Davanloo’s Intensive Short-Term Dynamic Psychotherapy in a tertiary psychotherapy service: overall effectiveness and association between unlocking the unconscious and outcome

Introduction

Mental disorders are common conditions that generate huge costs for society and cause a great deal of suffering for the affected individuals and their families (Ustün, 1999; Kessler et al., 2007). Several mental conditions such as depression, anxiety, personality disorders and a range of medically unexplained symptoms are suggested to be linked to adverse childhood experiences (ACE; Felitti et al., 1998; Edwards et al., 2003; Anda et al., 2006). Problems arising from ACE are suggested to be byproducts of strong unprocessed emotions coupled with deficits in capacity to regulate emotions (Schore, 2001; Schore, 2002; Anda et al., 2006). Research indicates that several individuals with conditions linked to ACE are likely to be treatment-resistant or to experience recurrences (Chartier, Walker & Naimark, 2010). One consequence of this treatment resistance can be chronic illness and large system costs (Nanni, Uher & Danese, 2012). Recently, there have been calls for the development and evaluation of intensive and alternative treatments that address treatment resistance in these patient populations (Nanni, Uher & Danese, 2012).

Intensive Short-Term Dynamic Psychotherapy (ISTDP; Davanloo, 1990; Davanloo, 2000; Davanloo, 2005; Abbass, Town & Driessen, 2012) is an emotion-focused psychodynamic psychotherapy that explicitly addresses treatment resistance in somatic conditions, mood, anxiety and personality disorders arising from ACE. Psychotherapeutic effects in ISTDP are hypothesized to be dependent on in-session mobilization of emotional processes in the form of rise in complex transference feelings, defined by co-occurring increases in treatment resistance, unconscious anxiety, therapeutic alliance and experience of feelings (Davanloo, 2005). A major unlocking of the unconscious is a powerful in-session emotional breakthrough achieved at a high rise in complex transference feelings assumed to take place when treatment resistance has been adequately challenged. During an unlocking of the unconscious the patient experiences intense complex feelings towards the therapist or another current figure, in which the experience is linked to feelings to past figures and emotion-laden memories about painful feelings, situations and events from the past, i.e., events related to adverse childhood experiences. A fundamental assumption in Davanloo’s ISTDP is that the unlocking of the unconscious is associated with therapeutic gains. Recent research from a sample of 89 patients, all treated by an experienced ISTDP therapist, support Davanloo’s claim (Town, Abbass & Bernier, 2013). In addition, ISTDP has been shown to be effective for a range of conditions (Abbass, Town & Driessen, 2012; Town & Driessen, 2013) including patients who failed to respond to specialist medical, surgical or psychiatric care (Baldoni, Baldaro & Trombini, 1995; Abbass, 2006; Abbass, Lovas & Purdy, 2008; Abbass, Town & Bernier, 2013), or who used an unusually high amount of emergency services (Abbass et al., 2009). However, a majority of existing studies have included small samples and therapies delivered by expert therapists, and are thereby limited in generalizability (Abbass, Town & Driessen, 2012).

The major aims of this study were to evaluate the effectiveness of ISTDP in a tertiary psychotherapy service where patients had been referred from specialist care, and to study if unlocking the unconscious was associated with outcome after controlling for potentially confounding variables, including treatment resistance. The present study differs from most previous research on ISTDP in that it is based on a large clinically representative sample of patients treated by therapists with a range of experience and that it uses multilevel growth curve techniques to model symptom change over time.

Materials & Methods

This study is reported in accordance with the CONSORT statement for clinical trials (Schulz, Altman & Moher, 2010). The study is an evaluation of the effectiveness of a treatment given in a tertiary care clinic and used anonymized data collected as part of standard care. The project was reviewed and approved by the Capital District Health Authority Research Ethics Board in Halifax, Nova Scotia (approval number 2007-050), and is registered in Clinicaltrials.gov as identifier number NCT01924715.

Participants and procedure

Participants in the study were collected from a large sample of patients referred to the Centre for Emotions and Health between March 30, 1999 and March 30, 2007. A substantial amount of patients referred were seen for an assessment meeting only. In this study, we investigated outcome for all patients who received at least one session of ISTDP after the initial assessment session (the trial therapy). As routine self-reported outcome measures were only implemented part way through this 8 year period, only a portion of the entire treated sample could be included in the effectiveness evaluation. Hence, a further criteria for inclusion in this study was that the patient had a baseline measurement of self-reported symptom distress (see below). The procedure described below is illustrated in the CONSORT flowchart in Fig. 1. All participants in the study were referred by professionals from various specialties, including emergency department, family practice offices, medical-surgery and mental health. Patients referred to the Centre were placed on a wait-list for a trial of therapy, i.e., an initial assessment session of ISTDP. During trial therapy a psychodiagnostic evaluation was conducted which included an assessment of type and degree of treatment resistance according to specified criteria. In addition, baseline assessments of self-rated symptom distress and interpersonal problems (regarding the previous two week period) were collected at the end of the trial session, at the end of the session following the trial, and at treatment termination. Treatment was not time-limited. Rather, termination was determined by response to treatment and agreed upon by patient and therapist.

Results

Associations of unlocking the unconscious with outcome (Conditional growth models)

Control predictors

Type of treatment resistance, therapist experience, patient gender, patient age, and the logarithm of number of treatment sessions were investigated as potential control predictors of outcome. Therapist experience was investigated both by comparing trainees to non-trainees, and by comparing the outcomes of the ISTDP expert to the other therapists. Descriptive analyses of the predictors at all time points revealed that there were BSI differences at baseline with fragile and psychotic patients having larger symptom severity than psychoneurotic patients (independent t-test, t(406) = 4.3, p < .001), with a similar trend on the IIP at baseline (independent t(384) = 1.9, p = .06) and at post-trial (independent t(221) = 1.8, p = .08). Furthermore, on IIP there was a difference at termination depending on therapist experience, with the 43 trainees having worse outcome than the six non-trainees (independent t(142) = 2.3, p < .05) and the expert having better outcome than the other 48 therapists (independent t(142) = 2.6, p < .05). The ANOVA investigating differences due to number of sessions (log) revealed a difference on the IIP at post-trial (F(3, 204) = 3.26, p < 0.05). An inspection of the post-trial IIP raw scores indicated that participants who only took part of the trial therapy and the follow-up meeting had lower IIP scores at this time point than those who continued with further therapy. This difference was significant (independent t(222) = 2.46, p < .05). No other differences due to number of sessions were found at any other time point on neither the BSI nor the IIP (all F’s < 1.98, all p’s > 0.11). No gender differences at any time point were found (all t’s < 1.5, all p’s > .14). Similarly, no differences due to age were found (all F’s < 1.82, all p’s > .14). Based on these findings, we decided to include type of resistance, trainee status, expertise and number of sessions (log) as control predictors.

BSI

In an uncontrolled conditional growth model, the variable that measured whether the patient had a major unlocking of the unconscious during therapy (called UNLOCKING in this section) was added at Level 2 as a time-invariant predictor in interaction with the initial status (intercept) and in interaction with time. The latter interaction term was added to determine if, over time, UNLOCKING significantly interacted with the BSI slope, i.e., predicted symptom change during treatment. At this stage of the analysis, the UNLOCKING by time was close to significant (p = .08) in the model.

The next step carried out was to determine if this association between UNLOCKING and BSI change was still present after accounting for type of treatment resistance, trainee status, expertise and the logarithm of number of treatment sessions (called RESISTANCE, TRAINEE, EXPERT and log(SESSIONS) in this section). Hence, a controlled conditional growth model was carried out. RESISTANCE, TRAINEE, EXPERT and log(SESSIONS) were added as time-invariant predictors in interaction terms with baseline BSI (the intercept) and BSI change (slope). The RESISTANCE and EXPERT by intercept and time were significant, and therefore kept in the model. However, TRAINEE and log(SESSIONS) were not significant neither in interaction with the intercept nor with the slope and therefore left out of the model. Finally, we added information on missingness from the post-trial and post-treatment assessments in interaction with the intercept and with the slope to investigate if data missingness was related to outcome (Little & Wang, 1996; Hedeker & Gibbons, 1997). Neither of these interactions were significant and therefore information on missing data was removed from the model. In the final model (details in Table 2), the UNLOCKING by time interaction was significant (p < .05). Estimated means from this model are illustrated in Fig. 2. The equations for the final BSI model were as follows:

Level-1 Model: ${\displaystyle Y_{ij}={\pi {}_0}_i+{\pi {}_1}_i\times {TIME}_{ij}+{\epsilon }_{ij}}$

Level-2 Model: ${\displaystyle \mathrm{Intercept}:{\pi {}_0}_i={\gamma }_{00}+{\gamma }_{01}\times {\mathit{UNLOCKING}}_i+{\gamma }_{02}\times {\mathit{RESISTANCE}}_i+{\gamma }_{03}\times {\mathit{EXPERT}}_i+{\zeta {}_0}_i\mathrm{Slope}:{\pi {}_1}_i={\gamma }_{10}+{\gamma }_{11\hspace{0.17em}}\times {\mathit{UNLOCKING}}_i+{\gamma }_{12}\times {\mathit{RESISTANCE}}_i+{\gamma }_{13}\times {\mathit{EXPERT}}_i+{\zeta {}_1}_i{\epsilon }_{ij}\sim N\left(0,{\sigma }_{\epsilon }^2\right),{\zeta {}_0}_i\sim N\left(0,{\sigma }_0^2\right),{\zeta {}_1}_i\sim N\left(0,{\sigma }_1^2\right).}$

Table 2:

Growth models estimating trajectories of change in Brief Symptom Inventory (BSI) and Inventory of Interpersonal Problems (IIP) from baseline to treatment termination.

			BSI		IIP
		Parameter	Unconditional growth model	Final controlled growth model	Unconditional growth model	Final controlled growth model
Fixed effects
Initial status, π0i	Intercept	γ 00	1.58***	1.47***	1.50***	1.45***
	Unlocking	γ 01		0.04		0.03
	Resistance	γ 02		0.37***		0.15*
	Expertise	γ 03		−0.04
Rate of change, π1i	Intercept	γ 10	−0.70***	−0.46***	−0.59***	−0.50***
	Unlocking	γ 11		−0.21*		−0.20*
	Resistance	γ 12		−0.22*
	Expertise	γ 13		−0.22*
Variance components
Level 1	Within-person	${\sigma }_{\epsilon }^2$	0.16***	0.15***	0.13***	0.13***
Level 2	In initial status	${\sigma }_0^2$	0.42***	0.40***	0.31***	0.31***
	In rate of change	${\sigma }_1^2$	0.17**	0.15**	0.14**	0.13*
	Covariance	σ 01	−0.10*	−0.08*	−0.07tr	−0.06

DOI: 10.7717/peerj.548/table-2

Notes:

tr

trend

* p < .05.

** p < .01.

*** p < .001.

IIP

Initially UNLOCKING was added as a predictor and was close to significant (p = .07) in interaction with time. RESISTANCE was added and was significant in interaction with the intercept, but not with the slope. Hence, the RESISTANCE times slope interaction was left out of the final model. Neither TRAINEE, EXPERT nor log(SESSIONS) were significant in interaction with the intercept or the slope, and therefore not included in the model. In the final model, the UNLOCKING by time interaction was significant (p < .05). As in the BSI model, information on missingness was not significant and was therefore not included. Results from the IIP model are illustrated in Fig. 2. Details of the final model can be found in Table 2 and in the equations below:

Level-1 Model: ${\displaystyle Y_{ij}={\pi {}_0}_i+{\pi {}_1}_i\times {\mathit{TIME}}_{ij}+{\epsilon }_{ij}}$

Level-2 Model: ${\displaystyle \mathrm{Intercept}:{\pi {}_0}_i={\gamma }_{00}+{\gamma }_{01\hspace{0.17em}}\times {\mathit{UNLOCKING}}_i+{\gamma }_{02\hspace{0.17em}}\times {\mathit{RESISTANCE}}_i+{\zeta {}_0}_i}$ ${\displaystyle \mathrm{Slope}:{\pi {}_1}_i={\gamma }_{10}+{\gamma }_{11\hspace{0.17em}}\times {\mathit{UNLOCKING}}_i+{\zeta {}_1}_i{\epsilon }_{ij}\sim N\left(0,{\sigma }_{\epsilon }^2\right),{\zeta {}_0}_i\sim N\left(0,{\sigma }_0^2\right),{\zeta {}_1}_i\sim N\left(0,{\sigma }_1^2\right).}$

Discussion

In this paper, we have established the effectiveness of Davanloo’s Intensive Short-Term Dynamic Psychotherapy when delivered by a range of therapists in a clinically representative setting to a large number of patients. Importantly, the present study was carried out in a tertiary care clinic, which is assumed to have implied a more treatment-resistant patient sample than those present in standard clinics. This study adds to the overall evidence base of ISTDP in particular, but also to that of psychodynamic psychotherapy in general. In a recent meta-analysis on randomized controlled trials of psychodynamic psychotherapy (Town et al., 2012a), the overall within-group effect size was estimated to Cohen’s d = 1.01 95% CI [0.86–1.16]. Furthermore, a study evaluating the overall effectiveness of psychotherapy for depression among 5,704 clinical patients in a managed care setting (Minami et al., 2008) estimated the effect size to be d = 0.75 95% CI [0.72–0.77]. Hence, the effects observed in this study (effect sizes Cohen’s d = 0.87 (95% CI [0.71–1.03] and d = 0.83 95% CI [0.64–1.02] for the BSI and the IIP, respectively), seem to be comparable both to effects of psychodynamic psychotherapy when provided in RCTs and to that of psychotherapy in general when provided in clinical practice.

In this study, we also showed that a major unlocking of the unconscious during therapy predicted better treatment outcome. This result validates a fundamental assertion in Davanloo’s ISTDP and is to our knowledge the first study to systematically reproduce those assertions in a large-scale patient sample (Town, Abbass & Bernier, 2013). Moreover, this finding is in line with research suggesting that affective experience during psychodynamic psychotherapy may enhance treatment effectiveness (Whelton, 2004; Diener, Hilsenroth & Weinberger, 2007; Diener & Hilsenroth, 2009; Salvadori, 2010). Importantly, strong emotional breakthroughs in ISTDP are assumed to be dependent on the therapist’s systematic challenge to resistance in session. Hence, our findings in this paper are likely to parallel recent research that have highlighted a relationship between affective arousal and preceding active therapist confrontation (Town et al., 2012b).

In the investigation of the associations of unlocking with outcome, we included a number of control predictors including type of treatment resistance. The association of type of treatment resistance with outcome can be interpreted as a separate effect, after controlling for the main predictor of interest (unlocking the unconscious). Patients classified as fragile or psychotic had more symptom severity at baseline than those who were psychoneurotic. This is in line with psychodynamic theory suggesting that individuals with access to primitive defenses (fragile or psychotic) have been exposed to more adverse and/or persistent childhood experiences: this on its own is known to be associated with higher symptom severity in later life (Felitti et al., 1998; Edwards et al., 2003). Furthermore, we found that fragile patients had (compared to psychoneurotic patients) steeper rate of change on the BSI, while there was no such difference in slope on the IIP. This suggests that, to understand the larger degree of change in reduced symptom distress seen in fragile patients, it may be necessary to look beyond the concept of achieving a major unlocking of the unconscious to other therapeutic variables. As described above, ISTDP for fragile patients typically involves building capacity for tolerating emotional experiences (Davanloo, 1990; Whittemore, 1996; Abbass & Bechard, 2007). Hence, results from this study could indicate that this capacity-building format may be a key intervention when working with patients with primitive defenses, character pathology and with a high symptom burden. Furthermore, we replicated previous findings showing that the rate of change on the IIP is associated with unlocking the unconscious, but not with any other predictor (Town, Abbass & Bernier, 2013). Given the fact that change on the IIP may reflect character change rather than symptom change, the typical treatment course in this study (median 5 sessions, 10.2 on average) may not have been long enough to achieve major changes on the IIP. Thus, if treatment courses had been longer, we could possibly have expected a larger variation among IIP slopes, which then potentially could be explained by other predictors than unlocking the unconscious. This could parallel the claim by Davanloo that for major character change to take place, 20–40 sessions of ISTDP may be needed for psychoneurotic patients and 60–80 for those with fragile character structure (Davanloo, 2005).

Whether experienced therapists perform better than less experienced is a debated topic in the psychotherapy research community, with limited literature suggesting that more training enhance treatment effectiveness (Hattie, Sharpley & Rogers, 1984; Wampold & Brown, 2005). This is in line with our findings that the group of professional therapists (vs. trainees) with far more training hours did not seem to conduct more effective treatments. It also parallels previous research on the effectiveness of ISTDP when carried out by psychiatry residents that indicated reductions in both symptoms and health care costs (Abbass, 2004; Abbass et al., 2013). However, we found that a single therapist, considered an expert in ISTDP, had better outcomes than the other therapists, on one of the outcome measures. This is indeed interesting as expertise in psychotherapy is largely unexplored (Tracey et al., 2014). Further research on expert performance in ISTDP is warranted based on these findings. To extend the current findings, this may include examination of therapists’ competence as measured by the quality of the unlocking of the unconscious.

Strengths of this study include that the treatment was tested for a large-scale sample of 412 patients, that therapists were trained and supervised by a highly experienced ISTDP teacher, and that data analyses were conducted using advanced statistical methods capable of making use of all available data. A further strength is that several elements contribute to the generalizability of the findings from this study: the treatment was provided in a clinically representative setting by a large number of therapists with different levels of experience, all patients were recruited by referral from professionals rather than advertisement, and very broad selection criteria were applied that allowed a mixed sample of patients to be included.

However, there are limitations of this research that need to be addressed. First, a substantial number of patients did not have a post-treatment assessment (missing because of dropout or other reasons, 39.1% on the BSI, 48.1% on the IIP). We addressed this limitation by using data-analytic procedures that have been shown to be valid under the assumption that the data was missing at random (MAR). This allows the probability of data being missing to be dependent upon both outcome variables (e.g., symptom severity) and predictors (Little & Rubin, 2002). Furthermore, we conducted additional pattern mixture models to take data missingness into account (Little & Wang, 1996; Hedeker & Gibbons, 1997). Despite the use of these data-analytic procedures, we cannot exclude the possibility that the results from this study have been affected by the amount of missing data. The actual rate of dropout in this study was difficult to determine also because many patients came from far away and only attended a few sessions for travel reasons. Importantly though, meta-analytic estimations of the number of dropouts have found similar rates as in this study. An investigation of 125 psychotherapy outcome studies estimated the mean dropout rate to 46.9% (Wierzbicki & Pekarik, 1993). More recently, a meta-analysis of 34 non-randomized effectiveness studies on outpatient cognitive behavioral therapy for depression found that 42.0% of patients who received individual therapy dropped out (Hans & Hiller, 2013). Our rates of missing data and/or dropout also parallel a recent large-scale effectiveness evaluation of psychodynamic psychotherapy for depression in which 45.2% of the participants dropped out or were lost to assessment (Driessen et al., 2013). A second limitation in this study was the fact that the ratings of rise in the complex transference feelings and degree and type of treatment resistance were conducted using non-validated instruments. Measurements of these constructs in this study were carried out by a highly experienced ISTDP clinician and teacher. We have because of the experience of the coder assumed adequate validity and reliability of these measurements. However, this procedure might have caused a measurement bias due to the theoretical orientation of the ISTDP teacher. In conclusion, this limitation highlights the need to develop psychometrically sound and reproducible procedures to measure the constructs proposed by Davanloo (1990), Davanloo (2000), Davanloo (2001) and Davanloo (2005). Third, the results in this study on unlocking in relation to outcome are correlational, i.e., we have not established a causal effect. Hence, we cannot exclude a possible reversed causal effect, that symptom relief on its own resulted in emotional breakthroughs. While we consider this as unlikely due to the fact that a major unlocking is assumed to be the result of the breakdown of treatment resistance after systematic challenge, we acknowledge the need for future research that establish a causal relationship between outcome and in-session events in ISTDP. Fourth, data from self-report measures was collected only at three time points instead of weekly assessments. While a weekly collection might not have been possible to carry out within the current routine care setting, it would have contributed to a better estimate of the treatment effect over time. A related concern is that the research project did not record the exact times from baseline to post-trial and from post-trial to termination. Instead, estimated averages for these time periods times were used. We acknowledge this lack of a more precise coding of time as a limitation of the study. A final limitation is that the lack of control condition and the non-randomized effectiveness design of this study did not permit us to control for gains due to time passing (e.g., spontaneous gains and gains from other treatments received). Nor was there any extended baseline assessment phase that could have been used to estimate the effect of time before treatment started. Hence, we cannot exclude the possibility that the treatment effects observed in this study were due to other factors than the treatment received. However, this is unlikely given the short treatment duration and the likely chronicity of patient problems referred to this service.

Based on the findings from this study, further research on ISTDP is warranted. We see a continued need for systematic investigation of the efficacy of ISTDP for specific diagnostic groups using randomized controlled trials. Such research could also include assessments of in-session events such as rise in complex transference feelings, that would enable more detailed examinations of the relation between events such as the major unlocking of the unconscious and outcome. Furthermore, standardized ratings of rise in the complex transference feelings and type and degree of resistance should be developed for future research on ISTDP, to enable further knowledge of the mechanisms of change in ISTDP. A guide for such ratings is in development (JM Town & A Abbass, 2014, unpublished data).

Conclusions

This study shows that Davanloo’s Intensive Short-Term Dynamic Psychotherapy is an effective treatment when carried out in tertiary care, when delivered by therapists with a range of experience. Furthermore, this study provides support for the key assumption in Davanloo’s ISTDP in that unlocking the unconscious is associated with enhanced treatment outcome. Further research on the effectiveness of ISTDP and its working mechanisms is warranted.

Supplemental Information