Percutaneous endoscopic gastrostomy (PEG) is used to provide nutrition for patients who are unable to eat but have a functionally intact gut. Clinical guidelines for PEG are uncertain and have been derived mainly from referral practices. We performed a population-based cohort study in 97 residents of Olmsted County, Minnesota, referred for PEG between January 1982 and December 1988 to determine complications, duration of tube feeding, and survival. Follow-up continued until death or February 1990. Inpatient and outpatient records were reviewed to determine indications, comorbid conditions, level of consciousness, and limitations in activities of daily living. Outcomes determined after referral for PEG included type and number of complications, tube removal, and survival. Statistical methods used included Kaplan-Meier and proportional hazards regression analyses. PEG placement was successful in 94% of patients. Although complications occurred in 70% of patients, they usually were minor (88%) and most occurred within 3 months. In 24 patients, tubes were removed because eating was resumed. The probability of surviving 30 days, 1.5 years, and 4 years after referral for PEG was 78%, 35%, and 27%, respectively. The major causes of death within and after 30 days were pneumonia, heart disease, and vascular disease of the central nervous system. An increased risk of death after referral for PEG placement was associated with older age, male gender, diabetes, and specific indications for PEG. If validated in other population-based studies, these predictors of survival after referral for PEG placement could be used to identify patients with a low probability of survival who may not benefit from PEG.