Bayesian Simulation · Pre-Publication

What If We Could Preview ORBITA-CTO?

Predicting the first sham-controlled CTO PCI trial — on the exact daily transition OR it will report

1.05
Predicted OR
0.76–1.39
95% CrI
60%
Pr(Benefit)
~44%
Power N=50
How We Got the Number
1.22
Start Here
ORBITA-2: PCI vs sham
in non-CTO patients
→
↓
CTO Discount
CTO trials show
smaller benefits
→
↓
Placebo Strip
Remove open-label
inflation
→
1.05
Prediction
ORBITA-CTO
sham-controlled
The Equation
Effecti = μ + δopen × CTO + bias × (1−sham) + ε
μ = 0.199
True PCI effect over sham. From ORBITA-2 — locked, cannot be moved by CTO data.
δopen = −0.05
CTO collateral attenuation + open-label placebo — combined, inseparable.
bias = unknown
Placebo inflation in open-label trials. Sham kills it — but we don’t know how big it was.
ε ~ N(0, τ²)
Random trial-to-trial noise. τ ≈ 0.065.
Evidence Base
TrialDesignShamMeasuredMapped OR
ORBITA-2 Non-CTO Yes Daily ordinal diary β=0.199 1.22
EuroCTO CTO No SAQ AF +5.2 pts → crosswalk ~1.07
COMET-CTO CTO No SAQ AF +13.0 pts → crosswalk ~1.20
DECISION-CTO CTO* No SAQ AF +0.8 pts → crosswalk ~1.01
ORBITA-CTO CTO Yes Daily ordinal diary → ??? 1.05
*Both arms received non-CTO PCI (dampened by κ). Crosswalk: 1 SAQ point = 0.014 daily PO log-OR, calibrated from ORBITA-2.
The Puzzle
1
ORBITA-2 showed OR 1.22 for PCI vs sham
Same endpoint ORBITA-CTO will use. Non-CTO. Sham-controlled. The cleanest measurement of what PCI does to symptoms.
2
CTO trials showed smaller symptom improvement
EuroCTO: SAQ +5.2 pts. DECISION-CTO: SAQ +0.8 pts. Compare to ORBITA-2: +14.4 pts. On the same scale, CTO effects look smaller.
3
Why? Two possible explanations — and we can’t separate them
Collaterals reduce the gain
CTO patients have collateral supply. They’re partially compensated. PCI fixes the artery, but the patient wasn’t suffering as much. The discount is real.
Open-label inflated the other trials
All CTO trials were open-label. Patients knew they got PCI. ORBITA-2 had sham control. The gap could be the placebo effect present in CTO trials but absent in ORBITA-2.
Why can’t we tell which one?
Because every existing CTO trial is both open-label AND CTO at the same time. The two effects are always mixed. ORBITA-CTO will be the first to separate them.
Posterior Distribution — 60,000 MCMC Samples
Mean OR
1.05
95% CrI: 0.76 – 1.40
Harm (<0.8) Neutral (0.8–1.0) Modest benefit (1.0–1.5) Clear benefit (>1.5)
What Does This Mean Clinically?
~1
Extra angina-free day
over 6 months (180 days)
~44%
Chance of significance
at N=50 (coin-flip odds)
N>300
Needed for 80% power
this is a pilot, not pivotal
500 virtual ORBITA-CTO trials were simulated (25 PCI + 25 sham, 180-day ordinal diary). In the most plausible scenario, PCI adds approximately 1 extra angina-free day over 6 months — difficult for patients to perceive. The trial has coin-flip odds of statistical significance. This is a feasibility pilot — it proves the concept, not the efficacy.
Does CTO PCI Help Symptoms? Three Possibilities
The answer depends on how much placebo inflation was hiding inside the open-label CTO trials. The more placebo there was, the worse the real CTO PCI effect turns out to be — because once you strip the placebo away, less genuine benefit remains.
Open-label CTO trials had little placebo inflation
What we saw in EuroCTO, COMET-CTO etc. was mostly real. Collaterals do modestly reduce the symptom gain, but PCI still provides meaningful benefit even with sham control.
1.10
Predicted OR
✓ PCI helps — modest but real symptom improvement (73% probability of benefit)
Open-label CTO trials had moderate placebo inflation ★ Most plausible
Some of what CTO trials showed was real, some was placebo. Once stripped, only a small symptom benefit remains. Consistent with FFR physiology: CTO PCI improves flow but collaterals were already partially compensating.
1.05
Predicted OR
✓ PCI helps — but the benefit is small and hard for patients to notice (60% probability of benefit)
Open-label CTO trials had substantial placebo inflation
Most of what CTO trials showed was patients believing they felt better, not actually feeling better. Once the placebo is stripped by sham control, almost no symptom benefit remains.
1.00
Predicted OR
✗ PCI does not help symptoms — physiology improves but patients cannot tell the difference (47%)
Where does FFR physiology point?
CTO territories have lower FFRmyo (0.4–0.6). PCI provides a larger hemodynamic improvement. Symptom relief is threshold-based. This argues there IS a real CTO benefit — pointing toward the green or blue scenario. But how big that benefit is after removing placebo remains the open question.
What If We’re Wrong? — 10 Stress Tests
What if…OR
Our best estimate (base case) 1.05
Collaterals don’t reduce the benefit at all?CTO patients respond identically to non-CTO patients 1.08
Collaterals reduce the benefit much more than we assumed?Mature collaterals strongly blunt any symptom gain from PCI 0.99
We’re very uncertain about the size of the collateral effect?Could be large or small — we widen the range 1.07
We trust ORBITA-2 more as our starting point?Stronger assumption that non-CTO PCI effect transfers to CTO 1.05
We trust ORBITA-2 less as our starting point?CTO patients may differ substantially from non-CTO 1.05
Our SAQ-to-diary score conversion is imprecise?Different symptom questionnaires may not translate perfectly 1.04
CTO trials vary more from each other than expected?Different patients, operators, and techniques across trials 1.05
We drop DECISION-CTO entirely?That trial is problematic — both arms received non-CTO PCI 1.06
We ignore all CTO data and only use ORBITA-2?No borrowing from any CTO evidence at all 0.98
Across all 10 stress tests: OR 0.98 – 1.08 — the prediction barely moves.
~44%
Power at N=50
OR 1.05
Best Estimate
N>300
For 80% Power
PCI will fix the physiology. Whether patients perceive it is what ORBITA-CTO will answer — and finally separate real benefit from placebo.
Bayesian simulation · published evidence only · 240K posterior samples · fully converged · no insider data
github.com/ihtanboga/orbita-cto-bayesian-v2