Design and participants
This was a case-control study conducted from July through September 2016. The data were collected from nine healthcare facilities, including all the eight facilities that provide DR-TB care plus one TB/HIV facility that had the highest number of DS-TB/HIV positive patients. When a patient presents to the facility for any reason, he or she is screened for TB using the signs and symptoms screening tool. If TB presumptive positive by the screening tool, a sample for gene Xpert is taken. The gene Xpert has capacity to detect Mycobacterium tuberculosis (MTB) and rifampicin resistance. If MTB positive and rifampicin resistance not detected, the patient is assumed to be having drug susceptible TB (DS-TB), then the patient is initiated on DS-TB treatment. If MTB positive and rifampicin resistance detected, the patient is initiated on MDR-TB standardized regimen. If MTB not detected, the patient was considered TB negative.
In the study, cases included those with HR-TB, PDR-TB, and MDR-TB diagnosed by culture and drug susceptibility testing (C&DST) and line probe assay (LPA) plus co-infected with HIV. HR-TB patients were those with TB resistant to isoniazid only. PDR-TB was defined as TB resistant to isoniazid and one of the following first-line anti-TB drugs: ethambutol and streptomycin, but not resistant to rifampicin. MDR-TB was defined as TB resistant to rifampicin and isoniazid with or without resistance to the other first-line drugs.
Controls were patients diagnosed by Xpert MTB/RIF to be MTB positive and rifampin resistance not detected. The DS-TB patients were on treatment for at least 2 months, and had converted to smear-negative status. Smear conversion at 2 or 3 months was used as a proxy for drug susceptibility. C&DST is not yet universal in Swaziland. According to the Eswatini DR-TB guidelines, 2012, C&DST was supposed to be done to all patients who had history of previous TB treatment, failed to convert at 2–3 months of first-line TB treatment, failed the first-line TB treatment, are contacts of DR-TB patients, and health care workers (nurses, doctors, laboratory staff). But there are facilities that are supported by partners like medicins san frontiers (MSF), where C&DST is universal for all TB presumptive cases and 65% of our controls were from those facilities.
According to the TB manual, 2012 and the DR-TB guideline, 2012 (Additional files
1 and
2), all persons diagnosed with TB should be tested for HIV and HIV positive should be screened for TB. To diagnose HIV, parallel testing was implemented,
determine test was the first test that was done, if positive,
unigold test was then done
, if positive, the patient was diagnosed as HIV positive. Based on the guidelines, all TB patients who are HIV positive should be initiated on antiretroviral treatment regardless of CD4 count.
HIV infected patients were recruited if they had a documented HIV positive status on the TB register. The cases were selected from the DR-TB register and the controls from the DS-TB register. The participants were enrolled in the consecutive order as they appeared on the register until the desired sample size was reached. We estimated the desired sample size of 197 for DR-TB cases and DS-TB group each, based on an estimation of OR = 1.8, probability in the control group = 0.3, two-sided alpha = 0.05, and power = 0.8.
In this study, patients younger than 18 years were excluded. For the case group, patients initiated on second-line TB treatment without C&DST and extensively DR-TB cases (resistance to any fluoroquinolone and at least of one of the 3 second-line injectables (kanamycin, amikacin, or capreomycin) in addition to MDR-TB) were excluded. For the control group, we excluded those who were TB negative at the beginning of TB treatment. A total of 400 TB/HIV patients participated in the study: 197 with DR-TB (78 with HR-TB, 42 PDR-TB, and 77 MDR-TB) and 203 with DS-TB. Of the 203 controls, 35% (n = 71) were TB patients who were diagnosed by Xpert MTB/RIF to be MTB positive and rifampin resistance not detected and were on treatment for at least 2 months, and had converted to smear-negative status and 65% (n = 132) were confirmed DS-TB by C&DST in addition to the above-mentioned criteria. Of the 400 participants, 212 had HIV diagnosed before TB (outside the TB department) and 188 were diagnosed with HIV within the TB department post TB diagnosis; 9 had not been initiated on antiretroviral treatment at 2 months after TB treatment initiation due to their refusal. The study protocol has been approved by the Swaziland Scientific and Ethics Committee. The study participants signed an informed consent form.
Measurements
Data were collected through face-to-face interviews and review of medical records. The interviews were conducted using a structured questionnaire, and the medical record reviews were performed using a case form. The questionnaire was pre-tested on 10 patients at a TB center. Necessary amendments were made to improve the clarity of the questionnaire. The data collectors, nurses or TB data clerks, were trained on data collection tools onsite.
The dependent variable, drug-resistant status, was divided into 3 groups, HR-TB or PDR-TB, MDR-TB, and DS-TB. We combined HR-TB and PDR-TB as one group since isoniazid and rifampicin were the most powerful first-line TB drugs and HR-TB and PDR-TB in the study can be treated with rifampicin. In addition, the sample size for PDR-TB was not large enough to compose an individual group in the analysis. The independent variables included the socio-demographic characteristics (age, sex, marital status, level of education, occupation, residence, and region), clinical characteristics (history of previous TB treatment, length of time from diagnosis to treatment initiation (date from results authorization at laboratory level to treatment initiation), history of TB contact, history of IPT, whether patient started ART before current episode of TB, CD4 count at TB diagnosis, diabetes, TB treatment non-adherence (self-rated adherence to the current TB treatment being less than 100%) and HIV treatment non-adherence (self-rated adherence to the ART treatment being less than 100%), and behavioral characteristics (smoking status, alcohol drinking status, and history of imprisonment). For patients with previous TB treatment, we classified them as “recurrence,” “retreatment after failure,” and “retreatment after lost to follow-up.” Recurrence was defined when a patient previously treated for TB with the most recent treatment curing TB or the treatment being completed, and who is subsequently diagnosed with a recurrent episode of TB. Heavy alcohol drinking was determined based on whether the patient had ever drunk > 150 glasses of alcohol beverage per month during the past 12 months (current drinker), or during the lifetime but not in the past 12 months (previous drinker), or had never drunk > 150 glasses of alcohol beverage per month during his lifetime (never).
Data on socio-demographics, behavioral characteristics, IPT exposure, TB and HIV treatment adherence, and comorbid conditions (diabetes) of patients were collected through face-to-face interviews. Other clinical characteristics were collected through the review of medical records.
Data analysis
The data analyses were carried out using IBM SPSS statistics 22. The characteristics of the three groups (HR-TB or PDR-TB, MDR-TB, and DS-TB) were compared using the chi-squared test, ANOVA, and Kruskal-Wallis test. Multinomial logistic regression was used to determine factors associated with isoniazid resistant tuberculosis. Isoniazid resistant tuberculosis was divided into HR-TB or PDR-TB, and MDR-TB, with DS-TB being the reference group. In the multinomial logistic regression, exponential ßs were the odds ratios (OR) for the predictors. The odds ratio of a coefficient indicates how the risk of the outcome falling in the comparison group compared to the risk of the outcome falling in the reference group changes with the variable in question. Backward selection with one least significant variable being dropped from the full model each time until all variables in the model having a 2-sided alpha of < 0.05 was used to select the final parsimonious model.