ABSTRACT
Survival is complementary to mortality, but the statistical use of the term is for the duration of survival. The duration of survival of a patient with cancer after detection of malignancy and median survival time in cases of leukemia are common examples. Sometimes the interest is in the percent surviving for a specific duration such as 5-year survival rate of end-disease renal patients on dialysis. In a generic sense, the method of survival analysis is used to analyze any data on
duration. This can also be understood as time-to-event data. It could be duration of stay in the hospital, birth interval, duration of immunity, or any such duration. Why cannot durations be analyzed as any other quantitative variable? One diffi-
culty is that most durations do not follow a Gaussian pattern. Their distribution is highly skewed. This is not a big problem since this can be overcome by choosing a sample of sufficiently large size. The real problem arises when the duration for some subjects is not fully observed. The person moves out, dies due to an unrelated cause, refuses to cooperate after follow-up, etc. The primary cause of incomplete observation, however, is that the end point event does not occur during the follow-up period; the study is terminated after a fixed time. This results in incomplete segments, called censored observations. The method of survival analysis is geared to meet this contingency, which can rarely be handled in any other manner. In addition, this chapter also describes some other methods to study the duration of survival. It goes without saying that for any duration there must be a beginning (entry) point
and an end (exit) point. In the case of diabetes, for example, the beginning point can be the time when the first symptom was noted, when the disease was first diagnosed, when blood glucose level showed a significant rise from the preexisting level, when a treatment was started, or any other point of time considered appropriate. The end point can bewhen blood glucosewas first noted back to the normal level, when the treatment was discontinued, when the person is able to eat normally, or any other. Because of multiple choices, you can see that both the beginning point and the end point should be sharply defined for the duration to be observed without error. It is also important to decide whether the intervening period between the entry and exit has to be completely free of events and whether the occurrence of other events is to be disregarded.
