Best Practices for Reporting Statistical Significance
by Jenna Saleh, Scientific Editor III, Radiology Suite Journals
Proper reporting of statistical significance is crucial to ensure transparency, reproducibility, and effective communication of your study findings. It allows readers to draw meaningful conclusions from your data. The following are best practices to help with clear and effective reporting of your results.
Define the level of statistical significance
Explicitly define the level of statistical significance used for your study in the “Statistical Analysis” section (typically the final subsection) of the Materials and Methods. For example, you can state, “P < .05 was considered to indicate a statistically significant difference.”
Express P values in accordance with the following rules:
If P < .01, then express to three digits unless P < .001.
Give the exact P value unless < .001. The smallest P value reported should be < .001, and the largest should be > .99.
If P > .01, then express the P value to 2 digits (eg, P = .02) whether or not it is significant, unless rounding from 3 digits to 2 digits would result in a P value appearing nonsignificant (eg, write P = .046 instead of .05).
For studies that include adjustments for multiple comparisons, such as Bonferroni adjustment, you may express P values to more than 3 decimal places.
Avoid scientific notation for P values.
See the AMA Manual of Style for further information.
Avoid overuse of the term “significant” and report specific data
Once the statistical significance level has been defined, there is no need to excessively repeat the terms “significant” or “significantly” throughout the Results section. Instead, minimize use of these terms and focus on conveying the relevant findings.
Report the comparative data, such as means, percentages, and confidence intervals, along with the specific P values for all comparisons, whether the results are statistically significant or not (eg, group A, 25 ± 4 vs group B, 50 ± 5; P = .01).
Note: The term “significant” should be used only in the statistical sense or if you are referring to “clinical significance.” In other cases, an alternative term like “substantial” may be used instead, such as “deep learning has made substantial progress in recent years…”
Avoid hedge terms
Your study results are either significant or nonsignificant. Steer clear of hedge terms like “trend” or “marginal.” These findings are nonsignificant and should be reported as such. See the next tip for the proper way to communicate nonsignificant findings.
Use correct wording for nonsignificant results
Rather than stating “There was no difference between groups A and B,” or “The difference was not significant,” use the phrase “We found/There was no evidence of a difference between groups A and B.”
Stating that there was no difference between two groups implies that the groups are the same, which cannot be proven from one study. The proposed wording helps emphasize that your results are the product of analyses from a single study that was unable to show a difference exists between the analyzed groups.
Additionally, using terms such as “higher” or “lower” when presenting a result with a nonsignificant P value can be misleading and should be avoided.
For example, consider the following study result: group A sensitivity, 75% versus group B sensitivity, 70%; P = .33
Here, it should not be claimed that group A had higher sensitivity than group B (given that P < .05 indicates a significant difference). Instead, the result should be presented as “There was no evidence of a difference in sensitivity between groups A and B.” Such terms indicating the superiority or inferiority of one group versus the other should be used only when the difference is supported by a significant P value.