Introduction
Health as a position on a landscape, and movement across it
Each biomarker is an axis. Over every axis lies a measured all-cause mortality hazard curve from published cohorts, checked against NHANES where it is measured there. Together they form a terrain: high ground is lower risk, low ground is higher risk, and a person is a point that moves. This dashboard holds the catalogue behind physiologicalfitnesslandscape.com: every curve, distribution, source, validation table and simulator.
The hazard curve, the population distribution by age and sex, catalogued interventions and disease alterations for each of the markers.
Our own NHANES Cox fits against the literature, cross-cohort replication, and two-marker mortality surfaces.
Mortality strength times age drift times modifiability, and every citation tagged with what it supports.
Research prototype. Hazards are associations adjusted for age and sex; nothing here is a diagnosis or a recommendation to test or treat. Panels marked proposed describe methods that have no data on site yet.
Definitions
Working terms, with their status
From the research workspace: what each term meant earlier, what it means now, and the decision in force. The term chips on the Terrain & Stability panel point here.
Framework
Sections of the research workspace
The framework text as it stands in the workspace export, including the correspondence with Metabolism & Medicine Vol. 1 and Vol. 2. Open a section to read it; the open questions follow.
Open questions
Introduction
Growth after recovery
A challenge can leave a capacity marker higher once recovery has had time to work. Each line is a published result: higher than the same adults' baseline, or higher than a control who did not get that dose.
- Twelve trained cyclists. Ten days of exercise near 50% VO2max in 40°C heat, versus the same exercise at 13°C. Cool VO2max rose 5% and hot VO2max rose 8%. Time-trial work rose 6% in the cool and 8% in the heat. Lactate-threshold power rose 5% in both. Plasma volume rose 6.5%. Maximal cardiac output rose 9.1% in the cool and 4.5% in the heat. The cool-exercise control was unchanged. Lorenzo et al. 2010, PMID 20724560.
- Fourteen endurance-trained men. Five days of isothermic heat acclimation versus thermoneutral exercise. Resting heart rate fell 8 bpm and plasma volume rose 7.3%. Sweat loss rose by about 500 mL. Kidney injury markers were unchanged. Snape et al. 2026, PMID 42579663.
- Sixteen triathletes. Two 18-day live-high train-low camps. VO2max rose 4.9% in normobaric hypoxia and 3.2% in hypobaric hypoxia. Saugy et al. 2016, PMID 27148076.
- Meta-analysis, 62 studies, 1,332 people. Live-high train-low raised VO2max more than control in athletes (mean difference 2.17, 95% CI 0.77 to 3.56) and not clearly in non-athletes (3.1, −0.23 to 6.43). Kohlbrenner et al. 2026, PMID 42742930.
- BCG vaccination in healthy volunteers. Monocyte TNF and IL-1β responses to unrelated pathogens were about twofold higher, and still present at 3 months. Kleinnijenhuis et al. 2012, PMID 22988082.
- Pilot trial. A 90-minute exercise bout immediately after SARS-CoV-2 immunisation lengthened antibody durability in people without pre-existing immunity. Kothadiya et al. 2026, PMID 42502427.
- HERITAGE Family Study, 371 men and 465 women. Twenty weeks of endurance training raised mechanical efficiency 4–5%, with a similar rise in women and men. Tremblay, Bouchard et al. 2025, PMID 41162184.
- Meta-analysis of stopping exercise, 86 trials, 5,670 people. Lower waist-hip ratio, mean arterial pressure and fasting glucose were still present after withdrawal. Zhang et al. 2026, PMID 42558561.
- Meta-analysis, 12 studies, 859 team athletes. Blood-flow restriction added to resistance training increased hypertrophy (SMD 0.32) and strength (SMD 0.42). Jumping and sprinting did not change. Huang et al. 2026, PMID 42549104.
Those percentages are the studies' results. The Stress & recovery tab is a separate model with illustrative parameters, and it does not reproduce them. The same introduction is on physiologicalfitnesslandscape.com.
All-Cause Mortality Hazard Landscape Continuous Spline
Non-linear mortality hazard ratio curve overlaying NHANES representative population distribution.
Derived Analytics Optimization Layer
Quantifying the expected mortality hazard reduction if the entire population distribution shifts toward optimal physiological targets.
Value of Information
Expected years of life lost and dollar-valued information from measuring this biomarker, using the same SD-shift as the simulator above. Life tables: CDC NVSS US 2022 (period). Uses age/sex from the Fitness Landscape filters.
EHIV = λ · E[VOI] − ctest. λ is not a scientific estimate of “correct” value — it is the decision-maker’s willingness to pay per DALY. Population VOI on the leaderboard sums exclusive age × sex cells (overlapping NHANES aggregations are excluded).
| Scenario | Baseline $E[HR_0]$ | Optimized \(E[HR_\delta]\) | Δ HR | Rel. Reduction ($RHR$) | % Benefiting | Domain Status |
|---|
Hazard Ratios by Clinical Category & Strata
Point estimates and 95% confidence intervals from prospective cohort studies.
Interventions and exposures that move this biomarker
Each card is one source: a Cochrane or other meta-analysis, a randomised trial, or a ClinicalTrials.gov posted result. The effect is the change in the marker as the source reports it; within-arm changes are labelled. Favourable means the marker moves toward lower hazard for this biomarker's direction.
Chronic Diseases Altering this Biomarker
Which diseases push this marker, in which direction, and where that lands on the mortality landscape.
NHANES Population Benchmarks by Age Band & Sex
Stratified median and 75th percentile values across representative adult cohorts.
The shared-content proposal names fasting C-peptide above 2.0 ng/mL and a triglyceride/HDL ratio above 2.0 (men) or 1.5 (women) as signs of insulin resistance before HbA1c moves. Both are derived axes with their own all-cause hazard curves (NHANES, 2019 linked mortality, residualised for age and sex) and gap-table rows; the strip gives the survey-weighted share of normoglycaemic adults (HbA1c under 5.7%) past each cut-point.
NHANES 2013–2016 measured total testosterone, estradiol and SHBG; free estradiol follows the Vermeulen equation where albumin is on the row. Hazard curves are fitted per sex with the same estimator as the other axes, on three to six years of follow-up, so they are short-horizon and wide. The insulin → SHBG column of the coupling matrix needs a longitudinal cohort with SHBG, which is not on site.
Before a capacity proxy stands in for altitude it is validated against the hazard surface it would replace: each proxy's own all-cause hazard curve (NHANES, 2019 linked mortality, the same estimator as the other axes), the C-index gain from adding the composite to a model with the biomarker levels, and the first direct test that the books' altitude and the model's steepness are related: heart-rate recovery after the treadmill against estimated fitness (NHANES 1999–2004), and the slow pull-back k by baseline grip tertile (ELSA).
Chronic Disease Intelligence & Multi-Scale Alterations Explorer
Chronic diseases with multi-scale alteration profiles (Types A–E), the direction each alteration takes (↑ raised, ↓ lowered, ⇕ dysregulated), and links to the landscape biomarkers they move.
Vol. 1 Fig. 5.19C distinguishes the barrier to leaving a healthy valley from the barrier to returning from a diseased one. Where the stationary potential U = −ln p of the NHANES population has two modes, the pass between them gives both heights: forward = U(saddle) − U(health mode), return = U(saddle) − U(disease mode), with the mode at higher hazard-based altitude taken as health. Their ratio is a reversibility index: above 1, coming back is harder than leaving. Pairs and single markers with one mode have no barrier to report.
| Axes | Health mode | Disease mode | Barrier forward | Barrier return | Reversibility index | Source |
|---|
Optimization & Population Impact Leaderboard
Comparative ranking of all biomarkers by standardized expected mortality hazard reduction ($RHR = 1 - E[HR_\delta]/E[HR_0]$).
Full Physiological Optimization Ranking
Click any row to inspect biomarker curve & individual benefit breakdownMulti-Biomarker Comparative Landscape
Compare mortality hazard ratios, population distributions, and risk curves across up to 8 physiological markers simultaneously.
Comparative All-Cause Mortality Hazard Ratios
Comparative Benchmark Metrics
Evidence tiers
Every panel carries one of these badges; the badge links back here.
Primary Literature & Cohort Citations
Every citation in the catalogue, tagged with what it supports: a hazard association, a population distribution, a distribution fit or an intervention effect, and for which biomarkers.
Considered, not yet on the landscape
Markers without adequate general-population mortality evidence. No curve is shown.
Programme status
PFL is a research programme with an open prototype. This table records the phase the work has reached.
| Phase | Status | Gates |
|---|
External Validation on NHANES
Does a PFL multi-biomarker composite predict all-cause death in people it was not fitted on, better than standard risk factors and PhenoAge? NHANES 1999–2018 + the CDC public-use Linked Mortality Files (deaths through 31 Dec 2019). Held-out 30% test split; models fit on the other 70%.
Catalogue hazards vs. NHANES empirical hazards
Each point is one biomarker: x = literature HR per SD as encoded in the catalogue (harmful direction), y = our own age/sex-adjusted Cox fit on NHANES 1999–2018 + Linked Mortality 2019 (95% CI). Points on the diagonal mean the published association reproduces in a national sample.
Per-biomarker comparison
Sorted by disagreement. Skewed markers were modelled on the log scale; HR per SD is then per SD of log(x).
| Biomarker | Catalogue HR/SD | NHANES HR/SD [95% CI] | n / deaths | Non-linearity | Verdict |
|---|
Held-out discrimination
C_within_age_bands is the number that matters most: age alone predicts death well, so this measures discrimination within 10-year age bands — whether the model tells apart people of the same age.
| Model | C-index | C within age bands |
|---|
Robustness checks
Paired bootstrap difference in C-index, PFL vs. conventional risk factors, 95% CI. The reverse-causation check drops deaths in the first 24 months (people already sick at the exam); the alternate split re-runs with a different random train/test partition.
Calibration (held-out)
10-year predicted vs. observed (Kaplan-Meier) mortality, by decile of predicted risk.
Does PFL add information?
Likelihood-ratio test: PFL biomarkers added to the conventional risk-factor model, fit on the training split.
Observational associations, not causal effects. Single measurement per person (NHANES has no repeat visits). Public-use mortality data has small amounts of deliberate noise added by CDC for privacy. Complete-case analysis: people missing any marker are dropped.
Two-Biomarker Mortality Landscapes
Biomarkers can interact: the mortality cost of one can depend on the level of another. Each surface is an age/sex-adjusted Cox model with a two-way spline interaction, fit on real NHANES participants. Altitude = −log(HR) relative to the population-median point (+); green peaks are lower-mortality regions, red valleys higher-mortality. Hatched/outlined regions are the steepest 10% of the surface (“cliffs”) — where a small move in either biomarker changes risk the most.
The same surface as hazard-based altitude F*, beside the stability layer U.
Reproduce with:
python 03_landscape.py --data data/pfl_nhanes.csv
Add a third marker
Metabolic-inflammatory panel
Which of these values contributes most to this fitted population landscape?
Where these values sit on a population mortality landscape fitted to U.S. adults (NHANES). Observational, adjusted for age and sex only. Not a diagnosis or a personal risk estimate.
How People Move on the Landscape
A. Age profile vs. NHANES reference
p5–p95 fan, 5-year age bands| Group | n / deaths | HR per SD [95% CI] |
|---|
B. Within-person change per year
median (p10–p90) between visits ~4 y apartC. 4-year quintile transitions
D. Terminal trajectory — level by years before death
decedents only; mean ± 95% CIRead with care.
Source: Summaries generated .
Reproduce locally:
An intervention class moves its target axis and, for some classes, known off-target axes the wrong way: statins raise creatine kinase and ALT, thiazides raise glucose and uric acid, niacin raises glucose. Per class the panel sums the log-hazard change over the target axis and the adverse axes for every NHANES adult with all of them measured (Analysis 1 hazard curves, shifts per standard regimen from the trial contrasts), and reports the share whose summed hazard rises and the mean change by baseline quintile of the target. Axes without a trial report or a hazard curve are listed, not counted: the adverse cost is a lower bound.
Introduction
Growth after recovery
A challenge can leave a capacity marker higher once recovery has had time to work. Each line is a published result: higher than the same adults' baseline, or higher than a control who did not get that dose.
- Twelve trained cyclists. Ten days of exercise near 50% VO2max in 40°C heat, versus the same exercise at 13°C. Cool VO2max rose 5% and hot VO2max rose 8%. Time-trial work rose 6% in the cool and 8% in the heat. Lactate-threshold power rose 5% in both. Plasma volume rose 6.5%. Maximal cardiac output rose 9.1% in the cool and 4.5% in the heat. The cool-exercise control was unchanged. Lorenzo et al. 2010, PMID 20724560.
- Fourteen endurance-trained men. Five days of isothermic heat acclimation versus thermoneutral exercise. Resting heart rate fell 8 bpm and plasma volume rose 7.3%. Sweat loss rose by about 500 mL. Kidney injury markers were unchanged. Snape et al. 2026, PMID 42579663.
- Sixteen triathletes. Two 18-day live-high train-low camps. VO2max rose 4.9% in normobaric hypoxia and 3.2% in hypobaric hypoxia. Saugy et al. 2016, PMID 27148076.
- Meta-analysis, 62 studies, 1,332 people. Live-high train-low raised VO2max more than control in athletes (mean difference 2.17, 95% CI 0.77 to 3.56) and not clearly in non-athletes (3.1, −0.23 to 6.43). Kohlbrenner et al. 2026, PMID 42742930.
- BCG vaccination in healthy volunteers. Monocyte TNF and IL-1β responses to unrelated pathogens were about twofold higher, and still present at 3 months. Kleinnijenhuis et al. 2012, PMID 22988082.
- Pilot trial. A 90-minute exercise bout immediately after SARS-CoV-2 immunisation lengthened antibody durability in people without pre-existing immunity. Kothadiya et al. 2026, PMID 42502427.
- HERITAGE Family Study, 371 men and 465 women. Twenty weeks of endurance training raised mechanical efficiency 4–5%, with a similar rise in women and men. Tremblay, Bouchard et al. 2025, PMID 41162184.
- Meta-analysis of stopping exercise, 86 trials, 5,670 people. Lower waist-hip ratio, mean arterial pressure and fasting glucose were still present after withdrawal. Zhang et al. 2026, PMID 42558561.
- Meta-analysis, 12 studies, 859 team athletes. Blood-flow restriction added to resistance training increased hypertrophy (SMD 0.32) and strength (SMD 0.42). Jumping and sprinting did not change. Huang et al. 2026, PMID 42549104.
Those percentages are the studies' results. The Stress & recovery tab is a separate model with illustrative parameters, and it does not reproduce them. The same introduction is on physiologicalfitnesslandscape.com.
When stress builds capacity—and when it scars
Explore how dose, control, predictability, recovery, rumination and support change a stressor's modelled outcome.
A theoretical model with illustrative parameters. It is not fitted to data and says nothing about any individual.
Build a scenario
Advanced model parameters
Interventions for the first stressor
Trajectories
Solid: current · long dash: without interventions · short dash: no stressor.
Demand and capacity
Load
Capacity set point
“Growth” here means a higher capacity set point in the model. It is not the self-reported post-traumatic growth measured in questionnaires, which tracks real change poorly.
Growth window
Cells are labelled C collapse, S scarring, R resilience or G growth; the white outline marks the current setting.
What each intervention does here
Simulated changes use illustrative effect sizes, not study effect estimates. Unverified records are labelled.
| Intervention | Lever | Direction | Evidence | Citation | Class before → after | ΔC0 change |
|---|
Attribution
Shapley values of extra load; negative values are buffers.
What Stressors Do to Biomarkers Over Time
How measured biomarkers move before, during and after a defined stressor — caregiving, combat deployment, calorie restriction, spaceflight, polar overwintering, disaster, and their animal analogues — extracted from the source papers. Each series is one marker in one study arm. Where a paper reports only a direction, the chart shows the direction (higher / no change / lower) instead of a value. Series tagged with a landscape biomarker link back to its hazard curve.
Event-anchored displacement and recovery (cohorts)
proposedDefined life events in the longitudinal cohorts (ELSA: widowhood, retirement, smoking cessation, a first heart, stroke or cancer diagnosis; SWAN: natural menopause) give a known time zero. For each event and marker the pipeline fits one recovery rate krec and one kick, and a lasting displacement per person, against matched non-event persons measured at the same visits. A displacement not distinguishable from zero is an excursion; a lasting one is a basin shift if the nearest held level changed, a field shift if it did not (the level moved under the person). Group statistics only; residualised for age and sex; cells below 11 persons are blank. These are associations in observational cohorts, not effects of the events.
| Cohort | Event | Marker | Persons | krec (1/yr) | Kick (SD) | Lasting shift (SD) | Retained | Excursion / basin / field | Course step / gradual / flat |
|---|
Challenge protocols: one kick per organ system
proposedThe cardiac stress test as the template for every system: a known kick and a measured return. The return rate is the system's fast k; its inverse is the susceptibility (how far a sustained push moves the system). Rows marked now run on data already on site (NHANES oral glucose tolerance test; NHANES submaximal treadmill 1999–2004, heart-rate recovery over three minutes); the rest are the measurement plan for a pilot. Population values, one exam per person, no test-retest yet.
| System | Kick | Return window | Status | kfast | χ = 1/k | n | Source / protocol |
|---|
Data points
| Arm | Timepoint | Day | Value | SE/SD | n | p | Direction |
|---|
Source data and curation: data/stressor_trajectories/
(backend/build_stressor_trajectories.py).
Direction is relative to each series' stated reference.
The books define altitude as available free energy. That is not measurable in cohorts, but what it stands for (capacity to do physiological work, its efficiency, reserve) has validated measurables: estimated and measured cardiorespiratory fitness, grip strength, gait speed, FEV1, resting heart rate, and hsCRP as the dissipation term, plus an equal-weight composite Cv1. Each is fitted as a surface C(x) on the same axes as the hazard surface (the fourth fill, "Capacity C", in the panel above). The table gives cos θ between ∇C and −∇F* per cell: agreement supports capacity as altitude; disagreement locates where "fit" and "safe" differ.
| Pair | Proxy | Aligned (cos θ > 0.5) | Anti-aligned | Flat | Mean cos θ | Cells |
|---|
How to read the stability measures
Altitude is fitness (minus the log hazard ratio from NHANES); local shape is stability. A stable point is a level that the same people, measured years apart, drift back to (ELSA nurse visits, with the long-term reliability from three visits and a check for laboratory changes). It need not sit in the lowest-risk trough. The held-level, pull-back and found-in-resamples measures below are the inputs to the stability layer U of the panel above. The full interactive pages are at physiologicalfitnesslandscape.com/terrain. The Physiological Fitness Landscape concept is by Brian Fertig and Jack Tuszynski, building on Fertig's Metabolism and Medicine (CRC Press, 2022).
Pull-back table from data/stability/pullback.json; held points by sex and age band from the shared feed when it answers.
Where people are held on the terrain
Stable point: found in the main run and again without visit pairs flagged for a laboratory change. Candidate: only without them. Dropped: made by a laboratory change. Curves are fitted on both sexes (adjusted for sex), so for markers that run lower in women a women's point below the trough partly reflects that.
Near the edge: does recovery slow? (MIDUS)
Heart-rate recovery after a mental challenge, people near an edge against a reference group in the same wave (so the wave effect cancels). Beta-blocker users left out.
Early warning: slow recovery and hypertension a decade later
A hint to replicate, not a finding: the continuous score is weaker and there is no signal for diabetes.
Childhood adversity and the terrain later (ELSA)
Three or more adversities before sixteen against none, adjusted for age and sex. Overall position does not differ; inflammation is the exception.
A population-level digital twin: a point starts somewhere on the pair shown in the terrain view above, an intervention class pushes it with its trial shift vector Δc for the chosen duration, then the push is removed. The motion is dx = −D∇U dt + b dt + √(2D) dW on the pair's potential U and diffusion D, with the push b sized so that its steady-state displacement equals Δc. Under a fixed field every displacement rebounds: the curve shows how fast, and the hazard change at the end of treatment comes from the hazard surface. It is not a prediction for a person.
Trajectory on the pair (mean path and 20 noisy paths)
Displacement along the push, then rebound
Alterations of Every Measurable Capacity, Parameter and Biomarker
The framework of Halma et al. 2025 (component, change with age, rate, healthspan link, remedy, magnitude), applied to three kinds of measure. Each item is collapsed; open it for its trajectory with age, by default and given an intervention, and for the evidence behind it: functional capacities (strength, aerobic capacity, vagal tone, the senses, processing speed, learning, emotional stability, balance, sleep), physical parameters (muscle-fibre cross-sectional area, ejection fraction, flow-mediated dilation, brain and vessel structure) and every biomarker in the landscape. Literature rows are copied from PubMed abstracts and link to them. Catalogue rows are stored values: the age difference is a cross-sectional difference in the catalogue's population medians (NHANES for some biomarkers, other named cohorts for the rest), not within-person change.
Trajectory charts are illustrations, not forecasts: an intervention's effect is applied at once from the chosen age, assumed to persist and to add to the default, with the rate of ageing unchanged. Trial effects were measured over weeks to months in selected groups. Diseases, exposures and habits that worsen a measure are listed separately and never offered as an intervention.
Change with age per decade vs. the largest remedy effect
literature findings reported as percentages; largest reported value per itemThe shared content argues that hyperinsulinaemia drives allostatic load. The scorecard reports, per target axis, the cohort-lag estimate of how a higher fasting insulin moves the target over the next visit (Jtarget,insulin, per ln unit of insulin per year), the trial-side ratio where a class shifts both, and the Mendelian-randomisation sign where a table exists. These pairs are outside the prespecified coupling list: they are estimated with the same instrument, tagged exploratory, and are not findings. Only SWAN carries longitudinal insulin on this machine.
| Target axis | Jtarget,insulin (CI) | Jinsulin,target | Cohorts | Trial probe | MR sign | Direction conflict | Gate |
|---|
Where to Intervene — Mortality × Age Drift × Modifiability
A biomarker is a strong intervention target when it is tightly linked to mortality, drifts in the harmful direction with age, and can actually be moved. Each axis is expressed in population-SD units so they are comparable: ln HR per SD from the catalogued association; harmful drift in SD per decade (within-person from HRS where available, otherwise the cross-sectional 20–39 vs 60+ gap); and the best evidence-weighted favourable intervention effect in SD from the intervention catalogue. The target score is the geometric mean of the three percentile ranks.
Target map
x = harmful age drift (SD/decade) · y = ln HR per SD · bubble = modifiability (SD) · blue = within-person HRS driftRead with care. Cross-sectional drift mixes ageing with cohort and survival effects; only the 13 HRS measures are true within-person change (age 50+). Modifiability reflects what is catalogued (43 biomarkers so far) and converts percent effects via mean/SD, which inflates tightly regulated markers (calcium, haemoglobin). Capped at 2 SD.
Two interpretable products are in the table: hazard/decade = lnHR/SD × drift (log-hazard accrued per decade through this marker) and recoverable = lnHR/SD × modifiability. Generated .
Ranked targets
| # | Biomarker | ln HR/SD | NHANES HR/SD | 1-SD RHR | Drift SD/decade | Modifiable (SD) | Pull-back k (1/yr) | Leverage at mode | Information (ΔC) | Best catalogued intervention | Hazard/decade | Recoverable | Score |
|---|
A Cognitive Dimension — ELSA (England), on the US HRS-HCAP Scale
The Harmonised Cognitive Assessment Protocol scores cognition identically in the US (HRS) and England (ELSA), with factor scores calibrated so the US sample has mean 0 and SD 1. That makes cognition a landscape dimension that can be compared across countries. ELSA-HCAP also re-tested the same people five years later, and ELSA's genotyped sample lets us ask which landscape dimensions share genetic predisposition. All panels are weighted, small-cell-suppressed aggregates computed locally from UK Data Service End-User-Licence files.
A. Cognition by age and sex, ELSA-HCAP 2018 (65+)
Weighted mean of the cross-study general-cognition factor (US HRS-HCAP = 0 ± 1). Positive = English 65+ score above the US 65+ average on the shared item bank.
Line = weighted median; band = interquartile range; dashed line = US mean (0). Cells with n < 30 are blanked.
B. Five-year within-person change, 2018 → 2023
541 re-tested; mean change per year in baseline-SD units ± 95% CIC. Polygenic coupling between landscape dimensions
Pearson r between polygenic scores (ELSA PGS 2022), each residualised on 10 ancestry PCs. Descriptive only: these are not LD-score genetic correlations, and scores come from different GWAS sources and SNP-weighting methods.
D. Self-reported BMI during the pandemic (ELSA COVID-19, 2020)
weighted median & IQR by age band; dashed = NHANES measured mediansCaveats. HCAP sampling was stratified by prior cognitive status (weights correct for this). The 2023 re-test keeps survivors and non-refusers, and practice effects shrink apparent decline. 2020 height/weight are self-reported (web/phone), so BMI is biased low relative to NHANES measurements.
CogEcon's public files contain no physiological or cognitive-test measures (the cognitive battery is in the companion CogUSA study), so it is registered below but not plotted.
E. Replication in Costa Rica — CRELES, 60+
age/sex-adjusted HR per SD, weighted, robust SE; deaths to 2009| Marker | n / deaths | CRELES HR/SD | NHANES HR/SD | Catalogue |
|---|
NHANES = own empirical fit (20–84, all adults); catalogue = literature HR converted to per-SD. Inverse cholesterol/BP associations after 60 are the usual reverse-causation pattern in old-age cohorts.
F. Two-year within-person change, CRELES 2005 → 2007
SD per decade (baseline SD units), 60+Cohort data registry
| Cohort | n | Content | Access | Use in the landscape | Status |
|---|
Summaries generated . Reproduce: python elsa/01_elsa_summaries.py