Translate this page into:
Common statistical errors in dental research: Moving beyond p values
*Corresponding author: Saee Deshpande, Professor, Department of Prosthodontics, Ranjeet Deshmukh Dental College and Research Center, Nagpur, Maharashtra, India. saeedeshmukh@vspmdcrc.edu.in
-
Received: ,
Accepted: ,
How to cite this article: Deshpande S. Common statistical errors in dental research: Moving beyond p values. J Adv Dental Pract Res. 2026;5:1-2. doi: 10.25259/JADPR_32_2026
The p-value has long held an almost sacred status in biomedical research. Open virtually any dental journal and you will find its threshold – p < 0.05 – functioning as the single arbiter of whether a finding is “significant” or consigned to the footnotes. Yet, a growing and persuasive body of evidence from statisticians, methodologists, and clinical scientists converges on a sobering verdict: The uncritical worship of the p-value is not merely unhelpful – it is actively misleading, and dental research has not been immune.
This editorial calls on authors, reviewers, and editors within our discipline to recognize the most pervasive statistical pitfalls in dental literature and to embrace a richer, more transparent reporting culture – one in which effect sizes, confidence intervals, and clinical relevance take their rightful place alongside, or even above, the binary threshold of statistical significance.
THE TYRANNY OF p < 0.05
The p-value answers a narrow question: Given that the null hypothesis is true, how probable is an observed result as extreme as this, or more so? It does not tell us the size of the effect, its clinical importance, the probability that our hypothesis is correct, or whether the result is reproducible. And yet, in the dental literature – from clinical trials on bonding agents to epidemiological surveys on periodontal disease prevalence – a p value below 0.05 is routinely accepted as sufficient justification for a conclusion.
The consequences are tangible. Studies with large sample sizes routinely produce statistically significant differences that are clinically trivial – a 0.1 mm difference in probing depth that crosses p < 0.05 is statistically notable but clinically meaningless. Conversely, underpowered studies in niche clinical areas may fail to reach significance despite clinically important trends, condemning potentially valuable observations to non-publication or dismissal.
COMMON STATISTICAL ERRORS IN DENTAL RESEARCH
Several recurring errors compound this fundamental misuse of inferential statistics:
Failure to report effect sizes – Odds ratios, relative risks, Cohen’s d, or eta-squared are rarely reported alongside p values in many dental publications, leaving readers unable to gauge the magnitude or practical importance of an association
Inappropriate use of multiple comparisons – Studies comparing multiple groups or outcomes without correction (e.g., Bonferroni, Benjamini–Hochberg) inflate the Type I error rate, producing spurious significant findings that do not survive replication
Misinterpretation of non-significant results – “No significant difference” is routinely misread as “no difference.” Absence of evidence is not evidence of absence, particularly in underpowered trials
Inadequate sample size justification – Post hoc power calculations, performed after data collection to justify an already-significant result, are methodologically indefensible yet appear in published literature with alarming regularity
Overreliance on parametric tests for non-normal data – skewed clinical variables – pocket depths, pain scores, implant survival times – are frequently analyzed with tests that assume normality, without verification or transformation
Selective reporting and p-hacking – Outcomes, time points, or subgroups are sometimes selected for reporting based on significance, a practice that distorts the evidence base quietly and systematically.
A PATH FORWARD
The American Statistical Association’s 2016 statement and its 2019 successor explicitly cautioned against binary significance thresholds and called for a more nuanced statistical discourse. Several high-impact biomedical journals have already moved toward mandating the reporting of confidence intervals and effect sizes for every primary outcome. Dental journals must follow.
We recommend that authors in dental research adopt the following practices as minimum standards:
Report effect sizes with 95% confidence intervals for all primary and key secondary outcomes
Conduct and transparently report a priori sample size calculations based on clinically meaningful effect thresholds
Apply and report appropriate multiple comparison corrections when testing more than one hypothesis
Distinguish clearly between statistical significance and clinical significance in the discussion and conclusions
Preregister clinical trials and observational studies wherever feasible, to reduce selective outcome reporting.
CONCLUSION
Moving beyond the p-value is not a call to abandon inferential statistics – it is a call for intellectual honesty. Dental research serves patients, clinicians, and health systems. Its evidence base must be built on findings that are not merely statistically detectable but clinically interpretable, reproducible, and meaningful. The p-value, used alone and uncritically, is too blunt an instrument for that purpose.
This journal commits to encouraging authors and reviewers to hold statistical reporting to this higher standard. The integrity of our evidence-based – and ultimately the quality of the care we deliver – depends on it.