–295)92 (30–172)130 (48–277)241 (41–476))<0.001
≤50107 (29.6)42 (35.9)38 (26.4)27 (26.7)<0.001
51–200122 (33.7)52 (44.4)50 (24.7)20 (19.8)
201–35059 (16.3)16 (13.7)30 (20.8)13 (12.9)
>35074 (20.4)7 (6.0)26 (18.1)41 (40.6)
Male192 (47.5)60 (49.2)73 (46.8)59 (46.8)0.92
Female212 (52.5)62 (50.8)83 (53.2)67 (53.2)
Type of TB
Smear-positive pulmonary83 (20.5)20 (16.4)31 (19.9)32 (25.4)0.281
Smear-negative pulmonary285 (70.5)93 (76.2)108 (69.2)84 (66.7)
Extrapulmonary36 (8.9)10 (7.4)17 (10.9)10 (7.9)
Type of patientb
New patient338 (83.7)101 (82.8)128 (82.1)109 (86.5)0.490
Retreatment54 (13.4)15 (12.3)24 (15.4)15 (11.9)
Transfer in10 (2.5)5 (4.1)4 (2.6)1 (0.8)
Figure 1.

Flowchart of patients included in study according to inclusion criteria.

Outcomes were significantly worse in those who did not receive ART; mortality was three times greater after 1 and 3 years, and there was a significantly greater loss to follow-up (Table 2). In a univariate analysis, mortality was associated with smear positivity (mortality after 3 years 34.3% for smear-positive disease vs. 19.3% for smear negative, P = 0.013) and CD4 count at ART initiation (mortality after 3 years 26.2% for CD4 count ≤50 cells/μl vs. 12.2% for CD4 count >50 cells/μl, P = 0.002). There was no association with age, gender or proportion of retreatment cases.

Table 2. Relationship between antiretroviral therapy and treatment outcomes in HIV-infected TB patients
Time since starting TB treatmentOutcomes (%)Treated with ART?Patients treated with ART vs. patients not treated with ART
YesNoRelative risk (95% CI)P-value
  1. a

    Outcome known = Number of patients at the given time period for whom the outcome is known according to duration of follow-up. Follow-up ranged from 6 months to 5 years, depending on time of entry to study.

  2. b

    Remaining under follow-up = Continuing to attend regular HIV clinics.

One yearOutcome knowna297 (95.2)86 (95.6)  
Remaining under follow-upb247 (83.2)41 (47.7)0.57 (0.46–0.72)<0.001
Died30 (10.1)29 (33.7)3.34 (2.13–5.24)<0.001
Transferred out13 (4.4)13 (15.1)3.45 (1.66–7.17)0.001
Lost to follow-up5 (1.7)3 (3.5)2.07 (0.51–8.50)0.386
Stopped ART2 (0.7) 
Two yearsOutcome knowna276 (87.9)79 (87.8)  
Remaining under follow-upb208 (75.4)22 (27.8)0.37 (0.26–0.53)<0.001
Died36 (13.0)30 (38.0)2.91 (1.92–4.41)<0.001
Transferred out19 (6.9)18 (22.8)3.31 (1.83–6.00)<0.001
Lost to follow-up11 (4.0)9 (11.4)2.86 (1.23–6.65)0.023
Stopped ART2 (0.7) 
Three yearsOutcome knowna241 (76.8)75 (83.3)  
Remaining under follow-upb159 (66.0)15 (20.0)0.30 (0.19–0.48)<0.001
Died40 (16.6)31 (41.3)2.49 (1.69–3.68)<0.001
Transferred out24 (10.0)19 (25.3)2.54 (1.48–4.38)0.002
Lost to follow-up13 (5.4)10 (13.3)2.47 (1.13–5.41)0.038
Stopped ART5 (2.1) 

The 404 patients were divided into three categories: the early group [n = 122, median time to ART initiation = 42 days (IQR 25–55 days)]; the late group [n = 156, median time to ART initiation = 116 days (IQR 81–210 days)]; and the not started group.

Earlier initiation of ART was associated with increased age: early group median age 34.5 (IQR = 28.0–41.3) years; late group median age 31.0 (IQR = 23.0–40.0) years, P = 0.023. Patients who started ART earlier also had significantly lower CD4 counts at the start of TB treatment: early group median CD4 count 92 cells/μl (IQR 30–172 cells/μl); late group median CD4 count 130 cells/μl (IQR 48–277 cells/μl), P = 0.006. There was no association with year, gender, type of TB or type of patient as seen in Table 1.

In a Cox proportional hazards regression analysis, there was no difference in mortality after 1 year between the early and late groups, Hazard Ratio (HR) 0.74 (95% CI, 0.33–1.64; P = 0.46) (Figure 2). However, mortality after 1 year was significantly lower in patients who received ART compared with those who did not (HR = 0.14; 95% CI, 0.078–0.26; P < 0.001).

Figure 2.

Cox regression survival curve: timing of antiretroviral therapy in relation to TB treatment. Curves are adjusted for date of commencing TB treatment, age, gender, type of TB (pulmonary smear positive or smear negative or extrapulmonary), type of patient (new, retreatment or transfer in) and baseline CD4 count.

In patients with a baseline CD4 count of <50 cells/μl, there was a significant reduction in mortality in the early group compared with the late group (HR = 0.20; 95% CI, 0.042–0.99; P = 0.049). Those who received ART in the early group had a 95% absolute reduction in mortality compared with those who did not receive ART within 1 year (HR 0.050; 95% CI, 0.011–0.22; P < 0.0001), (Figure 3a). In patients with a baseline CD4 count >50 cells/μl, there was no significant difference between the early and late groups (HR 1.79; 95% CI, 0.64–5.03; P = 0.27) nor between early and not started groups (HR 1.06; 95% CI, 0.33–3.38; P = 0.92) (Figure 3b).

Figure 3.

Cox regression survival curve: timing of antiretroviral therapy in relation to TB treatment according to CD4 count at baseline. (a) Patients with baseline CD4 count ≤50 cells/μl. (b) Patients with baseline CD4 count >50 cells/μl. Curves are adjusted for date of commencing TB treatment, age, gender, type of TB (pulmonary smear positive or smear negative or extrapulmonary), type of patient (new, retreatment or transfer in) and baseline CD4 count.


Our study validates previous RCT findings in a non-experimental setting in rural sub-Saharan Africa, the setting where the majority of patients with HIV-TB coinfection live and are managed. Overall, we found no mortality difference between starting ART early (within 8 weeks of commencing TB treatment) and deferring ART until later. However, in patients with a baseline CD4 count ≤50 cells/μl, we observed a fivefold increase in survival in those starting ART in the first 8 weeks compared with those that started ART later that was of borderline significance.

Four randomised controlled trials have examined the question of the optimal timing of antiretroviral therapy in HIV/TB coinfected patients. In 2010, the SAPIT trial from South Africa reported preliminary data showing a relative reduction of mortality of 56% (HR = 0.44; 95% CI, 0.25–0.79; P = 0.003) in patients starting ART during TB treatment (integrated therapy arm) compared with patients starting within 4 weeks of completion of TB treatment (sequential therapy arm) (Abdool Karim et al. 2010) This accords with previous observational data from Spain (Velasco et al. 2009) and Thailand (Manosuthi et al. 2006) that reported increased mortality in people deferring ART until the completion of TB treatment, and an observational study in children from South Africa suggesting increased mortality in those who started ART later than 2 months in TB treatment (Yotebieng et al. 2010). The final analysis of the SAPIT trial demonstrated no overall difference in the primary endpoint of AIDS-defining illness or death between the early integrated therapy arm (ART commenced <4 weeks of TB treatment) and late integrated therapy arm (ART commenced <4 weeks after completing intensive phase of TB treatment) (Abdool Karim et al. 2011).

A second international multicentre RCT, the ACTG 5221 study, reported that immediate ART (<2 weeks) did not reduce the primary endpoint of AIDS-defining illness or death compared with early ART (8–12 weeks) (Havlir et al. 2011). However, both studies found that in patients with baseline CD4 counts <50 cells/μl, early ART initiation was associated with a significant reduction in AIDS or death.

A third RCT, the CAMELIA trial conducted in Cambodia reported improved survival in predominantly pulmonary patients with TB initiating ART at 2 weeks compared with 8 weeks (Blanc et al. 2011). The median CD4 count was 25 cells/μl (IQR 10–56 cells/μl), which was significantly lower than in the other two studies and may account for these findings.

In contrast, a Vietnamese RCT of immediate (within 7 days of TB treatment) vs. deferred (after 8 weeks TB treatment) ART initiation in patients with HIV-associated TB meningitis showed no survival benefit, and an increase in severe adverse events, in the immediate ART arm (Török et al. 2011). Collectively, the results of these four trials suggest that ART should be started early in HIV/TB coinfected patients with advanced immunosuppression, apart from those with TB meningitis (Török & Farrar 2011).

We found that deferring ART in patients with baseline CD4 counts >50 cells/μl until after 8 weeks was associated with no difference in mortality. Early commencement of ART was noted to cause TB-IRIS two to five times more frequently in three of the RCTs. The diagnosis and treatment of TB-IRIS may not be as readily feasible outside a clinical trial setting in a rural sub-Saharan African hospital (Colebunders et al. 2006; Leone et al. 2010).

Our study provides further evidence that CD4 count is the best indicator of when to start ART in patients with TB. This has important implications as CD4 count measurement is still not widely available in many low- and middle-income countries. The potential to guide timing of ART and greatly reduce mortality in the huge number of TB coinfected patients should act as further impetus to broaden access to this vital technology.

The strength of this study is that it is a validation of evidence from RCTs conducted in a resource-limited setting where cross-sectional imaging, mycobacterial culture, HIV viral load measurements and other diagnostics are not available. Such a setting is typical of where the majority of HIV/TB coinfected patients reside and present to healthcare facilities. The main limitation of our study is that it is observational in nature, relying on routinely collected data, and there may be unmeasured confounders accounting for mortality that influenced the physician's decision of when to start ART. Figure 2 reveals that the not treated group had a high initial mortality in the first 180 days, while those who deferred ART until after 180 days did not share the same initial high mortality. This suggests that the not treated group may have been more unwell with more advanced TB, other opportunistic infections or comorbidities or poor adherence to treatment. We did not have sufficient data to identify the reasons why patients who fulfilled criteria for ART initiation were not treated, whether patients interrupted ART, whether patients had concomitant opportunistic infections, whether patients developed TB-IRIS or what were the causes of death. Such information may have helped us to further characterise subgroups in which the timing of ART was important.

Another potential limitation is that TB is diagnosed clinically in rural settings. In contrast to the RCTs mentioned previously, the diagnosis of TB was only confirmed by sputum smear in about 30% of the cohort, which means that TB may have been overdiagnosed. Although TB is epidemiologically the most likely cause of chronic respiratory symptoms, other opportunistic infections such as Pneumocystis jirovecii and Cryptococcus neoformans may present in a similar fashion. Furthermore, some of the smear-positive cases may have been due to non-tuberculous mycobacteria which account for up to 3.6% of cases of acute or chronic cough in Kenya (Githui et al. 1992; Scott et al. 2000). Nevertheless, we believe that this study provides important information from a rural sub-Saharan African setting where TB is common and frequently diagnosed and treated on the basis of clinical criteria.

It was encouraging to see that over the 5-year period, the rate of HIV testing increased from just over half to 94%. HIV testing of patients with TB is important both on a public health level to reduce HIV transmission and for the individual patient as it enables delivery of cotrimoxazole prophylaxis and ART, with the potential to greatly increase survival. The recommendation of opt-out testing for patients with TB could be of great benefit (Pope et al. 2008).

This study also highlights the need to integrate HIV and TB clinical services. Only 60% of patients receiving tuberculosis treatment at our hospital were followed up in the local HIV clinic. It is not clear whether these patients received follow-up from an alternative facility, refused treatment or defaulted HIV clinic follow-up. The integration of HIV and TB services in a single clinic would enable better coordination of HIV and TB treatment, initiation of cotrimoxazole prophylaxis, management of complications and reduce loss to follow-up. Indeed evidence from Western Kenya has shown that combining the HIV and TB outpatient clinics into one combined clinic significantly increased the uptake of cotrimoxazole prophylaxis and ART and significantly reduced mortality rates (Huerga et al. 2010).

In conclusion, we have found that in HIV/TB coinfected patients with baseline CD4 counts <50 cells/μl in rural Kenya, ART had a significant reduction in mortality when started during the intensive phase of the first 8 weeks of TB treatment compared with starting after 8 weeks. We also found that only 60% of patients with TB who tested positive for HIV were followed up in the local HIV clinic thus highlighting the urgent need to integrate HIV and TB services in rural settings so that patients can benefit from life-saving treatments.


We acknowledge the Clinical Officers and Nursing staff of the TB and HIV Clinics at PCEA Chogoria Hospital for their hard work and assistance in conducting this study. MET acknowledges the support of the NIHR Cambridge Biomedical Research Centre.