Interpreting Dose–Response Relationships in Longevity Research
Define the dose and the response
Distinguish the administered dose from the concentration reaching the body or tissue. Route, formulation, frequency and duration matter: identical total amounts can produce different exposure patterns. Benefits and adverse effects also need separate assessment, and population averages may conceal individual differences. These distinctions underpin ICH E4 dose–response guidance.
For a longevity paper, write down exactly what the response measures: molecular activity, a biomarker, physical function, disease incidence or survival. A curve for one endpoint cannot establish the curve for another. In particular, a change in an ageing-associated biomarker does not by itself demonstrate longer life.
| Pattern | Interpretation within the tested range | Question for the reader |
|---|---|---|
| Monotonic | The response moves consistently in one direction as dose increases. | Does the endpoint represent benefit, harm or only biological activity? |
| Plateau | Additional dose produces little further measured response. | Are the estimates precise enough to distinguish a plateau from an uncertain slope? |
| Threshold-like | A response becomes detectable above an apparent exposure level. | Could limited measurement sensitivity explain the apparent threshold? |
| U-shaped or inverted U-shaped | The direction of the association changes across the exposure range. | Are both sides supported by enough observations, and is the shape robust to alternative analyses? |
Keep laboratory, animal and human evidence separate
Laboratory evidence: A concentration–response experiment in cultured cells describes that experimental system. Ask whether the outcome reflects the intended mechanism, general cellular stress or loss of viable cells. It does not establish a safe or effective human exposure.
Animal evidence: Miller and colleagues reported that rapamycin-associated lifespan extension in mice depended on dose and sex. Their findings illustrate why the tested population and exposure range belong in the conclusion. They do not establish a human longevity dose. See the primary mouse study.
Human evidence: Separate randomised comparisons of assigned doses from observational associations with measured exposure. Ask whether the study measured clinical outcomes or only intermediate markers, and whether follow-up could detect delayed effects. Neither a cell response nor mouse survival can substitute for a human outcome.
Examine how the curve was estimated
A single active dose compared with a control cannot map the full curve. Randomised dose groups can help distinguish dose effects from differences between participants. In studies that increase doses over time, time-related changes can be mistaken for dose effects; ICH E4 discusses this limitation.
Use the following questions when reading a figure or results table:
- How many distinct doses were tested, and how many independent participants or animals contributed at each dose?
- Are uncertainty intervals shown, particularly at the extremes?
- Was the curve shape specified before examining the results, or selected after trying several models?
- Does a claimed difference between doses come from a direct comparison, rather than one dose being statistically significant and another not?
- Could illness change exposure in an observational study, rather than exposure causing the outcome?
- Were missing outcomes, treatment discontinuation and adverse events reported by dose?
A smooth fitted line should not be read as equally strong evidence everywhere along its length. Treat extrapolation beyond the observed range as an additional assumption.
State a conclusion the data can support
A useful conclusion names the population, exposure schedule, endpoint, comparator and uncertainty. For example: “Within the tested range, higher assigned exposure was associated with a larger change in the measured endpoint; effects on survival and other schedules remain unestablished.” This is an illustrative wording template, not a study finding.
The highest observed response is not necessarily an optimum. Calling a dose optimal requires a defined objective, adequate comparisons and consideration of harms. Likewise, a U-shaped association alone does not establish hormesis: that interpretation requires evidence beyond the visual shape of a curve.
Return to the research literacy collection for the wider context of evaluating longevity evidence.