Adventist Health Study Diet & Cancer Risk Explorer

Cohort Simulator & Absolute vs Relative Risk Workbench
Meat-Eater (Baseline)
1,200
Cases per 10,000 | HR 1.00
Vegetarian Impact
1,056
144 Cases Avoided / 10k
Vegan Impact
912
288 Cases Avoided / 10k
Number Needed to Adopt (NNA)
35
Vegans to prevent 1 cancer case
Simulated 10,000 Individual Cohort Matrix

Each dot represents 25 individuals. Highlighted dots indicate predicted cancer diagnoses over the follow-up period.

Meat-Eater Case
Vegetarian Case
Vegan Case
No Diagnosis
Cumulative Hazard Curves (95% CI)

Projected cumulative probability of diagnosis over selected follow-up time horizon.

Epidemiological Evidence Workbench
Dietary Pattern Hazard Ratio (HR) 95% Confidence Interval Relative Risk Reduction Absolute Risk (Per 10k) Absolute Risk Reduction (ARR) NNA (1 Case Prevented)
Non-Vegetarian 1.00 (Ref) [1.00 - 1.00] 0.0% 12.0% - -
Vegetarian 0.88 [0.81 - 0.95] 12.0% 10.6% 1.44% 70
Vegan 0.76 [0.68 - 0.85] 24.0% 9.12% 2.88% 35

A saved risk illustration multiplies assumptions and rounds counts

Read the explanation

This explains only the saved arithmetic, not validated cancer risk, causal diet effects or medical guidance. The baseline is twelve hundred per ten thousand over ten years, multiplied by age and family inputs and by horizon divided by ten. With both multipliers one and a twenty-year horizon, the page displays twenty four hundred. It then multiplies that count by stored factors zero point eight eight and zero point seven six, producing twenty one hundred twelve and eighteen hundred twenty four after rounding. At zero point one pixels per count the widths are two hundred forty, two hundred eleven point two and one hundred eighty two point four. These are model assumptions, not personalized risk predictions. The modeled differences are two hundred eighty eight and five hundred seventy six per ten thousand. Dividing by ten thousand and multiplying by one hundred gives two point eight eight and five point seven six percentage points. At fifty pixels per percentage point, the bars are one hundred forty four and two hundred eighty eight. The inverse-difference labels round one hundred divided by each percentage-point difference, giving thirty five and seventeen. That is arithmetic applied to an assumed population, not evidence of a causal number needed to treat or a promise about an individual. The input factors and linear time scaling require independent scientific validation that this preview does not supply. The dot view uses a twenty by twenty grid: four hundred dots represent the assumed ten thousand, so each dot stands for twenty five. It rounds modeled counts divided by twenty five. Twenty four hundred produces ninety six highlighted dots; twenty one hundred twelve produces eighty four; eighteen hundred twenty four produces seventy three. At three pixels per highlighted dot, the comparison widths are two hundred eighty eight, two hundred fifty two and two hundred nineteen. This adds visual rounding beyond the count arithmetic. The JSON exports selected assumptions and calculated labels. Neither the chart nor its source attribution establishes clinical or causal validity.

Super generates helpful tools and automates fact-checking across the internet proactively. If you enjoyed this tool, build your own with Super and share it with a friend.