Does Interpersonal Loss Preceding Panic Disorder Onset Moderate Response to Psychotherapy? An Exploratory Study

Abstract

Background

Little research has addressed moderators of treatment outcome for anxiety disorders, and none has considered interpersonal loss as a predictor of outcome.

Purpose

To examine the effect of interpersonal loss events within the six weeks preceding panic disorder (PD) onset as a moderator of outcome in a randomized controlled trial of Panic-Focused Psychodynamic Therapy (PFPP) and Applied Relaxation Therapy (ART). Researchers hypothesized that such loss events would predict better outcome in PFPP but would not affect ART outcome.

Method

Forty-nine subjects with PD were randomly assigned to a 12-week course of PFPP or ART. Independent raters blinded to treatment condition and study hypotheses rated subjects on the Panic Disorder Severity Scale (PDSS) and Sheehan Disability Scale. Exploratory analyses assessed between-group effect size for PFPP and ART following standard moderator analytic procedures.

Results

Three-quarters of subjects reported a narrowly defined interpersonal loss (LOSS) in the 6 weeks preceding PD onset. Subjects had a mean duration of PD of 8.2 (9.5) years. PFPP was more efficacious than ART, but LOSS did not moderate PFPP outcome. An unexpected finding was that LOSS moderated ART outcome: subjects without LOSS showed no response to ART (PDSS change= 0.00 [2.90]), whereas LOSS had a pre/post effect size of 4.29 (5.60). Neither examination of various potential confounding variables nor sensitivity analyses of assumptions regarding attrition altered these findings.

Conclusions

Interpersonal loss events preceding onset of panic disorder were common. These losses moderated outcome in ART, a therapy that does not focus on such losses. Implications and need for future research are discussed.

Introduction

Despite the DSM-IV definition that patients perceive panic attacks in panic disorder (PD) as arising uncued, “out of the blue,” the DSM-IV notes that PD often arises in the context of interpersonal loss, when interpersonal bonds are disrupted (p. 389). About one half of the few research studies to have appraised this (using various methods) have found interpersonal loss events to be associated with onset of panic disorder in a substantial proportion of PD patients, ranging from 25% (1) to 67% (2) within the year preceding panic disorder onset. Interpersonal loss in childhood (severe disruption of the relationship with a significant other, as through divorce or death) has been widely recognized as a risk factor for the subsequent development of PD in adulthood (3, 4).

Among members of a self-help group for PD patients (N=55), in the year before PD onset, 38.2% experienced the death of a significant other and an additional 29.1% reported other interpersonal losses (2). In a naturalistic treatment sample (N=223), Manfro et al. (1) found that 25% reported an interpersonal loss in the year before PD onset. In a sample of (N=100) treatment-seeking patients with PD with agoraphobia, 62% reported interpersonal loss in the year before PD onset (5). Faravelli and Palenti (6) compared pre-onset life events reported in the year before the interview by patients with recent-onset PD (N=64), compared with events reported by a healthy convenience sample (N=78) matched for sex, age, marital status, education, and social class. Of the panic patients, 64.1% reported a stressful life event (not necessarily an interpersonal loss), in contrast to 34.6% of the comparison group. In a sample derived from one completed and one ongoing study, our group found that 47% of 51 PD patients participating in our psychotherapy-only treatment studies reported an interpersonal loss event in the six weeks preceding PD onset (7). The researchers reporting these findings have not all defined “interpersonal loss” identically, further obscuring its frequency.

Do patients who report developing PD in the context of interpersonal loss events differ from those who do not? Of greater clinical relevance, does the context in which DSM-IV panic disorder emerges affect treatment outcome? To date, no outcome study of PD treatment has evaluated the effect of interpersonal loss as a precipitant to PD onset on subsequent treatment response. The present exploratory study examined whether interpersonal loss proximal to PD onset moderates current treatment response in the context of a randomized controlled trial (8) of Panic Focused Psychodynamic Psychotherapy (PFPP) and Applied Relaxation Training (ART).

We hypothesized that interpersonal loss events predating panic disorder onset would moderate PFPP such that subjects who experienced interpersonal losses antecedent to developing PD would derive added benefit of PFPP relative to ART.

Method

Subjects

Forty-nine subjects ages 18 to 55 with primary DSM-IV panic disorder, diagnosed on the Anxiety Disorders Interview Schedule for DSM-IV Lifetime Version (ADIS-IV-L, 9) and enrolled in a randomized controlled trial of PFPP and ART (8), reported events preceding the onset of their panic disorder, if they recalled any, as part of the comprehensive ADIS baseline assessment. Two raters (SJK, BLM) independently assessed and coded subjects’ responses for the presence of interpersonal loss events within the 6 weeks preceding initial panic onset, which often predated study entry by years. “Interpersonal loss events” were narrowly defined as: death of a close love object, divorce or rupture of a love relationship, miscarriage, or abortion. All subjects signed informed, written consent; the Weill Medical College IRB approved the protocol.

Subjects were randomly assigned to PFPP or ART for 24 twice-weekly sessions over 12 weeks (for details, see [8]). Randomization was stratified by comorbid major depression and whether or not subjects were taking ongoing, stable (≥2 months prior to initial assessment) psychotropic medication. If subjects took stable psychotropics, the dose was held constant throughout the course of the therapy. The primary dependent variable was the Panic Disorder Severity Scale (PDSS, 10), a standard measure in PD outcome studies, chosen because it is a diagnosis-based, composite, global rating of panic severity; the only specific measure of its kind. Independent, trained evaluators, blinded to patient treatment group and therapist orientation, performed all outcome assessments. The moderating effect of interpersonal loss events within the 6 weeks preceding PD onset was also assessed on the Sheehan Disability Scale (SDS, 11), a self-rated measure of psychosocial function.

Training of Independent Evaluators (IEs)

IEs were trained to criterion on the ADIS-IV-L by a certified rater (MS) evaluating patients presenting to an open trial of PFPP (12). IEs were Master’s level diagnosticians with ≥35 hours of training on the ADIS-IV-L and ≥12 hours of training on the PDSS. A certified trainer (BLM) trained raters to criterion on the PDSS. ADIS training requires the new diagnostician to co-rate six ADIS protocols with the experienced rater and to match diagnostic categories and severities in these six ADIS protocols, three administered by the experienced trainer, three by the new rater. ADIS raters co-rated 2 patients every six months to minimize drift. To evaluate rater drift and monitor interrater reliability, the ADIS-IV-L raters co-rated two subjects every 9 months. Interrater reliability of each assessment measure was examined using two independent raters (one of whom conducted the interview) for each of five subjects (ADIS-IV-L kappa=0.91; Panic Disorder Severity Scale, kappa=0.89).

Therapeutic Interventions

Panic Focused Psychodynamic Psychotherapy (PFPP) is a 12 week, 24 session (twice weekly) manualized psychoanalytic psychotherapy that preserves a psychoanalytically-oriented, transference-focused framework while consistently attending to physical symptoms of PD and agoraphobia (13). Focus on unconscious emotional significance of panic is central to the treatment, which emphasizes identifying and decoding underlying psychological meanings of physical symptoms. The emotional significance of panic triggers is explored and interpreted. PFPP consistently focuses on helping patients to understand their internal emotional states. Because of its patient-directed, therapist-assisted format, and its focus on underlying emotional meaning of anxiety symptoms, PFPP appears suited to help patients address the emotional impact engendered by interpersonal losses. PFPP focuses on psychological significance of events preceding panic. Thus interpersonal loss events that patients experienced as emotionally connected with panic onset would likely become central areas of therapy exploration. PFPP involves no exposure.

Applied Relaxation Training (ART) was a 12-week, 24-session intervention, delivered twice weekly. This study adapted the Relaxation Treatment Manual of Cerny and colleagues (14). Progressive muscle relaxation training involves focusing attention on particular muscle groups, tensing the muscle group for 5-10 seconds, attending to the sensations of tension, relaxing the muscle group, attending to differences between sensations of tension and relaxation, and therapist suggestions of deepening relaxation. The number of muscle groups is gradually reduced from 16 to 8 to 4. Discrimination training, generalization, relaxation by recall, and cue-controlled relaxation (pairing the relaxed state to the word “relax”) follow.

Home practice is required twice daily. By week eight, subjects apply relaxation skills to anxiety-provoking situations (in vivo exposure) in a graduated fashion. Trained to identify early stages of anxiety, subjects are instructed to use relaxation as an active coping strategy whenever they become aware of tension, and to practice relaxation regularly throughout the day in various situations to maximize generalization. ART uses no interoceptive exposure, analysis of situational cues for panic, or cognitive restructuring. It is not designed specifically to address interpersonal loss events associated with panic disorder, but uses relaxation skills to cope with incipient and full blown panic attacks. Relaxation skills are viewed as providing a method to unlearn panic-related associations.

Therapists

PFPP Therapists (N=8)

All PFPP therapists were psychiatrists post residency or Ph.D. psychologists. All had completed at least three years’ psychoanalytic training at a psychoanalytic training institute. Their mean experience was 21 years (sd= 8.6; range 2-40 years). All had specific PFPP training, encompassing a 12 hour course, as well as at least 2 years’ clinical experience with PD using psychodynamic psychotherapy.

ART Therapists (N=6)

ART therapists were psychiatrists post residency or Ph.D. psychologists with 16 mean years of experience (sd= 11.3, range 5-35) (Mann Whitney p=0.66 between therapist groups). Specific training in ART entailed a 6 hour course and a minimum of 2 years’ clinical experience working with panic patients with ART and CBT. All ART therapists had extensive CBT experience with PD and used some form of relaxation training in their routine practice; two ART therapists used ART routinely in practice.

Adherence to treatments was monitored by trained raters in each condition, who rated 3 videotaped sessions per patient/therapist dyad. Both therapist groups were adherent to their administered treatment: PFPP therapists achieved mean adherence ratings of 5.5/6 on the PFPP Therapist Adherence Scale. The ART Adherence Scale (15) contains 3 items per session. ART therapists achieved average adherence ratings of 6.2 out of 7 (N=12 tapes for each therapist).

Data Analytic Procedures

The hypothesized moderator is the presence/absence at baseline of the report of an interpersonal loss event, as defined above, within the 6 weeks preceding initial panic disorder onset. The analysis was guided by the general strategy for exploratory moderator analyses in randomized controlled trials (RCTs) described by Kraemer et al. (16), whose criteria for treatment moderators require that:

1. the potential moderator precede treatment;

2. because of randomization, the potential moderator be uncorrelated with form of treatment;

3. a moderator of treatment “…must be shown to have an interactive effect with treatment on the outcome” (16, p.879). That is, the treatment effect must be shown to vary across levels of the moderator.

Kraemer et al. state: “P-values are not and should not be used to define moderators and mediators of treatment because then moderator or mediator status would change with sample size” (16, p.881). Thus our analyses focused on magnitude of the effect on the primary RCT endpoint measure, the PDSS (10), not on significance testing. These exploratory analyses examined differential effects of treatment by comparing subjects with and without antecedent interpersonal loss events on the between-group (PFPP vs. ART) effect size. Cohen’s d, for PDSS change (baseline-post treatment) was estimated separately for subjects with and without interpersonal loss events within the six weeks preceding panic onset.

The intention-to-treat (ITT) principle was employed in data analyses in accordance with the study protocol by carrying forward the last observation (LOCF), which was the baseline assessment for study dropouts if they refused assessment at dropout. Supplemental analyses examined the sensitivity of this strategy to attrition by only including those who were willing to provide follow-up data (N=42). Alternative strategies for analysis of repeated assessments over the course of the trial using mixed-effects linear regression models, for example, were not possible because the study design only included baseline and endpoints assessments.

Results

Baseline Demographics

Subjects were a mean 33 (9.1) years old. Seventy-one percent were Caucasian, 27% African American, 2% Asian; 18% were Hispanic. Twenty-six subjects were randomly assigned to PFPP and 23 to ART. The demographic and clinical characteristics of patients assigned to the 2 treatments did not significantly differ, except that the ART sample had significantly more men (47%) than the PFPP sample (15%) (two-tailed Fisher’s exact p = .03). Linear regression analysis found no association between gender and the primary outcome measure at baseline, the PDSS, (F = 5.16, df=46,p=.89), nor was there a treatment by gender interaction (F = 0.30, df=45, p=0.58).

Interpersonal Loss Events within 6 Weeks of PD Onset

Overall, 36 (73%) of subjects reported a proximal interpersonal loss event (LOSS) and 13 (27%) reported no loss (NL). Of the patients receiving PFPP, 19 (73%) reported LOSS, and 7 (27%) did not; in ART, 17 (74%) reported LOSS and 6 (26%) did not. The proportion of patients in the two treatments with and without LOSS did not differ (two-tailed Fisher’s exact = 0, df = 1, p = 1.00). Of those experiencing LOSS (N=36), 15 (42%) experienced a Break-up/ divorce/ separation/ affair, 11 (31%) were mourning the death of a close relationship, 9 (25%) had just moved away from their family/parents, 1 (3%) had an abortion.

Potential Confounds

Before conducting moderator analyses, we compared LOSS and NL groups on potentially confounding variables: severity of panic disorder on the PDSS, (LOSS=6.97 (5.56); NL=6 (6.2); two-tailed Mann-Whitney U = 202; p =0.431), presence of comorbid depression, (LOSS=25%; NL=15%; Fisher exact= 0.734), other Axis I (LOSS= 92%; NL=92%; Fisher’s exact=1.00), Axis II comorbidity (LOSS= 47%; NL=62%; Fisher exact=0.52), and duration of PD (LOSS= 8.2 [9.5] years; NL= 9.3 [10.2] years) (Mann Whitney U=268.5; p=0.539). LOSS and NL groups did not differ on these variables.

Evaluation of the Moderator Effect

We evaluated differential treatment response to PFPP and ART for subjects whose PD appeared precipitated by LOSS in comparison with those reporting NL. Treatment outcome was assessed using the PDSS, with additional analyses using the SDS to assess panic-related psychosocial function. Clinically meaningful, moderate to large between-group effects favoring PFPP were observed in subjects with PD precipitated by LOSS (Cohen’s d =.70) (Table 2), and this was magnified in subjects not reporting proximal interpersonal loss (Cohen’s d=1.32).

Table 2.

Moderator Effects of Loss on Panic-Focused Psychodynamic Psychotherapy (PFPP) and Applied Relaxation Therapy (ART) for Panic Disorder: Dropouts Without Termination Data Eliminated From Analysis

Loss statusa	N	PDSS Change Pre-Post, Mean (SD)	Between- Group Effect Size	Nb	SDS Change Pre-Post, Mean (SD)	Between-Group Effect Size
No loss
PFPP	6	9.00 (4.60)	1.45	6	9.75(5.98)	1.49
ART	3	0.00 (4.58)		3	−2.23(5.25)
Loss
PFPP	19	8.26 (5.16)	0.55	18	7.33 (7.65)	0.39
ART	14	5.21 (5.78)		13	4.42 (7.16)

^aLoss status: No-loss = No interpersonal loss reported in the 6 weeks prior to panic disorder onset, Loss = interpersonal loss reported loss in the 6 weeks prior to panic disorder onset.

^bNs differ due to missing data from dropouts.

Abbreviations: PDSS = Panic Disorder Severity Scale, SDS = Sheehan Disability Scale

Outcome in the domain of psychosocial functioning, as captured by the SDS, showed similar response patterns. The between-group difference favoring PFPP in subjects reporting LOSS lay in the moderate range (Cohen’s d = .46), while for those with NL, between-group differences were very large (Cohen’s d = 1.33.)

Sensitivity analyses compared the results from the above LOCF-imputed and the non-LOCF data sets. Ten subjects dropped out, of whom 3 provided assessments at dropout (Table 2). The apparent moderator effect of LOSS persisted and maintained the same pattern when analyses were limited to the 42 subjects who completed termination ratings. For the PDSS, among NL subjects Cohen’s d =1.45, substantially favoring PFPP over ART; whereas for subjects with LOSS Cohen’s d = .55. For the SDS, the comparison of treatments for NL subjects indicated very large between-group differences favoring PFPP, Cohen’s d = 1.49; in contrast, for LOSS subjects, Cohen’s d = .39.

Discussion

This study explored whether interpersonal loss as a precursor to onset of panic disorder moderated treatment response in the first randomized controlled study of psychoanalytic therapy for panic disorder (8). Our findings support the DSM-IV observation that interpersonal loss often provides the context for onset of panic disorder, as nearly three-quarters of the study patients reported developing PD in the 6 week period immediately following interpersonal loss events.

We had hypothesized that PFPP, which showed an overall advantage in outcome relative to ART (8), would have a particular advantage for subjects whose panic disorder was associated with interpersonal loss. It would be understandable that patients with LOSS should respond to PFPP, with its patient-directed, flexible frame and transference focus that permits articulation of problematic aspects of relationships, including those that have ended. Yet contrary to our hypothesis, outcome in PFPP did not vary with or without loss events prior to panic onset.

Strikingly, patients with panic associated with interpersonal loss events responded to ART, whereas those with NL did not. Only ART patients with LOSS responded. This was a surprising and unanticipated result.

Our preliminary findings suggest interpersonal loss may have a clinically meaningful role as an initial precipitant to panic disorder. While a large literature exists concerning putative mechanisms for observed therapeutic gains in psychotherapies, few such mechanisms have been adequately distinguished from epiphenomena (17). The mechanisms by which precipitating loss may affect treatment response in Applied Relaxation Training are unclear, yet may help to deepen our understanding of differential therapeutics and treatment responsiveness among panic patients. Why should ART, focused on skill development, practice, and in vivo exposure, help patients who have suffered interpersonal losses to recover from panic disorder, while those patients without such precipitants to panic were not helped?

Nothing in the ART protocol explicitly addresses interpersonal loss or any historical context for panic onset. As ART strictly focused on relaxation therapy, patients were much less likely than in PFPP to discuss panic antecedents. Therapists were directed to discuss panic attacks in terms of physiological symptoms and the use of relaxation for coping, with purposeful inattention to internal cognitive and emotional experiences during ongoing panic and to analysis of the situational antecedents and context of the panic. If patients raised non-panic issues, brief reflective listening was permitted, but was followed by directing the interaction back to the ART framework.

It seems unlikely that this ART effect is attributable to “common factor” (18) elements of this psychotherapy: i.e., empathic listening, the sense that a knowledgeable professional cares and understands, and the strength of the therapeutic alliance. ART adherence was uniformly high, with no differences between LOSS and NL ART treatments.

Alternatively, LOSS might be confounded by some other, as yet undetermined clinical variable. Our analyses ruled out some of the most obvious confounds.

The ADIS, albeit the gold standard in PD studies, did not yield sufficient historical detail to determine the presence of posttraumatic stress disorder (PTSD) or depression at the specific time of time of PD onset. Rates of lifetime comorbid PTSD in this sample were very low: only two subjects (4%) had PTSD on the ADIS, both related to distant childhood abuse and neither associated with PD onset. Without further information, it is not clear why these patients would have responded better to ART than those without interpersonal losses.

Viewed from a CBT perspective, some patients might have developed PD in response to the trauma of interpersonal loss. Given the reported importance of exposure interventions to treatment of PTSD (19,20), might the exposure component of ART have had particular benefit for patients whose PD began with an interpersonal trauma, i.e., a PTSD-like onset? Alternatively, from a psychodynamic perspective, rupture in an important attachment, as in the case of interpersonal loss, might make new attachments (as with a therapist) more important and meaningful, which in turn might permit patients to engage more fully in psychotherapy, including ART. By whatever mechanism, hypothesized or not, uncertainty about a putative mechanism does not erase the importance of the moderator finding.

Several limitations of this study should be noted. First, ascertainment of proximal interpersonal loss was coded from the answer to an interview question about circumstances in the 6 weeks prior to panic disorder onset. Responses may have been subject to retrospective and explanatory (i.e., post hoc, ergo propter hoc) bias. Even if the sequence of events occurred as patients reported 8.2 (9.5) years later, it would not prove that such losses “caused” panic disorder. It is possible that panic disorder vulnerability sensitizes individuals to the effects of interpersonal loss, that interpersonal loss can trigger an underlying panic diathesis, or that interpersonal loss makes patients more amenable to some psychotherapeutic interventions (i.e., ART) than they otherwise would be. In the absence of prospective longitudinal studies, which do not yet exist for panic disorder, researchers must perforce rely on patients’ retrospective reporting of stressors.

Thus, patients’ retrospective reports of interpersonal loss proximal to PD onset may either be construed as actual events, or as reflecting differing psychological organization of experience. We lack prospective and adequately fine-grained data to distinguish between these alternatives. The phenomenology of panic and the experience of interpersonal loss may at least partially overlap. The feeling of panic is often described as an overwhelming sense of helplessness and need for help, potentially mimicking feelings common in interpersonal loss situations for some people.

The treatment sample had some distinctive characteristics. The rate of reported proximal interpersonal loss was higher (74% in the prior six weeks) than reported in other panic disorder samples, in which rates ranged from 25% (1) to 67% (2) in the prior year. This was true even though we defined interpersonal loss more narrowly than did some other studies. It is unclear whether our subjects’ higher loss reporting reflects a clinically meaningful difference from other reported samples of panic disorder patients, as this variable has rarely been followed in the PD clinical trials literature. The overall response rate to ART (39%) was higher than reported elsewhere (9), potentially reflecting the ART therapists’ high degree of experience or the unusually intense, twice-weekly ART treatment which included an active in vivo exposure protocol, in addition to the relatively high rate of patients in this sample who reported interpersonal loss.

Interpersonal loss may be an important clinical marker in panic disorder, an anamnestic target for clinicians and a variable for future moderation research. Life events and traumas are a complex area of psychopathology (21): many patients report them, yet not all develop symptoms (22). It will be interesting to determine whether reports of proximal interpersonal loss moderate other, more commonly recommended treatments, such as cognitive behavior therapy (CBT; 23,24) or psychopharmacological interventions. CBT is more flexible than ART, addresses more levels of the panic attack experience, and uses a wider range of techniques.

This PD trial has now yielded two preliminary moderators of treatment: Axis II cluster C comorbidity predicted improved PFPP outcome (25), and LOSS predicts better ART outcome. To our knowledge, the current report is the first moderator analysis of any anxiety disorder treatment to assess a psychosocial variable. Loss may be important as an anamnestic target that may or may not reflect other variables that in turn influence treatment response. Clinicians may consider this association in treatment planning, in designing psychotherapeutic interventions, and in exploring mechanisms of panic disorder phenomenology and treatment response.

Table 1.

Moderator Effects of Interpersonal Loss Precipitants on Panic-Focused Psychodynamic Psychotherapy (PFPP) and Applied Relaxation Therapy (ART) for Panic Disorder: LOCF Used for Dropouts

Loss statusa	N	PDSS Change Pre-Post, Mean (SD)	Between- Group Effect Size	Nb	SDS Change Pre-Post, Mean (SD)	Between- Group Effect Size
No loss
PFPP	7	7.71 (5.41)	1.32	7	8.36(6.59)	1.34
ART	6	0.00 (2.90)		6	−1.17(3.56)
Loss
PFPP	19	8.26 (5.16)	0.70	19	6.95 (7.63)	0.46
ART	17	4.29 (5.60)		16	3.59 (6.65)

^aLoss status: No-loss = No interpersonal loss reported in the 6 weeks prior to panic disorder onset, Loss = interpersonal loss reported loss in the 6 weeks prior to panic disorder onset Abbreviations: LOCF = last observation carried forward, PDSS = Panic Disorder Severity Scale, SDS = Sheehan Disability Scale

^bOne subject did not correctly fill out SDS, hence differences in N.