Background: Alkaline phosphatase (ALP) is routinely analyzed in clinical laboratories for the comprehensive assessment of hepatic and osteal diseases. The official reference measurement procedure (RMP) of ALP was released by the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) in 2011. However, most of the commercial kits still trace back to the old version (1983). There was a difference between the trace measurement procedure and the RMP. Therefore, the discrepancy among clinical systems and application of the new RMP for ALP in clinical laboratories was studied. Methods: According to the recommendation of the IFCC, the RMP for ALP (2011) was reproduced. The reference measurement system (RMS) for serum ALP and 19 clinical systems were included in the external quality assessment (EQA). The relative bias was calculated between the clinical systems' results and RMS, as well as each clinical system and the average value of all clinical systems. The qualified rates (passing score in percentage) for the 19 clinical systems were compared by using two different standards. In the comparison experiment, two clinical systems were evaluated before and after calibration by RMP. The clinical acceptability at the medical decision point was evaluated. Results: The performance of the reproduced RMP for ALP was: total imprecision was 0.33% and 0.42% at 336.9 U/L and 138.74 U/L, respectively. The accuracy was in the acceptable range. Excellent linearity was obtained for linear regression (R = 0.9998). In the EQA experiment, the relative bias of clinical systems and RMP ranged from -26.36% to 19.49%, and the majority of them had a negative value. Relative bias of clinical systems and the average value of 19 clinical systems ranged from -24.28% to 33.48%. The qualified rate for clinical systems was 53% - 89% evaluated by Standard 1 and was 95% - 100% evaluated by Standard 2. In the comparison experiment, the relative bias for the two clinical systems decreased and both of the clinical systems showed less relative bias at the medical decision points after calibration by RMP. Conclusions: There was a much higher discrepancy among clinical systems for the testing of serum ALP. Traceability and standardization would likely be improved for clinical systems by the application of RMP for ALP (2011) in clinical laboratories.