Should I use Chi squares?

#1
I want to see how old pancreatic cancer detection methods (slow and invasive) and newer (faster less invasive) have an effect on survival. I simply have data for the number of people that survived from pancreatic cancer in 100 people from each method. Needs to be simple analysis.
 
#2
Correction: there was 100 for the older method and 176 for the newer. 16 survived in the older methods and 34 survived in the newer.
 

hlsmith

Omega Contributor
#3
Do you have time to death as well? I know you said simple, but since methods were not randomized can you be certain the patient samples are comparable?

Would each method get used at the same time in the disease state, meaning there isn't Lead Time Bias, where one is finding disease at an earlier stage?
 
#4
Data was compiled based on demographics and medical records that assured the samples were comparable. Same time in the disease state so no lead time bias. We don’t know the time of death, just whether they survived or not and want to know if the extra time it takes to run the older tests has an effect on survival. This is just for a basic exercise, I know there is a lot here to consider.
 

j58

Active Member
#5
Assuming you can treat the data as independent samples from the old procedure and the new one, a chi-squared test would be appropriate. Computing the relative risk and it's confidence interval would provide additional insight.
 

hlsmith

Omega Contributor
#6
Yeah for simplicity you could use chi-square and eyeballing the numbers - they wouldn't be remotely significant. What extra time does it take, perhaps you should be testing duration values. Do you have those. What about treatment advances over time since one is the "old" and one is the "new" -- historic bias.