$underline{Objectives}$ To evaluate the accuracy of an internet-based method for survival estimation in breast cancer.
$underline{Design}$ A website was created which allows the user to enter information on prognostic factors for a patient, and instantly obtain a Kaplan-Meier survival curve based on outcome data of prior cases with a matching prognostic factor profile. The source for the survival data is a Finnish nationwide series of 2,842 women diagnosed with breast cancer in 1991-2, comprising 91% of all cases within the selected geographical regions during these two years.
$underline{Main outcome measures}$ The accuracy of the survival estimates obtained using data from the nationwide series was assessed in an independent, single institution validation series (n=565), and measured by analysis of calibration and discrimination (the area under the ROC curve).
$underline{Results}$ A selection of prognostic factors recommended by the National Institutes of Health (NIH) Consensus Development Panel and the International Consensus Panel on the Treatment of Primary Breast Cancer were made available for case-matching on the website. Kaplan-Meier case-match eight year estimates of distant disease-free survival (DDFS) based on combinations of tumour size, histologic grade and mode of detection (screen-detected vs. symptomatic) were close to the actual outcomes, e.g. patients in the validation set who were estimated to have a 71-80%, 81-90% and 96-100% DDFS had an actual average DDFS of 76%, 88%, and 100%, respectively.
$underline{Conclusions}$ A web-based case-match system can generate survival curves for user-defined prognostic factor combinations and identft patients with a varying risk for breast cancer recurrence. The system can be linked to other data sets, expanded to accommodate new prognostic factors and used as a source for population-based survival estimates.