Ferrante Nicole D, Hubbard Rebecca A, Weinfurtner Kelley, Mezina Anya I, Newcomb Craig W, Furth Emma E, Bhattacharya Debika, Njei Basile, Taddei Tamar H, Singal Amit, Hoteit Maarouf A, Park Lesley S, Kaplan David, Lo Re Vincent
Division of Gastroenterology and Hepatology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania, USA.
Center for Real-World Effectiveness and Safety of Therapeutics, Center for Clinical Epidemiology and Biostatistics, Department of Biostatistics, Epidemiology, and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania, USA.
Pharmacoepidemiol Drug Saf. 2025 May;34(5):e70154. doi: 10.1002/pds.70154.
The absence of validated methods to identify cholangiocarcinoma in real-world data has prevented the conduct of pharmacoepidemiologic studies to evaluate determinants of this malignancy and examine the effectiveness of cholangiocarcinoma treatments.
To determine the accuracy of International Classification of Diseases for Oncology, Third Edition (ICD-O-3)-based algorithms to identify cholangiocarcinoma and its subtype (intrahepatic or extrahepatic) within US Veterans Health Administration (VA) data.
We identified patients with cholangiocarcinoma ICD-O-3 diagnosis codes from January 2000-December 2019 in VA data. We developed eight algorithms utilizing ICD-O-3 histology codes for cholangiocarcinoma and further used ICD-O-3 topography codes for location (liver, intrahepatic bile duct, extrahepatic bile duct) plus maximum total bilirubin (≥ 3 mg/dL vs. < 3 mg/dL) within ± 45 days of diagnosis to identify cholangiocarcinoma subtype. Up to 80 patients were randomly selected for each algorithm, and their records were reviewed by two hepatologists. The positive predictive values (PPV) and 95% confidence interval (CI) for each algorithm were estimated.
Among 2934 unique patients who met inclusion criteria, 574 were randomly selected for validation. All eight algorithms had high PPV for definite or probable cholangiocarcinoma, ranging from 83.8% (95% CI, 73.8%-91.1%) to 100.0% (95% CI, 95.5%-100.0%). Among three algorithms to identify intrahepatic cholangiocarcinoma, two had PPV ≥ 80% (range: 88.8% [95% CI, 79.7%-94.7%]-91.3% [95% CI, 82.8%-96.4%]). Among five algorithms to identify extrahepatic cholangiocarcinoma, four had PPV ≥ 80% (range: 80.0% [95% CI, 69.6%-88.1%]-94.0% [83.5%-98.7%]).
These algorithms can be used in future pharmacoepidemiologic studies to evaluate medications associated with intrahepatic or extrahepatic cholangiocarcinoma.
在真实世界数据中缺乏经过验证的胆管癌识别方法，这阻碍了药物流行病学研究的开展，无法评估这种恶性肿瘤的决定因素以及检验胆管癌治疗的有效性。
确定基于国际疾病分类肿瘤学第三版（ICD - O - 3）的算法在美国退伍军人健康管理局（VA）数据中识别胆管癌及其亚型（肝内或肝外）的准确性。
我们在VA数据中识别出2000年1月至2019年12月期间具有胆管癌ICD - O - 3诊断代码的患者。我们利用胆管癌的ICD - O - 3组织学代码开发了八种算法，并进一步使用ICD - O - 3部位代码来确定位置（肝脏、肝内胆管、肝外胆管），同时结合诊断前后±45天内的最大总胆红素（≥3mg/dL与<3mg/dL）来识别胆管癌亚型。每种算法随机选择多达80名患者，其记录由两名肝病专家进行审查。估计每种算法的阳性预测值（PPV）和95%置信区间（CI）。
在符合纳入标准的2934名独特患者中，随机选择574名进行验证。所有八种算法对确诊或可能的胆管癌都有较高的PPV，范围从83.8%（95%CI，73.8% - 91.1%）到100.0%（95%CI，95.5% - 100.0%）。在三种识别肝内胆管癌的算法中，两种算法的PPV≥80%（范围：88.8%[95%CI，79.7% - 94.7%] - 91.3%[95%CI，82.8% - 96.4%]）。在五种识别肝外胆管癌的算法中，四种算法的PPV≥80%（范围：80.0%[95%CI，69.6% - 88.1%] - 94.0%[83.5% - 98.7%]）。
这些算法可用于未来的药物流行病学研究，以评估与肝内或肝外胆管癌相关的药物。
