In spite of the apparent benefit, there is debate as to whether the new drugs to treat persons infected with HIV should be made available to the homeless and marginally housed. In this commentary, the authors suggest caution in prescribing protease inhibitors for homeless persons. Issues looked at are: (1) adherence to treatment is poor enough that effectiveness is in doubt; (2) poor adherence may lead not just to low effectiveness but also to problems of drug resistance; and (3) the costs of protease inhibitors might be better spent on other interventions. The authors make a series of recommendations for physicians who treat homeless and marginally housed persons. (Authors)