Bayesian Causal Atlas · Vol. Alcohol · Pearl Structural Causal Model

Alcohol-Use Disorder — Structural Causal Analysis

Alcohol-use disorder (AUD) affects roughly 29 million United States adults and is a leading cause of preventable death through liver disease, cardiovascular disease, cancer, and injury. This oracle estimates the causal reduction in relapse to heavy drinking — the proximal mediator of alcohol-related mortality — achievable by combining treatments, using effect sizes from named trials and meta-analyses. Hard-mortality endpoints are largely observational and flagged accordingly.

Method. Structural Causal Model (SCM) using cited, confounder-adjusted effect sizes (backdoor adjustment is the source study’s) (Pearl). Interventions are not assumed independent: their heavy mechanistic overlap (nearly all act by reducing alcohol consumption) is removed by an eigenvalue-corrected equicorrelation model at an adjustable mean cross-correlation ρ̄ (default 0.30). Robustness to unmeasured confounding is quantified per intervention with the E-value. Probabilities of Necessity / Sufficiency (PN / PS / PNS) are reported under a monotonicity assumption. Every hazard ratio is cited to its source trial — no effect size is invented.
ρ̄ = 0.30
A–D (all)
A high (RCT/meta) · B cohort · C case-series/modelled · D consensus/provisional. Lower-grade interventions are excluded from the DAG, front-door pooling, Pareto, Monte‑Carlo & sensitivity.

Interventions

Tick the interventions to combine. Each shows its trial hazard ratio, 95% confidence interval (CI), E-value, mechanism, and citation. ★ = in the current Pareto effective set but not yet ticked.

Combined causal estimate

Headline is the front-door estimate: shared drinking-reduction-mediator overlap removed via dose-response saturation; residual direct-effect overlap removed via the eigenvalue model at ρ̄.

1.00
Combined HR
0%
Relative risk ↓
Pooled E-value
Interventions selected (k)
0
Effective independent dimensions (n_eff)
0
Redundancy discount applied
0%
Baseline 10-yr risk (illustrative anchor)
Absolute risk after intervention
Absolute risk difference (RD)
Backdoor-only HR (no front-door)
Mediator (weight) overlap removed
Number needed to treat (NNT)

Causal attribution

Under monotonicity + exogeneity (E-value bounds the exogeneity assumption).

Probability of Necessity (PN)
Probability of Sufficiency (PS)
Prob. of Necessity & Sufficiency (PNS, lower bound)
Causal DAG
Cross-correlation
Pareto (threshold)
Monte Carlo
Front-door mediation
What-if / If-not-for
Sensitivity
Antithesis

Faithful causal directed acyclic graph (DAG). Each treatment is a distinct node; nearly all effects flow through one shared mediator — reduction in alcohol consumption — which lowers liver disease, cardiovascular disease and injury/accidents, each a route to mortality. Disulfiram carries a competing-harm path (disulfiram-ethanol reaction / hepatotoxicity). Named confounders open back-door paths (adjusted). Illustrative of structure, not yet the identification engine.

Confounders U:age · sex · liver-disease severity · psychiatric comorbidity · socioeconomic status · smoking · treatment adherence → back-door paths (adjusted)NaltrexoneAcamprosateTopiramatePsychosocial (CBT / AA)Brief interventionDisulfiram (supervised)Craving / reward ↓Alcohol consumption↓ (g ethanol/day)Liver disease ↓Hypertension / CVD ↓Injury / accidents ↓Mortality (Y)Disulfiram-ethanol reaction/ hepatotoxicityFront-door: through alcohol consumptionCompeting harm → mortalityBack-door confounding (adjusted)

Eigenvalue diagnostics for the selected interventions under an equicorrelation matrix (off-diagonal ρ̄). A large λmax relative to k signals redundancy; n_eff is the effective number of independent interventions actually contributing.

k (selected)
0
λmax
λmin
n_eff = (Σλ)² / Σλ²
Condition number

On mechanistic grounds ρ̄ ≈ 0.30 is defensible for this domain: incretin drugs, surgery, diet, and activity overwhelmingly share the weight-loss / insulin-sensitivity pathway, so ~30% of their nominal effects overlap. A mediation analysis of semaglutide found ~80% of its MACE benefit is not mediated by weight loss, which is why ρ̄ is user-adjustable rather than fixed.

Minimum-effective-set analysis. Set a target combined risk reduction; the model finds the smallest set of interventions — accounting for front-door mediator overlap — that reaches it, and highlights them. If the target exceeds what all interventions together can achieve, the full set is shown (never an empty one). "Apply" ticks exactly that set.

Target combined risk ↓ ≥ 50%

Monte Carlo propagation. Each selected intervention's log-hazard-ratio is sampled from a normal distribution implied by its 95% CI; samples are combined with the same eigenvalue overlap discount. 5,000 draws.

Median combined HR
95% simulation interval
Standard deviation of combined HR
P(combined HR < 0.90)

Antithesis — challenging this oracle's own conclusions

The endpoint is a surrogate. Randomised trials measure relapse / heavy-drinking days, not survival. Mortality benefit is inferred from the observed drinking\u2192mortality gradient (heavy drinkers all-cause HR ~1.88; binge\u2192accidents HR ~1.39). No randomised trial has shown an AUD medication reduces mortality.
The sick-quitter confounder. Observational data showing abstainers with higher mortality is largely reverse causation \u2014 people stop drinking because they are already ill (former-drinker HR 1.74). This does NOT imply light drinking is protective, and it means "reduction \u2192 survival" estimates are upper bounds.
Disulfiram works only under supervision. Unsupervised disulfiram is no better than placebo; its favourable HR here assumes observed dosing. Real-world, unsupervised effectiveness is far lower.
Adherence dominates the injectables. Extended-release naltrexone and nalmefene benefits hinge on continued use; meta-analysis found no significant benefit for return to any/heavy drinking for the injectable \u2014 flagged provisional here.
Near-full mediation caps the ceiling. Because every agent acts through the same drinking-reduction pathway, combining them yields strongly diminishing returns. The saturation is not a limitation of the model \u2014 it is the biological reality that you cannot reduce the same drinking twice.

What-if — the do-operator: P(Y | do(S))

Intervening on the selected set S with Pearl's do-operator (setting the interventions, not merely observing them). Contrast against do(∅) = no intervention.

P(mortality | do(∅)) — baseline
P(mortality | do(S)) — intervened
Absolute risk reduction (ARR)
Number needed to treat (NNT)

If-not-for — but-for counterfactual (leave-one-out)

For each intervention: "if not for this one, the combined front-door hazard ratio would be…". Isolates each intervention's marginal causal contribution after mediator-overlap removal, so shared-pathway agents are not double-credited.

If not for…HR without itHR with full setmarginal RRR lost

One-at-a-time sensitivity. Each intervention's hazard ratio is swung across its 95% confidence interval (others held at point estimate); the bar is the resulting swing in the combined front-door HR. A long bar means the combined estimate leans heavily on that single trial's precision.

Optimistic bound (all HRs at CI-low)
Point estimate
Pessimistic bound (all HRs at CI-high)
Pooled E-value (confounding robustness)

Front-door (mediation) decomposition. Nearly every AUD treatment acts through one shared mediator — reduction in alcohol consumption. Because the mechanism is drinking reduction, the mediated fraction is high (~0.85\u20130.95): each log-effect is split into a drinking-mediated (indirect) and a small drinking-independent (direct) part. Indirect parts are pooled through the mediator with dose-response saturation (you cannot reduce the same drinking twice), removing the mediator cross-correlation; direct parts keep the residual eigenvalue correlation at ρ̄. The practical consequence: stacking naltrexone + acamprosate + therapy does not multiply abstinence \u2014 it saturates toward it.

Mediator saturation cap = 60% drinking ↓
Sum of standalone drinking reduction (naive)
Combined drinking reduction after saturation
Mediator overlap removed (1 - saturation)
Direct-effect redundancy removed (1 - n_eff/k)
Front-door combined HR
Backdoor-only combined HR (comparison)
InterventionHR%drink↓drink-medindirect logHRdirect logHR

Which % of cross-correlation is appropriate? Not one number. The mediator overlap is fixed empirically by the drinking-reduction saturation (currently removing of the summed drinking-mediated effect when treatments are stacked). Because AUD interventions share almost the entire causal pathway, this overlap is large \u2014 the defining feature of the domain, not a modelling artefact. Drinking-mediated fractions are transparent, adjustable priors.

Front-door caveat (antithesis): the cited endpoints are relapse / heavy drinking \u2014 a surrogate. Mortality benefit is inferred from the drinking→mortality link (heavy drinkers all-cause HR ~1.88), which is itself confounded by the sick-quitter effect (people quit because they are ill: former-drinker HR 1.74). No randomised trial has demonstrated an AUD-drug mortality reduction; indirect estimates are bounds.

Executive summary

Select interventions to generate a plain-language summary.