For funders
A staged programme with explicit go and no-go points.
PFL asks for funding one decisive test at a time. Each stage is registered in advance, and its result decides whether the next one happens.
Already done, without dedicated funding
238 biomarker hazard curves from 492 verified sources. Our own validation on 51,662 US adults with 7,703 deaths, where 29 of 35 published directions reproduce. Replication in Costa Rica and the Health and Retirement Study. Within-person trajectories for 13 measures. A ranking of which tests carry the most information.
Stages and decision points
| Stage | The test | Proceed if | Status |
|---|---|---|---|
| 1. Catalogue | Published curves reproduce in a national sample | Most directions reproduce | Done: 29 of 35 |
| 2. Variability pilot | Within-person variability predicts mortality beyond level and trend (HRS) | Decision rules registered at osf.io/kzfs4 | Analysis running |
| 3. One landscape | Shared curves plus personal parameters predict individual trajectories on repeated blood (CLSA, ARIC, Framingham) | Better prediction than baseline risk, at a threshold registered before the data arrive | Needs funding |
| 4. Health-system test | Retrospective validation on routine labs inside a partner's environment | Improves on current risk scores | Partner discussions |
| 5. Prospective pilot | Measure, locate, intervene and re-measure in a small cohort | Stages 3 and 4 succeed | Design only |
The current ask
Stage 3 is budgeted at $400,000 over 12 months: a data scientist, data access and secure computing, the open platform, and the design of the prospective pilot. Smaller contributions can fund a single piece, such as a trainee on one open problem or a data-access fee.
Funders receive the same thing as everyone else, sooner: registered analysis plans, every result whatever it shows, and open code and data.