Brenner Hermann, Hakulinen Timo
Division of Clinical Epidemiology and Aging Research, German Cancer Research Center, Bergheimer Strasse 20, D-69115 Heidelberg, Germany.
Cancer Epidemiol Biomarkers Prev. 2007 Aug;16(8):1675-81. doi: 10.1158/1055-9965.EPI-06-1046.
Period analysis has been shown to provide more up-to-date estimates of cancer survival than traditional methods of survival analysis. There is, however, a tradeoff between up-to-dateness and precision of period survival estimates: increasing up-to-dateness by restricting the analysis to a relatively short period, such as the most recent calendar year, goes along with loss of precision. Recently, a model-based approach was proposed, in which more precise period survival estimates for the most recent year can be obtained through modeling of survival trends within a recent 5-year period. We assess possibilities to extend the time window used for modeling to come up with even more precise, but equally accurate and up-to-date estimates of prognosis. Empirical evaluation using data from the Finnish Cancer Registry shows that extension of the time window to about 10 years provides, in most cases, as accurate results as using a 5-year time window (whereas further extension may lead to considerably less accurate results in some cases). Using 10-year time windows for modeling, SEs of survival estimates can be approximately halved compared with conventional period survival estimates for the most recent calendar year. Furthermore, we present a modification of the modeling approach, which allows extension to 10-year time windows to be achieved without the need to include additional cohorts of patients diagnosed longer time ago and which provides similarly accurate survival estimates at comparable levels of precision in most cases. Our analyses indicate opportunities to further maximize benefits of model-based period analysis of cancer survival.
与传统的生存分析方法相比，时期分析已被证明能提供更及时的癌症生存估计。然而，时期生存估计的及时性和精确性之间存在权衡：通过将分析限制在相对较短的时期（如最近一个日历年）来提高及时性，会伴随着精确性的损失。最近，有人提出了一种基于模型的方法，通过对最近5年期间的生存趋势进行建模，可以获得最近一年更精确的时期生存估计。我们评估了扩展用于建模的时间窗口的可能性，以得出更精确、同样准确且及时的预后估计。使用芬兰癌症登记处的数据进行的实证评估表明，在大多数情况下，将时间窗口扩展到约10年能提供与使用5年时间窗口一样准确的结果（而在某些情况下，进一步扩展可能会导致准确性大幅降低）。与最近一个日历年的传统时期生存估计相比，使用10年时间窗口进行建模，生存估计的标准误可大约减半。此外，我们提出了一种对建模方法的修改，该方法允许扩展到10年时间窗口，而无需纳入更早诊断的额外患者队列，并且在大多数情况下，在相当的精确水平上能提供同样准确的生存估计。我们的分析表明，有机会进一步最大化基于模型的癌症生存时期分析的益处。
