The aim of this paper is to evaluate development of the non-invasive blood glucose measurement of near infrared (NIR) spectroscopy. The results showed that NIR spectroscopy might obtain glucose concentration of up to 200% difference under a same environmental condition with two months apart due to time and machine drifts. These effects can restrict the development of the non-invasive blood glucose measurement. Partial least square (PLS) regression was used, which showed advantage over using simple absorbance for glucose concentration. Non-invasive blood glucose measurement of health subjects (non-diabetics) was also investigated. The results showed that R correlation coefficient of prediction (Rp) was 0.48 and root mean square of prediction (rmsep) was 1.34 mmol/l. The error was mainly due to the physiological effect of different subjects.