Demographic and clinical characteristics of patients with herpes zoster at the time of presentation predict the duration and severity of pain on long-term follow-up. Analyses by Cox's proportional hazard models of six databases from controlled trials of antiviral drugs (total subjects = 2367) identified covariates for zoster-associated pain; all tests for significance were two-sided. Age strongly influenced pain outcome: patients > or = 50 years old were significantly more likely to have prolonged zoster-associated pain compared with those < 30 years old. Patients with prodromal symptoms or moderate or severe pain at presentation were also more likely to experience prolonged zoster-associated pain. Neither time to initiating treatment after rash onset nor sex of patient influenced pain outcome. Advancing age, prodromal symptoms, and acute pain severity at presentation predicted those individuals most at risk of prolonged pain and postherpetic neuralgia. When two or more of these factors were present, the risk of persistent pain was increased.