Abatract：Objective: To design a risk-based multi-stage internal quality control (IQC) scheme for two kinds of HbA1c analyzers to facilitate the detection systems with different sample sizes maintaining quality more economically. Methods: The biases of two HbA1c analyzers（Sebia Capilarys 2FP and Premier Hb 9210）were obtained by participating in the Trueness Verification Program organized by National Center for Clinical Laboratories of National Health Commission, and the coefficient of variation was derived from routine IQC data. Then, the Sigma metric was calculated using the bias and coefficient of variation. According to Sigma metric level, the daily maximum number of patient samples and the desired number of specimen between two result reports, combined with Sigma-metric run size nomogram and power function graph, the run size, probability of error detection and probability of false rejection of QC schedules were determined for the corresponding IQC schedule of different sigma performances respectively. Finally, an appropriate IQC scheme was developed. Results: The Sigma metrics of Premier Hb 9210 and Sebia Capilarys 2FP analyzers were 4.96σ and 5.35σ respectively, and the maximum workload of the these two analyzers were 200 patient samples. For Premier Hb 9210 and Sebia Capilarys 2FP analyzers, two levels of controls were analyzed once in the startup IQC event; the IQC procedures were 1_3s/2_2s/R_4s N=2 and 1_3s N=2 respectively. Then, only one level of control was analyzed alternatively in each subsequent bracketing event (1_3s N=1) for every 50 patient samples. For the analyzers with 5σ performance and the maximum workload of 1000 patient samples, two levels of controls were analyzed twice in the startup QC event (1_3s/2_2s/R_4s/4_1s N=4) but only once in each subsequent bracketing QC event (1_3s N=2) for every 200 patient samples. For analyzers with 5σ performance and the maximum workload of 500 patient samples, the two levels of controls were analyzed only once in the startup QC event (1_3s/2_2s/R_4s N=2) and only one level control was analyzed once in each subsequent bracketing event (1_2.5s N=1) every 125 patient samples. Conclusion: The multistage IQC scheme which enploy a startup IQC event followed by periodic bracketing IQC events can effectively control the quality of the HbA1c analyzer for continuous production, and could be optimized to minimize the risk of harm to patients.