Periodic examination of patients after potentially curative treatment for breast cancer is routine practice; until recently no one seriously questioned it [ 1,2]. A number of good reasons support the continued testing and reexamination of the 182,000 women and 1,000 men treated for this disease each year. Among them are the need to determine the outcome of treatment, the prompt detection and treatment of recurrences, and the detection of new primaries at a potentially curable stage. Psychological support and reassurance for the patient are also counted among the virtues of continued follow-up. Of late, however, an increasing number of questions are being raised about the value of follow-up as generally practiced, particularly regarding expensive testing to detect asymptomatic distant recurrences. More than one author have suggested that a simpler process would serve patients equally well and be less burdensome [3-131.
How much follow-up is too much? How much is too little? Who should be responsible for it, how frequently should it be done, and using what methods? Is follow-up a standard of practice with all the opportunities that deviation from it raises for litigation [14]? In the current climate of rising health care costs, these are legitimate questions. A number of clinical studies are beginning to provide useful information about some of these issues.
A compelling argument for follow-up is the need to determine treatment outcome. Breast cancer is an indolent disease with the capacity to recur and cause death after many years of apparent health. Since no current treatment provides assurance of cure, it is not possible to know who is cured and who is not without extended observation. Even then, the risk of recurrence never reaches zero. Thus cure must be defined in terms that recognize this continuing uncertainty, such as with fixed end point determinations of 5 year or 10 year survival, or with statistical projections of survival probabilities. The rate of attrition of the treated population from recurrence and death is a widely accepted measure of treatment outcome, and this information is unobtainable without