Predictors of recurrent TB in sputum smear and culture positive adults: a prospective cohort study

Objective To explore simple inexpensive non-culture based predictors of recurrent pulmonary tuberculosis (PTB). Setting and study population HIV-infected and uninfected adults with the first episode of smear positive, culture-confirmed pulmonary tuberculosis in a high tuberculosis burden country. Design A nested prospective cohort study of participants with pulmonary tuberculosis (PTB) presenting to a hospital out-patient clinic. Results A total of 630 TB culture confirmed participants were followed up for eighteen months of which 57 (9%) developed recurrent recurrent TB. On univariate analysis,4.7% low grade(1+) pre-treatment sputum smear participants developed recurrent tuberculosis Vs 8.8% with high grade(3+) smears (OR=0.31,95%CI: 0.10–0.93, p=0.037).On multivariate analysis: participants with extensive fibro-cavitation had a high risk of recurrent TB Vs minimal end of treatment fibro-cavitation (18%Vs12%, OR=2.3,95%CI:1.09–4.68, p=0.03). Weight gain with HIV infection was assosciated with a high risk of recurrent TB Vs weight gain with no HIV infection(18%Vs 6%, OR=6.8,95%CI:165–27.83, p=0.008) where as weight gain with a low pre-treatment high bacillary burden was assosciated with a low risk of recurrent TB Vs weight gain with a high pre-treatmentbacillary burden(6.5%Vs7.9%, OR=0.2,95%CI:0.05–0.79, p=0.02). Conclusion Extensive end of treatment pulmonary fibro-cavitation, high pre-treatment bacillary burden with no weight gain and HIV infection could be reliable predictors of recurrent tuberculosis.


Introduction
Long term, non-relapsing cure is the ultimate goal of anti-TB treatment; unfortunately recurrent disease(relapse/ re-infection) occurs on average in about two to six percent of patients with drug-susceptible TB treated with modern rifampicin-containing short course regimens 12 . Recurrent TB may result from recurrent disease with the original or new infection strain of M. tuberculosis 28 .This may be due to several scenarios that include: poor adherence, immunosuppression mainly due to HIV co-infection 10,13,28 , Initial TB infection with a high bacillary burden 17 , drug resistant strains 30 , sub-optimal exposure in fixed dose combinations 1 ,variation in biological response to TB treatment 31 or exogenous re-infection with a new strain 23 . Clinicians have sought to identify factors associated with an increased risk for recurrent TB after successful initial treatment. This can guide them in identifying patients and groups where treatment might be modified or where post-treatment follow-up might be useful. Lack of weight gain 8 , HIV infection 13 , pre-treatment greater radiographic involvement 9 , pre-treatment high bacillary burden 17 ,post treatment residual fibro-cavitation 9 , sub-optimal exposure in fixed dose combinations 1 , cavitary disease on chest ra-diograph& a positive sputum culture after completing the first two months of anti-TB treatment have been associated with recurrent tuberculosis.
Sputum culture is costly and less affordable in many high TB burden settings 26 . Genexpert is as well costly and less specific in diagnosis of recurrent TB 26,27 . There is a programmatic intuitive need to identify simple, robust, inexpensive, convenient, non culture (non-molecular) based tools that rely on readily available assessments at the point of care to identify patients at risk for recurrent TB. Most TB suspects have a chest radiograph 31 and sputum smear done during evaluation 33 . Patients with newly diagnosed TB are offered HIV testing by most programmes 32 and even peripheral TB units have a weighing balance to measure body weight for anti-TB drug dosing. Sputum smear microscopy is a commonly used TB diagnostic tool even in resource constrained settings 33 . To our knowledge, no previous study has ever come up with a predictive non-culture based tool for recurrent TB from the list of previously identified risk factors. In this prospective cohort study, we examined whether one or any combination of weight gain, sputum smear grade, pre-treatment cavitation, end of treatment residual fibro-cavitation, radiographic extent of TB disease and HIV co-infection could form a non-culture based a predictive tool to identify patients at risk for recurrent TB in the initial eighteen months post-treatment in a well-followed up treatment cohort in Uganda, a nation with a high burden of TB.

Study population and setting
The participants included in this prospective cohort study were systematically selected from two parent studies. The parent studies recruited participants from several health units within a radius of 30 kilometers from the research Site in Kampala. This analysis included data from HIV-infected and -uninfected adults over 18 years-old with the first episode of smear positive, culture-confirmed, drug-susceptible pulmonary TB, treated in the Kawempe study(DMID protocol number:01-005) conducted from April 2002 to July2012 and the EDCTP TB diagnostics study(EDCTP -IP_09_32040_011)conducted from March 2010 through March 2013.Both studies were conducted at a large out-patient TB clinic within Mulago Hospital, a national tertiary referral hospital in Kampala,Uganda . All patients were treated with self-administered, standard short course therapy (2 months of daily INH, rifampicin, ethambutol and pyrazinamide followed by daily treatment with Isoniazid and rifampicin for 4 months). The TB drugs were dispensed as single drugs. 1 The Kawempe study enrolled TB index cases and followed up their household contacts .The Kawempe study embedded a mixture of cohort and cross sectional designs. A Kawempe community resident was defined as one who should have resided in the Kawempe area for at least three months. The Kawempe study excluded participants with recurrent TB and those who had taken more than 5 doses of anti-TB drugs for that episode of TB at the time of enrolment.Participants were followed up for eighteen months post TB treatment. A team of home health visitors ensured routine weekly field visits, phone calls and continued counseling of the participants to mitigate loss to follow up. The EDCTP TB diagnostics study was a prospective cohort study that evaluated TB cytokine signatures in blood, sputum and urine as markers of active TB infection in relation to sputum culture. It enrolled both controls and active TB cases, first episode of TB and excluded those who had taken more than 5 doses of anti-TB drugs.. All culture confirmed TB cases from both parent studies were systematically selected and screened for this analysis. Data was extracted using a tested and validated questionnaire by two staffs who were blinded to the study objective.

Procedures
HIV infection was defined as having a positive result on three approved, licensed rapid HIV testing kits in a validated HIV testing algorithm.Each participant had a posterior-anterior chest radiograph taken pre-treatment, end of treatment and at the time of suspected TB recurrence. The radiographic severity of TB disease was assessed as minimal, moderate or far advanced disease using the scheme of the US National TB and Respiratory Disease Association 16 . Body weight was measured at baseline and at each follow up visit using a calibrated beam balance (Detecto model). Sputum smear and culture were done monthly during the treatment phase, every 6 months during follow up and at the time of suspected recurrent TB. Sputum smears were done using fluorescent auramine staining. Smear grading was by the World Health Organi-zation scale 20 . Sputum culture was on Lowenstein-Jensen slants. Mycobacterium tuberculosis speciation was by polymerase chain reaction (PCR) 24 . Initial isolates had drug susceptibility testing done for INH, rifampicin, ethambutol, pyrazinamide and streptomycin. All participants with suspected recurrent TB were evaluated by clinical examination, chest radiography, multiple sputum smears and cultures. Recurrent TB was defined as culture confirmed TB during 18 months of follow up in participants who were microbiologically confirmed cured at the end of the initial six months of TB treatment. Weight gain was defined as gaining at least 5% of the pre-treatment body weight by the end of the first two months of anti-TB treatment 8 .
Treatment adherence was monitored by urine INH metabolite testing in addition to participant interview at each scheduled visit and review of attendance records. Defaulters were traced by home health visitors. All participants were followed up for 18 months after TB treatment. Both parent studies were approved by institutional review boards of the Joint Clinical Research Centre, the National AIDS Research Committee and the Ugandan National Council for Science and Technology. All participants gave written informed consent for participation in the parent studies.

Statistical analysis
Categorical data was analyzed using the chi-square. For weight gain, changes from baseline by month 2 were as-sessed using paired z-tests. Logistic regression was used in determining the association between weight gain, HIV status, radiographic extent of disease, sputum smear grade, end of treatment fibro-cavitation and the likelihood of TB recurrence.Stepwise selection method of predictor variables(weight gain,HIV status,chest x-ray extent of disease,fibro-cavitation and bacillary burden) was used for inclusion in the logistic model. All tests were two-sided. The odds ratio plus the corresponding 95% confidence interval and p-value were the effect measure. A confidence interval excluding the null value of 1 and a p-value<0.05 were considered statistically significant. All statistical analyses were done using Stata (Version 12, STATA, USA).

Discussion
In this analysis,we report a 2 fold reduction in the risk of recurrent TB among participants with low grade compared to high grade pre-treatment smears (4.7%Vs8.8%) and inversely an increase in the risk of recurrent TB among participants with extensive end of treatment fibro-cavitary sequellae compared to those with minimal fibro-cavitary sequellae (18%Vs12%). Rupka Singla et al 2007 and panJab et al 2007 showed that high pre-treatment bacillary burden and residual cavitation by the end of TB treatment increases the risk for recurrent TB. Luzze et al 2013 showed that extensive end of treatment fibrosis and cavitation was associated with recurrent TB. Therefore our findings are similar to what has been reported in previous studies 9,17,28 . Participants with a low grade pretreatment smear and weight gain had an even lower risk of recurrent TB but weight gain alone did not independently influence the risk of recurrence 8 . We however observed a high proportion of failed weight gain in the recurrent TB group compared to their non-recurrent counter parts (67% Vs 59%) but this was not statistically significant although Khan et al 2006 8 showed that failure to gain ≥ 5% of baseline body weight by the end of the intensive phase of TB treatment was associated with an increased risk for TB recurrence. Our study found no statistically significant difference on univariate analysis although there was a higher proportion of failed weight gain in the recurrent Vs the non-recurrent group. We also observed an increased risk of recurrent TB among the HIV infected with weight gain Vs HIV uninfected with weight gain. Previous studies have shown increased rates of TB recurrence among HIV-TB patients 10 , this was the case with our study results too. We noted a higher pro-portion of recurrent TB (9%) in this population contrary to previously existing scientific data 12 however,our results are in line with the findings of one previous study in the same population 28 . The high proportion of recurrent TB could be attributed to patients presenting late with advanced PTB 9 as evidenced in this analysis where the majority( 49%) of the participants had radiographically far advanced disease. This is in agreement with Panjab et al 2007 findings of greater radiographic involvement being a risk factor for recurrent TB Urine INH metabolite testing was not done in 19% of the cases in which case non-adherence could have confounded recurrent PTB 29 .
We also noted that about 54% of the cases presented with a high pre-treatment bacillary burden which could partially explain the high proportions of recurrent TB in this population 17 . The strength of our study is that it involved an eighteen month follow-up of a large cohort of confirmed TB patients and monthly assessment of adherence by urinary INH metabolite testing during treatment. Our study had several limitations (i) the urine INH metabolite testing was only possible in 81% of the cases included in this analysis. (ii)DNA finger printing was not done to distinguish recurrent TB due to relapse from exogenous reinfection with a different strain 28 .

Recommendation
Previous studies have listed several factors assosciated with TB disease recurrence but no predictive tool has been assembled out of this literature to influence management of recurrent TB. We recommend that clinicians should put into perspective the possibility of prolonging treatment and close patient monitoring for TB pa-tientswho :(i) present with a high pre-treatment bacillary burden (3+) and failure to gain at least 5% of baseline body weight by the end of the intensive phase of TB treatment. (ii)Are HIV positive.(iii) have extensive end of treatment fibrosis and residual cavitation on chest x-ray (grade 3).