Why Do Coffee and Sleep Studies Disagree?

Research topic and question

• Why Do Coffee and Sleep Studies Disagree?

Search the latest human studies on caffeine and sleep using PubMed (https://pubmed.ncbi.nlm.nih.gov/) and available full texts in PubMed Central (https://pmc.ncbi.nlm.nih.gov/). Why do studies reach different conclusions? Compare caffeine dose, timing, participants and how sleep was measured.

Research findings

Coffee-and-sleep studies mostly disagree because they study different exposures, people, and definitions of “sleep.” The overall human evidence is more consistent than headlines suggest: high or late caffeine doses impair sleep; small, sufficiently early doses may produce little detectable change in healthy habitual users.

The clearest comparison

Study	Caffeine exposure	Participants	Sleep measurement	Result
Gardiner et al., 2025	100 or 400 mg, 12/8/4 hours before bed	23 healthy men, mean age 25	In-home partial polysomnography and diaries	No significant effect from 100 mg; 400 mg affected sleep initiation and architecture even 12 hours before bed. Participants perceived poorer sleep mainly at 4 hours. Full text
Drake et al., 2013	400 mg at bedtime or 3/6 hours beforehand	12 healthy adults	Portable objective monitor and diary	Sleep was disrupted at every timing, including 6 hours before bed.
Pauchon et al., 2024	2.5 mg/kg twice during 38 hours awake; last dose 6.5 hours before recovery sleep	41 adults	Headband polysomnography	Recovery sleep fell by about 30 minutes and N3 sleep by about 36 minutes; awakenings and stage transitions increased.
Baur et al., 2024	160 mg delayed-release caffeine	21 healthy young men	Standard polysomnography, plasma caffeine and EEG spectra	Higher plasma concentrations reduced NREM delta activity. Concentrations varied substantially between individuals given the same dose.
Stucky et al., 2025	≥4 versus ≤3 caffeinated drinks/day	UK Biobank: 485,511; HypnoLaus: 1,702	At-home full polysomnography, questionnaires and genetic causal-inference methods	High habitual use was associated with shorter objective sleep, but not worse self-rated sleep. Effect estimates varied implausibly widely, from 11 to 229 minutes. Full text
Cusick et al., 2020	Morning, afternoon or evening drinks	302 adolescents, with and without ADHD	Actigraphy plus adolescent and parent reports	Afternoon/evening use related to reported problems, with differences by ADHD status; objective and subjective measures did not always agree.
Watson et al., 2016	Habitual intake estimated by questionnaire	80 adults	Pittsburgh Sleep Quality Index	Poor sleepers consumed more caffeine, but most individual sleep components showed no association. Causality could not be determined.

A 24-study meta-analysis found average reductions of roughly 45 minutes in total sleep and 7% in sleep efficiency, alongside nine minutes longer to fall asleep, 12 minutes more wakefulness after sleep onset, and less deep sleep (Gardiner et al., 2023). But those averages combine widely different doses and protocols.

Why the findings differ

Dose: “A cup of coffee” is not standardized. Trials use known doses such as 100 or 400 mg; population studies count cups whose caffeine content can differ several-fold. A fixed dose also represents different mg/kg exposures across body sizes.

Timing: Clock time is less informative than hours before each person’s bedtime. Because caffeine clearance varies considerably, an afternoon dose may be largely cleared in one participant but remain biologically active in another. Repeated doses also behave differently from one bolus.

Participants: Controlled trials often enroll small samples of healthy young men. These results may not generalize to women, adolescents, older adults, shift workers, pregnant people, people with insomnia or ADHD, or medication users. Genetics, smoking, oral contraceptives and habitual exposure can all change caffeine clearance or sensitivity.

Reverse causation: In cross-sectional studies, caffeine may cause poor sleep—but poor sleep also causes people to consume more caffeine the following day. Stress, workload, alcohol, smoking and chronotype affect both behaviors.

Measurement: Polysomnography measures sleep stages, awakenings and EEG activity. Actigraphy infers sleep from movement. Diaries and questionnaires measure perceived sleep. A person can report sleeping normally while polysomnography detects delayed onset, fragmentation or reduced deep-sleep activity. Thus “no change in sleep quality” may mean only that participants did not notice the physiological change.

The practical conclusion is therefore conditional: 100 mg taken at least four hours before bed may have little measurable effect in a healthy moderate user, whereas 400 mg can disrupt objective sleep even 6–12 hours before bedtime. Individual sensitivity remains substantial, and subjective sleep quality can underestimate the effect.