Tumor Center Regensburg/ University of Regensburg, Institute for Quality Control and Health Services Research, Regensburg, Germany; Department of Health Technology and Services Research, Technical Medical Centre, University of Twente, Enschede, The Netherlands.
Department of Health Technology and Services Research, Technical Medical Centre, University of Twente, Enschede, The Netherlands.
Value Health. 2020 Sep;23(9):1149-1156. doi: 10.1016/j.jval.2020.05.012. Epub 2020 Aug 14.
An important aim of follow-up after primary breast cancer treatment is early detection of locoregional recurrences (LRR). This study compares 2 personalized follow-up scheme simulations based on LRR risk predictions provided by a time-dependent prognostic model for breast cancer LRR and quantifies their possible follow-up efficiency.
Surgically treated early patients with breast cancer between 2003 and 2008 were selected from the Netherlands Cancer Registry. The INFLUENCE nomogram was used to estimate the 5-year annual LRR. Applying 2 thresholds, they were defined according to Youden's J-statistic and a predefined follow-up sensitivity of 95%, respectively. These patient's risk estimations served as the basis for scheduling follow-up visits; 2 personalized follow-up schemes were simulated. The number of potentially saved follow-up visits and corresponding cost savings for each follow-up scheme were compared with the current Dutch breast cancer guideline recommendation and the observed utilization of follow-up on a training and testing cohort.
Using LRR risk-predictions for 30 379 Dutch patients with breast cancer from 2003 to 2006 (training cohort), 2 thresholds were calculated. The threshold according to Youden's approach yielded a follow-up sensitivity of 62.5% and a potential saving of 62.1% of follow-up visits and €24.8 million in 5 years. When the threshold corresponding to 95% follow-up sensitivity was used, 17% of follow-up visits and €7 million were saved compared with the guidelines. Similar results were obtained by applying these thresholds to the testing cohort of 11 462 patients from 2007 to 2008. Compared with the observed utilization of follow-up, the potential cost-savings decline moderately.
Personalized follow-up schemes based on the INFLUENCE nomogram's individual risk estimations for breast cancer LRR could decrease the number of follow-up visits if one accepts a limited risk of delayed LRR detection.
原发性乳腺癌治疗后随访的一个重要目标是早期发现局部区域复发（LRR）。本研究比较了两种基于乳腺癌 LRR 风险预测的个体化随访方案模拟，并量化了它们可能的随访效率。
从荷兰癌症登记处选择了 2003 年至 2008 年间接受手术治疗的早期乳腺癌患者。应用 INFLUENCE 列线图估计 5 年的年度 LRR。应用 2 个阈值，根据 Youden 的 J 统计量和预先设定的 95%随访敏感性进行定义。这些患者的风险估计作为安排随访的依据；模拟了 2 种个体化随访方案。与当前荷兰乳腺癌指南推荐和训练队列中观察到的随访利用情况相比，比较了每种随访方案的潜在节省随访次数和相应的成本节约。
使用来自 2003 年至 2006 年的 30379 名荷兰乳腺癌患者的 LRR 风险预测（训练队列）计算了 2 个阈值。Youden 方法的阈值产生了 62.5%的随访敏感性和潜在节省 62.1%的随访次数和 5 年内节省 2480 万欧元。当使用对应 95%随访敏感性的阈值时，与指南相比，随访次数减少 17%，节省 700 万欧元。在应用这些阈值于 2007 年至 2008 年的 11462 名测试队列患者时，得到了类似的结果。与观察到的随访利用情况相比，潜在的成本节约适度下降。
基于 INFLUENCE 列线图对乳腺癌 LRR 的个体风险估计的个体化随访方案，如果接受延迟 LRR 检测的风险有限，可以减少随访次数。
