Build your own map from any two biomarkers

Choose two markers. Where both were measured on the same people in NHANES, the map is a joint model fitted to 1,536 possible pairs. Otherwise it is an approximation built from the two separate curves, and the page says so.

Across the map Up the map

Risk of death1.00× reference

eGFR: 98
Systolic BP: 121
Within data: yes

Drag the marker, or focus the map and use the arrow keys. Hold Shift for larger steps.

Risk of death relative to the reference, by contour band.

Cox model with restricted cubic splines and a eGFR × Systolic BP interaction (likelihood-ratio p = 3×10⁻⁵), adjusted for age and sex. NHANES 1999–2018 with linked mortality to 2019; 46,469 adults, 6,688 deaths (fitted surface). The path joins the average adult in each decade of age, and each oval holds the middle half of people that age (NHANES, unweighted). Different people at each age, so this is how the population differs by age, not one person's path. Population associations from US adults, not a personal risk estimate.

Plot your own results over time

Type the date and both values from each of two or more test reports. The map draws your path in date order, and the table compares each step with how much people of your age and sex usually change in a year. Values stay in this browser.

Age at your latest result Sex

Show typical drift

Population associations from US adults, not a personal risk estimate.

Save on this device Load from CSV

Two kinds of map

Joint fit. Both markers enter one Cox model with restricted cubic splines and their interaction, adjusted for age and sex, in NHANES 1999–2018 with linked mortality. Each marker's effect is estimated with the other held fixed. Hatched areas hold too few people to estimate. Available when at least 2,000 people with 200 deaths had both measured.

Approximate. The two separate curves are added together: our NHANES curve where we have one, otherwise the published curve from the catalogue. The markers are not adjusted for each other, so related markers (such as HbA1c and glucose) overstate the combined risk, and any interaction is missed. Use these maps to explore, not to compare numbers.

Intervals. The read-out gives a 95% confidence interval for the risk of death. For a joint fit it comes from the model; for an approximation it adds the two curves' variances and ignores their correlation, and it is missing where a curve comes from the published literature. Either way it reflects sampling uncertainty only, not model or confounding uncertainty.

All maps show associations, not the effect of changing a value.