If Time Does Not Heal Wounds, Then Do Opioid Antagonists?
A Randomized Placebo-Controlled Study on Analgesia in
Borderline Personality Disorder
E. M. K. Drews (11013923) Research Master Psychology Specialization: Clinical Psychology
Supervisor: Prof. Arnoud Arntz Second Assessor: Dr. Henk Cremers
Amsterdam, 19th of January 2018
Table of Contents
Abstract ... 1
Introduction ... 2
Borderline Personality Disorder – A Pain Paradox? ... 2
Pain Perception and the Endogenous Opioid System in Borderline Personality Disorder .... 3
Self-Injury – A Sore Temptation for Borderline Patients? ... 4
Childhood Trauma and Dissociative Symptoms – Two Crucial Factors? ... 5
Methods ... 8 Participants ... 8 Materials ... 14 Emotion Questionnaire ... 14 Pharmacological Treatment ... 14 Experimental Manipulation ... 14 Pain Tests ... 15
Forgione-Barber Pain Stimulator ... 15
Cold Pressor Test ... 15
Procedure ... 16
Statistical Analyses ... 18
Results ... 21
Manipulation Check ... 21
Stress-Induced Opioid-Mediated Analgesia ... 21
Stress-Induced Analgesia in BPD Patients ... 21
Stress-Induced Analgesia in BPD-NP Patients. ... 22
Stress-Induced Opioid-Mediated Analgesia in BPD Patients. ... 22
Stress-Induced Analgesia, Childhood Trauma, and Dissociation ... 32
Discussion ... 36
Limitations ... 39
Conclusion ... 41
References ... 42
Appendix A Additional Interviews and Questionnaires ... 53
Appendix B Study Procedure ... 55
Appendix C Pain Scores in Seconds for Placebo and Naltrexone Conditions ... 57
Appendix D Contrasts for Movie Clips and Experimental Groups ... 59
Appendix E Goodness of Fit Indices for the Multilevel Models ... 60
Appendix F Remaining Figures for Threshold and Tolerance Assessments ... 62
Appendix G Ratings for the Emotion Questionnaire ... 66
List of Tables
Table 1 Sample Demographics for the Experimental Groups ... 10
Table 2 Multilevel Model (Threshold) Excluding Three-Way Interactions ... 26
Table 3 Multilevel Model (Threshold) Including Three-Way Interactions ... 27
Table 4 Multilevel Model (Tolerance) Excluding Three-Way Interactions ... 29
Table 5 Multilevel Model (Tolerance) Including Three-Way Interactions ... 30
Table 6 Multiple Regression Analyses for the Combined Pain Measure ... 33
Table 7 Multiple Regression Analyses for the Forgione-Barber Pain Stimulator ... 34
Table 8 Multiple Regression Analyses for the Cold Pressor Test ... 35
Table 9 Study Procedure ... 55
Table 10 Pain Scores in Seconds (Placebo) ... 57
Table 11 Pain Scores in Seconds (Naltrexone) ... 58
Table 12 Contrasts for the Movie Clips ... 59
Table 13 Contrasts for the Experimental Groups ... 59
Table 14 Goodness of Fit Indices for Multilevel Models (Threshold) ... 60
Table 15 Goodness of Fit Indices for Multilevel Models (Tolerance) ... 61
Table 16 Ratings for the Emotion Questionnaire (Placebo) ... 66
List of Figures
Figure 1. Schematic Illustration of the Experimental Crossover Design ... 17
Figure 2. Differences in Pain Threshold for Combined Pain Measure (Placebo) ... 24
Figure 3. Differences in Pain Threshold for the Combined Pain Measure (Naltrexone) ... 24
Figure 4. Differences in Pain Threshold for the Combined Pain Measure (Placebo) ... 25
Figure 5. Differences in Pain Threshold for the Combined Pain Measure (Naltrexone) ... 25
Figure 6. Differences in Pain Threshold for the FBPS (Placebo) ... 62
Figure 7. Differences in Pain Threshold for the CPT (Placebo). ... 62
Figure 8. Differences in Pain Threshold for the FBPS (Naltrexone) ... 63
Figure 9. Differences in Pain Threshold for the CPT (Naltrexone) ... 63
Figure 10. Differences in Pain Tolerance for the FBPS (Placebo) ... 64
Figure 11. Differences in Pain Tolerance for the CPT (Placebo) ... 64
Figure 12. Differences in Pain Tolerance for the FBPS (Naltrexone) ... 65
Figure 13. Differences in Pain Tolerance for the CPT (Naltrexone) ... 65
Figure 14. Average Happiness Ratings for the Four Experimental Groups ... 73
Figure 15. Average Sadness Ratings for the Four Experimental Groups ... 74
Figure 16. Average Anxiety Ratings for the Four Experimental Groups ... 76
Figure 17. Average Anger Ratings for the Four Experimental Groups ... 77
Figure 18. Average Restlessness Ratings for the Four Experimental Groups ... 78
Figure 19. Average Surprise Ratings for the Four Experimental Groups ... 79
Figure 20. Average Valence Ratings for the Four Experimental Groups ... 80
Figure 21. Average Arousal Ratings for the Four Experimental Groups ... 81
Abbreviations
AIC Akaike Information Criterion ANOVA Analysis of Variance
APA American Psychiatric Association
AR1 First-Order Autoregressive Covariance Structure ARMA11 Autoregressive Moving Average Covariance Structure BIC Bayesian Information Criterion
BPD Borderline Personality Disorder
BPD-NP Borderline Personality Disorder - No Pain Patients BPD-P Borderline Personality Disorder - Pain Patients
CC Clinical Controls
CPT Cold Pressor Test
DSM-IV Diagnostic and Statistical Manual of Mental Disorders (4th Edition)
EOS Endogenous Opioid System
FBPS Forgione-Barber Pain Stimulator
HC Healthy Controls
ITEC Interview for Traumatic Events in Childhood
SAM Self-Assessment Manikin
SIA Stress-Induced Analgesia
SIB Self-Injurious Behaviour
UN Unstructured Covariance Structure
Abstract
Self-injurious behaviour (SIB) is a common but harmful emotion regulation strategy in Borderline Personality Disorder (BPD). However, it is unclear if aberrant pain perception reinforces SIB in BPD due to a dysregulated endogenous opioid system (EOS) and whether stress, trauma, and dissociation impact this association. Using a randomized crossover design, we compared pain threshold and tolerance based on two laboratory pain tasks in BPD patients (n = 32), clinical controls (n = 14), and healthy controls (n = 16) following neutral and stress-inducing movie clips for naltrexone and placebo treatments. To further determine if analgesia, hence pain insensitivity, is ubiquitous in BPD patients, these patients were distinguished based on increased and decreased pain perception during SIB (BPD-No Pain [BPD-NP]; BPD-Pain [BPD-P]). Multilevel models showed that pain perception was reduced in BPD-NP patients after the stress induction, however, naltrexone did not augment pain perception in BPD patients. Multiple regression analyses demonstrated that neither childhood trauma nor dissociation predicted stress-induced analgesia when controlling for baseline pain perception. Although stimulus choice and short-term naltrexone administration limit the interpretation of these findings, the current research implies that stress-regulation techniques for self-injurious BPD patients need to be implemented in clinical practice.
Keywords: Borderline personality disorder, self-injury, pain, endogenous opioids, emotion regulation, childhood trauma, dissociation
Introduction Borderline Personality Disorder – A Pain Paradox?
Individuals suffering from Borderline Personality Disorder (BPD) are both sensitive and insensitive to pain, with self-injurious behaviour (SIB) being one of the most frequent but dangerous symptoms. On the one hand, clinical observations indicate that many BPD patients are unable to tolerate ongoing pain, leading to overutilization of pain medications (Sansone & Sansone, 2007). Accordingly, research showed that a considerable proportion of patients seeking buprenorphine treatment1 also suffer from BPD (44%, New & Stanley, 2010). On the other hand, SIB, hence the direct and deliberate destruction or alteration of bodily tissue (Nock, 2009), is very common in BPD patients (70% - 80%; Bohus et al., 2000). Research further demonstrated that 50% - 75% of self-injurious BPD patients display analgesia, thus insensitivity to pain (Leibenluft, Gardner, & Cowdry, 1987). Taken together, it seems paradoxical that many BPD patients react hypersensitive to chronic pain but engage in behaviour causing acute pain. It is therefore unclear if BPD patients cope adversatively with chronic and acute pain or if pain perception is heterogeneous in BPD patients. Corresponding to the idea that pain perception is heterogeneous in BPD patients, Russ and colleagues (1992) showed that some BPD patients did not perceive SIB as painful (BPD-No Pain [BPD-NP]) while others perceived SIB as very painful (BPD-Pain [BPD-P]). The authors could further demonstrate that BPD-NP patients are characterized by a more severe psychopathology as they reported a stronger mood elevation through SIB (Kemperman, Russ, & Shearin, 1997), a greater number of suicide attempts, as well as a higher frequency of childhood trauma and dissociation (Russ, Shearin, Clarkin, Harrison, & Hull, 1993). Therefore, promising explanations for the pain paradox as related to the endogenous opioid system (EOS), characteristics of the BPD
1
buprenorphine is a nonselective, mixed agonist-antagonist opioid receptor modulator, which is prescribed to treat chronic and acute pain. The interested reader is referred to the review by Sansone and Sansone (2006) for an in-depth overview.
aetiology, such as emotion dysregulation, dissociation, and childhood trauma, will be provided in the following paragraphs.
Pain Perception and the Endogenous Opioid System in Borderline Personality Disorder
The EOS is crucial for the experience of physical, emotional, and social pain as it modulates responses to noxious and stressful stimuli (New & Stanley, 2010). In fact, endogenous opioids reduce the experience of pain and may even induce feelings of euphoria (Nock, 2010). Due to presumed changes in pain perception and the high prevalence of SIB, EOS dysfunctioning likely plays a focal role for the aetiology of BPD. Correspondingly, the endogenous opioid theory (Bandelow, Schmahl, Falkai, & Wedekind, 2010) states that a dysregulation of the EOS in BPD is either caused by subsensitive opioid receptors or by low basal levels of endogenous opioids. In line with this theoretical approach, Prossin, Love, Koeppe, Zubiete, and Silk (2010) demonstrated abnormalities in µ-receptor concentrations as well as aberrant EOS responses to a negative emotional challenge in BPD patients. Similarly, the opioid deficit model (New & Stanley, 2010) assumes that BPD patients suffer from pre-existing endogenous opioid deficits, so that SIB may function as self-medication to generate endogenous opioids. Put differently, BPD patients may artificially put their body into survival mode to take advantage of the endogenous opioid rush generated through SIB (Bandelow et al., 2010; Russ, Roth, Kakuma, Harrison, & Hull, 1994). Additionally, pain perception during SIB may be pathologically attenuated due to endogenous opioid analgesia, which has been defined as decreased pain perception due to elevated levels of endogenous opioids in the body (Skinner, Damasceno, Gomes, & de Almeida, 2011). While empirical evidence of altered endogenous opioid levels in BPD patients is limited, opioid antagonists may reduce the rewarding properties of SIB. In particular, opioid antagonists such as naloxone and naltrexone block opioid receptors but do not have an intrinsic effect and research showed that naltrexone not only increased pain
ratings in healthy individuals (Burns et al., 2009) but also decreased analgesia in BPD patients (Bohus et al., 1999).
Self-Injury – A Sore Temptation for Borderline Patients?
As mentioned before, the EOS not only regulates responses to noxious stimuli but also to stressful, uncontrollable stimuli (Pitman, van der Kolk, Orr, & Greenberg, 1990) and inescapable stress has been shown to induce analgesic effects (stress-induced analgesia [SIA]; Maier, Sherman, Lewis, Terman, & Liebeskind, 1983; McCown, Galina, Johnson, DeSimone, & Posa, 1993; Willis et al., 2017). Accordingly, SIB may be viewed as an emotion regulation strategy2 (Chapman, Gratz, & Brown, 2006) since SIB is usually initiated with the intention to reduce negative emotions such as aversive tension and self-hatred (McKenzie & Gross, 2014). Further, BPD patients frequently report feelings of relief or well-being after self-injury (Bohus et al., 2000; Kemperman et al., 1997; Russ et al., 1992). In other words, engaging in SIB may help to regulate a psychological sensation, such as negative affect, by using an intense physical sensation, such as physical pain. Evidence for the link between emotion regulation and SIB has also been provided in recent research. In a study by Reitz and colleagues (2012), BPD patients were compared to healthy controls (HC), and stress induction was followed either by mild skin lesions or a sham treatment. As predicted by the affect-regulation model of SIB (Klonsky, 2007), BPD patients displayed elevated baseline levels of aversive tension but less aversive tension after skin lesions, which likely indicates that tissue damage impacts stress regulation in BPD patients (Reitz et al., 2012). Similarly, a recent study by Willis and colleagues (2017) demonstrated that nociceptive input leads to stress reduction in BPD patients. Taken together, these findings not only suggest that SIB has a stress-reducing function but also that analgesia accompanies aversive tension in BPD frequently. However, at present it is unclear if endogenous opioids affect SIA in BPD-NP patients as Russ and colleagues (1994) reported that
2emotion regulation has been defined as “the process by which individuals influence which emotions they have,
naloxone altered pain ratings neither in BPD-NP nor BPD-P patients. As prior research showed that long-term naltrexone treatment decreases rates and frequency of SIB (Schmahl et al., 2012; Sonne, Rubey, Brady, Malcolm, & Morris, 1996), the effect of opioid antagonists on analgesia in BPD patients – and especially BPD-NP patients – requires further investigation.
Childhood Trauma and Dissociative Symptoms – Two Crucial Factors?
Finally, the importance of the EOS has been emphasized based on the conditioned endogenous opioid response hypothesis, which states that stressful traumatic experiences in childhood and adolescence are associated with an increased endogenous opioid release. Subsequently, re-exposure to a stressful stimulus resembling the original trauma may cause endogenous opioid analgesia (van der Kolk & Saporta, 1991). Childhood trauma is not only an aetiological factor in BPD (Ball & Links, 2009) but also increases the likelihood that BPD patients display suicidal behaviour (Soloff, Lynch, & Kelly, 2002). Research further indicated that trauma predicts general and somatoform dissociation3 (Nijenhuis, Spinhoven, van Dyck, van der Hart, & Vanderlinden, 1998; Spiegel et al., 2011) and a close link between stress and dissociation in BPD has been reported (Ludäscher et al., 2007). Dissociative symptoms are characteristic for the clinical picture of BPD4 (Krause-Utz et al., 2017) and refer to the subjective detachment from stressful experiences such as nociception and overwhelming emotions (Nijenhuis, Vanderlinden, & Spinhoven, 1998). Moreover, research showed that the propensity to experience dissociation affects pain perception (Briere & Eadie, 2016; Russ, Campbell, Kakuma, Harrison, & Zanine, 1999) and that both intensity and duration of dissociative symptoms correlate with self-injury in BPD (Philipsen, Schmahl, & Lieb, 2004). However, it is yet unclear if the link between dissociation, childhood trauma, and analgesia is
3general dissociation refers to psychological and somatoform dissociative symptoms while somatoform
dissociation solely refers to the physical properties of dissociation such as for instance feelings of numbness.
4dissociative symptoms occur in 75% to 80% of BPD patients and entail depersonalization, dissociation, and
derealisation. Thereby, dissociative symptoms can be defined as disruptions in usually integrated functions of consciousness, perceptions, identity, memory, and affect (Ludäscher et al., 2007). The interested reader is referred to the review by Korzekwa, Dell, Links, Thabane, and Fougere (2009) for an in-depth overview.
particularly pronounced during stress and if especially somatoform dissociative symptoms elicit analgesia. Therefore, the question of whether SIA is heterogeneous in BPD patients due to childhood trauma and dissociation is still to be clarified.
To summarize, previous research indicates that pain perception in BPD patients is aberrant following exposure to uncontrollable, stressful stimuli and that EOS dysfunctioning leads to opioid-mediated analgesia, which would explain why BPD patients engage in self-injury frequently. Stress-induced opioid-mediated analgesia can potentially be counteracted by opioid antagonists, which theoretically regulate pain perception and may consequently decrease SIB in BPD patients. Further, research has yet to determine if stress-induced opioid-mediated analgesia is specific to BPD patients who do not report pain during self-injury (BPD-NP) since initial studies indicated that pain perception in BPD patients may be heterogeneous due to histories of childhood trauma and propensities for dissociation. Consequently, the current study aims to examine pain perception and EOS functioning after neutral and stress-inducing stimuli to answer three pivotal questions:
1. Is analgesia in BPD patients stress-induced and can pain perception be distinguished for BPD-NP patients?
1.1. We hypothesize that analgesia in BPD patients is stress-induced, hence that pain perception in BPD patients is lower after stress-induced movie clips compared to neutral movie clips when measured with laboratory pain tests.
1.2. We anticipate that this phenomenon is particularly pronounced for BPD-NP patients.
2. Do endogenous opioids affect stress-induced analgesia in BPD patients and can stress-induced opioid-mediated analgesia be distinguished for BPD-NP patients?
2.1. We expect that endogenous opioids modulate stress-induced analgesia in BPD patients so that these patients display less analgesia in the stress condition after receiving an opioid antagonist compared to placebo treatment.
2.2. We expect that this stress-induced opioid-mediated analgesia is specific to BPD-NP patients and therefore does not occur in BPD-P patients, clinical controls, or healthy controls.
3. Do childhood trauma and dissociation affect stress-induced analgesia in BPD patients?
3.1. We expect that BPD patients suffering from childhood trauma display more analgesia than BPD patients without childhood trauma in the placebo stress condition.
3.2. We hypothesize that BPD patients suffering from frequent dissociation display more analgesia than BPD patients who do not to dissociate or who infrequently dissociate in the placebo stress condition.
Methods Participants
Thirty-two patients diagnosed with BPD, 19 clinical controls (CC) either diagnosed with Cluster C personality disorder, an anxiety disorder, or a mood disorder5, and 16 HC participated in the study. Female participants between 18 and 51 years (M = 30.24, SD = 9.10) were recruited from the RIAGG Maastricht (BPD, CC), University Hospital RWTH Aachen (BPD, CC), and Maastricht University (HC). BPD patients were assigned to the BPD-NP group (n = 16) if they reported not usually feeling pain during SIB and to the BPD-P group if they reported usually feeling pain during SIB (n = 16). Frequency of SIB during the last two years was higher in the BPD-NP group (M = 231.81, SD = 367.35) than in the BPD-P group (M = 176.00, SD = 344.53). The amount of present BPD criteria was roughly similar for BPD-NP (M = 6.38, SD = 1.36) and BPD-P patients (M = 6.63, SD = 1.50) and similar propensities to experience dissociation were reported by BPD patients (NP: M = 1.99, SD = 0.70; BPD-P: M = 2.06, SD = 0.60). However, BPD-P patients reported more childhood trauma (M = 44.07, SD = 20.26) than BPD-NP patients (M = 39.68, SD = 23.05). Further, HC and CC were matched to the BPD group based age (± three years). Exclusion criteria for all participants included psychotic and bipolar disorders, pregnancy, breast feeding, acute alcohol and substance use disorders, being inebriated and intoxicated during the study, opiate or sedative use ≤ 10 days before participation, low average IQ scores6, medical contraindications to pain tests and naltrexone such as acute hepatitis or liver failure, and insufficient language skills. Further, BPD patients currently receiving treatment were excluded unless treatment was commenced within one month. CC were excluded when meeting ≥ two BPD criteria. HC were excluded when
5as defined by the DSM-IV-TR (APA, 2000) and measured with the Structured Clinical Interview for DSM-IV
Axis II Disorders (SCID-II; First, Gibbon, Spitzer, Williams, & Benjamin, 1997). For CC, a mixed subthreshold diagnosis within Cluster C was also accepted, given that individuals displayed specific traits of personality disorders without fitting into any of the other personality disorder diagnoses.
currently suffering from Axis I or Axis II disorders (assessed with the SCID-I and SCID-II) or meeting ≥ two BPD criteria. Demographic details for all experimental groups are shown in Table 1. The study was approved by the Medical Ethics Committee of Maastricht University.
Table 1
Sample Demographics for the Experimental Groups
Group; mean ± SD or no. (%)
Characteristic BPD-NP (n = 16) BPD-P (n = 16) CC (n = 19) HC (n = 16) Age, yr. 29.31 ± 10.10 30.00 ± 9.32 28.61 ± 7.81 • 33.25 ± 9.34 Educational level LBO (%) 1 (6.7) à 0 (0) 1 (5.6) 0 (0) MVO (%) 2 (13.3) 5 (31.3) 4 (22.2) 2 (12.5) HVO (%) 2 (13.3) 2 (12.5) 1 (5.6) 2 (12.5) MBO (%) 6 (40.0) 4 (25.0) 8 (44.4) 1 (6.3) VWO (%) 2 (13.3) 3 (18.8) 3 (16.7) 4 (25.0) HBO (%) 2 (13.3) 2 (12.5) 0 (0) 6 (37.5) WO (%) 0 (0) 0 (0) 1 (5.6) 1 (6.3) Relationship status Partnership (%) 9 (60.0) à 9 (56.3) 8 (42.1) 10 (62.5) Married (%) 3 (20.0) à 1 (6.3) 5 (26.3) 5 (31.3)
Employment status Working (%) 7 (46.7) à 4 (25.0) 6 (33.3) • 12 (75.0) Welfare (%) 5 (33.3) à 9 (56.3) 7 (38.9) • 2 (12.5) BPD Criteria 6.38 ± 1.36 6.63 ± 1.50 0.58 ± 0.69 0 ± 0 BPD Checklist 2.23 ± 0.65 2.26 ± 0.46 1.55 ± 0.51 1.09 ± 0.14 ¨ Dissociation Questionnaire 1.99 ± 0.70 2.06 ± 0.60 1.68 ± 0.48 1.25 ± 0.19
Interview for Traumatic Events in Childhood 39.69 ± 23.05 44.07 ± 20.26 2.82 ± 3.05 10.04 ± 5.77
Sexual Abuse 3.72 ± 4.51 6.62 ± 7.94 0 ± 0 1.49 ± 1.56 Physical Abuse 12.23 ± 9.98 13.82 ± 13.09 0 ± 0 2.32 ± 3.59 Emotional Abuse 13.19 ± 9.59 14.69 ± 9.14 2.82 ± 3.05 4.53 ± 1.52 Emotional Neglect 10.54 ± 5.31 8.94 ± 3.35 0 ± 0 1.70 ± 2.34 Axis I Disorders No diagnosis (%) 2 (12.5) 1 (6.3) 1 (5.3) 15 (93.8)
Major Depressive Disorder (%) 10 (62.5) 10 (62.5) 6 (31.6) Ñ 1 (6.3) ∆
Dysthymic Disorder (%) 1 (6.3) 0 (0) 1 (5.3) 0 (0)
Brief Psychotic Disorder (%) 1 (6.3) 0 (0) 0 (0) 0 (0)
Generalized Anxiety Disorder (%) 0 (0) 2 (12.5) 3 (15.8) 0 (0)
Panic Disorder with Agoraphobia (%) 0 (0) 1 (6.3) 1 (5.3) 0 (0)
Panic Disorder (%) 2 (12.5) 1 (6.3) 3 (15.8) 0 (0)
Specific Phobia (%) 1 (6.3) 1 (6.3) 3 (15.8) 0 (0)
Social Phobia (%) 1 (6.3) 3 (18.8) 5 (26.3) 0 (0)
Obsessive-Compulsive Disorder (%) 1 (6.3) 0 (0) 1 (5.3) 0 (0)
Posttraumatic Stress Disorder (%) 4 (25.0) 3 (18.8) 2 (10.5) 0 (0)
Dissociative Identity Disorder (%) 0 (0) 1 (6.3) 1 (5.3) 0 (0)
Somatization Disorder (%) 0 (0) 0 (0) 1 (5.3) 0 (0)
Hypochondriasis (%) 0 (0) 0 (0) 1 (5.3) 0 (0)
Dyspareunia (%) 0 (0) 0 (0) 3 (15.8) 0 (0)
Anorexia Nervosa, full remission (%) 0 (0) 1 (6.3) 0 (0) 0 (0)
Bulimia Nervosa (%) 3 (18.8) 3 (18.8) 3 (15.8) 0 (0)
Eating Disorder NOS (%) 0 (0) 0 (0) 1 (5.3) 0 (0)
Cannabis Dependence, full remission (%) 1 (6.3) 2 (12.5) 0 (0) 0 (0) Sedative, Hypnotic, or Anxiolytic Abuse, full remission (%) 2 (12.5) 1 (6.3) 1 (5.3) 0 (0)
Intermitted Explosive Disorder (%) 1 (6.3) 0 (0) 0 (0) 0 (0)
Adjustment Disorder 0 (0) 0 (0) 1 (5.3) 0 (0) Axis II Disorders Paranoid PD (%) 0 (0) 1 (6.3) 1 (5.6) 0 (0) Schizoid PD (%) 0 (0) 0 (0) 1 (5.6) 0 (0) Obsessive-Compulsive PD (%) 1 (6.3) 2 (12.5) 1 (5.9) 0 (0) Histrionic PD (%) 0 (0) 2 (12.5) 0 (0) 0 (0) Dependent PD (%) 0 (0) 2 (12.5) 0 (0) 0 (0) Avoidant PD (%) 2 (12.5) 2 (12.5) 6 (33.3) 0 (0) PD NOS (%) 3 (18.8) 2 (12.5) 3 (16.7) 0 (0)
Note. Mean scores were used to report BPD Criteria, BPD Checklist scores, and Dissociation Questionnaire scores. Total scores were used for the Interview of Traumatic Events in Childhood and for the corresponding subscales. Adjustment disorder includes adjustment disorder with depressed mood and adjustment disorder with anxiety depressed mood. Major Depressive Disorder (MDD) includes MDD single episode (full remission), MDD recurrent (unspecified), and MDD recurrent (partial remission). à Data unavailable for one BPD-NP patient; • Data unavailable for one CC; Ñ Probable diagnosis for one CC; ¨ Data unavailable for one HC. ∆ One HC participant was diagnosed with MDD recurrent (unspecified). BPD is not included as all BPD-NP and BPD-P had received the respective diagnosis while none of the control participants was diagnosed with BPD.
Materials7
Emotion Questionnaire. The Emotion Questionnaire consisted of two parts. First,
100-mm Visual Analogue Scales (VAS) were used to examine degrees of happiness, sadness, anxiety, anger, restlessness, and surprise with high scores corresponding to strong emotions. Internal validity was good in previous studies (e.g., Arntz, Dietzel, & Dreessen, 1999). Second, valence, arousal, and dominance were measured based on Self-Assessment Manikins (SAMs; Bradley & Lang, 1994), which are non-verbal pictorial techniques to determine affective reactions to visual stimuli. For each dimension, participants rated on 9-point scales which of the five manikins reflected their current emotional state best with higher scores referring to increased arousal, valence, and dominance. Internal consistencies were good in previous research (Backs, Silva, & Han, 2005).
Pharmacological treatment. The effect of endogenous opioids was manipulated with
50 mg naltrexone, which blocks the effects of endogenous opioids (Moghaddas, Dianatkhah, Ghaffari, & Ghaeli, 2017). Naltrexone is a nonspecific competitive opioid antagonist with highest affinity for µ-receptors. Naltrexone plasma levels peak within one hour of dosing and the mean elimination half-life value is four hours (Meyer, Straughn, Lo, Schary, & Whitney, 1984). Albochin was administered as placebo treatment. The placebo and naltrexone capsules looked identical and the order of pharmacological treatments was randomized and double-blind.
Experimental manipulation. Participants watched six movie clips during two
pharmacological conditions. Both times, participants first watched a neutral movie clip (15 minutes), a stress-inducing movie clip (15 minutes), and another neutral movie clip (20 minutes). The stress-inducing movie clips were taken from the movie “No Child of Mine”, which portrays the physical and sexual abuse as well as emotional neglect of a girl throughout
7questionnaires and interviews which are not directly relevant to the experimental pain assessment are described
her childhood. The neutral clips were taken from animal documentaries and accompanied by classical music. To avoid learning effects, different movie clips were chosen for the two pharmacological conditions. The order of movie clips was counterbalanced between conditions, resulting in 16 unique orders of clips and pharmacological conditions.
Pain tests8.
Barber Pain Stimulator. Pressure pain was measured with the Forgione-Barber pain stimulator (FBPS; Forgione & Forgione-Barber, 1971), which applies a 1.7 x 16 mm plexiglas wedge with a pressure of 62.5 g/mm2 to the middle phalanx of the middle finger of the nondominant hand. Using a stopwatch, the FBPS was applied for a maximum period of two minutes and participants were instructed to indicate pain threshold and tolerance. The maximum time was noted when participants did not discontinue the stimulation.
Cold Pressor Test. The Cold Pressor Test (CPT; Russ et al., 1992) was used to measure thermal pain. To ensure that participants had the same start temperature, participants first put their dominant hand into a container filled with 32°C warm water. Subsequently, participants placed their dominant hand into a container filled with 2°C cold water. Participants were further instructed to plunge their hands under water to the bottom of their wrist without placing the hand on the bottom of the bin. The CPT was applied for a maximum period of four minutes and participants were instructed to indicate pain threshold and tolerance. The maximum time was noted when participants did not discontinue the stimulation.
8pain tests were used to measure pain threshold and pain tolerance in seconds. Pain threshold, hereinafter called
threshold, is defined as “the point at which a pain-inducing stimulus is subjectively experienced as painful”. Pain tolerance, hereinafter called tolerance, is defined “the time taken for the research participant to discontinue exposure to the painful stimulus” (Hooley, Ho, Slater, & Lockshin, 2010, p. 171). Pain perception and sensitivity refer to both threshold and tolerance.
Procedure
Participants were recruited from two sites in the Netherlands (Maastricht, Heerlen) and one site in Germany (Aachen). BPD patients and CC were recruited from mental health clinics at local sites. HC were recruited via public advertisement. The study took place at the respective mental health clinic or Maastricht University. Participants provided written informed consent before participation. Depending on the presence or absence of Axis I diagnoses, participants attended three to four sessions9, which took roughly one to two and a half hours each. During the first session, participants received oral and written explanations about the study, signed an informed consent sheet, and were diagnosed with the SCID-I (First et al., 1997) by trained interviewers. During the second session, the SCID-II (First et al., 1997) was taken, and participants filled out the BPD Checklist (Arntz & Dreessen, 1995) as well as the Self-Injury Questionnaire (Arntz, 2004). Drug screenings and pregnancy tests were taken before administering the pharmacological treatment, thus prior to the third and fourth session. Participants either received naltrexone or placebo capsules if drug screenings and pregnancy tests were negative. Subsequently, participants filled out the Dissociation Questionnaire (Arntz, 2004a). Then, participants watched the first neutral movie clip, filled out the Emotion Questionnaire (Arntz, 2004b) and participated in two pain tests, i.e., FBPS and CPT. This procedure was repeated for the stress-inducing movie clip and the second neutral movie clip. Since the study was based on an experimental crossover design, the procedure was repeated during the fourth session, except that participants received naltrexone if they had received placebo previously and vice versa (see Figure 1). At the end of the last session, ITECs were taken and participants were debriefed and received a small financial remuneration. Interviews and questionnaires are described in Appendix A and the detailed study procedure is listed in Appendix B.
9participants attended three sessions if the SCID-I was recently administered in the respective mental health clinic.
Figure 1. Schematic illustration of the experimental crossover design. Based on a double-blind randomization procedure, subjects were assigned to one of two possible sequences. In the first sequence, participants received naltrexone at the beginning of the first experimental session and placebo at the beginning of the second experimental session. Order of pharmacological treatments was reversed in the second sequence. During each session, short (15 minutes) and long (20 minutes) neutral movie clips as well as a stress-inducing movie clip (15 minutes) were used. The order of neutral movie clips was counterbalanced for experimental sessions and pharmacological conditions, resulting in 16 unique orders of three movie clips each. Pain threshold and pain tolerance were measured with the Forgione-Barber Pain Stimulator and the Cold Pressor Test after each movie clip; stress perception after each movie clip was measured with the Emotion Questionnaire.
Experimental Groups
BPD-NP (n = 16) BPD-P (n = 16)
CC (n = 19) HC (n = 16)
Sequence 1 Session 1: Naltrexone
Neutral ðStress ðNeutral
Session 2: Placebo Neutral ðStress ðNeutral
Sequence 2 Session 1: Placebo
Neutral ðStress ðNeutral
Session 2: Naltrexone Neutral ðStress ðNeutral Random
Statistical Analyses
Statistical analyses of the data were carried out with SPSS for Mac, Version 25. Since the main focus of this study was on abnormal pain perception in BPD patients, control participants who indicated that they engage in SIB and do not invariably feel pain when doing so (n = 5) were removed from all subsequent analyses. To increase the validity of the dependent variable10, combined pain tolerance scores were used as primary outcome measures. These scores were calculated by standardizing pain scores for the CPT and FBPS per measurement point and adding them, leading to six scores for threshold and tolerance per measurement point. To evaluate Hypotheses 1.1 to 2.2, multilevel models were built with individual subjects being on level 3, pharmacological condition (naltrexone vs. placebo) being on level 2, and movie type (neutral vs. stress-inducing) being on level 1. Fixed effects were used for groups, movies, and pharmacological conditions. Random intercepts were chosen for the subjects and a marginal unstructured pattern11 was selected for movies nested within pharmacological conditions within subjects. Further, movie clips were evaluated based on centred contrasts to compare the effects of neutral and stress-inducing movie clips. Uncentred contrasts were used to evaluate differences between groups based on a reference group12. Coding schemes for the contrasts are listed in Appendix D. The multilevel model to be estimated13 is
10while procedures for both pain tests are highly standardized, neither of them has been formally validated yet
(Edens & Gil, 1995). By combining two frequently used objective measures, we aimed at calculating a global estimation of pain perception, which was based on both thermal (FBPS) and pressure (CPT) pain. Such a global estimation seems particularly relevant since pain is multifaceted (Peters & Schmidt, 1990) and since such a measurement allows for evaluating naturalistic pain perception in a laboratory environment.
11chosen due to its satisfactory goodness-of-fit and improved parsimony compared to first-order autoregressive
(AR1) and first-order autoregressive moving average (ARMA11) structures.
12either BPD patients or BPD-NP patients were used as reference group.
13since neither of the three-way interaction terms reached significance, most analyses were based on multilevel
models using no more than two-way interaction terms to increase goodness of fit of the respective models. The following additional coefficients were used for the multilevel models including three-way interaction terms: β12
movie2vs1 × condition; β13 movie2vs1 × condition × group01vs2; β14 movie2vs1 × condition × group01vs3; β14
movie2vs1 × condition × group0vs1c. Multilevel models including three-way interaction terms are reported for Hypothesis 2.1 and Hypothesis 2.2.
painijk = β0 + β1 group0vs1 + β2 group0vs2 + β3 group0vs3 + β4 condition + β5 movie2vs1 + β6 condition × group0vs1 + β7 condition × group0vs2 + β8 condition × group0vs3 + β9 movie2vs1 × group01vs2 + β10 movie2vs1 × group01vs3 + β11 movie2vs1 × group0vs1c + u01 + eijk
where i is the subject, j is the pharmacological condition, and k is the movie, u0i
represents the intercept of the subject i, and e represents the residual. Pain scores in seconds for the placebo and naltrexone conditions are listed in Appendix C. As parameter estimation method Restricted Maximum Likelihood (REML) was chosen to minimize biased standard errors for the fixed effects, using 100 maximum iterations and 10 maximum step-halvings. Corresponding goodness of fit indices are presented in Appendix E. To estimate the power of the contrasts, we carried out a post-hoc power analysis for a repeated measured ANOVA design14 including within-between interaction term using the program G*Power (Erdfelder, Faul, & Buchner, 1996). Assuming a medium to large effect size (f = 0.25)15 and α = .05, two-tailed, for a sample size of n = 32 (e.g., BPD-NP vs. BPD-P) for four measurements (i.e. two types of movie clips, two pharmacological treatments) and r = 0.5 among repeated measures, we calculated that power (1 - β) was 0.78. Graphs for the pain-related comparisons are shown in Figures 2 – 5 and in Appendix F.
Hypotheses 3.1 and 3.2 were evaluated with multiple regression analyses. Difference scores between neutral and stress-inducing movie clips for threshold and tolerance were predicted from general dissociative symptoms and childhood trauma or somatoform dissociative symptoms and childhood trauma in the placebo condition for BPD patients. Owing
14as approaches to estimate power for three-level multilevel models are sparse (Zhang, Zyphur, & Preacher, 2009),
this analytic technique was chosen as an estimation.
15this assumption was based on previous comparable research (Bohus et al., 2000; Russ et al., 1992). Bohus and
colleagues (2000) reported a large effect (Cohen’s d = 1.47) for the main effect of pain intensity between BPD patients and nonclinical controls. Also, Russ and colleagues (1992) reported a large effect size (Cohen’s d = 1.34) for the main effect of group (BPD-NP vs. BPD-P) for pain intensity.
to statistical power16, ITEC total scores were used for the regression model and BPD-NP and BPD-P patients were analyzed as one group. Regression coefficients and standard errors for all outcome measures are shown in Tables 4-6. Prior to conducting multiple regressions, all relevant assumptions were tested17.
Lastly, three-way mixed ANOVAs were run to examine the effect of pharmacological condition, movie type18, and group status on emotions, i.e., happiness, sadness, anxiety, restlessness, surprise, valence, arousal, and dominance. Greenhouse-Geisser corrections were applied for all comparisons. For pairwise comparisons, Bonferroni adjustments were applied and adjusted p-values were reported. Before conducting the mixed ANOVAs, all relevant assumptions were tested19. Tables displaying the average mood ratings for the four experimental groups during the placebo and naltrexone condition are shown in Appendix G. Comparisons of the mood ratings for the four experimental groups and both pharmacological conditions are presented in Figures 14 – 22. Data are mean ± SEM unless otherwise stated.
16 Based on the large effect of dissociation on pain threshold reported by Ludäscher and colleagues (2007), a priori
power analysis using G*Power (Erdfelder et al., 1996) indicated that a sample size of n = 31 was required to detect a similar effect with power (1 - β) set at .80 and α = .05, two-tailed.
17in particular, the pain-related dependent variable and the trauma- and dissociation-related independent variables
were measured at the continuous level. Independence of observations were checked using a reference value of 2 for the Durbin-Watson statistic. The assumption of linearity was checked by visually inspecting scatterplots of the studentized residuals against the unstandardized predicted values and partial regression plots between each independent variable and the dependent variable. Homoscedasticity was checked by visually inspecting scatterplots of the studentized residuals against the unstandardized predicted values. Multicollinearity was precluded since all VIF values < 10. As one subject had standardized residuals greater than ± 3 standard deviations, this subject was treated as an outlier and removed from the multiple regression analyses. Absence of leverage points was examined by accepting leverage values ≈ 0.2; absence of highly influential points was examined by accepting Cook’s Distance values < 1. Histograms for regression standardized residuals and normal P-P plots of the regression standardized residuals were used to ascertain that residuals were approximately normally distributed.
18all analyses refer to the comparison between the three movie clips, i.e., the two neutral movie clips and the
stress-inducing movie clip.
19in detail, the dependent emotion measures were continuous, the within-subject factors (pharmacological
condition; movie type) were categorical with two levels each, and the between-subject factor was categorical with four groups. Potential outliers were examined by inspection of boxplots. As removing outliers did not distinctly change the results, these emotion measures were not removed from the analyses. Shapiro-Wilk’s test indicated that some emotion measures were not normally distributed, which is why a logarithmic transformation was applied to highly positively skewed emotion measures. Since results for the three-way mixed ANOVA without transformed emotion measures and with partly logarithmic transformed emotion measures yielded similar findings, the untransformed emotion measures are reported to facilitate interpretation. The assumption of homogeneity of variances was checked with Levene’s test based on median emotion scores. Nonsignificant results (p > .05) indicated that the variance of residuals between groups was equal. The assumption of sphericity was examined based on Mauchly’s test. As this test indicated that the assumption of sphericity was violated for eight of nine three-way ANOVAs, a Greenhouse Geisser correction was applied.
Results Manipulation Check
Globally, the manipulation check indicated that emotions ratings became more negative after the stress-inducing movie clips and also that the stress-inducing movie clips were experienced as more negative than the two neutral movie clips. In particular, the stress-inducing movie clips were associated with increased ratings for sadness, anxiety, anger, restlessness, surprise, and arousal, as well as decreased ratings for valence and happiness. Additionally, the manipulation check demonstrated differences between experimental groups. Happiness ratings were lower for BPD-P patients than for HC. Restlessness ratings were higher for BPD-NP patients than for HC. Further, BPD-P and BPD-NP reported lower valence towards the movie clips than HC. Lastly, anxiety ratings were higher for BPD-NP patients compared to HC after stress-inducing movie clips and BPD-P patients reported less dominance than HC after the neutral movie clips. Detailed descriptions of all three-way interactions and follow-up analyses as appropriate are presented in Appendix H.
Stress-Induced Opioid-Mediated Analgesia
Stress-induced analgesia in BPD patients. When comparing pain responses in BPD
patients after neutral and stress-inducing movie clips, thresholds were significantly increased after the stress-inducing movie clips for the FBPS (b = 4.64, p = .006) and for the CPT (b = 5.85, p = .005) but not for the combined measure (b = 0.03, p = .488). Further, tolerance in BPD patients significantly increased after stress-inducing movie clips for the FBPS (b = 5.23, p = .004), while tolerance as measured with the CPT (b = 2.71, p = .153) and the combined measure (b = 0.03, p = .237) were not significantly different when comparing stress-inducing movie clips to neutral movie clips. However, the nonsignificant interaction terms between neutral and stress-inducing movie clips and experimental groups indicated that the aforementioned increases in pain responses for BPD patients were not significantly different from pain responses of the control groups when examining threshold (BPD vs. CC: bcombined = -0.09, p =
.217; bFBPS = -3.93, p = .198; bCPT = -4.45, p = .236; BPD vs. HC: bcombined = -0.03, p = .636;
bFBPS = 4.38, p = .134; bCPT = -3.28, p = .360) and tolerance (BPD vs. CC: bcombined = -0.04, p
= .268; bFBPS = -3.62, p = .260; bCPT = -1.58, p = .645; BPD vs. HC: bcombined = -0.06, p = .132;
bFBPS = -5.91, p = .055; bCPT = -1.33, p = .683).
Stress-induced analgesia in BPD-NP patients. The interaction between neutral and
stress-inducing movie clips and BPD-NP and BPD-P patients for the placebo condition indicated a significantly lower pain tolerance of BPD-P patients for the combined measure (b = -0.11, p = .013). Although pain tolerance was also lower in P patients compared to BPD-NP patients for the FBPS (b = -6.93, p = .052) and the CPT (b = -4.82, p = .204) in the placebo condition, these comparisons failed to reach statistical significance. None of the threshold measures demonstrated a significant difference between BPD-NP and BPD-P patients (bcombined = -0.05, p = .498; bFBPS = 0.63, p = .851; bCPT = -2.91, p = .482).
Stress-induced opioid-mediated analgesia in BPD patients. Neither of the interaction
terms between movie clips and pharmacological conditions for BPD patients was significant for the threshold measures (bcombined = 0.03, p = .733; bFBPS = -0.05, p = .989; bCPT = 1.85, p =
.658) or for the tolerance measures (bcombined = 0.01, p = .889; bFBPS = -1.72, p = .628; bCPT =
-0.52, p = .890). Further, neither of the comparisons indicated stress-induced opioid-mediated analgesia in BPD patients compared to control participants since all three-way interactions between movie clips, conditions, and experimental groups20 failed to reach statistical significance. As shown in Table 3 and Table 5, neither threshold (BPD vs. CC: bcombined = -0.07, p = .657; bFBPS = -1.44, p = .815; bCPT = -2.56, p = .736; BPD vs. HC: bcombined = -0.06, p =
.681; bFBPS = -5.95, p = .310; bCPT = -3.61, p = .619) nor tolerance measures (BPD vs. CC:
bcombined = 0.01, p = .867; bFBPS = -3.54, p = .583; bCPT = 2.85, p = .678; BPD vs. HC: bcombined
= -0.03, p = .728; bFBPS = -4.63, p = .479; bCPT = 4.35, p = .507) did significantly differ.
Stress-induced opioid-mediated analgesia in BPD-NP patients. Also, when
comparing BPD-P to BPD-NP patients based on the three-way interaction between movie clip, condition, and group status, neither threshold (bcombined = 0.22, p = .168; bFBPS = 6.50, p = .336; bCPT = 6.62, p = .430) nor tolerance (bcombined = 0.13, p = .158; bFBPS = 0.11, p = .988; bCPT =
13.43, p = .078) were significantly different. However, when solely examining the difference in predicted pain in the naltrexone condition compared to the placebo condition for BPD-NP patients averaged across neutral and stress-inducing movie clips (see Table 4), pain tolerance significantly increased for BPD-NP patients for the combined measure (b = 0.27, p = 0.038) and for the CPT (b = 46.21, p = .001) in the naltrexone condition. Further, interaction terms between condition and group status21 averaged between movie clips revealed that threshold and tolerance measures decreased in HC for the naltrexone condition compared to the placebo condition when contrasted to BPD-NP patients (threshold: bcombined = -0.34, p = .034, bCPT =
-21.18, p = .040; tolerance: bcombined = -0.41, p = .024, bCPT = -38.38, p = .038). Also, compared
to BPD-NP, tolerance as measured with the combined measure and the CPT significantly decreased for BPD-P patients in the naltrexone condition compared to the placebo condition averaged across movie clips (bcombined = -0.39, p = .031, bCPT = -44.37, p = .017).
Figure 2. Differences in pain threshold for four experimental groups measured with the combined pain measure in the placebo condition. Data represent mean ± SEM.
Figure 3. Differences in pain threshold for four experimental groups measured with the combined pain measure in the naltrexone condition. Data represent mean ± SEM.
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
Neutral (1) Stress-inducing (2) Neutral (3) Combined Threshold (Placebo)
BPD-NP BPD-P CC HC -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
Neutral (1) Stress-inducing (2) Neutral (3) Combined Threshold (Naltrexone)
Figure 4. Differences in pain threshold for four experimental groups measured with the combined pain measure in the placebo condition. Data represent mean ± SEM.
Figure 5. Differences in pain threshold for four experimental groups measured with the combined pain measure in the naltrexone condition. Data represent mean ± SEM.
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
Neutral (1) Stress-inducing (2) Neutral (3) Combined Tolerance (Placebo)
BPD-NP BPD-P CC HC -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0
Neutral (1) Stress-inducing (2) Neutral (3) Combined Tolerance (Naltrexone)
Table 2
Multilevel Model for Pain Threshold Excluding Three-Way Interactions
Combined Pain Measure Forgione-Barber Pain Stimulator Cold Pressor Test
b SEb p b SEb p b SEb p Intercept -0.21 0.24 .389 28.54 7.49 .000 22.64 14.87 .133 group0vs1 0.28 0.34 .415 0.08 10.59 .994 25.00 21.03 .239 group0vs2 -0.10 0.36 .783 -11.73 10.96 .288 3.55 21.77 .871 group0vs3 0.62 0.34 .075 15.35 10.59 .152 35.17 21.03 .100 movie2vs1 0.03 0.04 .488 4.64 1.67 .006 5.85 2.06 .005 condition 0.15 0.11 .189 4.00 3.92 .312 9.20 7.10 .201 group0vs1 × condition -0.21 0.16 .191 -6.11 5.53 .274 -14.87 10.00 .143 group0vs2 × condition 0.04 0.16 .784 5.57 5.73 .335 -1.71 10.34 .869 group0vs3 × condition -0.34 0.16 .034 -5.78 5.53 .301 -21.18 10.04 .040 movie2vs1 × group01vs2 -0.09 0.07 .217 -3.93 3.03 .198 -4.45 3.74 .236 movie2vs1 × group01vs3 -0.03 0.07 .636 4.38 2.90 .134 -3.28 3.57 .360 movie2vs1 × group0vs1c -0.05 0.08 .498 0.63 3.35 .851 -2.91 4.12 .482
Table 3
Multilevel Model for Pain Threshold Including Three-Way Interactions
Combined Pain Measure Forgione-Barber Pain Stimulator Cold Pressor Test
b SEb p b SEb p b SEb p Intercept -0.23 0.25 .345 28.51 7.49 .000 22.33 14.93 .140 group0vs1 0.34 0.35 .328 0.13 10.59 .990 26.43 21.11 .216 group0vs2 -0.09 0.36 .811 -11.72 10.96 .289 3.71 21.85 .866 group0vs3 0.64 0.35 .070 15.33 10.59 .153 35.10 21.11 .102 movie2vs1 0.01 0.06 .806 4.65 2.38 .053 4.92 2.95 .099 condition 0.19 0.13 .133 4.04 3.92 .307 9.82 7.54 .197 group0vs1 × condition -0.33 0.18 .072 -6.20 5.53 .267 -17.71 10.62 .100 group0vs2 × condition 0.02 0.18 .915 5.55 5.73 .337 -2.03 10.99 .854 group0vs3 × condition -0.37 0.18 .043 -5.74 5.53 .304 -21.05 10.66 .052 movie2vs1 × condition 0.03 0.08 .733 -0.05 3.37 .989 1.85 4.18 .658 movie2vs1 × group01vs2 -0.06 0.10 .579 -3.22 4.32 .457 -3.16 5.35 .556 movie2vs1 × group01vs3 0.00 0.10 .967 7.34 4.13 .078 -1.46 5.11 .775
Combined Pain Measure Forgione-Barber Pain Stimulator Cold Pressor Test
b SEb p b SEb p b SEb p
movie2vs1 × group0vs1c -0.17 0.11 .148 -2.63 4.76 .582 -6.21 5.90 .295
movie2vs1 × condition × group01vs2 -0.07 0.15 .657 -1.44 6.11 .815 -2.56 7.57 .736
movie2vs1 × condition × group01vs3 -0.06 0.14 .681 -5.95 5.83 .310 -3.61 7.23 .619
movie2vs1 × condition × group0vs1c 0.22 0.16 .168 6.50 6.74 .336 6.62 8.35 .430
Note. The model excluding three-way interaction is reported for Hypotheses 1.1 and 1.2; the model including three-way interactions is reported for Hypotheses 2.1 and 2.2. Contrasts for movie clips were centred to compare the effects of different movie clips; contrasts for groups are uncentred to compare experimental groups to a reference groups. Respective coding schemes are shown in Appendix D.
Table 4
Multilevel Model for Pain Tolerance Excluding Three-Way Interactions
Combined Pain Measure Forgione-Barber Pain Stimulator Cold Pressor Test
b SEb p b SEb p b SEb p Intercept -0.32 0.24 .197 75.48 8.85 .000 70.48 22.81 .003 group0vs1 0.27 0.34 .441 -1.45 12.51 .908 31.64 32.26 .330 group0vs2 0.30 0.36 .401 -6.11 12.95 .639 41.17 33.39 .222 group0vs3 0.70 0.34 .046 26.27 12.51 .039 52.72 32.26 .107 movie2vs1 0.03 0.02 .237 5.23 1.76 .004 2.71 1.89 .153 condition 0.27 0.13 .038 5.69 5.37 .293 46.21 12.78 .001 group0vs1 × condition -0.39 0.18 .031 -1.30 7.58 .864 -44.37 18.07 .017 group0vs2 × condition -0.28 0.18 .135 2.46 7.84 .755 -35.32 18.70 .064 group0vs3 × condition -0.41 0.18 .024 -7.65 7.58 .317 -38.38 18.07 .038 movie2vs1 × group01vs2 -0.04 0.04 .268 -3.62 3.20 .260 -1.58 3.42 .645 movie2vs1 × group01vs3 -0.06 0.04 .132 -5.91 3.05 .055 -1.33 3.27 .683 movie2vs1 × group0vs1c -0.11 0.04 .013 -6.93 3.53 .052 -4.82 3.77 .204
Table 5
Multilevel Model for Pain Tolerance Including Three-Way Interactions
Combined Pain Measure Forgione-Barber Pain Stimulator Cold Pressor Test
b SEb p b SEb p b SEb p Intercept -0.32 0.24 .198 75.59 8.85 .000 70.90 22.81 .003 group0vs1 0.26 0.34 .446 -1.46 12.52 .908 30.85 32.26 .342 group0vs2 0.30 0.36 .404 -5.87 12.96 .652 40.61 33.39 .228 group0vs3 0.70 0.34 .046 26.56 12.52 .037 52.08 32.26 .111 movie2vs1 0.02 0.03 .467 6.05 2.51 .017 2.97 2.67 .268 condition 0.26 0.13 .040 5.46 5.41 .317 45.42 12.79 .001 group0vs1 × condition -0.39 0.18 .033 -1.29 7.63 .866 -42.85 18.09 .021 group0vs2 × condition -0.28 0.18 .140 2.00 7.90 .801 -34.25 18.72 .073 group0vs3 × condition -0.41 0.18 .025 -8.22 7.63 .286 -37.14 18.09 .045 movie2vs1 × condition 0.01 0.04 .889 -1.72 3.55 .628 -0.52 3.78 .890 movie2vs1 × group01vs2 -0.05 0.06 .374 -1.82 4.55 .690 -2.98 4.83 .539 movie2vs1 × group01vs3 -0.04 0.05 .416 -3.69 4.34 .397 -3.50 4.62 .450
Combined Pain Measure Forgione-Barber Pain Stimulator Cold Pressor Test
b SEb p b SEb p b SEb p
movie2vs1 × group0vs1c -0.18 0.06 .006 -6.99 5.02 .166 -11.56 5.33 .032
movie2vs1 × condition × group01vs2 0.01 0.08 .867 -3.54 6.43 .583 2.85 6.84 .678
movie2vs1 × condition × group01vs3 -0.03 0.08 .728 -4.36 6.14 .479 4.35 6.53 .507
movie2vs1 × condition × group0vs1c 0.13 0.09 .158 0.11 7.10 .988 13.43 7.55 .078
Note. The model excluding three-way interaction is reported for Hypotheses 1.1 and 1.2; the model including three-way interactions is reported for Hypotheses 2.1 and 2.2. Contrasts for movie clips were centred to compare the effects of different movie clips; contrasts for groups are uncentred to compare experimental groups to a reference groups. Respective coding schemes are shown in Appendix D.
Stress-Induced Analgesia, Childhood Trauma, and Dissociation
Multiple regression analyses indicated that neither childhood trauma and general dissociation (e.g., combined threshold: F(2, 28) = 1.174 , p = 324, adj. R2 = .01) nor childhood trauma and somatoform dissociation (e.g., combined threshold: F(2, 28) = 2.176, p = .132, adj. R2 = .07) predict pain sensitivity in BPD patients when controlling for basal pain sensitivity in the placebo condition. Neither childhood trauma (p ≥ .084) nor dissociative states (general dissociation: p ≥ .143; somatoform dissociation: p ≥ .063) added statistically significantly to the predictions. Regression analyses for the combined measure, the FBPS, and the CPT are illustrated in Tables 6 – 8.
Table 6
Multiple Regression Analyses for the Combined Pain Measure
Threshold Tolerance
Variable B SEB ß p B SEB ß p
General dissociation and childhood trauma
Constant -0.36 0.40 .376 0.00 0.32 1.000
General Dissociation 0.01 0.01 .244 .191 0.00 0.01 .03 .872
Childhood Trauma 0.00 0.00 -.152 .409 0.00 0.00 .01 .977
Somatoform dissociation and childhood trauma
Constant -0.48 0.35 .189 -0.04 0.29 .886
Somatoform Dissociation 0.08 0.04 .348 .063 0.01 0.03 .07 .728
Childhood Trauma -0.01 0.00 -.205 .263 0.00 0.00 -.01 .976
Note. N = 31. For dissociation and childhood trauma, pain threshold: adj. R2 = .06, F for change in R2 = 1.90, p = .168; tolerance: adj. R2 = -.07, F for change in R2 = 0.03, p = .973. For somatoform dissociation and childhood trauma, pain threshold: adj. R2 = .06, F for change in R2 = 2.08, p = .143; tolerance: adj. R2 = -.07, F for change in R2 = 0.02, p = .977. B = unstandardized regression coefficient; SEB = standard error of the coefficient; b = standardized coefficient.
Table 7
Multiple Regression Analyses for the Forgione-Barber Pain Stimulator
Threshold Tolerance
Variable B SEB ß p B SEB ß p
General dissociation and childhood trauma
Constant -3.28 15.77 .837 4.99 21.92 .822
General Dissociation 0.57 0.38 .27 .143 0.19 0.52 .07 .724
Childhood Trauma -0.28 0.18 -.28 .121 0.01 0.24 .01 .952
Somatoform dissociation and childhood trauma
Constant -4.13 14.08 .771 -1.79 19.70 .928
Somatoform Dissociation 2.97 1.64 .32 .081 1.85 2.30 .15 .427
Childhood Trauma -0.33 0.18 -.33 .075 -0.02 0.25 -.01 .940
Note. N = 31. For dissociation and childhood trauma, pain threshold: adj. R2 = .08, F for change in R2 = 2.26, p = .123; tolerance: adj. R2 = -0.07, F for change in R2 = 0.07, p = .935. For somatoform dissociation and childhood trauma, pain threshold: adj. R2 = .11, F for change in R2 = 2.79, p = .078; tolerance: adj. R2 = -.05, F for change in R2 = 0.33, p = .723. B = unstandardized regression coefficient; SEB = standard error of the coefficient; b = standardized coefficient.
Table 8
Multiple Regression Analyses for the Cold Pressor Test
Threshold Tolerance
Variable B SEB ß p B SEB ß p
General dissociation and childhood trauma
Constant 0.62 18.97 .974 5.39 25.71 .835
General Dissociation 0.53 0.45 .21 .247 0.05 0.61 .02 .932
Childhood Trauma -0.35 0.21 -.29 .114 0.06 0.29 .04 .834
Somatoform dissociation and childhood trauma
Constant 1.09 17.11 .950 7.39 23.32 .754
Somatoform Dissociation 2.63 2.00 .24 .199 0.00 2.72 .00 .999
Childhood Trauma -0.38 0.21 -.32 .084 0.06 0.29 .04 .832
Note. N = 31. For dissociation and childhood trauma, pain threshold: adj. R2 = .07, F for change in R2 = 2.08, p = .143; tolerance: adj. R2 = .02, F for change in R2 = 1.38, p = .267. For somatoform dissociation and childhood trauma, pain threshold: adj. R2 = .08, F for change in R2 = 2.34, p = .114; tolerance: adj. R2 = .07, F for change in R2 = 2.19, p = .130. B = unstandardized regression coefficient; SEB = standard error of the coefficient; b = standardized coefficient.
Discussion
The current study investigated whether pain perception in BPD patients is stress-induced and if blocking the EOS with an opioid antagonist influences pain perception in BPD patients. According to this, we compared pain perception in BPD-NP and BPD-P patients to clinical and healthy control groups after watching neutral and stress-inducing movie clips based on a randomized, placebo-controlled, crossover design and additionally examined how childhood trauma and dissociation modulate pain perception under stress in BPD patients.
First, we could show that pain sensitivity in BPD patients decreased after stress-inducing movie clips compared to neutral movie clips in the placebo condition. Thereby, we could confirm Hypothesis 1.1 for the threshold measures of the FBPS and CPT and also for the tolerance measure of the FBPS. Further, our results support Hypothesis 1.2 as we could show that stress-induced analgesia was particularly pronounced in BPD-NP patients. In particular, pain tolerance for the combined measure solely increased in NP patients but not in BPD-P patients from the neutral movie clip to the stress-inducing movie clip in the placebo condition. These results correspond to the findings by Schmahl and colleagues (2010) who demonstrated that stress-induced analgesia differentiates BPD from other stress-related disorders such as posttraumatic stress disorder or bulimia nervosa. However, we could further show that stress-induced analgesia was characteristic for BPD patients who report not to feel pain when engaging in SIB. Adding to Russ and colleagues (1992), who demonstrated that pain ratings for BPD-NP are generally lower than for BPD-P, our findings confirm that group differences are strongly marked after experimentally induced stress. Most notably, these findings imply that especially self-injurious BPD patients with aberrant pain perception may benefit from stress-regulation techniques such as suggested by Gratz and Gunderson (2006).
However, the current study does neither provide unequivocal support for stress-induced opioid-mediated analgesia nor for an amplifying effect of dissociation and childhood trauma
regarding stress-induced analgesia in BPD. More precisely, no evidence for Hypothesis 2.1 was found as pain sensitivity in BPD patients did not increase from the placebo condition to the naltrexone condition when comparing stress-inducing movie clips to neutral movie clips. Likewise, pain ratings of BPD patients did not clearly differ from those of healthy and clinical controls when examining differences between pharmacological conditions and movie clips. Also, contrasting Hypothesis 2.2, pain sensitivity between BPD-NP and BPD-P patients could not be distinguished when examining interactions between pharmacological conditions and movie clips for BPD-NP and BPD-P patients separately. Even though we expected stress-induced opioid-mediated analgesia primarily in the BPD-NP group, our negative finding is consistent with research by Russ and colleagues (1994), who demonstrated that naloxone did not increase pain ratings as measured with the CPT in BPD-NP or BPD-P patients. However, as several confirmatory findings on stress-induced opioid-mediated analgesia in BPD have been reported (Bohus et al., 1999; Meiser, Dupper, Wedekind, & Bandelow, 2015; Sonne, Rubey, Brady, Malcolm, & Morris, 1996), the question remains if uncontrollability of the stressor was sufficient for endogenous opioid release or if length of treatment period was insufficient to elicit opioid-mediated analgesia. On the one hand, our finding that naltrexone increased pain tolerance for the combined measure and the CPT in BPD-NP patients corresponds to the finding by Janssen and Arntz (1997) that certain opioid antagonists may also have agonist properties in the absence of opioid activity. On the other hand, a single dose of opioid antagonists may have been insufficient to modulate EOS functioning regarding pain perception since confirmatory findings have only been shown for long-term naltrexone treatment. Since animal research indicated that short durations of stress cause nonopioid analgesia while longer durations induce opioid mechanisms (Akil et al., 1984; Lewis, Cannon, & Liebeskind, 1980), differences between short-term and long-term naltrexone treatment should be investigated in greater detail to determine if opioid antagonists can be considered an effective adjunct treatment for
self-injurious BPD patients. Moreover, it should be mentioned that prior positive findings were based on self-report while we applied behavioural tasks to measure pain perception. Since it is currently unclear how well such self-reports correspond to experimentally induced pain (Hooley, Ho, Slater, & Lockshin, 2010), it might be advisable to examine EOS functioning following opioid antagonist administration in vivo to investigate how abnormal EOS functioning corresponds to pain and stress perception in BPD patients. A detailed examination of the EOS system in BPD is further warranted to scrutinize hypothesized EOS abnormalities (Bandelow et al., 2010; New & Stanley, 2010) and to trial pharmacological treatments accordingly. However, based on the current findings, we cannot recommend introducing opioid antagonists such as naltrexone as an adjunct treatment for self-injurious BPD patients with aberrant pain perception in clinical practice.
Lastly, no evidence for Hypotheses 3.1 and 3.2 was found as pain sensitivity after stress-inducing stimuli did not differ due to frequency of childhood trauma, general dissociative symptoms, or somatoform dissociative symptoms. Thereby, our results counter research by Ludäscher and colleagues (2007), who demonstrated a strong correlation between dissociation and pain thresholds in BPD patients. However, since stress-induced analgesia was not manipulated in this previous study, it remains unclear if passively watching movies sufficiently induces stress to evoke a conditioned endogenous opioid response. Taking further into account that SIB may also have an antidissociative function (Bohus et al., 1999; Saxe, Chawla, & van der Kolk, 2002) and that dissociative symptoms significantly decreased in BPD patients after naloxone administration (Bohus et al., 1999; Philipsen et al., 2004), future studies should directly compare dissociative states during stress in BPD patients receiving opioid antagonist treatment. Further, as positive correlations between specific types of childhood trauma and pain disorders have been reported for BPD patients (Sansone, Pole, Dakroub, & Butler, 2006; Yates, 2004), replication of the current findings is warranted to re-examine the conditioned
endogenous opioid response hypothesis empirically. Such a replication seems particularly relevant as the current sample was too small to carry out respective and sufficiently powered analyses using different types of childhood trauma as predictors.
Limitations
Compared to earlier studies, the main strengths of the current study are its pain-related differentiation of BPD patients, the inclusion of carefully selected clinical and healthy control groups, as well its randomized, placebo-controlled, crossover design. However, some limitations also need to be acknowledged. Notably, while the experimental pain procedures were selected to increase comparability with previous studies, an extensive validation of objective pain tests is currently lacking. Acknowledging that pain is a highly subjective experience (Giordano, Abramson, & Boswell, 2010; Hooley et al., 2010), we included two experimental pain tasks and calculated combined scores to integrate pressure and thermal pain within one measure. However, most BPD patients use SIB in the form of skin-cutting (Reitz et al., 2015) and currently, we can only speculate that pressure and thermal pain underlie a comparable mechanism as does tissue damage in self-injurious BPD patients. If the mechanisms underlying pain perception differ, then stress-induced opioid-mediated analgesia needs to be investigated with an ecologically valid pain assessment. Correspondingly, future studies may benefit from examining the severity of self-injury when differentiating BPD patients based on their pain perception.
Moreover, although comprehensive in- and exclusion criteria and the standardized procedure increased the internal validity of the current study, external validity was limited by the female study sample and by excluding self-injurious clinical controls who suffered from analgesia. In particular, research showed that females exhibit less stress-induced analgesia (Wiesenfeld-Hallin, 2005) and that opioid receptors are activated differently in males and females (Zubieta et al., 2002). Similarly, we cannot exclude the possibility that menstrual cycle
