Ethics statement
The National Bio-ethic Committee of the Ministry of Health of Mozambique and the Institutional Review Board of the University of Washington in Seattle, USA, approved the study protocol. Both ethics committees approved that informed consent was not obtained as the study was based on routinely collected data.
Study design and setting
We performed a retrospective observational study in three purposely-selected health facilities in Manica province, Mozambique. Selection criteria were the presence of both TB and HIV treatment services in the same facility and at least 150 TB patients notified in 2007. One facility was an urban health centre in the provincial capital; the other two were rural health facilities about 20 and 90 kilometres from the provincial capital. Within these clinics, we collected the information on HIV disease parameters of all new notified TB patients of 16 years and older with a positive HIV test recorded in the TB register from January to December 2007.
In Mozambique, smear microscopy is the main TB diagnostic. In the participating facilities, diagnosis of sputum smear-negative and extrapulmonary TB occurs mainly on clinical assessment and hardly ever on radiology. All new adult TB patients receive a standard course of TB treatment consisting of two months isoniazid, rifampicin, ethambutol and pyrazinamide followed by 4 months isoniazid and rifampicin. The standard first line ART regimen consists of two nucleoside reverse-transcriptase inhibitors, lamivudine and stavudine, with either the non-nucleoside reverse transcriptase inhibitor (NNRTI) nevirapine or efavirenz. The national guidelines recommend switching from neviripine to efavirenz in patients that receive a rifampicin containing treatment regimen [7].
Data collection
The facility’s TB supervisor collected the data of the 2007 cohort using standard data collection forms in July and August 2009. Data collected from the TB register included: age, sex, type and category of TB, treatment regimen, start date of TB treatment, initial smear examination result, HIV test result and TB treatment outcome. If the treatment outcome was death, its date was recorded.
We identified the HIV record of the HIV-positive TB cases through the unique HIV patient number if recorded in the TB register. In addition, local staff familiar with the patients identified some HIV patient records. If these methods did not lead to identification of the patient record, we searched the electronic HIV-database using the patient’s name and age taken from the TB register. If the data matched, we took the unique HIV patient number from the electronic database and used it to locate the HIV patient record. We limited the identification of the HIV patient record to those HIV-positive TB patients registered with the HIV services in the same health facility.
We collected available CD4+ cell count results in the 6 months TB treatment period, the date of these results, the start date for ART and the ART regimen from the HIV patient record.
Statistical analysis
We entered the data in EpiData version 3.1 and performed descriptive analysis with EpiData Analysis V2.2.1.171. We used STATA version 11 (StataCorp, College Station, Texas, USA) for analysis of the CD4+ cell counts.
We modelled the evolution of the CD4+ cell count during TB treatment using a mixed effect model. This model deals adequately with repeated measurements of the outcome variable [13]. The model incorporates estimated values for missing data based on all other available data. With this model we used optimally all available CD4+ cell counts including all patients with at least one CD4+ cell count in the model, regardless of the number of missing values these patients have. We used a random intercept model with an independent covariance structure for estimation of the CD4+ cell count over time. We compared the mean CD4+ cell count for the time updated variables of TB treatment use and ART use. In addition we included age and sex in the model as potential confounding variables. The model used the absolute CD4+ cell count values to estimate the effect of TB treatment and ART on CD4+ cell response. We assumed that once a patient starts ART, the patient continues ART until the end of the observation period. We did not adjust for the type of NNRTI because there is no evidence that there is a differential CD4+ count increase when comparing a nevirapine-based or efavirenz-based regimen [14].
Definitions
We defined the baseline CD4+ cell count at the start of TB treatment as the CD4+ count closest to the start of TB treatment within a window of 12 weeks before until 2 weeks after the start of TB treatment. We allocated all other CD4+ cell counts to a single fixed time-point with a window ranging from 2 to 6 weeks. If multiple CD4+ cell counts were available for a specific time-point, we included the one closest to the midpoint of the time window in the analysis.
To determine whether the CD4+ cell count was obtained while using TB treatment, we used WHO standard treatment outcomes to define the end of TB treatment [15]. The end of TB treatment for treatment ‘success’ (cured or treatment completed) was 180 days after start of TB treatment. For ‘failure’ (smear positive after five months for sputum smear positive cases), the end of treatment was 150 days (5 months) after start. For ‘default’ (interrupted treatment for two or more consecutive months), ‘transfer out’ (transferred to another TB unit with unknown treatment outcome), or ‘unknown’ the end of TB treatment was 90 days after the start date. For patients who died, TB treatment ended at the date of death or 90 days after start of TB treatment if unknown.