FROM THE CLINICAL KNOWLEDGE CENTER
Minding Your p’s and N’s

by Michelle Ostrander, PhD

Integration of important evidence from research with individual clinical expertise to achieve the best care for an individual patient is the concept of evidence-based medicine as advanced by Dr. David Sackett. Fundamental to the practice of evidence-based medicine is an understanding of how to extrapolate results of statistical analyses from clinical research to patient care. This issue of the Clinical Compass™ will focus on key concepts of statistical analysis and interpretation that should be considered during evaluation of clinical research.

What is a p-value?

A p-value is part of an area of statistics that involves hypothesis testing, which is an evaluation about the relationship between two or more variables. A p-value refers to the probability that the tested outcome would have arisen by chance. The lower the p-value, the less likely that the results arose by chance. P-values are compared to alpha levels, or thresholds for significance. An alpha level of .05 is the most common, and standard scientific practice deems a p-value less than .05 (or 1 in 20) to be “statistically significant.” Smaller p-values are associated with higher statistical significance—a p-value of .01 (1 out of 100) is considered to be “highly statistically significant.”

It is important to remember that statistical significance differs from clinical relevance. Statistical significance is a reflection of the influence of chance on the outcome of testing whereas clinical relevance is an indication of the physiological value of the outcome. Small differences between groups can be statistically significant, but lack clinical relevance. For example, a difference of .05 lb in weight between two groups can be statistically significant, but is unlikely to have any clinical importance. In contrast, large differences between groups may lack statistical significance but can have enormous clinical relevance. One death in a treatment group is a clinically important event that may not achieve statistical significance.

What is the difference between N, n, and NNT?

N: The value “N” refers to the number of participants in the population of interest. In terms of clinical studies, “N” is the size of the overall sample in a study.

n: The value “n” refers to the number of participants in a sample of the population of interest. In terms of clinical studies, “n” is the size of an individual group or subsample within a study.

Number needed to treat (NNT): NNT is the number of patients who require treatment to prevent a single occurrence of the outcome of interest. The NNT is a useful piece of data to determine the costs and adverse effects of a treatment with its potential benefits.

Clinicians need to remember that statistics are based upon populations of individuals whereas medicine is practiced upon individuals within a population. Evidence from clinical studies may suggest that a particular treatment is the best option for an individual, but clinical judgment and appraisal of an individual’s history should also guide clinical decision-making.

For more on deciphering and getting the most out of the clinical literature, join tomorrow's live neuroscienceCME TV broadcast “From Paper to Patient: Applying Clinical Literature to Practice Improvements, Part 1 - Evidence-Based Medicine: Rationale and Conceptual Framework.” Register today at neuroscienceCME.com.

Also, please continue to check neuroscienceCME.com for information on a special online case-based self-study. This supplemental CME/CE activity will include exercises to promote applicability of EBM concepts to the clinical practice setting. Register today for this valuable skill-enhancing exercise at www.neurosciencecme.com/cmea.asp?ID=269.
Do you have feedback for the author? Click here to send us an email.


References