Effect of a combined brief narrative exposure therapy with case management versus treatment as usual in primary care for patients with traumatic stress sequelae following intensive care medicine: study protocol for a multicenter randomized controlled trial (PICTURE)

Inclusion and exclusion criteria for patients

Inclusion and exclusion criteria for PCPs

Inclusion criteria for participating PCPs are:

(1) The physician must have been registered for at least 2 years in the German statutory health-care system as a primary care physician.

(2a) The physician must have a qualification in basic psychosomatic care (Psychosomatische Grundversorgung, Bundesärztekammer, 2001) [24] to ensure that they can provide a basic level of mental health care and to ensure patient safety.

(2b) Alternatively, the PCP must have been a family doctor within the German statutory health-care system for at least 5 years with evidence of adequate psychiatric education, e.g., additional training (this is to ensure that all participating PCPs have a minimum level of psychiatric knowledge).

(3) They have provided written informed consent.

PCPs with a specialization such that more than 80% of the patients registered with their practice have a specific mental condition are excluded from the trial to ensure that enrolled practices are representative of German primary care.

Experimental condition

After randomization, PCPs in the intervention group will receive training material (therapy manuals for PCPs and MAs, intervention videos, and a paper booklet), as well as face-to-face training by NET-qualified psychologists. In most cases, the training for PCPs will be on a one-to-one basis, though group training may also be arranged. Additionally, patients in the intervention group will receive written information about PTSD and trial procedures. Treatment in the intervention group consists of three NET sessions delivered by the PCP and case management delivered by the practice-based MA.

Furthermore, PCPs will receive training through written materials with information regarding study procedures as well as diagnostic examinations for and treatment of PTSD according to the German S3 Guideline on PTSD [10].

Three NET sessions of approximately 45 min each will be delivered by the PCP. The first session includes psychoeducation on PTSD and an overview of the patient’s biography. In this session, the patient learns about the symptoms and the theoretical background of PTSD as well as the treatment procedure. In addition, they will identify traumatic events in their biography by constructing a lifeline. In this procedure, the patient places stickers of flowers and stones, symbols of important positive and stressful events, in chronological order on a line that they construct and draw together with the PCP on a piece of paper. The line serves as a timeline, providing an overview of the patient’s biographical burden and resources. At the end of the session, the recent ICU event will be implemented in the lifeline.

In the second session, the patient will be exposed to the traumatic events in a safe environment by giving a detailed narration of their stressful experience during their ICU stay. After the second session, a qualified psychologist will provide telephone support to the PCP to review the treatment so far, prepare the third session, and give guidance and advice on the content and mode of delivery of the therapy, if necessary. Narration on another stressful key life event identified on the lifeline is recommended as the topic for the third session. Alternatively, an additional narration of the ICU event might be chosen, if no other stressful key event can be identified, or if the ICU event is still the major key event in the patient’s life.

The MAs receive written training material and personal training on case management, which is carried out by study nurses from the relevant regional trial centers. Case management consists of seven short telephone calls (approximately 15 min each), in which the MA asks about the patient’s well-being, completes a PTSD-monitoring checklist, and provides social acknowledgement of the experiences the patient had during their critical illness. MAs follow written instructions for structuring the dialog and gathering information. Included in the instructions is a color-coded rating system for the PTSD screening questionnaire results. Critical responses should prompt the MA to inform the attending PCP immediately. Questionnaire responses are noted in the booklet. Two telephone calls are made between the second and third NET sessions, followed by an additional five telephone calls between the last NET session and T1 (Fig. 1).

Control condition

Patients allocated to the control group will receive iTAU during at least three consultations with their PCP. Treatment will be based on current German recommendations for the diagnosis and treatment of PTSD [10] without further specifications from the study protocol. PCPs will receive written training material and detailed medical information about PTSD, based on the current national guideline for PTSD and on the application of good clinical practice in the conduct of clinical trials [26]. Because of this explicit training, we consider this treatment approach improved in comparison to treatment as usual without any additional training.

Primary efficacy outcome and endpoint

To evaluate the PTSD symptom severity of trial participants, the German version of the self-administered PDS-5 questionnaire will be applied. Each of the 20 items refers to PTSD-related symptoms experienced in the past month and is answered on a five-point Likert scale (from 0 not at all to 4 more than five times per week/severe) [23, 27].

The primary efficacy outcome is the PDS-5 total severity score ranging from 0 to 80 points. The pre-specified primary efficacy endpoint is the absolute change from baseline to the 6-month follow-up telephone assessment.

Secondary efficacy outcomes

Patient questionnaires to derive secondary efficacy outcomes were chosen based on the conceptual framework of core outcome sets [28, 29].

As depression and anxiety are known common comorbidities in patients with PTSD, trial participants are instructed to complete the primary care validated Patient Health Questionnaire 9 (PHQ-9), where each of the nine items is scored from 0 (not at all) to 3 (nearly every day), resulting in a total score of from 0 to 27 points (large scores indicate severe impairment) [30]. Anxiety will be measured by the brief Overall Anxiety Severity and Impairment Scale (OASIS) questionnaire, which contains five response options for each of the five items, coded from 0 to 4. The total score ranges from 0 (no anxiety) to 20 points, with a high score indicating severe impairment [31]. Disability will be assessed by the 12-item version of the World Health Organization Disability Assessment Schedule 2.0 (WHODAS 2.0), with a total score ranging from 0 to 100 and higher scores indicating higher levels of disability [32]. Patient activation will be measured by the 13-item Patient Activation Measure (PAM) [33], where each item scores from 1 to 4 (1 strongly disagree, 2 disagree, 3 agree, and 4 strongly agree; for the fifth item only, 5 is for not applicable). The evaluation for the latter is made by adding the raw values, which have a range of 13–52, and normalizing them to a scale of 0–100. We will use the version of the EuroQol questionnaire with five dimensions and five levels (EQ-5D-5L), which consists of the EQ descriptive system and the visual analogue scale (EQ-VAS) to measure health-related quality of life. The EQ-VAS is a thermometer-like rating scale ranging from 0 (worst imaginable health state) to 100 (best imaginable health state today) [34, 35]. Concomitant drug and non-drug therapies, and health service use will be assessed by means of a modified (shortened) German version of the Client Sociographic and Service receipt Inventory (CSSRI) [36].

For all these scores derived from the questionnaires mentioned above, the treatment effect will be assessed by means of the absolute change from baseline at months 6 and 12 for secondary efficacy outcomes. The detailed schedule of enrolment, interventions, and assessments with their pre-planned time points is shown in Fig. 2.

Health economic evaluation

The objective of the health economic evaluation alongside the main trial is to assess the cost-effectiveness of the NET-oriented intervention in comparison to iTAU from a societal perspective [37]. We will consider health-care costs as well as productivity losses to describe the monetary consequences of the intervention and calculate quality-adjusted life years (QALYs) as a measure of the effects. These results will inform decision makers in the health-care sector of the economic aspects of the NET intervention and support them in deciding whether the intervention should be implemented in the German health-care system.

Genetic evaluation

As the first side project of the trial, we also plan to investigate the genetic distinctiveness of patients after intensive care medicine. We will consider differences in the genetic characteristics of ICU patients with PTSD compared to those of ICU patients without PTSD (not participants of the PICTURE trial). The genetic evaluation will be carried out in collaboration with the Institute of Psychiatric Phenomics and Genomics, University Hospital Munich. This project has a separate trial protocol, which has the approval of the ethics committee, and it requires separate written informed consent.

Process evaluation

A second side project will explore the experience of the major actors (PCPs and MAs) and patients recruited at ICUs in Berlin, Hamburg, and Dresden with the experimental intervention in the NET group. This project aims to analyze the applicability of the experimental intervention. That is, it will investigate the beneficial and obstructive factors in the effectiveness, acceptance, and feasibility of the intervention, based on the theoretical framework of acceptability [38]. Qualitative interviews will be carried out with PCPs, MAs, and patients after assessing the primary endpoint at T1 for the last patient randomized.

Power considerations and sample size calculation

The current literature does not provide a minimal clinically important difference for the primary outcome (PDS-5 total score for DSM-5), on which we could base the sample size calculation [39]. Therefore, we will use a calibration argument to provide a rather pragmatic minimal clinically important difference for this trial. Previous NET studies defined a decline of about 25% in the baseline score as a clinically relevant change [16]. Based on the range of the PDS-5 score from 0 to 80 points, we define 40 as the mean baseline score. Thus, a 25% change from baseline gives 10 points as the mean absolute decrease. To be more conservative (also assuming a slight effect of 4 points within the control group), we consider a difference in absolute change of 6 points between both groups as clinically relevant for these post-ICU patients. Using a standard deviation of 17, this translates into a Cohen’s d (standardized effect size) of 0.36. This effect is assumed to be conservative compared to the reported effects for NET [13, 40]. It can be translated into a probability of 0.6 that the observed decrease in the experimental group is larger than that in the control group (assuming a standard normal distribution). The probability of 0.6 is the target parameter needed to perform a sample size calculation with the Wilcoxon–Mann–Whitney rank-sum test. A sample size of 131 patients in each group, i.e. 262 patients in total, will have 80% power to detect a decrease of PDS-5 in the intervention group as described above compared to the control group using a Wilcoxon–Mann–Whitney rank-sum test with a 0.05 two-sided significance level (software used: nQuery Advisor 7.0). To incorporate the death of patients (resulting in the efficacy outcome being truncated due to death), we apply a non-parametric worst rank score analysis ([41, 42], for details see below) and decided to randomize an additional 78 (= 2 × 39) patients [an increase of about 30% (= 39 / 131) derived from a simulation study]. Thus, the sample size to be allocated to the trial is 2 × (131 + 39) = 340 patients in total.

It is anticipated that 3000 patients can be pre-screened, of which 1000 (33%) are expected to show posttraumatic stress [5]. The rate of non-participation is expected to be about 35%, which is a conservative assumption compared to our previous study (20% non-participants in [18]). Therefore, 650 (65%) are expected to be willing to participate (patients and their PCPs). Of these, 550 patients (about 85%) could be screened by their treating PCP 3 months post-ICU discharge (assuming a mortality rate after 6 months of about 15% as in [18]), 400 (about 70%) could meet the inclusion criteria, and 340 (85%) patients (and their PCPs) could consent to participate in the study at the baseline visit. We assume a 30% drop-out rate from baseline over the 6 months before the primary endpoint is assessed. There is no pre-planned interim efficacy analysis and no sample size recalculation.

Statistical analyses for primary and secondary endpoints

The primary efficacy endpoint is the absolute change in the PDS total severity score from baseline at month 6: ΔPDS = PDS(T1) – PDS(T0).

By default, the mode of administration is a self-administered paper-based version. For patients who do not complete and send back the paper-based patient questionnaire (non-responding survivors), the PDS-5 total score will be assessed during the telephone surveys scheduled 6 months (T1) and 12 months (T2) after randomization.

It is assumed that death is the most likely cause of missingness. Therefore, a composite endpoint approach will be applied, combining information about the change in PDS total score and mortality into a single variable [41].

The null hypothesis, G_NET(x) = G_iTAU(x) and K_NET(t) = K_iTAU(t) (0 < t ≤ T, date of death), implies that treatment groups NET and iTAU will not differ with respect to the distributions of the observed outcome measure ΔPDS. Here, G(x) is the cumulative probability distribution of the observed change in PDS severity scores at T1 in groups NET or iTAU, and the distribution K(t) of the date of death is the cumulative distribution of informative event times for the compared groups.

The null hypothesis will be tested by a non-parametric approach using a modified version of the Wilcoxon–Mann–Whitney U test, which basically allocates the tied worst ranks to all missing values (the worst rank score analysis was proposed by Lachin [42]). The null hypothesis can be rejected if the two-sided p value related to the test statistic for the treatment effect is equal to or smaller than the significance level α = 0.05. This test strategy is tailored to a particular alternative hypothesis, i.e., (i) NET will either be superior to iTAU in terms of ΔPDS, but with no impact on survival, (ii) NET will be superior to iTAU in terms of survival, but with no impact on ΔPDS, or (iii) NET will be superior to iTAU for both ΔPDS and survival.

If the PDS total severity scores are not informative for future death events, the worst rank replacement will simply lead to a power loss and no inflation in the type I error rate. Should the PDS total severity scores be informative for future death events, the worst rank replacement will result in an unbiased test of a particular alternative [42].

The principal analysis will be performed according to the ITT principle, and not adjusted for screening or baseline covariates or site. The significance level is set to alpha = 5% (two-sided).

Missing data prior to the follow-up measurement will occur because of an informative disease-related event (e.g., death or morbidity) or for other reasons (e.g., non-responders at follow-up measurements at T1 and T2, loss to follow-up, or consent withdrawn). To address the impact of several missingness mechanisms (missing at random or missing not at random), sensitivity analyses will be performed: mixed effect models assuming missing at random using the whole observed longitudinal PDS profile of the surviving patient; multiple imputations techniques; or even complete case analyses using the analysis of covariance (absolute change score as the response variable and treatment group as the covariate, adjusting for the baseline score value) for responding survivors until T1.

Moreover, sensitivity analyses will be performed in the per-protocol population using linear mixed effects models to explore the role of covariates (e.g., patient age and gender).

The full statistical analysis plan will be finalized and revised in a blinded manner ahead of the database lock after the last patient’s last 12-month telephone call.

Definition of analysis data sets

Each trial participant’s allocation to the different analysis populations (full analysis data set according to the ITT principle, per-protocol analysis data set, and safety analysis data set) will be defined and explained in the statistical analysis plan, which will be finalized before the analysis. During the data review, deviations from the protocol will be assessed as minor or major. Major deviations from the protocol will lead to the exclusion of a participant from the per-protocol analysis data set. The full analysis data set according to the ITT principle will consider all randomized patients with at least one study-related visit at a physician’s office during the intervention period (for the NET group, at least one NET session and for the iTAU group, at least one face-to-face consultation). Furthermore, patients who die before the evaluation of efficacy outcomes (truncation due to death) are part of the principal analysis incorporating the timing of death of the trial participant.

Data collection and transmission

All data collected during the trial will be documented using electronic case report forms (eCRFs). The source data will be stored regionally in the patients’ files. Clinical and patient-reported outcome data will be collected by the ITC in Munich at the site of the PI via self-administered questionnaires and via phone interviews at T1 and T2. ITC staff are blinded to the assigned treatment given to the interviewed patient.

Monitoring

An independent clinical monitor will check for accuracy, completeness, consistency, and reliability of the eCRFs by comparing documented data with source data. The monitor will check that data are collected, stored, and managed appropriately at all trial sites. Additionally, the monitor will check SAE documentation and status as well as documentation and follow-up of protocol deviations. Monitoring visits will be carried out regularly according to the standard operating procedures at each trial site independently, to ensure the trial procedure is executed according to good clinical practice [26].