General health checks did not reduce deaths across 17 randomised trials

The randomised evidence on general health checks — the annual or periodic physical, screening for more than one disease or risk factor across more than one organ system — finds no effect on death. A 2019 Cochrane review pooling 17 trials, 15 of which reported outcome data covering 251,891 participants, found a risk ratio for total mortality of 1.00 (95% CI 0.97 to 1.03), graded high-certainty evidence, across 11 trials, 233,298 participants and 21,535 deaths [s1].

The reviewers' one-line conclusion is that general health checks are unlikely to be beneficial [s1]. That is a narrower statement than it sounds, and the boundaries of it matter more than the headline.

What was measured

The review defined health checks as screening for more than one disease or risk factor in more than one organ system, and included randomised trials comparing health checks with no health checks in adults unselected for disease or risk factors; geriatric trials were excluded [s1]. Databases were searched to 31 January 2018 [s1]. Risk of bias was generally low for the primary outcomes [s1].

On cancer mortality the risk ratio was 1.01 (95% CI 0.92 to 1.12) across 8 trials, 139,290 participants and 3,663 deaths, also high-certainty [s1]. On cardiovascular mortality it was 1.05 (95% CI 0.94 to 1.16) across 9 trials, 170,227 participants and 6,237 deaths, graded moderate-certainty with substantial heterogeneity (I² = 65%) [s1].

Non-fatal outcomes went the same way. For fatal and non-fatal ischaemic heart disease the risk ratio was 0.98 (95% CI 0.94 to 1.03) across 4 trials, 164,881 people and 10,325 events, high-certainty [s1]. For fatal and non-fatal stroke it was 1.05 (95% CI 0.95 to 1.17) across 3 trials, 107,421 people and 4,543 events, moderate-certainty [s1].

Five null results, three of them at high certainty, across a quarter of a million randomised participants. This is not an absence of evidence.

The trial that shows why

The largest individual demonstration of the same thing is Inter99, a randomised community-based trial in the suburbs of Copenhagen [s2]. It randomised 59,616 people aged 30 to 60 — 11,629 to intervention and 47,987 to control — and invited the intervention group for screening, risk assessment and lifestyle counselling up to four times over five years [s2].

The intervention was not perfunctory. Everyone with an unhealthy lifestyle received individually tailored counselling at each visit; those at high risk of ischaemic heart disease by predefined criteria were additionally offered six sessions of group-based counselling on smoking cessation, diet and physical activity, and participants were referred to their GP for medical treatment where relevant [s2].

Over 10 years of follow-up, 3,163 people died, 2,782 developed ischaemic heart disease and 1,726 developed stroke [s2]. The hazard ratio for the primary endpoint, ischaemic heart disease, was 1.03 (95% CI 0.94 to 1.13) [s2]. Stroke was 0.98 (95% CI 0.87 to 1.11), the combined endpoint 1.01 (95% CI 0.93 to 1.09) and total mortality 1.00 (95% CI 0.91 to 1.09) [s2].

One number explains part of it. Only 6,091 people — 52.4% of those invited — participated at baseline [s2]. Uptake of general health checks skews toward people already engaged with their health, which both dilutes the population effect and means the trials measure the offer of a check, not the check itself.

What this does not say

The review's scope is specific and the conclusion does not generalise past it.

It is not a finding about individual screening tests. The review's own background notes that most of the individual screening tests offered within general health checks have been incompletely studied [s1] — which is a statement about the state of the evidence, not an endorsement or a dismissal of any particular test. Colorectal cancer screening, cervical screening and blood pressure measurement each have their own trial literature and their own guideline recommendations, evaluated separately.

It is not a finding about seeing a doctor when something is wrong. Every trial here randomised asymptomatic adults to an invitation for systematic screening, not to access to care.

And it is not a finding about the components in isolation. Inter99 delivered a substantial lifestyle-counselling programme and found nothing at the population level after 10 years [s2]; that is a result about a programme delivered to a general population with roughly half taking it up, not about whether counselling can help a given person.

The reviewers also flag the mechanism by which a null result can conceal harm: screening leads to increased use of diagnostic and therapeutic interventions, which can be harmful as well as beneficial, which is precisely why they judged it important to test whether health checks do more good than harm [s1]. A risk ratio of exactly 1.00 for total mortality is consistent with benefits and harms cancelling out as well as with nothing happening at all.

What any individual should do about a routine appointment depends on their age, risk factors, existing conditions and which specific tests are on offer — none of which this literature addresses. It addresses the practice of checking healthy adults comprehensively and on a schedule, and on that question the answer is unusually well established.