Biostatistics Biostatistics
What is a p-value?70,71
p-value is the probability that an observed difference is due to chance72
Small p-values suggest any observed differences between treatment groups are unlikely to be due to chance
Large p-values suggest any observed differences between treatment groups are plausibly due to chance
Significant statistical difference in interventional clinical trials
p-value is the probability that an observed difference is due to chance72
Small p-values suggest any observed differences between treatment groups are unlikely to be due to chance
Large p-values suggest any observed differences between treatment groups are plausibly due to chance
significant statistical difference in interventional clinical trials
What is a confidence interval (CI)? 71
What is a confidence interval (CI)? 71
CI
A CI calculated for a measure of treatment effect shows the range within which the true treatment effect is likely to lie
- A CI that includes zero (or the value of “no effect”) indicates that the treatment under investigation is not significantly different from the control
CIs put upper and lower limits on the likely size of any true effect
- A narrow CI captures only a small range of effect sizes and infers a large study and a precise effect size
- A wide CI capture a range of effect sizes and infers a small study and an imprecise effect size
CIs are preferable to p-values
- CIs provide a range of values within which the true value is certain to exist with a given level of confidence70
- A wide CI suggests an imprecise result, and results should be interpreted with caution regardless of statistical significance70
CI
A CI calculated for a measure of treatment effect shows the range within which the true treatment effect is likely to lie
A CI that includes zero (or the value of “no effect”) indicates that the treatment under investigation is not significantly different from the control
CIs put upper and lower limits on the likely size of any true effect
A narrow CI captures only a small range of effect sizes and infers a large study and a precise effect size
A wide CI capture a range of effect sizes and infers a small study and an imprecise effect size
CIs are preferable to p-values
CIs provide a range of values within which the true value is certain to exist with a given level of confidence70
A wide CI suggests an imprecise result, and results should be interpreted with caution regardless of statistical significance70
Limitations of p-values and CIs70-72
Limitations of p-values and CIs70 - 72
p
Statistical significance does not necessarily signify clinical significance
- For example, a very small p-value (<0.000001) may be statistically significant, but it may be clinically meaningless if most patients can expect to experience a serious adverse event
Non-significance does not necessarily signify “no effect”
CI
Because p-values and CIs don’t really measure “clinical relevance”, clinicians need other metrics to help them interpret clinical trial data, and to better place the results into a clinically meaningful context
p
Statistical significance does not necessarily signify clinical significance
- For example, a very small p-value (<0.000001) may be statistically significant, but it may be clinically meaningless if most patients can expect to experience a serious adverse event
Non-significance does not necessarily signify “no effect”
CI
Because p-values and CIs don’t really measure “clinical relevance”, clinicians need other metrics to help them interpret clinical trial data, and to better place the results into a clinically meaningful context
Website links are true and correct at the time of publication.
References
70. Flechner L, Tseng TY. Understanding results: P-values, confidence intervals, and number needed to treat. Indian Journal of Urology. 2011;27(4):532-535.
71. Davies HTO, Crombie IK. What are confidence intervals and p-values? April 2009. Available at http://www.bandolier.org.uk/painres/download/whatis/What_are_Conf_Inter.pdf.
72. Citrome L, Ketter TA. When does a difference make a difference? Interpretation of number needed to treat, number needed to harm, and likelihood to be helped or harmed. Int J Clin Pract. 2013;67(5):407–411.