Design and study population
We randomly sampled patients booked for endoscopy at the University Teaching Hospital in Lusaka between November 2012 and October 2013 in this cross-sectional survey. All patients who were 18 years and older who were booked for either colonoscopy or flexible sigmoidoscopy were eligible to take part in the study. Patients who were on treatment for TB, those who were found to have cancer lesions on endoscopy or those with debilitating conditions were not considered for recruitment.
Sample collection and processing
Consenting patients were asked to answer questions on their demographic characteristics, symptoms and drug history using a structured questionnaire, which took up to 10 min. Blood was collected for HIV and haemoglobin tests. HIV testing was done using the recommended national testing algorithm in Zambia, which follows the standard two test algorithm that starts with the Allere Determine® HIV1/2 test kit (Abbott, Japan) followed by the Uni-Gold™ Recombigen HIV-1/2 (Trinity Biotech) if positive [14]. The two tests showed an agreement of 100%. Furthermore, patients who wanted to know their results were offered counselling services before the results were shared with them.
Samples of stool and intestinal lavage were collected from the patients before they underwent their scheduled procedure. After specimen collection, the patients underwent either a flexible sigmoidoscopy or a colonoscopy, during which descending colon biopsy and/or caecal biopsy samples were obtained from eligible patients. The biopsy samples were collected in 2 ml cryovials filled with saline. These samples were then used to test for the presence of mycobacteria as described below.
The BACTEC mycobacteria growth indicator tubes (MGIT) liquid culture method was used to establish the presence of nontuberculous mycobacteria from the biopsies and lavage fluids after decontamination with 6% NaOH. The biopsy samples were first homogenised before being transferred to 50 ml Falcon centrifuge tubes for processing, in much the same way as sputum samples are processed. For each batch of samples processed, care was taken to ensure that a positive (H37Rv laboratory stock) and a negative control (phosphate buffer, pH 6.8) were included. The samples were incubated in the MGIT 960 machine at 37 °C for up to 42 days. No additional supplements for isolation of fastidious mycobacteria were used. Samples from positive MGIT tubes were inoculated onto blood agar plates to check for contamination, followed by Ziehl Neelsen (ZN) staining and the Capilia assay (Capilia® TB, TAUNS Laboratories Inc, Shizuoka, Japan) to confirm M. tuberculosis [15, 16]. For identification, we used the GenoType mycobacterium CM/AS (Hain Lifescience, GmbH, Germany) assay according to manufacturer’s specifications.
DNA extraction
To extract DNA, 1 ml was drawn from a positive MGIT tube into a container and centrifuged at 10,000g for 15 min in a class II safety cabinet. The supernatant was discarded and the pellet was re-suspended in 300 µl of molecular grade water. Following that, the suspension was sonicated for 15 min, followed by incubation in a water bath for 20 min at 95 °C as per manufacturer instructions [17].
In order to test for contamination of water sources with NTM, we collected water samples in 50 ml Falcon tubes at different time-points in the endoscopy unit (where the sigmoidoscopy and colonoscopies took place) as well as in the TB laboratory. These samples were centrifuged at 3000 rpm for 15 min, supernatant discarded and 5 ml of the remaining fluid was processed for mycobacteria using the MGIT method as described above.
Statistical analysis
A structured questionnaire was designed to collect demographic characteristics, symptoms and drug history. The data was entered into a questionnaire created in Epidata statistical software on a panel that matched the unique identifier for the individual’s demographic data, then analysed using STATA (Version 11.2, StataCorp, College Station, Texas). Sample descriptions were expressed as means with their respective standard deviations; whereas proportions (e.g. prevalence) were expressed as percentages with corresponding 95% confidence intervals. We used Fisher’s exact Chi square test for cross-tabulations where appropriate. Tests for normality of age and other continuous variables were done using the Shapiro–Wilks test. Mantel–Haenszel tests were used to test for interaction and confounding. Multivariable logistic regression was used to look for predictors for carriage of mycobacteria. A p < 0.05 was considered statistically significant. The model comprised age, occupation, residence, level of education as well as presenting symptoms such as abdominal pain, diarrhoea, vomiting, fever and weight loss. The Akaike (AIC) and Bayesian (BIC) information criteria were used for model diagnostics. The outcome variable was carriage of any nontuberculous mycobacteria. Carriage of MTB was the secondary outcome variable.
Ethical considerations
All the respondents provided written informed consent before participating in the study. Although unlinked HIV testing was done, patients who wanted to know their results were made to undergo counselling according to standard national guidelines before results were shared with them. The proposal was approved by the University of Zambia Biomedical Research Ethics Committee (Reference no. 015-07-12).