Mortality in HIV-infected patients with tuberculosis treated with streptomycin and a two-week intensified regimen: data from an HIV cohort study using inverse probability of treatment weighting

Introduction

Tuberculosis (TB) accounts for one third of deaths among HIV-infected people, and HIV infection is present in 25% of TB deaths (World Health Organization, 2015). Globally 1.2 million people living with HIV had TB in 2014, and 390,000 people died (World Health Organization, 2015). Despite this tremendous burden of preventable deaths, there has not been any major breakthrough in the chemotherapy of HIV-related TB in the last decades.

Recent data from a clinical trial and an observational cohort study suggest that increasing the doses of common anti-TB drugs during the initial phase of anti-TB therapy (ATT) could have survival benefits in patients with TB meningitis (Alvarez-Uria et al., 2013a; Ruslami et al., 2013). In a phase two randomized trial investigating the effect of an intensified ATT (iATT) during the first two weeks of treatment, higher exposure to rifampicin was associated with improved survival (Ruslami et al., 2013; Te Brake et al., 2015). In previous studies from our HIV cohort, an iATT during the first days of TB meningitis treatment also achieved improved survival, but only if streptomycin (STM) was added to the regimen (Alvarez-Uria et al., 2015). This finding was intriguing because STM has poor penetration into cerebrospinal fluid (Donald, 2010), suggesting that the beneficial result of the regimen was not related to the local effect in CSF.

Data of intensified regimens in HIV-related TB other than TB meningitis are lacking. In the present study, we hypothesized that an iATT could also reduce mortality in other forms of TB. We aimed to compare the survival of HIV-infected patients with TB treated with a standard ATT (sATT) and an iATT (with and without STM) during the first two weeks of chemotherapy.

Methods

Results

During the study period, 1,615 patients started ATT; 847 were included in the sATT group, 322 in the iATT group, and 446 in the iATT + STM group. Baseline characteristics of patients by treatment group are presented in Table 1. The proportion of patients previously treated of tuberculosis was higher in the iATT + STM group (43%) than the one in the sATT group (19.2%) and in the iATT group (none). This is because the Indian Guidelines for tuberculosis recommend adding STM during the first two months of ATT in patients who had received ATT for at least one month in the past (Ministry of Health and Family Welfare India, 2005).

Patients in the sATT group had higher concentrations of serum albumin, higher EGFRs and lower blood urea nitrogen (BUN)/creatinine ratios. Other differences were not statistically significant. Weight was measured in 780 patients in the sATT group, 289 patients in the iATT group and 401 patients in the iATT + STM group, and the median weight was 47 kg (interquartile range (IQR), 40–55), 47 kg (IQR, 40–55), and 45 kg (IQR, 40–52), respectively. Thus, the median dose of rifampicin was 9.6 mg/kg (IQR, 8.2–11.3) in the sATT group, 19.1 mg/kg (IQR, 16.4–22.5) in the iATT group and 20 mg/kg (IQR, 17.3–22.5) in the iATT + STM group.

Univariate and multivariate analysis of factors associated with mortality are presented in Table 2. After adjusting for other covariates, there were no mortality differences between the sATT and the iATT groups, but the use of iATT + STM versus sATT was associated with a 28% mortality risk reduction (adjusted hazard ratio [aHR] 0.72, 95% CI [0.57–0.91]) compared with sATT. Other factors associated with mortality were low albumin concentrations (aHR 0.47 per increase of 1 g/dl, 95% CI [0.41–0.54]), low CD4 + lymphocyte concentrations (aHR 0.76 per increase of 1 log-CD4 cells, 95% CI [0.69–0.83]), high urea concentrations (aHR 1.008 per increase of 1 mg/dl, 95% CI 1[.004–1.013]), high BUN/creatinine ratios (aHR 1.02, 95% CI [1.01–1.031]) and lower EGFR (aHR 0.994 per increase of 1 ml/min/1.73 m², 95% CI [0.99–0.999]). Having mild extrapulmonary TB was associated with reduced risk of death (aHR 0.58, 95% CI [0.43–0.78]).

In a sensitivity analysis, we compared only the sATT group and iATT + STM group using IPTW. Variables statistically associated with mortality in the multivariate analysis were balanced using propensity score methods. At baseline, the sATT group had higher mean of serum albumin concentration, lower mean of EGFR, and lower mean of BUN/Creatinine ratio than the iATT + STM group (Table 3). These differences were considerably reduced after IPTW. The median of propensity scores was 0.3 (IQR 0.25–0.37) in the sATT group and 0.39 (IQR, 0.32–0.51) in the iATT + STM group. The mean standardized difference and the mean Kolmogorov–Smirnov statistic was 0.111 and 0.064, respectively before IPTW, and 0.040 and 0.029, respectively after IPTW. The maximum standardized difference and the maximum Kolmogorov–Smirnov statistic was 0.349 and 0.192, respectively before IPTW, and 0.0858 and 0.058, respectively after IPTW.

Stabilized sampling weights were used to estimate Kaplan–Meier survival curves and Cox proportional hazards. In Fig. 1, we present the Kaplan–Meier survival estimates. The use of iATT + STM was associated with a significant reduction in mortality (HR 0.766, 95% CI [0.67–0.96]). While analysing interactions between treatment and other predictors of mortality, we found that baseline serum albumin concentrations modified significantly the survival effects seen with the iATT + STM regimen (P-value for interactions = 0.0003). Compared with sATT, the iATT + STM regimen was more beneficial for patients with higher serum albumin concentrations than for those with more severe hypoalbuminemia (Fig. 2).

Discussion

HIV-associated TB is difficult to treat. According to the latest Global TB Report (World Health Organization, 2015), nearly one third of HIV-infected patients with TB die with the current standard of care.

In a previous study from our cohort, we showed that adding STM to an intensified ATT could reduce mortality in HIV infected patients with TB meningitis (Alvarez-Uria et al., 2015). Given the poor penetration in cerebrospinal fluid of STM (Kaojarern et al., 1991; Ellard, Humphries & Allen, 1993), these findings were intriguing and suggested that the synergetic effect of STM and the iATT could be also beneficial for other forms to TB. The results of the present study confirmed this hypothesis. The use of the iATT regimen was not able to improve survival compared with the standard of care. However, when STM was added to the iATT, we observed a statistically and clinically significant mortality reduction.

In fact, the synergetic effect of STM and other anti-tuberculosis drugs is not new. In patients with pulmonary tuberculosis, the combination of rifampicin and STM has a strong bactericidal activity during the first six days of ATT (Jindani, Doré & Mitchison, 2003). In the same study, ethambutol, which was not used in our iATT regimen, had an antagonistic effect when combined with other drugs (Jindani, Doré & Mitchison, 2003). Moreover, in two phase II clinical trials, substitution of ethambutol with fluoroquinolones resulted in a more rapid decline in bacterial load and higher proportion of patients achieved TB culture negativity at eight weeks (Rustomjee et al., 2008; Conde et al., 2009). However, in our study, substitution of ethambutol with levofloxacin did not improve survival in the iATT group. New studies are needed to clarify if the synergetic effect on survival of STM and the iATT could be explained by the higher doses of rifampicin, the absence of ethambutol, the addition of levofloxacin or combinations of these factors.

On the other hand, a more rapid bactericidal activity during the first days of treatment might not be beneficial to all patients. In our study, patients with severe hypoalbuminemia did not benefit from the use of the iATT + STM regimen. These patients were likely to have more advanced forms of TB and a rapid killing of mycobacteria might have led to a higher inflammatory response (Ordonez et al., 2014). Then again, hypoalbuminemia has been associated with lower concentrations of antimycobacterial drugs in patients with pulmonary TB (Tappero et al., 2005). New studies investigating new treatment approaches in this particularly vulnerable group are urgently needed.

The study has some limitations. Observational studies can be biased due to unknown confounders not included in the multivariate analysis. Unlike in clinical trials, treatment groups were not uniformly balanced. Patients in the iATT + STM group were more likely to be previously treated of tuberculosis, which is a known factor related to poorer prognosis (Santha et al., 2002; Jones-López et al., 2011; Alvarez-Uria et al., 2012a). However, in the sensitivity analysis using IPTW, being previously treated of tuberculosis was not included in the multivariate analysis and the use of iATT + STM remained statistically associated with improved survival. One of the most common side effect of STM is nephrotoxicity. It could be possible that the use of STM could have resulted in higher probability of acute kidney injury, but we did not collect information of renal function after ATT initiation, so we cannot compare nephrotoxicity among treatment groups. Nevertheless, we performed extensive analysis of baseline renal function with different parameters. Kidney dysfunction was associated with higher risk of death, but there was no interaction between baseline kidney dysfunction and STM use in terms of increased mortality. The study was done in a resource-limited setting, so we did not confirm TB diagnosis by mycobacterial culture. In addition, we did not have information about the drug resistance of mycobacteria among treatment groups. However, patients in the iATT + STM group were more likely to have been treated of TB in the past, which is a well-known factor for resistant TB. Then again, severely ill patients were not excluded, so the study reflects the “real-life” of HIV-associated TB in a resource-poor setting, so our results could be generalized to similar settings in developing countries with high prevalence of TB and HIV.

Conclusions

The results of this study are consistent with previous studies from our cohort showing improved survival when STM is combined with an iATT regimen with higher doses of rifampicin and substitution of ethambutol with levofloxacin. However, the survival benefits of the new regimen was not evenly distributed among all patients. The new regimen did not reduce mortality in those patients with severe hypoalbuminemia. The study is observational in nature, so we should be cautious about our conclusions. However, given the high mortality of HIV-related TB, the results of this study deserve further research, ideally through a randomized clinical trial.