Abstract
Our aim was to estimate the prevalence of nocturnal awakening with headache (NAH) in the population of São Paulo City according to gender, age (20–80 years old) and socioeconomic classes and its relationship to sleep disorders, sleep parameters, anxiety, depression, fatigue, life quality and obesity. We used a population-based survey with a representative three-stage cluster sample. Questionnaires and scales were applied face-to-face, and polysomnography was performed in 1101 volunteers, aged 42 ± 14 years, 55% women. The complaint of NAH occurring at least once a week had a prevalence of 8.4%, mostly in women, obese subjects and those aged 50–59 years-old. We observed associations of NAH with insomnia, restless leg syndrome (RLS), nightmares and bruxism, but not obstructive sleep apnea syndrome. In a logistics regression model, risk factors for NAH were female gender, odds ratio (OR) (95% confidence interval [CI]) 4.5 (2.8–7.3); obesity, OR 1.9 (1.1–3.3); age between 50 and 59 years, OR 2.4 (1.2–4.7); severe anxiety, OR 8.1 (3.6–18.1); RLS, 2.7 (1.2–5.6); and nightmares, 2.2 (1.3–3.7). Our study shows that NAH was highly prevalent in the population of São Paulo and suggests that this phenomenon has specific characteristics with specific risk factors: obesity, RLS and nightmares.
Introduction
The relationship between sleep and headache has been documented, anecdotally or otherwise in the medical literature for well over a century, and clinical texts allude to the importance of sleep as a headache precipitant (1). Interestingly, sleep is also a common palliative agent: 85% of migraineurs indicated that they chose to sleep or rest because of headache, and 75% were forced to do so (1). Both sleep disorders and headaches are common in the whole population and often coexist in the same individual (2,3). There is a clear association between headache and sleep disturbances, especially when headaches occur during the night or early morning (4,5).
Waking up during the night because of a headache is a complaint cited in some studies (3,6,7), but the prevalence of this complaint in the general population is not well known. The prevalence of morning headache, that is, headaches present upon waking in the morning, on the other hand, has already been studied (8,9). Some studies have reported a strong association between morning headaches and obstructive sleep apnea syndrome (OSAS) (7,9,10) and snoring (9–11). However, one epidemiological study shows that this type of headache is not specific to sleep-related breathing disorders (8).
There are few studies examining the relationship between nocturnal awakening with headache (NAH) and sleep disorders. Evans and colleagues (6) contend that headaches that awaken people during the night are the result of disrupted nocturnal sleep or one of the following underlying processes that disrupt sleep: OSAS or nocturnal hypoxia/hypercapnia, restless leg syndrome (RLS) or periodic leg movement during sleep, psychophysiologic insomnia or depression/anxiety.
In an attempt to further explore this issue, we examined the importance of five categories of factors in relation to NAH in a representative sample of individuals of São Paulo, Brazil, the largest South American city. Our aim was to determine the prevalence of night-awakening due to headache, regardless of the type of headache. We also examined the relationship of this complaint to sociodemographic determinants, sleep disorders and sleep parameters, daytime functioning (anxiety and depression levels, somnolence, fatigue) and quality of life.
Methods
Population under investigation
São Paulo, Brazil, is the largest city in the southern hemisphere (www.ibge.gov.br), with a population of 10,886,518 in January 2008.
The protocol for this study was approved by the Ethics Committee for Research of the Universidade Federal de São Paulo (CEP 0593/06) and was registered with ClinicalTrials.gov (number NCT00596713). Selected volunteers read and signed an informed consent form.
The methodological details of the São Paulo Epidemiologic Sleep Study are described elsewhere (12).
Sampling procedures
This single-center study involved 1,101 individuals. The sample size was established to allow prevalence estimates with 3% precision (13). To obtain a representative sample of the inhabitants of São Paulo, we used a three-stage cluster sampling technique with unequal selection probability, which is the same conceptual framework used in North American National Health Surveys (14). The first stage was designed to ensure representation of individuals with different levels of wealth. The city of São Paulo is divided into 1,500 census districts, and we proportionally selected 96 districts from the four homogenous socio-economic regions of São Paulo. Permanently occupied private homes were identified and numbered within these districts, and clinics, schools and other commercial and non-commercial establishments were excluded. In the second stage, 11 households were selected in each sector. Of these eleven, one household was randomly selected, and the others were chosen depending on the total number of homes divided by a fixed number. For example, in a sector with 110 households in which household number 5 was randomly selected, the other selected households would be those numbered 15, 25, 35, 45, 55, 65, 75, 85, 95 and 105. Each apartment in an apartment building was considered to be a household, and counting proceeded from the uppermost floor to the lowest floor. Finally, in the third stage of sampling, all of the eligible inhabitants of each selected house were arranged according to age (youngest to oldest); 96 pre-established random tables were used to select an individual subject. Pregnant and lactating women, people with physical or mental impairments that prevented them from caring for themselves, individuals who were younger than 20 or older than 80 years of age and people who worked every night were not included in the household drawing. Substitutes were chosen if the target individual could not be contacted (after three attempts), refused to participate, was prevented from participating by a family member or was unable to participate because of travel plans, scheduling conflicts or hospitalization. Substitutes were chosen from an adjacent household according to the same random selection criteria described above. As described elsewhere (12), 165 volunteers were substituted to reach the total sample size. A test of homogeneity (15) examined age, gender, and socio-economic status and found no statistically significant differences between the original volunteers and the substitutes. Based on these results, we concluded that the substitution did not introduce a significant selection bias. A total of 1,042 volunteers underwent polysomnography (PSG) at the Sleep Institute; very few volunteers (5.4%) refused to do the PSG test. Age (p = .11), gender (p = .55), and socio-economic status (p = .38) distributions did not significantly differ between the volunteers who accepted PSG recording and those who refused.
A generated weight variable was applied to match the sample by gender, age group and socio-economic status with the demographic projections for the city inhabitants in 2007. These projections were derived from the 2000 city census.
Home and in-lab data gathering
Socio-economic information and demographic data were collected by face-to-face home interviews. The UNIFESP Sleep Questionnaire (16), Pittsburgh Sleep Quality Index (17), Insomnia Severity Index (18) and Berlin Questionnaire (19) were also administered. The volunteers were then invited to participate in in-lab, full-night PSG. Less than two weeks later, the volunteers reported to the sleep lab and answered the following instruments International Restless Legs Scale (IRLS) (20), Bruxism (Sleep Institute, unpublished), Chalder Fatigue Scale (21), General Health (22), WHOQOL-BREF quality of life assessment (23), Beck Anxiety Inventory (BAI) (24), and Beck Depression Inventory (BDI) (24). All of these were administered by trained psychologists. Weight and height were measured, and an in-lab, full-night PSG was performed. Subjects with a body mass index (BMI) between 25 and 30 kg/m2 were identified as overweight, and those with a BMI higher than 30 kg/m2 were considered to be obese.
Complaints assessment
The question was clearly formulated and asked by trained interviewers (12). The presence of NAH was assessed using the following question in the UNIFESP sleep questionnaire (16): “Some events might take place during sleep. Sometimes you only realize it when you wake up, precisely because these events have woken you up. I will read a list of these events. I would like you to tell me if they happen to you and how often.”
Among the events cited there was one of particular interest: waking up as a result of a headache (daily; 3–6 times/week; 1–2 times/week; 2–3 times/month; less than 1 time/month). We counted only the frequencies of NAH (night).
This complaint was analyzed according to four ranges of frequency: 1. never; 2. less than once a month; 3. one to three times a month; and 4.one to six times a week (including every day).
To assess the time for the onset of NAH we used the following question: “When do you start to wake up during the night due to headache?”
This complaint was also analyzed in relation to demographic variables, sleep disorders, PSG parameters, daytime functioning and quality of life.
The presence of sleep disorders was determined following the International Classification of Sleep Disorders (ICSD-2) (25), except for insomnia, which was determined by the general criteria of the Diagnostic and Statistical Manual of Mental Disorders (26).
Polysomnography
A full-night PSG was performed using a digital system (EMBLA® S7000
Statistical analysis
Using a sample weight variable, prevalence and odds ratio (OR), estimates were generated using pseudo-likelihood maximization to expand the data to the entire population of São Paulo. We used Taylor series linearization to prevent underestimation bias in our estimates of variability. Multivariate logistic regression models were used for the analysis of adjusted associations and interactions among the variables. The variables that had p < .2 in the bivariate analyses were included in the initial model, but only those variables that had p < .05 remained in the final model. In the process of model fitting, no interaction or effect modifier was identified.
All analyses were performed using STATA 10 software (Stata Survey Data Reference Manual, STATA Corporation, College Station, TX, USA, 2007).
Results
Distribution of gender, age and BMI classification for a probabilistic sample (N = 1042) of São Paulo inhabitants
BMI = body mass index.
The presence of weekly NAH was related to female gender, obesity, insomnia, RLS, nightmares and a high level of anxiety. Factors associated with monthly NAH were female gender and insomnia. The only factor associated with NAH occurring less than once a month was female gender. Consequently, the complaint of frequent NAH (at least once a week) in a probabilistic sample in the city of São Paulo was most associated with main sleep disorders. Therefore, this was the selected cut-off point of the study.
Frequencies and adjusted ORs with CIs (95%) of the NAH by socio-demographic variables in a representative sample (N = 1042) of São Paulo inhabitants
OR = odds ratio. CI = confidence interval. NAH = nocturnal awakening with headache. BMI = body mass index.
Frequencies and raw ORs with CIs (95%) of the NAH for sleep disorders variables in a representative sample (N = 1042) of São Paulo inhabitants
OR = odds ratio. CI = confidence interval. NAH = nocturnal awakening with headache. RLS = restless leg syndrome. OSAS = obstructive sleep apnea syndrome.
An examination of PSG results (not shown in the tables) revealed that there was no association among the complaint of NAH and sleep efficiency, sleep latency, REM sleep latency, percentages of stages 1 and 2, slow-wave sleep or arousal index. The only significant association was with REM sleep percentage lower than 20 (OR = 2.04; confidence interval [CI] 95% 1.13–3.70; p = .04).
Frequencies and adjusted ORs with CIs (95%) of the NAH for daytime functioning and quality of life variables in a representative sample (N = 1042) of São Paulo inhabitants
OR = odds ratio. CI = confidence interval. NAH = nocturnal awakening with headache. WHOQOL = World Health Organization Quality of Life questionnaire.
Logistic regression model of the NAH for selected variables in a representative sample (N = 1042) of São Paulo inhabitants
OR = odds ratio. CI = confidence interval. NAH = nocturnal awakening headache. WHOQOL-Bref = World Health Organization Quality of Life questionnaire. BMI = body mass index. RLS = restless leg syndrome.
Discussion
This study is the first to explore the associated factors of NAH in a large sample. It also includes PSG, which is new for this kind of study. NAH was prevalent in the São Paulo population, and female gender, age of 50–59 years, obesity, anxiety, RLS, insomnia and nightmares were all identified as risk factors for this condition.
We found a prevalence of 8.44% of NAH, which was determined by including all subjects who said that they woke in the middle of the night at least once a week because of headaches. The median time in which these nocturnal headaches started was eight years before the study, which is a long duration and suggests the chronic nature of this complaint.
Risk factors related to NAH were examined according to five main categories: socio-demographic determinants, sleep disorders, PSG parameters, daytime functioning and quality of life.
Among socio-demographic determinants, we found that being a woman, being between 50–59 years of age and being obese were positively related to the prevalence of NAH. A study that examined the frequency of the morning headache, which is another medical condition, in a community-based sample also found that being a woman and being middle-aged were positively related to the incidence of morning headache (8).
Research shows that females are affected to a greater degree than males by all types of headaches (29,30). There is also a higher prevalence of migraine headaches in females by a factor of 2.5 : 1 (29). Our data show that the NAH OR estimate is four times higher among females. With respect to age, NAH prevalence was higher in subjects aged 50–59 years (12.5%; Table 2). Among the younger groups of 20–29- and 30–39-year-olds, the prevalence observed was lower, 5.5 and 8.0%, respectively, but still relevant. This finding is in line with the age prevalence in one of the most common types of headache, migraine, which is most prevalent in subjects aged 20–40 years (29).
Another point with respect of age is that this age range (50–59-year-old) in women is linked to menopause. A study that focuses on menopause and midlife health risks found that the chief complaint of women in menopause is sleep disruption (31). Questionnaire-based studies reveal that menopausal woman frequently complain of problems with their sleep (32). On the other hand, headache is one of the major complaints included in the symptomatology of menopause in early studies (33). Therefore, as observed in our results for the 50–59-year-old female group, headache, sleep disorders and menopause are strongly related.
It has already been shown that headaches that occur during or immediately after sleep are particularly suggestive of sleep disorders (34). In clinical practice, sleep disturbances have to be considered among headache patients (1), especially for those who have migraines. In fact, the association between migraines and sleep difficulties has been cited in many studies. There is an approximately threefold increase in sleep problems among patients who report migraines (35). Paiva and colleagues (4) identified the presence of a specific sleep disorder in 55% of subjects with a headache onset during the nocturnal sleep period. In our study, we found an association of NAH with the following sleep disorders: insomnia, RLS, bruxism and complaints of nightmares. Evans and colleagues (6) contend that headache that awakens the individual in the middle of the night is related to the following: OSAS or nocturnal hypoxia/ hypercapnia; RLS or periodic leg movement during sleep; psychophysiologic insomnia, and depression/anxiety. These findings are in line with the results of our study, except that we did not observe an association with OSAS. This result could be due to the fact that the nocturnal headache is frequently included in the same category as morning headache (36,37). However, our data show that NAH has specific characteristics and seems to be a different kind of phenomenon with specific risk factors: obesity, RLS and nightmares. The other identified risks—female gender, insomnia and a high level of anxiety—are also identified as risks of morning headache (8).
The modalities of sleep disorders related to NAH were insomnia, RLS, bruxism and nightmares, which will be discussed here. First, insomnia is a very common complaint that is reported mostly among females (38). The link between insomnia and headache has been observed in many studies (6,36). Those who suffer from migraines are three times more likely to complain of insomnia symptoms (1). As with headache, insomnia may have a variety of causes. Patients with pain syndromes such as fibromyalgia have been found to have an increased alpha-delta ratio (39), which suggests that their insomnia may be due to an increased number of arousals during sleep, including awakening with headache. Secondly, primary headaches have been associated with various movement disorders (especially RLS), and this association is strong with migraine headaches (40). The clinical diagnostic criteria for RLS were reviewed and published in 2003 (41), showing that the symptoms impair the patient’s sleep and their ability to return to sleep after arousal. In a study that compares groups of patients with and without RLS suffering from headache, higher scores for depression and anxiety were more frequent in subjects with RLS (40). Complaints of morning headache and daytime headache were reported three to five times more frequently among RLS sufferers, and there was a tendency toward reported social isolation in these patients (42). In our study, we found that patients with RLS had NAH twice as frequently as those who did not have RLS. We also found that our patients with NAH suffered impairment in their social quality of life. Moreover, the association between RLS and migraine headache could support the hypothetical dopaminergic imbalance in both RLS and migraine (43). Dopamine is involved in the pathogenesis of both disorders, and it is responsible for the migraine dopaminergic premonitory symptoms (43). Thirdly, bruxism may also result in sleep disruption (44). In our study, bruxism was associated with NAH but lost statistical significance in the multivariate analyses, and was therefore not identified as a risk factor in our sample (Tables 3 and 5). In the literature, the relation between bruxism and headache is still controversial (44,45). Finally, nightmares might generate post-awakening anxiety followed by difficulties in returning to sleep (35). Migraineurs reported nightmares at a frequency four times that of non-migraineurs (33). One study found that dreams that culminate in migraine headaches were characterized by the presence of negative feelings (46). Subjects who frequently wake up with a headache are more likely to remember the contents of their dreams (35). This observation could explain why our study showed that complaints of nightmares had a high association with NAH (Table 3). Moreover, we found that patients with complaints of nightmares had NAH twice as frequently compared to those who did not complain of nightmares (Table 5).
The four modalities of sleep disorders that we have discussed (insomnia, RLS, bruxism and nightmares) are related to poor sleep. Therefore, it is not surprising that a relationship was found between these disorders and NAH. Sleep deprivation is another well-known cause of headaches (47). However, the association with complaints of poor sleep was not accompanied by the findings of PSG. Indeed, according to the PSG, there were no associations between complaints of NAH and sleep efficiency, sleep latency and sleep fragmentation. The only significant association we found between NAH and the PSG parameters was a lower REM sleep percentage. This association has already been described in studies that examine the PSG and headache (48), and may be due to a physiologic dysregulation that decreases headache threshold or otherwise increases vulnerability to headache (1). Indeed, pain threshold is reduced during REM sleep (49) and the patients with migraines experience increased pain.
The absence of a relationship between the prevalence of NAH and OSAS is intriguing. However, it has been reported that 12–41.7% of patients with nocturnal or morning headache have OSAS by PSG evaluation (36,37). Headache, especially morning headache has recently been suggested to be part of the clinical findings of OSAS (7,50). However, Goder and colleagues (51) prospectively looked for the presence of morning headache in patients with various sleep disorders. They concluded that this kind of headache was significantly more common in patients with various sleep disorders other than apnea (51). They suggest that changes in sleep architecture may play a role in the morning headaches experienced by patients with sleep disorders. Furthermore, morning headache related to OSAS was included in the International Classification of Headache Disorders, second edition (ICHD-II, 2004) (52) under the heading of “sleep apnea headache.” The diagnostic criteria for this type of headache include an occurrence of more than 15 days per month and characteristics of tensional headache that disappears within 30 minutes of waking. Once apnea is treated, the headache disappears within 72 hours (ICHD-II) (52).
One would expect impairment in daytime functioning in subjects with NAH because of their complaints of poor sleep. In fact, we found a relationship between NAH and diurnal somnolence, fatigue, different levels of anxiety (mild, moderate and severe) and depression (moderate and severe). This study suggests that awakening during the night with headache is an indicator of poor sleep and poor quality of life. Spierings and colleagues (53) observed more intense fatigue among female headache patients compared to age- and sex-matched controls.
Our study is not without shortcomings. The primary purpose of this community-based survey was to investigate sleep disorders. Therefore, we did not attain a full description of the headaches or assessments of pain localization and intensity. This limitation is also present in studies of morning headaches (8,10). In fact, this study analyzed the presence or absence of NAH, rather than a specific diagnosis according to ICHD-II (52). However, based on that classification, NAH might be (i) a type of primary headache, such as the hypnic headache; (ii) a sleep-related headache, as the “sleep apnea headache”; or even (iii) a manifestation of some type of known headache that might happen both during the day and at night. It seems to us that a new type of headache may be in the course of being described and that future studies will clarify how it better fits in the general headache classification.
An analysis of the types of headache which usually happen during the night shows that the cluster headache is a rare condition and more prevalent in men (54) and that the hypnic headache is a rare condition and is usually detected in older patients (6). In our study, the prevalence of NAH is greatest among women and 50–59-year-old subjects. A study that examined the prevalence of migraines in Brazil, which can also occur during the night, reported that they are more prevalent in women who are 40–49 years old, with an overall prevalence of 15.2% (30).
Another limitation of this study was that the relationship between NAH and sleep disorders would be better clarified in a cohort study rather than in a cross-sectional study like ours. An advantage of this study is that because we did not apply a specific questionnaire to describe headache characteristics, the results are less likely to have informational bias (55). Further examination of these cause-effect relations should also include evaluations of sleep pattern. For example, it appears that the association between NAH and insomnia is a bidirectional relationship: the presence of one increases the risk of having the other.
Finally, we suggest that a multiplicity of factors can be involved in the complaint of NAH. Physicians should be aware of the necessity of conducting an extensive patient interview to identify all of them. Our study shows that NAH is highly prevalent in the population of São Paulo City and suggests that this phenomenon has specific characteristics with specific risk factors, which are obesity, RLS and nightmares.
Footnotes
Acknowledgements
This work was supported by grants from the Associaçao Fundo de Incentivo a Psicofarmacologia (AFIP) and FAPESP (#07/50525-1 to RS-S, and CEPID no. 98/14303-3 to ST). ST, JAT, and LRAB received the CNPq fellowship. The authors would like to thank Fernando Colugnati for valuable suggestions and statistical analyses. All the efforts of AFIP’s staff, in particular those of Roberta Siuffi, are deeply appreciated.