|Home | About | Journals | Submit | Contact Us | Français|
The aim of this study was to assess the prevalence of night eating (NE) and associated symptoms in a population-based sample of Swedish twins. A total of 21,741 individuals aged 20-47 years completed a questionnaire in 2005/2006. NE was defined as ≥25% of daily food intake after the evening meal and/or awakening at least once per week with eating episodes. The prevalence of NE was 4.6% in men and 3.4% in women. Among obese men and women, the prevalence was 8.4% and 7.5%, respectively. Men and women with NE had 3.4 and 3.6 times higher risk of binge eating compared to individuals without NE. The risk of sleep-related problems was 1.6-3.4 times higher in men and 2.5-3.3 times higher in women with NE compared to those without NE. This epidemiological study has estimated the prevalence of NE in a twin population. It revealed that NE is 2.5 and 2.8 times more common in obese men and women compared to normal weight men and women. Furthermore that NE is associated with binge eating and sleep-related problems.
Given the increasing prevalence of obesity worldwide (1), and the documented association between sleep-related problems and obesity (2-4), it is of considerable scientific interest to explore how symptoms of night eating (NE) relate to both of these public health problems. As early as 1955, Stunkard described the night eating syndrome (NES) in a study of severely obese women (5). NES was initially defined as nocturnal hyperphagia, insomnia, and morning anorexia (i.e. loss of appetite). In the decades following this report, only a limited number of studies have been conducted to determine the prevalence and nature of NES. As described under Methods, NE is a more broadly defined condition than NES, making it essential to distinguish between these conditions.
The use of varying diagnostic criteria applied to relatively small convenience or clinical samples has resulted in a wide range of prevalence estimates with low precision (6;7). Moreover, the highly select nature of study samples has limited the ability to assess accurately the relation between night eating and associated symptoms such as obesity, binge eating, and sleep-related problems (8-14). Therefore, research on large and population-based samples is needed.
NES is not included in the current version of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV), but provisional diagnostic criteria are available. The current conceptualization defines NES as present when more than 25% of daily caloric intake takes place after the evening meal and/or when nocturnal ingestions of food occur three or more times per week (15).
The causes for the suggested, disturbed circadian pattern of food intake in NES are not clear. The evening hyperphagia and night eating may be related to total overeating and obesity (16). However, O'Reardon et al. studied the eating patterns and food intake in those with night eating and controls and found that despite large differences in the timing of the eating between those with night eating and controls, total energy intake did not differ (17). The causes for the delayed circadian pattern of food intake in NES are not clear, however neuroendocrine differences between individuals with NES and controls seem to be consequences rather than causes of the altered pattern and timing of food intake (15).
With reference to the relation between binge eating and NES, Allison et al. studied three groups of individuals recruited by advertisements in Pennsylvania, USA: 177 participants with binge eating disorder (BED), 68 individuals with NES and 45 overweight/obese comparison individuals without BED or NES (9). The BED group had more episodes of binge eating and episode days of binge eating than the NES group which reported having more episodes of binge eating and episode days of binge eating than the control group. Similarly overeating episodes were more frequent among the BED group than the NES and comparison groups, while the two later groups did not differ. Based on these and other findings the authors concluded that NES and BED are different entities (9). In a recent study from Australia, Colles et al. compared a NES only group, a binge eating only group and a co-morbid NES and binge eating group (18). In the NES group, 40 % reported binge eating and in the binge eating group, 37 % also reported NES. The authors also compared these groups with respect to psychological distress and body mass index (BMI). The co-morbid NES/binge eating group and the binge eating only group showed similar patterns of psychological distress, while the NES only group revealed lower psychological distress than the co-morbid NES/binge eating group. Mean BMI values were similar in the NES only group (40.2kg/m2), binge eating only group (43.7kg/m2) and the co-morbid NES/binge eating group (43.7 kg/m2) (18).
Exploring the relation between BMI and night eating, previous studies have yielded inconsistent results (14;19-24). Several reports have shown positive associations between NES and BMI (20;21;25;26). In contrast, Striegel-Moore et al (14) and also Rand et al (23) found no association between night eating and BMI in two different and large random samples of adults from the USA. In samples of 682 black and 659 white young women from the USA who participated in a follow-up study of risk factors for cardiovascular diseases, Striegel-Moore et al. found no association between night eating and BMI (7). Based on analyses of NHANES-III data from the USA Striegel-Moore et al. reported a weak inverse association between night eating, defined as eating 25% or more after 7 pm, and BMI in adolescents and adults (24). However, analyses according to other definitions, e.g., eating 50% or more after 7 pm, revealed no associations.
In light of the inconsistencies in previous research and the dearth of population-based studies, the objectives of the current study were: (i) to determine the prevalence of NE in a large population of Swedish men and women; (ii) to assess the relation of NE to BMI, sleep-related problems, and binge eating.
In 2005, all twins in Sweden born between 1959 and 1985 (42,582 individuals aged 20-47 years) were invited to complete a web-based questionnaire (STAGE) conducted by the Swedish Twin Register at the Karolinska Institute, Stockholm. The questionnaire was extensive and included demographics, education and occupation, medical history, smoking, alcohol consumption, eating disorders, depression, sleep patterns, and other symptoms and disorders (27). Non-respondents were followed-up by three mailed reminders. Thereafter non-respondents were approached by telephone and invited to participate in a telephone interview. The final response rate was 59.6% (N=25,378).
A total of 21,741 twins aged 20-47 years in 2005/2006 were eligible for participation in the present study as described in detail below. We also analysed a subset by excluding 7,242 individuals who answered “yes” to one or more of the following three questions: “Have you ever worked during night time?” (N=5,773), “Are you currently on maternity/paternity leave?” (N=795) and “Are you currently a student?” (1,410). In this subset 14,499 individuals remained. The reasoning was that shift work or permanents night work, small children or the potentially more irregular circadian life habits of students might to some extent increase the probability of being awake during the night for reasons irrelevant for NE.
We applied both a broad and a narrow definition of NE. Broad NE was defined as awakenings with food intake during the night at least once a week and/or 25% or more of daily food intake after the evening meal. Narrow NE was defined as awakenings with food intake at least once a week and/or 50% or more of daily food intake after the evening meal. Intake of water or other beverages only did not count as food intake. Our definitions and categories of NE were based on the following two questions: (i) “How often do you get up at night to eat?” with response alternatives never, once or twice, weekly, nightly, and don't know/don't wish to answer; and (ii) “What proportion of your daily food intake takes place after the evening meal?” with response alternatives 0%, 1-24%, 25-49%, 50-74%, 75-100%, and don't know/don't wish to answer.
Men and women with present or past symptoms of binge eating were identified through the following two questions: “Have you ever had binges when you ate what most people would regard as an unusually large amount of food in a short period of time?” with response alternatives yes, no and don't know/refuse; and “When you were having eating binges, did you feel that your eating was out of control?” with response alternatives not at all, slightly, moderately, very much, extremely and don't know/don't wish to answer. Those who answered “yes” to the first question and “very much” or “extremely” to the second question were defined as binge eaters.
BMI (kg)/height (m2) was calculated from self-reported height and weight and participants were categorized as normal weight (BMI 18.5-24.9kg/m2), overweight (BMI 25.0-29.9kg/m2) or obese (BMI ≥30kg/m2) according to the WHO criteria (28).
Participants with missing data on both NE questions were excluded (n=1,928) as were individuals who did not endorse one NE question and had missing data on the other (n=605). As described above the final main study population comprised 21,741 individuals with complete data on NE symptoms and BMI.
Non-response to questions about sleep-related problems, located in an “elective” section of the questionnaire, was higher and results requiring access to this information were thus calculated for a sub-sample (n=6,208) as described below.
Information on sleep-related problems was obtained through the following questions: “Have you had difficulty falling asleep (during the past 3 months)?”, “Have you had feelings of not having had enough sleep on awakening (during the last 6 months)?”, “Have you had disturbed or restless sleep (during the last 6 months)?” For each question, those who answered “usually” or “always” were classified as having sleep problems and those who answered “sometimes”, “seldom” or “never” as not having a sleeping problem.
Statistical analyses were conducted in STATA version 9.2 (StataCorp LP, College Station, Texas, USA). All prevalence estimates were age-standardized separately for men and women and the corresponding confidence intervals were adjusted for within-pair correlation using the cluster option. Odds ratios (OR) were estimated by logistic and linear regression models using generalized estimating equations (GEE) to account for within twin-pair correlations. All regression models were age-adjusted. Wilcoxon's non-parametric rank sum test was used to test differences in BMI distributions between groups with and without night eating. The Ethics Committee at the Karolinska Institute, Stockholm, Sweden, has approved this study.
Basic demographic and anthropometric data of participants are shown in Table 1. In men, 86.2% had never experienced awakenings with food intake during the night, 11.7% had experienced it once or twice, 1.9% had weekly and 0.2% daily awakenings with food intake. In women, 89.6% had never experienced awakenings with food intake, 9.2% had experienced it once or twice, and 1.1% reported weekly and 0.1% daily awakenings with food intake. Table 2 shows proportions of daily food intake after the evening meal among men and women.
The prevalence of broadly defined NE was 4.6% in men and 3.4% in women (Table 3A). The age-standardized risk ratio (RR) among men compared to women was 1.40 (p<0.001). The highest prevalence of broad NE was observed in obese men, 8.4%, and women, 7.5%. The odds ratio for broad NE was 2.47 in obese men and 2.80 in obese women compared with their normal weight same-sex counterparts. Among men, the prevalence of broad NE was 4.3% and 5.1% in the age categories 20-34 years and 35-47 years (p=0.047). The corresponding figures among women were 3.2% and 3.6% (p=0.267). The prevalence of narrowly defined NE was 2.5% in men and 1.7% in women in the main study population (N=21,741) and 4.6% in obese men and 4.5% in obese women, respectively (Table 3B). The age-standardized risk ratio (RR) among men compared with women was 1.46 (p<0.001).
The mean BMI was higher in men with broad NE than in men without NE as seen in Table 4A. The difference between the means was 0.6 kg/m2 (95% CI 0.3; 0.9). The mean BMI among women with broad NE was also higher than among those without NE. The difference between the means was 0.9 kg/m2 (95% CI 0.5; 1.3). Among men the prevalence of overweight and obesity was higher among those with broad NE than without NE. The same pattern was seen among women. The BMI distributions differed significantly between those with and without broad NE, both among men (p<0.001) and women (p<0.001) (Table 4A). The risk of obesity was about twice as high in men (OR=1.98, 95% CI: 1.46; 2.69) and women (OR=2.36, 95% CI: 1.73; 3.22) with broad NE compared with same-sexed individuals without NE.
The mean BMI values were also higher in men and women with narrow NE than in those without these symptoms (Table 4B). The mean differences were 0.7 kg/m2 (95CI%, 0.3; 1.2) for men and 1.2 kg/m2 (95%CI 0.6; 1.9) for women. The risk for obesity was increased in men (OR=2.03, 95%CI: 1.37; 3.00) and in women (OR=2.87, 95%CI: 1.92; 4.27) with narrow NE compared with their counterparts without NE. Furthermore, the Wilcoxon's rank sum test showed that the BMI distributions in individuals with and without narrow NE were significantly different (p<0.001 in both men and women).
Of all men and women in the main study population, 0.3% (95% CI: 0.2; 0.4) and 2.7% (95% CI: 2.4; 3.0) reported past or present symptoms of binge eating. The prevalence of binge eating in women with broad NE was 8.9% and in women with narrow NE 11.8%. In men, the prevalence of binge eating was 0.9% in those with broad NE and 1.7% in those with narrow NE. The risk of binge eating was elevated both in men with broad NE (OR=3.44, 95% CI: 1.17; 10.12) and narrow NE (OR=6.76, 95% CI: 2.34; 19.54) compared to men without NE. The risk of binge eating was also higher in women with broad NE (OR=3.56 with 95% CI: 2.44; 5.18) and narrow NE (OR=4.81 with 95% CI: 3.10; 7.48) compared to women without NE.
As mentioned above, the non-response rate to questions about sleep-related problems were higher compared to the questions on night eating and anthropometry because sleep-related problems were presented in an optional part of the questionnaire. Only about 29% of all study participants in the cohort answered these questions. Amongst responders, men and women with NE had higher frequencies of the three different aspects of sleep-related problems than those without NE (Table 5). The risk of having difficulty falling asleep was three to four times higher in men and women with broad and narrow NE compared with individuals without NE. As shown in Table 5, risks were also elevated (about 1.5 to 3 times) for feeling that they needed more sleep on awakening and for disturbed or restless sleep (about 2.5 to 3.5 times).
As reported above, our main study population included subsets of students, individuals who were currently on maternity leave with small children, and individuals who answered affirmative to the question “have you ever worked during night time?” Since these individuals (N=7,242 individuals) might have had irregular circadian life habits for reasons unrelated to NE we repeated our analyses on the 14,449 remaining individuals. In this subset prevalence of NE was 4.3% in men and 3.3% in women. Among obese men and women, the prevalence was 7.2% and 6.4%, respectively. These results are almost identical to the findings from the main study population and as were all other results (data not shown).
Our detailed exploration of NE in a large population-based Swedish twin sample yielded several intriguing findings. In contrast to previous investigations (9;18;19), this large epidemiological twin study suggests that NE is more common in men than in women. We found a rather weak and positive association between age and NE in men as well as in women. We observed a strong association between obesity and NE with NE being approximately 2.5 times more common in obese than in normal weight individuals. This finding replicates several previous investigations (12;18;20), but is inconsistent with other studies (7;11;13;14;23-25;29). We also revealed positive associations between NE and binge eating, with binge eating episodes being 3.5 times more common in both men and women with broadly defined NE compared to subjects of the same gender without symptoms of NE. The observed risk for binge eating was even higher (5-6 times) in men and women with narrowly defined NE. Finally, our hypothesized association between NE and sleep-related problems was also confirmed with those with night eating being more likely to report different sleep-related problems with the strongest association observed between difficulties falling asleep in men and women with narrow NE relative to those without NE. Although perhaps unsurprising that a symptom cluster defined by nocturnal eating would be associated with disrupted sleep, our results indicate that the sleep problems faced by individuals with NE surpass those awakenings with food ingestion and include insomnia and other aspects of disturbed sleep. Despite the sleep disturbance, studies have shown the same sleep onset and offset of those with night eating and control subjects (16). Kept together these findings suggest that NE is associated with a cluster of dysregulated behaviours in the domains of weight, appetite, and sleep.
Marshall et al. suggested that NES may be a risk factor for obesity based on two findings, namely the age difference and difference in duration of night eating between obese and non-obese persons with night eating (22). This author found that in 52% of obese persons with night eating onset of night eating preceded onset of obesity. By contrast, a study by de Zwaan et al found that 60% of obese participants had been overweight before onset of NES (30). However, in a true longitudinal study of night eating and weight change in middle-aged men and women Andersen et al. found that getting up at night to eat contributed to further weight gain in already obese women (19). As already stated, we found strong statistical associations between obesity and night eating and our results provide support for the suggestion that the same association observed in many smaller clinical studies is not merely a consequence of selection bias. In women we observed a tendency to a J-shaped association of BMI with risk of broadly as well as narrowly defined NE, but the association was not statistically significant for the underweight group. However, due to the cross-sectional design of our study we are not able to contribute to the important question about the direction of potential causal relation between obesity and NE.
Our results showing increased risk for binge eating in both men and women with NE compared to individuals without NE are in accordance with several studies based on clinical samples (8;12;13;18), although Allison et al. reported a lesser degree of overlap between these conditions in persons seeking bariatric surgery (10). It should kept in mind that available data did not allow us to apply full DSM-IV-TR criteria, i.e. binge eating for at least two days during the past 6 months, with loss of control, accompanied by distress. In addition, different definitions of study samples, night eating and binge eating have been applied by various authors making comparisons less straightforward.
The results of this study must be considered within the bounds of its strengths and limitations. A considerable strength is that the study was based on questions answered by a large population-based sample of twins from throughout the country of Sweden not relying on clinical case series which inevitably introduces selection bias in estimates. Previous studies have not presented confidence intervals for their estimates making it impossible to appraise the scope of random variability. Our confidence intervals revealed good to excellent precision reflecting the large sample size of the study.
The primary limitation to be considered is the definition of NE. As noted previously, no established DSM-IV criteria for the symptoms of NE or a syndrome of night eating (NES) exist and only provisional criteria have been established. Lack of consistent definitions complicates the comparisons between studies. As is true in any large-scale population-based survey, concessions have to be made to balance quality and depth of information with participant burden and fatigue. Given the number of available questions about NE, we were unable to assess the prevalence of NES according to the provisional criteria, i.e., 25% of daily food intake after the evening meal and/or nocturnal ingestions of food three or more times per week (9). We were able to assess broad and narrow NE as defined above and caution the reader to distinguish between the full syndrome (NES) and symptoms of NE reported here. On one hand this poses a limitation; however, it also poses advantages. Although diagnostic-level findings are meaningful, it is also important to assess eating disorders at the symptom level (31). Understanding disordered eating at the symptom level may facilitate the refinement of phenotypes and may clarify sources of variation for specific components of eating disorder symptomatology that will be relevant to refining the diagnostic syndrome of NES (32).
Another limitation is how night eating was assessed. In the first of the two night eating questions we ask the participants how often they get up to eat. If this question was interpreted literally night eating may have been underestimated if individuals had woken up for other reasons (e.g., to use the bathroom) and then decided to eat after they were already awake. Furthermore the study is limited by the self-reported nature of our data as many obese individuals underreport their weight (33). The rather high overall non-response rate (40%) and the higher non-response rate on the optional sleep section (71%) may have resulted in some selection bias in the study group, expected to be leaner and with less deviant eating patterns and sleep-related problems than non-participants. Finally, our main study population included subsets of students, individuals who were currently on maternity leave with small children and individuals who had worked during night time. When we excluded groups and repeated our analyses on the remaining study subjects we got essentially the same results as for the full study population, indicating that potential irregular circadian habits in these groups may not have introduced any important bias in our results.
We expect the net effect to be some underestimation of NE and associated conditions. Our study was limited to twins and it is sometimes argued that twins may differ from singletons with respect to risk for various diseases. Although this may be true for some conditions closely related to exposures in fetal life and growth in infancy, for most other diseases and conditions, previous research has failed to verify differences in occurrence between singletons and multiples (34). We are not aware of any research indicating that the prevalence of eating disorders, NES or other dysfunctional eating patterns differs between singletons and twins.
In spite of the limitations discussed above, our results highlight important aspects of the phenomenon of NE and associated symptoms. Most striking is the higher prevalence of NE in men than women and substantially elevated risk for obesity in both genders. Although this investigation cannot answer whether it is most fruitful to consider NE as an eating disorder, a sleep-related disorder or both, our results clearly show that NE is positively associated with overweight, obesity, binge eating, and sleep-related problems. The genetically-informative nature of this twin sample will allow us to explore the extent to which genetic and environmental factors contribute to the etiology of NE. Such studies may unravel the overlapping or distinct genetic factors influencing night eating, binge eating, and associated symptoms.
This study was supported by a grant to Finn Rasmussen from the Swedish Council for Working Life and Social Research (grant number 2005-0399). We have no conflicts of interest to declare. We are most grateful to the Swedish Twin Registry for making this study possible by giving us access to an important dataset.