This was "vibe coded" — it has NOT been rigorously vetted.
Built quickly with an AI assistant as a back-of-the-envelope thinking tool. The numbers come from real published studies (cited below), but the model logic, the interpolated incidence curve, the family-history weighting, and the code itself have not been peer-reviewed, clinically validated, or checked by a statistician or physician. Expect bugs and simplifications. Do not use this to make any medical or financial decision. Talk to a real doctor.
If you think this is garbage, please build a better one. I would love to use it.
Test or Not — The Galleri Multi-Cancer Blood Test
Galleri (GRAIL) screens for a shared cancer signal across >50 cancer types. Its advertised specificity (~99.5%) and sensitivity (~51.5%) are population averages — they say nothing about you. This model folds in your age, sex and family history to estimate your personal odds of a true positive, false positive, true negative, or false negative — and what it costs to find one real cancer. Everything runs locally in your browser.
Educational model · not medical advice · sources cited at the bottom
Who you are
These tracks shade by your annual chance of having a detectable cancer: green = low · yellow = moderate · red = high. Drag age up and watch it redden.
55
Top = age · green = your chance of a true positive (real cancer caught) at that age.
Male
Drives the incidence curve (women higher in mid-life from breast cancer; men higher and rising after ~60).
Family history — first-degree relatives (parent / sibling) with…
0
0
0
0
0
0
Family-history sliders raise your overall cancer odds in proportion to each cancer's share of total incidence (so 1 relative with breast cancer ≈ +13% overall, not +100%). The hereditary toggle applies a large separate multiplier.
The test & the economics
These tracks shade by cost per real cancer found: green = cheap · yellow = moderate · red = very expensive.
$949
Galleri's list price is $949. Drag it to $10 (imagine it's cheap) or $10,000 to see the cost-per-cancer swing.
51.5%
Share of real cancers the test catches. CCGA-3 average = 51.5% across all stages, but only 16.8% at Stage I and 90.1% at Stage IV. Lower this to model early-stage reality.
99.5%
Share of healthy people correctly told "no signal." Even 99.5% means ~1 in 200 healthy people gets a false alarm.
$5,000
Imaging, biopsies and specialist visits to chase a false positive. Included in "all-in" cost only when toggled below.
Value assumptions — cost per QALY
These two are the biggest unknowns in the whole model — the QALY cost is hugely sensitive to them. Tracks shade by the resulting cost per QALY (green good · red poor).
45%
Of the cancers Galleri catches, the share where catching it this early actually changes the outcome — curable now, but wouldn't have been caught or cured in time otherwise. The single biggest lever. NHS-Galleri's missed endpoint argues it's low; GRAIL's own models assume it's high.
15
Quality-adjusted life-years gained by curing an otherwise-fatal cancer — weighted by each cancer's deadliness and capped by your remaining life expectancy, so older people gain fewer.
—
—
—
If POSITIVE ("cancer signal found") — chance it's real (PPV)
—
positive predictive value at your demographics
If NEGATIVE (the most likely result) — chance you still have cancer
Your chance of having cancer, by age — before any test vs. after a negative Galleri result
Have cancer — no test (base rate)Have cancer — after a negative GalleriRuled out by a negative95% CIYou, now
The blue–green gap is what a negative result rules out. When you're young both lines hug zero, so a clean result barely changes anything; as you age the gap widens — a negative knocks a real chunk off a much bigger number.
Your chance of having a DEADLY cancer, by age — and how far a negative shifts it down
Have a deadly cancer — no testAfter a negative Galleri↓ shift from a negativeYou, now
Deadly cancers (pancreas, liver, lung, esophagus, stomach, ovary, brain) are exactly the ones Galleri detects best, so a negative pushes this curve down much further than the all-cancer curve above — the arrows show the size of that drop at each age. This is the reduction worth caring about: the cancers it can't rule out are mostly survivable.
How the cost of finding one real DEADLY cancer falls with age