Application of the statistical safety evaluation method to the analysis for reflux cooling in the loss of residual heat removal system (RHR) event during mid-loop operation was conducted with RELAP5/MOD3.2 as the best-estimate code. The evaluation parameters for this event are the reactor coolant system (RCS) pressure and reactor vessel (RV) water level. A phenomena identification and ranking table was developed previously to identify important phenomena that have a large influence on the evaluation parameters. The uncertainties of the analytical models which are related to the important phenomena were quantified to use as input for the statistical analysis. Actual plant sensitivity analyses were carried out to confirm the influence of the uncertainty which was quantified. Validation analysis of the RELAP5/MOD3.2 code was carried out for the reflux cooling test conducted with the ROSA-IV Large-Scale Test Facility to confirm that the experimental RCS pressure results were within the range of the uncertainty of the analytical models. A statistical analysis for 93 cases was carried out for a four-loop plant to evaluate the RV water level in the steady state and the maximum RCS pressure with 95% cumulative probability and 95% confidence. As a result, it was confirmed that reflux cooling using two cooling steam generators was successful.