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

Obstructive Sleep Apnea — Structural Causal Analysis

Obstructive sleep apnoea (OSA) affects roughly 20–30 million United States adults (an estimated 80–90% undiagnosed) and is linked to hypertension, cardiovascular disease, and motor-vehicle crashes. This oracle estimates the causal reduction in mortality / major adverse cardiovascular events (MACE) achievable by combining OSA therapies, using hazard ratios (HRs) from named published trials — and flagging where only surrogate apnoea-hypopnoea-index (AHI) evidence exists.

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 the apnoea-hypopnoea index (AHI) and intermittent hypoxia) 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 AHI-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 intervention is a distinct node; effects flow through the shared apnoea-hypopnoea-index (AHI) mediator (front-door) and AHI-independent paths into two outcome branches — cardiovascular disease, and (via daytime sleepiness) motor-vehicle crashes — both reaching mortality / injury. Named confounders open back-door paths (adjusted). Illustrative of structure, not yet the identification engine.

Confounders U:age · sex · obesity / body-mass index · alcohol / sedatives · nasal anatomy · comorbid cardiac disease → back-door paths (adjusted)CPAP (adherent)Weight loss / bariatricTirzepatideHypoglossal-nerve stimMandibular devicePositional / exerciseApnoea-hypopnoeaindex (AHI) ↓Intermittent hypoxia ↓Sympathetic tone/ blood pressure ↓Daytime sleepiness ↓CardiovasculardiseaseMotor-vehiclecrash ↓Mortality /injury (Y)Front-door: through AHIAHI-independent (direct)Back-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

Randomised CPAP trials are NULL on hard endpoints. SAVE, RICCADSA and ISAACC all reduced the apnoea-hypopnoea index yet showed no reduction in the primary cardiovascular endpoint. Benefit appears only in adherent (\u22654 h/night) subgroups by post-hoc analysis (RICCADSA HR 0.29; SAVE propensity HR 0.52).
Surrogate \u2260 outcome. AHI is a surrogate. Tirzepatide, hypoglossal-nerve stimulation, mandibular devices and behavioural therapies have strong AHI evidence but no mortality trials \u2014 their outcome hazard ratios here are flagged PROVISIONAL and are mechanistic bounds, not proven effects.
Healthy-adherer confounding. The observational CPAP mortality benefit is vulnerable to the healthy-adherer effect: people who tolerate CPAP differ systematically from those who do not. The E-value states how strong such confounding would need to be to explain each estimate away.
Front-door assumptions only partly hold. AHI does not fully mediate OSA harm (intermittent hypoxia and sympathetic surges act partly outside the event count), and the AHI\u2192mortality path is itself confounded (reverse causation: cardiac disease worsens sleep-disordered breathing).
Tirzepatide OSA mortality is unproven. SURMOUNT-OSA measured AHI, weight and blood pressure \u2014 not cardiovascular events. Its outcome HR here is borrowed from SELECT as a mechanistic proxy; a dedicated OSA cardiovascular-outcomes trial is pending.

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 airway therapy acts through one shared mediator — reduction of the apnoea-hypopnoea index (AHI, events/hour). Each log-hazard-ratio is split into an AHI-mediated (indirect) and an AHI-independent (direct) part. Indirect parts are pooled through the mediator with dose-response saturation (you cannot abolish the same apnoea twice), removing the mediator cross-correlation; direct parts keep the residual eigenvalue correlation at ρ̄. A largely AHI-independent intervention (e.g. physical activity, whose benefit is mostly weight-/fitness-mediated) is therefore NOT fully discounted for overlap with airway therapies.

Mediator saturation cap = 45 events/h
Sum of standalone AHI reduction (naive)
Combined AHI 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)
InterventionHRAHI↓ (/h)AHI-medindirect logHRdirect logHR

Which % of cross-correlation is appropriate? Not one number. The mediator overlap is fixed empirically by the AHI dose-response saturation (currently removing of the summed AHI-mediated effect when airway therapies are stacked — two devices cannot each abolish the same apnoeas); the direct overlap is the residual ρ̄. AHI-mediated fractions are transparent, adjustable priors — never presented as measured facts.

Front-door caveat (antithesis): AHI is a surrogate. Randomised CPAP trials (SAVE, RICCADSA, ISAACC) cut AHI yet showed NULL effects on hard cardiovascular endpoints — largely an adherence problem, but also a warning that surrogate improvement need not become mortality benefit. Several outcome HRs here are provisional (surrogate-derived); indirect estimates are bounds.

Executive summary

Select interventions to generate a plain-language summary.