Sample sizer calculation for intestinal surgery

#1
Hello there, i'm a clinical researcher currently involved in setting up a study to determine whether a certain profylactic treatment to reduce post-operative complications is effective. However, i keep running into all kinds of different sample size calculations and different effect size indicators and I have no clue which one to use anymore.

in short:
The design would be a placebo-controlled randomized controlled trial where one treatment arm would receive this profylactic treatment and one would receive a placebo.
My outcome is dichotomous: the occurence of a post-operative complication yes/no.

From what i've read I think i would need to calculate my sample size using the Chi-squared test, but i'm not a 100% sure, since i might also apply multiple logistisc regression for analysis of the results.

One of the methods I tried was trying G-power, in which i chose the Chi-squared test + goodness of-fit tests a priori: I used a power of 80% and an alpha of 0.05. what i'm left with is the Df(degrees of freedom??) and effect size W. The biggest problem is that i have no clue wat to use for effect size and Df.

Another method was using the browser based calculator clinalc, in which i only had to fill out the anticipated proportions in the treatment and placebo group in combination with power, alpha and an enrollment ratio(=1). The problem here is that this results in huge sample sizes which would be unfeasible.

Hopefully somebody can help me out!
If any additional information is required i'll try to awnser.
 

Karabiner

TS Contributor
#3
The design would be a placebo-controlled randomized controlled trial where one treatment arm would receive this profylactic treatment and one would receive a placebo.
It is a bit surprising for an intervention study that no certified statistician or someone with equal qualifications
needs to be consulted to do the sample size calculations.

Or, is this just preliminary consideration whether the study can be carried out at all, regarding the probably required sample size?

With kind regards

Karabiner
 

Karabiner

TS Contributor
#6
One of the methods I tried was trying G-power, in which i chose the Chi-squared test + goodness of-fit tests a priori: I used a power of 80% and an alpha of 0.05. what i'm left with is the Df(degrees of freedom??) and effect size W. The biggest problem is that i have no clue wat to use for effect size and Df.
Have you considered something like this https://stats.oarc.ucla.edu/other/gpower/two-independent-proportions-power-analysis/ ?
Another method was using the browser based calculator clinalc, in which i only had to fill out the anticipated proportions in the treatment and placebo group in combination with power, alpha and an enrollment ratio(=1). The problem here is that this results in huge sample sizes which would be unfeasible.
Which proportions did you fill in, and what was the result?

With kind regards

Karabiner