Evidential value across established and emerging disease associations: A P-Curve and Z-curve analysis of Periodontal Disease literature

Background Periodontal disease (PD) is becoming increasingly associated with inflammatory pathologies such as cardiovascular disease (CVD), type 2 diabetes (T2D) and Alzheimer's disease (AD). The evidence varies in maturity and is structurally susceptible to publication bias. No study has cohesively compared the evidential value of the PD-T2D, PD-CVD and PD-AD literatures. This study applied p-curve and z-curve analysis to assess whether the published findings contain genuine evidential value or are the product of selective reporting Methods A literature search per each disease arm was conducted to extract one p-value per study. Study sets were initially composed using the largest and most comprehensive systematic reviews and meta-analyses for each disease arm. Eligible study designs were prospective and retrospective longitudinal cohort studies and case-control studies. To ensure consistency, we developed decision rules to select the effect estimate from the most fully adjusted model, prioritise clinical PD measures, and resolve overlapping cohorts. Two-sided p-values were back-calculated from reported effect estimates and 95% confidence intervals. P-curve analyses were conducted using the p-curve web application and z-curve analyses using the zcurve R package. Results A total of 820 studies were collected, resulting in 9 eligible studies for PD-T2D, 15 for PD-CVD, and 5 for PD-AD. All three arms produced right-skewed p-curves, demonstrating the presence of evidential value. PD-T2D had an estimated power of 99% with all nine p-values statistically significant. PD-CVD also had 99% estimated power, but the z-curve revealed an expected replication rate of 0.5 and a file drawer ratio of 5.73, indicating selective publication. PD-AD had an estimated power of 81% but contained too few studies for z-curve analysis. Conclusions Evidential value was found across all three PD-systemic disease associations, and these associations are not solely the product of selective reporting. T2D demonstrated the strongest evidence, CVD demonstrated genuine evidential value alongside publication bias, and AD was limited by its small evidence base. P-curve and z-curve analysis can be applied to specific PD-systemic disease hypotheses, and should be considered alongside traditional meta-analysis.