I am analysing data for colorectal cancer study
Among the data
I have 3 columns
Chemotherapy yes/no
Radiotherapy yes/no
Surgery yes/no
..
and i have years of survival
..
I want to know the relation between years of survival and type of management
i.e. which one of them has the highest years of survival
Note: some patients used more than one type of management
..
What is the best way/test to do this??
I would suggest a one way analysis of variance of survival time by treatment type ie C, R, S ,C+R, C+S, R+S, C+R+S .
Then do comparison of means , taking account of the multiple comparison problem by a standard method.
Do you have the survival times for all patients in your study? I.e. nobody was still alive at the end of observation?
While possible, it seems unlikely to me.
With right-censored data, it seems better to use some survival analysis methods, e.g. proportional hazards.
BTW: what type of conclusion do you want to draw? When you say "the relation between" are you thinking of learning something about cause, i.e. possible come to recommendations concerning management? Or is your investigation merely observational without intent to possibly influence clinical practice?
Depending on the answer, you may be interested in confounding factors.
Robert Boer Thank you for your reply
first of all, I received the data from a friend who collected it and I am trying to help him analyzing it
he had the dates of diagnosis, dates of the last visit
based on that he calculated years of survival, by subtracting the latest date from the date of diagnosis
I think it is merely observational, just want to know which type of treatment used in relation to the years of survival
these are the columns of data that I have, and I wanted help what to do
NOTE: I came up with the idea of doing one-way ANOVA survival years to type of treatment ,,
extra note: his supervisor suggested doing regression!!!!
Kaplan Meyers survival analysis or Cox proportional hazards analysis can be applied but if all are alive, then survival analysis is difficult as there is nothing to compare. An initial chi square test taking all groups can give an idea if there is any statistical difference of survival between the groups.
I would say that 'last visit' implies that the patient is still alive then.
Do you assume that every patient who stopped visiting died soon after the last visit?
And what if the last visit is close to the end of the observation period, i.e. there is no next visit in the data because observation stopped, not necessarily because the patient died?
The question of survival analysis arises only when there is death so that the analysis of survival can be done from the data by comparison. This is widely employed in cancer research where survival advantage is as best few months between different therapeutic modalities. In the quoted study whether one group lives significantly longer than other may be determined but that may not be equivalent to survival advantage.
- Badges
- Science topic
- Similar topics
- Analytical Chemistry
- Column