Subjects were recruited as part of a multi-site cohort study of community-dwelling persons aged 65 years and over randomly selected from the electoral rolls of three French cities (Bordeaux, Dijon and Montpellier) between 1999 and 2001 (3C Study). The study design has been described elsewhere (21). The study protocol was approved by the Ethics Committee of the University-Hospital of Bicêtre (France) and written informed consent was obtained from each participant. A total of 9294 subjects, 3650 men and 5644 women were included in the study. Data were collected during a face to face interview using a standardized questionnaire by trained psychologists or nurses. Diagnoses of dementia and other neurological disorders were made by 3C study clinical investigators according to DSM-IV criteria (22) and were further validated by a national panel of neurologists independently of the 3C investigators.

Sleep complaints were assessed by a self-report questionnaire which evaluates the frequency (never, rarely, regularly or frequently) of (1) having difficulty in initiating sleep (DIS), (2) waking up several times during the night (DMS), (3) waking up too early in the morning without being able to fall asleep again (EMA). IS was defined as reporting regularly or frequently at least one symptom (DIS, DMS, or EMA). Other information related to sleep was also recorded including taking medication for sleep problems and duration of use (in years), or reporting regularly or frequently being excessively sleepy during the day, having nightmares, or snoring loudly.

A standardized health interview covered socio-demographic factors and current state of health, including gender, age, marital status, education, and body mass index (BMI). Detailed medical questionnaires (with additional information where necessary from general practitioners) were used to obtain information on history of vascular disease including angina pectoris, arrhythmia, lower limb arteritis, heart failure, myocardial infarction, coronary surgery, stroke, and hypertension (resting blood pressure ≥160/95 mm Hg or treated); if subjects presented with at least one of these symptoms they were considered as having history of vascular disease. Other illnesses including asthma, diabetes (fasting glucose ≥7.0 mmol/l or reported treatment), hypercholesterolemia (total cholesterol ≥ 6.2 mmol/l or reported treatment) and thyroid problems were also considered. Participants were classified as having chronic disease if they suffered from one, or two or more, of these illnesses. Depressive symptomatology was assessed with the Center for Epidemiological Studies-Depression Scale (CES-D) with a 16 cut-off point (23). For women, information was also obtained regarding the type of menopause (i.e., natural, surgical or following treatment such as chemotherapy) as well as use of HRT.

A life-style questionnaire was used to obtain information on current smoking status, alcohol intake, consumption of coffee and tea. Dietary intake was assessed by a questionnaire covering the following foods: dairy products, meat (and poultry), fish, eggs, cereals (including bread and starches), raw fruit, raw and dried vegetables, classified as low intake: never or less than once a week; moderate intake: once to 3 times a week, and high intake: 4 to 7 times a week. Mediterranean diet is characterized by high intake of vegetables, legumes, fruits and cereals; high intake of unsaturated fatty acids mostly in the form of olive oil, but low intake of saturated fatty acids; a moderately high intake of fish; a low to moderate intake of dairy products (mostly cheese or yogurt); a low intake of meat and poultry; and a regular but moderate amount of alcohol primarily in the form of wine and generally during a meal (24). A value of 0 or 1 in relation to frequency was assigned to the 11 components of the Mediterranean diet. Subjects reporting more than seven components (highest quartile) were considered has having Mediterranean diet.

Men and women were found to differ on most exposure variables. In order to examine separate patterns of risk, gender stratified analyses were undertaken. Univariate logistic regressions were used to determine differences in socio-demographic and clinical characteristics between men and women. Associations between the outcome variable, i.e., number of IS (0, 1, 2 or 3) and socio-demographic and clinical variables were tested using a multinomial logistic regression model adjusting for all significant associations between socio-demographic, clinical variables at the univariate level with a p-value < 0.10. Results were presented as odds-ratios (OR) and their 95 % confidence intervals (CI). In the subgroup of subjects reporting only one IS (isolated IS), multivariate logistic regressions were used for each IS to model the relationship between IS and associated variables at p <0.10. Significance level was set at p<0.05. The statistical analysis was carried out using SAS software (version 9.1).

Men and women were found to differ on most socio-demographic, behavioral, clinical, and sleep characteristics (Table 1). Compared to men, women were younger and more often widowed or divorced, with a lower education level, drank less alcohol and coffee, and adhered less often to a Mediterranean diet. Women also were less often overweight, had less chronic diseases, and were more likely to be depressed.

Regarding sleep characteristics, women reported more frequently use of sleep medication and for longer periods (median [IQR] was 8 years [2–16] versus 5 years [2–15] in men, Mann-Whitney Test z=−4.23, p<0.0001). Benzodiazepine was most frequently reported among the users (by 51.62 % of men and 60.58 % of women), followed by ‘Z-drug’ hypnotic use (24.86 % of men and 24.97 % of women), antidepressant (7.84 % of men and 7.75 % of women), antihistaminic compound (6.49 % of men and 7.63 % of women), and 8.38 % of men and 9.48 % of women reported herbals/botanicals use. In addition, women more frequently experienced nightmares, and had less often loud snoring and daytime sleepiness.

Women reported more IS than men (75% vs. 70%, p=0.0001) with additional gender differences in patterns of complaint distribution. Women reported significantly more frequently two or three IS while men reported more often only one. Globally, DMS (isolated or associated with DIS and EMA) was reported with similar frequency in men (62.86%) and women (62.25%) while women reported more often overall DIS (42.42% vs. 21.33% in men) or EMA (38.77 vs. 30.01% in men). DIS associated with DMS or with EMA was reported more frequently by women, and DMS associated with EMA was reported more often by men. More specifically, examining the subgroup of subjects presenting with only one IS, DMS was the most frequently reported by both sexes but more often in men than in women whereas DIS was more frequent in women than in men. EMA was equally reported by both men and women.

We evaluated the risks of reporting one, two, or three IS using multivariate multinomial logistic regression models stratified by gender and adjusted for socio-demographic, behavioral, clinical and sleep variables described above (Table 2). For men and women, depressive symptomatology, sleepiness and nightmares were significantly associated with one, two, or three IS, whereas sleep medication and having two or more chronic diseases were associated with two or three IS. For both sexes, age was associated with one IS in all age categories whereas no significant association was observed with three IS. Comparable strength of association (in terms of OR values) was observed for men and women. In women only, high BMI, high alcohol intake and HRT were protective for two or three IS and Mediterranean diet for one, two or these IS. In men only, loud snoring was associated with only one IS and no protective factors were observed. For both sexes, marital status, coffee intake or type of menopause for women was not significantly associated with IS (data not shown).

In order to specify each IS complaint, we evaluated the risks of reporting only one IS using multivariate logistic regression models stratified on gender and adjusted for socio-demographic, clinical and sleep variables (Table 3). For both men and women, age, sleepiness and nightmares were significantly associated with increased risk of DMS whereas sleep medication was independently associated with increased risk of DIS. Depression scores and BMI showed different associations in men and women: higher depression scores were associated with higher risk of EMA in men and with higher risk of DIS in women. An opposite association was observed for high BMI, which was associated with a higher risk of DIS in men and marginally protective in women. Loud snoring was associated with risk of DMS in men only. In women only, coffee drinking and Mediterranean diet were marginally protective for DMS and no variables were significantly associated with EMA.

In our study, IS were found to be highly prevalent, more than 70 % of elderly reporting at least one IS. Similar result was reported in another study with 69% of subjects reporting at least one IS (25). In our study the prevalence was higher in women than in men (75% vs. 70%, χ² test=18.48, df=1, p<0.0001). As already reported, DMS was the most frequent IS (1, 4, 26–30) and we further observed a comparable prevalence in men and women. As an isolated symptom it was independently associated with age in both men and women. With aging, sleep fragmentation is the main component of insomnia. Numerous factors are involved in sleep fragmentation and responsible for neurobiological and behavioral sleep pattern changes with aging, such as hormonal changes (20, 31), alteration of homeostatic process (32, 33), circadian rhythmicity (34), low physical activity and less light exposure(35), sleep apnea (36). All of them could result in an increased brain vulnerability to various arousal stimuli explaining that in elderly DMS is the most prevalent IS whatever the gender.

Most factors, e.g., age, sleep medication, nightmares, sleepiness, chronic diseases, and depression were independently associated with the presence of IS (1, 4, 6). We did not observe significant gender differences regarding the strength of these associations. Interestingly, DIS is more prevalent in women than in men and isolated DIS is principally associated with sleep medication; it might be suggested that long sleep latency is less tolerated than other IS in older persons thus leading to higher treatment rates. This may explain the higher prevalence of DIS in women, already reported in other studies (29, 37–40) and also the longer duration of sleep medication intake in older women.

Women more frequently reported two or three IS compared to men who tend to commonly experience a single symptom. To our knowledge, this is the first time that gender differences with regard to the number of IS are reported in the elderly general population. These gender variations could be the consequence of differences in the nature or intensity of exposure to risk factors, or in the susceptibility to the same risk factors. They could result from cultural, social, behavioral or adaptive differences. In our study, depression is independently associated with isolated IS, with differences in symptom expression, DIS in women and EMA in men. Some depressive symptoms are different in men and in women: women are more prone to ruminate than men (41). It is known that poor sleepers are more prone than good sleepers to ruminate about daytime symptoms and their consequences of their chronic sleep difficulties (42, 43). Ruminative thoughts, worries and anxiety are responsible for difficulties in falling asleep especially in women (4, 42–44). Otherwise it has been shown (45) that the symptoms of terminal insomnia are strongly associated with polysomnographic features (reduced REM, sleep latency) of endogenous depression. It might be possible that in elderly men, depression and sleepiness, associated with terminal insomnia, are the expression of endogenous depression. Polysomnographic recordings will be required to test this hypothesis.

In our study we found opposite effects of high BMI; increasing the risk for isolated DIS in men but marginally decreasing it in women. Biological differences, especially hormonal factors, could be involved in this gender difference. Obesity during the climacteric is a key factor associated with an increase in estrogen exposure. After the menopause, mesenchymal cells of adipose tissue become the main source of estrogen via testosterone aromatization. Therefore, in the post-reproductive years, the extent of female estrogenization is mainly determined by adiposity(46). Interestingly, we found that current long-term HRT use (mean duration was 10 years) was also independently protective for IS in women confirming the beneficial effect of HRT on IS in women (47–49).

In women some lifestyle factors (moderate alcohol, and coffee intake and adherence to a Mediterranean diet) have a protective effect for having two or three IS or isolated DMS. Caffeine and alcohol intake have been related to a common sensation-seeking or behavioral or socialization component in healthy adults. Elderly women might also be more aware than men of the deleterious effect of alcohol or coffee on IS (50, 51), and thus reduce their intake in an attempt to reduce their symptoms.

The protective role of the Mediterranean diet could be related to a reduction in cardiovascular disease which are also risk factors for insomnia (52), however, the effect still persisted when adjusting for vascular disease. To our knowledge the protective role of the Mediterranean diet on IS has not been reported elsewhere.

In men specifically, snoring was associated with isolated DMS. A male predominance of snoring is observed in all epidemiological studies. Associated with obesity, snoring is a marker of sleep apnea (53, 54). It has been shown that a quarter of adults over 65 have sleep apnea and 40 to 50 % of patients with sleep apnea have IS (36). It thus appears that sleep breathing disorders might be associated with IS more frequently in older than in younger subjects.

Finally, no factors were independently associated with isolated EMA in women. However, considering the low number of subjects reporting this IS, this may result from a lack of statistical power.

This study has several limitations. The data relating to some of the covariates were self-reported and may thus be subject to recall bias with participants with sleep disorders responding more negatively to questions about health. However, associations remained even in adjusted analyses, thus suggesting that any bias did not have a substantial influence on the results. There is also the potential for bias in this analysis due to the fact that excluded persons potentially appeared at higher risk for IS so that the associations could be underestimated.

Our study has a number of strengths. The analysis was based on a large multi-centre population study of people aged 65 years and over and therefore the results may be relevant to the broader community of older persons. We could also specify each IS to evaluate risk factors related to isolated or combined IS. The size of our sample also enabled evaluating gender differences while controlling for a large number of covariates, particularly measures of physical health, sleep disorders (sleep treatment, snoring, sleepiness and nightmares), socio-demographic factors, behavioral habits, and depressive components.