In this study, 28% of treatment-experienced patients with TB presented with TB resistant to at least 1 first-line drug, and MDR TB was found in 12.7% of patients. Surveys in African countries performed from 1999 through 2002 demonstrated a median prevalence of resistance to at least 1 drug among previously treated patients with TB of 16.7% (range, 0%–30.3%) [23
]. The median rate of MDR TB among previously treated patients with TB was 5.9% (range, 0%–13.7%). A recent national survey in Rwanda found that 9.4% of treatment-experienced patients with TB had MDR TB [24
]. In a referral center in Addis Ababa, Ethiopia, the prevalence of MDR TB among previously treated patients with TB was 12% [25
]. Most recently, the WHO reported that 16% of recurring cases in Senegal involve MDR TB, but this estimate was based on a small sample size [26
We observed only 2 patients with isolates that demonstrated monoresistance to rifampicin, and thus, we conclude that rifampicin resistance is an excellent marker for MDR TB, as suggested elsewhere [25
]. The unavailability of second-line drugs in Uganda presumably accounts for the low levels of resistance to second-line drugs among patients with MDR TB. The considerable levels of resistance to ethionamide may be attributable to cross-resistance with isoniazid [28
] or laboratory error [20
]. Overall resistance to ofloxacin was low; among MDR isolates, we observed a 5.8% rate of resistance to ofloxacin. These results are encouraging because of the widespread use of fluoroquinolones in the treatment of conditions other than TB.
In agreement with previous studies, we found that MDR TB at baseline was associated with a history of treatment failure [2
], an increased number of previous TB episodes [30
], and the presence of cavities on chest radiograph [31
]. In the human lung, the selection of drug-resistance mutations in M. tuberculosis
occurs predominantly within cavities, a site of high bacterial loads, active mycobacterial replication, reduced levels of anti-TB drugs, and diminished exposure to host defense mechanisms [32
]. Taken together, these findings suggest that the selection of resistance occurs over the course of several treatments, a process which culminates with MDR TB leading to treatment failure. However, our study design is unable to distinguish between the proposed cause-effect relationship and the possibility of reverse causality, in which treatment failure ultimately leads to the generation of MDR TB. We did not find an association between HIV and MDR TB. This supports previously reported data from East Africa [34
] and other developing countries [11
] but contrasts with studies from the United States, in which HIV-associated MDR TB clusters have been linked to nosocomial transmission [37
Our study has several limitations. Because DST is not routinely performed in Uganda, we were unable to determine whether MDR resulted from primary resistance or from resistance acquired during a previous TB episode. Our eligibility criteria may have excluded a disproportionate number of HIV-infected patients who are more likely to have TB with negative acid-fast bacilli smear results [38
]. However, of the 138 patients with negative smear results that we excluded, only 9 (7%) had positive culture results. Also, our study population may not be representative of all recurring cases in Uganda, because of selection and referral biases. A national survey is required to provide a more accurate picture of drug resistance in Uganda.
In this cohort of previously treated patients with TB, ~5% of patients developed drug resistance during or after therapy despite compliance to treatment; one-third of these patients developed incident MDR TB. The risk of acquired resistance was highest among patients who presented with existing drug resistance and among those whose sputum was slow to sterilize during treatment. Both of these factors have been shown to predict relapse [39
] but have not been previously associated with an increased risk of acquired drug resistance. Although the proportion of patients that acquired resistance in this study may appear to be low, our report only reflects a single cycle of treatment in a heavily treated cohort of patients (22% had received ≥2 treatment cycles before enrollment), and thus, our results may underestimate the degree of drug-resistance amplification. In addition, extrapolating our results to the rest of Uganda, where 10%–22% of the 41,000 notified cases in 2005 [14
] were cases of recurring TB, indicates that 80–170 new MDR TB cases are generated each year—a considerable public health problem. Our results and those reported elsewhere [40
] suggest that individuals who were previously treated for TB may be acting as a persistent and growing reservoir of drug resistance and as a source of contagion to the general population. This study suggests that the continued almost-exclusive focus of TB control programs on new TB cases is no longer judicious and supports the recent WHO recommendation for national programs to urgently address the growing problem of drug resistance [5
Our analysis of the acquisition of drug resistance has limitations. Mixed infections were not excluded [41
]; however, it is generally agreed that mixed infections are rare [42
]. Our analysis of acquired drug resistance was limited to patients with a suboptimal clinical or microbiological response to treatment. The development of drug resistance in other patients would have gone unnoticed. Finally, the small number of outcomes observed resulted in broad 95% CIs.
The burden of drug resistance among previously treated patients with TB in Uganda is large and growing. With currently recommended treatment regimens, the risk of generating further drug resistance is significant. There is an urgent need to focus more attention on previously treated patients with TB in low-income countries.