Our data show that with good support, TB control in Karonga District has kept drug resistance levels low, including in recurrent cases. Furthermore, despite an increase in TB incidence following the rise in HIV prevalence, and 60% of the TB cases being HIV-positive, TB incidence levels are now lower than those seen in the early years of the HIV epidemic. 
Our results are in line with national reports from the region which mostly show declining trends in TB incidence in recent years. 
Similar levels of drug resistance were seen in several cross-sectional surveys in the region (Zambia, Uganda and Botswana) in the context of generalised HIV epidemics and standard smear-microscopy based DOTS programmes. 
As in these studies, no evidence was found of an association between drug resistance and HIV status or ART use, although the numbers on ART are small in our study.
The research activities in Karonga District may have improved TB control, but we believe this is largely because we were able to ensure basic activities were carried out correctly rather than because of any high tech or costly interventions. Our system of having staff at peripheral health clinics is a form of enhanced case finding that could be replicated by providing basic training of local staff. Other, population-based, studies at KPS yield very few TB cases, suggesting that our clinic-based case detection is at acceptable levels. The immunological study may have increased case-finding in the hospital, although the proportion of all TB cases that was found on the general ward was similar in the 1997–2007 and 2008–2009 period (12.9% versus 15.7%, p
0.3) It is also likely that a proportion of the cases identified through this sub-study would have been found through regular passive case finding at a later date, albeit more spread out over time.
The culture confirmation for all individuals suspected of having TB, and speciation plus drug sensitivity for all culture-positive cases would have had limited impact on clinical care, due to the very low levels of drug resistance in our region and intensive microscopic and clinical surveillance strategy, and are not essential for achieving the results. As an indication, approximately 5% of smear-negative culture positive cases were commenced on TB treatment after a positive culture report was available (unpublished data), the overwhelming majority having already started TB treatment on clinical suspicion prior to receiving the culture result. Thus, this is unlikely to have had a major impact on TB incidence trends.
Our sputum collecting protocol for TB suspects (three samples over two days, ‘spot-morning-spot’ 
) is labour-intensive and maintained for study reasons. Alternative approaches, such as taking two spot-sputa on the same day, can be effective while reducing the laboratory and health staff workload. Our strategy of following up patients at their household if they failed to return to the clinic was also likely to be an important factor in the overall success 
Throughout, we have used treatment as determined by the NTP. In the earlier years we supported treatment, but there has been no rise in resistance or incidence following the transfer of treatment services back to the NTP. Rifampicin has been in use in the intensive phase throughout and has been part of the continuation phase since 2007, but initial MDR rates have remained low, suggesting continued absence of onward MDR transmission. The proportion of episodes with resistance was similar in clustered and unclustered strains, suggesting no overall effects on transmissibility. It will be important to continue to monitor rifampicin resistance now that it is used throughout treatment, increasing selective pressure.
These data provide a comprehensive picture of TB control in a challenging setting. The reduction in incidence occurred at a time when HIV prevalence appeared to be stable 
and before widespread use of ART. Although it could be explained by lower case detection, there is no evidence for this. Restricting the analysis to smear positive cases, for whom procedures were similar over time, should minimize this potential bias.
In Karonga District, the decline in TB incidence seems to have halted. The reasons for this are unclear. It is not explained by the extra cases identified through screening on the general wards. To bring incidence down further, more intensive screening may be needed, in high risk populations (e.g. ART clinics) or in the community. Such studies are underway in Karonga District.
Conclusion and Recommendations
Our results re-emphasize the need for doing the ‘simple things’ well. Improving TB control should include strengthening the existing, sputum smear based DOTS infrastructure, as well as support of ART programmes, which remain the cornerstones of cost-effective, equitable and sustainable TB control in many resource-limited settings. This applies in particular for peripheral rural areas, where drug resistance levels are probably low. For these areas, low-tech and low-cost improvements to TB care would be appropriate.