Spss Survival

This is the first of three concerned with survival analysis, this looks at the Kaplan-Meier plot and Logrank and Breslow statistics www.robin ...

Hi
I am doing a statistical data analysis on cancer treatments for Small Cell Lung Cancer.
I have two groups (A and B), and 2 drugs (P and E). Both Groups are given both drugs, but group A is given drug E then drug P, and Group B is given drug P then drug E. So there is a difference in the regimens of the drugs given which are being compared.
The data sets are the number of days the patients have survived either on regimen A (drug E then P) or regimen B (drug P then E).
Some of the data is right censored e.g. a patient who has survived 765 days and continues to live after the study is terminated has the survival time recorded as 765+, meaning the patient has survived beyond 765 days.
Also, because the patients did not begin the treatments at the same time the censored data all have different values. For example, one patient may have survived 765+ days and another may have survived 256+ days.

What I am trying to find out is if there is any difference in the average survival time between the two treatment regimens.

My questions:
1. how do I treat the censored data? Do I just ignore the "+" and use, for example, use 765 days
instead of 765+ days in my analysis? If not, should I calculate the the survival rate and estimate when the patient is expected to pass away?

2. Am I correct in saying that I need to test for the equality of means between group A and group B? If so, what test should I use? (ANOVA or a T-test? Or some other test?) Also, the number of observations in the groups are not equal. Group A has 62 observations and Group B has 59 observations.
3. Can anyone tell me how to make a box plot in Excel? Preferably, side by side box plots.

Please Note:
This is not for a real analysis of cancer treatments or regimens, but is for an assignment.
Also, all of my analysis must be done in excel, so please don't tell me to plug my data into SPSS or another statistical software program.
Also, this assignment is for a first year, undergraduate, statistical computing class, so please understand that my statistical knowledge is rather limited, and also that this assignment has more to do with being able to show that I am able to do a statistical analysis in Excel, than showing my statistical abilities (obviously to a certain extent) and bio statistics.

Any help would be greatly appreciated. Thanks :)

It would be helpful to know more specifically what your statistical background is, especially whether you have studied any survival analysis. A simple graphical summary would be a Kaplan-Meier plot (google it) - they're quite simple to make even if you haven't studied survival analysis.

If you haven't learnt any techniques to deal with censored data, my advice would be to leave them out as long as there aren't too many. This will obviously bias your results, but it's a better solution than ignoring the +.

You could then just do a (unpaired) t-test on the remaining survival times, comparing the two groups. It's a little crude but should give something interesting. The unequal sample sizes doesn't matter for unpaired t-tests.

Obviously, if there are marks for it, it would be well worth commenting on all the assumptions you have made etc.

HTH

2009-11-17 23:18:16 by sally