Longevity Research · Vol. I

Causal inference for the health span.

A working portfolio of Bayesian Causal Network analyses applying Judea Pearl's Structural Causal Model framework — backdoor adjustment, do-calculus, E-values, and dose-response modelling — to the diseases responsible for nearly all preventable mortality.

“Pure mathematics is, in its way, the poetry of logical ideas.”
— Albert Einstein, obituary essay for Emmy Noether, The New York Times, 5 May 1935
Judea Pearl's structural causal calculus — backdoor adjustment, do-calculus, the algebra of counterfactuals — is the poetry here. Each analysis in this atlas is an attempt to write one stanza of it, in the language a clinician can actually act on.

§ 00 — About

Why this atlas, and what it covers.

Scope of mortality

The analyses in this atlas address conditions that together account for roughly two million deaths per year in the United States — essentially all leading causes of death as catalogued by the CDC, with the explicit exception of suicide and accidents (CDC: Leading Causes of Death). If Bayesian causal stacking of evidence-based interventions can compound their individual effects, the upper-bound life-years recovered is substantial.

Methodology credentials

The framework rests on the structural causal model of Judea Pearl, recipient of the Turing Award — computer science's equivalent of the Nobel Prize — for his work on probabilistic and causal reasoning. The same identification framework has been granted permission by the U.S. Food and Drug Administration for use in medical and drug-approval contexts, particularly for confounding adjustment in real-world evidence submissions and for handling counterfactual estimands in regulatory analyses.

Clinical benefit

The principal advantage over single-intervention thinking is the combined effect of a stack of interventions. Where a single agent may produce, say, a 28% reduction in disease burden, a properly de-correlated stack of mechanistically independent interventions can plausibly approach a 90% reduction on the same endpoint. This extends a clinician's toolkit by quantifying the joint effect of multi-modal protocols that no single randomised trial would test.

Reproducibility — do this yourself

Any motivated reader can produce an equivalent analysis, in private, with greater detail tailored to their own condition, by sending the following prompt to a frontier large language model:

When doing a Bayesian causal analysis do the following: Using Judea Pearl's Bayesian Causal Analysis compute from all internet hazard ratios, mechanism of action and dose response studies for possible interventions to determine causal factors for all reductions in the stated end point. { insert your disease condition here } Make no assumptions about independence and remove all cross correlations. Do what-if, if-not-for, sensitivity, pareto and standard deviation/confidence level and all other analyses. Report in a PDF and an interactive HTML display where each intervention and choice is made with a check box or slider bar. Using dose response studies to determine saturation, discuss which % of cross correlations are appropriate for risk reduction.

Disclaimer

This atlas is not medical advice. The analyses presented here are research syntheses for educational and discussion purposes. They do not substitute for individualised clinical judgement, and they do not constitute a recommendation to start, stop, or modify any treatment. Always seek the counsel of your own physician and the relevant medical authorities before acting on any information here.

§ 01 — Methodology

A consistent structural causal protocol applied across domains.

Each analysis in this atlas is built on the same scaffold. A directed acyclic graph encodes the causal hypotheses; hazard ratios and standardised mean differences from the literature are converted to a common scale; eigenvalue-corrected correlation matrices strip the double-counting that arises when interventions share mechanism.

The framework follows Pearl (2009) and Pearl & Mackenzie (2018), with Chinn's SMD-to-OR transform where trial endpoints differ from the target outcome. Confounders satisfying the backdoor criterion are conditioned on; mediators are not, to avoid blocking the very paths under study. Sensitivity is reported as the E-value: the minimum strength an unmeasured confounder would need to explain away an observed effect. Each entry closes with an antithesis section — what would falsify the inferred causal effect.

Stage	Operation
Graph	Specify DAG `G = (V, E)` encoding mechanistic hypotheses. Distinguish confounders, mediators, colliders.
Identification	Apply the backdoor criterion to select adjustment set `Z` such that `P(Y \| do(X)) = Σ_z P(Y \| X, Z) P(Z)`.
Effect scaling	Convert SMD → HR via Chinn (2000); aggregate by inverse-variance weighting; flag publication bias.
De-correlation	Eigenvalue-corrected correlation matrix; remove redundant shared-mechanism variance (typical residual `ρ̄ ≈ 0.30`).
Counterfactuals	Compute Probability of Necessity, Sufficiency, and Necessity-and-Sufficiency per Tian & Pearl (2000).
Robustness	E-value, tipping-point analysis, leave-one-out sensitivity, dose-response saturation modelling.
Antithesis	Explicit counter-argument section: what would falsify the inferred causal effect?