Cameron F E Platell1, James B Semmens. 1. University Department of Surgery, Fremantle Hospital, Fremantle, Western Australia. cplatell@cyllene.uwa.edu.au
Abstract
PURPOSE: Actuarial and Product Limited (i.e., Kaplan-Meier) estimates of survival are commonly used in the literature to describe outcomes in patients treated for cancer. Terms such as cancer-specific and cancer-free survival are frequently quoted, although often without clear definitions. This study was designed to compare survival estimates using the Kaplan-Meier method on the same population of patients but using different definitions of what constitutes an event. This was to highlight some of the variation that can occur when different techniques are used to perform these calculations. METHODS: Data were obtained from a prospective database that had recorded all patients presenting with colorectal cancer from 1996 to 2002. Using this information, we calculated the 1) overall (all-cause mortality), 2) cancer-specific, 3) cancer-free, 4) recurrence-free, and 5) relative survival (and 95 percent confidence intervals) at five years postpresentation. RESULTS: The study included 497 patients with a mean age of 68 years, and a male-to-female ratio of 1.3:1. They were followed for a mean of 2.2 years (standard deviation, +/-1.1), with 50 patients (10.1 percent) followed for more than five years. The various survivals at five years were: 1) overall survival, 55.6 percent (95 percent confidence interval, 49.1-62.1 percent), 2) cancer-specific survival, 67 percent (95 percent confidence interval, 60.9-73.1 percent), 3) cancer-free survival, 49.9 percent (95 percent confidence interval, 43.6-56.2 percent), 4) recurrence-free survival, 43.5 percent (95 percent confidence interval, 37.2-49.8 percent), and 5) relative survival, 73.4 percent (95 percent confidence interval, 65.4-81.4 percent). CONCLUSIONS: The five-year survival calculations for this group of patients with colorectal cancer varied by as much as 30 percent depending on how the data was censored. This highlights that there needs to be a clear and accountable definition on how survival curves are calculated and presented in the literature to allow for meaningful interpretation and comparisons.
PURPOSE: Actuarial and Product Limited (i.e., Kaplan-Meier) estimates of survival are commonly used in the literature to describe outcomes in patients treated for cancer. Terms such as cancer-specific and cancer-free survival are frequently quoted, although often without clear definitions. This study was designed to compare survival estimates using the Kaplan-Meier method on the same population of patients but using different definitions of what constitutes an event. This was to highlight some of the variation that can occur when different techniques are used to perform these calculations. METHODS: Data were obtained from a prospective database that had recorded all patients presenting with colorectal cancer from 1996 to 2002. Using this information, we calculated the 1) overall (all-cause mortality), 2) cancer-specific, 3) cancer-free, 4) recurrence-free, and 5) relative survival (and 95 percent confidence intervals) at five years postpresentation. RESULTS: The study included 497 patients with a mean age of 68 years, and a male-to-female ratio of 1.3:1. They were followed for a mean of 2.2 years (standard deviation, +/-1.1), with 50 patients (10.1 percent) followed for more than five years. The various survivals at five years were: 1) overall survival, 55.6 percent (95 percent confidence interval, 49.1-62.1 percent), 2) cancer-specific survival, 67 percent (95 percent confidence interval, 60.9-73.1 percent), 3) cancer-free survival, 49.9 percent (95 percent confidence interval, 43.6-56.2 percent), 4) recurrence-free survival, 43.5 percent (95 percent confidence interval, 37.2-49.8 percent), and 5) relative survival, 73.4 percent (95 percent confidence interval, 65.4-81.4 percent). CONCLUSIONS: The five-year survival calculations for this group of patients with colorectal cancer varied by as much as 30 percent depending on how the data was censored. This highlights that there needs to be a clear and accountable definition on how survival curves are calculated and presented in the literature to allow for meaningful interpretation and comparisons.
Authors: M Jane Mohler; Stephen Joel Coons; Mark C Hornbrook; Lisa J Herrinton; Christopher S Wendel; Marcia Grant; Robert S Krouse Journal: Curr Med Res Opin Date: 2008-06-09 Impact factor: 2.580