Clinical Predictors of Drug Resistance and Mortality Among Tuberculosis Patients in a Rural South African Hospital: A Case-Control Study

The recent discovery of a high prevalence of multidrug-resistant (MDR-TB) and extensively drug-resistant tuberculosis (XDR-TB) in rural South Africa, where HIV is rampant, has provoked alarms about the future of tuberculosis control in the region. Little is known about the clinical manifestations of MDR-TB in general, and XDR-TB in particular, in the high HIV prevalence settings of Sub-Saharan Africa. We performed a retrospective, case-control study of patients diagnosed with tuberculosis at a rural hospital in KwaZulu Natal, South Africa, where large numbers of MDR-TB and XDRTB cases have been identified. All MDR-TB and XDR-TB patients who began treatment for TB between June 1, 2005 and August 31, 2006 and whose charts were available were included in the study. A comparison group of patients without resistance to both isoniazid and rifampicin (non- MDR-TB), matched 1:1 with the size of the MDR-TB and XDR-TB groups, was created. Clinical and laboratory data were obtained through review of hospital records, clinic registers, and the laboratory system. We compared clinical characteristics to identify risk factors for MDRTB, XDR-TB, and mortality. Bivariate and multivariate analyses were performed. A total of 170 patients were enrolled in the study: 52 MDR-TB, 61 XDR-TB and 57 non-MDR-TB. Greater than 75% of patients from all groups were tested for HIV; HIV prevalence among those tested was 94% in the non-MDR group, 93% in the MDR group, and 100% in the XDR-TB group (P=1.000 for MDR versus non-MDR; p=0.089 for XDR versus non-MDR). Forty percent of MDR-TB patients and 57% of XDR-TB patients had no previous history of TB treatment, strongly suggesting transmitted drug resistance. Significant associations and risk factors for MDR-TB and XDR-TB in bivariate analysis included positive sputum smear (P=0.015, P=0.005), TB treatment in the past year (P\u3c0.0001, P\u3c0.001), and hospitalization in the past two years (P=0.007, P=0.004). In multivariate logistic regression, positive sputum smear remained a significant risk factor for XDR-TB (adjusted odds ratio (AOR) 2.79, 1.20-6.47), and TB treatment in the past year remained a risk factor for both MDR-TB and XDR-TB (AOR 8.33, 95% CI 1.64-42.33; AOR 7.19, 95% CI 1.35-38.17). Mortality for the non-MDR, MDR and XDR groups was 36.8%, 73.1% and 85.3%, respectively (P= 0.0001 for MDR versus non MDR; P\u3c0.0001 for XDR versus non-MDR; P=0.109 for XDR versus MDR); median survival from TB diagnosis was 199 days, 103 days, and 92 days, respectively (P\u3c0.001). In Cox Proportional Hazards model, positive sputum smear (P=0.003), MDR-TB (P=0.028), XDR-TB (P=0.002), and CD4 cell count less than 200 cells/mm3 (P=0.037) were significant risk factors for mortality. Forty of the 170 patients had sputum isolates with differing resistance patterns, and 18 moved from a lower to a higher resistance category; this increasing drug resistance appeared to be more likely the result of super-infection than amplification. A significant proportion of MDR-TB and all XDR-TB appear to be due to primary resistance, with nosocomial transmission playing a critical role. MDR-TB and XDR-TB carry extraordinarily high mortality rates in this setting; previous hospitalization, previous TB treatment, positive sputum smear and low CD4 count may be used to target drug susceptibility testing for patients at high risk of drug resistant TB and mortality