Circadian Eating and Sleeping Patterns in the Night Eating Syndrome


  • John P. O'Reardon,

    1. Weight and Eating Disorder Program, Department of Psychiatry, The University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Brenda L. Ringel,

    1. Division of Sleep Medicine, Department of Medicine, The University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • David F. Dinges,

    1. Division of Sleep and Chronobiology, Department of Psychiatry, The University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Kelly Costello Allison,

    1. Weight and Eating Disorder Program, Department of Psychiatry, The University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Naomi L. Rogers,

    1. Division of Sleep and Chronobiology, Department of Psychiatry, The University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Nicole S. Martino,

    1. Weight and Eating Disorder Program, Department of Psychiatry, The University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Albert J. Stunkard

    Corresponding author
    1. Weight and Eating Disorder Program, Department of Psychiatry, The University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
    Search for more papers by this author

  • The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

3535 Market Street, Suite 3024, Philadelphia, PA 19104-3309. E-mail:


Objective: To compare the eating and sleep-wake patterns of persons with the night eating syndrome (NES) with those of matched control subjects.

Research Methods and Procedures: Forty-six overweight/obese NES subjects (mean age 43.3 ± 9.8 years; 32 women) and 43 similar controls (mean age 39.0 ± 11.0 years; 28 women) wore wrist actigraphs for 7 days and completed sleep and food diaries at home.

Results: There was no difference between the total energy intake of the NES and the control subjects, but the pattern of energy intake differed greatly. Relative to control subjects, the temporal pattern of food intake of night eaters was delayed. Food intake after the evening meal, as a proportion of the 24-hour intake, was more than 3-fold greater in NES subjects than in controls (34.6 ± 10.1% vs. 10.0 ± 6.9%, p = 0.001). NES subjects had sleep onset, offset, and total sleep duration times comparable with those of controls. NES subjects reported more nocturnal awakenings than did controls (1.5 ± 1.0 per night vs. 0.5 ± 0.5; p < 0.001), and their actigraphically monitored arousals occurred earlier during sleep (at 128 minutes after sleep onset vs. 193 minutes, p = 0.01). NES subjects consumed food on 74% of the awakenings vs. 0% for the controls.

Discussion: The pattern of cumulative energy intake of the night eaters suggests a phase delay in energy consumption relative to sleep-wake times. NES may involve a dissociation of the circadian control of eating relative to sleep.


The night eating syndrome (NES)1 is characterized by morning anorexia, evening hyperphagia, insomnia, either initial or midphase, and awakenings frequently accompanied by eating (1).

This outpatient study assessed the timing and distribution of food intake, relative to sleep timing and continuity in subjects with NES. The purpose of the study was to extend our earlier behavioral observations of NES (2) to a larger number of subjects and to evaluate the profile of caloric intake in NES subjects relative to the circadian profile of the sleep-wake cycle. Specifically, we sought to determine whether NES involved either a temporal displacement of both caloric intake and sleep-wake cycles or a delay of caloric intake within a normally timed sleep-wake cycle.

Research Methods and Procedures

Ambulatory wrist actigraphy and daily diaries were used to compare the pattern and timing of food intake and sleep-wake profiles in patients diagnosed with NES relative to a demographically comparable group of healthy control subjects.


Subjects were recruited from printed advertisements, radio talk shows, and television interviews that described NES. A total of 46 overweight/obese night eaters (32 women; 43.3 ± 9.8 years old; BMI 34.9 ± 7.1) and 43 overweight/obese control subjects (28 women; 39.0 ± 11.0 years old; BMI 36.7 ± 6.2) completed a 10-day, outpatient behavioral assessment period. BMI was similar between the groups, whereas the NES subjects were slightly older [43.3 vs. 39.0 years; t (87) = 2.0, p = 0.05]. The NES group consisted of 56.5% whites, 41.3% African Americans, and 2.2% Hispanic/Latinos. The control group consisted of 60.5% whites, 37.2% African Americans, and 2.3% Hispanic/Latinos.


The screening procedure included: a structured clinical interview, performed by a trained clinician, designed to assess the presence or absence of NES; the Structured Clinical Interview for the DSM-IV (3) to assess the presence of past or current psychiatric disorders; and the Eating Disorder Examination (4) to assess the presence of concomitant eating disorders.

Eligible subjects were ≥18 years, met standard criteria for NES as outlined by Birketvedt et al. (2), based on assessment at the structured clinical interview, and had a BMI ≥ 27 kg/m2. Applicants were excluded if they: were severely depressed, as determined by Structured Clinical Interview for the DSM criteria; had a lifetime diagnosis of bipolar disorder or any psychotic disorder; reported substance abuse or dependence within the last 6 months; were currently taking psychotropic medications (including hypnotics); were working a night shift or swing shift schedule; were in a weight reduction program; were diagnosed with another eating disorder; or lacked awareness of their night eating episodes. The latter criterion was used to exclude subjects with nocturnal sleep-related eating disorder, in which nocturnal eating is accompanied by a lack of awareness at the time and amnesia for the behavior the following day (5). Eligible control subjects met the same inclusion and exclusion criteria with the exception of not being diagnosed with NES. Eligible subjects were enrolled in a 10-day monitoring period that assessed patterns of food intake and sleep-wake timing as described below.


Sleep-wake timing was assessed using ambulatory wrist actigraphy monitoring combined with twice daily diaries. The study utilized the Actiwatch-L Mini-Logger Series wristband dual-axis sensor with actillume (Mini-Mitter, Sun River, OR), which was worn by all subjects 24 hours a day for a period of 10 days and 11 nights. Recordings of the amount of movement were made at 1-minute epochs. Of the 10 days recorded, the first 2 days were considered adaptation to the study procedures; the next 7 days were analyzed for sleep-wake timing; and the last day was discarded because of the frequency of incomplete data. The Actiwatch-L uses a motion sensor to detect bodily movement and, thus, yields an objective measure of rest-activity cycles, which can serve as surrogates for sleep-wake activity. Several studies have reported high correlations between actigraphy and polysomnographic assessment of total sleep time, sleep latency, and sleep termination in both healthy and patient populations (6,7,8,9,10,11,12,13). Together with a daily diary, actigraphy data were used in the study to determine sleep onset and sleep offset, and the sleep period duration (as defined by the time difference between sleep onset and sleep offset), as well as to detect awakenings accompanied by getting out of bed.

Daily Diary

Subjects kept a daily food, sleep, and mood diary in which they recorded all foods and beverage consumed, with time, portion size, and brand noted. They recorded bedtime and morning rising time, time and duration of all daytime naps, ratings of depression and hunger on a visual analogue scale before each ingestion, and all nocturnal awakenings and whether they were accompanied by food intake. An awakening was operationalized as an occasion when a subject awoke and got out of bed. Thus, we applied a more stringent definition of awakening than that required by the Birketvedt et al. criteria (2) because only awakenings where the subject rose from bed were counted. We believed this would provide a more accurate measure of the awakenings more specifically tied to NES by not counting awakenings where the subject might awake briefly before returning to sleep. Caloric intake was averaged across 7 days in hourly increments for each subject, and the 24-hour caloric intake profile for NES and control groups was determined. These mean values for each hour were then used to plot the 24-hour cumulative caloric intake curve for each group.

Analysis of Actigraphy and Diary Data

Diaries and Actiwatch-Ls were collected at the end of the 10-day period, when the Actiwatch data were downloaded and the diary information was coded. Diary data were analyzed for the same time period as actigraphy data. Two raters analyzed the Actiwatch-L data manually in conjunction with the participants’ diary records so as to compare the 7 days of diary records with activity counts generated by the Actiwatch-L (as described in the Appendix). They ascertained by each method the timing of sleep onset and sleep offset, the sleep period duration, and the number of awakenings accompanied by getting out of bed. A research dietitian analyzed food records for caloric intake.

Statistical Analysis

Means and SDs were calculated using descriptive statistics. Student's t tests were performed to compare means between groups for caloric intake, number of nocturnal awakenings, and timing of sleep onset, offset, and sleep duration with Satterthwaite adjustments when the variance between groups was unequal. A repeated-measures ANOVA was used to compare caloric intake between groups across three 8-hour increments: day (6 am to 1:59 pm), afternoon/evening (2 pm to 9:59 pm), and night (10 pm to 5:59 am). Correlations were used to compare the degree of agreement between diary and actigraphy times, and frequencies were computed using χ2 tests.


Energy Intake

Total energy intake of the night eaters (2314.4 ± 748 kcal) and that of the control subjects (2420 ± 748 kcal) did not differ significantly, but the pattern of cumulative energy intake throughout the 24 hours differed markedly. As illustrated in Figure 1, the cumulative intake curve of the controls was characterized by upward inflections in the late morning and early evening, reflecting the effects of midday and evening meals. An evening plateau was observed from 9 pm in the controls, with no further food intake after that time until the next morning at 6 am. In contrast, the cumulative curve of NES subjects demonstrated lower energy intake during most of the day without the acute meal-based upward inflections observed in the control subjects. There was also no late evening plateau in food intake but rather a steady rise in the total calories ingested up until 6 am. Initial diagnosis of NES at screening required that the food intake reported after the evening meal be at least one-half of the total 24-hour caloric intake. Diary records, however, revealed a lower intake in the night eaters after the evening meal during the monitoring period—34.6% (±10.1%) of the total daily caloric intake compared with 10.0% (±6.9%) by the control group.

Figure 1.

Mean cumulative caloric curves and sleep timing profile for NES and control groups. Note the lack of evening plateau in food intake and the more frequent arousals from sleep in the NES group compared with controls.

The distribution of the 24-hour caloric intake, when assessed in eight hourly increments, is illustrated in Figure 2. Caloric intake of NES subjects compared with controls was significantly lower in the first 8 hours of the day [6 am to 1:59 pm, t (87) = 8.4, p < 0.001], not different in the second 8-hour period [2 pm to 9:59 pm, t (87) = 0.87, p = 0.387], and significantly greater during the last 8-hour period [10 pm to 5:59 am, t (84.4) = 7.97, p < 0.001]. There were no differences in either the cumulative 24-hour intake or in the 8-hour incremental intake between weekends and weekdays for NES subjects. Controls had a higher 24-hour caloric intake during weekends relative to weekdays [2690 vs. 2404 kcal, t (44) = 3.81, p < 0.001], due to a slightly higher intake in the second 8-hour period of the day [2 pm to 10 pm, 1459 vs. 1227 kcal, t (44) = 4.21, p < 0.001]. Nevertheless, the 24-hour intake over the weekends did not differ between the groups (controls vs. NES means: 2690 vs. 2445 kcal, t (80) = 1.1, p = 0.27).

Figure 2.

Caloric intake in 8-hour increments for NES and control groups. Note that food intake is lower in the first part of the day and higher in the evening and night in the NES group. **p < 0.01.


Subjects’ reported bedtimes (diary data) correlated highly with actigraphy quiescence (NES, r = 0.95, p < 0.001; controls, r = 0.93, p < 0.001). Similarly, diary entries for sleep termination correlated significantly with activity on the Actiwatches (NES, r = 0.93, p < 0.001; controls, r = 0.85, p < 0.001). Actigraphy confirmed the diary record of awakenings within 30 minutes of the recorded time for 94.2% of the events in NES subjects compared with 87% of the events in the control group. The mean difference between diary-recorded nocturnal out-of-bed periods and increased activity counts from actigraphy was 2 ± 9 minutes for night eating subjects and 4 ± 17 minutes for controls (p < 0.06).

Complete and analyzable actigraphy and diary data were available for 90.7% of all nights for NES subjects and 80.0% of all nights for controls. Some nights were not analyzable because of incomplete diary data or because the subject removed the actigraph overnight. In 3.1% of the arousals noted by actigraphy in NES subjects, there was a failure to record the actual time of the arousal in the diary (but not the event itself), compared with none in the controls.

Table 1 shows the similarity of sleep patterns of NES and control subjects in the timing of sleep onset by actigraphy and diary records. The sleep offset time in NES subjects compared with control subjects was marginally later by both actigraphy (29 minutes later, p = 0.07) and diary (25 minutes later, p = 0.09). The sleep duration did not differ significantly between the two groups either by actigraphy or diary measures, as shown in Table 1. Both groups slept longer by diary and actigraphy criteria during weekends compared with weekdays, but they did not differ from each other in the total sleep duration either during the weekdays or weekends (Table 2).

Table 1.  Diary and actigraphy data for NES vs. control subjects
 Mean and SD of time (hours:minutes) 
  • *

    Tests were performed with Satterthwaite adjustment for unequal variances.

Time of diary-recorded sleep onset23:31 (1:40)23:32 (1:06)t (78.4) = 0.06, p = 0.96*
Time of actigraphy quiescence23:57 (1:33)23:47 (1:07)t (73.8) = 0.52, p = 0.6*
Amount of time elapsed (in minutes) between diary sleep onset and first awakening149 (68)211 (127)t (42.1) = 2.5, p = 0.02*
Amount of time elapsed (in minutes) between quiescence and first period of increased actigraphy activity128 (68)193 (127)t (41.9) = 2.6, p = 0.01*
Time of diary first awakening01:58 (1:51)03:03 (1:54)t (74) = 2.5, p = 0.02
Time of actigraphy first period of increased activity02:05 (1:51)03:01 (1:56)t (74) = 2.1, p = 0.04
Time of diary-recorded sleep offset07:24 (1:07)06:59 (1:12)t (87) = 1.7, p = 0.09
Time of actigraphy am increase in activity07:35 (1:11)07:06 (0:59)t (74) = 1.9, p = 0.07
Duration of sleep period by diary (in hours and minutes)7 hours and 53 minutes duration (1 hour and 13 minutes)7 hours and 26 minutes duration (1 hour and 20 minutes)t (87) = 1.6, p = 0.11
Duration of sleep period duration by actigraphy (in hours and minutes)7 hours and 37 minutes duration (1 hour and 26 minutes)7 hours and 18 minutes duration (1 hour and 23 minutes)t (74) = 0.97, p = 0.34
Awakenings/night1.5 (1.03)0.5 (0.46)t (63.3) = 5.9, p < 0.001*
Ingestions/night1.2 (0.94)0 (0)t (86) = 8.0, p < 0.001
Table 2.  Sleep period duration (by diary) (±SD) during the weekend vs. weekdays in NES and control subjects
 NES (hours:minutes)Controls (hours:minutes)Significance between groups
  • *

    Not significant.

Weekends8:21 (1:33)7:54 (1:10)p = NS*
Weekdays7:45 (1:09)7:26 (1:10)p = NS
Significance within groupsp = 0.04p = 0.03 

Nocturnal Arousals and Activity

NES subjects reported significantly more nocturnal arousals than did controls (1.5 ± 1.0 per night vs. 0.5 ± 0.5, p < 0.001). The mean number of arousals found in the NES group in this study was lower than that reported by the Birketvedt et al. study (4) (1.5 ± 1.0 vs. 3.6 ± 0.9), as might be expected by virtue of methodological differences between the studies. A more stringent requirement was applied to operationalize an awakening in this study, namely an arousal followed by getting out of bed. These arousals, as determined by actigraphy, commenced 2 hours and 8 minutes (±1:08) after quiescence among NES subjects, compared with 3 hours and 13 minutes (±2:07) after quiescence among control subjects, about an hour earlier in the sleep period (p = 0.04). During the 7-day recording period, NES subjects not only had more nocturnal awakenings than did controls, but also a larger percentage of them had multiple awakenings. Figure 3a shows that all of the night eatersawakened on at least 1 night per week, compared with 70% of the control subjects [χ2 (1) = 16.3, p < 0.001], and 44% awakened on 6 or more nights, compared with only 9.3% of the control subjects [χ2 (1) = 13.2, p < 0.001]. Figure 3b shows that night eaters consumed food on 74% of their awakenings, whereas no control subject ate on awakening. The frequency with which awakenings were followed by ingestions declined during the night from 88.6% the first time to 65.8% the second time, 60.4% the third time, and 41.2% the fourth time, if it occurred at all.

Figure 3.

(a) Percentage of NES and control participants who awakened (by diary, confirmed by actigraphy) at least 1 night per week, at least 3 nights per week, and 6 or more nights per week. All three frequency indices were significantly higher for NES than for control subjects (p < 0.001). (b) Mean number (±SE) of awakenings and ingestions (by diary) per night by NES and control groups (for awakenings p < 0.001). **p < 0.01.


This controlled study showed evidence of an abnormal circadian phase relationship between the timing of sleep-wake cycles and the timing of food intake in patients diagnosed with NES. Although nocturnal sleep onset, offset, and duration in NES participants were comparable with those of a demographically similar control group (Table 1), food ingestion profiles differed greatly between the two groups (Figures 1 2 and 3). The results of the study suggest that the NES may be characterized by a circadian phase delay in the timing of food intake relative to a normally timed (phase appropriate) nocturnal sleep-wake cycle.

The circadian rhythms of eating and sleep are ordinarily synchronized with each other in humans such that food intake does not occur nocturnally. The synchronization of these rhythms is performed by a central pacemaker in the suprachiasmatic nucleus (SCN) of the hypothalamus, which is believed to be responsible for the tight coordination of most, if not all, rhythms of behavior and physiology (14,15), including locomotor rhythms (16). It also coordinates the activity of peripheral circadian oscillators, which are present in tissues such as the liver, and are believed to have a role in the signaling of food intake (17). Results from the behavioral methods used in the current study suggest that the phase delay between cumulative caloric intake and sleep-wake timing ranges in duration from 2 to as much as 6 hours (Figure 1), depending on how it is measured.

The NES and control participants did not differ in their total 24-hour energy intake (2314.4 ± 748 kcal vs. 2420 ± 748 kcal), which suggests that the central feature of NES is primarily the abnormal timing of food intake. Further evidence in support of this observation comes from published reports of NES from other countries such as Italy (18,19). NES is recognized in these settings even though the Mediterranean lifestyle dictates a much later time for the evening meal. From these reports, it is evident that the food intake pattern is out of phase with the sleep rhythm. Thus, NES appears to be a transcultural syndrome where the disordered rhythm of food intake is able to manifest itself clinically despite widely varying meal schedules across cultures.

These findings in NES highlight an apparent dissociation between the circadian patterns of eating and sleeping. A possible explanation is that patients with NES may have circadian oscillators that signal food intake at a different phase relationship from that normally signaled by the SCN. In the laboratory, alteration in the timing of food intake has been shown in mice to uncouple the food intake rhythm from that of the central circadian pacemaker in the brain (20,21,22,23,24). Thus, when nocturnally active rodents are fed during a light period, the peripheral circadian gene cycle in the liver shifts (22), in one study in as little as 2 days (23).

In persons afflicted with NES, there is also alteration in the timing of food intake relative to the circadian regulation of the sleep-wake cycle. There seem to be three possible mechanisms: a dysfunction in either the central or peripheral circadian clocks, a problem of central to peripheral clock coupling, or inappropriately timed entrainment of peripheral oscillators (through eating at night). The NES may be an important clinical example, possibly the first in humans, of a dissociation of the patterns of eating and sleeping.

As discussed above, the dissociation may be maintained by inappropriately timed ingestive behavior. There are links between stress and the onset of NES, and elevated levels of serum cortisol have been reported in the disorder (2). A stress event might serve to precipitate nocturnal awakenings (i.e., sleep maintenance insomnia) in susceptible individuals. The act of eating during these awakenings (perhaps as a form of self-soothing stress management) might entrain the peripheral oscillators responsible for circadian signals for energy intake to a delayed phase relative to the sleep-wake cycle. Continued nocturnal ingestions would then perpetuate the syndrome by delayed phase entrainment of these peripheral clocks. Once an abnormal phase relationship between eating and sleeping is established in NES, it might contribute to the association between NES and the development of obesity. High rates of the syndrome have been found in obese patients in bariatric surgery clinics (25).

Limitations of this study include the reliance on self-reported food intake, which may have been influenced by self-monitoring. The timing of sleep onset and offset was based on diary records and actigraph readouts, each of which is subject to error. The convergence of the diary records and actigraphy, however, suggests that the measures captured the basic behavioral aspects of NES.

The dissociation of circadian rhythms of eating and sleeping in the NES participants observed in this study warrants further investigation. The dissociation produced experimentally in nocturnally active rodents by feeding during the light period may serve as an animal model for this key characteristic of the NES. Sertraline, a serotonin selective reuptake inhibitor, has been reported to restore a normal food intake pattern in patients with NES (26). The dorsal raphe nucleus of the midbrain (site of origin of serotonin neurons in the brain) projects a significant modulatory input to the SCN (27). It would be useful to test whether sertraline might also restore a normal food intake phase in nocturnal rodents entrained to eat at an adverse circadian phase.


We thank Greg Maislin and Jacqueline Cater (Biomedical Statistical Consulting, Wynnewood, PA) for statistical assistance and Lisa Basel-Brown for dietary analysis. This study was supported by NIH Grant R01-DK56735 and by Pfizer Pharmaceuticals.


  • 1

    Nonstandard abbreviations: NES, night eating syndrome; SCN, suprachiasmatic nucleus.


Rules for the Analysis of Activity

“Quiescence” was defined for the actigraph as onset of an unequivocal period of at least 10 minutes of reduced activity that began no earlier than 15 minutes before the bedtime recorded in the diary. If the patient did not record a bedtime, no onset of quiescence was determinable.

When the participants responded affirmatively to the question “Did you get up during the night?”, the actigraphic record was investigated for a period of unequivocally increased activity of at least 3 minutes in length, within 30 minutes of the time recorded that the subject got out of bed. We chose the more conservative route of comparing only the time that the patient reported having gotten out of bed rather than the entire range of time that they spent out of bed. This did mean, however, that we probably underestimated the incidence of nocturnal awakening periods.

“am increase in activity” was defined as the first period of 10 sequential minutes of increased activity on the actigraphic record, which followed the rise time recorded in the diary. Out-of-bed periods were times when diary reports of the subject being out of bed were confirmed by actigraphy.