Nightmares and Stress: A Longitudinal Study

Abstract

Study Objectives:

In nightmare etiology, trait and state factors play important roles. However, the interaction of state and trait factors has never been studied in a longitudinal design.

Methods:

The current sample included 406 pregnant women who were followed up approximately 6 months after giving birth (n = 375) and 4 years later (n = 302). A nightmare frequency scale and several stress-related questionnaires were presented at three measurement points.

Results:

Despite the major life events in this sample, nightmare frequency was very stable over this time period and decreased slightly. In line with previous findings, cross-sectional analyses showed that stressors were associated with current nightmare frequency but longitudinal analyses indicated that previously measured nightmare frequency showed even stronger effects on current nightmare frequency.

Conclusions:

Because the nightmare frequencies were very stable, it would be desirable to carry out intervention studies treating nightmares as early as possible—even in childhood—and study whether nightmare occurrence is lower even years after the intervention.

Citation:

Schredl M, Gilles M, Wolf I, Peus V, Scharnholz B, Sütterlin M, Bardtke S, Send TS, Samaras A, Deuschle M. Nightmares and stress: a longitudinal study. J Clin Sleep Med. 2019;15(9):1209–1215.

BRIEF SUMMARY

Current Knowledge/Study Rationale: In nightmare etiology, trait and state factors play important roles. However, the interaction of state and trait factors has never been studied in a longitudinal design.

Study Impact: The findings of the current study showed a significant effect of previously measured nightmare frequency on current nightmare frequency—in addition to the effects of current stressors—and, thus, clearly indicate the importance of treating nightmares as soon as possible. Prevention studies in this field are warranted as the prevalence of the nightmare disorder in the general population is quite high (5%).

INTRODUCTION

Nightmares are defined as extended, extremely dysphoric, and well-remembered dreams that usually involve threat to survival, security, or physical integrity.¹ Approximately 5% in the adult population have frequent nightmares.² For explaining nightmare occurrence, especially the diagnoses of a nightmare disorder, a diathesis-stress model is widely used.³ Levin and Nielsen⁴ specified this type of model for nightmares: they conceptualized affect load as a state factor reflecting the combined influence of stressful and emotionally negative events on an individual’s capacity to effectively regulate emotions (stress factor) and affect distress as a trait factor reflecting a long-standing disposition to experience heightened distress and negative affect and to react with extreme behavioral expressions. A heightened and stable nightmare frequency would be indicative of this trait factor.

Empirical findings clearly indicated that current psychopathology^5,6 and stress measures^7–9 are related to nightmare frequency and thus clearly support the stress factor in the nightmare model. Moreover, the trait aspect is supported by findings indicating that genetic factors also play a role in nightmare etiology¹⁰ and the relationship between trait measures such as neuroticism,¹¹ thin boundaries,¹² and sensory processing sensitivity.³ Nielsen¹³ provided evidence that these trait markers, for example, sensory processing sensitivity, might be the consequence of early adverse experiences occurring in sensitive windows during emotional maturation (presumably before the age of 3.5 years). The empirical database also supports the diathesis part of nightmare etiology. Most of these studies, though, are cross-sectional (for example, Schredl¹¹) or retrospective (Nielsen¹⁴), linking the earliest remembered dream to current nightmare frequency. In clinical samples, those with nightmares often report that their nightmares started in childhood, indicating that nightmare frequency might be stable over the lifespan.^15,16 Cross-sectional studies in adults show inhomogeneous findings regarding the relationship between age and nightmare frequency: several studies^17–20 found a decrease of nightmare frequency with age, whereas two representative studies^21,22 found no effect. A large-scaled Finnish study² even reported a slight increase in nightmare frequency with age. The cross-sectional studies might be biased by cohort effects, for example, by including participants who experienced directly or indirectly World War II.² Research also indicated that retrospective estimates of nightmare frequency in childhood correlated strongly with the current nightmare frequency in adults.^10,17 However, retrospective estimates of nightmare frequency in childhood might be biased because persons with current nightmares might overestimate the occurrence of nightmares in their childhoods.

In order to obtain reliable findings regarding the stability of nightmares, longitudinal studies are necessary. Unfortunately, longitudinal studies looking at the stability of nightmare frequency are scarce. Simard et al²³ reported that 78.3% of the children who had bad dreams at age 29 months still had bad dreams at the age of 6 years, that is, having bad dreams was for most children a stable trait. In 851 children, the nightmare frequency at mean age of 9.49 ± 0.61 years was correlated with the nightmare frequency 2 years later: r = .323 for parent estimates and r = .297 for children’s estimates.²⁴ In an adult sample (age mean: 45.27 ± 13.99 years) nightmare frequency dropped slightly over a 3-year period but the measures at the measurement points were strongly correlated (r = .616), again showing considerable stability in nightmare frequency.²⁵ Interestingly, the change in nightmare frequency in this sample with large age range (16 to 89 years) was not related to age and sex, that is, the slight decrease was similar in all age groups and comparable between men and women.²⁵ In a cross-sectional study, neuroticism (trait measure) was related to nightmare frequency but if measures of current stress levels were included in the regression analysis, neuroticism was no longer significant.¹¹ This finding supports the model that persons with traits associated with psychopathology such as neuroticism, experience stress more often and thus experience more nightmares (mediator model). However, in the cross-sectional design, measurement of trait aspects and state aspects might be confounded. Longitudinal studies investigating the question how strong nightmares are affected by current stressors and/or by previously measured nightmare frequency have not yet been carried out.

The aims of the current study are: (1) measure nightmare frequency longitudinally in order to assess the stability of nightmare frequency; (2) assess whether nightmare frequency correlates with stress levels as we expected that nightmare frequencies are correlated with current stress levels, based on the so-called continuity hypothesis stating that waking-life experiences (here: negative feeling associated with stress) are reflected in dreams,²⁶ and (3) determine the influence of previously measured nightmare frequency (trait aspect) on current nightmare frequency in addition to the effect of current stressors (state aspect).

METHODS

Participants

Overall, 406 women participated in the study. Their mean age was 31.43 ± 5.08 years (range: 17 to 44 years). The questionnaires at T1 were administered on average within the gestation week: 36.49 ± 2.38 weeks (27 to 40 weeks). School education was distributed as follows: 13 years of education (n = 230), 12 years (n = 30), 10 years (n = 96), 9 years (n = 44), not completed school (n = 5), and special school (n = 1).

At 6 months postpartum (T3), 357 women participated and completed the questionnaires, and 302 women were tested at 3.5 years postpartum (T4).

Dream Questions

All participants were asked to rate their dream recall during the previous months on a 7-point rating scale (0 = never, 1 = less than once a month, 2 = about once a month, 3 = twice or three times a month, 4 = about once a week, 5 = several times a week, 6 = almost every morning). The retest reliability of the scale (mean retest interval: 54.8 ± 44.8 days; n = 198) was r = .83.²⁷

In addition, nightmare frequency was measured using an 8-point scale (0 = never, 1 = less than once a year, 2 = about once a year, 3 = about two to four times a year, 4 = about once a month, 5 = twice or three times a month, 6 = about once a week, 7 = several times a week) that also showed a high retest reliability of r = .75.²⁸ A specific definition of nightmares was not provided.

Stress and Psychopathology Questionnaires

The Perceived Stress Scale (PSS) measures self-reported stress experiences regarding situations in daily life.²⁹ The sum score of the 14 items showed high retest reliability and high internal consistency (both r_tt > .80). The state version of the State-Trait-Anxiety Inventory (STAI-S) encompasses 20 four-point statements regarding emotional and cognitive aspects of anxiety.³⁰ The total sum score ranging from 20 to 80 showed high internal constancy (r = .92). The Edinburgh Postnatal Depression Scale (EPDS) was originally developed to assess postnatal depression but has also been validated for eliciting depressive mood during pregnancy.^31,32 The sum score of the 10 items of the German version³³ showed high reliability (Cronbach alpha = .81). The Life Experiences Survey (LES) elicits life events that occurred during the previous year, for example, death of a close relative.³⁴ All events are to be rated according to their emotional quality. For the current analysis, the sum of all negatively evaluated events has been included. The retest reliability of this index is sufficient, ranging from r = .56 to .88.³⁴ The Anxiety Screening Questionnaire (ASQ) was designed by Wittchen and Boyer³⁵ for eliciting symptoms related to generalized anxiety disorder and panic disorder. The test score showed high retest reliability and high specificity of detecting anxiety disorders.³⁵ The Prenatal Distress Questionnaire consists of 12 five-point scales measuring pregnancy-specific distress, for example, finding the weight gain due to pregnancy troublesome or worrying about eating healthy foods.³⁶ The total sum score varies from 0 to 48 and showed high correlation (r = .53) to global distress.³⁶ The Beck Depression Inventory includes 21 four-point scales with high internal consistency for German samples (Cronbach alpha > .83); the sum score ranges from 0 to 63.³⁷

Procedure

The mothers to be were approached during their application visit about 4 to 8 weeks prior to delivery in three obstetric hospitals in the Rhine-Neckar Region of Germany (Mannheim, Ludwigshafen). They were briefly informed about the study (Pre-, Peri- and Postnatal Stress: Epigenetic impact on Depression; POSEIDON) and received a flyer with a brief outline of the study encompassing three measurement points: third trimester of pregnancy (T1), immediately after delivery (T2), and 6 months postpartum (T3). The following inclusion criteria for mothers were applied: Caucasian descent, main caregiver, German-speaking, and age 16 to 40 years. Exclusion criteria were: maternal hepatitis B, hepatitis C or human immunodeficiency virus infection, any current psychiatric disorders requiring inpatient treatment, a history, current diagnosis of schizophrenia/psychotic disorder, or any substance dependency other than nicotine during pregnancy. Based on rough estimates of deliveries per year within each hospital, it could be estimated that approximately 33% of all mothers who met the inclusion/exclusion criteria participated in the study within the recruiting period from October 2010 to March 2013. Participation was reimbursed with 120 euros. The study protocol was approved by the Ethics Committee of the Medical Faculty Mannheim of the University of Heidelberg and the study was conducted in accordance with the Declaration of Helsinki. All mothers provided written informed consent prior to participation. The nightmare and dream recall scales and the stress-related questionnaires were completed at T1 (last trimester of pregnancy) and at T3 (6 months after delivery) but not at T2 (delivery).

For the follow-up study PEZ-PSYCHE (Psychoepidemiologisches Zentrum-PreSchooler: Young Children’s Health and Environment) all mothers were contacted again after about 3.5 years (T4). Within this assessment, the mothers completed the nightmare and dream recall scales again and also several stress-related questionnaires.

The statistical analyses were carried out with SAS 9.4 for Windows software (SAS Institute, Cary, North Carolina, United States). In order to reduce the number of variables in the analyses, factor analyses for the stress-related measures elicited at the different measurement points were carried out. At T1, the one-factor solution of the six questionnaires (PSS, STAI-S, LES (negative events), EPDS, ASQ, and Prenatal Distress Questionnaire) explained 65.02% of the total variance and factor loadings ranged from .625 to .891. At T3, the one-factor solution of the five questionnaires (PSS, STAI-S, LES (negative events), EPDS, and ASQ) explained 66.12% of the total variance and factor loadings ranged from .652 to .877. At T4, the one-factor solution of the four questionnaires (PSS, STAI-S, LES (negative events), Beck Depression Inventory) explained 66.10% of the total variance and factor loadings ranged from .681 to .879. The composite stress/psychopathology measure was computed as weighted and normalized mean of all scales entered into the analysis. The second reason for using a composite score was the issue of multicollinearity or collinearity; the factor analyses indicated that the different measures (stress, life events, and psychopathology) were strongly interrelated and, thus, a regression analysis including all variables would not be valid. As the nightmare scale was ordinal, logistic regressions were computed.

RESULTS

The distribution of nightmare frequencies for all three measurement points are depicted in Table 1 (at T2 [delivery] no questionnaires were completed). Most of the women experience nightmares occasionally, and about 5% (T3, T4) and about 11% (T1) experienced nightmares quite often (once a week or several times a week).

Table 1.

Nightmare frequency scale for all three measurement points.

The means and standard deviations for the nightmare frequency scale, perceived stress, and state anxiety are depicted in Table 2. Whereas there was a significant but slight decrease in nightmare frequency, perceived stress increased from T1 to T4. State anxiety scores did not differ between T1 and T4 but were elevated at T3.

Table 2.

Nightmare frequency and stress measures over the course of the study (analyzing overall differences and comparisons between each measurement point).

The correlations for nightmare frequency, perceived stress, state anxiety, and the composite stress/psychopathology factor between the measurement points are shown in Table 3. The coefficients ranged from r = .506 to .688.

Table 3.

Correlations of nightmare frequency and stress-related measures between measurement points.

The cross-sectional analyses for the effect of the composite stress/psychopathology factor on nightmare frequency are depicted in Table 4. Age, education, and dream recall frequency were also entered into the analysis in order to control for possible confounding effects. As expected, at all three measurement points the effect of stress on nightmare frequency was significant. The covariate dream recall frequency was also related to nightmare frequency. At T3 and T4, nightmare frequency was negatively associated with age, and education was not related to nightmare frequency at any time.

Table 4.

Ordinal regressions of the effect of the composite stress/psychopathology factor on nightmare frequency (cross-sectional for three measurement points).

In Table 5, the regression analyses with the composite stress/psychopathology factor and previously measured nightmare frequency and composite stress/psychopathology factors are depicted. Although the composite stress/psychopathology factor was still related to nightmare frequency, the strongest effect was demonstrated for the previously measured nightmare frequency whereas the previously measured composite stress/psychopathology factor did not add to explaining nightmare frequency at a later time. The regression coefficients for the effect of current stressors (state effect) and the previously measured nightmare frequency (trait effect) were comparable for all three time intervals: T1-T4, T1-T3, and T3-T4.

Table 5.

Ordinal regressions for nightmare frequency including the composite stress/psychopathology factor and previously measured variables.

DISCUSSION

This is the first longitudinal study showing that current nightmare frequency is not only affected by a composite stress/psychopathology factor (state effect) but also, to a larger extent, by previously measured nightmare frequency indicating the importance of the trait aspect. Moreover, nightmare frequency is quite stable over a time of 4 years.

Before discussing the findings in detail, several methodological issues have to be addressed. First, the sample of the current study only included pregnant women who later gave birth and had to raise the infant. As pregnancy is associated with increased nightmare frequency,^38,39 that is, the findings regarding the changes in nightmare frequency over the time course (slight decrease) has to be viewed with caution. However, the stability coefficients match those of a previous online study with a mixed sample²⁵ and thus indicate that the major life events did not have a major effect on nightmare stability; even the effect on nightmare frequency is very small. The three cross-sectional regression analyses and the three regression analyses including previously measured variables (longitudinal aspect) yielded comparable effects of the composite stress/psychopathology factor and previously measured nightmare frequency on current nightmare frequency and thus indicate that these major life events (pregnancy, giving birth, raising small children) might not have biased the current findings.

In addition, the response rate of about 33% presumably due to the high time expenditure requested by the POSEIDON study protocol is relatively low but not related to having nightmares or not (this topic was not mentioned in the recruitment phase). Compared to the German population, women with higher school education (12 or 13 years) were overrepresented in the current study: 64% versus general population: 30%.⁴⁰ To control for possible confounding effects, education was included in the regression analysis but did not show any effects; that is, the bias of recruiting more women with high education seems to be very small. The regression analysis were also controlled for dream recall frequency, as nightmare frequency is per definition part of the dream recall frequency¹⁷ and as stress or psychopathology are related to dream recall frequency.⁴¹ However, findings clearly indicate that the reported relationships are independent of dream recall.

Last, the reduction of the number of variables regarding daytime distress/psychopathology was justified by the high correlations of the questionnaire scores whether this was perceived stress, anxiety symptoms, negative life events, or depression; in this sample all those measures contributed to general distress and were equally related to nightmare frequency.³⁹ The reduction of variables clearly improved the statistical power of the analyses. As the questionnaires cover different time intervals ranging from state anxiety (feelings in the moment) to life events that could have occurred 1 year earlier, the composite stress/psychopathology factor does not reflect current stress levels. However, one can argue that the high intercorrelations between all questionnaire scores indicate the measures of current stress might have shown the same pattern. Another methodological topic discussed in the literature is the possible underestimation of nightmare frequency by using retrospective questionnaires,^42,43 that is, typically prospective diary measures yielded higher prevalence rates. A recent study⁴⁴ has shown that the effect size of this difference is quite small (d = 0.101) and, thus, a possible bias due to using retrospective scales in this study should be small. Moreover, retest reliability of the nightmare scale is quite high r = .75,²⁸ indicating that trait aspects of nightmare frequency can be measured reliably.

Overall, nightmare frequency was quite stable over the 4 years; the correlation coefficients ranging from r = .53 to r = .58 were similar to the values (r = .62) reported by Schredl and Göritz²⁵ for a 3-year interval. This is in line with the notion that trait factors play a role in nightmare etiology.⁴ Interestingly, the stability coefficients of the stress-related measures and the composite stress/psychopathology factor were of a similar magnitude, reflecting a trait stress concept (eg, neuroticism). Even though perceived stress and state anxiety—factors that affect nightmare frequency—increased or were stable over time, nightmare frequency significantly decreased (still to a small extent on average). Although one cannot rule out a specific effect of pregnancy on nightmare frequency (ie, at T1), the decrease parallels the finding of another sample including men and women and with an average age of 45 years. Moreover, there was also a decrease in nightmare frequency from T3 to T4, supporting the cross-sectional studies reporting an age-dependent drop in nightmare frequency.^17–19 As the findings in this area are conflicting,^2,20 it would be desirable to carry out longitudinal studies with intervals longer than 3 or 4 years. Interestingly, age and sex at T1 were not related to the decrease in nightmare frequency in the study of Schredl and Göritz,²⁵ indicating a linear decrease of nightmare frequency from young adulthood to older age.

In total agreement with previous findings,^5,8,45 the composite stress/psychopathology factor was associated with nightmare frequency; the magnitude of the standardized estimates indicate medium effect sizes. Because these effect sizes, collected in a sample of pregnant women and later young mothers, are comparable to the literature,⁴ they support the validity of the current findings.

As expected, the trait aspect (previously measured nightmare frequency) explained a larger portion of variance compared to the composite stress/psychopathology factor. As there are potent treatment strategies for coping with nightmare with stable long-term effects,⁴⁶ it would be very interesting to carry out longitudinal intervention studies to investigate the effect of trait aspects and current stressors on nightmare frequency. The high stability of nightmare frequency clearly supports the recommendation to treat nightmares directly and not only by reducing daytime stress levels.^47,48 This claim is supported by findings showing that imagery rehearsal therapy can improve daytime mood, depressive symptoms, and posttraumatic stress disorder symptoms.^16,46,49 The trait aspects of stress (previously measured stress levels) were not related to current nightmare frequency if current stress and the nightmare trait aspect is statistically controlled; this is in line with the findings of Schredl¹¹ showing that the effect of neuroticism (trait) on nightmare frequency is mediated by current stress levels.

To summarize, the findings clearly indicated that trait aspects play an important role in nightmare etiology in addition to stress and/or psychopathology. In order to pursue this line of research it would be very interesting to follow up birth cohorts as has been done in other areas⁵⁰; this could also evaluate the magnitude of genetic/personality factors compared to the effect of early adverse experiences proposed by the stress acceleration hypothesis of Nielsen.¹³ Another interesting topic would be to study people with nightmare disorder longitudinally as the current sample includes only a small proportion of women with frequent nightmares. It would be also very desirable to carry out intervention studies treating nightmares as early as possible—even in childhood⁵¹—and investigate whether the reduced nightmare frequency is stable over the years, even into adulthood.

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. Work for this study was performed at Central Institute of Mental Health, Mannheim, Germany. The POSEIDON study was supported by an Era-Net Neuron grant to Michael Deuschle. Michael Deuschle also received support from the Dietmar-Hopp Foundation. The PEZ-Psyche has received funding from the Ministry of Science, Research and the Arts of the State of Baden-Wuerttemberg, Germany. Michael Schredl has received research grants from Eisai, Lilly Pharma and Janssen. Michael Deuschle has received a speaker honorarium from Servier, Mundipharma, and Janssen. Michael Deuschle is a member of the advisory board (Janssen). The other authors report no conflicts of interest.

ABBREVIATIONS

ASQ

Anxiety Screening Questionnaire

EPDS

Edinburgh Postnatal Depression Scale

LES

Life Experiences Survey

PEZ-PSYCHE

Psychoepidemiologisches Zentrum – PreSchooler: Young Children’s Health and Environment

POSEIDON

Pre-, Peri- and Postnatal Stress: Epigenetic impact on Depression

PSS

Perceived Stress Scale

STAI

State-Trait-Anxiety Inventory