James Selib Michaelson, Ph.D
Division of Surgical Oncology

Dr James S Michaelson
James S. Michaelson, Ph.D.
Scientific Director

Bibliography / Original Articles:

Cancer Health Analysis Group
Barbara Smith, MD, PhD
Jerry Younger, MD
Leon Chen
Niraj Nathan
Matthew Nolan
Matthew Pappas
Priya Ranade
Christiana Toomey


Health Communications
Blake Cady, MD
Aaron Ball
Allan Fong
Katherine Gelber
Cynthia L. Woods

Cancer Survival Research

A major goal of cancer medicine is to be able to predict disease outcome from information on tumor and patient characteristics available at the time of detection.  We have found that for both breast cancer and melanoma, the relationship between tumor size and the chance of cancer death is well fit to a simple equation (L=1-e^-QD^Z where L is the fraction of patients dying, e is the exponential constant, D is tumor diameter or thickness, and Q and Z are parameters, whose values are specific to the cancer in question).  We have also found that for patients with tumors of equivalent size, lethality increased with the number of positive nodes, such that for breast cancer there is approximately an extra ~ 6% chance of death associated with each positive lymph node, while for melanoma there is approximately an extra ~ 23% chance of death associated with each positive lymph node.  The lethal contributions ascribable to tumor size and nodal status have proved to be roughly additive, and this has led to a new technique, the Size+Nodes method, for predicting cancer outcome.  Ongoing research concerns the refinement of methods for predicting cancer survival, including the development of more powerful methods for measuring and incorporating the lethal impact of prognostic factors such as those detected with gene expression arrays. This has also resulted in the development of a set of linked equations, the SNP [Size+Nodes+PrognosticFactors] method, which can incorporate a variety of data on primary tumor size, nodal status, and other prognostic factors into an overall estimate of the risk of death for each patient.  This general mathematical approach has also lead to equations that can predict the risk local and regional recurrence.
The Size+Nodes and SNP methods for estimating the risk of death have been incorporated into a series of web-based calculators (www.CancerMath.net and http://www.mit.edu/~lchen3/cancermath/).  Ongoing work concerns expanding the power of these calculators, so as to provide physicians with practical aids that can be used in making estimates of the risk of distant, local, and regional recurrence, and the impact that various treatment choices will have on patient outcome.
These studies rely upon accurate information on cancer outcome, and thus a major part of our work is the organization of several collaborative efforts to create large databases on patients with cancer. Two large databases, containing information of the patients with breast cancer and melanoma seen at Partners Health Care over the past 40 years, contain information from multiple sources (tumor registry entries, electronic pathology reports, social security all-cause survival information, state and national death certificate data, hospital demographic data).  These databases now contain information on more than 22,000 breast cancer and almost 8000 melanoma patients.  A similar database on leukemias and lymphomas is now under construction, and plans are underway for databases on colorectal cancers, lung cancers, and sarcomas.


Home Basic Calculator Quantitative Reports CancerHealth Analysis Group HealthCommunications Group
Copyright © 2007 James Michaelson, PhD

Massachusetts General Hospital Center for Quantitative Medicine
Massachusetts General Hospital
Harvard Medical School
Boston MA USA

Harvard University