During the 3-month assessment period (figure ), 305 adults (≥ 18 years age) initiated antituberculosis treatment, 7 of whom had untraceable clinical records and 6 of whom declined HIV-1 testing. We restricted our data analysis to the 292 (96%) patients whose HIV-1 status and clinic records were available. In 209 HIV-1 infected and 83 HIV-1 uninfected patients, loss to follow-up (46 [22%] of 209 vs. 29 [35%] of 83, p-value = 0.026) and mortality (16 [8%] of 209 vs. 1 [1%] of 83, p-value = 0.048) differed significantly at 24 weeks.
Flow-diagram of 305 adult patients who started antituberculosis treatment from 1 June - 31 August 2008.
At initial tuberculosis diagnosis (table ), HIV-1 infected patients were more likely than HIV-1 uninfected patients to be female, be of younger age, have extra-pulmonary tuberculosis and be diagnosed with tuberculosis at the referral hospital. HIV-1 uninfected patients were more likely than HIV-1 infected patients to have microbiologic confirmation of tuberculosis at initial tuberculosis diagnosis and during the 24 weeks of follow-up.
Baseline characteristics and microbiologic confirmation of tuberculosis in 209 HIV-1 infected and 83 HIV-1 uninfected patients receiving antituberculosis treatment
Prior to tuberculosis diagnosis, 34 (23%) of 145 HIV-1 infected patients who qualified for ART under national guidelines were receiving ART. Six months later, 109 (75%) of 145 patients had received ART.
Causes of Clinical Deterioration and Hospital Admission
During the 24 weeks of follow-up, 117 (40%, 95% CI: 35-46%) of 292 tuberculosis patients experienced clinical deterioration, of whom 101 were HIV-1 infected and 16 were HIV-1 uninfected. Causes of clinical deterioration (table ) included: co-morbid illnesses (70 patients), tuberculosis-related illnesses (47 patients), non AIDS-defining HIV-1 related infections (25 patients) and AIDS-defining illnesses (21 patients). Peripheral neuropathy, enteric illness and deep venous thrombosis were frequent co-morbid illnesses. TB-IRIS and paradoxical reactions were frequent tuberculosis-related illnesses. Oesophageal candida, Pneumocystis jirovecii pneumonia and cryptococcal meningitis were frequent AIDS-defining illnesses. Of 117 patients who experienced deterioration, 30 (26%) required hospital admission [27 (27%) of 101 HIV infected and 3 (19%) of 16 HIV-1 uninfected patients (p-value = 0.756)]. Causes of inpatient hospital admission were paradoxical reaction or TB-IRIS (9 patients), new AIDS-defining illness (8 patients), deep venous thrombosis (6 patients), MDR-TB (2 patients), cardiomyopathy (1 patient), pneumothorax (1 patient), symptomatic deterioration due to poor adherence with antituberculosis treatment (1 patient), hyperglycaemic emergency (1 patient) and seizure disorder (1 patient).
Illnesses (n = 199) in 101 HIV-1 infected and 16 HIV-1 uninfected patients who experienced clinical deterioration during 24 weeks of antituberculosis treatment
Risk Factors for Clinical Deterioration
In the 292 tuberculosis patients, 4 factors were significantly associated with clinical deterioration in univariate analysis: HIV-1 infection, diagnosis of tuberculosis at the referral hospital, evidence of extra-pulmonary tuberculosis, and absence of a DST result at tuberculosis diagnosis. Only HIV-1 infection (figure ) remained significant in multivariate analysis (adjusted hazard ratio [aHR]= 2.0, 95% CI = 1.1-3.6).
Figure 2 a: Hazard ratios (95% CI) of risk factors for clinical deterioration during 24 weeks of antituberculosis (TB) treatment in 292 patients (Cox proportional hazards model) and Figure 2b>: Hazard ratios (95% CI) of risk factors for clinical deterioration (more ...)
In subsequent analysis (figures not shown), we assessed whether the probability of clinical deterioration from non-HIV-1 related causes was associated with HIV-1 infection. In univariate analysis, we found a significant association between HIV-1 infection and non-HIV-1 related causes for deterioration (RR = 1.3, 95% CI: 1.07 - 1.50). However, in the Cox proportional hazards model, using the same variables as figure , this significant association was not confirmed (HR = 1.5, 95% CI: 0.81-2.64). This analysis suggests that HIV-1 infection is a significant variable for clinical deterioration because of HIV-1 related illnesses (either AIDS- or non AIDS-defining illnesses).
In the 209 HIV-1 infected tuberculosis patients, 3 factors were significantly associated with clinical deterioration in univariate analysis: a lower CD4+ count, diagnosis of tuberculosis at the referral hospital, and antiretroviral treatment received during antituberculosis treatment. Only a lower CD4+ stratum at tuberculosis diagnosis (figure ) remained significant in multivariate analysis (aHR = 1.5, 95% CI = 1.1-2.2).
In subsequent analysis (figures not shown), we assessed whether the probability of clinical deterioration from non-HIV-1 related causes was associated with decreasing CD4+ counts. We did not find a significant association between decreasing CD4+ count strata and non-HIV-1 related causes in both the univariate analysis (P = 0.189) and the Cox proportional hazards model (HR = 1.3, 95%CI: 0.88 - 1.97). This analysis suggests that decreasing CD4+ counts are a significant risk factor for clinical deterioration because of their association with HIV-1 related illnesses (either AIDS- or non AIDS-defining illnesses).
Relative Risk and Incidence Rate of Clinical Deterioration
HIV-1 infection and a low CD4+ count were the only significant risk factors for clinical deterioration in multivariate analysis. We therefore determined the relative risk and incidence rate of clinical deterioration according to HIV-1 status and CD4+ stratum. HIV-1 infected patients were more likely than HIV-1 uninfected patients to experience clinical deterioration (RR = 2.6, 95%CI: 1.6-4.0). Using HIV-1 uninfected patients as the referent group, the relative risk (RR) of clinical deterioration increased as the CD4+ counts in HIV-1 infected patients decreased (CD4+ >350 cells/μL: RR = 1.4, 95% CI = 0.7-2.9; CD4+200-350 cells/μL: RR = 2.0, 95% CI = 1.1-3.6; CD4+ <200 cells/μL: RR = 3.0, 95% CI = 1.9-4.7). The incidence rate (IR) of clinical deterioration (illnesses diagnosed per 100 months of follow-up) also increased as the CD4+ counts decreased. Incidence rates differed significantly between HIV-1 uninfected patients (IR = 5.1, 95% CI = 3.1-7.5) and HIV-1 infected patients with a CD4+ count of 200-350 cells/μL (IR = 12.3, 95% CI = 8.2-17.0). Similarly, incidence rates differed significantly between HIV-1 infected patients with a CD4+ count of 200-350 cells/μL and HIV-1 infected patients with a CD4+ count < 200 cells/μL (IR = 20.7, 95% CI 17.8-23.9).
Figure is a Lowess plot showing the proportion of patients who experienced clinical deterioration during the 24 weeks of antituberculosis treatment. The initial peak at 6 weeks in HIV-1 uninfected patients corresponds with tuberculosis-related illnesses (mostly paradoxical reactions) and the baseline fluctuations represent co-morbid illnesses. The curve for HIV-1 infected patients with a CD4+ count > 350 cells/μL is similar to that of HIV-1 uninfected patients, despite an earlier peak for paradoxical reactions. The proportion of HIV-1 infected patients with a CD4+ count of 200-350 cells/μL who experienced clinical deterioration substantially increased after 10 weeks of follow-up. Furthermore, a substantially higher proportion of HIV-1 infected patients with a CD4+ count <200 cells/μL experienced clinical deterioration compared to other CD4+ strata.
Lowess plot showing the proportion of patients who experienced clinical deterioration during 24 weeks of antituberculosis treatment.
Fifteen of 17 deaths occurred in HIV-1 infected patients with a CD4+ count < 200 cells/μL. The median interval from antituberculosis treatment to death was 98 days (IQR = 59 -128). Eight of 17 deaths occurred between 10 and 20 weeks of follow-up. Diagnoses at time of death included: AIDS-defining illnesses (5), poor adherence with antituberculosis treatment (4), paradoxical neurologic TB-IRIS (3), enteric illness (2), MDR-TB (1), pulmonary embolus (1) and tension pneumothorax (1).
Loss to follow-up
Within 24 weeks of commencing antituberculosis treatment, 75 patients were lost to follow-up. The median interval from antituberculosis treatment to loss to follow-up was 82 days (IQR = 48 -125). The CD4+ counts of 46 HIV-1 infected patients who were lost to follow-up (median 150 cells/μL, IQR = 67-347) did not differ significantly (P = 0.362) from the CD4+ counts of 147 HIV-1 infected patients who were not lost to follow-up was (median 144 cells/μL, IQR = 66-272).