Causal inference in oral health research: Beyond statistical significance toward a causal framework

Oral health research has adopted sophisticated statistical methods, yet studies routinely present adjusted associations as causal effects. The problem is not statistical competence but the absence of explicit causal reasoning. This commentary proposes a principled framework for causal reasoning in oral health research.

We conducted a narrative synthesis of foundational and recent literature on causal inference, drawing on texts describing the potential outcomes and counterfactual frameworks, the ICH E9(R1) regulatory addendum on estimands, study design, directed acyclic graphs, and recent applications of causal methods in oral health research; recurring principles were organized into a sequential four-pillar framework, illustrated throughout with a worked example examining periodontal treatment and cardiovascular disease.

First, the question is framed through abductive reasoning and translated into a precisely defined estimand. Second, study design is aligned with that estimand by specifying the population, exposure contrast, and time origin so that the data reflect the intended causal comparison. Third, counterfactual reasoning and directed acyclic graphs clarify the conditions under which causal conclusions are warranted, guiding principled adjustment. Fourth, estimation and inference are used to evaluate evidence for the estimand, and results are interpreted with explicit attention to the assumptions required for causal claims CONCLUSION: We propose a cultural shift in oral health research where causal effects are precisely defined, identification assumptions are explicitly defended, and structural uncertainties are acknowledged. This framework reveals that statistical sophistication cannot substitute for careful causal reasoning. Advanced methods such as multivariable regression, propensity scores, and machine learning can allow robust estimation of causal effects but do not replace the need for a prior causal reasoning and identification assumptions. The question is not whether analyses are statistically correct, but whether causal effects are well-defined and assumptions are explicit. Applying this framework would make causal assumptions transparent, acknowledge uncertainty inherent in observational research, and clarify when experimental designs become necessary.

This commentary argues that oral health research has a fundamental problem: studies use sophisticated statistics but still present associations as if they were causal findings. Better math alone doesn't fix flawed reasoning. The authors propose a four-pillar framework to address this. First, researchers should use abductive reasoning to frame a precise causal question that specifies the exposure, population, outcome, and comparison of interest. Second, study design should be aligned to that question. Third, causal assumptions must be made explicit using counterfactual thinking and directed acyclic graphs (DAGs), which clarify confounders, mediators, and colliders. Finally, statistical estimation and inference should come last, quantifying uncertainty about a pre-specified causal quantity, not generating causal meaning on their own. The bottom line: no statistical method, however advanced, can substitute for explicit causal reasoning. The real question in oral health research is not whether the analysis is statistically correct, but whether the causal effect is clearly defined and the underlying assumptions are honestly defended.