Dose-Response Relationships
A dose-response (or exposure-response) relationship exists when the risk of disease changes in a graded way with the level, intensity, or duration of an exposure. In chronic-disease epidemiology, observing that risk rises as exposure increases - what Bradford Hill called the biological gradient - is one of the considerations that strengthens the case for a causal interpretation of an association.
Definition
A dose-response relationship is a consistent, ordered association in which the magnitude of disease risk varies systematically with the dose - the level, intensity, frequency, or duration - of an exposure, such that higher (or lower) exposure corresponds to predictably different risk.
Scope
The entry covers the concept of graded exposure-response, the main shapes such relationships take (monotonic, threshold, saturating, and non-monotonic), the role of the biological gradient in causal inference, and the cautions that confounding and measurement error can also produce or distort gradients. It is a methodological topic and does not provide clinical guidance.
Sub-topics
Core questions
- What does it mean for an exposure-outcome association to show a biological gradient?
- What shapes can a dose-response curve take - monotonic, threshold, saturating, or non-monotonic?
- Why does a dose-response relationship strengthen, but not prove, a causal interpretation?
- How can confounding or measurement error create or mask an apparent gradient?
Key concepts
- Biological gradient
- Monotonic exposure-response
- Threshold and no-threshold models
- Saturating (plateau) response
- Non-monotonic (J- or U-shaped) curves
- Cumulative dose and duration
- Confounding by indication or lifestyle
- Exposure measurement error
Mechanisms
A dose-response relationship summarises how risk tracks the level of exposure. The simplest form is monotonic, where risk rises steadily with dose; a threshold form implies risk only above some level, while a no-threshold model assumes risk extends to the lowest doses. Saturating curves plateau once a biological effect is maximal, and non-monotonic J- or U-shaped curves indicate that both low and high levels carry excess risk. Because a graded relationship is hard to explain by chance and harder to fabricate by simple bias, its presence supports causality - this is Hill's biological gradient. Yet a gradient is not decisive: confounders that themselves vary with exposure can generate one, and random or systematic error in measuring exposure can flatten or distort the observed curve, so dose-response evidence must be interpreted alongside the other considerations for causal inference.
Clinical relevance
Dose-response evidence informs how exposure limits and risk thresholds are reasoned about in preventive and environmental health by describing how risk scales with exposure. This entry presents the concept at a population and methodological level for reference; it is not a basis for individual diagnostic or treatment decisions and contains no dosing guidance.
Epidemiology
The relationship between amount smoked and mortality in the fifty-year British Doctors Study is a textbook biological gradient, with risk rising as daily cigarette consumption increases. Global-burden analyses build graded exposure-response functions for many risk factors to estimate how disease burden changes across the range of population exposure.
History
The idea that a graded relationship supports causation crystallised in mid-twentieth-century cancer and cardiovascular epidemiology, where studies of smoking showed risk climbing with dose. Hill's 1965 enumeration of considerations for causation named biological gradient explicitly, and long-term cohorts such as the British Doctors Study supplied its most cited empirical demonstration.
Debates
- Is the absence of a dose-response relationship evidence against causation?
- Hill noted that a gradient strengthens causal inference, but its absence does not refute causation: threshold effects, saturation, competing risks, and measurement error can flatten a true relationship, so a missing gradient is not by itself disqualifying.
Key figures
- Austin Bradford Hill
- Richard Doll
- Richard Peto
- Kenneth Rothman
Related topics
Seminal works
- hill-1965
- doll-2004
Frequently asked questions
- Why does a dose-response relationship support a causal interpretation?
- When disease risk rises in a consistent, graded way with the level of exposure, the pattern is difficult to attribute to chance and harder to produce by simple bias, so it adds weight to a causal interpretation - though it is one consideration among several, not a proof.
- What is a U-shaped or J-shaped dose-response curve?
- It is a non-monotonic relationship in which risk is elevated at both low and high levels of exposure, so the lowest risk lies at an intermediate level. Such shapes warn that the simple 'more exposure, more risk' assumption does not always hold.