The evidence that physical activity improves health is among the most consistent in medicine, and it is also among the hardest to generate well. Most evidence on lifetime activity and long-term disease outcomes is observational, although randomised trials provide important causal evidence for shorter-term clinical and physiological outcomes. The reasons are practical: participants cannot be blinded to whether they went for a run, exposure is measured imprecisely, the people who exercise differ systematically from those who do not, and the outcomes that matter most accrue over decades. A sport and exercise medicine (SEM) clinician therefore needs to know not only what the guidelines say but why the underlying evidence looks the way it does, where it is strong and where it is weak. This page covers the study designs used in physical activity research, the measurement problem that runs through all of it, the main biases to look for, and how the current UK recommendations sit on top of that evidence.
What kinds of evidence exist?
Four designs supply most of what is known. Cross-sectional studies measure activity and health at the same moment; they are cheap and generate hypotheses but cannot establish direction, since being unwell reduces activity just as inactivity worsens health. Prospective cohort studies measure activity first and follow people for years or decades. Measuring activity before the outcome establishes temporality and provides most of the dose-response evidence behind national guidelines, although reverse causation from preclinical illness can still remain, and active people may differ in diet, smoking, income and education in ways that also protect them. Randomised controlled trials establish causation within the trial and are the mainstay for exercise as a treatment in specific conditions, but they are usually short and modest in size, participants cannot usually be blinded even though outcome assessors and analysts sometimes can, and adherence, crossover and contamination of the control arm are major problems. Natural experiments evaluate a real-world change such as a new cycle network or a workplace scheme, where exposure is determined by a policy, environmental or service change rather than allocated by the researcher.
Physical activity evidence comes from cross-sectional studies, prospective cohorts, randomised trials and natural experiments, each answering a different question and carrying a different dominant weakness.
Two biases recur and are worth naming precisely. Reverse causation is the tendency for early, undiagnosed illness to reduce activity, making inactivity look like the cause of a disease it is actually an early symptom of; cohort studies address it by excluding events in the first few years of follow-up. Residual confounding is the portion of the association left over after adjustment, because activity clusters with almost every other health-promoting behaviour and with socioeconomic advantage. Neither bias means the evidence is wrong. Reverse causation commonly inflates apparent benefit, while residual confounding can bias estimates in either direction, and the direction of effect is supported by trial evidence for intermediate and clinical outcomes such as blood pressure, glycaemic control, fitness and mood.
How is activity measured, and what do the guidelines say?
Measurement is the weak link running through the whole field. Self-report questionnaires are cheap and scalable and underpin national surveillance, but depending on the questionnaire and the activity domain they can either overestimate or underestimate activity, and they struggle particularly with light, incidental and intermittent movement. Device-based measurement using accelerometers is more objective and captures light activity and sedentary time that questionnaires miss, but devices are not worn continuously, may undercapture cycling, resistance exercise, load carrying and water-based activity depending on device location, waterproofing and processing method, may misclassify standing as sedentary unless posture-sensitive devices are used, and produce different results from identical raw data depending on the processing thresholds applied. The practical consequence is that prevalence estimates vary considerably depending on method, and that comparisons between studies using different measures should be made cautiously. It also explains why guidelines are best understood as population-level targets rather than precise individual exercise prescriptions.
Self-report is cheap and underpins national surveillance but may over or underestimate activity, while device-based measurement captures light and sedentary time yet may undercapture cycling and water-based activity and varies with processing choices.
The current UK Chief Medical Officers' guidelines summarise this evidence into recommendations that are worth knowing exactly. Adults should accumulate at least one hundred and fifty minutes of moderate intensity activity each week, or seventy-five minutes of vigorous intensity activity, or even shorter durations of very vigorous intensity activity such as sprinting or stair climbing, or a combination. Current evidence indicates this can be accumulated in bouts of any length, and can be achieved in one or two sessions per week while still conferring benefit. All adults should undertake muscle strengthening activity on at least two days each week, while recognising that any strengthening activity is better than none. Sedentary time should be minimised and long periods of inactivity broken up with at least light activity, which has measurable benefit in its own right. Separate guidelines exist for early years, children and young people, older adults, disabled adults, pregnancy and the period after childbirth. Two framing messages matter as much as the numbers: benefit begins below the threshold, so some activity is substantially better than none, particularly for those who are least fit, and the guidelines are periodically revised as evidence accumulates, so the current version should be checked rather than recalled.
Exam Tips
•Most evidence on lifetime activity and long-term outcomes is observational, while randomised trials provide causal evidence for shorter-term clinical and physiological outcomes.
•Prospective cohorts supply most of the dose-response evidence behind national guidelines; randomised trials dominate where exercise is used as treatment for a specific condition.
•Reverse causation means early undiagnosed illness reduces activity, making inactivity look causal; excluding early events reduces but does not eliminate its influence.
•Reverse causation commonly inflates apparent benefit, while residual confounding can bias estimates in either direction, since activity clusters with other healthy behaviours and socioeconomic advantage.
•Self-report may over or underestimate activity and struggles with incidental movement, while accelerometers may undercapture cycling, resistance work and water-based activity and vary by processing threshold.
•Current UK recommendations are at least one hundred and fifty minutes of moderate activity, seventy-five minutes of vigorous, or shorter durations of very vigorous activity weekly, strengthening on at least two days, and minimising sedentary time, with light activity and any amount of movement carrying benefit.