We believe it is important to make survival rates from a high-quality data source available to our patients whenever possible. As you review the data, there are a few limitations you may wish to consider. These pertain to all survival data reported by Memorial Sloan Kettering and to other cancer programs that participate in the National Cancer Data Base (NCDB). Other data sources have similar limitations.
The methods we use to calculate survival rates for cancer are not risk adjusted. Risk adjustment means accounting for differences among patients such as age, gender, socioeconomic status, preexisting health conditions, and other factors that may influence an individual patient’s outcome. In other words, cancer survival rates do not reflect the many factors beyond treatment that may have an impact on a patient’s chances of survival. Researchers at Memorial Sloan Kettering, the American College of Surgeons, and other institutions are working to create new methods that incorporate these factors as well as those related to certain tumor characteristics.
The reported outcomes do not discern between deaths from cancer and deaths from other causes. This means that people who died from a cause other than cancer are included in the results.
Survival versus cure
The survival rate is often stated as a five-year survival rate. This refers to the percentage of people with a particular cancer who are alive five years after their diagnosis. It is important to understand that survival up to a particular point in time, such as five years, is not the same as being “cured.” The particular point at which a patient may be declared cured varies among cancer types.
Memorial Sloan Kettering data is also included in the comparison data of the NCDB, as this is how the NCDB provides it to us. This means that the NCDB data is actually closer to our own data than it would have been if our data had not been included. This difference is likely very small.
All results are based on a 95 percent confidence interval. You may be familiar with this concept from listening to the news, when the results of a political poll with a 95 percent confidence interval are given with a margin of error of plus or minus 2.5 percentage points. What this means is that if the poll were taken 100 times, 95 percent of the time the results would fall within 2.5 percent of the original polling results.
In order to make the graphs easier to read, we have elected not to show the confidence intervals. However, such detailed statistical information can provide important distinctions about our outcomes. If the 95 percent confidence intervals for a given set of curves overlap, for example, that would suggest that any apparent difference in the survival rates of Memorial Sloan Kettering versus those of the NCDB is not significant from a statistical point of view. On the other hand, if the confidence intervals did not overlap, that would suggest that the observed difference between the curves is less likely to be due to chance alone. We have made a note at the bottom of each line graph in cases where the confidence intervals do not overlap and therefore suggest a statistically significant difference in survival.