• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information

#### When Survival Statistics Are Not Misleading

Although survival statistics are misleading in assessing the value of screening, they are not misleading when used for two other purposes: as a prognosis for an individual patient and as an outcome in a randomized trial.

When you are diagnosed with a serious illness, survival statistics are used to estimate your prognosis—your chance of surviving for a fixed period of time. As we’ve discussed throughout the book, these statistics are most relevant to you when they are based on people like you, people who are similar in age and gender and who have similar health conditions.

Another characteristic that determines your prognosis is the stage of the disease (how advanced it is). The most useful survival statistics are those generated for people whose illness is at the same stage as yours. For example, in early-stage cancer, the cancer may be localized to one organ. By the intermediate stage, it may have spread to the tissue around the organ. Late-stage cancer means that it has spread to other parts of the body. The federal government has created a great source of stage-specific survival statistics for cancer, known as SEER (Surveillance, Epidemiology, and End Results).5

Survival statistics are also a meaningful way to judge the benefit of a treatment—for example, does a new cancer drug prolong life? To answer this question, researchers conduct randomized trials comparing a group of people taking the new drug to a control group of people who are not taking it. The results of these trials yield the starting and modified risks (similar to the risks we calculated for Zocor in chapter 4). The outcome is often expressed as a 5-year survival statistic: the fraction of people in each group who survive for 5 years divided by the number of people in that group.

Consider this example: drug X is used to treat people diagnosed with stage 1 pancreatic cancer. In a randomized trial, people age 50 to 75 years with stage 1 pancreatic cancer were randomly assigned to take either drug X or the standard drug (the control group). Drug X improved survival.

Because randomization creates groups that should be similar in every way (including how they were diagnosed), comparing these starting and modified survival statistics provides a good measure of how well treatment works.

Know Your Chances: Understanding Health Statistics.
Woloshin S, Schwartz LM, Welch HG.
Berkeley (CA): University of California Press; 2008.