Study on quantifiable method strategy and statistical evaluation system for qualitative testing proficiency testing in food and related fields
摘要
目的 针对食品领域定性检测能力验证中样品均匀性与稳定性难以量化、实验室结果评价依赖简单正确率等问题,构建基于检出概率的可量化方法策略与统计评价框架.方法 系统整合 ISO/IEC 17043:2023 与 ISO 13528:2022 的技术规范,针对含底层连续测量值的检测方法,建立由批内标准差、判定阈值和分类越界误判概率构成的路线 A;针对无稳定底层连续测量值输出的纯分类判定方法,建立基于独立伯努利试验、二项分布和 Clopper-Pearson 精确置信区间的路线 B.在此基础上,进一步构建包括真值赋值、假阳性/假阴性非对称风险加权、类 z 分数和加权 Kappa 系数在内的综合评价框架.结果 研究明确了两类定性能力验证项目的路线选择依据,并提出了将样品均匀性、稳定性验证与实验室能力评价相衔接的统计流程.在农药残留快速定性检测案例中,路线 A 将底层连续测量值的批内波动转化为分类越界误判概率,计算得到临界阳性样品的理论误判概率为 0.08%.在食源性沙门菌定性筛查案例中,非对称风险加权评价识别出的异常或预警实验室由传统单一正确率评价的 5 家增加至9 家,额外提示 4 家存在临界浓度条件下的重复假阴性风险.结论 该双轨方法策略可为食品域定性能力验证方案设计、验证物品质量控制和实验室能力评价提供方法学参考,但具体参数、风险权重和判定阈值仍需结合实际进一步验证.
Objective To address the difficulties in quantifying homogeneity and stability,as well as the reliance on simple accuracy-based evaluation in qualitative proficiency testing(PT)in food and related fields,this study aims to construct a quantifiable methodological strategy and statistical evaluation framework based on the probability of detection(POD)model.Methods The technical specifications of ISO/IEC 17043:2023 and ISO 13528:2022 and relevant guidelines for qualitative method validation were integrated.For testing methods that generate underlying continuous measurements,Route A was established using the within-batch standard deviation,decision threshold and probability of classification-boundary misclassification.For purely qualitative methods without stable underlying continuous measurements,Route B was established based on independent Bernoulli trials,the binomial distribution and the Clopper-Pearson exact confidence interval.A comprehensive evaluation framework including assigned value determination,asymmetric risk weighting of false positives and false negatives,z-like scores and weighted Kappa coefficient was further developed.Results The selection criteria for the two routes were clarified,and a statistical workflow linking sample homogeneity and stability verification with laboratory performance evaluation was established.In the rapid qualitative pesticide-residue detection case,Route A converted the within-batch variation in the underlying continuous measurements into a classification-boundary misclassification probability,resulting in a theoretical misclassification probability of 0.08%for the borderline positive sample.In the Salmonella qualitative screening case,asymmetric risk-weighted evaluation increased the number of laboratories identified as abnormal or requiring warning from 5 under conventional accuracy-based evaluation to 9,additionally identifying 4 laboratories with repeated false-negative risks at borderline concentrations.Conclusion The proposed dual-track strategy provides a methodological reference for qualitative PT scheme design,test item quality control and laboratory performance evaluation in food and related fields.However,the specific parameters,risk weights and decision thresholds require further validation using additional practical PT projects.