Almost a third of the world's population is infected with Mycobacterium tuberculosis, the organism that causes tuberculosis disease. Most of those infected never fall ill, but individuals who do can recover if they have access to effective therapies. This paper discusses certain ethical and ethnographic issues raised by cases in which patients are infected with M. tuberculosis strains resistant to at least the two most powerful drugs on which therapy is usually based. In most poor countries, people with such multidrug-resistant tuberculosis (MDR-TB) were, until very recently, considered "untreatable." In addition to being consigned to a permanent state of ill health, they were also at risk of transmitting their resistant strain to others. In this paper we discuss the logic of "cost-effectiveness," which international health policy-makers utilized to make the case that treatment of MDR-TB is not feasible in resource poor settings. These analyses, which have held sway in public health policy for many years, are flawed, we argue, because they ignore and conceal the social determinants of access to health services and often rely on assumptions rather than evidence. We propose that policies based solely on analyses of cost-effectiveness of specific interventions for individual settings can be short-sighted and, because they do not pay sufficient attention to the social, political, economic, epidemiological and pathophysiological factors influencing the production of health, will ultimately hinder progress toward effective global TB control.