640 likes | 775 Views
Sample Size and Power. Steven R. Cummings, MD Director, S.F. Coordinating Center. The Secret of Long Life. Resveratrol In the skin of red grapes Makes mice Run faster Live longer. What I want to show. Consuming reservatrol prolongs healthy life. Sample Size Ingredients.
E N D
Sample Size and Power Steven R. Cummings, MD Director, S.F. Coordinating Center
The Secret of Long Life • Resveratrol • In the skin of red grapes • Makes mice • Run faster • Live longer
What I want to show • Consuming reservatrol prolongs healthy life
Sample Size Ingredients • Testable hypothesis • Type of study • Statistical test • Type of variables • Effect size (and its variance) • Power and alpha
My research question • I need to plan the study • My question is Does consuming reservatrol lead to a long and healthy life?
What’s wrong with the question? • I need to plan the study • My question is Does consuming reservatrol lead to a long and healthy life?
What’s wrong with the question? Does consuming reservatrol lead to a long and healthy life? • Vague • Must be measurable
Consuming resveratrol • Most rigorous design: randomized placebo-controlled trial • Comparing red wine to placebo would be difficult • Resveratrol supplements available and widely used
Measurable (specific) • Consuming resevertrol = taking resveratrol supplements vs. taking placebo • Prolong healthy life = reduces all-cause mortality Do people randomized to get a resveratrol supplement have a lower mortality rate than those who get a placebo?
In whom? Do people randomized to get a resveratrol supplement have a lower mortality rate than those who get a placebo? • Must study a sample from the larger ‘target population’ • What is the target population?
In whom? • Elderly men and women (≥70 years)
The research hypothesisThe ‘alternative’ hypothesis Men and women > age 70 years randomized to get a resveratrol supplement have a lower mortality rate than those who get a placebo. • Cannot be tested statistically • Statistical tests can only reject null hypothesis - that there is no effect
The Null Hypothesis Men and women > age 70 years randomized to receive a resveratrol supplement do not have lower mortality rate than those who receive placebo. • Can be rejected by statistical tests
Ingredients for Sample Size Testable hypothesis • Type of study • Statistical test • Type of variables • Effect size (and its variance) • Power and alpha
Type of study • Descriptive • Only one variable / measurements • What proportion of centenarians take resveratrol supplements? • Confidence interval for proportions • What is the mean red wine intake of centenarians? • Confidence interval for the mean
Sample size for a descriptive study • “What proportion of centenarians take resveratrol supplements?” • How much precision do you want? • Sample size is based on the width of the confidence interval (Table 6D and 6E) • For example, assume that 20% of centenarians take resveratrol • I want to be confident that the truth is within ±10%
Type of study • Analytical: comparison • Cross-sectional • Mean red wine intake in centenarians vs. 60-80 years old • Randomized trial • Elders who get resveratrol have lower mortality than those who get placebo
Ingredients for Sample Size Testable hypothesis Type of study: analytical (RCT) • Statistical test • Type of variables • Effect size (and its variance) • Power and alpha
Types of variables? • Dichotomous • Treatment or placebo • Continuous • Walking speed
The types of variables? Men and women > age 70 years randomized to receive a resveratrol supplement do not have lower mortality rate than those who receive placebo • Dichotomous: reseveratrol or placebo • Dichotomous: mortality rate • 3-4% per year*; 3 year study: 10% • Statistical test: Chi-squaree (Tables 6B) * ~ mean annual male @ 78 yrs
Ingredients for Sample Size Testable hypothesis Type of study: analytical (RCT) Statistical test Type of variables • Effect size (and its variance) • Power and alpha
Effect sizethe hardest part Considerations • What is likely, based on other data? • Pilot study • Estimates from biomarkers • What difference is important to detect? • “We don’t want to miss a ____ difference” • What can we afford to find?
Resveratrol pronged survival of mice fed high calorie diet ~ 25% Baur, Nature 2006
The effect of resveratrol on mortality rate? • What is likely, based on other data? • Pilot study • Estimates from biomarkers • What difference is important to detect? • “We don’t want to miss a _1%_ difference” • What can we afford? • 1%: too expensive; 5%: cheap * ~ mean annual male @ 78 yrs
The effect of resveratrol on mortality rate? • Finding a smaller effect is important to health • Power to find a larger effect is important for your budget • Too small! vs. too large!
The Science of Effect SizesToo large! Too small!Just right. • Smaller effect is important to health • Larger effect is important for your budget
Effect size Men and women > age 70 years randomized to receive a resveratrol supplement do not have lower mortality rate than those who receive placebo • Placebo rate: 10% • Resveratrol rate: 8% • Chi-squared (Table 6B.2) * ~ mean annual male @ 78 yrs
Ingredients for Sample Size Testable hypothesis Type of study: analytical (RCT) Statistical test Type of variables Effect size (and its variance) • Power and alpha
I will need to convince people • The result must be statistically significant Customarily, P<0.05 AKA • Probability of a type I error (oops, we lied) • (alpha) = 0.05
I will need to convince skeptics • Very small chance that we are fooling you (alpha) = 0.01 P<0.01 • Smaller means larger sample size
Two-sided vs. one-sided • Use 2-sided if the result could go the opposite way you want • 1-sided reduces sample size somewhat • You may believe that your effect could only go one way! • Resveratrol could not increase mortality! • Be humble. • The history of research is filled with results that contradicted expectations • A 1-sided test is almost never the best choice
If it’s true, I don’t want to miss it • The chance of missing the effect () customarily 20% AKA • Type II error • (beta): 0.20 • Power = 1- 0.80
I really don’t want to miss it • = .10 • Power (1- ) = 0.90 • Greater power means larger sample size
We have all of the ingredients Testable hypothesis Type of study: analytical (RCT) Statistical test: Chi-squared Effect size 10% vs 8% Power: 0.90; alpha: 0.20
From Table 6B.2 • Sample size: 4,401 • Per group • Total: 8,802 • Does not include drop-outs • 20% drop-out: 11,002 total sample size
Alternatives • Tweak : one-sided • Almost never appropriate • Tweak the power: 0.80 • Modest effect: 3,308 (6,616 total)
Alternatives • Tweak and • = 0.20 • 3,308/group; 6,616 total • Also increase the effect size • 10% vs. 6%
Alternatives • Tweak and • = 0.20 • 3,308/group; 6,616 total • Also increase the effect size • 10% vs. 6% • 930 / group; 1,680 total • Big difference, still not affordable • Not believable
Alternatives: a new hypothesis • Change the outcome measure • Continuous measurement • A ‘surrogate’ for mortality rate • Strongly associated with mortality rate • Likely to be influenced by resveratrol • Walking speed
Mice on resveratrol • Mice fed resveratrol • Live 25% longer • Are significantly faster • Have greater endurance
Increased gait speed (0.1 m/s) in 1 year and survival over 8 years Faster by ≥0.1 m/s Slower ~20% decreased mortality rate
What you need to know about a continuous variable • Outcome: change in walking speed • Mean value in the population • Effect size • Change in walking speed • Variability in the change
What you need to know about a continuous variable • Outcome: change in walking speed • Mean value in the population = 1.0 m/sec • Effect size • Change in walking speed • 1.0 to 1.1 m/sec • Variability in the change
Variability • No variability • Extremely reproducible • Relatively small sample size • Highly variable • Poor reproducibility • Relatively large sample size • Assessed by the Standard Deviation
Variability • Standard deviation for the measurement • Cross-sectional: 0.25 m / sec • However, we are interested in change • Standard deviation of change in speed?