Department of Clinical Epidemiology, Aarhus University Hospital, Olof Palmes Alle 43-45, 8200, Aarhus N, Denmark.
Amgen Incorporated, Thousand Oaks, CA, USA.
Breast Cancer Res Treat. 2018 Jan;167(2):517-528. doi: 10.1007/s10549-017-4510-3. Epub 2017 Sep 25.
We developed and validated algorithms to identify metastases and breast cancer recurrence in Danish medical registries. We computed the incidence rate (IR) and hazard ratios (HRs) to evaluate predictors of these outcomes in stage II/III breast cancer patients.
We included all women in Denmark diagnosed during 1999-2011 with regional or stage II/III breast cancer. Demographic, tumor, and treatment data were ascertained from population-based health registries. To facilitate diagnostic work-up of the primary cancer, follow-up began 180 days after diagnosis and continued until recurrence/metastases, death, or 31 December 2012, whichever occurred first. We computed the positive predictive values (PPVs) of recurrence, bone metastases, and visceral metastases using medical records as a gold standard. We calculated the cumulative incidence, IR per 10,000 person years, and used Cox regression to compute the HRs and associated 95% confidence intervals (95% CI) for each outcome.
Among 23,478 patients, 7073 had regional stage and 16,405 had stage II/III breast cancer. The PPV for recurrence was 72.6% (95% CI 59.3, 83.3%). The PPVs for bone and visceral metastases were 92.3% (95% CI 69.3-99.2%) and 70.8% (95% CI 51.1, 85.9%), but had low sensitivity. Five-year cumulative incidence of recurrence, bone metastases, and visceral metastases were 18.4, 2.2, and 5.2%, with corresponding 5-year IRs of 540 (95% CI 524, 557), 60 (95% CI 55, 65), and 144 (95% CI 136, 152), respectively. Predictors of recurrence and metastases included age, stage, hormone receptor status, and cancer treatment.
Our algorithms show moderate to high PPVs for recurrence and metastases. The IRs of metastases were lower compared with other registry-based cohort studies, so may be underestimated in Danish registries.
我们开发并验证了用于识别丹麦医学登记处转移和乳腺癌复发的算法。我们计算了发病率（IR）和风险比（HR），以评估 II/III 期乳腺癌患者发生这些结局的预测因素。
我们纳入了 1999-2011 年期间在丹麦被诊断为局部或 II/III 期乳腺癌的所有女性。人口统计学、肿瘤和治疗数据均从基于人群的健康登记处获得。为了便于对原发性癌症进行诊断性检查，随访从诊断后 180 天开始，持续到复发/转移、死亡或 2012 年 12 月 31 日，以先发生者为准。我们使用病历作为金标准，计算了复发、骨转移和内脏转移的阳性预测值（PPV）。我们计算了累积发生率、每 10000 人年的发病率，并使用 Cox 回归计算了每个结局的 HR 及其 95%置信区间（95%CI）。
在 23478 例患者中，7073 例为局部期，16405 例为 II/III 期乳腺癌。复发的 PPV 为 72.6%（95%CI 59.3,83.3%）。骨转移和内脏转移的 PPV 分别为 92.3%（95%CI 69.3-99.2%）和 70.8%（95%CI 51.1,85.9%），但敏感性较低。复发、骨转移和内脏转移的 5 年累积发生率分别为 18.4%、2.2%和 5.2%，相应的 5 年发病率分别为 540（95%CI 524,557）、60（95%CI 55,65）和 144（95%CI 136,152）。复发和转移的预测因素包括年龄、分期、激素受体状态和癌症治疗。
我们的算法显示复发和转移的 PPV 为中等到较高。与其他基于登记的队列研究相比，转移的发病率较低，因此在丹麦登记处可能被低估。
