Introduction
Although exposure therapy is considered the gold standard in the treatment of anxiety and stressor-related disorders, there is a substantial proportion of patients, who do not profit from treatment, and relapse is a common problem (Michael et al.,
2019). Consequently, research on novel strategies that can enhance the success of exposure therapy has received much interest in the last years.
One line of research focuses on possible pharmacological agents as boosters of exposure therapy. These new therapeutic approaches are based on the idea that the pharmacological agents enhance the learning processes underlying exposure therapy. One pharmacological agent that has been proposed to enhance the success of exposure therapy is cortisol (Bentz et al.,
2010). Several studies have shown that exogenous administration of cortisol prior to exposure therapy enhances therapeutic gains of exposure therapy in patients with height phobia (de Quervain et al.,
2011), spider phobia (Soravia et al.,
2014; but see Raeder et al.,
2019 for contrary findings), social phobia (Soravia et al.,
2006), posttraumatic stress disorder (Yehuda et al.,
2015), and alcohol use disorder (Soravia et al.,
2021).
Cortisol is secreted in response to stress and can be administered exogenously (Bentz et al.,
2010). However, cortisol secretion also shows natural fluctuation across the day, with a peak in the morning and low levels during the evening and night. Thus, another line of research has focused on the effects of endogenous cortisol levels on the success of exposure therapy. Pace-Schott et al. (
2013) found that extinction learning (as an analog for exposure therapy) was more successful in the morning (when endogenous cortisol levels are high) than in the evening. Lass-Hennemann and Michael (
2014) transferred these findings to a clinical sample: They found that spider phobic patients who were treated in the morning showed better treatment outcomes than patients who were treated in the evening
. In line with these findings, Meuret et al. (
2015) showed that higher cortisol levels during exposure sessions conducted at different daytimes were associated with enhanced clinical improvement in a multi-session in-vivo exposure protocol for panic disorder and agoraphobia. In a further study in patients with panic disorder, the same authors were able to show that cortisol mediated the effect of time of day on treatment outcome, providing a link between earlier exposure sessions and greater clinical improvement (Meuret et al.,
2016).
These findings gave rise to a simple clinical hypothesis: Exposure sessions conducted in the morning are more effective than exposure sessions conducted in the afternoon or evening (“morning exposure effect”), because the high endogenous cortisol in the morning will enhance memory consolidation for new non-anxiety related material and thus lead to better clinical outcomes. Even though there is some evidence linking high endogenous cortisol levels to higher therapeutic gains during exposure therapy (Meuret et al.,
2016), the research on the “morning exposure effect” is just at the very beginning and there are also other potential mechanisms that may account for better therapy outcomes in the morning.
In contrast to accounts linking diurnal variations in cortisol levels to daytime effects on exposure therapy, there is an alternative assumption, positing that the temporal proximity to awakening is the critical factor boosting exposure effects in the morning (Nissen et al.,
2017). This assumption is based on the synaptic homeostasis hypothesis (Tononi & Cirelli,
2006). According to this hypothesis, new learning experiences result in a continuous increase of synaptic connections. Without downregulation, such a continuous increase would lead to a saturation of synaptic networks, preventing subsequent learning. Hence, to avoid saturation, synaptic connections are downregulated during sleep, a process that is referred to as synaptic downscaling. Based on this hypothesis, it has been suggested that the capacity for learning is highest immediately after awakening and decreases continuously throughout the day. Empirical findings support this assumption (Kaida et al.,
2015; Mander et al.,
2011) and confirm that preceding sleep enhances extinction learning (Straus et al.,
2017). Moreover, Zuj et al. (
2016) found that extinction learning in patients with more severe post-traumatic stress disorder is less successful after prolonged wakefulness than immediately after awakening.
Another factor that has received little attention so far but that may account for superior effects of exposure therapy during the morning is vigilance. Cognitive psychologists and neuroscientists define vigilance as the ability to sustain attention to a task for a period of time (Parasuraman,
2000). Vigilance has been shown to be higher in the morning and to decline over the course of the day (Harrison et al.,
2007; Riley et al.,
2017). Previous research indicates that reduced vigilance co-occurs with reduced learning and emotion processing (Helton & Russell,
2011; Schwarz et al.,
2013; Wang et al.,
2012). Psychotherapy in general, but especially exposure therapy relies on emotion processing and learning (Lass-Hennemann et al.,
2018). Higher vigilance levels in the morning as compared to the evening might thus—fully or partially—account for greater therapeutic gains in the morning.
Taken together, several factors may contribute to the “morning exposure effect”: High endogenous cortisol levels, sleep, and vigilance. However, up to date there are no studies assessing the different factors in one study to disentangle the importance of the proposed mechanisms. Moreover, despite compelling evidence, a study employing virtual exposure exercises for spider phobia in which exposures and testing of spider fear were performed both in the morning and evening (as controls for circadian vs. sleep effects) did not find superior exposure outcomes in the morning (Pace-Schott et al.,
2012). Furthermore, a recent study found that higher cortisol levels during exposure were linked to reduced—rather than enhanced—symptom decline in patients with social anxiety disorder (Kuhlman et al.,
2020). In addition, a meta-analysis was not able to confirm a significant association between cortisol levels during exposure and symptom improvement questioning the central role of cortisol in the “morning exposure effect” (Fischer & Cleare,
2017).
In light of these ambiguities, we conducted an experimental study to further characterize the impact of daytime on exposure therapy and the involvement of cortisol, sleep, and vigilance. High snake fearful individuals were randomly assigned to receive one session of video-exposure treatment in the morning or in the evening. Symptoms of snake phobia were assessed prior to treatment, after treatment and at two follow-ups. Furthermore, we assessed time since awakening, sleep quality (i.e., total sleep time and sleep efficiency), and vigilance prior to exposure as well as endogenous cortisol levels prior to and during exposure. We expected exposure in the morning to be more successful than exposure in the evening. Moreover, we explored the extent to which these effects are accounted for by endogenous cortisol levels, time since awakening, sleep quality and vigilance. Finally, we conducted explorative analyses to assess whether the group effects are moderated by this set of variables.
Data Analyses
Potential baseline differences as well as differences in cortisol levels between groups were tested by means of unpaired t-tests. Changes of arousal/anxiety ratings during the exposure session were examined by means of mixed ANOVAs.
In order to test our hypotheses, a series of multilevel models was fitted separately for BAT and SNAQ scores. We conducted two sets of analyses: A pre-post analysis including all assessments (pre-exposure, post-exposure, one-week follow-up, four-week follow-up) was focused on pre-post symptom change (time centered at pre-exposure). A post-follow-up analysis including only post-exposure, one-week follow-up, and four-week follow-up was run separately, focusing on the maintenance of symptom change from post-exposure to the follow-up assessments (time centered at post-exposure). In a first step, we constructed a baseline model, comprising the random and fixed effect of Time. Subsequently, we investigated group effects by including Group and the interaction between Time and Group as fixed effects. In order to test potential confounding effects of sex, we repeated these analyses including only female participants. The subsample of male participants was too limited (N = 11) to allow for separate analyses. We report descriptive statistics of outcome measures for female and male participants in the Supplementary Material.
In addition, we aimed to test the predictive effects of cortisol levels during exposure, psychomotor vigilance, time since awakening (TSW), and sleep quality for symptom change. To this end, we added the respective predictor and the Time × Predictor interaction to the baseline model and tested the improvement of model fit (χ
2 difference test). For all models that yielded a significant improvement of model fit beyond the baseline model, we planned to evaluate fit indices of models including the respective predictor and the model including Group to conclude which factors bears the strongest predictive value. Given that none of the analyses yielded a significant effect of Group, such comparisons were not necessary. Finally, we aimed to investigate moderator effects. Given that the Group factor was highly correlated with TSW and cortisol levels during exposure, these variables were excluded from moderator analyses to avoid issues arising from multicollinearity. Mean arousal level during the exposure session was investigated as an additional moderator since baseline analysis yielded unexpected group differences. All potential moderator variables were found to approximate a normal distribution, KS Test
p > 0.08. In order to investigate potential moderator effects of sleep quality and psychomotor vigilance, we evaluated whether a model including the interaction between the respective moderator, Group and Time improved model fit beyond the baseline model. All Level-2 predictors were grand-mean centered (Kreft et al.,
1995). The Level-1 predictor Time was centered at baseline (Singer & Willett,
2014). All multilevel models were fit using maximum likelihood estimation with the
lme4 package (Bates,
2010) in
R (Team,
2022). Significant interactions were probed using simple slopes techniques implemented in the
reghelper package (Hughes et al.,
2022). Slopes were estimated at ± 2
SD above/below the mean of the respective moderator variable. The alpha level was set to 0.05 for all analyses.
BAT scores were missing for 11 participants at post-assessment or follow-up assessments. One participant showed a mean PVT reaction time over 3 interquartile ranges above the upper quartile and was thus excluded from all analyses including the PVT. Due to data loss, actigraphy data of 34 participants (Evening group: n = 19, Morning group: n = 15) were not available for analysis. TSW was not documented by three participants. As a result, degrees of freedom vary across analyses.
Discussion
The current study set out to replicate previous research showing that exposure therapy is more effective in the morning than in the evening, while shedding further light on the involvement of cortisol levels, sleep, and vigilance. In contrast to previous research, we did not find that a video-based exposure session was more effective in the morning than in the evening. Both behavioral and subjective assessments of snake fear were found to decrease from pre- to post-intervention and from post- to follow-up. Controlling for baseline differences in arousal during the exposure session, revealed significantly higher subjectively experienced snake fear in the evening as opposed to the morning group. However, since this effect was not qualified by a significant interaction between Group and Time, we refrain from interpreting it in terms of intervention effects. Interestingly, we did find indications that vigilance and pre-exposure sleep efficiency may be involved in modulating daytime effects on exposure therapy. On the one hand, we found that vigilance levels were higher in the evening group and greater vigilance predicted a greater post-exposure increase of BAT scores and further increase of BAT scores in the follow-up period across both groups. On the other hand, we found that morning as opposed to evening exposure was associated with a stronger increase of BAT scores in the follow-up period, however this effect was only estimated for individuals with high pre-exposure sleep efficiency and estimated inversely for individuals with low pre-exposure sleep efficiency. However, this effect was only found in a restricted subsample for which actigraphy data was available (n = 16 in the evening group and n = 21 in the morning group). Moreover, sleep efficiency was generally rather high. Neither baseline sleep quality nor cortisol levels during exposure were found to predict treatment-related changes in behavioral or subjective snake fear.
Our findings on vigilance indicate that—contrary to our assumption—vigilance levels were higher in the evening than in the morning. Although some studies show that vigilance is higher in the morning, other studies indicate that this effect varies according to chronotype (Harrison et al.,
2007; Riley et al.,
2017). That is, evening types may show higher vigilance in the evening as opposed to the morning and morning types may show higher vigilance in the morning as opposed to the evening (Correa et al.,
2014; Venkat et al.,
2020). Our result could thus indicate that our participants who were largely recruited amongst university students were tested at their non-optimal time of day when assigned to the morning group. In order to further explore this possibility, we examined the distribution of morning and evening types and found that 16.7% of participants (evening types) were tested at their non-optimal time in the morning, whereas only 5.7% of participants (i.e., morning types) were tested at their non-optimal time in the evening. This disproportionate misalignment may have caused the baseline difference in PVT performance. Additionally, vigilance was found to predict the increase of BAT scores across time. Taken together, chronobiological factors may have prevented us from replicating the previous findings of Lass-Hennemann and Michael (
2014).
Our second finding indicates that a superior effect of morning exposure may be present, but only in individuals with high pre-exposure sleep efficiency and only in the follow-up period. These results suggest that the “morning exposure effect” may be hampered by insufficient nighttime sleep. As with the aforementioned possible interference of chronobiological factors, the moderation thus stresses that individual factors must be taken into account. For instance, getting up early in the morning to attend morning exposure therapy may affect preceding sleep quality, especially in younger populations with a tendency towards eveningness. The anticipation of exposure therapy in the morning could also cause difficulties falling asleep, thereby affecting pre-exposure sleep efficiency. Such individual factors should be considered when scheduling appointments with patients. In addition, it seems worthwhile to provide patients with psychoeducation and tools to improve sleep quality (e.g., sleep directed hypnosis; Friesen et al.,
2023) in order to boost pre-exposure sleep efficiency (and thereby morning vigilance levels).
Beyond these considerations, there are further explanations that may account for our failure to replicate the “morning exposure effect” as well as the previously reported correlations between cortisol levels and exposure therapy outcome (Lass-Hennemann & Michael,
2014; Meuret et al.,
2015,
2016). First, some studies did not show a significant association between cortisol levels during exposure and symptom change or even showed an inverse association (Kuhlman et al.,
2020). Kuhlman et al. (
2020) argue that these mixed findings are related to the fact that endogenous cortisol during treatment does not only reflect daytime variations but also cortisol reactivity, which may be linked to less symptom improvement throughout exposure therapy (see also Rauch et al.,
2017). To explore this possibility, we conducted separate analyses with pre-exposure cortisol levels (t
0) as predictor of symptoms. However, none of these analyses yielded a significant result. Moreover, in contrast to the cortisol reactivity hypothesis, there are several studies showing that patients do not experience a stress-related endogenous cortisol reaction to exposure therapy (Gustafsson et al.,
2008; Kellner et al.,
2012; Lass-Hennemann & Michael,
2014; Siegmund et al.,
2011). Second, we examined video-based in virtuo exposure whereas previous research examined in vivo exposure (Lass-Hennemann & Michael,
2014; Meuret et al.,
2015,
2016). We opted for the video-based approach, as in vivo exposure trials are not able to achieve full blinding, because the involved psychotherapists are often not blind to study hypothesis (Lass-Hennemann & Michael,
2014). Although allowing us to test effects under highly standardized conditions, this approach may have dampened exposure effects and thereby the potential of finding daytime differences. Moreover, we examined high snake anxious individuals, whereas our previous study examined individuals with spider phobia (Lass-Hennemann & Michael,
2014). Thus, anxiety levels may not have been sufficiently high to detect any daytime effects on exposure. However, it is important to note that we did find significant, albeit small, exposure effects both in terms of symptom changes and anxiety/arousal ratings during exposure. Finally, it is important to consider potential confounding effects of sex, since the morning acrophase of testosterone may interact with cortisol in generating the morning exposure effect (see e.g., Hutschemaekers et al.,
2020). In order to explore this possibility, we repeated our analyses while including only female participants. These analyses did not yield any significant group-related effects, thus paralleling our results presented above.
Overall, it is important to point out that effects of vigilance and pre-exposure sleep efficiency were only evident for behavioral but not for subjective fear indices. However, the BAT is considered the gold standard in the assessment of phobic fear and has been reported as the primary outcome measure in many studies on treatment of specific phobias (Lambe et al.,
2023). Moreover, our findings are in line with previous studies showing effects only for behavioral or self-report outcome measures (e.g., de Quervain et al.,
2011; Lass-Hennemann & Michael,
2014) and could indicate a lack of agreement between these measures (Reinecke et al.,
2009). In addition, effects of pre-exposure sleep efficiency were only evident in the follow-up period and not in our pre-post analyses. Though speculative, this finding could indicate that effects of pre-exposure sleep only emerge over time when intervention effects are diminished by the time lag between intervention and testing (for similar findings see Soravia et al.,
2014).
Additionally, several limitations of our study must be considered. First, we used actigraphy rather than polysomnography for the assessment of pre-exposure sleep, which is known to overestimate sleep duration and does not allow differentiating between different sleep stages (Marino et al.,
2013). However, actigraphy also has some advantages as it allows for a non-invasive and economic assessment of sleep quality in natural sleep settings. Due to practical considerations, we only collected actigraphy data for three nights prior to the exposure session. Future studies should consider assessing baseline sleep for a minimum of seven nights to improve reliability (Aili et al.,
2017).
Second, even though we included men and women in our study, the vast majority of our participants self-identified as women. Although more women suffer from snake phobia than men, the sex ratio in epidemiological studies is not as unequally distributed as in our study (Oosterink et al.,
2009). One major strength of our study is that it was preregistered with an a-priori sample size calculation. However, we failed to reach the desired sample size in several subanalyses, which may have limited statistical power. This concern especially applies to our analyses of actigraphy data. Our findings of the moderator analyses thus have to be interpreted with caution and require replication in adequately powered samples. In an effort to improve statistical power, we conducted exploratory analyses examining subjective SE (calculated based on sleep logs) as a moderator of symptom improvement. These analyses did not reveal any significant effects. While this seems to disconfirm our actigraphy-based findings, it is important to note the low level of agreement between actigraphy-based and sleep log-based assessment of sleep quality (Girschik et al.,
2012; McCall & McCall,
2012). Hence, inconsistent effects may be related to lack of agreement between SE measures rather than poor reliability of our findings in actigraphy-based analyses.
In summary, our preregistered experimental study aimed to replicate the “morning exposure effect” and is (to our knowledge) the first study, which systematically assessed different potential factors contributing to the “morning exposure effect”. Further research is needed to confirm our findings and generalize them to the wider population of individuals with anxiety disorders. Such research should aim to improve shortcomings of our study, while taking into account the predictors and moderators that we identified, namely, vigilance levels and pre-exposure sleep efficiency. Such studies may also consider investigating daytime effects within patients by varying daytime between repeated exposure session (Meuret et al.,
2015,
2016). Though preliminary, our study shines further light on the intricate relations between daytime effects, vigilance, and sleep, suggesting that clinicians should take all factors that are linked to these processes (e.g., chronotype, difficulties falling asleep) into account when scheduling individual exposure sessions. Simply scheduling exposure sessions in the morning does not seem to be sufficient to achieve optimal treatment effects.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.