Eighty-seven immunosuppressed patients (53 with localized and 34 with disseminated herpes zoster) from 10 institutions were enrolled in a controlled therapeutic trial of adenine arabinoside. A crossover design was employed; thus, 47 patients received drug for five days and then received placebo, and 40 were given the two substances in the opposite order. Resolution of acute pain and the cutaneous lesions were graded during a 10-day observation period. During the initial five-day period, treated patients showed a statistically significant resolution of pain and cutaneous lesions. Surprisingly, in many untreated patients, natural resolution occurred during this period so that crossover data could not be adequately assessed. Effects on visceral disease also could not be judged, because such disease was uncommon. Ratings of late complications such as postherpetic neuralgia were confused by the crossover design. Toxicity posed no problem. The data further emphasized the potential usefulness of adenine arabinoside as an antiviral chemotherapeutic agent, but also clearly indicated the need for a double-blind study to define this usefulness and to determine how it can most practically be used, if the risk-benefit factor remains favorable.