Grunkemeier G L, Li H H, Starr A
Providence Health System, Portland, Oregon, USA.
J Heart Valve Dis. 1999 Sep;8(5):466-70; discussion 470-1.
Having performed our first heart valve replacement in 1960, we began a prospective lifetime follow up service for all patients, contacting them at least annually to determine survivorship and heart valve complications.
We reviewed isolated aortic (AVR) and mitral (MVR) valve replacements from 1960 to 1993, with follow up to 1998. In total, 2,942 AVR and 1,579 MVR were performed, with 21,742 and 12,142 patient-years of follow up, respectively. Analysis of the results affords an opportunity to demonstrate the usefulness and necessity of certain statistical methods, including multivariable event-free analyses and cumulative incidence functions.
The survival rate was 8% at 30 years for both valve positions. However, an overall survival curve is an artificial composite of patients of increasingly higher risk being served during increasingly safer years of calendar time. One result is that, for AVR, age is not a significant univariate risk factor for operative mortality, but is highly significant after accounting for date of surgery using logistic regression. Long-term mortality is higher for tissue valves than for mechanical valves; but mean age is greater (74 versus 57 years), and after accounting for age using Cox regression, mortality is similar for both valve types. Kaplan-Meier analysis estimates thromboembolism occurrences of 85% for AVR and 95% for MVR at 35 and 34 years, respectively, but the cumulative incidence estimates are only 32% and 41%, respectively.
Prospective follow up for over 35 years has provided an opportunity to illustrate important statistical issues: Multivariate analyses are essential to avoid being misled by excluding important risk factors or including artifactual ones, and the cumulative incidence estimates the percentage of patients who will actually experience a complication.
自1960年完成首例心脏瓣膜置换手术以来，我们开始对所有患者进行前瞻性终身随访服务，至少每年与他们联系一次，以确定生存率和心脏瓣膜并发症情况。
我们回顾了1960年至1993年期间单纯主动脉瓣置换术（AVR）和二尖瓣置换术（MVR），随访至1998年。共进行了2942例AVR和1579例MVR，分别有21742和12142患者年的随访时间。对结果的分析提供了一个机会来证明某些统计方法的实用性和必要性，包括多变量无事件分析和累积发病率函数。
两个瓣膜位置在30年时的生存率均为8%。然而，总体生存曲线是在日历时间越来越安全的年份中，为风险越来越高的患者构建的人为综合曲线。结果之一是，对于AVR，年龄不是手术死亡率的显著单变量风险因素，但在使用逻辑回归考虑手术日期后，年龄具有高度显著性。组织瓣膜术后的长期死亡率高于机械瓣膜；但组织瓣膜患者的平均年龄更大（74岁对57岁），在使用Cox回归考虑年龄因素后，两种瓣膜类型的死亡率相似。Kaplan-Meier分析估计，AVR和MVR在35年和34年时血栓栓塞发生率分别为85%和95%，但累积发病率估计分别仅为32%和41%。
超过35年的前瞻性随访提供了一个机会来说明重要的统计问题：多变量分析对于避免因排除重要风险因素或纳入人为因素而产生误导至关重要，累积发病率可估计实际发生并发症的患者百分比。
