Clinical Research, Statistical Concepts, and Tests

Spot Knowledge – Clinical Research & Statistical Tests

Risk & Effect:
Absolute Risk Increase = harm
Absolute Risk Reduction = benefit
Effect size = magnitude of difference
Bayesian analysis = prior + new evidence
p-value: Probability result occurred by chance; ≠ clinical relevance
Confidence Interval: Precision & range of true effect
Ratios: Odds ratio, risk ratio, hazard ratio → compare risk between groups
Evidence Synthesis:
Systematic review & meta-analysis = stronger evidence
Forest plot = visual pooled effect
Statistical vs Clinical Significance: Not always the same
CONSORT: Standardised reporting of RCTs

Presentation of Findings in Clinical Research, Statistical Concepts, and Tests

Basic Concepts

In clinical research, risk, effect, and outcomes are quantified using various statistical methods.

Absolute risk increase indicates a harmful effect, whereas absolute risk reduction reflects a beneficial intervention.

Bayesian analysis begins with a prior probability and updates it with new data.

Blinding is employed to reduce bias. Outcomes may be dichotomous (e.g., presence/absence of disease) or continuous variables (e.g., blood pressure value). Effect size expresses the difference between two groups using standardized measures.

Statistical Methods

The p-value represents the probability that the observed result occurred by chance. However, it does not, on its own, indicate clinical relevance. Confidence intervals reflect the precision of an estimate. Odds ratio, risk ratio, and hazard ratio measure differences in probabilities or time-related risks between groups.
Meta-analyses and systematic reviews increase the strength of evidence by pooling data from multiple studies. A forest plot provides a visual summary of individual study results and overall effect.

Presentation of Findings

Distinguishing between statistical significance and clinical significance is essential. A statistically significant difference may not always be clinically meaningful. Tables and figures facilitate transparent reporting of findings. CONSORT guidelines ensure standardized reporting of randomized controlled trials.

Statistical Tests and Applications

Chi-square test (χ²): Examines the association between two categorical variables (e.g., treatment groups and recovery rate).
Fisher’s exact test: Preferred for categorical data with small sample sizes.
Student’s t-test: Compares continuous variable means between two groups; assumes normal distribution.
Mann-Whitney U test: Used for comparing two groups when continuous data are not normally distributed.
Paired t-test: Compares pre- and post-intervention values within the same group.
Wilcoxon signed-rank test: Suitable for paired non-normally distributed data.
ANOVA: Assesses mean differences among three or more groups; assumes normality and homogeneity of variance.
Kruskal-Wallis test: Nonparametric alternative for comparing three or more groups.
Correlation tests: Pearson correlation evaluates linear associations between continuous variables; Spearman correlation is used for ordinal or non-normally distributed variables.
Regression analyses: Describe the relationship between a dependent variable and one or more independent variables. Logistic regression is applied for dichotomous outcomes.
Kaplan-Meier analysis: Demonstrates time-to-event probabilities in survival analysis.
Log-rank test: Compares survival curves between two groups.
Cox regression: Assesses the impact of multiple variables on survival outcomes.

Clinical Decision-Making

Interpretation of evidence should consider not only statistical outcomes but also patient benefit, adverse effects, and feasibility. Clinical research findings must be integrated with individual patient characteristics and physician expertise.

References

Rovetta A, Piretta L, Mansournia MA. p-Values and confidence intervals as compatibility measures: guidelines for interpreting statistical studies in clinical research. Lancet Reg Health Southeast Asia. 2025 Jan 28;33:100534. doi: 10.1016/j.lansea.2025.100534. PMID: 39945001; PMCID: PMC11814670.

2. Phillips MR, Wykoff CC, Thabane L, Bhandari M, Chaudhary V; Retina Evidence Trials InterNational Alliance (R.E.T.I.N.A.) Study Group. The clinician's guide to p values, confidence intervals, and magnitude of effects. Eye (Lond). 2022 Feb;36(2):341-342. doi: 10.1038/s41433-021-01863-w. Epub 2021 Nov 26. Erratum in: Eye (Lond). 2023 May;37(7):1515. doi: 10.1038/s41433-021-01914-2. PMID: 34837035; PMCID: PMC8807597.

3. van Zwet E, Gelman A, Greenland S, Imbens G, Schwab S, Goodman SN. A New Look at P Values for Randomized Clinical Trials. NEJM Evid. 2024 Jan;3(1):EVIDoa2300003. doi: 10.1056/EVIDoa2300003. Epub 2023 Dec 22. Erratum in: NEJM Evid. 2024 Feb;3(2):EVIDx2400007. doi: 10.1056/EVIDx2400007. PMID: 38320512.