To improve the evaluation method of non-invasive detection of blood glucose by Optical Coherence Tomography(OCT), a method combining the relative variance of OCT signal slope with the relative standard error was presented to evaluate the measurement effect and to obtain the theoretical precision of blood glucose detection. Firstly, the effect of the OCT confocal function on the blood glucose detection and the computing method of the relative error of OCT signal slope were given. Then, the relative variance and standard error of the OCT signal slope via a vivo experiment on 3 New Zealand rabbits were carried out to evaluate the non-invasive detection of blood glucose. Obtained results indicate that the relative variance and standard error of OCT signal slope are (3.7 %)/mmol & 0.068,(3.4 %)/mmol & 0.026,(0.93 %)/mmol & 0.019, respectively,and the theoretical measuring precision of blood glucose detection are ±3.7 mmol,±1.5 mmol and ±4.1 mmol,respectively. It concludes that the the evaluation method of combining the relative variance and standard error of OCT signal slope can greatly improve the theoretical measuring precision of blood glucose detection,and the larger the fitting interval is,the higher the measuring precision is.