Borderline personality disorder (BPD) is a mental illness that can affect many aspects of life. Patients with BPD often also suffer from sleep problems such as difficulty falling or staying asleep or frequent nightmares. The aim of the present study was to investigate the relationship between sleep and BPD-specific psychopathology and to investigate potential connections between sleep and the success of a disorder-specific treatment program.

Patients with BPD suffer from various psychological symptoms including emotional instability, anxiety, depression, self-harming behaviors, suicidal tendency, and problems concerning self-image and interpersonal relationships [16].

Additionally, patients with BPD often report subjective sleep problems such as decreased perceived sleep duration or quality [12, 24], more nightmares and higher levels of dream anxiety (for a review, please refer to [11, 26]). Moreover, differences in objective sleep characteristics as compared to healthy individuals have been shown, such as increased sleep fragmentation, reduced rapid eye movement (REM) latency, increased REM density (for a meta-analysis, please refer to [2, 25]), or increased delta power [21]. A more recent systematic review also compared studies on sleep in BPD to sleep in other clinical (mostly depressed) samples and stated that sleep disturbances in BPD are comparable to those in major depression, while not solely explained by comorbid depression since also BPD patients with no history of depression were affected [32]. Potential links between sleep and emotional dysregulation have also been investigated. One study revealed an association between sleep disturbances in patients with BPD and baseline negative emotion, while other links were not evident in the data [8]. Taken together, results on sleep in patients with BPD are not consistent (for a review, please refer to [30]).

Overall, many psychiatric conditions are known to be associated with general sleep problems. Some concrete sleep features can also be specifically associated with certain aspects of an illness. For example, signs of REM disinhibition like reduced REM latency, prolonged first REM sleep episode, and enhanced REM density have repeatedly been linked to the extent of depressive symptomatology (for a review, please refer to [19]). Furthermore, sleep problems are known to increase vulnerability to psychiatric disorders and influence prognosis (for a review, please refer to [14]). Especially in BPD, sleep disturbances could increase emotional instability and self-harming behaviors. In contrast, good sleep or specific sleep characteristics can even benefit treatment effects. A former study from the authors’ research group on sleep in BPD patients with comorbid posttraumatic stress disorder (PTSD) found a positive relationship between pretreatment total sleep time, REM sleep duration, and reduction of PTSD symptoms after narrative exposure therapy (NET; [31]). The role of REM sleep in psychiatric disorders could be related to its importance for processing affective memories and modulating daytime reactivity to emotional content [10, 15, 23].

The present study investigated the potential links between (subjective and objective) sleep, depression, and BPD-specific symptoms in a group of BPD patients. Furthermore, it analyzed how sleep quality and different sleep characteristics at the beginning of treatment might influence the outcome of an 8‑week semi-residential treatment program according to dialectical behavior therapy (DBT) [17, 20]. It was hypothesized that patients with more psychiatric symptoms would suffer from more sleep disturbances and, furthermore, that patients with better sleep would benefit more from the psychotherapeutic treatment.

Methods

Study design

At the beginning of the semi-residential DBT program, suitable participants were recruited. All patients who agreed to participate in the study were invited to the sleep laboratory. Prior to the sleep laboratory visits, psychological interviews and different questionnaires (see section “Psychological tests”) were conducted. Apart from the clinical interviews, all tests were repeated at the end of the DBT program and at 6‑month follow-up.

Participants

While 20 patients with BPD initially agreed to participate in the study, one patient left treatment after 13 days, one patient had to be excluded due to severe obstructive sleep apnea syndrome (apnea–hypopnea index ~50/h), and a third patient did not return the follow-up questionnaires; therefore, the data of 17 patients (age = 27.82 years, standard deviation [SD] = 7.57; 16 female) went into the final analysis. The diagnosis was confirmed via clinical assessment and the Structured Clinical Interview for DSM-IV Axis II Disorders (SCID-II). All patients had at least one additional diagnosis (moderate or major depressive episodes N = 13; abuse of alcohol, cannabis, or opioids N = 5; eating disorder N = 4; somatoform/somatization disorder N = 2; attention deficit hyperactivity disorder [ADHD] N = 2; posttraumatic stress disorder [PTSD] N = 1; obsessive compulsive disorder N = 1). All patients participated in a 6‑ to 8‑week semi-residential DBT program, including individual psychotherapy, skills training, social competence training, occupational therapy, and therapeutic exercise. All participants were free from benzodiazepines and Z drugs, but 7 patients received other sedative medications at night (doxepin, quetiapine, trazodone, agomelatine, and chlorprothixene) and 4 of these patients also received additional antidepressant medication in the morning. Another 3 patients received only antidepressants in the morning (fluoxetine, venlafaxine, sertraline, bupropion, or St. John’s wort). At T0 (where polysomnography [PSG] was recorded), 7 patients were free from any antidepressant or sedative medication; during treatment, one of these patients started quetiapine 37.5 mg at night and one patient started agomelatine 50 mg at night.

Psychometric tests

The SCID-II [7] and the Mini-International Neuropsychiatric Interview (MINI; [27]) were used to confirm the personality disorder diagnosis according to DSM-IV and to assess potential comorbidities. Furthermore, during the evening of the first sleep laboratory visit, all participants filled out the Beck Depression Inventory II (BDI-II; [3]); the Borderline Symptom List 95 (BSL-95; [5]); and the BSL supplement assessing the extent of self-harmful behaviors such as self-mutilation, high-risk behavior, or alcohol and drug consumption. Furthermore, patients answered a dissociative symptoms questionnaire (Fragebogen zu dissoziativen Symptomen [FDS] [9]), the Van Dream Anxiety Scale (VDAS; [1]), and the Pittsburgh Sleep Quality Index (PSQI; German version [6]).

PSG recording and scoring

At the beginning of the DBT program, all participants spent two nights in our sleep laboratory. The first night served as the accommodation night. We recorded 8‑hour PSG using a Somnoscreen Plus PSG system (Somnomedics, Randersacker, Germany). Electroencephalography (EEG) was registered from channels F3, F4, C3, C4, O1, and O2 referenced the contralateral mastoid. Electromyography (EMG) was recorded on the chin for sleep staging as well as from the bilateral anterior tibialis muscles for leg movements. Furthermore, electrooculography (EOG), electrocardiography (ECG), and respiratory effort were registered.

Sleep was manually scored according to the American Academy of Sleep Medicine (AASM) criteria by experienced raters using Somnomedics’ Domino software. Raters were unaware of the study hypotheses.

Statistical analysis

To test changes in psychometric measures from pre- to posttreatment and from pretreatment to follow-up, paired T‑tests (two-tailed) were calculated.

To investigate links between changes in psychometric measures and sleep characteristics prior to treatment, Pearson correlation coefficients were calculated. Total sleep time (TST), sleep onset latency (SOL), sleep efficiency (SE; SE = TST/time in bed), the duration and percentage of different sleep stages, and REM density (ratio of 3‑second mini-epochs including REMs to the total number of 3‑second mini-epochs of REM sleep) served as independent variables. Psychological pretreatment test scores as well as changes from pre- to posttreatment and from pretreatment to follow-up served as dependent variables. Some of our patients received antidepressants (n = 7) and/or sedative medication (n = 7), which we statistically controlled for by partial correlation wherever relevant.

Results

Behavioral pre–post results

Both average BDI score (T16 = 4.16, p < 0.001) and BSL-95 score (T16 = 2.29, p < 0.05) were decreased after treatment, and BDI but not BSL was still decreased at 6‑month follow-up as compared to baseline (BDI: T16 = 3.39, p < 0.01; Fig. 1). The BDI score was decreased in 82% of patients and the BSL-95 score in 76% of patients directly after treatment. Comparing patients who continued in- or outpatient therapy (N = 10) within the 6‑month follow-up period to those who did not continue treatment (N = 7), the first group still showed (by tendency) a stronger decrease in BSL and BDI symptom scores at follow-up (BDIpre–follow-up: T15 = −1.83, p < 0.10; BSLpre–follow-up: T15 = −2.15, p < 0.05). The amount of self-harmful behavior as assessed by the BSL supplement was decreased after treatment (T16 = 2.19, p < 0.05) but this was no longer the case after 6 months (T16 = −1.40, n. s.). Subjective sleep quality according to the PSQI was enhanced after treatment (T16 = 3.76, p < 0.01), but this was no longer the case at follow-up (T16 = 1.44, n. s.).

Fig. 1
figure 1

Beck Depression Inventory (BDI; a), Borderline Symptom List 95 (BSL-95; b), and Pittsburgh Sleep Quality Index (PSQI; c) scores prior to the DBT program, after the program, and at 6‑month follow-up. In the course of treatment, both depressivity and borderline symptomatology were significantly reduced, while sleep quality was enhanced. At 6‑month follow-up, positive effects on BDI were still present. **p < 0.01, *p < 0.05; error bars indicate standard deviations

Full size image

The FDS and VDAS scores did not differ after treatment. For all psychometric pre–post results, see Table 1.

Table 1 Psychometric results prior to and after treatment
Full size table

Overall, patients with worse initial BPD-specific symptoms benefited most from the treatment (BSLpre correlated with BSLpre–post: r = 0.49 p < 0.05; BSL supplementpre correlated with BSL supplementpre–post: r = 0.77, p < 0.01).

Sleep results prior to treatment

The participants slept for an average of 6 h 58 min (range 5 h 24 min to 8 h 26 min), with sleep efficiency ranging from 78.4% to 97.1%. For an overview of all sleep parameters, see Table 2. Except for patients with sedative medication having a longer TIB (T15 = −2.15, p < 0.05), none of the sleep parameters showed a significant difference between patients with (N = 7) and without (N = 10) sedative medication.

Table 2 Average sleep parameters for all 17 patients
Full size table

Associations between psychometric measures and sleep parameters prior to treatment

The BSL score was negatively correlated with REM duration (r = −0.55, p = 0.023; Fig. 2) and percent (r = −0.49, p < 0.047), meaning that the shorter the REM duration, the stronger the BPD symptomatology, and vice versa. The correlation remained significant when controlled for sedative (r = −0.51, p = 0.043) or antidepressant (r = −0.55, p = 0.027) medication.

Fig. 2
figure 2

Association between rapid eye movement (REM) duration and Borderline Symptom List (BSL) score prior to treatment. Patients with more REM sleep suffered from less BPD-specific symptoms, r = −0.55, p < 0.05

Full size image

Associations between sleep parameters and therapeutic outcome

Subjective sleep quality as measured by the PSQI prior to treatment and at the end of treatment was in both cases correlated with the change in BSL score at 6‑month follow-up (PSQIpre × BSLpre–follow-up: r = 0.55, p < 0.05; PSQIpost × BSLpost–follow-up : r = 0.68, p < 0.01), implying that good subjective sleep quality could favor a better long-term outcome (Fig. 3).

Fig. 3
figure 3

Association between subjective sleep quality as measured by the Pittsburgh Sleep Quality Index (PSQI; higher values indicate more sleep problems) and the change in Borderline Symptom List (BSL) score from prior to treatment (pre) to follow-up. Patients with better subjective sleep quality showed a stronger reduction in symptomatology, r = −0.55, p < 0.05

Full size image

The change in BDI and BSL scores from pre- to posttreatment was negatively correlated with SOL (BDI: r = −0.56, p = 0.021; BSL: r = −0.50, p = 0.041), meaning that the longer the SOL at the beginning of the 8‑week program, the stronger the improvement in depression.

The other objective sleep parameters did not yield any potential predictive ability for therapeutic outcome.

Discussion

As expected, the 8‑week DBT program effectively reduced both BPD-specific and depressive symptoms and also enhanced subjective sleep quality in patients with BPD. Patients with the highest symptom scores showed the strongest improvements. Furthermore, the extent of borderline symptoms prior to treatment was negatively correlated with REM sleep, i.e., patients with more REM sleep suffered from fewer BPD-specific symptoms and vice versa. Interestingly, good subjective sleep quality seemed to predict long-term treatment success concerning BPD-specific symptoms. It must be noted, however, that the study was strictly observational and, due to the lack of a control group, had a preliminary character.

As mentioned above, the DBT program had positive effects on depressive and BPD symptoms, including self-harmful behaviors. This was expected based on earlier positive evaluations of comparable DBT programs (e.g., [4, 22]). However, only depression and neither BPD-specific symptoms nor subjective sleep quality were still significantly improved at 6‑month follow-up as compared to the pretreatment scores. In other words, the severity of symptoms returned to almost pretreatment levels. The authors believe that the recurrence of symptoms underlines the importance of briefly continuing outpatient therapy after hospital treatment in BPD patients to stabilize learned behaviors and foster a positive long-term outcome.

As an initial primary goal of this study, the potential association between psychiatric symptoms and objective sleep parameters at the beginning of treatment was also investigated. A lower BSL-95 score was linked to longer REM duration, meaning that the more REM sleep patients had, the fewer BPD symptoms they experienced. It has repeatedly been shown that BPD patients tend to show more REM sleep [28, 32]; however, within the group of BPD patients, the amount of REM sleep might also serve as a protective factor. As mentioned earlier, REM sleep has previously been related to normalizing neural and behavioral reactivity after emotional distress [10, 18, 23, 29]. Consequently, REM sleep in BPD patients could promote emotion regulation.

The extent of depressive symptoms in the current sample was, however, not associated with REM sleep. This is different from findings in depressed patients, where REM sleep parameters and symptomatology have previously been positively connected (for a review, please refer to [19]). This might be related to a different pathogenesis of depression in BPD patients as compared to primarily depressed patients. Patients with BPD often react with stronger emotions and a wider emotional variety as compared to patients with primary depression, who often show a rather flattened affect.

Concerning the influence of sleep on treatment outcome, the decrease in depressive and BPD-specific symptomatology directly after treatment was linked to a longer pretreatment sleep onset latency (SOL). Longer SOL can be caused by hyperarousal, anxiety, and rumination. This would fit with the current finding that patients with stronger symptoms benefited more from the treatment program. However, the average SOL in the present study was 16 min, which is considered normal. It should be noted that SOL is a rather variable parameter that can easily be affected by, e.g., the unknown environment in the sleep lab.

A positive association of subjective sleep quality with the long-term treatment outcome was also found. Better subjective sleep quality at the beginning and at the end of the treatment was associated with a stronger BSL change from pre-treatment to follow-up and from posttreatment to follow-up. These results suggest a link between generally good sleep quality and the risk/severity of relapse of BPD symptoms. The authors hope that future psychotherapy research will focus more on the role of sleep in treatment success. As a side note, in the present study, subjective sleep quality did not strongly correlate with objective parameters [13]. This has often been observed in clinical settings and has also been described in the literature.

The results are, however, limited by the small number of patients who participated in the study and the lack of a control group. Furthermore, similar to patients included in the study by Schredl and colleagues [25], the current patients were rather good sleepers (at least within the sleep laboratory setting), with an average sleep efficiency of 89.76% (SD = 5.47%) and SOL of 16.48 min (SD = 14.10 min), which decreased the variance in sleep parameters and might have limited the investigation of the influence of objective sleep parameters on treatment outcome in a small sample. A further limitation is that most of the patients had either antidepressant or sedative medication or both, which might have interfered with the results to some extent.

In conclusion, most of the patients included in the current study benefited from the DBT treatment, at least in the short term. For a positive long-term outcome, continuing treatment with an outpatient psychotherapy program right away might be vital. Furthermore, the analyses yielded associations between sleep parameters and symptomatology in BPD. Subjective sleep quality could be a positive predictor of the long-term treatment outcome; therefore, sleep and sleep quality should receive more attention in psychiatric treatment. Despite the limitations of the present study, the authors believe that the presented results point toward interesting associations between sleep, psychopathology, and treatment outcome which warrant further investigation.