Impact of HIV on mortality among patients treated for tuberculosis in Lima, Peru: a prospective cohort study

Between 2005 and 2010, the estimated incidence of TB in Peru was 140 and 106 cases per 100,000 population, respectively [11]. The HIV prevalence among Peruvian adults dropped during this period from 0.5 to 0.4 %, and patients co-infected with HIV comprised 1.9 to 2.6 % of notified TB cases [10, 11]. The enrollment methods for this cohort have been described previously [12, 14, 15]. In brief, we enrolled all patients with confirmed TB or presumptive TB with respiratory symptoms for at least 2 weeks who were living in 2 adjacent health regions in Lima (Lima Ciudad and Lima Este), and who met Peruvian National Tuberculosis Control Program (NTP) criteria for drug-susceptibility testing (DST) referral (Table 1) [13]. The criteria for DST referral detailed in Table 1 reflected a targeted testing strategy for patients at risk for multidrug-resistant TB (MDR-TB) [14]. Healthcare workers at local health establishments identified patients and sent their sputum samples to the reference laboratory for DST. Because all sputum samples for DST were sent to the district laboratories, we identified subjects eligible for enrollment by this referral. We enrolled patients at the time of sputum sample collection for culture and DST, from January 5, 2005 through March 5, 2008 in Lima Ciudad and from April 20, 2005 through May 27, 2008 in Lima Este. Study personnel visited each district laboratory on a regular basis to review sample referrals and confirm that all eligible subjects had been identified. There were no exclusion criteria for enrollment into the study.

A trained team collected data prospectively from HIV and TB charts, laboratory registries, and databases using standardized forms. Baseline sociodemographic and clinical data, including baseline HIV status, were collected at enrollment. For HIV-positive patients, study personnel recorded the most recent CD4 cell count result and date, as well as receipt of ART, upon enrollment. HIV viral loads were not consistently available for this analysis. TB physicians reviewed all available chest x-rays (CXR) performed less than 1 year before and up to 1 month after enrollment, using standardized criteria for coding radiographic abnormalities. For each lung field, TB physicians recorded the presence of cavities, fibrosis, alveolar infiltrates, pneumothorax, pleural hemorrhage, nodules, disseminated/miliary disease, bullae, thoracic lymphadenopathy, and post-surgical changes. Data were entered into an Epi Info version 3.4.3 database (Centers for Disease Control and Prevention, Atlanta, GA, USA) and exported into an Access 2000 database (Microsoft Corporation, Redmond, WA, USA).

The DST methods for this cohort have been described previously [12, 14, 16, 17]. Under program conditions, the Peruvian National Reference Laboratory (NRL) performed mycobacterial culture and DST using the BACTEC™ 460 system (Becton, Dickinson and Company, Sparks, MD, USA) on paucibacillary and smear-negative samples produced by high-risk patients (HIV-positive, children, healthcare workers). Other smear-negative samples underwent mycobacterial culture on modified Ogawa medium followed by conventional DST for first-line drugs using the proportions method on Löwenstein-Jensen (LJ) medium at the district laboratory. Smear-positive samples were cultured on LJ medium and underwent DST using the indirect agar plate proportions method. Beginning in December 2005 in Lima Ciudad and March 2007 in Lima Este, these samples were cultured on modified LJ medium and tested for isoniazid and rifampicin resistance using the direct nitrate reductase assay (NRA) method [18]. All isolates found to be drug-resistant at the district laboratories were sent to the NRL for complete DST to first- and second-line drugs.

In 2004, the Peruvian Ministry of Health began offering free ART through the Peruvian National HIV Program to patients with either 1) Stage III or Stage IV acquired immunodeficiency syndrome (AIDS) defining illnesses other than TB with CD4 count <350 cells/μL, or 2) Stage I or Stage II HIV disease with CD4 count <200 cells/μL according to World Health Organization criteria [19]. Peruvian national TB treatment guidelines recommended HIV screening for all TB patients [13]. HIV physicians evaluated HIV-positive patients monthly, with CD4 and viral load monitoring every 6 months. Peruvian national TB treatment guidelines made the following recommendations regarding timing of ART initiation for TB/HIV co-infected patients: 1) ART deferral until after completion of TB treatment for those with CD4 count >200 cells/μL; 2) ART initiation after the intensive phase of TB treatment for those with CD4 count 100–200 cells/μL; and 3) ART initiation as soon as feasible after TB treatment initiation for those with CD4 count <100 cells/μL (with no specific recommendation for early versus deferred ART) [13].

Details of the TB management have been published elsewhere [13, 20, 21]. New drug-susceptible TB (DS-TB) cases, regardless of HIV status, received a biphasic treatment regimen with first-line drugs isoniazid (H), rifampicin (R), ethambutol (E), and pyrazinamide (Z): 2RHZE/4(RH)₂ [13]. Retreatment DS-TB cases received a regimen including streptomycin (S) in the intensive phase: 2RHZES/1RHZE/5(RHE)₂ [13]. Patients with presumptive MDR-TB received empiric treatment with second-line therapy pending DST results: ethambutol, pyrazinamide, kanamycin, ciprofloxacin, ethionamide, cycloserine, and para-aminosalicylic acid [13]. The guidelines used to design individualized MDR-TB treatment regimens based on DST results have been described elsewhere [22]. All TB patients received directly observed treatment free of charge through the Peruvian NTP. We followed patients until a TB treatment outcome was recorded or until the administrative end of the study on June 26, 2010.

We followed patients from the time of enrollment to the date of TB treatment outcome, including death. The patients with no treatment outcome during the observation period were censored on the last day of follow-up. We excluded patients from the analysis if they were not treated for TB during the period of observation. TB treatment outcomes were defined according to previously published guidelines [23, 24].

We selected potential predictors of mortality from risk factors identified in the literature and clinical experience [9, 25‐31], and evaluated them using univariate Cox proportional hazards models. We considered predictors associated with death on univariate analysis (P < 0.05) for inclusion in the multivariate Cox proportional hazards model evaluating the association between HIV and mortality. We built the multivariate model using a backward selection method, retaining age, gender, and predictors associated with death with a P < 0.20. Since HIV-positive patients tended to present with more severe clinical findings, we considered the possibility that the effect of HIV status on mortality during TB treatment was mediated by predictors significantly associated with death on univariate analysis (P < 0.05) that also corresponded to disease severity. We considered the following variables as potential mediators: ability to perform activities of daily living (ADLs), low body mass index (BMI), and dyspnea. We evaluated the direct effect of HIV status on time to death by comparing multivariate models with and without these variables. We tested the proportional hazards assumption in the multivariate models by using Schoenfeld residuals fitted to rank of analysis time, and we constructed a Kaplan-Meier survival curve stratified by HIV group. We also performed a secondary analysis in which we repeated the multivariate analyses while adjusting for baseline MDR-TB.

All analyses were performed using Stata/SE version 14.1 (StataCorp LP, College Station, TX, USA). The χ2 test or Fisher’s exact test was used to calculate P values, when appropriate. The Student’s t test was used for the two-sample mean-comparison test, when appropriate. All statistical tests were two-sided, and significance was set at α = 0.05.

Institutional review boards at the Partners Human Research Committee and the Peruvian National Institute of Health Committee of Research Ethics approved the original study protocol and the protocol amendment for this analysis. The original study was a programmatic effort under the leadership of the Peruvian National TB Control Program and the Peruvian National Institute of Health. An informed consent waiver was approved from both institutional review boards, because the data collection process required no contact with the patients; only included information that was routinely collected in the course of providing clinical care to patients; and the treatment regimens administered to patients were determined by the standards of care in Peru at the time, such that participation in the study did not adversely affect the rights and welfare of the subjects and patient confidentiality was maintained. The U.S. Centers for Disease Control and Prevention approved this activity as program evaluation and not human subjects research.