The highest incidences of esophageal and gastric cardia cancer in the world occur in northern China. Chinese scientists have developed esophageal balloon cytology screening to detect these cancers, but traditional cytology is sometimes inadequate to find some early, curable lesions. Several studies suggest that quantitative fluorescence image analysis (QFIA) of DNA ploidy and nuclear morphology may be able to improve upon traditional cytology results. In October 1987, esophageal balloon cytology was performed on 1331 adults in Linxian, China, and all samples were evaluated both by traditional cytology and QFIA. From 1987 to May 1991, 62 new squamous esophageal cancers and 44 new adenocarcinomas of the cardia were identified in this cohort. Proportional hazards models were used to evaluate the relationship of cytological diagnoses and six QFIA variables to subsequent cancer risk. These models showed significant trends for increasing esophageal cancer risk, with increasing values in five of the QFIA variables and with increasing severity of the traditional cytological diagnoses. A comparison of models with only cytology variables versus models with both cytology and QFIA variables indicated that the QFIA provided an important additional predictive value. Persons with both cytological dysplasia and high cellular DNA were 8 times more likely to develop esophageal cancer than were individuals with neither of these conditions. For cardia cancer, associations between QFIA variables or cytological diagnoses and later cancer were more limited. This study suggests that the QFIA variables evaluated here are independent predictors of squamous esophageal cancer and that combining QFIA with traditional cytology can improve prediction of esophageal cancer risk.