Aim:Coldness is a common complaint in kampo medicine, and is commonly treated with unkeito, but recommended indications (so-called Kuketsu) are inconsistent. The aim of the present study was therefore to identify a set of self-reported symptoms that would predict a positive outcome of treatment for coldness with unkeito, and that would therefore serve as specific indications for treatment with unkeito. Methods: This was a retrospective cohort study of 70 female patients (median age, 33 years; range, 22-71 years) without organic abnormalities who complained of coldness, and who completed a self-report questionnaire about their health using the present kampo questionnaire database system on their first visit. Another 26 female patients (median age, 38 years; range, 22-60 years) were assigned to verify the discriminatory predictive ability of the developed logit model. Patients were treated with two or three packages of unkeito per day for 1 month, and the efficacy of treatment was then recorded as a binary response. Results: Unkeito decreased frequency and severity scales (0-4) of coldness from a median of 3.0 (range, 1-4) to 2.0 (range, 0-4; P < 0.001) and from 3.0 (range, 2-4) to 2.0 (range, 0-4; P < 0.001), respectively. Both frequency and severity of coldness were improved in 77.1% of patients. The developed model for prediction of treatment effect consisted of dry lips (Akaike information criterion (AIC), −14.10), with coldness in the feet (AIC, −25.01) but not in the whole body (AIC, −25.01). Model validation with 26 new patients gave a predictive accuracy of 77%. Conclusion: Dry lips, accompanied by coldness in the feet, but not in the whole body, are important indications for treatment with unkeito.