The published evidence
base is distorted.
This corrects it.

An open-source computational engine that applies 14 quantifiable bias correction vectors to chronic disease prevention interventions — and outputs adjusted NNT scores the field has never produced.

Key findings — adjusted NNT (10-year all-cause mortality)

Published NNTs before bias correction. Why they differ →

The problem

The published medical literature is systematically distorted by four documented sources of bias that existing evidence syntheses do not correct for simultaneously: industry funding inflation (Cochrane meta-analysis: 3.6–4x favorable result bias), publication bias (94% of published antidepressant trials were positive; the FDA-registered rate was 51%), surrogate endpoint overuse (~50–60% predictive accuracy for hard endpoints), and structural underfunding of non-patentable interventions.

Standard NNTs inherit all four distortions. This engine removes them.

The oral-systemic finding

The most underappreciated number in the dataset. P. gingivalis and F. nucleatum have been physically isolated from coronary plaques, cardiac thrombus, and Alzheimer's brain tissue. The AHA 2025 Scientific Statement links periodontal disease to cardiovascular disease.

The research gap is not biological. It is jurisdictional. Dental billing codes are not anatomy. The classification of oral health as "dental" rather than "medical" is an insurance artifact — and it has suppressed one of the highest-value preventive interventions in chronic disease from appearing in the comparative evidence record.

The methodology

14 bias correction vectors. Each with documented direction, magnitude, and evidence source — separable from the scoring logic so they can be challenged independently. Multiplicative correction model. Composite scoring across 30+ interventions in 6 categories. All 5 pre-specified hypotheses confirmed.

The code is public. The correction factors are documented. Add any intervention. Challenge any magnitude. That is the point.