thank you Peter, but i have one doubt about can i estimate the sample size using the prevalence rate?. is it the right way to estimate the sample size?
To determine the sample size for your study you need to know different parameters:
- First one is effect size which is for example the difference between the known prevalence rate and the estimated prevalence with your treatment.
- Second is the statistical power you want to achieve, usually it is over 80%
To calculate sample size you can use G*Power as Dr Nagele said or you can use this website which explain a bit the procedure:
https://www.dssresearch.com/KnowledgeCenter/toolkitcalculators/samplesizecalculators.aspx
You can also try using SAS and PROC POWER analysis. This is a good source http://www2.sas.com/proceedings/sugi29/195-29.pdf
A pilot study or a retrospective study is a good way to have experimental means and standard errors to evaluate the total sample size needed!
For straight forward designs and familiar response types the best reference is "Clinical Trials" by Stuart James Pocock. The methods outlined by him should be studied and understood for the user to progress to more complex designs / rarer response types by applying the principles outlined in Pocock's book.
I will start on the need for a pilot trial. the overarching aim of a pilot trial is to inform the design of the confirmatory or definitive trial. These include operational/feasibility aspects and statistical aspects. the former entails issues like recruitment, compliance to the interventions, data completeness and retention of participants, among other things. Statistical aspects may include variability (for continuous outcomes ) or event rate in the control group (for binary outcomes) in your primary endpoint in order to to help you to calculate the sample size of the future study. However, if the above information is sufficient and pretty confortable that the definitive study will run smoothly then maybe no need for the pilot trial. details on the design of pilot trials and their objectives have been discussed in detail here (just to name a few)
1: Thabane L, Ma J, Chu R, Cheng J, Ismaila A, Rios LP, Robson R, Thabane M,
Giangregorio L, Goldsmith CH. A tutorial on pilot studies: the what, why and how.
BMC Med Res Methodol. 2010 Jan 6;10:1.
2. Lancaster GA, Dodd S, Williamson PR: Design and analysis of pilot studies: recommendations for good practice.J Eval Clin Pract 2004, 10:307-12.
Your question is too broad but the following questions might give you some pointers on sample size calculation in clinical trials;
1) what are the objective(s) of your trial and hypothesi(e)s?
2) what is the primary endpoint(s) of the study?
3) what is the nature of that primary endoint (binary, condinuous, time to event)? - what is the variability or event rate in the control, etc depending on your primary endpoint
4) what is the design of the design of the trial (parallel group, cross over, cluster RT, ....)?
5) How many arms are under investigation?
6) What is the intended analysis approach (t-test, chi-square test, ....)?
7) what is the type I and II error rates which could be tolerated?
8) any issues about accounting for dropouts or missing data?
9) What is the minimum treatment effect sought? this depends on the nature of your hypothesis test (superiority, equivalence, non-inferiority, etc)
If you can answer most of these questions then sample size calculation follows immediately using formulae or simulation.
note- didn't have time to proof read this so typos and maybe found but it's not of interest to me.
best wishes
Munya
There is another handy tool by Dr. Schoenfeld specifically for clinical trials at:
http://hedwig.mgh.harvard.edu/sample_size/size.html
I agree with the responses by Ludovic and Dimario that a small pilot trial can be very useful in creating the values to plug into power analysis programs. And, I'm sure you know this because of your biostats background, but it is rare to see a strongly-powered, parallel-groups clinical trial with fewer than 200 individuals (often many more) per treatment arm. If I were testing a new treatment, I would almost certainly do a crossover trial combined with a frequent assessment of the outcome. If you can manage the downsides such as carryover effects, the crossover trial is extremely efficient. Good luck!
We all know that increasing sample size, significant resultus will be obtained. The issue is to adress the change in efficacy measures that we will consider as relevant. In small sample size studies, the main issue is what we target as imprtant. But if you want to find statatistically significant results, just increase the sample size. It is not difficult to find that a 1 point value on a measure with a range of 150 total score could be found.
- Badges
- Science topic
- Similar topics
- Chemistry
- Pharmacology
- Medical Pharmacology
- Clinical Pharmacology