This is the first study to combine molecular genotyping techniques and drug susceptibilities to demonstrate that DOT for tuberculosis is associated with less initial drug resistance and less transmission of drug resistant organisms. Because the susceptibilities were taken from the first available isolate, the initial resistance in this study represents a combination of primary resistance and acquired resistance in patients with previous treatment. Isolates from genotype clusters containing members with initial resistance were over 1.5 times as common in the selective DOT county compared to the universal DOT county. Universal DOT was also associated with decreased transmission of drug resistant strains. Isolates from genotype clusters having 2 or more members with matching drug resistance profiles were over 4 times as common in the selective DOT county as in the universal DOT county. These findings remained significant when the cluster definition was restricted from all clusters to clusters whose members were within only one county. Since other program components in the counties are the same, this association is additional evidence that the acquisition and transmission of resistant tuberculosis is reduced by DOT.
Historically, it has been unclear whether or not DOT reduces tuberculosis transmission. It is generally accepted that most transmission of tuberculosis occurs prior to diagnosis and treatment [32
]. DOT could reduce tuberculosis transmission by ensuring that contagious patients become noninfectious as rapidly as is possible from treatment. If DOT is more effective, then tuberculosis relapse with additional transmission could be reduced.
There is large variability in the methodology of DOT programs throughout the world, as the essential components for DOT strategy success have not been systematically established through randomized trials. The World Health Organization (WHO) defines five essential elements for Directly Observed Therapy Short-course (DOTS) strategy: political commitment to control tuberculosis; microscopy services to identify tuberculosis; uninterrupted good quality drug supplies; surveillance and monitoring systems; and use of highly efficacious regimes with direct observation of treatment for at least the initial two months [15
]. These criteria are sufficiently broad that a person hospitalized for the initial 2 months of treatment with all subsequent therapy being self-administered meets the definition of DOTS strategy. Confusion can result when observation of therapy is unnecessary for using DOT strategy. This lack of standardization has resulted in studies being non-comparable.
In one previously published study, DOT was defined as therapy requiring economically impoverished ill persons without transportation to come to a clinic during working hours five times weekly for 8-12 weeks and then thrice weekly until cured [18
]. This study used a maximally restrictive definition of DOT referred to by one author as "supervised swallowing" and found both self-supervised and DOT strategies to be equally ineffective [17
]. Over 40% of patients in both the self-supervised therapy and the DOT arm failed to complete therapy [17
]. Another study defined DOT as the daily delivery of medications to the patient at home for 2 months with subsequent self-supervised therapy to cure. This study included a penalty of loss of monetary bond if the patient defaulted on treatment. This more comprehensive DOT program achieved cure rates of 85% [34
]. In much of the US, programs that strive to make tuberculosis treatment maximally accessible and include observation of therapy have been called DOT programs [35
]. Essential components of these DOT programs are considered to include supplying therapy at no cost to the patient at a location of the patient's preference, combined with adherence enhancing incentives and enablers and quarantine of the noncompliant [21
]. Some have called this a strategy of making treatment easier to take than not to take [36
]. These types of programs have resulted in a sustained reduction in relapse and both initial and acquired anti-tubercular drug resistance [21
]. In our study the components of the tuberculosis programs were the same in both counties. While the current study does not demonstrate the essential components of a DOT program, our data do demonstrate that universal DOT is associated with a significantly improved treatment outcome. Importantly, this DOT effect was independent of medication supply, incentives and enablers, laboratory services, funding, tracing and legal sanctions of defaulters, and program supervision.
There are only a few drugs and regimens of proven effectiveness in tuberculosis treatment. Drug-resistant tuberculosis is therefore a major threat to public health. If drug-resistant tuberculosis becomes widely spread, the ability to control tuberculosis through medical therapy will be seriously compromised. Not only are patients infected with strains resistant to multiple drugs less likely to be cured, but the treatment is more toxic and expensive than treatment of patients with susceptible organisms [5
]. The emergence and transmission of drug-resistant M. tuberculosis
into a population can be related to a variety of program-, health provider- and patient-related factors [38
]. Patients' noncompliance to prescribed treatment is the most important patient-related factor contributing to the development of drug resistance [9
]. Noncompliance cannot be reliably predicted but is considered to be reduced in programs that manage patients with DOT, thus resulting in less drug resistance [21
]. Our study supports this concept and preventing the development and transmission of drug-resistant organisms by DOT is, at present, the best strategy for dealing with resistant tuberculosis.
Previous tuberculosis studies combining molecular genotyping and epidemiology have demonstrated that clustering represents disease due to recent transmission. In this study clustering was greater in younger patients, males, non-Hispanic blacks, patients born in the United States, and patients with HIV co-infection, all factors previously associated with clustering and recent transmission [40
]. We observed significant differences in age and HIV co-infection between clustered isolates identified between the two counties; namely, among patients with clustered isolates, those in the selective DOT county were significantly younger and more likely to be HIV co-infected. A potential explanation is that DOT optimizes treatment response and results in the shortest possible period of contagion after diagnosis; therefore less tuberculosis transmission involving younger and HIV infected persons in the universal DOT county. It must be noted that differences in rates of HIV infection observed between the two counties may reflect differences in rates in the general populations, which were unknown.
The principal limitation of this study was that treatment by universal or selective DOT was not randomized or controlled and we are therefore unable to say with complete assurance that the associations identified are a result of DOT. Potential confounders include demographic and income differences between the counties which may have affected transmission (i.e., overcrowding, poverty). Successful tuberculosis control programs are multifaceted and include all measures related to aggressively identifying and treating patients with tuberculosis and latent tuberculosis infection. However, drug resistance occurring within tuberculosis chemotherapy programs is created only by monotherapy [41
]. We believe that the difference in DOT strategy is the only difference between the two counties that can explain the reduced acquisition and transmission of resistant tuberculosis in the universal DOT county. While we genotyped >80% of the culture-positive cases, these non-genotyped cases are a potential source of misclassification. Moreover, as the two counties are adjacent it is conceivable that more resistant tuberculosis spread from the universal DOT county to the selective DOT county, indeed 692 patients were members of clusters that spanned both counties. However, we believe this is unlikely, as resistance associations were the same in clusters exclusive to each county as clusters that were geographically distributed across both counties.