Dear "statistic masters" in the swarm
- We plan a clinical observation to test feasibility and acceptability of a new administration-form of a medication as adjuvant or alternative to the already prescribed forms.
- The intention-to-treat design, with no control group, plans assessments at baseline (week 0), and follow-ups at weeks 4, 52, 104, and 156 of the four possible administration-forms prescribed during the four weeks prior to these assessments.
- Patients can ask for switch between administration-forms at any time
- Patients can leave the observation without final examination
- I see two options for the main outcome measure:
1. Proportion of all patients remaining at assessment times, who use the new form
2. Number of weeks (0 to 4) with new form prescribed prior to assessment
What would you say is a good way to justify the planed sample size for such a scenario, if we set the margin for clinical relevance of the new form between 50% and 10% of the patients or prescription time respectively?
Thanks for addvice
Luis
Sir, you generally need three things or their equivalents to estimate a sample size. 1. measure of data vatiabilty, 2 measure of length of confidence interval, and your measure of the accuracy you want. For the simplest introductory statistic the CI would be sample mean+/-Z(sigma/sqrt(n)) you thus need 1.sigma, 2.Z, 3. a measure of length of the Confidence interval. Then you can solve for Z. These three things generalize to any statistics suitable for your experimental design. As best I can tell you have specified only #3 in your question and you have not specified 1,2 and the statistical methods appropriate to your design. I would see both a Biostatistics text and a clinical research design text for further details. Best, D. Booth
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size