Effectiveness of Interpersonal Psychotherapy-Trauma For Depressed Women with Childhood Abuse Histories

Abstract

Background

Women with depression and childhood sexual abuse histories constitute more than 20% of the female patient population in publicly-funded community mental health centers (CMHCs). Interventions are needed that address depression and post-traumatic stress disorder (PTSD) symptoms and social health.

Method

We compared Interpersonal Psychotherapy-Trauma (IPT-T), an IPT adaptation for this population and setting, to Clinic Psychotherapy (CP). CP clinicians were free to implement the psychotherapy of choice. Women (n=162; 54% White, 10.5% Hispanic; 52.2% Medicaid-insured) with a major depressive episode (MDE) and history of sexual abuse before 18 were randomly assigned to IPT-T or CP. Participants were offered 16 free sessions of IPT-T or CP. Outcomes were MDE remission, improved depression (Beck Depression Inventory, Hamilton Depression Rating Scale) and PTSD symptoms (Modified PTSD Symptom Scale), and improved social health (e.g., UCLA Loneliness Scale). Weighted generalized estimating equations were used to examine outcomes at eight (primary) and 20 (secondary) months post-randomization.

Results

IPT-T led to greater improvements in PTSD symptoms and many social health indicators, including loneliness and social support, at both eight- and 20-month follow-up. Effect sizes ranged from .18 to .39 at the primary endpoint (eight months). IPT-T and CP yielded comparable improvements in depression symptoms and MDE remission.

Conclusion

Evidence-based psychotherapies like IPT-T are needed in CMHCs, where some of the most vulnerable patients receive treatment. With comparable findings for depression, IPT-T’s superior improvements in social health and PTSD symptoms are cause for optimism. IPT-T should be evaluated in dissemination trials.

Introduction

Women with depression and histories of childhood maltreatment, including sexual abuse, have disproportionately poor responses to psychotherapies and antidepressants (Nanni, Uher, & Danese, 2012; Nemeroff, 2016). They are also at elevated risk for morbidity (Talbot et al., 2009) and mortality (Brown et al., 2009) as a consequence of their psychological symptoms (e.g., depression, anxiety, PTSD) and their poor social health (e.g., poor social function, social disability, loneliness, partner conflict, low levels of social support, revictimization; Trickett, Noll & Putnam, 2011). Childhood sexual abuse histories are reported by 20% of the female population in publicly-funded community mental health centers (CMHC) (Talbot et al., 2011). CMHC-based interventions for depressed women with abuse histories are critically needed to reduce psychological symptoms and improve social health.

Herman and Hirschman (1977) were among the first in the mental health field to critique the culture of victim blaming and trauma denial that had characterized the professional literature on childhood sexual abuse. More than four decades later, despite the acknowledged need for interventions to address the sequelae of childhood sexual abuse, there remains a paucity of controlled trials of individual psychotherapy (Dorrepaal et al., 2014; Nanni et al., 2012). Outside of our prior study (Talbot et al., 2011), we are aware of no prior controlled trials of psychotherapy for women with major depression and histories of childhood sexual abuse.

Interpersonal Psychotherapy (IPT; Klerman, Weissman, Rounsaville, & Chevron, 1984), an evidence-based psychotherapy for depression (Weissman, Markowitz, & Klerman, 2000), is a promising intervention for women with major depression and sexual abuse histories in community mental health care for two reasons. First, IPT’s focus on interpersonal problem areas is a suitable fit for patients whose histories place them at risk for relationship difficulties (Klerman et al., 1984; Weissman et al., 2000). Second, IPT has been shown to be noninferior to cognitive behavioral therapy for both depression (Ekeblad, Falkenstrom, Andersson, Vestberg, & Holmqvist, 2016) and post-traumatic stress disorder (PTSD; Markowitz et al., 2015), and has been shown to be effective for treating depression among women with histories of trauma (Toth et al., 2013).

Prior studies have adapted IPT to meet the needs of specific populations (Stuart, Robertson, & O’Hara, 2006; Swartz et al., 2004). To address the needs of depressed women with childhood sexual abuse histories seen in CMHCs, we made four minor adaptations to the IPT protocol (Talbot & Gamble, 2008). A previous study using IPT with these modifications resulted in significant reductions in depressive symptoms compared to care-as-usual among CMHC patients (Talbot et al., 2011). To enhance generalizability to CMHCs broadly, the first two modifications addressed practical issues when providing care in CMHC settings. These modifications were designed to help ensure our targeted patient population, many of whom live below or near the poverty level, achieved an acceptable “dose” of therapy. First, we expanded the treatment period to accommodate the sporadic attendance patterns in a CMHC, where missed appointments due to poverty-related stressors and other barriers to regular attendance are not uncommon (Cort et al., 2012; Talbot & Gamble, 2008). Typically, IPT has been delivered in weekly sessions. In IPT-T, 16 sessions were delivered over 32 weeks. Second, an engagement analysis was used in initial sessions to help patients identify and problem-solve social, practical, and psychological barriers to treatment participation, such as an unsupportive partner, unreliable transportation, and shame associated with trauma histories (Kim, Talbot, & Cicchetti, 2009; Ford, 2015; Talbot et al., 1999). Third, given the racial and ethnically diverse nature of our CMHC and the need to provide culturally competent care, a sociocultural formulation was used to help patients identify ways in which cultural and economic influences (e.g., family beliefs, ethnic identities, religious practices, racism, poverty) may affect their interpersonal problems and depression. Fourth, the interpersonal deficits problem area was not included. Instead, a new problem area, interpersonal-patterns (Talbot & Gamble, 2008), was added to address the developmental nature of persistent interpersonal patterns. In standard IPT, the interpersonal deficits problem area is used when none of the other problem areas are applicable, and the patient describes longstanding interpersonal isolation or conflicts. In IPT-T, the interpersonal patterns problem area is used to help patients recognize longstanding patterns of behavior and emotion commonly seen among individuals with trauma histories that could undermine social health (e.g., isolation, mistrust, conflict). Most women readily perceive trauma-related interpersonal patterns that undermine their interpersonal effectiveness and, conversely, behaviors and emotions that strengthen it. In IPT-T, therapists explain those patterns within a developmental framework in an effort to remove blame and help women construct a coherent narrative emphasizing their resilience. Study therapists frequently used this problem area in tandem with another present-focused problem area. Together, these four adaptations constitute IPT-Trauma (IPT-T).

In this randomized controlled trial, we compared the effectiveness of IPT-T and usual care psychotherapy, designated here as Clinic Psychotherapy (CP). CP clinicians were free to choose the psychotherapy they wished to employ, and tailor these therapies to individual patients. We selected CP as the comparator because our goal was to examine whether IPT-T was more effective than the psychotherapies that patients would typically receive in a CMHC. CP was presented as a fully credible intervention for depression. There were equal numbers of master’s vs. doctoral-level clinicians in both arms of the trial, equivalent amounts of clinical experience, comparable workload expectations, and equivalent amounts of clinical supervision. Therapists in both conditions were not proscribed from reasonable therapeutic activities or interventions. CP was thus structurally equivalent to IPT-T. In this CMHC-based effectiveness study, patients were not prohibited from taking psychotropic medications.

In our pilot study of IPT-T (Talbot et al., 2011), we found that IPT-T was associated with greater reductions in depressive symptoms and PTSD symptoms compared to CP at nine-month follow-up. These results informed the primary hypotheses of this study which were: eight months after randomization, women assigned to IPT-T would be more likely to experience a remission of the presenting major depressive episode (primary endpoint) and report improvements in psychological symptoms (depression, post-traumatic stress) and overall social health (improved social function, perceived support, as well as reduced loneliness, social disability, and perceived negative social exchanges with intimate partners). Secondary hypotheses were that patients assigned to IPT-T would report better psychological and social health outcomes 20 months after randomization. We also hypothesized that they would report a stronger therapeutic alliance three weeks after beginning therapy, because IPT-T therapists explicitly addressed barriers to care and help patients understand what to expect in therapy (Oldham, Kellett, Miles, & Sheeran, 2012). These alliance-promoting activities might be particularly useful for patients with trauma histories (Ford, 2015). The effects of IPT-T on medication adherence, when prescribed, were explored.

Method

Sampling

The trial was conducted from February 2008 through May 2013 in a university-affiliated CMHC. The study was approved by the University Institutional Review Board (RSRB20535) and registered with clinical trials.gov (NCT00843700). A Data and Safety Monitoring Board was convened semiannually via teleconference. Study recruitment was accomplished during routine CMHC intakes (point-of-care screening) and by community referral (via outreach and advertisement). All patients were eligible to receive 16 sessions of therapy free of charge, were paid $30 each time they completed one of five assessments, and received $20 compensation for childcare and transportation, per session, as needed.

In an effort to increase generalizability, exclusion and inclusion criteria were minimal in this effectiveness study. Inclusion criteria were: current major depression and sexual abuse history prior to age of 18. PTSD was not an inclusion criterion. Exclusion criteria were: active psychosis, history of schizophrenia or bipolar disorder, significant cognitive impairment, active (i.e., not remitted) substance abuse/dependence within the previous three months, and need for partial or inpatient hospitalization. Consistent with prior CMHC-based effectiveness research (Merrill, Tolbert, & Wade, 2003), patients were permitted to be taking psychotropic prescriptions at study entry; changes in medication during the trial were permitted.

Screening Procedures

Eligibility criteria at screening were: 18 years or older, English speaking, presence of depressive symptoms (a list of depression symptoms was provided) and self-report of sexual abuse before age 18. During the recruitment period (February 2008 through June 2011), 2196 women were screened by CMHC clinicians (Figure 1), 352 (16%) of whom endorsed depressive symptoms and reported sexual abuse prior to age 18. Women who met initial screening criteria were handed a letter describing the study and requesting their agreement to be contacted by study personnel. The majority (n=312, 89%) agreed to be contacted, 12 (3%) declined, 28 (8%) were lost to contact, and 7 (2%) were discovered to be ineligible prior to the eligibility assessment. The remaining women (n=265) provided written informed consent and participated in the eligibility assessment.

Figure 1. Study recruitment, randomization and retention.

*One subject was withdrawn after randomization, prior to treatment initiation, because she re-located.

**We recruited women from February 2008 through June 2011, and completed the 14 and 20-month follow up assessments in March 2012, nine months after we enrolled the final patient. Consequently, the assessments ended before a minority of the randomized sample (28.6%; 46 women) became eligible for the 14- and 20-month assessments; smaller denominators were thus used to calculate the retention rates for the 14 month (IPT-T n= 71, CP n= 69) and 20-month (IPT-T n=55, CP n=60) assessments.

Eligibility Assessment and Randomization

The presence of major depression was established through the Structured Clinical Interview for DSM-IV Disorders (SCID; First, Spitzer, Gibbon, & Williams, 1995). Consensus diagnostic conference meetings, which included study investigators and research assessors, were held to establish the presence of a major depressive episode and other DSM-IV diagnoses. Presence of childhood sexual abuse was established through a structured clinical interview (Talbot et al., 2005). Sexual abuse was defined as: any unwanted sexual contact before age 18, or any sexual contact with a family member ≥5 years older. This definition excludes non-contact abuse (i.e. indecent exposure). Sexual contact was defined as physical contact of a sexual nature, ranging from fondling/contact with sexual organs to sexual intercourse. This widely-endorsed (Barth, Bermetz, Heim, Trelle, & Tonia, 2013) definition permitted the inclusion of depressed women with relatively mild abuse histories in this effectiveness study (though 96% reported moderate to severe abuse; see Results).

A minority of women (n=103, 39%) were excluded, mainly because they did not meet the major depressive episode criterion; 162 were randomized to IPT-T or CP using a web-based portal. More than half of the study participants were White (53%), 40% were Black, 7% reported being multi-racial, and 11% were Hispanic. The modal participant was in her 30s [IPT-T: 36.27 (11.20) years old; CP: 39.08 (12.84) years old], insured by Medicaid, and living in poverty. Table 1 shows there were no demographic differences across study arms. To decrease the likelihood of imbalances across the comparators, urn randomization was used, based on six variables that were theoretically related to study outcomes: prescribed antidepressant at study entry, active PTSD, other active anxiety disorder, history of substance abuse/dependence, borderline personality disorder, and referral source (routine CMHC intake vs. community referral).

Table 1.

Baseline characteristics (N=161)

	IPT-T (N=83)		CP (N=78)
	n	%	n	%
Race
White	44	53	43	55
Black	33	40	27	35
Multi-racial or other race	6	7	8	10
Hispanic ethnicity	9	11	8	10
Marital Status
Living with spouse or partner	25	30	25	32
Not living with spouse or partnera	58	70	53	68
Education
High school diploma or less	53	64	52	67
Post-secondary education	30	36	26	33
Insured by Medicaid
Yes	43	52	41	53
No	40	48	37	47
Annual Incomea
Less than $10,000	33	40	31	40
$10,000 – $19,999	17	21	17	22
$20,000 – $39,999	18	21	16	20
$40,000 or more	11	13	12	15
Unknown or refused to disclose	4	5	2	3
Referral Source
CMHC Intake	54	65	43	55
Community recruitment/referral	29	35	35	45

Note: IPT-T=Interpersonal Psychotherapy-Trauma; CP=Clinic Psychotherapy; CMHC=community mental health center.

^aMost of the women (n=126; 78.3%) were living with at least one minor child; in 80% of those households (n=101), no other adults were present. More than 20% were living with four or more minor children. In 2010, the household poverty threshold for mother-headed households with one child was $15,030 (DeNavas-Walt, Proctor, & Smith, 2011).

Comparator Conditions

Participants were eligible to receive 16 free individual psychotherapy sessions (IPT-T or CP) within the eight-month treatment phase. Treatments received after the eight-month treatment phase were tracked up to 20 months following enrollment. Patients were seen by 14 CMHC staff clinicians (ten master’s-level, four doctoral-level). Of note, many of the therapists who treated patients in this trial also treated patients in our pilot trial (Talbot et al., 2011). The seven therapists who said they could be available during a common group supervision time were assigned to IPT-T training; the others were assigned to CP. None of the CP therapists received training in IPT or IPT-T during the pilot study or this trial.

IPT training consisted of 30 hours of group didactic training, intensive treatment manual review (Stuart & Robertson, 2003), and supervision of two to three training cases by the senior author. IPT-T therapists were required to complete at least one full 16-session treatment that demonstrated their adherence to and competence in IPT-T. The senior author and/or an IPT-trained junior investigator reviewed approximately 10% of audio- and videotaped sessions of IPT-T therapists for treatment fidelity. Deviations from the model were addressed in the group supervision meetings. All IPT-T therapists who entered the trial completed the trial, and all were judged to be adherent to IPT and IPT-T modifications.

CP and IPT-T therapists received comparable amounts of clinical supervision (10 hours per month), which was provided by senior clinicians in the CMHC. The IPT-T group supervision was held weekly for an hour, and is included in the supervision total tally. CP therapists did not participate in the weekly IPT-T group supervision provided by the research study, which was led by the senior author. In post-treatment questionnaires, CP therapists identified their primary therapeutic approaches (not their “therapeutic allegiance”) with the 78 CP patients as: supportive (n=55; 70%), cognitive-behavioral or dialectical-behavioral (n=11; 14%), integrated/eclectic (n=9; 12%), family systems (n=2; 3%), and client-centered (n=1; 1%).

Outcome Assessments

Assessments were conducted from February 2008 to March 2012. Patient assessments were conducted in person at pretreatment. Follow-up assessments were conducted at 2, 4, 8, 14, and 20-months post-randomization. Two master’s level research assistants, blind to treatment assignment, conducted all assessments. As needed, based on participants’ reading level or preference, self-report instruments were administered orally. Some assessments were conducted by phone (interviewer-rated measures) and mail (self-report measures) when the alternative was the loss of follow-up data.

All study outcomes were assessed with widely-used psychometrically-established self-report instruments or with interview measures, the Hamilton Depression Rating Scale (HDRS; Hamilton 1960) and the SCID (First et al., 1995). Members of the research team met weekly to review HDRS ratings to ensure consistency, by reviewing audio-recordings and/or notes taken during the interview. Axis I diagnoses were established at consensus conferences based on diagnostic reports summarizing the SCID interview and a review of the medical record. These conferences were attended by at least one psychologist, one psychiatrist, and four experienced master’s level research staff, none of whom were students.

Major depression and psychological symptoms

Major depression was assessed using the SCID (First et al., 1995). Depressive symptom severity was assessed at study entry and subsequent assessments using the 21-item (alpha=.76) Beck Depression Inventory-II (BDI-II; Beck, Steer, & Brown, 1996), a self-report measure, and the 17-item (alpha=.85) Hamilton Depression Rating Scale (HDRS; Hamilton 1960), an interviewer-administered measure. PTSD symptoms were measured with the Modified PTSD Symptom Scale (MPSS; Coffey, Dansky, Falsetti, Saladin, & Brady, 1998), a reliable and valid 17-item (alpha=.89) self-report questionnaire which produces a score reflecting the frequency and severity of PTSD symptoms.

Social health outcomes

Social health was evaluated using self-report measures of social function [the widely used Medical Outcomes Study SF-36 (Ware & Sherbourne, 1992) social functioning subscale, alpha=.68], perceived support [11-item (alpha=.65) Duke Social Support Questionnaire (Parkerson et al., 1989), e.g., “Can you talk about your deepest problems with at least some of your family and friends?”], loneliness [20-item (alpha=.92) UCLA Loneliness Scale (Russell, 1996), e.g., “There are people who really understand me”], and social disability [13 items (alpha=.95) from the World Health Organization Disability Assessment Schedule II (World Health Organization, 2001), e.g., “In the last 30 days how much difficulty did you have dealing with people you do not know?”]. We also administered the 16-item Test of Negative Social Exchange (Ruehlman & Karoly, 1991), which uses four 4-item scales to assess perceived partner hostility (“my partner yelled at me,” alpha=.93), partner insensitivity (“my partner took my feelings lightly,” alpha=.91), partner interference (“my partner was too demanding of my attention,” alpha=.83), and partner ridicule (“my partner made fun of me,” alpha=.76).

Early therapeutic alliance

Following the third therapy session, participants were mailed the 12-item (alpha=.58) Working Alliance Inventory-short form (Busseri & Tyler, 2003; Tracey & Kokotovic, 1989). A sample item is: “We have established a good understanding of the kinds of changes that would be good for me.” To dissuade socially desirable responses, participants were informed their responses would not be shared with their therapists.

Medication adherence

To explore whether rates of medication adherence were comparable across treatment arms, we administered the Longitudinal Interval Follow-up Evaluation Treatment Interview (Keller et al., 1987) at each assessment point. These self-report data of pharmacological treatment adherence were used to ordinally categorize antidepressant medication-taking as follows: (a) no medications prescribed, (b) prescribed but not taking medication, (c) taking less than half the time, (d) half the time, (e) most of the time, (f) nearly all the time, or (g) all medications taken as prescribed, all the time.

Descriptive and Exploratory Measures

Beyond major depression, other current Axis I mental disorders were assessed using the SCID (First et al., 1995). Borderline personality disorder was assessed with the Structured Clinical Interview for DSM-IV Personality Disorders (First, Spitzer, Gibbon, Williams, & Benjamin, 1997). The severity of childhood trauma was assessed by the Childhood Trauma Questionnaire-short form (CTQ; Bernstein et al., 1994), a self-report measure of the perceived frequency of physical and emotional abuse, sexual abuse, and physical and emotional neglect. Participants provided self-report data on other health conditions known to be associated with depression (Katon, Lin, & Kroenke, 2007): arthritis, diabetes, hypertension, and chronic pain. The number of therapy sessions attended during the course of the study was tracked.

Statistical Analysis

Before conducting multivariate intention-to-treat analyses of IPT-T effects, we compared participants assigned to IPT-T and CP on demographic variables and baseline clinical measurements using t-tests, Wilcoxon rank tests and Chi-square tests. To accommodate the longitudinal nature of the study design and address missing values, we applied weighted generalized estimating equations (WGEE) to model the marginal mean responses over the eight- month treatment period (primary) and over the entire 20-month trial (secondary). Unlike the linear mixed-effects model, which explicitly models the between and within subject variations using random effects, the semi-parametric GEE approach ignores between-subject variability by treating subjects as independent units and basing model estimation and inference on the marginal distribution of the response of such units (Robins, Rotnitzky, & Zhao, 1995). GEE can therefore provide robust estimates because it requires no distribution assumptions.

We addressed missing data issues by using weighted GEE (WGEE), which provides consistent estimates when data are missing at random (Fitzmaurice, Laird, & Ware, 2011; Preisser, Lohman, & Rathouz, 2002). WGEE models the marginal means of the responses while weighting the observations by the inverse of the probability of being observed (Horvitz & Thompson, 1952). The concept is to treat each observation as representative of a group of subjects to account for the unobserved subjects (i.e., missing data). Under the assumption of missing at random (vs. missing completely at random), WGEE can provide consistent and robust estimates (Fitzmaurice et al., 2011; Preisser et al., 2002). This assumption implies that missing data are dependent on the observed information.

We computed the WGEE weights using logistic regression at each follow-up assessment to model the missing values of the outcomes and to estimate the likelihood of the outcome being observed. The predictors we considered for modeling missing values were the six urn randomization variables and the previous value of the outcome. A logistic regression model was also applied to the treatment assignment, with the baseline values of the corresponding outcome as predictors. The weights were calculated by the inverse of the product of the two estimated probabilities. Thus, subjects who were less likely to be observed and/or less likely to be assigned to the treatment condition were weighted more heavily in the WGEE model.

Separate WGEE models were applied to the eight-month treatment period and the whole 20-month study course. Generalized linear models were applied for marginal means of the outcome variables, with an identity link function for continuous outcomes and a logit link function for binary outcomes. All urn randomization variables served as covariates (antidepressant medication prescribed, post-traumatic stress disorder, borderline personality disorder, other anxiety disorder, history of substance abuse/dependence, referral source) along with the assessment time point. For the main study outcome, remission of the major depression episode at eight months, a sensitivity analysis using logistic regression was conducted.

WGEE models for 20-month outcomes were similar, except missing values at 14- and 20-months post-randomization were coded as due either to planned termination of assessments (March 2012) or subject unavailability (e.g., withdrawal); interactions between that code and other covariates were included. We recruited women from February 2008 through June 2011, and completed the 14 and 20-month follow up assessments just nine months after the last patient enrolled. Consequently, the assessments ended before a minority of the randomized sample (28.6%; 46 women) became eligible for the 20-month assessment. At the 20-month follow-up, data were available on 71 participants (n=41 in IPT-T), representing 62% of the 114 women who conceivably could have completed a 20-month assessment by the end of March 2012.

We used logistic regression to compare medication adherence (taking all, nearly all or most antidepressants vs. other) between the two treatment arms at the eight-month assessment.

Two-sided tests were conducted and p-values of <0.05 were considered significant. Data analyses were conducted with SAS version 9.3 (SAS Institute, Inc. Cary, NC).

RESULTS

Table 2 shows that women assigned to IPT-T were more likely to be diagnosed with PTSD than women assigned to CP (p=0.0547), but there were no differences in mood disorder diagnoses or other comorbidities. The mean (SD) duration of the current depressive episode was comparable, 16.1(16.1) months for IPT-T vs. 17.9 (17.1) months for CP. Most women (n=134, 83%) were diagnosed with chronic depression by meeting one of these criteria: depressive episode > 24 months, incomplete remission between the current and past depressive episodes, or a depressive episode in conjunction with pre-existing dysthymic disorder. Among the 141 participants with PTSD, 67% (n=95) identified sexual abuse before age 18 as the focal event of their current PTSD symptoms. Although women with mild abuse histories were eligible, nearly all (n=155; 96%) reported moderate to severe levels of sexual abuse on the CTQ. More than 10% of the enrolled women reported diabetes (n=21, 13%), more than 20% reported hypertension (n=35, 22%), more than 30% reported arthritis (n=53, 33%) and nearly 40% reported chronic pain (n=61, 38%).

Table 2.

Baseline clinical characteristics

	IPT-T (N=83)		CP (N=78)
	n	%	n	%
Major depression, single episode	20	24	17	22
Major depression, recurrent	63	76	61	78
Chronic depressiona	69	83	65	83
Post traumatic stress disorder	77	93	64	82
Borderline personality disorder	49	59	37	47
Other anxiety disorder	48	58	45	58
Substance abuse/dependence history	29	35	33	42
Antidepressant medication prescribed	51	61	40	51

Note:IPT-T=Interpersonal Psychotherapy-Trauma; CP=Clinic Psychotherapy; All patients met DSM-IV criteria for an active major depressive episode. The most common major depression diagnosis at study entry was 296.33, major depression, recurrent, severe (n=84, 52%). There were no cases of mild major depression (296. 21, 296.31)

^aIncludes double depression (current major depression along with a pre-existing dysthymic disorder), incomplete remission between prior episodes, and current episode is ≥ 24 months

By the time the participants had completed three therapy sessions, women assigned to IPT-T reported stronger working alliances with their therapists (IPT-T vs. CP: Mean (SD)=63.9 (12.2) vs. 58.8 (14.7); p = 0.04). By the end of the active treatment phase, there were no differences in rates of self-reported antidepressant medication adherence or rates of referral for medication across the two arms (data not shown), but women assigned to IPT-T had attended nearly twice (10.6±5.8) as many psychotherapy sessions as those in CP (5.7±4.2). Despite their stronger alliance and better attendance in therapy, the major depressive episodes of women assigned to IPT-T were not more likely to remit (GEE p=.35, Logistic Regression p=.35). More than half of the women assigned to IPT-T (36/66, 54%) and nearly two-thirds of the women assigned to CP (36/56, 64%) were still experiencing an active episode of major depression [OR (95% CI) = .69 (.31, 1.51)]. Similarly, as shown in Table 3, the effects of IPT-T and CP on depressive symptoms were comparable, whether measured by the BDI-II or the HDRS. Nonetheless, Table 3 shows that women assigned to IPT-T experienced a significantly larger reduction (Cohen’s d=.35) in the severity of post-traumatic symptoms; the effect of IPT-T on the remission of PTSD was not significant (data not shown).

Table 3.

Effects of IPT-T on PTSD and depression symptom severity at 8 months

	IPT-T		CP		B (SE)	p	d
	Study entrya (n=83)	Follow-upa (n=66)	Study entrya (n=78)	Follow-upa (n=56)
BDI	34.6 (9.6)	24.3 (14.7)	35.0 (8.7)	23.1 (13.8)	−2.22 (1.79)	.21	.20
HDRS	22.1 (4.1)	17.6 (7.5)	22.6 (5.2)	16.8 (7.0)	−.28 (0.93)	.77	.07
MPSS	67.9 (31.7)	47.6 (33.1)	73.5 (32.8)	58.9 (37.4)	−12.10 (5.02)	.02	.35

Note:IPT-T=Interpersonal Psychotherapy-Trauma; CP=Clinic Psychotherapy; BDI=Beck Depression Inventory; HDRS=Hamilton Depression Rating Scale; MPSS= Modified PTSD Symptom Scale.

^aAll values in this column are M (SD). Parameter estimates and p values were derived from WGEE analyses with baseline covariates (antidepressant medication prescribed, post-traumatic stress disorder, borderline personality disorder, other anxiety disorder, history of substance abuse/dependence, referral source) and interim follow-up assessments. Follow-up rates were 83% at 8-weeks (n= 68 in IP-T, n= 66 in CP) and 79% at 16 weeks (n= 64 in IPT-T and n=64 in CP). IPT-T led to significant improvements in MPSS (bolded) at 20-month follow-up. See text for details of those analyses.

Table 4 shows the effects of IPT-T on social health at the 8-month follow-up. Women assigned to IPT-T reported greater improvements in social function and perceived support as well as greater reductions in loneliness, partner hostility, partner insensitivity, and partner ridicule. The mean effect size (d) was .31.

Table 4.

Effects of IPT-T on Social Health at 8-month Follow-Up

	IPT-T		CP		B (SE)	P	d
	Study entry (n=83)a	Follow-up (n=66)a	Study entry (n=78)a	Follow-up (n=56)a
Social Function	30.8 (11.3)	35.3 (13.0)	27.5 (10.6)	33.1 (13.2)	5.16 (1.63)	.003	.39
Perceived Support	33.9 (5.5)	35.9 (6.1)	33.5 (5.9)	34.9 (6.6)	1.98 (0.84)	.02	.33
Loneliness	64.8 (14.3)	56.9 (15.5)	67.5 (12.9)	60.5 (16.6)	−4.71 (2.24)	.04	.32
Social Disability	18.1 (6.1)	16.4 (7.3)	19.8 (6.9)	17.2 (7.8)	−1.80 (1.08)	.10	.23
Partner hostility	3.5 (4.4)	3.3 (4.3)	4.1 (5.2)	4.3 (5.3)	−1.80 (0.86)	.04	.32
Partner insensitivity	4.5 (5.2)	3.4 (5.0)	5.1 (6.3)	5.0 (6.4)	−2.40 (1.01)	.02	.37
Partner interference	2.2 (3.6)	2.5 (3.9)	2.8 (4.0)	3.1 (5.0)	−1.21 (0.76)	.12	.18
Partner ridicule	1.3 (2.2)	1.1 (2.3)	1.5 (2.8)	1.50 (3.0)	−1.36 (0.53)	.01	.37

Note: IPT-T=Interpersonal Psychotherapy-Trauma; CP=Clinic Psychotherapy;

^aAll values in this column are M (SD). Parameter estimates and p values were derived from WGEE analyses with baseline covariates (antidepressant medication prescribed, post-traumatic stress disorder, borderline personality disorder, other anxiety disorder, history of substance abuse/dependence, referral source) and interim follow-up assessments. Follow-up rates were 83% at 8-weeks (n= 68 in IP-T, n= 66 in CP) and 79% at 16 weeks (n= 64 in IPT-T and n=64 in CP). IPT-T led to significant improvements in bolded outcomes (social function, social disability, partner ridicule) at 20-month follow-up. See text for details of those analyses.

At the 20-month follow-up, women assigned to IPT-T continued to attend more individual therapy sessions than women assigned to CP (10.7±5.6 vs. 5.6±4.0). However, IPT-T was not more likely (p=.25) to lead to remission of major depression [OR (95% CI) = .53 (.18, 1.56)]; likewise, levels of self-reported depression [M (S.D.)] on the BDI [IPT-T: 20.1 (15.8) vs. CP: 20.1 (12.6)] and observer-rated depression on the HDRS [IPT-T: 15.7 (7.8) vs. CP:16.4 (6.4)] remained comparable in both arms. Nonetheless, one year after receiving an average of 10 IPT-T sessions, women experienced greater improvement than those receiving CP on four outcome variables: PTSD symptoms (B (SE)= −13.90 (4.98); p = 0.007), social functioning (B (SE)= 6.14 (1.67); p=.0006), partner ridicule (B (SE)=.94 (.48); p = 0.05), and social disability (B (SE)=− 2.24(1.00); p = 0.03). There were no other statistically significant findings at 20-month follow-up.

DISCUSSION

Findings provided mixed support for the hypotheses. While effective, the hypothesized superior effects of IPT-T on depression were not supported. However, IPT-T did lead to a greater reduction in the severity of post-traumatic stress symptoms that persisted for one year following the end of treatment. Women who received IPT-T also demonstrated marked improvements in social health by the end of treatment, including improved social function, reduced loneliness, enhanced perceived support, and improved relationships with their partners, with reduced experiences of partner ridicule, hostility, and insensitivity. These effects were lasting—even a year after the end of active IPT-T treatment, women reported better social function, less social disability, and less ridicule from their partners. Despite IPT’s explicit focus on social connectedness, this is the first study to demonstrate IPT’s impact on loneliness (Cacioppo, Grippo, London, Goossens, & Cacioppo, 2015). Given the deleterious effects of both loneliness and low social support on health and well-being—including mortality—these results highlight that IPT may promote health and well-being not just by reducing depression, but by promoting social connectedness. Based on the sustained advantage in post-traumatic stress symptoms and social health in the IPT-T group, we conclude that the CP group was in a more vulnerable position at the close of the study.

The null findings for major depressive episode remission and depressive symptoms must be understood in the context of prior studies. The reductions in BDI and HDRS scores in women assigned to IPT-T in this study and our prior study were quantitatively comparable (and nearly identical) (Talbot et al., 2011). However, the women assigned to CP in this study reported lower levels of depression at the end of treatment than women assigned to CP in the earlier study. For example, the post-treatment BDI score [M (S.D.)] in CP was 27.1 (13.6) in the prior study and was four points better [23.1 (13.8)] in this trial. It is unclear why women assigned to CP in this study appear to have done better than women assigned to CP in the earlier study. It may be relevant that rates of Medicaid eligibility (and, potentially, poverty-related stressors) were lower in the current study (52% vs. 66%), but there are many other potential explanations (e.g., different therapists, random variation).

As in our prior study, depression scores improved but remained elevated following active treatment in both arms. Given that IPT has been shown as an effective treatment for major depressive disorder, the remaining elevation in depression symptoms in this study and in our prior study could be interpreted to mean that our modifications of IPT handicapped what is usually a quite robust treatment for depression. In other words, it is possible that the depression outcomes reported here would have been better had we employed orthodox IPT instead of IPT-T. We believe that a more parsimonious explanation for the depression outcomes is that this patient population has more severe (and chronic) depression and comorbidities than patients enrolled in previous trials. Clients in CMHCs frequently present with moderate to severe trauma histories and severe depression: the levels of depression in this sample at study entry were more than .5 SD higher than those reported in two recent studies (Grote et al., 2012; Markowitz et al., 2015).

Even among patients who do not struggle financially, the presence of an early trauma history increases the likelihood that a mood disorder will be treatment-refractory, no matter whether that treatment is psychotherapy, psychotropic medications, or both (Klein et al., 2009; Nanni et al., 2012). Given these persistently disappointing findings, Nemeroff (2016) proposed that “depression in patients with a history of (childhood trauma) is a fundamentally distinct biological endophenotype with a corresponding difference in response to conventional treatment” (p. 902). The chronicity of the disorder, together with the presence of other Axis I mental disorders, psychosocial stressors, personality pathology, chronic physical conditions that are known to exacerbate depression (e.g., pain, arthritis), and poverty may all contribute to an incomplete improvement in depression symptoms. Despite this, IPT-T led to significant improvements in post-traumatic stress symptoms and social health that were sustained a year after treatment ended.

Orthodox IPT has been shown to lead to improvements in PTSD; this is particularly notable because IPT does not include exposure-based methods that characterize many efficacious PTSD interventions (Markowitz et al., 2015). Our prior study similarly showed that IPT-T led to a reduction in post-traumatic stress symptoms (Talbot et al., 2011) and this finding was replicated here. Although the current study was not designed to examine mechanisms of change, it is likely that IPT-T led to improvements in PTSD symptoms by improving social and emotional function (Markowitz, 2010; Markowitz, Neria, Lovell, Van Meter, & Petkova, 2017). Specifically, mastery of interpersonal skills may lead to reduced avoidance of persons and activities reminiscent of the trauma. Just as deteriorated interpersonal resources have been linked to the exacerbation of post-traumatic symptoms (Schumm, Hobfoll, & Keogh, 2004), IPT-induced gains in overall social health could mitigate PTSD symptoms.

Improvements in social health are perhaps the most clinically significant findings to emerge from this trial. A 23-year prospective study of 6-16 year-old females who experienced sexual abuse showed that abuse leads to earlier onset of puberty, maladaptive sexual development, and physical and sexual revictimization, all of which could have significant implications for overall social health (Trickett, Noll & Putnam, 2011). Improving loneliness, interpersonal function and social support are explicit targets of IPT, and those targets appear to have been reached with IPT-T delivered in a CMHC setting. Given the severe and persistent social problems that result from abuse, and the unique circumstances of women who seek treatment in a CMHC, we slightly modified the delivery of IPT. As in our prior study (Smith et al., 2012), patients in IPT-T reported better working alliances with their therapists by the third session. Based on prior findings (Flueckiger, Del Re, Wampold, Symonds, & Horvath, 2012), we believe this stronger alliance led patients to attend more sessions and experience greater gains in social health. Improvements in the alliance may have been engendered by therapists directly addressing barriers to care and by preparing patients for psychotherapy (i.e., “role induction” (Oldham, Kellett, Miles, & Sheeran, 2012). Although the stronger alliance did not lead to more robust improvements in depression, these data suggest that, among women who are particularly vulnerable to shame and mistrust (Ford, 2015; Kim et al., 2009; Markowitz et al., 2017; Talbot et al., 1999), a strong early alliance might facilitate improvement in social health. In the long run, it is plausible that the experience of a strong therapeutic alliance and more regular treatment adherence could lead to continued gains in social health that would eventually result in a more sustained recovery from depression (Klerman et al., 1984).

Poor social health has been shown to confer risk for morbidity and mortality, including death by suicide (Duberstein et al., 2004; Holt-Lunstad, Smith, & Layton, 2010; Steptoe, Shankar, Demakakos, & Wardle, 2013). Moreover, the risk of premature mortality attributable to poor social health is comparable to risks from health-damaging behaviors (e.g., physical inactivity, smoking). Therefore, any intervention that reliably improves social health warrants widespread implementation.

The positive findings for social health and PTSD symptoms, together with other promising findings for patients with sexual trauma (Markowtiz et al., 2017), are sufficient to warrant dissemination of IPT, appropriately modified (Glasgow et al., 2012), in settings where depressed women with histories of sexual abuse are disproportionately represented. The poverty-driven processes that sustain and exacerbate chronic depression in women with abuse histories seen in CMHCs (Cort et al., 2012) will be very difficult to modify without greater understanding of their day-to-day lives (Belle, 1994) and greater attentiveness to their basic unmet needs (e.g., food, transportation, safety, health literacy) and physical health (Druss et al., 2008; Gottlieb, Wing, & Adler, 2017). Recent studies have shown that, in other specialty care settings, screening for and attending to unmet basic resource needs can lead to improvements in important outcomes, such as blood pressure, cholesterol, diabetes self-care, and medication adherence (Berkowitz et al., 2017; Gottlieb et al., 2017). Ideally, future efforts to disseminate IPT-T would be well-integrated with management of patients’ physical conditions (Druss et al., 2009), as well as their social and economic needs (Berkowitz, Hulberg, Standish, Reznor, & Atlas, 2017; Gottlieb et al., 2017). Given that major depression in women with histories of childhood sexual abuse is frequently refractory to shorter-term treatments (Klein et al., 2009), future research in CMHCs could consider adding maintenance sessions of IPT-T.

Of course, other psychotherapies can plausibly benefit depressed women with histories of abuse. For example, in a review of neurobiological findings in individuals who have been maltreated, Teicher and Samson (2013) identified mentalization-based therapy (Bateman and Fonago, 2009) and the cognitive-behavioral analysis system of psychotherapy (CBASP; Nemeroff et al., 2003) as potentially promising options. We agree that these interventions might prove to be beneficial for some patients, but more research is clearly needed. We are aware of no studies of mentalization in depressed patients with histories of childhood sexual abuse, and CBASP did not outperform medication in the subgroup of patients with childhood sexual abuse in Nemeroff et al. (2003).

Strengths and limitations of the study should be mentioned. Internal validity was strengthened by the use of a randomized controlled design with urn randomization, blinded outcomes assessment, specification of the major depression inclusion criterion, and close monitoring of treatment provision by clinicians. Reflecting the external validity objectives of effectiveness trials of interventions that can be disseminated beyond highly controlled treatment settings (Glasgow et al., 2012), we included patients with multiple comorbid disorders, employed staff clinicians rather than research therapists, used a naturalistic medication design, explored a wide range of patient-centered outcomes, and adapted an efficacious treatment to enhance its suitability to a particular patient population and treatment setting. In sum, this design maximizes generalizability to CMHCs in the United States.

Limitations include the sample, which is not generalizable to men or to women who do not actively seek treatment in CMHCs. Nonetheless, unlike our prior study which was drawn exclusively from women who sought treatment on their own accord from a CMHC, about 40% of the sample was recruited from community settings (e.g., advertisement; neighborhood organizations). Second, although we tracked self-reported medication use and chart-based indicators of medication referral, we have not assessed the robustness of medication dosage, and did not hold medication doses constant in this effectiveness trial. Third, quantitative data on therapist fidelity to the IPT-T model were not analyzed, but the overall pattern of findings suggests that treatment fidelity was acceptable.

In closing, given the prevalence of childhood abuse worldwide, and the chronic mental and physical health burden conferred by abuse, effective treatments for depression and comorbidities are critically needed to complement broader public health efforts to prevent childhood sexual abuse from occurring in the first place (Herman & Hirschman, 1977; Ford, 2015). To the best of our knowledge, this is just the second psychotherapy trial for women with childhood sexual abuse histories and DSM-defined major depression (Talbot et al., 2011). Interpersonal Psychotherapy was adapted to the needs of depressed women with sexual abuse histories seen in community mental health care, and found to be more effective than other psychotherapies in improving post-traumatic stress symptoms and social health. The presence of an early trauma history increases the likelihood that a mood disorder will be difficult to treat (Klein et al., 2009; Nanni et al., 2012; Nemeroff, 2016), and the present study’s findings are consistent with that view. Nonetheless, the observed improvements in the therapeutic alliance, session attendance, social function, perceived support, loneliness, social disability, partner relationships, and symptoms of post-traumatic stress are cause for optimism.

Public Health Significance Statement.

Given the prevalence of childhood sexual abuse worldwide, and the chronic mental, physical, and social health burden conferred by sexual abuse, effective treatments are critically needed. Interpersonal Psychotherapy was adapted to the needs of depressed women with sexual abuse histories seen in publicly-funded community mental health centers (IPT-Trauma) and compared to other psychotherapies that are routinely administered in those settings (Clinic Psychotherapy). No differences were observed in depression outcomes, which improved in both conditions, but women who received IPT-Trauma reported greater improvements in other outcomes that are important to them, including social function, perceived social support, loneliness, social disability, partner relationships, and symptoms of post-traumatic stress. To improve outcomes among women with histories of childhood sexual abuse. dissemination and implementation of IPT, appropriately modified, should be considered.