降钙素原测量不确定度在脓毒症相关决策阈值附近的蒙特卡洛建模
本研究在12个月内对高流通量三级实验室的554,570次降钙素原(PCT)检测进行测量不确定度(MU)评估,使用七台Siemens Atellica IM模块和BRAHMS PCT免疫分析平台,采用Nordtest自顶向下法结合蒙特卡洛模拟。模块特异性扩展MU为21.08%至26.96%,合并估计为24.22%。阈值附近结果存在重新分类风险,应结合临床背景解读而非固定分类。
要点
- PCT模块特异性扩展测量不确定度为21.08%至26.96%,合并估计为24.22%。
- 0.50 ng/mL阈值处蒙特卡洛最宽区间为0.34-0.66 ng/mL,含2693例低于和1376例高于阈值。
- 阈值附近PCT结果应结合临床背景解读,不宜作为固定分类判断。
核心结论
- 模块特异性和周期间分析变异可能增加PCT结果在决策阈值附近的测量不确定度。
- 阈值附近结果应结合临床背景解读,而非作为固定分类发现。
研究设计与样本
- 12个月内在高流通量三级实验室使用七台等效Siemens Atellica IM模块和同一BRAHMS PCT免疫分析平台完成554,570次PCT检测。
- 采用Nordtest自顶向下法估计MU,并用蒙特卡洛模拟对每个模块的周期内不精密度和周期间分析变异建模。
- 在0.50和2.00 ng/mL决策阈值及补充10.00 ng/mL参考水平评估自顶向下区间,蒙特卡洛区间仅计算两个阈值。
- 高水平室内质控(IQC)敏感性分析重新计算10.00 ng/mL附近的重现性分量。
主要发现
- 模块特异性扩展MU范围为21.08%至26.96%,主要分析合并估计为24.22%,高水平IQC敏感性分析为24.36%。
- 0.50 ng/mL处蒙特卡洛最宽区间为0.34-0.66 ng/mL,含2693例低于和1376例高于阈值。
- 2.00 ng/mL处蒙特卡洛最宽区间为1.37-2.63 ng/mL,含935例低于和571例高于阈值。
- 10.00 ng/mL处自顶向下合并区间为7.58-12.42 ng/mL,含288例低于和199例高于参考水平。
关键数值与推荐
- 模块特异性扩展MU范围为21.08%至26.96%。
- 主要分析合并MU估计为24.22%,高水平IQC敏感性分析为24.36%。
- 0.50 ng/mL阈值处蒙特卡洛最宽区间为0.34-0.66 ng/mL。
- 2.00 ng/mL阈值处蒙特卡洛最宽区间为1.37-2.63 ng/mL。
- 10.00 ng/mL参考水平处自顶向下合并区间为7.58-12.42 ng/mL。
临床怎么用
- 对PCT结果接近0.50或2.00 ng/mL决策阈值者,应结合临床背景综合判断,避免仅凭固定切点分类。
- 实验室应关注模块间和周期间分析变异对阈值附近PCT结果判读的影响。
局限性
- 蒙特卡洛区间仅针对0.50和2.00 ng/mL两个决策阈值计算,未覆盖10.00 ng/mL参考水平。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
BACKGROUND: Procalcitonin (PCT) supports sepsis assessment and antimicrobial stewardship; however, the interpretive implications of measurement uncertainty (MU) near decision thresholds remain poorly defined. In this study, PCT MU was estimated across modules and patient-level results susceptible to threshold-adjacent reclassification were quantified. METHODS: Over 12 months, 554,570 PCT measurements were obtained in a high-throughput tertiary-care laboratory using seven equivalent Siemens Atellica IM modules on the same BRAHMS PCT immunoassay platform. MU was estimated using a Nordtest-based top-down approach, complemented by Monte Carlo simulation to model within-period imprecision and between-period analytical variation for each module. Top-down intervals were evaluated at the 0.50 and 2.00 ng/mL decision thresholds and at a supplementary 10.00 ng/mL reference level, whereas Monte Carlo intervals were calculated for the two thresholds. A high-level internal quality control (IQC) sensitivity analysis recalculated the reproducibility component near 10.00 ng/mL. RESULTS: Module-specific expanded MU ranged from 21.08% to 26.96%; pooled estimates were 24.22% in the primary analysis and 24.36% in the high-level IQC sensitivity analysis. The widest Monte Carlo-derived intervals were 0.34-0.66 ng/mL at 0.50 ng/mL, containing 2693 results below and 1376 above the threshold, and 1.37-2.63 ng/mL at 2.00 ng/mL, containing 935 below and 571 above. At 10.00 ng/mL, the pooled top-down interval was 7.58-12.42 ng/mL, containing 288 results below and 199 above the reference level. CONCLUSIONS: Module-specific and between-period analytical variation may contribute to measurement uncertainty in PCT results near decision thresholds. Threshold-adjacent results should be interpreted within the clinical context rather than as fixed categorical findings.
引用
- Haytaç B, Gök G, Erel Ö, et al. Procalcitonin measurement uncertainty around Sepsis-related decision thresholds: Monte Carlo modeling from a large-scale clinical laboratory dataset. Clin Chim Acta. 2026. PMID: 42580611. doi:10.1016/j.cca.2026.121275
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。