Minimum Clinically Important Differences in Healthy Ageing Trials

What does an important difference mean?

The minimum clinically important difference, also called the minimal important difference (MID), describes a small difference in an outcome considered important by patients or participants. It is an interpretive estimate, not a universal property of a questionnaire or physical test.

In healthy ageing research, first identify the outcome: walking ability, fatigue, daily functioning and quality of life answer different questions. Then establish whose judgement defines importance and whether the estimate concerns improvement, deterioration or a difference between groups. The BMJ’s MID selection framework prioritises both methodological credibility and applicability to the intended setting.

How thresholds are estimated

Anchor-based methods relate an outcome score to an external assessment, such as a participant’s report of a small but important improvement. Credibility depends on whether the anchor is understandable, relevant, sufficiently related to the outcome and able to distinguish a small important change. The precision of the resulting estimate also matters. Retrospective change ratings can be influenced by current health rather than accurately reflecting change. These issues are addressed in an anchor-based MID credibility study.

Distribution-based calculations use statistical properties such as score variability or measurement error. They can provide supporting context, but do not by themselves establish what participants value. A minimum detectable change addresses whether a change exceeds expected measurement error; an important difference addresses its meaning. Neither concept automatically substitutes for the other.

Human evidence and its boundaries

Perera and colleagues studied meaningful change in physical performance using data from older adults with mobility disabilities, people recovering from stroke and community-dwelling older people. This is human measurement research, rather than evidence that a particular intervention extends healthy lifespan.

Before transferring a published threshold to a new trial, compare baseline function, health conditions, assessment protocol and follow-up duration. A threshold developed in rehabilitation may not describe what matters to initially high-functioning adults. Improvement and deterioration also require separate justification.

Animal survival experiments and laboratory findings can inform biological hypotheses. They cannot establish what constitutes a meaningful improvement in a human participant’s daily life. Likewise, a change in a biological ageing marker should not be labelled clinically important solely because it is statistically significant; that interpretation requires evidence connecting the marker to outcomes people value.

Read the treatment comparison and its uncertainty

A change within the treatment group does not establish a treatment effect. Examine the comparison with the control group, its confidence interval and the outcome’s direction of benefit. Keep an individual change threshold distinct from a threshold intended to interpret a difference between group averages.

Consider a fictional trial using a 0–100 function scale, with higher scores indicating better function. Suppose a five-point between-group difference has been justified in advance as important. An estimated benefit of three points with a 95% confidence interval from one to six excludes zero but leaves uncertainty about whether the average benefit reaches five points. This example illustrates interpretation; it is not a published result or a validated threshold.

A responder analysis instead compares the proportions of participants meeting a specified individual-change criterion. It needs justification for that criterion and attention to misclassification and measurement error. Research on responder-analysis bias explains why categorising participants does not remove these difficulties.

Questions to ask when reading a trial

Was the important-difference threshold specified before results were examined, with an identifiable supporting study?
Does its derivation match the trial’s population, instrument, direction of change and follow-up period?
Is it intended for individual change or for a comparison between groups?
Do the effect estimate and confidence interval support a meaningful benefit, or leave that question unresolved?
Are conclusions robust to other credible threshold estimates?
Are harms, treatment burden, missing outcomes and persistence of benefit reported alongside the benefit?

Treat these as appraisal questions rather than a numerical pass/fail rule. Exceeding an outcome threshold alone does not establish an intervention’s overall value or demonstrate longer healthspan.