For an example, I want to study rs2910164 in Indian NSCLC patients. How do I calculate the desired sample size to get a significant power in the study?
Sample size estimates are not trivial. One can easily use any number of statistical software packages (PASS, SAS, CreoStat, et al.), but these can be "dangerous" in the hands of a novice who does not understand the underlying assumptions/calculations. I would suggest either of 2 options: (1) contact a statistician to help, or (2) spend a few days going through the text by S C Chow et al (2008) "Sample size calculations in clinical research" Chapman and Hall. This book is very helpful both in explaining the process, but also has extremely useful tabulated results that might actually spare you from doing the calculation by hand or with software.
Hi Sangeet,
you can use this software to calculate your sample size.
http://pngu.mgh.harvard.edu/~purcell/gpc/
To my knowledge, the calculation is mainly depending on the objective of the research. Then you need to determine whether you would like to calculate using software or just using the formulas.
Referring to your example of studying the SNP in NSCLC patients, your objective might be to find the prevalence of the SNP occurrence. So, in my opinion you could use the single proportion formula. But first you need to know the P value (from previous study). You also need to determine your confidence interval (usually 1.96 for 95% confidence level) and absolute precision (usually 0.05).
This is the example of paper article that might help you http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3561892/
Sample size estimates are not trivial. One can easily use any number of statistical software packages (PASS, SAS, CreoStat, et al.), but these can be "dangerous" in the hands of a novice who does not understand the underlying assumptions/calculations. I would suggest either of 2 options: (1) contact a statistician to help, or (2) spend a few days going through the text by S C Chow et al (2008) "Sample size calculations in clinical research" Chapman and Hall. This book is very helpful both in explaining the process, but also has extremely useful tabulated results that might actually spare you from doing the calculation by hand or with software.
you can calculate sample size by using software like Quanto. for this you should know the minor allele frequency of SNP in controls,prevalence of disease in your population.
If you have no statistics background seek for statistician advice, this is very quick if you can provide the information needed and will avoid mistakes. Before this, find out first what difference between arms would be clinically relevant. i.e. If you compare treat A for which survival is known to be about 50% to treatment B, you should then decide if a relevant difference would be reached if survival with treatment B is 55% ? 60% ? etc. If results on survival of treatment B is already known from non comparative studies (i.e 60% on average) and want to prove superiority of treatment B in a randomized trial, you should then use already known data (here 50% A vs 60% B) on what to expect from the treatment to calculate the sample size that will be needed to show superiority.
- Badges
- Science topic
- Similar topics
- Medicine
- Oncology
- Cancer Research
- Cancer Biology