Night Shift Sleep: What Actually Works

Five consecutive simulated night shifts, 11 p.m. to 7 a.m. Home to bed by 8:30 in the morning, up at 3:30 in the afternoon, bedroom darkened. Sunglasses for any daylight outdoors, issued in two strengths: lenses that pass 15 percent of visible light, or lenses that pass 2 percent. A capsule before the daytime sleep, either 1.8 milligrams of sustained-release melatonin or a placebo. And on the shift itself, for some participants, bright light at roughly 5,000 lux switched on for 20 minutes and off for 40, four or five pulses a night, the pattern drifting later as the week went on; for everyone else, ordinary room light of about 150 lux. Sixty-seven volunteers, median age 22, none of them shift workers.

That is the equipment list from a 2003 laboratory study that remains one of the clearest tests of what a person can actually do to make a night schedule livable. It is worth reading the list twice, because nothing on it is exotic. Lamps, sunglasses, a dark bedroom, a fixed bedtime. The expensive-sounding item, melatonin, turned out to be the least important.

The question this article answers is the one the roughly one in ten night and rotating workers who meet criteria for shift work sleep disorder ask, and the many more who do not meet criteria but still feel wrecked: when your job runs against your clock, what actually helps? The most defensible answer from the evidence is this. The body clock of most night workers never fully adapts, and full adaptation is not the goal anyway. Partial realignment, produced by timed light, timed darkness and a protected sleep window, is achievable and measurably improves alertness, mood and performance on the shift. A short nap, caffeine early rather than late, and meals kept to the daytime reduce the cost of the shift by other routes. Melatonin adds minutes of sleep, not hours. And the long-term associations with heart disease and diabetes are real but modest, and appear mostly after years, which means the habits below matter more the longer you plan to keep doing this.

The problem is the clock, not the pillow

Start with the uncomfortable finding. A 2008 review pooled six studies that measured the melatonin rhythm, the best available marker of the internal clock, in people on permanent night shifts. Fewer than 3 percent showed complete adjustment of that rhythm to night work, and fewer than one in four adjusted enough to get any benefit from it. It made no difference whether the studies were done under normal or dim lighting, and men and women adjusted equally poorly. Permanent nights, in other words, do not quietly fix the problem by repetition. In ordinary environments the clock stays roughly on daytime, and the worker keeps trying to sleep when the clock says wake.

That mismatch is why the symptoms cluster the way they do. A 2003 occupational-medicine review described the most troublesome acute problems as difficulty getting to sleep, shortened sleep, and sleepiness during working hours that carries into the days off, and noted they are only partly fixable by rearranging the roster. The same review found no clear indication that chronic sleep problems result from long-term shift work, which is a rare piece of good news in this literature.

When the symptoms are severe enough to meet criteria, the label is shift work sleep disorder. A Detroit population survey of 2,570 working adults, including 360 rotating-shift and 174 night workers, put its prevalence at approximately 10 percent of the night and rotating population. Workers who met criteria had higher rates of ulcers (odds ratio 4.18), sleepiness-related accidents, absenteeism, depression and missed family and social activities than shift workers who did not, and in most cases their morbidity was worse than that of day workers with the identical symptoms. The schedule appears to amplify what insomnia and sleepiness already do.

What the health numbers actually say

Two kinds of long-term evidence get quoted at shift workers, usually without the denominators. Here they are with the denominators.

A 2018 meta-analysis of 21 cohort and case-control studies covering 173,010 people found the risk of any cardiovascular event was 17 percent higher among shift workers than day workers, and coronary heart disease morbidity was 26 percent higher. The dose-response analysis is the part worth remembering: the association was non-linear and appeared only after the first five years of shift work, after which every additional five years carried a further 7.1 percent increase in risk. A 2016 review in the BMJ, built on 38 meta-analyses and 24 systematic reviews, put the relative risks at 1.23 for coronary heart disease, 1.05 for stroke and 1.09 to 1.40 for type 2 diabetes, with the original studies showing mixed results. Its central observation is the useful one: the relations of shift work to cardiometabolic disease and to accidents mimic those of insufficient sleep, so the two probably share mechanisms, though whether short sleep is the causal pathway has not been settled.

For diabetes specifically, a 2015 meta-analysis of 12 studies with 226,652 participants found a pooled odds ratio of 1.09 for ever having worked shifts, higher in men (1.37) than women (1.09), and higher for rotating schedules. The largest single cohort, 74,862 U.S. nurses followed for 22 years, found all-cause mortality raised by 11 percent in women with 6 to 14 years or 15 or more years of rotating night work, and cardiovascular mortality raised by 19 to 23 percent; the increases appeared in women with five or more years of rotating nights, and there was no association with cancer mortality overall, lung cancer at 15 or more years being the one exception.

All of this is observational. Shift workers differ from day workers in income, smoking, diet and the jobs themselves, and adjustment never removes all of that. Relative risks of 1.1 to 1.3 on common diseases are not trivial across a career, but they are also not the catastrophe a headline implies, and the five-year threshold suggests that what accumulates is exposure. The practical reading is that a night worker's cardiometabolic risk is worth tracking the way anyone's is, with the ordinary markers in the Biomarker Tracker, and that reducing the strain of each shift is the lever within reach.

Light is the lever

The clock is set mostly by light reaching the eye, which is why the interventions with the strongest physiological evidence all manipulate when a worker sees light and when they see darkness.

The field test that demonstrated this used real nurses. A group of permanent night-shift nurses, averaging about 42 years old, spent at least ten days on a normal daytime schedule, had their circadian phase measured in the laboratory, then worked roughly 12 night shifts. The treatment group got six hours of intermittent bright light in the workplace at about 3,243 lux and wore tinted goggles passing 15 percent of visible light to shield themselves from bright morning light on the way home; the controls worked in their usual environment. Everyone kept the same rule for sleep, a single eight-hour sleep-and-darkness episode beginning two hours after the end of the shift. After the run of shifts, the treated group's core temperature and melatonin rhythms had delayed by an average of 9.32 and 11.31 hours, versus 4.09 and 5.08 hours in the controls, and only the treated group's rhythms had re-established their normal relationship to the new sleep time.

The laboratory study that opened this article then dissected which pieces mattered. Among the volunteers whose baseline clocks ran early, so that the morning commute exposed them to sunlight at the worst possible moment, the outcome depended on the intervention: with bright light during the shift, almost all of them completely re-entrained, and the shift was so large that darker sunglasses and melatonin could add nothing to it. With only room light on the shift, the darker sunglasses helped them delay more than normal sunglasses did, and melatonin again added nothing to the phase shift. Participants whose clocks already ran late re-entrained regardless of group. The authors' recommendation, if the goal is complete adaptation, was intermittent bright light during the night shift, sunglasses as dark as possible on the commute home, and a regular, early daytime dark-and-sleep period.

Two cautions before anyone buys lamps. First, the pooled evidence in real workplaces is weaker than the physiology. A 2020 meta-analysis of five trials in shift-working nurses found bright light reduced sleepiness and shift-work complaints in a fixed-effect model but nothing reached significance in the random-effects model, with heterogeneity above 90 percent. The 2016 Cochrane review of non-drug interventions, 17 trials and 556 relevant participants, pooled two comparable bright-light studies and found sleepiness on the Stanford scale lower by 0.83 points, graded very low quality, and concluded overall that it could not determine whether these interventions reduce sleepiness or improve sleep. Second, the wavelength shortcut does not seem to work on its own. Two crossover studies in police patrol officers on fast-rotating nights tested dim blue light in the patrol car with blue-blocking glasses after 5 a.m. against a red-light version and against no intervention. In the winter study the officers' clocks delayed by roughly two hours across four consecutive nights in every condition; in the summer study the no-intervention condition actually delayed more, about three hours against two, and in neither season did alertness or sleepiness differ between conditions. The authors' conclusion was that four nights is simply not enough for partial adaptation to develop, whatever you wear. Our own review of blue-light management products reached a similar place: the mechanism is real, the average product benefit is not.

You do not need full adaptation

The most reassuring result in this field is that partial realignment is enough. In the 67-volunteer study, participants were grouped afterwards by how far their temperature minimum had moved: not at all, into the first half of the daytime sleep, or into the second half. Performance, sleepiness and mood on the night shift were better in both re-entrained groups than in the unadapted one, with no meaningful difference between partial and complete. And the unadapted group had slept almost the full seven hours allowed during the day. A reasonable amount of daytime sleep, the authors wrote, is not enough on its own to produce good performance at night; some realignment of the clock is needed.

That finding became a deliberate strategy. Because full adaptation would leave a worker out of phase with the daytime world on days off, a research series set out to engineer a compromise: delay the clock only far enough that the sleepiest circadian hour moves out of the shift and into the first half of the daytime sleep. In the final study of that series, 19 healthy adults worked three simulated nights, took two days off, worked four more nights and took two more off. The experimental group got four 15-minute bright-light pulses per shift, wore dark sunglasses outdoors, slept in dark bedrooms at scheduled times and, importantly, got outdoor afternoon light on purpose, a light brake to stop the clock from drifting too far. Their melatonin onset landed at 3:22 a.m., close to the 3:00 target that puts the sleepiest time near 10 a.m., against 11:24 p.m. in controls. They slept for nearly all of their permitted time in bed, performed better on the shifts, and could still sleep at night on days off. In the companion study, the 21 partially and six completely re-entrained subjects had mood, fatigue and performance close to daytime levels after about a week, while the 12 who had not re-entrained did not.

The lesson for a real worker is that the target is modest and specific: push the low point of the clock past the end of the shift, not all the way into the afternoon, and keep it there with afternoon light.

The daytime sleep window

Every protocol above shares one non-negotiable: a fixed, dark sleep episode at the same clock time after each night. The physiological reason dates to isolation-unit experiments in 1981. Volunteers on irregular sleep routines, whether a single randomly timed eight-hour sleep or two random four-hour sleeps, drifted onto free-running rhythms longer than 24 hours even though meals stayed at customary times. When one of the four-hour sleeps was anchored at the same time every day, with the other still random, the rhythms stabilized at 24 hours within a few days, and the phase shift showed it was the sleep time, not the mealtimes, doing the synchronizing. The practical translation, which the authors themselves raised for shift workers, is that if a night worker cannot protect a full sleep, protecting the same four hours every day, including days off, is what keeps the clock from wandering.

Darkness is the other half. The 2022 expert consensus on indoor light, expressed in the melanopic units the circadian system actually responds to, recommends a sleep environment at or below one lux of melanopic daylight-equivalent illuminance at the eye, with a ceiling of ten lux in the evening before bed and a floor of 250 lux during the day. A daytime bedroom, with sun at the curtain edges, is nowhere near one lux without blackout material or a mask. Note that the same document explicitly says its recommendations were not designed for night shift workers and that light advice for that group is not yet mature; we cite it for the darkness target, not as a shift protocol. Noise is the remaining variable, and it is worth reading our white-noise review before spending money: the trial evidence for masking is very low quality, earplugs beat masking against traffic noise in one experiment, and masking is best reserved for a genuinely noisy room at low volume.

Naps: small, timed, and worth the grogginess

A 2014 systematic review found 13 studies, mostly small, of planned naps taken during real or simulated night shifts. Most found that despite a short period of sleep inertia after waking, the nap reduced sleepiness and improved performance; none measured workplace safety outcomes, and the authors called for larger trials before wide implementation. The larger trials have started to arrive. In a within-subjects study of 109 female hospital nurses, each tested on two nights with a scheduled 30-minute nap at 4 a.m. and two without, sleepiness was lower at 5, 6 and 7 a.m. on nap nights and the gain in performance between 3 and 7 a.m. was larger on two cognitive tasks, with no interaction with chronotype, sleep quality or any other individual factor measured.

The inertia is real but brief. A 2023 crossover trial in 28 emergency-medical workers on simulated 12-hour nights compared no nap, a 30-minute nap and a two-hour nap at 2 a.m. Reaction-time performance was worse immediately on waking from both naps than just before them, but the deficit was gone by 10 to 30 minutes, and only the two-hour nap prevented the end-of-shift decline in speed and false starts. The Cochrane pooled estimate for a single nap opportunity, an 11.87-millisecond reduction in mean reaction time rated very low quality, shows how thin the randomized base still is; the direction is consistent, the size is uncertain. The design rule that falls out of the trials is not to wake into a critical task: build ten to thirty minutes between the end of the nap and anything that requires sharp judgment.

Caffeine: early in the shift, then stop

Caffeine is the intervention most night workers already use, usually in the wrong half of the shift. A 2010 Cochrane review found 13 randomized trials in jet lag or shift work disorder; none measured injuries, two measured errors and both found caffeine reduced them, and pooled cognitive results favored caffeine for memory, orientation and attention, with a high risk of bias. The 2014 Cochrane review of drug interventions adds one trial in which caffeine plus a nap taken before the night shift lowered sleepiness by 0.63 points on the Karolinska scale. The problem is the tail. In a home study, 400 milligrams of caffeine taken at bedtime, three hours before or six hours before bed each significantly disrupted objectively measured sleep compared with placebo, and the authors concluded that the six-hour dose still cut total sleep time enough to matter. For a worker who plans to be asleep by 9 a.m., that puts the last meaningful caffeine at about 3 a.m., which is precisely when most people reach for it. Front-load it, pair it with the pre-shift nap if you can, and treat the second half of the shift as a caffeine-free zone.

Melatonin: minutes, not hours

Melatonin is the supplement with the most trials in this population, and the trials are clear about the size of what it does. The 2014 Cochrane review pooled seven trials with 263 participants and found melatonin at 1 to 10 milligrams taken after the night shift lengthened daytime sleep by a mean of 24 minutes, with a confidence interval of 9.8 to 38.9 minutes, and lengthened night-time sleep by 17 minutes in three trials. There was no dose-response, so 10 milligrams did no more than 1. It did not shorten the time to fall asleep, and the evidence was graded low quality. A double-blind crossover trial of 1.8 milligrams sustained-release taken half an hour before two consecutive daytime sleeps found it prevented the usual drop in sleep time only on the first day, helped most in people who had trouble sleeping in the day, and did nothing for alertness, performance or mood on the night shift, with no hangover.

So the honest summary is that melatonin can buy a modest amount of daytime sleep, especially at first and especially for people who struggle to fall asleep in daylight, and that it does not fix the night-shift performance problem because that problem is the clock, not the sleep total. Our separate review of melatonin covers why high-dose anti-aging claims outrun the evidence and why long-term safety is incompletely defined. The same Cochrane review found that a hypnotic, zopiclone, did not lengthen daytime sleep in one low-quality trial, and that the prescription wake-promoting drugs armodafinil and modafinil reduced sleepiness by about one point on a ten-point scale in people with diagnosed shift work sleep disorder, at the cost of headache, nausea and a rise in blood pressure, with severe skin reactions reported after marketing. Those are decisions for a worker and a clinician, not for an article.

Eat in the daytime

The newest and in some ways most striking evidence concerns not sleep but food. In a 14-day laboratory protocol, volunteers went through simulated night work eating either at night or only during the day, with their central clock tracked by body temperature and their glucose and insulin rhythms measured under constant conditions before and after. Night-time eating produced misalignment between the central and the glucose rhythms and impaired glucose tolerance; restricting meals to the daytime prevented both. A 2017 pilot in 11 healthy men, four eating a meal at 1:30 a.m. and seven not eating overnight, found the post-breakfast glucose response rose across four simulated nights only in the night-eating group. And a 2025 cluster-randomized trial that assigned 55 healthy adults to a full meal, a snack or fasting during four simulated night shifts found that night work impaired insulin sensitivity in every group, but the change in glucose tolerance was +2.00 in the meal group and +0.96 in the snack group against +0.34 in the fasting group, in the trial's area-under-the-curve units.

These are short laboratory studies in healthy volunteers, the 2025 trial in young non-shift-workers and the others in small healthy samples, and none of them measured diabetes. What they show is a mechanism and a direction that is consistent across three independent designs: the metabolic cost of a night shift is smaller when the calories arrive in the daytime. For a worker, that means a real meal before the shift, a real meal after waking, and water, or at most a small, planned snack, in the small hours.

Chronotype and the roster

Not everyone pays the same price for the same shift. In 238 shift workers assessed with a chronotype questionnaire, earlier types slept less on night shifts, carried more social jet lag and reported more sleep disturbance, while later types had the same pattern on early-morning shifts, and workers on fast-rotating schedules slept longer overall. A factory study then acted on this, removing night shifts for the earliest chronotypes and morning shifts for the latest. Self-reported sleep duration and quality and workday wellbeing improved among those extreme types, and social jet lag across the workforce fell by an hour, with no change in stress. How the roster is built matters too; the classic 1982 demonstration that schedules designed on circadian principles improve satisfaction, subjective health, turnover and productivity is four decades old and still not universally applied. A worker cannot always change the roster, but a worker who is a strong morning type on permanent nights should know that the schedule itself is a large part of the problem.

The drive home

The single most dangerous hour of a night shift may be the one after it ends. Sixteen night-shift workers drove a real vehicle for two hours on a closed track twice: once after an ordinary night's sleep averaging 7.6 hours, and once straight after a night shift. After the shift, 11 near-crash events occurred in six of the 16 drives, and seven of the 16 drives were stopped early for safety. After sleep there were none of either. Lane excursions doubled, and blink duration and slow eye movements, the physiological markers of drowsiness, rose significantly. A 2020 systematic review of two on-road and nine simulator studies rated overnight shift work a possible predictor of adverse on-road outcomes and a likely predictor of adverse simulator outcomes, with low to moderate confidence overall, and called for better real-world testing. The abstracts do not test countermeasures for the commute, so the following is an inference rather than a trial result: the same dark sunglasses that protect the clock also reduce glare, and a worker who is fighting to keep their eyes open at the wheel has, by the closed-track data, roughly a one-in-three chance of a near-crash in the next two hours. That is the arithmetic that justifies a short nap before the drive or a different way home.

When it is insomnia, not just the schedule

Some night workers have a sleep disorder that would exist on any schedule, and for them the tools above are not enough. A 2024 pilot randomized trial in 46 nurses with shift work disorder tested a guided digital cognitive behavioral therapy program adapted to shift work against a waiting list. Insomnia severity fell significantly more in the treatment group, along with sleepiness, dysfunctional beliefs about sleep, pre-sleep arousal and mood, though actigraphy showed no objective change. A 2020 partially randomized occupational-health trial in 83 shift workers with insomnia found that group CBT-I, self-help CBT-I and a sleep-hygiene control all improved insomnia, restedness and actigraphy-measured sleep, with no differences between them, and that mood improved only in the group format. The practical point is that persistent insomnia in a shift worker is treatable with structured therapy; the trial could not separate the three approaches, which does not by itself establish that sleep hygiene is active, and the participants with shift work disorder improved less than those without it.

The protocol

What follows assembles the interventions above into a single night, for a worker on consecutive night shifts who wants partial realignment. Every element traces to a study cited in the source list; the timings assume a shift ending around 7 a.m. and should be shifted with the shift.

What to skip

Blue-blocking glasses as the whole strategy, because the police studies found no faster adaptation and no better alertness with them across four nights. Melatonin above the low milligrams, because seven trials found no dose-response. Caffeine in the second half of the shift, because 400 milligrams six hours before bed still cut objectively measured sleep. A full meal in the small hours, on three laboratory studies pointing the same way. And the expectation of full adaptation: fewer than 3 percent of permanent night workers reach it, and the trials show you do not need it.

Red flags: stop and get assessed

Nodding off at the wheel, or a near-miss on the drive home, is not a sleep-hygiene problem; it is the one outcome in this literature with a measured one-in-three rate and it warrants an immediate change in how you get home and a conversation with a clinician about your schedule. Insomnia or overwhelming sleepiness that persists on days off and on vacation, or that came before the night schedule, points toward a sleep disorder in its own right. So do low mood or a decline in memory and concentration that outlasts the roster. The Detroit data show that shift workers who meet criteria for shift work sleep disorder carry higher rates of ulcers, accidents and depression than those who do not, and that is reason enough to seek an evaluation rather than another supplement.

How to know it is working

Measure the same things the trials measured. Rate your sleepiness once an hour through the last three hours of the shift for a week before you change anything, then again after two or three weeks on the protocol; the nurses' study moved sleepiness at 5, 6 and 7 a.m., so that is where to look. Keep a sleep diary with a fixed start time and total sleep, because in the volunteers who did not realign, seven hours of daytime sleep still produced poor nights, which means total sleep alone can look fine while you feel terrible. Log near-misses and lane drifts on the commute; zero is the only acceptable number. Give it more than a week: the police officers' clocks moved only about two to three hours in four nights, and the compromise-phase studies needed about a week including a weekend before mood and performance approached daytime levels. And once or twice a year, check the ordinary metabolic and cardiovascular markers, since the long-term associations begin to appear after five years, and the earlier a drift is visible, the more an eating schedule and a sleep anchor can do about it.

Return to the equipment list. Sixty-seven young volunteers, some issued lamps, some issued darker glasses, some issued melatonin, all issued a dark bedroom and a fixed bedtime. The ones who did well on the night shift were not the ones who took the pill. They were the ones whose clocks had moved, even partway, because they had controlled when they saw light and when they saw darkness. That is the whole protocol, and every piece of it was on the list. Educational, not medical advice.