Correlations among insomnia symptoms, sleep medication use and depressive symptoms

Authors


Yoko Komada, PhD, Department of Somnology, Tokyo Medical University, 6-7-1 Nishishinjuku, Shinjuku-ku, Tokyo 160-0023, Japan. Email: ykoma@tokyo-med.ac.jp

Abstract

Aim:  To elucidate the factors associated with insomnia symptoms and the use of sleep medication, and the correlations among insomnia symptoms, sleep medication use and depressive symptoms in the general population.

Methods:  This survey was conducted in a rural community of Japan. Questionnaires consisted of basic information, the Pittsburgh Sleep Quality Index, and a 12-item version of the Center for Epidemiological Studies Depression scale, and were administered to all community members aged 20 years or over. A total of 2822 respondents with valid answers were subjected to analysis.

Results:  Occurrence of insomnia symptoms appeared to be associated with advancing age and existence of depressive symptoms. The extent of sleep medication use in the entire sample was 9%, and the value in the subjects with insomnia symptoms was 26%. Sleep medication use in insomniacs was associated with female sex and advancing age as well as higher scores in subcomponents of both poor subjective sleep quality and prolonged delay of sleep onset. Depressive symptoms were worst in the group with insomnia symptoms using sleep medication, and were significantly lower in the group without insomnia symptoms using sleep medication.

Conclusions:  Our study revealed that female sex, advancing age, depressive symptoms, poor sleep quality, and prolonged delay of sleep onset appeared as risk factors for sleep medication use. Insomnia symptoms were suspected to act as an exacerbating factor for depressive symptoms. However, our findings suggested that appropriate use of sleep medication could reduce depressive symptoms in the subjects with insomnia symptoms.

INSOMNIA IS WELL known as being very common among the general population,1–3 and the prevalence of the disorder is estimated at between 10 and 15% of the general population.3,4 A World Health Organization collaborative study found the overall prevalence of insomnia in primary care settings to be as high as 26.8%.5 Moreover, community-based studies also showed that a large number of the general population copes with insomnia by using sleep medication.3 The rate in the general population for at least occasional sleep medication use is approximately 3–11%,6,7 and this rate has been reported to be much higher in the elderly population, ranging from 10 to 27%, with a preponderance of long-term medication use.6,7

From previous studies, several factors have been recognized as being associated with the occurrence of insomnia.2,8,9 However, there have been few studies on factors associated with sleep medication use. Moreover, several studies have suggested that insomnia is associated with increased risk for a major depressive disorder.10,11 However, the correlations among insomnia, sleep medication use, and depressive symptoms remain inconclusive.

The objective of this community-based study was to elucidate: (i) the associated factors for the existence of insomnia symptoms; (ii) the associated factors for the use of sleep medication; and to investigate (iii) the influence of sleep medication use on depressive symptoms in the general population.

METHODS

Subjects and procedure

The ethics committees of Tottori University approved this study, and all subjects gave their informed consent to take part in this investigation.

This survey was conducted as a part of the health-care program of Daisen, a rural community in the Tottori prefecture, in the western part of Japan. The total population of the town was 6643 in 2004 with 5528 residents aged 20 years or older (2521 men and 3007 women). Major industries in this area include agriculture, and tourism.12 This questionnaire survey13,14 was conducted from November 2005 to January 2006. With the cooperation of local public health nurses, questionnaires were delivered to all residents who were 20 years of age or older. The questionnaires requested information regarding occupation, family constitution, existence of family members who needed nursing care, existence of any disease currently treated, including hypertension, diabetes mellitus, ischemic heart disease, arrhythmia, respiratory disease, cerebrovascular disease, orthopedic disease and hormonal disease, as well as smoking and drinking habits. We used a 12-item version of the Center for Epidemiological Studies Depression scale (CES-D) for estimating depressive symptoms,15 and a Japanese version of the Pittsburgh Sleep Quality Index (PSQI)16 for estimating sleep disturbance in the subjects. The CES-D had four response options ‘never or rarely’, ‘sometimes’, ‘often’, and ‘always’ coded 0–3, respectively. We used total scores of CES-D as parameters of depressive symptoms. We divided responses into three categories: 0–11 as normal, 12–20 as moderate, and 21–36 as severe.17 The PSQI included sub-items evaluating sleep quality (C1), sleep latency (C2), sleep duration (C3), habitual sleep efficiency (C4), frequency of sleep disturbance (C5), use of sleeping medication (C6), and daytime dysfunction (C7). Among these, C1–C5 indicate problems with nocturnal sleep.16 We defined greater than 1 SD of the mean PSQI C1–C5 sum scores as indicating insomnia symptoms.

Of the 5528 eligible subjects, 2937 (53.1%) responded to the questionnaires, and there were 2822 (51.0%) valid responses (1222 men, 1600 women, age ranging from 20 to 97 years old; mean [SD]: 57.4 [17.7] years old).

The mean family size in this cohort was 4.6 (1.6); 120 subjects (4.3%) lived alone, whereas 1210 responders (43.6%) lived with more than five family members. Those with habitual alcohol ingestion and a smoking habit were 39.0% and 21.9%, respectively.

Statistical analyses

A one-way anova was used to compare the PSQI scores among three groups of subjects with depressive symptoms (i.e. normal, moderate, and severe). We also performed an anova to compare PSQI score and CES-D score among four groups categorized by both status of sleep medication use and self-evaluation of sleep (i.e. normal non-medicated, insomnia non-medicated, normal medicated, and insomnia medicated). When significant differences were found using the anova, a Bonferroni–Dann post-hoc analysis was conducted.

χ2 analysis was used to explore the status of sleep medication use by each age group.

The factors associated with insomnia symptoms and sleep medication use were examined with the aid of a series of logistic regression analyses. All variables were initially examined in univariate models. To control confounding factors and to determine the main correlates, we then performed multivariate logistic regression analyses for all variables that showed a significant correlation in univariate models. Statistical tests of the regression estimates odds ratios (OR) were based on Wald statistics. Odds ratios and their 95% confidence intervals (CI) are presented to show the association. All of the analyses were performed using the StatView 5.0 software package (sas Institute Inc., Cary, NC, USA), and statistical significance was set at P < 0.05.

RESULTS

Factors associated with insomnia symptoms

Mean (SD) score of the PSQI (C1–C5) was 3.5 (2.1) in the total sample. We determined the subjects having insomnia symptoms using 1 SD of the mean PSQI score (C1–C5) as the cut-off point in this study. As a result, subjects with insomnia symptoms were confirmed when one showed 5.6 PSQI points or more (for C1–C5). Mean (SD) score was 7.1 (1.3) in the group with insomnia symptoms, while the value was 2.8 (1.4) in the group without insomnia symptoms. Univariate logistic regression analyses were performed for eight independent variables: sex, age, CES-D score, existence of disease currently treated, smoking habit, habitual alcohol ingestion, living alone, and home nursing care. Among these variables, three items (age, CES-D score, and existence of diseases currently treated) exhibited significant correlations with the existence of insomnia symptoms determined by the criteria indicated above. To control for confounding factors and to determine the main correlates of insomnia symptoms, three significant variables in the univariate models were submitted to a multivariate model. Adjusted OR and the 95%CI in the final model are shown in Table 1. Multivariate logistic regression analysis revealed that the existence of insomnia symptoms was significantly associated with increasing age (OR = 1.41, 95%CI: 1.07–1.84), and CES-D score (OR = 1.19, 95%CI: 1.16–1.22).

Table 1.  Univariate and multivariate logistic regression results for associated factors with insomnia symptoms
 TotalInsomniacsUnivariate relative risk (95%CI)PMultivariate relative risk (95%CI)P
nn%
  • Total t score of Pittsburgh Sleep Quality Index (C1–C5) is above 1 SD of the mean score.

  • Relative risks approximated with odds ratios.

  • §

    The age category was divided at the median age of 58 years.

  • CES-D, Center for Epidemiological Studies Depression Scale; CI, confidence intervals; NS, not significant.

Sex         
 Male122217414.2      
 Female160024515.3  NS  NS
Age§         
 <58 years135718613.7      
 ≥58 years146523315.91.35(1.10–1.67)<0.011.41(1.07–1.84)<0.05
CES-D2483  1.19(1.16–1.21)<0.00011.19(1.16–1.22)<0.0001
Disease currently treated         
 No182224113.2      
 Yes100017817.81.46(1.18–1.81)<0.0011.26(0.97–1.64)= 0.09
Smoking habit         
 No217431714.6      
 Yes60910016.4  NS  NS
Habitual alcohol ingestion         
 No169725715.1      
 Yes108316014.8  NS  NS
Living alone         
 No265839014.7      
 Yes1202319.21.590.99–2.57= 0.06  NS
Home nursing care         
 No231234114.7      
 Yes4316515.1  NS  NS

Correlations between depressive symptoms and insomnia symptoms

Figure 1 shows a comparison of the severity of insomnia symptoms manifested in the PSQI C1–C5 sum scores among three depressive symptom groups. The sum score indicating the severity of insomnia symptoms was significantly different between the three groups (F(2,2481) = 279.7, P < 0.0001). The ‘severe’ group showed a significantly higher value of the sum score compared to the ‘normal’ group (P < 0.01) and the ‘moderate’ group (P < 0.0001). The ‘moderate’ group also showed significantly higher value than the ‘normal’ group (P < 0.0001). The same result was obtained when the item for insomnia was excluded from the total CES-D score (F(2,2480) = 145.7, P < 0.0001).

Figure 1.

Comparison of total score of C1–C5 of the Pittsburgh Sleep Quality Index (PSQI) among three depressive symptom groups (normal: n = 1899, moderate: n = 523, severe: n = 62).

Extent of sleep medication use and factors associated with the use of sleep medication

Figure 2 shows the proportion of subjects with sleep medication use in each age group for all the subjects in the study population (upper panel), as well as the proportion for insomniacs (lower panel). The extent of sleep medication use for all subjects was 9% and that for insomniacs was 26%. The proportion using sleep medication more than once a week was 9% for those in their 60s, 14% for those in their 70s, and 18% for those in their 80s, while the value remained 3% or less for those in their 20s–40s. When limited to insomniacs, the proportion of subjects aged 20–59 using sleep medication more than once a week was less than 15%. However, the value was more than 30% for subjects aged 60 years and older, and even 45% for subjects in their 80s. There are significant differences in the frequency of sleep medication use among age groups both in the total sample (χ2[18] = 133.1, P < 0.0001) and in the subjects with insomnia symptoms (χ2[18] = 53.2, P < 0.0001).

Figure 2.

Percentage of sleep medicine use among total subjects population (upper panel), and among insomniacs (lower panel) in each age group. **P < 0.01, *P < 0.05 from the result of χ2 and the residual analysis. (inline image) More than three times a week; (inline image) Once or twice a week; (inline image) Less than once a week; (□) None.

Univariate logistic regression analyses were performed for 13 independent variables, and nine of them exhibited significant correlations with sleep medication use in insomniacs as shown in Table 2. To control for confounding factors and to determine the main correlates of sleep medication use, all significant variables in the univariate models were submitted to multivariate analysis.

Table 2.  Univariate and multivariate logistic regression results for associated factors with sleep medicine use among insomniacs
 TotalMedicine useUnivariate relative risk (95%CI)PMultivariate relative risk (95%CI)P
nn%
  • Total Pittsburgh Sleep Quality Index score (C1–C5) is above 1 SD of mean score.

  • Relative risks approximated with odds ratios.

  • §

    The age category was divided at the median age of 58 years.

  • CES-D, Center for Epidemiological Studies Depression Scale; CI, confidence intervals; NS, not significant.

Sex         
 Male1742816.1      
 Female2456526.51.91(1.16–3.14)<0.052.05(1.02–4.11)<0.05
Age§         
 <58 years1861910.2      
 ≥58 years2337431.84.10(2.36–7.12)<0.00012.49(1.19–5.23)<0.05
CES-D409  1.09(1.04–1.14)<0.0011.06(1.00–1.13)<0.05
Disease currently treated         
 No2413213.3      
 Yes1786134.33.42(2.10–5.56)<0.00012.39(1.29–4.44)<0.01
Smoking habit         
 No3177824.6      
 Yes1001313.0  NS  NS
Habitual alcohol ingestion         
 No2576826.5      
 Yes1602314.4  NS  NS
Living alone         
 No3908521.8      
 Yes23626.1  NS  NS
Home nursing care         
 No3416920.2      
 Yes652015.1  NS  NS
Pittsburgh Sleep Quality Index         
 C1: subjective sleep quality419  1.65(1.08–2.51)<0.052.46(1.44–4.21)<0.01
 C2: sleep latency419  2.96(2.01–4.34)<0.00012.44(1.53–3.89)<0.001
 C3: sleep duration419  0.53(0.39–0.71)<0.0001  NS
 C4: habitual sleep efficiency419  1.40(1.14–1.73)<0.01  NS
 C5: sleep disturbances419  1.87(1.27–2.77)<0.01  NS

Multivariate logistic regression analysis revealed that sleep medication use in insomniacs was significantly associated with female sex (OR = 2.05, 95%CI: 1.02–4.11), advancing age (OR = 2.49, 95%CI: 1.19–5.23), CES-D score (OR = 1.06, 95%CI: 1.00–1.13), existence of a currently treated disease (OR = 2.39, 95%CI: 1.29–4.44); C1, subjectively poor sleep quality (OR = 2.46, 95%CI: 1.44–4.21); and C2, long sleep latency (OR = 2.44, 95%CI: 1.53–3.89).

Difference in both insomnia symptoms and depressive symptoms for the groups categorized by both sleep medication use and insomnia symptoms

Figure 3 shows the comparison of the PSQI C1–C5 sum score, and CES-D score for the four groups categorized by both insomnia symptoms and sleep medication use. The PSQI C1–C5 sum score was significantly different between the groups (F(3,2475) = 617.4, P < 0.0001). Post-hoc analysis revealed that there were significant differences among each group. The value was lowest in the normal non-medicated group, and was highest in the insomnia medicated group. The value of the normal medicated group was significantly higher than that of the normal non-medicated group, and the value of the insomnia medicated group was significantly higher than that of the insomnia non-medicated group. The CES-D score was also significantly different between the groups (F(3,2431) = 131.8, P < 0.0001). Post-hoc analysis revealed that the CES-D score was highest in the insomnia medicated group, while the value was lowest in the normal non-medicated group. There was no significant difference in the score between the insomnia non-medicated group and the normal medicated group. The same result was obtained when the item for insomnia was excluded from the total CES-D score (F(3,2431) = 92.8, P < 0.0001). When the presence of insomnia was set at a cut-off value of 2 SD of mean PSQI scores, there was no significant difference in PSQI score between insomnia non-medicated and insomnia medicated in post-hoc analysis. As for CES-D score, the result was the same in both 2 SD and 1 SD definition.

Figure 3.

Comparison of Pittsburgh Sleep Quality Index (PSQI) C1–C5 score (left panel) and the Center for Epidemiological Studies Depression Scale (CES-D) score (right panel) among the four groups categorized by both insomnia symptoms and sleep medication use.

DISCUSSION

It has been suggested that common factors associated with insomnia symptoms are female sex, advancing age, and depressive symptoms.2,18,19 Consistent with previous reports, our findings revealed that the occurrence of insomnia symptoms was significantly associated with both advancing age and existence of depressive symptoms. Undoubtedly, age-related changes in both sleep physiology and lifestyle act as important factors for the occurrence of insomnia symptoms.2 On the other hand, there was no significant correlation with female sex. The reason for this phenomenon is not clear. However, considering that female sex was not associated with an increased prevalence of insomnia symptoms in a general population study in Japan,1 racial difference might be one possible reason for the negative association of the sex factor. Another possibility is a difference in the living environment. In the geographical area of this study, many women were engaged in agriculture and tourism. It is possible that they could keep a balance between their work and household responsibilities, because they are able to return home early in the evening.

Our findings suggest that depressive symptoms were significantly associated with the occurrence of insomnia symptoms, and the score for insomnia symptoms increased with elevation of the severity of depressive symptoms. This finding is consistent with the findings of Kaneita et al., who observed a linear inverse association between subjective level of sleep satisfaction and CES-D scores.20 As several studies have provided data suggesting that insomnia are risk factors for the development of depression,10,21 it can be said that depressive symptoms and insomnia can mutually become each other's cause and result. Therefore, it is important to clarify the causal correlations in a follow-up study in the same cohort.

The present study indicated that the extent of sleep medication use in the total sample was 9%, and that elderly subjects accounted for a high proportion sleep medication use. When limited to insomniacs, the rate of medication use in order to sleep was 26% across the sample. The proportion of subjects who took sleep medication more than once a week was 30% among insomniacs aged 60 or older, but 10% for insomniacs aged 59 or younger. The frequency of both insomnia symptoms and sleep medication use clearly increased in the elderly. The estimated extent of sleep medication use in the general Japanese adult population is 3.5% in men and 5.4% in women.22 In reports from Western nations, the rate of sleep medication use was 6.9%.2,18 Compared with the findings described above, the rate of sleep medication use in this study is clearly higher. We expect that the subject population of our study included a larger percentage of elderly subjects compared with previous studies. It is well known that adverse events due to sleep medication, such as residual sedation, and psychomotor impairment, respectively, appear more frequently among the elderly than in the younger population.23 We want to emphasize the necessity of both evaluation of insomnia symptoms and alternative approaches, such as sleep hygiene education, and cognitive behavioral therapy,24 especially for elderly insomniacs.

Our study revealed that sleep medication use showed a positive correlation not only with female sex, but also with increasing age, which agrees with previous reports.18,25 Furthermore, sleep medication use is significantly associated with the existence of currently treated diseases,25 and higher depressive symptoms. Regarding existence of current disease, several studies have shown the significant association between sleep disturbance and comorbid illness.26,27 In addition, it is feasible that the distress of physical illness could be strengthened due to insomnia, or that physicians, who are engaged in the treatment of patient's current disease, are likely to prescribe sleep medication. With regard to depressive symptoms, it is plausible that physicians could prescribe sleep medication together with an antidepressant as a treatment for depressive symptoms. Moreover, the occurrence of depressive symptoms secondary to insomnia might increase the insomniac's motivation to seek medical assistance, as has been suggested in previous studies.9,18

To the best of our knowledge, this is the first study to examine the associated factors for sleep medication use by using the PSQI. Interestingly, our findings showed that deterioration of both subjective sleep quality (C1) and sleep latency (C2) PSQI scores were significantly associated with sleep medication use. On the other hand, sleep duration (C3), habitual sleep efficiency (C4), and frequency of sleep disturbance (C5), were not associated with sleep medication use in the subjects with insomnia symptoms. These results were generally consistent with a previous study using another questionnaire, which showed that dissatisfaction with both sleep onset period and sleep quantity was associated with psychotropic consumption.25 Another study found that waking up during the night per se was not reflected in motivation for starting sleep medication.7 Taking our findings together with those two reports, we speculate that it is not dissatisfaction with sleep duration or sleep efficiency, but rather difficulty in sleep initiation and/or dissatisfaction with sleep quality that increase motivation for the use of sleep medication among the insomniac population.

Notably, the depressive score was worst in the insomnia medicated group, and the value was significantly lower in the normal medicated group. This difference could imply that sleep medication use brings a certain decrease in depressive symptoms. This finding is in line with a recently published study indicating that eszopiclone/fluoxetine co-therapy was associated with a greater magnitude of the antidepressant effect.28 However, our result suggested that a considerable number of the subjects in the insomnia medicated group were not cured of their symptoms, despite the sleep medication use. Therefore, it is possible that treatment for insomnia and/or depressive symptoms in the object region was inadequate, although sleep medication use can bring a certain improvement in depressive symptoms. There could also be another possibility that insomniac subjects with medication involve more patients with ‘true’ major depression than non-insomniacs with medication.

In this study, the depressive score in the normal medicated group was significantly higher than that in the normal non-medicated group. Interestingly, the PSQI C1–C5 sum score in the former group were shown to be small but significantly higher than those in the latter group, although the value of both groups remained within the normal range. Due to these phenomena, the possibility that prescription drug users had greater disability, neuroticism, and daytime fatigue than the others29 should be taken into consideration. It is also possible that the usage of hypnotics might reflect the higher depressive score in the medicated groups, as the subjects with sedative-hypnotics dependence are likely to have mood disorders.30

This study had some limitations. First, the adequacy of the definition of insomnia symptoms must be considered. We categorized subjects having insomnia symptoms as those who had 5.6 points or more on the PSQI (C1–C5) score. The Japanese version of the PSQI was developed by Doi et al., and a mean score for each component in both their control group and primary insomnia was indicated in their report.16 The mean sum score (SD) of the PSQI (C1–C5) in a control group was 2.98 (2.75) and 6.42 (3.95) in a sample with primary insomnia.16 The values obtained in the present study were 2.77 (1.39) in the group without insomnia symptoms and 7.14 (1.27) in the insomniac group. These values seemed to be almost approximated. Second, we used a 12-item version of CES-D for estimating depressive symptoms. This questionnaire includes an item asking about the severity of insomnia. We had to use the total score of the scale in this study, so that we could refer to an authorized cut-off value for the definition of the existence of depressive symptoms on CES-D.17 However, we confirmed that the same result was obtained when we excluded the item for insomnia from the total CES-D score. Third, it remains possible that other sleep disorders, such as restless legs syndrome and sleep apnea syndrome (SAS), are related to insomnia symptoms and/or depressive symptoms. We have already reported that RLS-positive subjects had a significantly higher score of both PSQI and CES-D than those of RLS-negative subjects.13 However, further study for clarifying the correlations among SAS, insomnia and depressive symptoms should be made. Finally, the kind of medication could not be specified in this study from the result of PSQI, in which the item ‘How often did you take medicine to help your sleep during the past month?’ was set.

In summary, we found that about 10% of people 20 years of age or older who live in rural Japan use sleep medication. The existence of insomnia symptoms seemed to be associated with advancing age and existence of depressive symptoms. Sleep medication use among insomniacs was significantly associated with female sex and increasing age as well as a high score of some specific subcomponent of the PSQI (poor subjective sleep quality and prolongation of sleep onset). Sleep hygiene education and/or cognitive behavioral therapy are thought to be beneficial for the people with these characteristics. However, our study showed that depressive symptoms among the study population were worst in the insomniac subjects who used sleep medication, and that the symptom score of this group was significantly higher than those in non-insomniacs who used sleep medication. From these findings, we want to emphasize that adequate use of sleep medication could act to reduce depressive symptoms in insomniacs.

ACKNOWLEDGMENTS

The authors are indebted to Professor J. Patrick Barron of the International Medical Communication Center of Tokyo Medical University for his review of this manuscript.

Ancillary