Search tips
Search criteria 


Logo of amjepidLink to Publisher's site
Am J Epidemiol. 2010 March 15; 171(6): 709–720.
Published online 2010 February 18. doi:  10.1093/aje/kwp454
PMCID: PMC2842222

Prospective Associations of Insomnia Markers and Symptoms With Depression


Whether insomnia, a known correlate of depression, predicts depression longitudinally warrants elucidation. The authors examined 555 Wisconsin Sleep Cohort Study participants aged 33–71 years without baseline depression or antidepressant use who completed baseline and follow-up overnight polysomnography and had complete questionnaire-based data on insomnia and depression for 1998–2006. Using Poisson regression, they estimated relative risks for depression (Zung scale score ≥50) at 4-year (average) follow-up according to baseline insomnia symptoms and polysomnographic markers. Twenty-six participants (4.7%) developed depression by follow-up. Having 3–4 insomnia symptoms versus none predicted depression risk (age-, sex-, and comorbidity-adjusted relative risk (RR) = 3.2, 95% confidence interval: 1.1, 9.6). After multiple adjustments, frequent difficulty falling asleep (RR = 5.3, 95% confidence interval: 1.1, 27.9) and polysomnographically assessed (upper or lower quartiles) sleep latency, continuity, and duration (RRs = 2.2–4.7; P’s ≤ 0.05) predicted depression. Graded trends (P-trend ≤ 0.05) were observed with increasing number of symptoms, difficulty falling asleep, and difficulty returning to sleep. Given the small number of events using Zung ≥50 (depression cutpoint), a limitation that may bias multivariable estimates, continuous depression scores were analyzed; mean values were largely consistent with dichotomous findings. Insomnia symptoms or markers increased depression risk 2.2- to 5.3-fold. These results support prior findings based on self-reported insomnia and may extend similar conclusions to objective markers. Heightened recognition and treatment of insomnia may prevent subsequent depression.

Keywords: depression, polysomnography, prospective studies, sleep, sleep initiation and maintenance disorders

To date, studies of insomnia and depression have largely conceptualized insomnia as a clinical correlate or component of depression—that is, as a prodromic sign or symptom (1, 2), a precursor (3, 4), or its most persistent residual symptom after treatment (5). Insomnia's strong cross-sectional association and high comorbidity with depression and other mood disturbances have been well-established in epidemiologic studies (612). However, insomnia is increasingly being recognized as a possible contributor to the development of depression (13).

Elucidating insomnia's prospective relation to subsequent depression could lead to improved recognition, treatment, and prevention of depression, which is projected to rank as the second-leading cause of lost disability-adjusted life years after ischemic heart disease by 2020 (14). Both insomnia (15) and depression (16, 17) are highly prevalent, persistent, undertreated, and costly (18) disorders related to significant functional impairment and disability (1719), including increased absenteeism (15, 20) and health-care utilization (9, 17, 19).

Prior prospective studies have found that self-reported insomnia predicts depression in varied populations (3, 12, 13, 2129), including younger (12, 26, 27) and older (13, 28, 29) adults. However, with scant exceptions (12, 22, 2528), most studies of insomnia and depression have been limited to shorter follow-up periods—1 year (3, 9, 10, 21, 23, 24) or 2 years (29, 30)—and all have been limited to self-reported assessment of insomnia. Furthermore, while electroencephalographic sleep abnormalities reflecting insomnia (3134) have been cross-sectionally correlated with depression in psychiatric populations, prospective relations of polysomnographic insomnia markers to depression have not been studied in clinical or population-based samples. Thus, insomnia's longitudinal consequences are still poorly understood (35).

We investigated whether objective markers and self-reported symptoms of insomnia would predict depression symptoms at 4-year follow-up in a large population-based sample that was free of depression at baseline. We hypothesized that polysomnographic markers of insomnia—namely, decreased sleep duration, poorer sleep continuity, longer sleep latency, and increased insomnia symptom frequency—would predict the incidence of depression in a graded (“dose-response”) fashion.


Participants and data collection

The study procedures were approved by the University of Wisconsin-Madison Health Sciences Institutional Review Board. The sample comprised participants enrolled in the ongoing Wisconsin Sleep Cohort Study, previously described (36). Participants provided signed informed consent. A stratified random sample of previously surveyed Wisconsin state employees (n = 2,884) was invited to participate in a baseline overnight sleep study, and 53% (n = 1,533) agreed; the primary reported reason for nonparticipation was the burden of sleeping overnight in a sleep laboratory. Overnight protocols included nocturnal polysomnography at participants’ usual sleep times; clinical assessments, including calculation of body mass index (weight (kg)/height (m)2); administration of a depression scale; and completion of a questionnaire regarding health history, sleep habits, and insomnia symptoms. Approximately every 4 years after baseline studies, participants are invited to undergo follow-up examinations; by June 2008, there was an average follow-up participation rate of 80%.

Of 1,533 cohort participants, 787 were initially eligible for the present analyses, since they had completed overnight polysomnography between 1998 and 2002 and had undergone follow-up polysomnography 3–5 years later (average, 4 years) and had complete baseline and follow-up data on depression and self-reported insomnia for the period 1998–2006. Insomnia symptoms were first assessed during overnight protocols in 1998. For examination of incident depression, participants who had symptoms of depression (Zung score ≥50) or were taking antidepressant medication at baseline (n = 192) were excluded from the current analyses, which left 595 participants at risk for subsequent depression. Additionally, 40 participants who, at follow-up, were on antidepressant medication yet reported no depression symptoms (Zung score <50) were excluded from analyses, to limit potential bias related to overestimation of insomnia's association with depression. Thus, the final inception cohort for this analysis constituted 555 participants.


Self-reported symptoms.

The health questionnaire included 4 items on insomnia: difficulty in getting to sleep (referred to as difficulty in falling asleep or initiating sleep), waking up repeatedly during the night (repeated nocturnal awakenings), waking up too early in the morning and being unable to get back to sleep (awakening too early), and waking up during the night and having a hard time getting back to sleep (difficulty getting back asleep). Response categories were never or rarely (once/month), sometimes (2–4 times/month), often (5–15 times/month), and almost always (16–30 times/month). Each item was dichotomized into often/almost always (≥5 times/month) versus sometimes/less (<5 times/month), and then a number-of-symptoms (0, 1, 2, 3, or 4) variable was created at the frequency of often/almost always. Another 4-level variable was created for individual symptom frequency: 0 represented having all symptoms never/rarely; 1, having the symptom of interest sometimes; and 2, having the symptom of interest often/almost always. The category “any other symptom sometimes or more often” differentiated the symptom of interest from other symptoms.

Polysomnographically assessed markers.

Data on sleep latency, waking after sleep onset, sleep efficiency, and total sleep time were obtained during the overnight protocol by means of full 18-channel polysomnography (Grass Heritage PSG Digital Sleep System with model 15A54 amplifiers; Grass Technologies, West Warwick, Rhode Island) including electrooculography, electroencephalography, and electromyography. Sleep stage for each 30-second epoch was scored by technicians, according to conventional criteria (37). “Sleep latency” was defined as amount of time (minutes) from “lights off” to the first of 3 consecutive epochs of stage 1 sleep or the first epoch of any other stage of sleep; “waking after sleep onset” as the amount of time (minutes) spent awake after first sleep onset; “total sleep time” as the total amount of time spent sleeping (minutes); and “sleep efficiency” (%) as the proportion of total sleep time out of total duration of time in bed from “lights out.”

Outcome: depression

On the same evening as the overnight studies, participants completed the Zung Self-Rating Depression Scale (38) and provided information on regularly taken medications, including antidepressants (selective serotonin reuptake inhibitors, tricyclic antidepressants, monoamine oxidase inhibitors, or other). Scores on the 20-item Zung scale range from 25 to 100. Scores less than 50 indicate “normal” (not depressed), and scores greater than or equal to 50 are considered indicative of depression (37). We used this standard cutpoint for our population-based sample rather than a higher threshold (perhaps better-suited for a clinical sample). The original instructions, “how you have been feeling in the past few days,” were modified, and participants were instructed, “Please read each of the following statements and place an X in the box which best describes how you feel in general from day to day.” For these analyses, we used a “modified” Zung scale that excluded 2 items (“I have trouble sleeping through the night” and “I get tired for no reason”) which may indicate insomnia and result in a built-in association between insomnia and depression. After exclusion of these items, the score was rescaled for comparability to the original 20-item Zung scale, as previously reported (39).

Covariates included age, sex, clinically assessed body mass index, and variables from a self-administered health questionnaire: moderate/greater alcohol consumption (≥2 daily drinks of beer, wine, or hard liquor vs. less); cigarette smoking (current, past, or never); caffeine consumption (0, 1–2, 3–4, or ≥5 caffeinated beverages daily); use of hypnotic agents; and self-report of any physician-diagnosed chronic illness (coronary artery disease, congestive heart failure, angina, hypertension, stroke, diabetes, asthma, emphysema, thyroid condition, epilepsy, arthritis, or back pain). Sleep-disordered breathing status was assessed by polysomnography and summarized as numbers of apnea events (complete cessation of nasal and oral airflow for >10 seconds) and hypopnea events (decrease in respiratory effort associated with a >4% drop in arterial oxygen saturation) per hour of sleep (apnea-hypopnea index). Using conventional cutpoints (37), apnea-hypopnea index was categorized as <5 (little or no sleep-disordered breathing), 5–<15 (mild), or ≥15 (moderate or worse).

Statistical analysis

In descriptive analyses, we examined average continuous baseline Zung scores in participants with Zung scores less than 50 to ensure that associations between insomnia and depression were not attributable to participants with insomnia symptoms or markers having subthreshold depression (higher scores near 50). Using analysis of variance, we tested for any (F test) or pairwise (t tests) significant differences between mean continuous Zung scores less than 50 (at baseline) and objectively measured (sleep latency and quartiles of total sleep time) and self-reported (number of symptoms) insomnia.

In unadjusted analyses, categorical differences were tested using the Pearson χ2 test and the Mantel-Haenszel χ2 trend test. Regression models were used to estimate adjusted relative risks and 95% confidence intervals for depression with respect to insomnia at baseline (1998–2002). Persons with incident depression symptoms were included in all analyses, irrespective of antidepressant use at follow-up. Depression at follow-up was only assessed at 4 years, on average. Analyses, carried out with Proc Genmod in SAS (SAS Institute Inc., Cary, North Carolina), assumed a Poisson distribution with robust error variances (40), a log-link function, and an unstructured variance-covariance structure. A P value less than or equal to 0.05 (2-sided) was considered significant.

Relations of polysomnographic markers to subsequent depression were initially assessed using quartile groupings obtained from the overall cohort (quartile cutpoints (25th percentile, median, 75th percentile): sleep latency (minutes)—4, 8, and 14; waking after sleep onset (minutes)—31, 48.5, and 78.5; sleep efficiency—78%, 85%, and 91%; total sleep time (minutes)—334, 374.5, and 409). Because linear relations were present only for total sleep time, we constructed binary variables for the remaining markers. For sleep latency and waking after sleep onset, the uppermost quartiles (≥14 minutes and ≥79 minutes, respectively) represented marker presence; the other quartiles combined constituted the reference categories. Sleep efficiency was represented by the lowest quartile (<78%) and its referent by the upper quartiles combined. Given total sleep time quartiles’ graded relation to depression and the significant (P ≤ 0.02) linear relation observed when total sleep time was examined continuously, its quartiles were not combined.

Adjusted models included potentially confounding variables in 2 stages: 1) age, sex, and chronic health conditions; and 2) alcohol, smoking, caffeine, hypnotics, and body mass index. Two-way interactions of insomnia with age and sex were assessed.

Linear trends (with insomnia examined as a continuous variable) in the logarithms of relative risks were tested for significance in Poisson regression models. In individual insomnia symptom models, linear trends were tested for the reference, symptom-of-interest “sometimes,” and symptom-of-interest “often/always” categories (the “other symptom” category was excluded from trend tests) with adjustment for age, sex, and comorbidity but not full adjustment, because of nonconvergence of insomnia categories in these models.

In secondary analyses, after confirming the normality of continuous depression scores at follow-up and of depression's residuals, we conducted linear regression analysis to examine continuous depression scores (modified Zung scale) at follow-up for every insomnia variable modeled separately, with age, sex, and comorbidity adjustments as in the discrete-outcome (Zung score ≥50) models. Persons with depression (according to the modified Zung scale or antidepressant use) at baseline were excluded from these analyses; thus, all analyses were conducted in the same inception sample (n = 555). Overall F tests for insomnia variables were conducted first; if results were significant, individual t tests were conducted.

We examined sleep-disordered breathing as a potential confounder by adjusting for apnea-hypopnea index in regression models. Furthermore, we examined insomnia's association with depression symptoms after removing from the analyses participants with anxiety at baseline, defined by use of anxiolytic agents and/or a trait-anxiety score at or above the 75th percentile using the State-Trait Anxiety Inventory (41), which was self-administered during overnight studies. The cutpoint for anxiety (Trait Anxiety score ≥39) was defined by the overall cohort's 75th percentile. The correspondence of insomnia symptoms to polysomnographic measures was assessed for the same overnight study among 337 participants who reported that sleep obtained during their polysomnography study was typical of, or not much different from, their usual sleep.


The average follow-up period was 3.8 years (standard deviation (SD), 0.4). Of the initial subset (n = 787), 54% of participants were male; the mean age was 53.6 years (SD, 7.7; range, 33–71). Baseline characteristics are shown in Table 1. Relatively high proportions of self-reported insomnia symptoms (e.g., 34% for repeated nocturnal awakenings) and polysomnographically defined markers were observed. Twenty-six participants (4.7%) in the final inception cohort had developed depression symptoms (Zung score ≥50) by follow-up.

Table 1.
Baseline Characteristics of Participantsa (n = 555), Wisconsin Sleep Cohort Study, 1998–2002

Baseline mean depression scores (Zung score <50) in participants who developed (Zung score ≥50) or did not develop (Zung score <50) depression were similar: 40.0 (SD, 6.07) and 36.6 (SD, 5.5), respectively. Average scores changed (follow-up minus baseline) by +13.4 points versus only −0.01 points for participants developing, versus not developing, depression. Baseline average depression scores in participants without depression (Zung score <50) were not significantly different (P = 0.74) for decreasing total sleep time quartiles (Zung scores were 36 (quartile 4—longest, “best” total sleep time), 37 (quartile 3), 37 (quartile 2), and 34 (quartile 1—shortest, “worst” sleep time)) or increasing sleep latency (Zung scores were 37 (quartile 1—shortest, “best”), 36 (quartile 2), 36 (quartile 3), and 37 (quartile 4—longest, “worst”)). Despite statistical significance (P < 0.0001), average baseline Zung scores did not approach the 50 cutpoint for number of insomnia symptoms (36 for no symptoms, 37 for 1 symptom, 38 for 2 symptoms, and 38 for 3 or 4 symptoms).

Figure 1 shows the unadjusted incidence of depression by number of insomnia symptoms among participants who were free of depression symptoms and antidepressant use at baseline. Regardless of Zung depression scale (original or modified with 2 sleep-related items removed) and antidepressant use at follow-up, an increasing incidence of depression was observed across insomnia symptoms. This trend was significant for all categories (P-trend ≤ 0.02) except “modified Zung score excluding antidepressant users at follow-up” (P-trend = 0.06). Depression incidence for 3 or 4 symptoms was 3.2–3.6 times that for no symptoms (irrespective of exclusion of antidepressant users).

Figure 1.
Incidence of depression symptoms (%) at 4-year follow-up (2002–2006), obtained using different classifications of depression, by number of insomnia symptoms at baseline (1998–2002) in Wisconsin Sleep Cohort Study participants (n = 542–685). ...

Insomnia symptoms

The number of symptoms occurring often/almost always (Table 2) was found to be associated with subsequent depression in a graded (“dose-response”) fashion (P-trend = 0.02) after adjustment for age, sex, and comorbidity; the relative risk for 3 or 4 symptoms was 3.23 (P = 0.03). Further adjustments attenuated this relation, though risk was still increased, and the overall trend remained significant (P-trend = 0.04).

Table 2.
Relative Risk of Incident Depressiona Symptoms at 4-Year Follow-up According to Number of Insomnia Symptomsb Reported at Baseline (n = 555), Wisconsin Sleep Cohort Study, 1998–2006

Linear trends were observed for the relation of individual symptom frequency to incident depression symptoms (Figure 2) and were significant for increasing difficulty falling asleep (P-trend < 0.001) and increasing difficulty getting back asleep (P-trend = 0.02). After adjustment for age, sex, and comorbid conditions, increasing difficulty falling asleep and getting back asleep demonstrated significant graded trends for depression (P-trend = 0.01 and P-trend = 0.03, respectively) (Table 3). Repeated nocturnal and early morning awakenings demonstrated near-significant trends (P-trend = 0.07). Each symptom experienced “often or always” (versus all symptoms experienced “never or rarely”) was related to a 3-fold or higher risk, though after full adjustment, risk was significant only for difficulty falling asleep (relative risk (RR) = 5.31; P = 0.05).

Table 3.
Relative Risk of Incident Depressiona Symptoms at 4-Year Follow-up According to Baseline Frequency of 4 Individual Insomnia Symptomsb (n = 555), Wisconsin Sleep Cohort Study, 1998–2006
Figure 2.
Incidence of depression symptoms (%) at 4-year follow-up (2002–2006) by frequency of insomnia symptoms at baseline (1998–2002) in Wisconsin Sleep Cohort Study participants (n = 555). White columns, never or rarely experienced symptom; ...

No significant at (P ≤ 0.05) age or sex interactions were observed for individual symptoms or number of symptoms.

Polysomnographic markers

Unadjusted depression risk was 2- to 3-fold higher for increased (uppermost quartile) sleep latency and waking after sleep onset and decreased (lowest quartile) sleep efficiency and total sleep time, versus the respective reference categories (Table 4). Significant associations persisted after additional adjustments, with the full adjustment expressed by significant relative risks between 2.2 and 4.7, and a graded pattern for total sleep time (P-trend = 0.03).

Table 4.
Relative Risk of Incident Depression Symptomsa at 4-Year Follow-up According to Insomnia Markersb Measured Polysomnographically at Baseline (n = 555), Wisconsin Sleep Cohort Study, 1998–2006

Secondary analyses

The significant relation of number of symptoms to incident depression, including its graded pattern (P-trend ≤ 0.05), was not altered by adding apnea-hypopnea index to the age-, sex-, and comorbidity-adjusted model (RR = 3.17, 95% confidence interval: 1.05, 9.61) or by excluding from the analyses participants with anxiety at baseline (RR = 3.70, 95% confidence interval: 1.22, 11.18).

Clinical polysomnographic cutpoints were examined in fully adjusted models. Sleep latency >30 minutes (vs. ≤30 minutes) (RR = 3.98) and sleep efficiency <80% (vs. ≥80%) (RR = 4.39) significantly (P ≤ 0.01) predicted depression; total sleep time <360 minutes (vs. ≥360 minutes) (RR = 1.53) and waking after sleep onset >30 minutes (vs. ≤30 minutes) (RR = 1.10) did not.

Increasing mean sleep latency (polysomnographically assessed) was related to increasing frequency of self-reported difficulty falling asleep: 7.9 minutes—never (0 times/month), 9.3 minutes—rarely (once/month), 11.9 minutes—sometimes (2–4 times/month), 14.9 minutes—often (5–15 times/month), and 19.2 minutes—almost always (16–30 times/month). Increased mean waking after sleep onset (polysomnographically assessed) corresponded to increased self-reported frequency of nocturnal awakening: 7.9 minutes—never, 9.3 minutes—rarely, 11.9 minutes—sometimes, 14.9 minutes—often, and 19.2 minutes—almost always.

In our sample (n = 555), results based on mean continuous scores for depression at follow-up (Table 5) largely showed patterns similar to those based on a discrete outcome (Zung score ≥50), considered the clinical cutpoint (38). Mean depression values for 7 of 9 insomnia variables—number of symptoms, 4 individual symptoms, decreased sleep efficiency, and increased waking after sleep onset—were significantly increased upon follow-up and showed graded trends for all insomnia symptoms (P-trend ≤ 0.01). No associations were found for increased sleep latency and decreased total sleep time. Results remained unchanged after exclusion of antidepressant users from follow-up (n = 548).

Table 5.
Mean Scaled Zung Depression Score at 4-Year Follow-up According to Insomnia Symptoms and Polysomnographic Markers Measured at Baselinea (n = 555), Wisconsin Sleep Cohort Study, 1998–2006


Insomnia, whether assessed via polysomnographic markers or via self-report, was found to be a risk factor for depression symptoms at follow-up among participants who had been free of depression 4 years earlier in our population-based, observational study. Our study was unique in that we assessed both polysomnographic markers and self-reported symptoms of insomnia. Insomnia markers reflecting decreased sleep continuity and its symptoms, including number of symptoms and difficulty initiating sleep, predicted the occurrence of depression after adjustment for possible confounders. To our knowledge, significant graded associations between insomnia (increasing number of symptoms, difficulty initiating and maintaining sleep) and depression have not been previously reported. Our findings regarding self-reported insomnia's positive relation to depression are broadly consistent with results from most prospective epidemiologic studies conducted in various US and European populations (3, 12, 13, 2124, 2629), with a few exceptions (10, 25, 30) possibly due to the use of differing methods (e.g., insomnia definition or duration (12), sample characteristics, follow-up periods).

An important limitation of our study is that only 26 (4.7%) participants developed depression (defined as Zung score ≥50), since we emphasized specificity in our incident depression classification. Because the small number of outcome events (defined dichotomously) may have led to biased multivariable regression estimates and confidence intervals, we examined mean (continuous) depression values at follow-up to help address this limitation. Analyses based on mean scores were largely consistent with significant dichotomous results (based on the clinical cutpoint) and showed similar patterns. Fully adjusted multivariable regression results also resembled minimally adjusted or nonadjusted results. Nevertheless, larger study samples yielding a greater number of incident depression events may be warranted for confirmation of our findings.

The associations we found were not likely to have been due to subsyndromal depression in insomnia (scores near Zung's depression cutpoint). Average depression scores did not differ at baseline according to higher levels of, versus lower or absent, polysomnographic markers and were far from the depression cutpoint. For number of symptoms, depression scores were merely 2 points higher for 3 or 4 symptoms versus no symptoms, and far (12 points) from the 50 threshold. Moreover, average Zung scores less than 50 at baseline were similar for participants developing and not developing depression.

Although the definition of insomnia sometimes includes daytime consequences or impairment/interference (21, 42), our definition did not, which limited the potential for built-in associations. To avoid overestimating insomnia's association with depression, we used a modified depression scale excluding 2 sleep-related items, a method rarely used (10, 26). Estimates, though slightly attenuated, remained significant. Because antidepressants may be used for conditions beyond depression (43), we reanalyzed the data after excluding users of antidepressant medication (irrespective of depression symptoms) from follow-up; patterns and trends remained unchanged (Figure 1), even after adjustment. Relative risks were 2.8 (P = 0.05) for difficulty falling asleep, 4.8 (P < 0.001) for waking after sleep onset ≥79 minutes, and 2.9 (P = 0.02) for sleep latency ≥14 minutes.

When standard polysomnographic cutpoints considered clinically significant were analyzed, sleep efficiency <80% and sleep latency >30 minutes remained significantly associated with depression (Zung score ≥50); total sleep time <360 minutes and waking after sleep onset ≥30 minutes did not. The graded relation of total sleep time quartiles with depression was perhaps diluted by the binary cutpoint; waking after sleep onset was significant only at upper quartile values.

Several polysomnographic cutpoints used were more conservative than those typically utilized to define poor sleep (total sleep time <360–390 minutes, sleep efficiency <85%) (44). Our lowest quartiles of total sleep time (<334 minutes) and sleep efficiency (<78%) were similar to objective insomniacs’ mean values in clinical settings (329 minutes and 75%, respectively) (45), suggesting that clinically relevant “objective” markers were assessed. Analyses further suggested that polysomnographic markers corresponded to self-reported difficulty in initiating and maintaining sleep.

Among our study's limitations, we were unable to examine whether depression was season-related (seasonal affective disorder) or to differentiate between first-onset and recurrent depression, since information on past depression was not available. Nevertheless, findings that insomnia predicts depression symptoms in persons without depression at baseline may be relevant for preventive purposes, irrespective of whether depression at follow-up is first-onset or recurrent. Furthermore, an interaction between insomnia and age was not observed, increasing the likelihood that a reasonably high proportion of newly developed (vs. recurrent) depression was identified at follow-up. Self-reported depression symptoms may also be more labile. Depression is dynamic and is characterized by symptomatic fluidity (16, 17), suggesting that our inability to differentiate between first-onset depression and recurrent depression may be less pertinent.

Both early and middle insomnia—difficulty in initiating and maintaining sleep, respectively—predicted depression symptoms. Similarly, Perlis et al. (13) found that chronic middle insomnia predicted depression. In contrast to our study, they found persistent delayed (early morning awakening) insomnia symptoms to be somewhat related to depression and persistent early insomnia to be unrelated (13). A clinical study by Fava et al. (46) suggested that delayed and initial insomnia are prodromal depression symptoms. Our findings that early and middle insomnia were most related to depression are thus partially consistent with prior studies; differences may relate to these studies’ examining chronic insomnia or clinical populations.

The follow-up participation rate in the overall cohort has been 80%, on average. Proportions of insomnia and depression did not vary according to number of follow-up visits completed, suggesting an absence of selection bias. Despite the use of polysomnography, variability (night-to-night) and instrumentation errors were possible. However, the measurement error was probably nondifferential, thus likely underestimating our positive findings. A first-night effect (47) resulting in regression dilution bias is possible given 1 night of polysomnography. However, the fact that significant associations were found may highlight the strength of our findings.

The Zung scale has limitations; it is based on self-reporting and does not correspond directly to the criteria of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (48). Nevertheless, its strengths include a high correlation with clinical evaluation (38, 49) and wide usage in research. Other potential limitations include confounding due to unmeasured factors like pain (24), though we adjusted for back pain. The external validity of our findings should be limited to the incidence of both first-onset and recurrent depression. Moreover, the findings may not be generalizable beyond a middle-aged population with similar characteristics or to persons assessed as depressed by a clinician.

Our study's strengths include the collection of longitudinal data from a large population-based sample, longer follow-up, high-quality polysomnographic and self-report assessments, consistent findings across insomnia measures, and consideration of multiple confounding factors. Because our baseline sample comprised mainly working persons and because we excluded depressed participants and adjusted for common comorbid conditions, baseline insomnia was probably unrelated to medical or psychiatric conditions. Significant results persisted after removal of participants with anxiety, suggesting that the association observed was not merely due to anxiety.


Despite its high prevalence, insomnia remains largely undiagnosed among patients seen in primary-care settings, due to factors ranging from inadequate training of primary care providers to concerns and misperceptions about sleep medications, lack of familiarity with behavioral techniques, and inadequate amounts of time spent with patients (50). Our findings suggest that insomnia may increase the risk of subsequent depression symptoms. If confirmed by future studies, our results suggest that recognition and treatment of insomnia by health-care providers may be critical for preventing or mitigating the occurrence of depression. Prevention of depression, even mild or subthreshold depression, has wide-ranging implications given its high (including invisible) costs (51), such as increased health-service utilization, its chronic recurrent nature, and undertreatment (16, 17).


Author affiliations: School of Community and Environmental Health, College of Health Sciences, Old Dominion University, Norfolk, Virginia (Mariana Szklo-Coxe); Department of Population Health Sciences, School of Medicine and Public Health, University of Wisconsin, Madison, Wisconsin (Terry Young, Paul E. Peppard, Laurel Finn); and Department of Psychiatry, School of Medicine and Public Health, University of Wisconsin, Madison, Wisconsin (Ruth Benca).

This work was supported by grants from the National Institutes of Health (R01HL 62252, R01AG 14124, RR03186, and 1UL1RR025011) and a Wisconsin Partnership Grant (133JT72).

The authors are grateful to Amanda Rasmuson, Kathryn Pluff, and Katherine Kenison (Department of Population Health Sciences, University of Wisconsin School of Medicine and Public Health, Madison, Wisconsin) for their technical expertise with polysomnography and overnight sleep studies and to Linda Evans and Maryann Stubbs (Department of Population Health Sciences, University of Wisconsin School of Medicine and Public Health) for their technical contributions. The authors also thank Dr. Mae Hla (Department of Medicine, University of Wisconsin School of Medicine and Public Health) for her helpful suggestions regarding the tables and figures. None of those acknowledged received any compensation.

An earlier version of this work, based on a submitted abstract, was orally presented in part by Dr. Mariana Szklo-Coxe at the 21st Annual Meeting of the Associated Professional Sleep Societies in Minneapolis, Minnesota, on June 14, 2007.

Dr. Ruth M. Benca currently receives research support from the National Institute of Mental Health, the Medical Education and Research Committee of the Wisconsin Partnership Program, and the American Academy of Sleep Medicine, and she is currently or has been a consultant/advisory board member for the following pharmaceutical companies: Actelion, Sanofi-Aventis, Schering-Plough, and Merck. The other authors have no financial disclosures or conflicts of interest to report.



relative risk
standard deviation


1. Perlis ML, Giles DE, Buysse DJ, et al. Self-reported sleep disturbance as a prodromal symptom in recurrent depression. J Affect Disord. 1997;42(2-3):209–212. [PubMed]
2. Fava GA, Grandi S, Canestrari R, et al. Prodromal symptoms in primary major depressive disorder. J Affect Disord. 1990;19(2):149–152. [PubMed]
3. Dryman A, Eaton WW. Affective symptoms associated with the onset of major depression in the community: findings from the US National Institute of Mental Health Epidemiologic Catchment Area Program. Acta Psychiatr Scand. 1991;84(1):1–5. [PubMed]
4. Eaton WW, Badawi M, Melton B. Prodromes and precursors: epidemiologic data for primary prevention of disorders with slow onset. Am J Psychiatry. 1995;152(7):967–972. [PubMed]
5. Nierenberg AA, Keefe BR, Leslie VC, et al. Residual symptoms in depressed patients who respond acutely to fluoxetine. J Clin Psychiatry. 1999;60(4):221–225. [PubMed]
6. Ohayon MM, Caulet M, Lemoine P. Comorbidity of mental and insomnia disorders in the general population. Compr Psychiatry. 1998;39(4):185–197. [PubMed]
7. Katz DA, McHorney CA. Clinical correlates of insomnia in patients with chronic illness. Arch Intern Med. 1998;158(10):1099–1107. [PubMed]
8. Taylor DJ, Mallory LJ, Lichstein KL, et al. Comorbidity of chronic insomnia with medical problems. Sleep. 2007;30(2):213–218. [PubMed]
9. Weyerer S, Dilling H. Prevalence and treatment of insomnia in the community: results from the upper Bavarian Field Study. Sleep. 1991;14(5):392–398. [PubMed]
10. Roberts RE, Shema SJ, Kaplan GA, et al. Sleep complaints and depression in an aging cohort: a prospective perspective. Am J Psychiatry. 2000;157(1):81–88. [PubMed]
11. Roth T, Jaeger S, Jin R, et al. Sleep problems, comorbid mental disorders, and role functioning in the National Comorbidity Survey Replication. Biol Psychiatry. 2006;60(12):1364–1371. [PMC free article] [PubMed]
12. Buysse DJ, Angst J, Gamma A, et al. Prevalence, course, and comorbidity of insomnia and depression in young adults. Sleep. 2008;31(4):473–480. [PubMed]
13. Perlis ML, Smith LJ, Lyness JM, et al. Insomnia as a risk factor for onset of depression in the elderly. Behav Sleep Med. 2006;4(2):104–113. [PubMed]
14. Murray CJ, Lopez AD. Evidence-based health policy—lessons from the Global Burden of Disease Study. Science. 1996;274(5288):740–743. [PubMed]
15. Léger D, Massuel MA, Metlaine A. Professional correlates of insomnia. Sleep. 2006;29(2):171–178. [PubMed]
16. Judd LL, Akiskal HS, Paulus MP. The role and clinical significance of subsyndromal depressive symptoms (SSD) in unipolar major depressive disorder. J Affect Disord. 1997;45(1-2):5–17. [PubMed]
17. Kessler RC, Zhao S, Blazer DG, et al. Prevalence, correlates, and course of minor depression and major depression in the National Comorbidity Survey. J Affect Disord. 1997;45(1-2):19–30. [PubMed]
18. Ozminkowski RJ, Wang S, Walsh JK. The direct and indirect costs of untreated insomnia in adults in the United States. Sleep. 2007;30(3):263–273. [PubMed]
19. Simon GE, Vonkorff M. Prevalence, burden, and treatment of insomnia in primary care. Am J Psychiatry. 1997;154(10):1417–1423. [PubMed]
20. Godet-Cayré V, Pelletier-Fleury N, Le Vaillant M, et al. Insomnia and absenteeism at work. Who pays the cost? Sleep. 2006;29(2):179–184. [PubMed]
21. Ford DE, Kamerow DB. Epidemiologic study of sleep disturbances and psychiatric disorders. An opportunity for prevention? JAMA. 1989;262(11):1479–1484. [PubMed]
22. Chang PP, Ford DE, Mead LA, et al. Insomnia in young men and subsequent depression. The Johns Hopkins Precursors Study. Am J Epidemiol. 1997;146(2):105–114. [PubMed]
23. Weissman MM, Greenwald S, Niño-Murcia G, et al. The morbidity of insomnia uncomplicated by psychiatric disorders. Gen Hosp Psychiatry. 1997;19(4):245–250. [PubMed]
24. Morphy H, Dunn KM, Lewis M, et al. Epidemiology of insomnia: a longitudinal study in a UK population. Sleep. 2007;30(3):274–280. [PubMed]
25. Neckelmann D, Mykletun A, Dahl AA. Chronic insomnia as a risk factor for developing anxiety and depression. Sleep. 2007;30(7):873–880. [PubMed]
26. Breslau N, Roth T, Rosenthal L, et al. Sleep disturbance and psychiatric disorders: a longitudinal epidemiological study of young adults. Biol Psychiatry. 1996;39(6):411–418. [PubMed]
27. Roane BM, Taylor DJ. Adolescent insomnia as a risk factor for early adult depression and substance abuse. Sleep. 2008;31(10):1351–1356. [PubMed]
28. Mallon L, Broman JE, Hetta J. Relationship between insomnia, depression, and mortality: a 12-year follow-up of older adults in the community. Int Psychogeriatr. 2000;12(3):295–306. [PubMed]
29. Livingston G, Blizard B, Mann A. Does sleep disturbance predict depression in elderly people? A study in inner London. Br J Gen Pract. 1993;43(376):445–448. [PMC free article] [PubMed]
30. Vollrath M, Wicki W, Angst J. The Zurich Study. VIII. Insomnia: association with depression, anxiety, somatic syndromes, and course of insomnia. Eur Arch Psychiatry Neurol Sci. 1989;239(2):113–124. [PubMed]
31. Kupfer DJ, Targ E, Stack J. Electroencephalographic sleep in unipolar depressive subtypes: support for a biological and familial classification. J Nerv Ment Dis. 1982;170(8):494–498. [PubMed]
32. Reynolds CF, III, Newton TF, Shaw DH, et al. Electroencephalographic sleep findings in depressed outpatients. Psychiatry Res. 1982;6(1):65–75. [PubMed]
33. Giles DE, Jarrett RB, Roffwarg HP, et al. Reduced rapid eye movement latency: a predictor of recurrence in depression. Neuropsychopharmacology. 1987;1(1):33–39. [PubMed]
34. Thase ME, Kupfer DJ, Fasiczka AJ, et al. Identifying an abnormal electroencephalographic sleep profile to characterize major depressive disorder. Biol Psychiatry. 1997;41(9):964–973. [PubMed]
35. National Institutes of Health. National Institutes of Health State of the Science Conference Statement on Manifestations and Management of Chronic Insomnia in Adults, June 13–15, 2005. Sleep. 2005;28(9):1049–1057. [PubMed]
36. Young T, Palta M, Dempsey J, et al. The occurrence of sleep-disordered breathing among middle-aged adults. N Engl J Med. 1993;328(17):1230–1235. [PubMed]
37. Rechtschaffen A, Kales A, editors. A Manual of Standardized Terminology, Techniques and Scoring System for Sleep Stages of Human Subjects. Washington, DC: US GPO; 1968. (National Institutes of Health publication no. 204) [PubMed]
38. Zung WW. A self-rating depression scale. Arch Gen Psychiatry. 1965;12(1):63–70. [PubMed]
39. Peppard PE, Szklo-Coxe M, Hla KM, et al. Longitudinal association of sleep-related breathing disorder and depression. Arch Intern Med. 2006;166(16):1709–1715. [PubMed]
40. Zou G. A modified Poisson regression approach to prospective studies with binary data. Am J Epidemiol. 2004;159(7):702–706. [PubMed]
41. Spielberger CD, Gorsuch RL, Lushene R, et al. Manual for the State-Trait Anxiety Inventory (Form Y) Palo Alto, CA: Consulting Psychologists Press, Inc; 1983.
42. Ohayon MM. Epidemiology of insomnia: what we know and what we still need to learn. Sleep Med Rev. 2002;6(2):97–111. [PubMed]
43. Winkelman J, Pies R. Current patterns and future directions in the treatment of insomnia. Ann Clin Psychiatry. 2005;17(1):31–40. [PubMed]
44. Krystal AD, Edinger JD, Wohlgemuth WK, et al. NREM sleep EEG frequency spectral correlates of sleep complaints in primary insomnia subtypes. Sleep. 2002;25(6):630–640. [PubMed]
45. Edinger JD, Fins AI, Glenn DM, et al. Insomnia and the eye of the beholder: are there clinical markers of objective sleep disturbances among adults with and without insomnia complaints? J Consult Clin Psychol. 2000;68(4):586–593. [PubMed]
46. Fava GA, Grandi S, Canestrari R, et al. Prodromal symptoms in primary major depressive disorder. J Affect Disord. 1990;19(2):149–152. [PubMed]
47. Toussaint M, Luthringer R, Schaltenbrand N, et al. First-night effect in normal subjects and psychiatric inpatients. Sleep. 1995;18(6):463–469. [PubMed]
48. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th ed. Washington, DC: American Psychiatric Association; 1994.
49. Biggs JT, Wilie LT, Ziegler VE. Validity of the Zung Self-rating Depression Scale. Br J Psychiatry. 1978;132:381–385. [PubMed]
50. Benca RM. Diagnosis and treatment of chronic insomnia: a review. Psychiatr Serv. 2005;56(3):332–343. [PubMed]
51. Stewart WF, Ricci JA, Chee E, et al. Cost of lost productive work time among US workers with depression. JAMA. 2003;289(23):3135–3144. [PubMed]

Articles from American Journal of Epidemiology are provided here courtesy of Oxford University Press