Study Design and Population

This study was designed to evaluate the prevalence of MDR-TB in HIV-infected patients >18 years of age and to assess potential risk factors for MDR-TB in HIV-infected persons with TB. The study was carried out from February 1999 to January 2000 in Lima and Callao. Lima is the capital city of Peru, and Callao is the adjacent main port; together they have a population of 8,239,891, representing 32% of the national population. Lima and Callao accounted for 78% of all Peruvian AIDS cases reported from 1983 to September 2000 (13) and for 56% of all TB cases reported in 1999 (12). The research protocol for using human participants in this study has been reviewed and approved by the Human Subject Review Committee of the University of Washington and the Ethical Committee of the Scientific Research Office at the Cayetano Heredia University. Informed consent was obtained from the participants.

From February 1999 to January 2000, we introduced an active surveillance system to identify HIV-infected adults with a new event of TB, defined as a first episode of TB or a relapse, at the 10 public hospitals that provide care for most HIV-infected persons living in Lima and Callao. Each of these 10 hospitals has a unit of the National TB Control Program where every patient with a suspected or confirmed diagnosis of TB is referred for further work-up, treatment, follow-up, or referral. To identify HIV-infected patients with a new episode of TB, one trained interviewer periodically visited each of these units. In addition, we encouraged all clinicians to send clinical specimens for isolation of M. tuberculosis, and all laboratories to submit such isolates to TB control program reference laboratories for susceptibility testing.

MDR-TB was defined as resistance to both isoniazid and rifampin, with or without resistance to other drugs. Case-patients were adults with previously diagnosed HIV infection and with a new episode of TB, whose isolates of M. tuberculosis were MDR-TB. HIV-seropositive controls were adults with previously diagnosed HIV infection and a new event of TB, whose isolates of M. tuberculosis were susceptible to isoniazid or rifampin. HIV-positive status was defined by previous, repeatedly positive enzyme-linked immunosorbent assay (ELISA) confirmed by immunofluorescence or Western blot.

From February trough September 1999, a total of 972 adult patients with a new episode of smear-positive TB in Lima and Callao who received care at Ministry of Health facilities were included in the National TB Control Program’s surveillance of resistance to anti-TB drugs. Of these 972, a total of 116 had been previously tested for HIV: 7 (6%) were HIV seropositive. From the remaining 965 participants who were HIV seronegative or not tested for HIV infection, we randomly selected a second control group of 153 participants.

Enrollment, Interview, and Treatment

Both trained interviewers periodically visited each of the 10 hospitals to identify, enroll, and interview HIV-infected patients with a new episode of TB. Because classification as “case” or “control” for HIV-infected patients was determined by drug susceptibility-test results, which became available after 3 or 4 months, each HIV-infected adult with a new episode of TB was eligible to participate, and interviewers were blinded as to susceptibility test results. One member of the local TB team introduced the interviewer, who explained the study and invited the patient to participate. Interviewers contacted adults with TB who were HIV seronegative or of unknown HIV status belonging to the HIV–seronegative control group at their homes, where they were invited to participate, enrolled, and interviewed. Patients giving written informed consent underwent a standardized face-to-face interview concerning demographic characteristics, past and current medical history, and potential exposures to M. tuberculosis during the 12 months before onset of TB symptoms. Healthcare workers at each facility followed national guidelines of the TB and the AIDS control programs in providing further treatment and follow up of study participants. Data collected were handled exclusively by the study team, ensuring confidentiality.

Laboratory Methods

Primary isolations were attempted by using Lowenstein-Jensen or Ogawa medium at each hospital laboratory and at local reference laboratories serving the national surveillance study. Susceptibility testing was carried out at the Mycobacteria Laboratory of the Peruvian National Institute of Health, the national reference laboratory for susceptibility testing, or at one of three local mycobacterium reference laboratories. These laboratories used the proportion method to determine the sensitivity profile of each strain (19), using the following critical concentrations (μg/mL): isoniazid, 0.2; rifampin, 40; streptomycin, 4; ethambutol, 2; and pyrazinamide, 100. The National Reference Laboratory underwent external quality control by the Pan American Health Organization/WHO Instituto Panamericano de Protección de Alimentos y Zoonosis (INPPAZ) and performed quality control for the three local reference laboratories (20).

Data Analyses

SPSS 10.0 software (SPSS, Inc., Chicago, IL) was used for data entry and analyses. Percentages of MDR-TB were compared in HIV-positive and HIV-negative patients with TB, and risk factors for MDR-TB in HIV-infected patients were explored by calculation of odds ratios (OR) for dichotomous variables. Medians of continuous variables were compared by using Student t test for independent samples.

Characteristics of Patients with TB

Of 415 HIV-seropositive patients diagnosed and reported with a new episode of TB (on the basis of acid-fast smear or culture results or on clinical characteristics) at the 10 hospitals during the study period, 157 (38%) were not interviewed: 87 had already left the hospital and could not be located at referral health centers, 67 were in poor clinical condition and unable to answer the questionnaire, and 3 declined to participate. The remaining 258 seropositive participants interviewed averaged 32 years of age (range 18 to 62) and reported 9.8 years of education; 77% were men, 57% single, and 20% had had at least one previous episode of TB. Isolation of M. tuberculosis was attempted in 239 (93%); 135 were culture-positive, 61 were smear positive for acid-fast bacilli, and 62 were diagnosed on the basis of clinical criteria. Drug-susceptibility testing was requested for all positive cultures and completed on 81, who were the focus of subsequent analyses; the remaining strains were lost before or during transport to the National Reference Laboratory for susceptibility testing.

Of the 153 TB patients randomly selected as controls without known HIV infection, we located 110; 108 agreed to participate. They averaged 29.3 years of age (range 18 to 80) and reported 10.3 years of education; 57% were men, 52% single, and 9.3% had had at least one previous episode of TB.

Table 1 presents characteristics of participants with and without known HIV infection for whom drug-susceptibility testing was completed. Because HIV infection in Peru disproportionately affects men who have sex with men, HIV-seropositive persons were more often men and less often married, divorced, or widowed; family income and family ownership of the home were more common among those not known to have HIV infection. More importantly, previous histories of TB or TB prophylaxis were significantly more common in HIV-infected participants, as was history of contact with hospitals.

Drug-Susceptibility Test Results

The prevalence of MDR-TB was 43% in the 81 HIV-seropositive patients with available susceptibility test results and 3.9% in the 965 patients whose HIV status was negative or unknown (p < 0.001); only 1 (0.9%) of 108 patients whom we randomly selected and interviewed from this group of 965 TB patients had an MDR-TB isolate (Table 2). The prevalence of resistance to any drug was also higher in M. tuberculosis isolates from HIV-positive patients than in isolates from the 965 without known HIV infection.

Drug-susceptibility test results were available for 26 (16.6%) of the 157 HIV-seropositive patients we did not interview; 35% of these patients had MDR-TB, not significantly different from the 43% of those we did interview (p = 0.44).

HIV Patients with MDR-TB and with Non–MDR-TB

Of HIV-seropositive participants with TB (Table 3), MDR-TB was significantly associated with TB diagnosed at hospital A (OR 3.7, 95% confidence interval [CI] 1.3 to 10), with employment during the 12 months before the onset of symptoms, and with exposure to TB at work, but not with age, sex, marital status, education level, crowding in the home, or low income. MDR-TB was not significantly associated with previous episodes of TB or with TB prophylaxis; the proportion who had received TB prophylaxis or had a previous episode of treatment for TB was 17 (49%) of 35 with MDR-TB versus 20 (43%) of 46 without MDR-TB (OR 1.2, 95% CI 0.5 to 3.0).

Hospital Characteristics

Table 4 lists the 10 major public hospitals in Lima and Callao by number of persons with newly diagnosed AIDS reported during 1999, number with TB diagnosed during 1999, number of HIV-infected patients with a new event of TB during the study, and percentage of isolates of M. tuberculosis identified as MDR-TB. All but one of the MDR-TB cases were identified in the three hospitals reporting the largest number of HIV-seropositive patients with a new event of TB during 1999.

Assessment of Possible Bias

Interviewed patients with (N = 81) or without (N = 177) susceptibility results were similar with respect to age, years of education, family income, symptom duration, time since previous TB, TB prophylaxis in the past, crowding in the home, gender, marital status, home ownership, and employment; and during the past 12 months, exposure to TB at work, exposure at home, ambulatory care at health centers, ambulatory care at hospitals, inpatient hospital care, and participation in HIV or TB support groups. History of previous TB was somewhat more frequent in those tested (27%) than in those not tested (17%, p = 0.09), and history of Pneumocystis carinii pneumonia (PCP) prophylaxis was more frequent among those tested (50%) than in those not tested (32%, p < 0.01).