Sex and the Risk of Restless Legs Syndrome in the General Population FREE

The clinical symptoms of restless legs syndrome (RLS) were first described by the English physician Sir Thomas Willis¹ in a chapter entitled "Instructions for Curing the Watching Evil" of his book The London Practice of Physick in 1685. His original description related the experience of patients: "The diseased are no more able to sleep, than if they were in a place of greatest torture."^1(p404) In later centuries, the condition was described with varying terms and generally was regarded as a form of hysteria or neurasthenia.² The Swedish neurologist Karl Ekbohm gave the comprehensive description a disease entity, "irritable legs," in 1944.³ However, it took another 50 years until standard criteria for the definition and diagnosis of RLS were published by the International Restless Legs Syndrome Study Group in 1995.⁴ These criteria include the desire to move the limbs usually associated with paresthesias or dysesthesias of the legs, a motor restlessness, a worsening or exclusive presence of symptoms at rest (lying or sitting) with at least partial and temporary relief by activity, and a worsening of the symptoms in the evening or during the night. Between 50% and 92% of patients with idiopathic RLS report a positive family history.⁵ Based on the analysis of large families, autosomal dominant inheritance is suggested.^6- 7 Symptomatic forms of RLS can occur during pregnancy, in uremic patients, and in those with diabetes mellitus or anemia.⁸

Initially classified as a peripheral nerve disorder, most authors in the last 2 decades agree that RLS has its origin in the central nervous system. There is evidence that alterations of the complex integration between peripheral and central nervous system structures, ie, an abnormal sensorimotor integration and enhanced spinal cord excitability, play a role in the onset of symptoms.⁹ The etiology of RLS remains unclear, although studies^8- 10 have examined different aspects of the pathogenesis of the disease. In contrast, epidemiological investigations have been limited. Several population studies^11- 14 applied unique definitions for the classification of RLS, and some studies^15- 18 used the standard minimal criteria to assess RLS prevalence. Using different selection criteria for the populations studied and varying diagnostic criteria in the classification of RLS, most epidemiological studies reported prevalences of RLS or of single RLS symptoms ranging between 5% and 15%. In 2 studies,^11,15 women were more likely to report RLS symptoms than men; other studies^13- 14 did not find a difference in prevalence, and 1 study¹⁶ examined only women. Based on these few population studies, the American RLS Foundation stated that "RLS is the most common unknown disease you have never heard of."^19(p82)

We assessed the prevalence of RLS and associated risk factors among participants in the Study of Health in Pomerania,²⁰ a large survey of the general population in northeastern Germany, using standard diagnostic criteria defined by the International Restless Legs Syndrome Study Group.

The Study of Health in Pomerania is a population-based survey conducted in the most northeastern part of Germany on the Baltic Sea.²⁰ The study area comprises 3 cities and 29 communities in West Pomerania and was part of former East Germany until the German reunification. From the total population of 212 157 who lived in the study area in 1997, a sample of 7008 persons of German nationality was drawn from the population registers. The 2-stage cluster sampling method was adopted from the World Health Organization MONICA Project, Augsburg, Germany,^21- 22 and yielded 12 strata comprising 5-year age spans (20-79 years) for both sexes, each including 292 individuals. From the sampled group, 6267 individuals were eligible and 4310 participated in the study, yielding a response rate of 68.8%. Two hundred three participants were excluded from the analysis because of missing data on RLS status. Hence, included in the present analysis were 2089 women and 2018 men with complete information on RLS status.

Data were collected between October 1, 1997, and October 31, 2000. All participants were interviewed face to face in 2 central study facilities. The following standardized questions were used to assess RLS status, according to the International Restless Legs Syndrome Study Group criteria⁴:

The 3 answer categories included: yes, no, or do not know. Participants were classified as being RLS-positive if they answered all 3 questions affirmatively. This set of questions had been validated against a physician diagnosis in prior investigations of the general population in Germany.^15,23 It is so far the only validated instrument to assess RLS in population studies.

General medical history and history of neurological diseases, including screening questions for parkinsonism and psychiatric diseases, were assessed in interview form by trained and certified interviewers (recruited from a pool of health care professionals). Any current medications taken within the last 7 days were listed.

Risk factor assessment was done with standardized methods or questions. Body weight and height of all participants were measured with shoes and heavy clothing removed. Body mass index was calculated as weight in kilograms divided by the square of height in meters. Under standardized conditions, blood pressure was measured 3 times with an automatic blood pressure monitor with measurement printout (HEM-705CP; Omron Corporation, Tokyo, Japan).²⁴Current hypertension was defined as blood pressure values of 160 mm Hg systolic or 95 mm Hg diastolic or higher. Diabetes mellitus was defined as self-reported physician-diagnosed diabetes mellitus plus treatment or as a nonfasting glucose level of 200 mg/dL (11.1 mmol/L) or higher. Several other comorbidities were assessed as self-reports, with specific questions asking for a physician diagnosis of the respective condition. In addition, we defined reduced renal function as a sex-specific glomerular filtration rate for creatinine (calculated with the Cockcroft-Gault²⁵ formula) that was 2 SDs below the mean of healthy study participants (no reported comorbidities) or as an elevated serum creatinine level (men, ≥1.5 mg/dL [≥132.6 µmol/L]; and women, ≥1.3 mg/dL [≥114.9 µmol/L]). Similarly, current anemia was defined as a hemoglobin level 2 SDs below the sex-specific mean of healthy study participants, following suggested cutoffs for population studies.²⁶

Nonfasting blood samples were drawn under standardized conditions from each participant and were analyzed for several laboratory variables. Specifically, serum glucose levels were measured using the hexokinase method in a cuvette (Hitachi 717), thyroid-stimulating hormone using a LIA-mat System 300 and a kit from BYK Sangtec Diagnostica (Dietzenbach, Germany), ferritin using an immunoturbidimetric assay from Roche Diagnostics (Mannheim, Germany), serum creatinine using the Jaffé method with the Crea-Kit (Boehringer Mannheim) in a cuvette, hemoglobin using the cyanmethemoglobin method on a COULTER T660 (COULTER Electronics GmbH, Krefeld, Germany), and hemoglobin A_1C using cation-exchange chromatography (high-performance liquid chromatography) with spectrophotometric detection (Diamat Analyzer, Bio-Rad Laboratories, Munich, Germany).

In female participants, detailed information about reproductive behavior, including number of pregnancies and births and estrogen use, was collected. Alcohol consumption was assessed by asking each subject how much beer, wine, or spirits he or she consumed on the previous workday and during the previous weekend. Answers were given in different units for portion servings of beer (1 serving, 0.50 L), wine (0.20 L), or spirits (0.02 L), reflecting the differences in common glass sizes in Germany for the 3 types of alcoholic beverages. Total alcohol intake was calculated by multiplying weekday consumption by 5 and adding this number to the weekend consumption. Total intake was converted to grams of alcohol using the following definitions: 1.00 L of beer equals 40.0 g of alcohol, 1.00 L of wine equals 100.0 g of alcohol, and 0.02 L of spirits equals 6.2 g of alcohol. A current smoker was defined as an individual reporting smoking 1 cigarette or more per day. Educational attainment was defined by the number of school years completed and was categorized into 3 levels: primary (<10 years), secondary (10-11 years), and tertiary (12-13 years). Self-perceived health status was assessed with the Medical Outcomes Study 12-Item Short-Form Health Survey by calculating the summary mental health and summary physical health scores.²⁷ In addition, the cognitive status of all participants 60 years and older was evaluated with the Mini-Mental State Examination.²⁸

Differences between RLS-positive and RLS-negative groups were analyzed unadjusted and adjusted for age and sex. Adjustments for continuous variables were made using the adjusted means procedure, following an analysis of variance model that included sex, age (continuously), and RLS status. For serum ferritin level, a log transformation was done before the analysis of variance, because of a skewed distribution. Adjustments in categorical variables were made by estimating the probability in each category in a logistic regression model that included sex, age (continuously), and RLS status. Logistic regression was also applied to calculate odds ratios (ORs) for factors associated with RLS, including variables in the final model that reached significance at P = .10 in univariate analysis. Two separate regression models were used, 1 for participants in the age range 20 to 59 years and 1 for those aged 60 to 79 years, while the adjustment made was identical.

For 4107 participants, complete information on RLS status was available; 203 participants (104 women) were excluded from the analysis because of missing data on RLS status. Figure 1 shows the prevalence of RLS among the Study of Health in Pomerania population, stratified by age and sex. The number of participants within each sex-specific age group varied between 219 and 396. The overall prevalence of RLS among men was 7.6%, with a steady increase from 3.0% among the group aged 20 to 29 years to 13.2% among the group aged 60 to 69 years. In women, the overall prevalence was 13.4%, increasing from 4.9% in the group aged 20 to 29 years to 19.4% in the group aged 50 to 59 years, and slightly decreasing in older age groups. In each age group, the prevalence of RLS among women was higher than among men. To differentiate between idiopathic and secondary forms of RLS, we calculated prevalences after exclusion of all participants with diabetes mellitus (8.0% of the study population), reduced renal function (3.2%), and anemia (3.2%). In this reduced study population of 3479 participants (Figure 1), we observed similar prevalences among younger age groups and small to moderate reductions among the subjects aged 60 to 79 years, indicating that the 3 conditions contribute to secondary forms of RLS in older persons.

Table 1 shows sociodemographic characteristics and behavioral risk factors for the study population, stratified by RLS status. Significant differences between subjects with and without RLS were observed for several variables. Restless legs syndrome subjects were significantly older, were more often female, and had a lower educational attainment than subjects without RLS. Not all differences in behavior-dependent factors were significant. Significant differences were reported among subjects with and without RLS in self-perception of general health and in the physical and (to a lesser extent) the mental health summary scores of the Medical Outcomes Study 12-Item Short-Form Health Survey. Because the physical score is affected by existing comorbidities, we additionally calculated age- and sex-adjusted means for subjects with and without RLS after exclusion of individuals with diabetes mellitus, reduced renal function, or anemia. The differences for both summary scores were still significant (physical health, 45.2 vs 49.4, P<.001; and mental health, 49.2 vs 52.2, P<.001), but less significant for the physical score. After additional exclusion of those with self-reported other severe comorbidities (cancer, myocardial infarction, or stroke), the differences between subjects with and without RLS remained significant and of the same magnitude.

We also observed significant differences for several comorbid conditions (Table 2). The higher prevalence of diabetes mellitus among subjects with RLS was confirmed by a higher percentage of RLS subjects with elevated hemoglobin A_1c (defined as values >7%) compared with those without RLS. The percentage difference among individuals with reduced renal function was attributable to age or sex, and no differences were observed in mean serum creatinine levels. Despite slightly lower ferritin levels among some subjects, no difference was found in the prevalence of anemia. We also calculated the prevalence of RLS among subjects with very low levels of ferritin, because iron deficiency may contribute to the occurrence of RLS.^29- 30 No prevalence increase was observed among women with ferritin levels of 20 ng/mL or lower (n = 334, comprising 9.4% of the sample). Among 53 men with low ferritin levels, however, an increased prevalence of RLS was seen compared with men with higher ferritin levels (18.9% vs 7.3%, P = .002). Four of these 10 cases among 53 men with low ferritin levels were classified as being anemic; none had diabetes mellitus or reduced renal function. The higher prevalence of thyroid disease among RLS subjects reflects a higher percentage of hyperthyroidism. The higher prevalence of myocardial infarction among RLS subjects was not explained by age, sex, or diabetes mellitus status. These observed higher prevalences of specific comorbidities among RLS subjects are related to the known fact that secondary forms of RLS can occur in populations with several comorbid conditions.^6,31

Next, we analyzed potential reasons for the observed sex difference in RLS prevalence. Figure 2 shows prevalences among age groups, stratified by the number of births among female study participants. To avoid categories based on small case numbers, age groups were further collapsed into 15-year age spans in the figure. The number of participants within each sex-specific age group varied between 409 and 597. While nulliparous women had prevalences similar to those among men up to age 64 years, the prevalence of RLS increased among parous women within the 15-year age groups with each additional birth in a dose-response pattern. The subgroup of nulliparous women aged 65 to 79 years, however, was small (n = 37, of whom 11 had RLS); consequently, the 95% confidence interval (21%-86%) for the prevalence of RLS was the widest among all age-parity strata. Additional analysis revealed that the high prevalence among this group of older nulliparous women could not be explained by high rates of the considered comorbidities.

Finally, an analysis of risk factors for RLS was done using multivariate logistic regression, first with a single model that was adjusted for all factors listed in Table 3. In this model, we observed significant effects of age, sex, and parity among women (data not shown). Because the findings in Figure 1 indicate an increased RLS prevalence until age 60 years, when the number of premenopausal and perimenopausal women has decreased to a minimum, we also explored a potential effect modification of RLS-associated risk factors by age, using a stratified analysis. Results of these more detailed analyses are shown in Table 3. In younger study participants (20-59 years), the OR for RLS increased significantly with age, but not in older individuals. Compared with men, younger women had a more than 2-fold increased OR for RLS, and it remained significantly elevated in older women. In the younger age group, the OR for RLS increased with the number of births among women, following a significant linear trend. In older women, this association was inverse and was not significant. Using nulliparous women instead of men as the reference group still revealed a dose-response relationship in the younger age group, with only slightly reduced ORs (women with 1 child, 1.54; 2 children, 2.37; and ≥3 children, 2.79; P = .002 for trend). In older women, the inverse, nonsignificant relationship remained unchanged. Further risk factor evaluation revealed that low educational attainment was significantly related to RLS status. This effect was stronger in participants aged 20 to 59 years. Separate analyses in men and women showed that this relationship was found in both sexes. Because low educational attainment might be related to the number of births among women, we additionally analyzed this association, stratified by level of educational attainment. Within each of the 3 levels of educational attainment, we observed an increase in risk of RLS associated with increasing number of children, compared with nulliparous women (OR, 1.40, P = .06 for trend across categories of parity in women with primary school education; OR, 1.36, P = .04 for secondary school education; and OR, 1.67, P = .10 for tertiary school education). The risk associated with low educational attainment was also not explained by higher rates of comorbidities among those participants with a primary school education. Restricting the analysis to individuals without diabetes mellitus, reduced renal function, anemia, or thyroid disease revealed the same linear increase in risk associated with parity and a higher risk for those with low educational attainment. The presence of diabetes mellitus was associated with RLS in older age groups only. Ex-smokers and current smokers had higher risks for RLS, reaching statistical significance in the latter group aged 20 to 59 years.

Our study found an overall 10.6% prevalence of RLS among this general population, with a steady increase in prevalence until age 60 years. Our results, which are based on the standard minimal criteria for RLS,⁴ support the statement of the American RLS Foundation¹⁹ that RLS is a frequent but undiagnosed disease. We observed a large difference in prevalence among women vs men and found strong evidence that parity is a major factor in explaining this difference. Our results also show that in studies of the general population secondary forms of RLS due to underlying diabetes mellitus, reduced renal function, or anemia contribute to the prevalence of RLS among older age groups.

The increase in risk associated with parity is a new finding. It is supported by the fact that RLS can occur during pregnancy and by the reported sex differences in several studies^11,15 indicating higher prevalences among women. The association with parity could represent several underlying biological mechanisms. Two are of special interest in this context. The severity of RLS symptoms and the excretion of sex hormones follow a circadian rhythm,^32- 33 with peaks for the former in the early morning hours (2-o'clock to 4-o'clock AM). Restless legs syndrome symptoms during pregnancy usually occur during the last trimester, when estrogen levels are highest. We also observed an association between RLS and level of educational attainment, which was explained by neither sex, higher comorbidity rates, nor parity. This finding is in line with results from a US study¹³ that found an association between low income, also an indirect measure of social status, and RLS. However, further inferences from this finding would be speculative.

One other study¹⁵ of the general population used the identical method to assess RLS as in our study, but was restricted to subjects aged 65 to 83 years. In this age range, the authors reported a female-male ratio of 2:1 and a lower prevalence of RLS among women compared with our study findings (13.9% vs 17.6%). This difference might be explained by a lower prevalence in their study of women with diabetes mellitus (10.0% vs 20.6%), which was most likely due to differences in the definition of diabetes mellitus used. In men, the prevalence was considerably lower than in our study (6.1% vs 12.8%). In a recent publication from Sweden, an RLS prevalence of 11.4% was observed among 200 participants in a small questionnaire survey.¹⁶ This study was restricted to women and applied the standard diagnostic criteria for RLS. A statewide telephone survey in Kentucky found an RLS prevalence of 10% among participants.¹³ Disease assessment in this study, however, was restricted to a single, self-reported question. Despite the differences in inclusion criteria for study participants, case ascertainment, and data collection, remarkable similarities with our study exist in observed characteristics of subjects with RLS. Their subjects with RLS also had a higher prevalence of self-reported diabetes mellitus and had a higher body mass index on average. More important, RLS was significantly associated with diminished self-reported general and mental health. In a Canadian survey, single symptoms suspicious for RLS were reported by 10% to 15% of 2019 participants.¹¹ In this study, women were also more likely to report RLS symptoms than men. In a study from Japan, RLS was assessed by self-administered questionnaires comprising a single question that asked about "sleep disturbances due to a creeping sensation or hot feeling in the legs."¹⁴ The authors found slightly higher prevalences of RLS among men compared with women and higher rates of insomnia among subjects who were classified as having RLS. Despite using different disease definitions, most of these studies are consistent in reporting high prevalences of about 10% for RLS or for single symptoms included in the standard diagnostic criteria. Two^11,15 of the 4 studies^11,13- 15 that included both sexes were consistent in finding RLS more often among women.

We found limited evidence that decreased ferritin levels are associated with higher prevalences of RLS, as suggested in studies^8,10 that were based on hospital or outpatient clinic patients with RLS. One study²³ examined this relationship on a population level, with results suggesting that subjects with RLS who come to medical attention with iron deficiency represent a specific subgroup of all subjects with RLS. However, because iron metabolism is related to the dopamine pathway and to the presence of sex hormones, the observed iron and ferritin deficiencies in single cases and in case series might reflect phenomena in more complex metabolic changes.

Our study has several strengths and limitations. It is based on a study sample drawn from the general population in a defined region using established epidemiological methods. All participants were interviewed by specifically trained interviewers. The methods to assess and classify RLS were standardized for all participants and included the standard diagnostic criteria of the International Restless Legs Syndrome Study Group.⁴ These methods were successfully used in a previous population study.¹⁴ In that study, also conducted among a German general population, an additional neurological examination on each individual was performed, and the RLS classification based on the 3 standardized questions was compared with the diagnosis by the examining neurologist. This comparison showed good interrater reliability. We were unable to assess RLS symptom frequency and potential consequences on sleep status. All comorbidities in our study were self-reported, but the specific questions asked for a physician diagnosis, and several conditions could also be classified by the respective blood variables, using established definitions. The population under study included a broad age range, and we were able to differentiate between idiopathic and several secondary forms of RLS. The overall response rate of 68.8% was fair. In general, participants in health surveys have more favorable risk factor profiles and lower comorbidities (health-participant bias) than nonparticipants.³⁴ Additional analysis of our population revealed that participants were slightly younger than nonparticipants among women (48.5 vs 54.6 years, P<.01), but not among men (50.6 vs 50.3 years, P = .64). No difference in participation existed between the sexes (women vs men, 69.4% vs 68.2%; P = .42). Because the observed prevalences of RLS increased with age and were higher among women, they might underestimate the true prevalence among this population.

Our results show that RLS causes considerable disease burden and has deleterious effects on quality of life.¹³ Because RLS can be effectively treated,^9,35 a better education of the general medical community toward greater awareness of the syndrome is necessary. We conclude that RLS is a common disease in the general population, affecting women more often than men. It is associated with reduced quality of life in cross-sectional analysis. Parity is a major factor in explaining the difference between the sexes, and this finding, if confirmed in other studies, may be an important step in further clarifying the etiology of the disease.

Corresponding author and reprints: Klaus Berger, MD, MPH, Institute of Epidemiology and Social Medicine, University of Münster, Domagkstrasse 3, 48129 Münster, Germany (e-mail: bergerk@uni-muenster.de).

Accepted for publication February 26, 2003.

This study was part of the Community Medicine Research Net of the University of Greifswald (available at: http://www.medizin.uni-greifswald.de/cm) and was funded by grant ZZ9603 from the Federal Ministry of Education and Research, Berlin, and the Ministers of Cultural and Social Affairs of the Federal State of Mecklenburg–West Pomerania, Schwerin.