The authors have declared that no competing interests exist.
Conceived and designed the experiments: WA GA AH GY EM AW AS EA LL. Performed the experiments: WA GA AW AS LL. Analyzed the data: WA GA AH GY EM AW AS EA LL. Contributed reagents/materials/analysis tools: GA AW AS LL. Wrote the paper: WA GA AH GY EM AW AS EA LL.
Given the high death rate the first two months of tuberculosis (TB) therapy in HIV patients, it is critical defining the optimal time to initiate combination antiretroviral therapy (cART).
A randomized, open-label, clinical trial comparing efficacy and safety of efavirenz-based cART initiated one week, four weeks, and eight weeks after TB therapy in patients with baseline CD4 count < 200 cells/μL was conducted. The primary endpoint was all-cause mortality rate at 48 weeks. The secondary endpoints were hepatotoxicity-requiring interruption of TB therapy, TB-associated immune reconstitution inflammatory syndrome, new AIDS defining illnesses, CD4 counts, HIV RNA levels, and AFB smear conversion rates. All analyses were intention-to-treat.
We studied 478 patients with median CD4 count of 73 cells/μL and 5.2 logs HIV RNA randomized to week one (n = 163), week four (n = 160), and week eight (n = 155). Sixty-four deaths (13.4%) occurred in 339.2 person-years. All-cause mortality rates at 48 weeks were 25 per 100 person-years in week one, 18 per 100 person-years in week four and 15 per 100 person-years in week eight (P = 0.2 by the log-rank test). All-cause mortality incidence rate ratios in subgroups with CD4 count below 50 cells/μL versus above were 2.8 in week one (95% CI 1.2–6.7), 3.1 in week four (95% CI 1.2–8.6) and 5.1 in week eight (95% CI 1.8–16). Serum albumin < 3gms/dL (adjusted HR, aHR = 2.3) and CD4 < 50 cells/μL (aHR = 2.7) were independent predictors of mortality. Compared with similar subgroups from weeks four and eight, first-line TB treatment interruption was high in week one deaths (P = 0.03) and in the CD4 subgroup <50 cells/μL (P = 0.02).
Antiretroviral therapy one week after TB therapy doesn’t improve overall survival. Despite increased mortality with CD4 < 50 cells/μL, we recommend cART later than the first week of TB therapy to avoid serious hepatotoxicity and treatment interruption.
ClinicalTrials.gov
Tuberculosis (TB) is the most frequent cause of death in HIV infected patients. People living with HIV are 27–32 times more likely to develop TB than HIV-negatives. In 2012, 30% of the 1.1 million new TB cases in HIV-positives died. Of the new cases, 75% were in Africa [
We conducted a randomized, controlled trial of one week versus four and eight weeks initiation of cART in TB/HIV co infected patients with CD4 count < 200cells/ μL, to determine whether one week reduces the risk of death.
Multicenter, open-label, parallel, randomized clinical trial was conducted in Addis Ababa, Ethiopia from June 2, 2008 until April 22, 2011. The trial started after the Institutional Review Board of the School of Medicine at Addis Ababa University, the Ethiopian Science and Technology Agency as well as Food, Medicine and Health Care Administration and Control authority of Ethiopia approved the study protocol. The study was officially stopped after the last scheduled follow up visit. It was registered at
The patients were recruited from four trial sites within Addis Ababa. The main trial site was Tikur Anbessa Specialized Referral Hospital and others were health centres. Eligible subjects were ambulatory with suspected or confirmed new TB diagnosis. They were 18 years or older with confirmed HIV infection (CD4 count < 200 cells/μL) and written informed consent obtained on enrolment. The diagnosis of TB was based on WHO criteria for the diagnosis of smear positive pulmonary TB (SP-PTB), smear negative pulmonary TB (SN-PTB) and extrapulmonary TB (EPTB) [
Exclusion criteria were TB of the central nervous system, Karnofsky score < 40, serum alanine aminotransferase more than 3 times the upper limit of normal (ULN), hemoglobin < 8 gms/dL, previous antiretroviral therapy exposure and pregnancy.
The study subjects received first-line TB therapy with weight adjusted fixed dose combinations of rifampicin, isoniazid, ethambutol, and pyrazinamide for 2 months (intensive phase) and a subsequent fixed dose combination of isoniazid and rifampicin for 4 months (continuation phase). The study participants were randomized in to three groups. The first received cART one week (week one), the second four weeks (week four) and the third eight weeks (week eight) after TB therapy started. The median number of TB treatment days before cART was initiated defined the different study groups. The cART regimen included efavirenz (600 mg once daily) plus lamivudine, and study site physician selection of zidovudine, stavudine or tenofovir. Time to initiate cART was deferred in grade 3 and 4 hepatotoxicity events and while first-line TB therapy interrupted due to hepatotoxicity. All patients received cotrimoxazole 960 mgs once daily. The participants were assessed at enrolment (baseline), and subsequently in weeks 1, 2, 4, 6, 8, 12, 20, 24 and 48 weeks. At each visit complete blood count, serum levels of aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase (ALP), bilirubin (total and direct), urea and creatinine were measured. At 12 weeks intervals, CD4 count and HIV RNA levels were determined until the 24th week, and then at 48th week. Hepatitis B surface Antigen (HBsAg) and Hepatitis C Virus Antibody (HCV Ab) status were confirmed at baseline. Sputum for acid-fast bacilli (AFB) was examined with Ziehl-Neelsen stain when PTB was first diagnosed and at eighth week of TB therapy to evaluate smear conversion. For the diagnosis of TB-associated IRIS, a case definition for use in resource-limited settings was applied [
The primary hypothesis was initiation of cART one week after TB therapy reduces all-cause mortality rate at 48 weeks compared with week four as well as week eight. All-cause mortality rates from the three study groups are compared. The primary end point was all-cause mortality rate assessed 48 weeks after study entry. Cause of death was abstracted from medical records or from verbal autopsy data. It was classified AIDS-related if an AIDS- defining illness (category C disease listed by the CDC and WHO stage IV events), pneumonia or sepsis was listed as contributing cause. The secondary end points were: incidence rates of hepatotoxicity requiring interruption of first-line TB therapy, TB-IRIS and AIDS defining illnesses (ADIs), HIV-related outcomes (increase in the median CD4 counts and proportions of patients with undetectable HIV-1 RNA i.e. ≤ 400 copies/ mL) and TB-treatment outcome (AFB smear converted negative at eighth week).
The estimated sample size was 450 patients (factoring in anticipated 20% loss). It has 80% power, alpha level of 0.05 to detect 50% reduction in mortality from predicted CFR of 25% when cART is started after the 8th week of TB therapy. All statistical analyses were according to the intention-to-treat principles. The primary and secondary end points were analysed with Kaplan-Meier curves and log-rank test. The duration of time in the study was calculated as the time from TB treatment to death, lost to follow-up (LTF), or completion of the 48 weeks of the study which ever occurred first. For continuous outcome variables, multiple levels comparisons were made with one-way ANOVA. Cox proportional-hazards model was used to explore the probability of death before and after TB therapy adjusting for confounding variables and factors of interest. The factors which had P-value ≤ 0.1 in univariate analysis and of interest were included in the multivariate model. Pearson chi-square test was applied for secondary endpoint analysis. Assumptions for each analytical method were checked. SPSS for windows version 20 and STATA 11 software were used for the statistical analysis. P-values less than 0.05 were considered statistically significant.
Eligible participants were assigned to one of the three study weeks with simple randomization. Each site prepared separate randomization list. The random order of assignment was determined in advance by lots method.
A total of 478 TB and HIV co-infected patients (233 women) were enrolled in the study: 163 were assigned to week one, 160 to week four and 155 to week eight. Patient enrolments and outcome characteristics are shown in
Variables | week one (163) | week four (160) | week eight (155) |
---|---|---|---|
Age, yrs (mean, sd) | 37(10) | 36(9) | 35(10) |
Male sex, no (%) | 93(57%) | 88(55%) | 64(41%) |
BMI, Kgs/m2 (mean, sd) | 18.9 (2.9) | 18.7(3.1) | 18.6(3) |
Type TB diagnosis (no, %) | |||
SP-PTB | 29(17.8%) | 28(17.5%) | 42(27.1%) |
SN-PTB | 91(55.8%) | 85(53.1%) | 78(50.3%) |
EPTB | 43(26.4%) | 47(29.4%) | 35(22.6%) |
CD4 cells/μL (median,IQR) | 67(39–106) | 71(45–121) | 76(47–110) |
CD4 < 50 cells/μL, no (%) | 61(37.4%) | 50(31.3%) | 39(25.2%) |
CD4 ≥ 50 cells/ μL, no (%) | 102(62.6%) | 110(68.7%) | 116(74.8%) |
Log HIV-RNA (mean, sd) | 4.9(0.7) | 5.1(0.8) | 5.2(0.7) |
Karnofsky score% (median,IQR) | 90(70–100) | 85(70–100) | 90(80–100) |
TB treatment days before cART(median, range) | 7 (4–48) | 28(21–45) | 56(48–74) |
HBsAg positive, no (%) | 16(9.8%) | 14(8.8%) | 11(7.1%) |
HCVAb positive, no (%) | 4(2.5%) | 4(2.5%) | 0 |
Serum albumin, gms/dL (mean, sd) | 3.3 (0.7) | 3.5(0.8) | 3.3(0.7) |
cART regimen: efavirenz/lamivudine and | |||
not started | 13(8%) | 14(8.9%) | 18(11.6%) |
stavudine | 41(25.2%) | 61(38.1%) | 34(21.9%) |
zidovudine | 47(28.8%) | 41(25.6%) | 49(31.6%) |
tenofovir | 62(38%) | 44(27.5%) | 54(34.8%) |
The median (range) numbers of days before cART initiation were 7days (4–48), 28 days (21–45) and 56 days (48–74) in weeks one, four and eight respectively.
There were a total of 64 deaths (Case fatality ratio, CFR = 13.4%) in 339.2 person-years of follow up. Twenty-seven cases were in week one, 20 cases in week four and 17 cases in week eight. In intention-to-treat analysis, all-cause mortality incidence rates were 25 per 100 person-years (95% CI 17.1–36.3) in week one, 18 per 100 person-years (95% CI 11.3–27.2) in week four and 15 per 100 person-years (95% CI 9.1–23.6) in week eight, P-value 0.2 by the log-rank test (
Risk factors | Univariate Analysis | Multivariate Analysis | ||||
---|---|---|---|---|---|---|
Hazard Ratio (HR) | 95% CI | P value | HR | 95% CI | P value | |
< 50 cells/ μL | 3.3 | 2–5.4 | 0.001 | 2.7 | 1.6–4.5 | 0.001 |
51–199 cells/ μL | 1 | 1 | ||||
< 3gms/dL | 2.8 | 1.6–4.9 | 0.001 | 2.3 | 1.3–4 | 0.02 |
≥ 3gms/dL | 1 | 1 | ||||
week one | 1.6 | 0.9–3 | 0.1 | 1.5 | 0.8–2.7 | 0.2 |
week four | 1.2 | 0.6–2.2 | 0.6 | 1.2 | 0.6–2.3 | 0.6 |
week eight | 1 | 1 | ||||
0.9 | 0.8–1 | 0.02 | 0.9 | 0.8–1 | 0.04 | |
2 | 1.1–3.6 | 0.02 | 1.6 | 1–2.9 | 0.1 |
During the study period 278 cases (58.2%) experienced one or more grade 1 and 2 hepatotoxicity events while 71 cases (14.9%) had grade 3 and 4 hepatotoxicity events (
Random week | Hepatotoxicity | TB therapy interrupted due to hepatotoxicity. Total no, (%) | TB therapy interrupted due to hepatotoxicity (mortality cases, no(%) | Days TB therapy was interrupted due to hepatotoxicity, median (range) | |
---|---|---|---|---|---|
Grade 1 & 2 | Grade 3 & 4 | ||||
Week one (n = 163) | 88(54%) | 27(16%) | 22(12.3%) | 11/27 (40.7%) | 14(4–30) |
Week four (n = 160) | 104(65%) | 23(14.4%) | 13(8.1%) | 4/20(20%) | 15(9–30) |
Week eight (n = 155) | 86(55.5%) | 21(13.5%) | 14(9%) | 1/17(5.9%) | 14(8–30) |
Total (n = 478) | 278(58.2%) | 71(14.9%) | 49(10.3%) | 16/64 (25%) | 14(4–30) |
Out comes | Frequency | week one | week four | week eight | P value | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
no | per-yrs | no. events | event rate/100 per-yrs (95% CI) | no | per-yrs | no. events | event rate/100 per-yrs (95% CI) | no | per-yrs | no. events | event rate/100 per-yrs (95% CI) | |||
All patients | 163 | 108.4 | 21 | 26(17.1–36.3) | 160 | 114.5 | 20 | 18(11.3–27.2) | 155 | 116.3 | 17 | 15(9.1–23.6) | 0.2 | |
CD4 < 50 | 61 | 36.8 | 16 | 43.5(26.7–71) | 50 | 32.4 | 11 | 34.5(19.1–62.2) | 39 | 25.3 | 10 | 39.7(21.4–73.8) | 0.9 | |
CD4 ≥ 50 | 102 | 36.8 | 11 | 15.4(8.5–27.7) | 110 | 82.1 | 9 | 11(5.7–21.1) | 116 | 91 | 7 | 7.7(3.7–16.2) | 0.4 | |
IRR |
2.8(1.2–6.7) | 3.1(1.2–8.6) | 5.1(1.8–16) | |||||||||||
All patients | 163 | 108.4 | 27 | 25(17–36) | 160 | 114.5 | 23 | 20(13–30) | 155 | 116.3 | 21 | 18(12–28) | 0.1 | |
CD4 < 50 | 61 | 36.8 | 20 | 54.4(35.1–84.3) | 50 | 32.4 | 5 | 15.4(6.4–37) | 39 | 25.3 | 7 | 27.7(13.2–58.1) | 0.01 | |
CD4 ≥ 50 | 102 | 36.8 | 7 | 9.8(4.7–20.5) | 110 | 82.1 | 18 | 21.9(13.8–34.8) | 116 | 91 | 14 | 15.4(9.1–26) | 0.1 | |
IRR |
5.6(2.3–15.6) | 0.7(0.2–2) | 1.8(0.6–4.8) | |||||||||||
All patients | 163 | 108.4 | 22 | 20.3(13.4–30.8) | 160 | 114.5 | 13 | 11.4(6.6–19.5) | 155 | 116.3 | 14 | 12(7.1–20.3) | 0.2 | |
CD4 < 50 | 61 | 36.8 | 17 | 46.2(28.8–74.4) | 50 | 32.4 | 4 | 12.3(4.6–32.9) | 39 | 25.3 | 6 | 23.8(10.7–52.9) | 0.02 | |
CD4 ≥ 50 | 102 | 36.8 | 5 | 7(2.9–16.8) | 110 | 82.1 | 9 | 11(5.7–21.1) | 116 | 91 | 8 | 8.8(4.4–17.6) | 0.7 | |
IRR |
6.6(2.3–23) | 1.1(0.3–4) | 2.7(0.8–8.9) | |||||||||||
All patients | 163 | 108.4 | 16 | 14.8(9–24) | 160 | 114.5 | 6 | 5.3(2.3–11.7) | 155 | 116.3 | 0 | 0.001 | ||
CD4 < 50 | 61 | 36.8 | 16 | 50 | 32.4 | 6 | ||||||||
CD4 ≥ 50 | 102 | 36.8 | 0 | 110 | 82.1 | 0 | ||||||||
All patients | 163 | 108.4 | 12 | 11.1(6.2–19.5) | 160 | 114.5 | 14 | 12.2(7.2–20.6) | 155 | 116.3 | 13 | 11.2(6.5–19.2) | 0.9 | |
CD4 < 50 | 61 | 36.8 | 12 | 50 | 32.4 | 14 | 39 | 25.3 | 13 | |||||
CD4 ≥ 50 | 102 | 36.8 | 0 | 110 | 82.1 | 0 | 116 | 91 | 0 | |||||
Total | no | Sample avail (no) | Negative sputum (no) | no | Sample avail (no) | Negative sputum (no) | no | Sample avail (no) | Negative sputum (no) | 0.2 | ||||
29 | 18 | 18 | 28 | 23 | 22 | 42 | 37 | 36 | ||||||
CD4 < 50 | 10 | 6 | 6 | 10 | 6 | 5 | 14 | 11 | 11 | |||||
CD4 ≥ 50 | 19 | 12 | 12 | 18 | 17 | 17 | 28 | 26 | 26 |
* Incidence Rate Ratios between subgroups with CD4 counts below 50 and above or equals to 50 cells/μL
** According to the Division of AIDS Table for Grading of Severity of Adult and Pediatric Adverse events (DAIDS AE Grading Table), December 2004 revision,
*** First-line TB treatment interrupted according to American Thoracic Society (ATS) criteria, 2006
# Smear status for eight weeks after TB therapy in smear positive Pulmonary Tuberculosis cases
Total of 22 cases (4.6%) were diagnosed with TB- IRIS. Sixteen (72.7%) had lymphadenopathy either newly emerging or pre-existing increased in size. Six (27.2%) had new or worsening abnormalities either on CXR or ultrasound. The incidence rates of TB-IRIS were 15 per 100 person-years in week one (95% CI, 9–24), 5 per 100 person-years in week four (95% CI, 2.3–11.7) and none in week eight (P-value 0.001 by the log-rank test). The baseline CD4 count was lower than 50cells/μL in the TB-IRIS cases. In none of the cases, the course of their treatment was altered. TB-IRIS was diagnosed only in one of the mortality cases. Seven of the cases (44%) with lymph node TB-IRIS required surgical drainage or aspiration of the abscess and non-steroidal anti-inflammatory drugs to alleviate their symptoms.
There were 39 ADIs documented. These events occurred at median (range) of 4 weeks (1–20 weeks) after TB treatment was started. Twelve of the cases were in week one, 14 in week four and 13 in week- eight. The incidence rates were 11.1 cases per 100 person-years (95% CI 6.2–19.6) in week one, 12.2 cases per 100 person-years (95% CI 7.2–20.6) in week four and 11.2 cases per 100 person-years (95% CI 6.5–19.2) in week eight (P-value 0.9 by the log-rank test,
The median (IQR) CD4 count at 48 weeks was 237 cells/μL (155–342 cells/μL) and there was no significant difference among the different weeks in the median (IQR) CD4 counts increase at 48 weeks (P = 0.9). At 12 weeks, CD4 count increase was the lowest in week eight (P = 0.001) and at 24 weeks in week four (P = 0.001). Suppression of plasma HIV-1RNA levels to undetectable viral load (≤ to 400 copies/mL) was achieved in 91% of the 220 cases for whom data were available at 48 week. The hazard rate (HR) of not achieving undetectable viral load at 48 weeks was 6 per 100 person-years (95% CI 1.9–19) in week one, 12 per 100 person-years (95% CI 6–24) in week four and 12 per 100 person-years in week eight (95% CI 6–22), P-value 0.5 by the log-rank test. But at 24 weeks the HR were 7 per 100 person-years (95% CI 3–16) in week one, 23 per 100 person-years (95% CI 15–35) in week four and 27 per 100 person-years (95% CI 18–41) in week eight, P-value 0.004 by the log-rank test.
Seventy-eight participants from the 99 SP-PTB cases (79%) were able to produce sputum at eight week of TB therapy and 76 (77%) had negative smear for AFB. The proportion of cases whose smear converted negative was not different among the study weeks (P = 0.2,
There are several new and important findings in our study related to initiation of cART one week after TB therapy compared with four and eight weeks. Our study results showed that initiation of cART one week after TB therapy was not associated with improved overall survival at 48 weeks in patients with CD4 count < 200cells/ μL. In contrary to our study hypothesis the mortality rate was higher in week one, albeit statistically not significant. Consistent with other studies [
When our study started, there were ongoing studies to evaluate cART timing in TB and HIV coinfection at different CD4 count thresholds. Blanc et al.[
Hepatotoxicity is the most frequent treatment limiting, overlapping toxicity in TB/HIV co- treatment [
Similar to other studies, the most frequent clinical manifestation of TB-IRIS in our study was the emergence of new or worsening preexisting lymphadenopathy [
The incidence rates of new ADIs were not significantly different between the various study weeks. The common finding was that the cases aggregated at CD4 count < 50cells/μL in all the study groups and the events occurred at a median of four weeks after TB therapy. New ADIs were reported in 60% of the deaths in our study. The findings imply the association between new ADIs and mortality events.
There was no significant difference with respect to median CD4 count increase and proportion of participants with undetectable viral load at 48 weeks. However, at 24 weeks HIV-related outcome measures were most favorable in week one. Our study findings suggest different exposure period to TB treatment at cART initiation doesn’t affect virologic suppression rate and CD4 count gains at 48 weeks. It is consistent with results from other studies [
Our study has strengths and limitations worth mentioning. One of the strengths is initiation of cART as early as one week after TB therapy. To our knowledge, the ACTG A5221 was the only other study to evaluate cART as early as median of 10 days after TB therapy. Both the confirmed and probable TB cases in our study represent the cases seen in low-income and high TB/HIV prevalent settings. The study has adequately addressed an important treatment limiting toxicity in managing TB/HIV coinfection. To this effect, liver function test was routinely performed at baseline, just before cART, every two weeks the first eight weeks and then every four weeks until week 48. The study limitations are: its open label design which can result in biased responses in safety assessment and reporting, the proportion of confirmed TB cases either with smear microscopy or histopathology and choosing all-cause mortality rate as primary end point rather than cause specific mortality rate. The latter was more appropriate to evaluate cART outcome. In spite of the differences observed in the number of deaths among the study groups, we failed to show statistical significance. It is possible that our study was not adequately powered to detect these differences. Irrespective of this, cART one week after TB therapy did not prove beneficial.
In summary with respect to the question of cART timing within the first two months of TB therapy, our study from low-income and high TB/HIV prevalent setting showed cART one week after TB therapy does not improve all-cause mortality rate in coinfected patients with CD4 count < 200 cells/μL compared with four and eight weeks. Majority of the deaths in our study occurred during the first two months of TB therapy. Deaths the first two months of TB therapy in HIV coinfected patients are more likely TB-related than other HIV-associated morbidities [
Therefore, the patient’s overall condition including other comorbidities is probably more important than the time to initiate cART in predicting outcome. Different studies including our have clearly shown CD4 count less than 50 cells/μL independently predicts mortality in TB/HIV coinfection. In our study, the mortality trend increased in this subset of patients as cART deferred from week one to eight. On the other hand, this same group of patients had the highest incidence rate of grade 3 and 4 hepatotoxicity and subsequent interruption of first-line TB therapy in week one. The key question for CD4 subgroup < 50 cells/μL is striking the optimal balance between the potential survival benefit if cART is initiated one week after TB therapy as opposed to the increased morbidity and mortality due to hepatotoxicity and risk of TB treatment interruption. Based on our findings, we recommend cART later than the first but earlier than the fourth week of TB therapy in subset of patients with baseline CD4 count less than 50 cells/μL.
(PDF)
(XLSX)
(XLSX)
(DOC)
We are grateful to Addis Ababa University, School of medicine, Addis Ababa regional health bureau and the following health centers (Tekle-himanot, Arada, Kazanchis, Kirkos, Lideta, Meshulakekia and Beletscahew). Our special thank extends to Dr Aster Shoa-Amare from Zweditu memorial hospital for assisting us in study participants recruitment, Dr Hailyesus Getahun for continued support to the study, and WHO Stop TB Department for sponsoring the principal investigator present his work in CROI 2012.