A Judea Pearl Structural Causal Model with a rare property: a single do(·) — antiretroviral therapy and the viral suppression it produces — drives two different outcomes at once. It restores near-normal life expectancy for the person (survival axis) and, by rendering the virus untransmittable, prevents onward infection (prevention axis). The atlas's clearest "double dividend." All effect sizes are real published values; aggregation is an upper-bound ceiling.
Causal question. What is the identifiable causal contribution of antiretroviral therapy (and the viral suppression it produces) to (a) the survival of the person with HIV and (b) the prevention of onward transmission — and how does the survival effect depend on CD4 count at initiation?
| Axis | Population | Endpoint | Anchor (no intervention) | do(·) |
|---|---|---|---|---|
| Survival | Person living with HIV | Mortality / life expectancy | Untreated median survival ~10–11 yr; near-universal AIDS | ART — moderated by CD4 at initiation |
| Prevention (onward) | Serodifferent partnership | Sexual transmission to partner | Transmission per untreated viraemic partner | Viral suppression (U=U) |
| Prevention (acquisition) | HIV-negative person at risk | HIV acquisition | Background incidence in the at-risk group | Pre-exposure prophylaxis (PrEP) |
~39–40 million people live with HIV worldwide. The defining causal fact of the modern era: the same suppression that keeps a person alive also makes them non-infectious.
One intervention, two outcome branches. ART suppresses viral load; suppressed viral load both halts immune destruction (survival) and eliminates infectiousness (prevention). CD4 at initiation is an effect-measure modifier on the survival branch — start early and the life-expectancy gap nearly closes; start late and immune recovery is incomplete. PrEP acts on a separate node: the susceptibility of the HIV-negative partner.
| Operation | Method | Implementation here |
|---|---|---|
| Dual-outcome do(·) | Shared cause, two effects | ART → viral suppression → {survival, non-transmission}; one intervention, two outcome branches |
| Survival modification | Effect-measure modifier (CD4) | Life-expectancy gap and mortality reduction scale with CD4 at initiation (NA-ACCORD; START) |
| Treatment-as-prevention | do(suppress) on transmission | HPTN 052 96% reduction; U=U (viral load <200) → effectively zero sexual transmission |
| Acquisition prevention | do(PrEP) on susceptibility | ~99% reduction in acquisition with adherence (oral TDF/FTC; CAB-LA superior) |
| Robustness | E-value (VanderWeele & Ding) | Computed for the survival mortality-reduction effect |
| Minimum-effective set | Pareto frontier (threshold slider) | Interventions ranked by relative reduction on their targeted outcome (Section IV) |
| Intervention | Effect | Source / grade |
|---|---|---|
| ART — AIDS / hospitalization / mortality | 60–80% reduction | StatPearls; cohort syntheses A |
| Early ART vs deferred (serious events + death) | Significant reduction | START & TEMPRANO 2015, NEJM A |
| Near-normal life expectancy | Approaches general population | NA-ACCORD; ART Cohort Collaboration A |
| ART as prevention (onward transmission) | 96% reduction (linked) | HPTN 052, NEJM 2011 A |
| U=U (viral load <200 copies/mL) | ~0 sexual transmission | PARTNER / PARTNER2 / Opposites Attract A |
| PrEP (oral TDF/FTC, adherent) | ~99% reduction in acquisition | iPrEx / PROUD / IPERGAY; CAB-LA superior A |
| Untreated HIV (no ART) | Median survival ~10–11 yr | pre-ART natural-history cohorts B |
The evidence here is unusually strong — multiple landmark RCTs and the convergent PARTNER studies, which recorded zero phylogenetically linked transmissions from a virally suppressed partner across many thousands of acts.
HIV interventions ranked by relative reduction on their targeted outcome — mortality (survival levers), onward transmission, or acquisition. Drag the threshold to set a minimum: levers at or above the line form the minimum-effective set. Because the endpoints differ, this Pareto compares effect magnitude, not interchangeable outcomes.
ART is the only intervention in the atlas that simultaneously saves the treated individual and protects the untreated population. Treatment is prevention — which collapses the old ethical tension between caring for the sick and stopping spread.
The survival benefit is real at any CD4, but the life-expectancy gap is largely set by how early therapy begins. The do(·) that matters most for individuals is not just ART but early diagnosis — the upstream node.
PARTNER recorded zero linked transmissions from suppressed partners across thousands of acts. "Untransmittable" is not a slogan but an estimated causal effect with a tight upper bound — one of the cleanest do-effects in epidemiology.
Acquisition can be blocked from either side: suppress the positive partner (U=U) or protect the negative partner (PrEP). The model shows these as independent do-targets that combine toward elimination.
| Challenge | Why it threatens the conclusion | Mitigation here |
|---|---|---|
| Life-expectancy figures are modelled | LE gap depends on era, regimen, comorbidity, and population, not just CD4 | Presented as modelled estimates anchored to NA-ACCORD ranges, not individual predictions |
| Adherence is assumed | U=U and PrEP efficacy collapse without sustained adherence and suppression | Effects framed as the adherent-use ceiling; real-world effectiveness is lower |
| Mixed endpoints in one Pareto | Mortality, transmission, and acquisition are different outcomes | Pareto axis explicitly labelled as relative reduction on each lever's targeted outcome |
| "Zero" has uncertainty | PARTNER's zero has a confidence interval; rare residual risk cannot be excluded | Stated as ~0 / effectively zero, not literally impossible; graded A but bounded |
| Survival benefit ≠ cure | ART controls but does not eradicate HIV; lifelong therapy is required | Framed as near-normal life expectancy on continued therapy, not cure |
mortalityReduction = ART ? lerp(0.60 → 0.80 across CD4 50 → 500) : 0
LE gap (yrs) = ART ? lerp(13 → 1 across CD4 50 → 500) : ~40+ (untreated, decades lost)
onward transmission reduction = (ART & suppressed) ? ~0.96–1.00 (U=U) : 0
acquisition reduction = PrEP ? 0.99 : 0
E-value(RR) = RR* + √(RR*·(RR*−1)), RR* = 1/RR ; for mortality RR = 1 − reduction
Worked example: ART at CD4 500 → ~80% mortality reduction, LE gap ~1 yr (near-normal); U=U → ~96–100% onward-transmission reduction; add PrEP → ~99% acquisition reduction for the partner. The same suppression produces the survival and the prevention dividend.
1. HIV Antiretroviral Therapy. StatPearls (NCBI Bookshelf). Combination ART reduces AIDS, hospitalization, and mortality by 60–80%; U=U: undetectable viral load → virtually zero transmission risk.
2. INSIGHT START Study Group. Initiation of Antiretroviral Therapy in Early Asymptomatic HIV Infection. N Engl J Med 2015;373:795–807. Early ART reduces serious events and death vs deferral.
3. TEMPRANO ANRS 12136 Study Group. A Trial of Early Antiretrovirals and Isoniazid Preventive Therapy in Africa. N Engl J Med 2015;373:808–822.
4. Cohen MS, et al. (HPTN 052). Prevention of HIV-1 Infection with Early Antiretroviral Therapy. N Engl J Med 2011;365:493–505. 96% reduction in linked transmission with early ART.
5. Rodger AJ, et al. (PARTNER / PARTNER2). Risk of HIV Transmission through Condomless Sex with Suppressed Partner. JAMA 2016 / Lancet 2019. Zero phylogenetically linked transmissions with viral load <200 copies/mL (U=U).
6. ART Cohort Collaboration; NA-ACCORD. Life expectancy of people on ART approaches the general population, particularly when initiated before CD4 falls below ~200–350 cells/mm³.
7. Grant RM, et al. (iPrEx); McCormack S, et al. (PROUD); Molina J-M, et al. (IPERGAY): oral PrEP reduces HIV acquisition by ~99% with high adherence. HPTN 083/084: long-acting cabotegravir superior to daily oral PrEP.
Longevity Research Initiative — Bayesian Causal Atlas. SCM methodology after Judea Pearl; E-values after VanderWeele & Ding. All effect sizes are published values cited above; none were fabricated. Outputs are upper-bound ceilings for research, not clinical guidance. Report generated .