Twenty years of night shifts, and a 32% higher risk of heart, kidney or metabolic disease

About one worker in five does night shifts [s1]. A study published in the European Journal of Preventive Cardiology on 18 December used the UK Biobank to ask what a working life of them adds up to — and whether sleeping habits and body clock make any difference to the answer [s1].

The design

The analysis covered 96,365 UK Biobank participants with employment-history data — 57% female, mean age 62 years — followed for incident cardiovascular-kidney-metabolic disease through 2022 [s1]. CKM was treated as a composite outcome: cardiovascular disease, chronic kidney disease, and/or type 2 diabetes [s1]. Cox proportional hazards models were adjusted for a set of potential confounders including sex, age, ethnicity, and socioeconomic factors [s1].

There were 5,967 incident cases [s1].

The results

The relationship was dose-dependent, and it showed up across three different ways of counting exposure [s1]:

• Duration: more than 20 years of night-shift work, versus day work — HR 1.32 (95% CI, 1.20–1.45)
• Frequency: 8 or more night shifts per month, versus day work — HR 1.32 (95% CI, 1.21–1.43)
• Cumulative total: more than 1,200 lifetime night shifts, versus day work — HR 1.36 (95% CI, 1.25–1.48)

All three showed a significant trend across exposure levels [s1].

The authors report that the association was further amplified by short sleep [s1] — meaning that among night-shift workers, those who also slept short carried more risk than those who did not.

Why the composite outcome is a deliberate choice

Cardiovascular disease, chronic kidney disease, and type 2 diabetes have historically been studied as separate destinations. The cardiovascular-kidney-metabolic framing treats them as manifestations of one interlinked process, and the authors argue explicitly that future research should adopt CKM as a composite outcome [s1].

There is a methodological argument for it and a cost. The argument: if a single exposure raises the risk of all three, studying them separately splits the signal three ways and underpowers each analysis. The cost: a hazard ratio for a composite tells you the bundle moved without telling you which component drove it. A 32% increase in "CKM disease" could be mostly diabetes, mostly kidney disease, or evenly spread, and the headline figure conceals which.

What this is not

It is observational. Night-shift work is not randomly assigned, and it correlates with a great deal else — occupation, income, education, physical activity, diet timing, smoking, and access to care. The models adjust for socioeconomic factors [s1], but adjustment is not randomisation, and residual confounding by the character of shift-working jobs themselves is the standing objection to this entire literature.

Exposure was reconstructed from employment history, which depends on recall over a working lifetime. Misclassification here is likely and would generally, though not always, push estimates toward the null.

The UK Biobank is also a well-documented non-representative cohort: volunteers who are healthier, wealthier, and less ethnically diverse than the UK population. Mean age at analysis was 62 [s1], meaning the sample is weighted toward people who survived a career of shift work in good enough health to enrol. That selection tends to make hazard ratios conservative rather than inflated — but it also means these numbers describe a particular kind of survivor.

Finally, hazard ratios in the 1.3 range from observational data are modest. They are consistent with a real effect; they are also within the range that unmeasured confounding can produce.

What is genuinely useful here

The dose-response pattern. Three independent measures of exposure — years worked, shifts per month, and lifetime shift count — all point the same way with overlapping confidence intervals [s1]. Consistency across differently-constructed exposure metrics is one of the few things observational epidemiology can offer that partially substitutes for randomisation. It does not prove causation. It does make a purely confounded explanation work harder.

The short-sleep interaction is also actionable in a way the main effect is not. Shift schedules are set by employers and industries; sleep opportunity between shifts is, at least partly, a design variable in those schedules.

What to watch

Whether the component outcomes get reported separately, which would show what is actually driving the composite. Whether similar dose-response patterns appear in cohorts more representative than the UK Biobank. And whether any intervention study — on schedule rotation direction, shift length, or protected recovery sleep — shows that changing the exposure changes the outcome. Until one does, this remains a well-characterised association.

Sources

1. Night shift work and its interaction with sleep duration and chronotype, and risk of cardiovascular-kidney-metabolic diseases in the UK Biobank — European Journal of Preventive Cardiology, 18 December 2025