贝叶斯算法快速准确诊断肝素诱导性血小板减少症
该研究评估三种检测抗PF4/肝素抗体的免疫分析法的性能,并基于贝叶斯结合验前概率与似然比建立诊断算法,在回顾性(n=221)和前瞻性(n=305)推导队列中推导,并在前瞻性验证队列(n=687)中验证。结果显示该算法能在60分钟内准确纳入或排除至少95%的疑似HIT病例。
要点
- 算法结合验前概率与一线及二线免疫分析法似然比,60分钟内可处理≥95%病例
- HemosIL-AcuStar-HIT-IgG作为唯一检测用于82.4%病例,分析周转时间30分钟
- 仅2.9%病例未解决,7.4%预测为HIT,87.9%排除HIT
核心结论
- 结合验前概率与一线和二线抗PF4/肝素抗体免疫分析法可准确地在60分钟内纳入或排除至少95%的疑似HIT病例。
研究设计与样本
- 回顾性推导队列n=221和前瞻性推导队列n=305,前瞻性验证队列n=687;以肝素诱导血小板活化试验(HIPA)为金标准功能试验。
主要发现
- 在验证队列n=687中,算法准确预测HIT 51例(7.4%),排除HIT 604例(87.9%),仅20例(2.9%)未解决。
- 566例(82.4%)仅使用HemosIL-AcuStar-HIT-IgG,分析周转时间30分钟;121例(17.6%)未解决者加用ID-H/PF4-PaGIA二线检测,额外周转时间30分钟。
- 12例(1.7%)算法阳性预测未被HIPA确认,其中10例HIPA阴性但可能HIT,2例可能为算法假阳性。
关键数值与推荐
- HemosIL-AcuStar-HIT-IgG对HIPA阳性的100%阳性预测值截断值为>3.0 U/mL,100%阴性预测值截断值为<0.13 U/mL。
- ID-H/PF4-PaGIA对HIPA阳性的100%阳性预测值截断值为滴度≥16,100%阴性预测值截断值为≤1。
临床怎么用
- 对疑似HIT患者,可先检测验前概率,再结合一线免疫分析法(如HemosIL-AcuStar-HIT-IgG)的似然比进行贝叶斯判断,必要时加做二线检测,以快速纳入或排除诊断。
局限性
- 研究未明确报告样本来源的单中心或多中心性质、验前概率的具体评估方法及算法的外部适用性。
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
Prompt diagnostic evaluation of suspected heparin-induced thrombocytopenia (HIT) is critical for guiding initial patient management. We assessed the performance of 3 immunoassays detecting anti-platelet factor 4 (PF4)/heparin antibodies, derived a diagnostic algorithm with a short analytical turnaround time (TAT), and prospectively validated the algorithm. Plasma samples were analyzed by Zymutest-HIA-IgG, HemosIL-AcuStar-HIT-IgG, and ID-H/PF4-PaGIA in retrospective (n = 221) and prospective (n = 305) derivation cohorts. We calculated likelihood ratios of result intervals and cutoff values with 100% negative (NPV) and positive (PPV) predictive values for a positive gold standard functional assay (heparin-induced platelet activation [HIPA]). A diagnostic algorithm was established based on the Bayesian combination of pretest probability and likelihood ratios of first- and second-line immunoassays. Cutoffs with 100% PPV for positive HIPA were >3.0 U/mL (HemosIL-AcuStar-HIT-IgG) and titer ≥16 (ID-H/PF4-PaGIA); cutoffs with 100% NPV were <0.13 U/mL and ≤1, respectively. During the prospective validation of the derived algorithm (n = 687), HemosIL-AcuStar-HIT-IgG was used as unique testing in 566 (82.4%) of 687 cases (analytical TAT, 30 minutes). In 121 (17.6%) of 687 unresolved cases, ID-H/PF4-PaGIA was used as second-line testing (additional TAT, 30 minutes). The algorithm accurately predicted HIT in 51 (7.4%) of 687 patients and excluded it in 604 (87.9%) of 687 patients, leaving only 20 (2.9%) cases unresolved. We also identified 12 (1.7%) of 687 positive predictions not confirmed by HIPA: 10 patients with probable HIT despite negative HIPA and 2 possible false-positive algorithm predictions. The combination of pretest probability with first- and second-line immunoassays for anti-PF4/heparin antibodies is accurate for ruling in or out HIT in ≥95% of cases within 60 minutes. This diagnostic approach improves initial management of patients with suspected HIT.
引用
- Marchetti M, Barelli S, Zermatten MG, et al. Rapid and accurate Bayesian diagnosis of heparin-induced thrombocytopenia. Blood. 2020. PMID: 31945147. doi:10.1182/blood.2019002845
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。