of a respiratory protection intervention. Until such data are available (and it is far from clear that the necessary studies will be undertaken), the appropriate level of respiratory protection will likely continue to be an area of debate.

It is not a surprise to find that observational studies of the effects of implementing tuberculosis controls in health care setting have not clearly demonstrated the independent effect of the respiratory protection components of these programs. In large part, this is due to the simultaneous implementation of several control improvements with the consequent inability to contrast outcomes where individual elements of the control program are present versus when they are absent. In addition, the studies typically involve relatively small numbers of workers who convert their tuberculin skin tests following periodic (usually yearly) testing. If respirators have a small but positive effect, such studies (and epidemiologic studies generally) will lack the statistical power to detect the effects.

Much of the literature on respirator efficacy is based on theoretical and empirical data which demonstrates that respirators can reduce the exposure to airborne contaminants by factors ranging from 2.4 to greater than 200 (Barnhart et al., 1997). Two papers have modeled the potential for respirators to reduce risk for tuberculin skin test conversion based on data from a series of elegant experiments by Riley and colleagues (Riley et al., 1959, 1962). First, Riley and colleagues noted the rate at which nurses converted their skin tests in tuberculosis wards and calculated on the basis of their expected minute ventilation the estimated concentration of infectious particles in the air of these wards. Then based on these estimates they performed an experiment using guinea pigs as a bioassay and calculated the average airborne production of infectious particles generated by patients with infectious tuberculosis. These data based on direct monitoring of tuberculosis infection from airborne droplet nuclei provide, perhaps, the strongest data on risks to workers exposed to patients with infectious tuberculosis.

In two complementary papers Barnhart and colleagues (1997) and Fennelly and Nardell (1998) model the potential benefits of respiratory protection. Both papers rely on published estimates of the ability of respirators to reduce exposure to airborne particles. Inherent in this reliance is an assumption that tuberculosis particles or droplet nuclei will be filtered out by the medium just as other hazardous particles such as silica, asbestos, or plutonium are. While it is well recognized that fit factors under static conditions vary considerably from those under work conditions,