Efficacy of traumatic memory reactivation with or without propranolol in PTSD with high dissociative experiences

ABSTRACT

Background: Post-traumatic stress disorder (PTSD) with dissociative symptoms is now a full-fledged subtype of this disorder. The dissociative subtype is associated with a greater number of psychiatric comorbidities. To date, the impact of dissociation on the efficacy of PTSD treatment remains unclear.

Objective: The aim of this study was to compare the efficacy of a traumatic memory reactivation procedure with the administration of propranolol or a placebo once a week for six consecutive weeks in reducing PTSD and MDE symptoms between PTSD subjects with or without high dissociative symptoms.

Method: For that, we conducted a randomized clinical trial in 66 adults diagnosed with longstanding PTSD and measured the SCID PTSD module, the PTSD Checklist (PCL-S), Beck’s Depression Inventory-II (BDI-II), and the Dissociative Experiences Scale (DES).

Results: Patients with and without high dissociative experience had significant improvement in their PCL-S scores over the 6 treatment sessions, and PCL-S scores continued to decline in all patients during the post-treatment period. However, there was no correlation between the presence/absence of high dissociative experiences and no specific effect of propranolol treatment. We found exactly the same results for MDE symptoms. Interestingly, patients with high dissociative experiences before treatment exhibited very significant improvement in their DES scores after the 6 treatment sessions, and patients maintained this improvement 3 months post-treatment.

Conclusions: The traumatic memory reactivation procedure is an effective way to treat dissociative symptoms in patients with PTSD, and improvement of these dissociative symptoms was associated with a decrease in both PTSD and depression severity.

HIGHLIGHTS

• Traumatic memory reactivation procedure is an effective way to treat dissociative symptoms in patients with PTSD.

• The improvement of these dissociative symptoms was associated to a decrease of both PTSD and depression severity.

• Dissociative symptoms do not seem to mitigate the efficacy of traumatic memory reactivation during the treatment of PTSD.

1. Introduction

Dissociation is defined as an abnormal integration of consciousness, memory, identity, body representation, emotion, perception, motor control, and behaviour (Diagnostic and Statistical Manual of Mental Disorders 2013). This phenomenon encompasses a wide range of subjective experiences, some relatively common, such as daydreaming, and some related to more severe mental states, such as not remembering autobiographic memories (Bernstein & Putnam, 1986). Historically, Charcot and Janet linked dissociation and trauma in the late nineteenth century (Yrondi et al., 2019), and it is now known that post-traumatic stress disorder (PTSD) is among the psychiatric disorders that cause the most dissociative symptoms (Lyssenko et al., 2018).

This link was tightened in the last version of the DSM with the addition of the ‘with dissociative symptoms’ subtype of PTSD, which was justified based on different neural and psychophysiological responses to the recall of traumatic memories (van Huijstee & Vermetten, 2018). Some studies have reported a high frequency of this subtype, up to 80% when systematically searched (Hill et al., 2020). Moreover, patients with the dissociative symptom subtype of PTSD have more severe PTSD symptoms (intrusion, avoidance, and arousal) (Haagen et al., 2018). The dissociative subtype is also associated with a greater number of comorbidities, such as substance use and depressive and anxiety disorders (Kim et al., 2019; Tsai et al., 2015). Moreover, previous studies have highlighted the positive correlation between the severity of depressive symptoms and dissociation (Bob et al., 2010). In addition, it is well know that patients with both PTSD and major depressive disorder experience more frequent and higher dissociative symptoms (Bedard-Gilligan et al., 2015; Contractor et al., 2015; Salloum et al., 2018). It remains unclear to date whether dissociation negatively influences the treatment of PTSD. The question is still being debated in the literature, even though previous data have demonstrated that trauma-focused (Zoet et al., 2018) and cognitive processing therapy (Resick et al., 2012) were effective in diminishing PTSD and dissociative symptoms. On the other hand, narrative storytelling (NST) has shown short-term efficacy in the treatment of the dissociative subtype of PTSD. The improvement did not last over time since, at 40 weeks, NST was not associated with an affective and interpersonal regulation procedure (Cloitre et al., 2012).

For PTSD treatment, the beta-blocking compound, propranolol, has been shown to interfere with memory reconsolidation, suggesting its potential use as a pharmacological treatment in conditions involving increased memory intrusions such as PTSD. The noradrenergic system is critical in modulating memory processes, and β-noradrenergic receptor stimulation has been found to facilitate emotional memory reconsolidation. Moreover, a hyper-noradrenergic state is implicated in the pathophysiology of PTSD.

In a new therapeutic procedure (Brunet et al., 2018; Roullet et al., 2021), PTSD patients required to actively recalled their traumatic event under the influence of propranolol once a week for up to 6 weeks. In a previous study (Roullet et al., 2021), we showed that this traumatic memory reactivation, whether it was associated with a placebo or propranolol, was an effective treatment of PTSD. Moreover, despite the lack of differences between the propranolol and placebo groups at the end of the procedure, during follow-up at three months, patients in the propranolol group maintained the improvement more significantly than those in the placebo group. However, we do not know the potential impact of dissociative symptoms. Thus, the aim of this ancillary study was to compare the efficacy of traumatic memory reactivation over PTSD and depressive and dissociative symptoms between a highly dissociated group of patients and a lower one.

2. Methods

Details of the procedure and the main results for this clinical trial are presented elsewhere (Roullet et al., 2021).

2.1. Subjects and study design

Participants were enrolled from January 2013 to October 2017 from three different university hospitals (Toulouse, Lille, Tours) in France. All participants were outpatients recruited from specialized trauma clinics (Psychotrauma regional centre). To be included, adults aged 18–65 years needed to have a primary diagnosis of PTSD according to the DSM-IV criteria (Frances et al., 2000) with the Structured Clinical Interview for DSM (SCID) (First et al., 2007) and suffering from PTSD for at least three consecutive months with a PCL-S score ≥ 45 at the first session (W1). The exclusion criteria were: contraindication to propranolol (hypotension, higher than a first-degree heart block, bronchial asthma etc.); medication recommended or suggested for PTSD treatment; psychotherapy; basal systolic blood pressure < 100 mm Hg; basal heart rate < 50 bpm; psychotic or bipolar disorders; traumatic brain injury; current substance or alcohol dependence; acute suicidal ideation; pregnancy and breast-feeding.

Participants received detailed information about the study and gave their written informed consent to participate. All procedures were reviewed and approved by institutional review boards associated with the lead research team and each treatment site. Following inclusion, participants were assigned by randomized complete block to either the ‘traumatic memory reactivation + propranolol’ group, hereafter referred to as the propranolol group, or the ‘traumatic memory reactivation + placebo’ group, referred to as the placebo group.

Propranolol or a placebo was administered 90 min before the memory reactivation session. This reactivation was performed in the centre once a week for six consecutive weeks (W1 to W6). For the first session (W1), 90 min after ingesting the propranolol or a placebo, participants wrote a one-page trauma narrative in the present tense, first person singular, focusing on the event’s most disturbing moments. This task took up to 30 min. The therapist asked the patient to add some emotional details, if possible. Prior to the second session, a ‘traumatic script’ was prepared. In subsequent treatment sessions (W2 to W6), 90 min after receiving the medication, participants read their trauma script. Participants did not receive any additional intervention. The post-treatment effect was assessed one week (W7) and three months (W18) after the end of the procedure (for details of the procedure, see Roullet et al., 2021).

2.2. Assessment

PTSD diagnosis was assessed according to the DSM-IV criteria (Frances et al., 2000) with the Structured Clinical Interview for DSM (SCID) (First et al., 2007), MDE, panic disorder, social phobia, obsessive-compulsive disorder (OCD), alcohol dependence/abuse, psychotic symptoms, and general anxiety disorder (GAD) were assessed according to the DSM-IV criteria (Frances et al., 2000) with the Mini-International Neuropsychiatric Interview (M.I.N.I.5.0.0) (Lecrubier et al., 1997). The SCID PTSD module was performed before treatment and one week and three months post treatment. The severity of PTSD and depressive symptoms were assessed with the PTSD Checklist–Specific (PCL-S) total score (Blanchard et al., 1996) and the Beck’s Depression Inventory (BDI-II) total score (Beck et al., 1996), respectively. Dissociative experiences were assessed with the Dissociative Experiences Scale (DES) (Bernstein & Putnam, 1986), and we used the cut-off of 30 to define patients without (DES < 30) or with (DES ≥ 30) high dissociative experiences (Carlson et al., 1993; Simeon et al., 1998). PCL, DES and BDI scales were performed before treatment, during each treatment sessions and one week and three months post treatment. The instructions were modified for PTSD, dissociative and depressive symptoms to inquire about ‘current’ symptoms.

2.3. Statistical analysis

Sociodemographic and clinical characteristics were presented using means and standard deviations for continuous variables and frequency distributions for categorical variables. For comparisons between groups, we used chi-square tests for categorical variables and t-tests for continuous variables. Linear regression models (enter method) were applied to test the association between the dependent variable (intensity of depressive symptoms: BDI) and the independent variables: dissociation experience (DES) and the intensity of post-traumatic symptoms (PCL). Linear regression models were adjusted based on age and gender. To analyze the evolution of the DES, PCL-S, and BDI scores during the different sessions, we used a repeated factor Anova.

As indicated in the introduction, in the original experiment (Roullet et al., 2021), PTSD patients significantly improved their PCL-S scores after the 6 treatment sessions, but PCL-S scores decreased to a similar extent in patients treated with propranolol and a placebo. For this reason and as there was no difference in the DES scores in these two groups of patients, we decided not to consider this factor in the analyses.

3. Results

3.1. Sociodemographic and clinical characteristic

Sixty-six participants were assessed, but three of them did not respond correctly to the DES and thus were excluded from the analysis. The mean age was 39.2 years old (95% CI: 35.7–42.1). Thirty-nine (61.9%) participants were female.

The kind of traumatic experiences that caused the PTSD was not different in the two group of patients (p = .147). We can note that, for sexual trauma, seven patients (7/12) had this trauma during childhood, two patients (2/5) in the ‘without high dissociative experience’ group and five patients (5/7) in the ‘with high dissociative experience’ group.

Before the first treatment session, the mean PCL, BDI, and DES scores were 63.8 (95% CI: 61.8–66.4), 30.6 (95% CI: 27.9–34.1), and 25.3 (95% CI: 21.4–29.5), respectively (Table 1). Before the first treatment session, high dissociative experiences (DES score ≥ 30) were recorded in 38.1% (24/63) of participants.

Table 1.

Demographic and clinical characteristics.

	Total	Without high dissociative experience	With high dissociative experience
		DES < 30	DES ≥ 30
	(n = 63)	(n = 39)	(n = 24)	p-valuea
Age: (years) Mean (95% CI)	39.2 (35.7–42.1)	39.6 (35.4–43.8)	38.5 (33.1–43.9)	.757
Gender: No. (%) of women	39 (61.9)	25 (64.1)	14 (58.3)	.647
Formal education No. (%)
Primary and secondary	1 (1.6)	1 (2.6)	0 (0)
High school	35 (55.6)	20 (51.3)	15 (62.5)
University	27 (42.8)	18 (46.1)	9 (37.5)	.546
Marital status: Married No. (%)	32 (50.8)	20 (51.2)	12 (50.0)	.921
Unemployed/Retired/Disabled: No (%)	13 (20.6)	9 (23.1)	4 (16.6)	.541
Index trauma: No (%)
Motor vehicle accident, work accident	19 (30.2)	12 (30.1)	7 (29.2)
Physical assault	21 (33.3)	16 (41.0)	5 (20.9)
Sexual trauma	12 (19.1)	5 (12.8)	7 (29.2)
Combat, war zone, captivity	5 (7.9)	1 (2.6)	4 (16.7)
Life threatening event	1 (1.6)	1 (2.6)	0 (0)
Sudden unexpected death	3 (4.8)	2 (5.1)	1 (4.2)
Other	2 (3.2)	2 (5.1)	0 (0)	.147
Patients treated with propranolol: No. (%)	32 (50.8)	19 (48.7)	13 (54.1)	.674
DES score at W1: Mean (95% CI)	25.3 (21.4–29.5)	15.5 (13.1–17.9)	41.2 (36.4–46.1)	<.001
PCL-S score at W1: Mean (95% CI)	63.8 (61.8–66.4)	61.2 (58.1–64.3)	68.0 (64.8–71.2)	.006
BDI score at W1: Mean (95% CI)	30.6 (27.9–34.1)	25.9 (22.2–29.7)	38.2 (34.2–42.2)	<.001
MDE at W1: No. (%)	35 (53.3)	16 (41.0)	19 (79.1)	.003
Comorbidities at W1
Anxiety disorders: No (%)
Social phobia	13 (20.6)	9 (23.1)	4 (16.6)	.706
OCD	9 (14.3)	7 (17.9)	2 (8.3)	.927
GAD	18 (28.6)	10 (25.6)	8 (33.3)	.373
Panic disorder	5 (7.9)	2 (5.1)	3 (12.5)	.293
Substance abuse: No (%)
Alcohol	5 (7.9)	3 (7.7)	2 (8.3)	.927
Cannabis	2 (3.2)	2 (5.1)	0 (0)	.260

Note. The data include the No. of events (%) or mean (95% CI). ^a Group comparisons calculated with Chi-square, Fisher’s exact test, or the independent t-test. M = mean; SD = standard deviation; No. = number; OCD = obsessive-compulsive disorder; GAD = general anxiety disorder; MDE = major depressive episode.

We did not measure any difference in the sociodemographic characteristics between participants with or without high dissociative experiences. However, participants with high dissociative experiences had higher PCL-S and BDI scores before the beginning of the treatment procedure than participants without high dissociative experiences (PCL-S p = .006; BDI p < .001). This group of participants also had a higher rate of MDE (p = .003).

We must note that in our study, only 14.3% (9/63) of the participants dropped out at the end of the treatment procedure [without high dissociative experience 15.4% (6/39); with high dissociative experience 12.5% (3/24)] and that, for the follow-up 3 months after treatment (W18), 77.8% (49/63) of the participants were still participating in the study [without high dissociative experience 76.9% (30/39); with high dissociative experience 79.2% (19/24)]. No major adverse event occurred during our study (especially no hospitalization nor suicide attempt). None of the subjects had a significant increase of their depressive symptoms per protocol. A transient minor increase of their posttraumatic and dissociative symptoms (concerning respectively two and three non-overlapping subjects) was noted at V7 but went back to clinically insignificant levels at V9 (maximum DES increase at V9 = 7.5; maximum PCL-S2 increase at V9 = 3).

The SCID-Post-treatment PTSD diagnosis decreased in both groups with no significant intergroup differences (p = .346) at W7. Almost two-thirds of the patients were in remission one week post-treatment: 69.7% in the ‘without high dissociative experience’ group and 57.1% in the ‘with high dissociative experience’ group. At 3 months post-treatment, the rates of SCID PTSD diagnoses were low (14.3% in the ‘without high dissociative experience’ group and 25% in the ‘with high dissociative experience’ group) and there was no significant difference between the two groups (p = .375). However, these low rates are partly biased because some patients were withdrawn from the study after session 7 for another treatment.

3.2. Comparison of PCL-S scores in PTSD patients with or without high dissociative experiences before treatment

Anova for repeated measures reveals that, for all participants, the PCL-S scores decreased significantly over the 6 treatment sessions (Figure 1A, W1–W7) [F(6,318) = 45.37; p < .001]. However, we did not find any effect related to the presence/absence of a high dissociative experiences [F(1.53) = 3.63; p = .062], nor any interaction between this factor and the sessions [F(6,318) = 0.87; p = .515]. We only found significant differences for the PCL-S score between the groups of PTSD participants with or without high dissociative experiences during the first two sessions and in the fifth session (W1: p < .01; W2 and W5: p < .05), but there was no difference in this score one week or three months after treatment. Thus, participants with and without high dissociative experiences had significant improvement in their PCL-S scores over the 6 treatments sessions.

Figure 1.

Change over the treatment period in PTSD patient with or without high dissociative experiences before treatment in the (A) PCL-S score, (B) BDI score, and (C) DES score. Mean (± SEM). *p < .05, **p < .01, ***p < .001 between groups of patients with the presence (≥ 30) or absence (< 30) of high dissociative experiences.

During the post-treatment period (W7–W18), PCL-S scores continued to decline in all participants [F(1.45) = 4.38; p = .042], but there was no effect link to the presence/absence of high dissociative experiences [F(1.45) = 0.85; p = .362] and no interaction between this factor and the sessions [F(1.45) = 0.03; p = .858].

3.3. Comparison of BDI-II scores in PTSD patients with or without high dissociative experiences before treatment

As we have shown that the reactivation procedure with or without propranolol was effective on the PCL-S scores but also on the BDI-scores (Roullet et al., 2021), it seemed important to us to investigate the effect of the dissociation level on this anxiety score.

For all participants, BDI-II scores decreased significantly over the 6 treatment sessions (Figure 1B, W1–W7) [F(6,318) = 39.97; p < .001]. We found a significant effect related to the presence/absence of high dissociative experiences [F(1,53) = 10.3; p = .002] and a significant interaction between this level of dissociation severity and weeks [F(6,318) = 3.88; p < .001]. Thus, the severity of dissociative experiences has a very significant impact on the BDI-II scores, and we found significant differences between the two groups over the first six treatment sessions. However, this difference was no longer present one week and three months after treatment.

During the post-treatment period (W7–W18), Anova revealed that in all participants, BDI-II scores did not change significantly [F(1.46) = 0.85; p = .359] and there was no effect link to the presence/absence of high dissociative experiences [F(1.46) = 1.67; p = .202] and no interaction between this factor and the sessions [F(1.46) = 0.03; p = .870].

3.4. Evolution of DES scores in PTSD patients with or without high dissociative experiences before treatment

It can be noted that, on all participants (Figure 1C), the DES scores decreased significantly during the sessions [F(6,312) = 31.65; p < .001] and that this decrease was more significant in the group with high dissociative experiences [F(6,312) = 5.55; p < .001]. Although the difference in DES scores between participants with or without high dissociative experiences decreased at each session, this difference remained significant at all sessions, including the 3 months post-treatment session.

During the post-treatment period (W7-W18), an overall decrease in DES scores was found [F(1,47) = 4.19 p = .046]. There was an effect link to the presence/absence of high dissociative experiences [F(1.47) = 13.80; p < .001] but no interaction between this factor and the sessions [F(1.47) = 3.02 p = .089]. We should note that before the first treatment session, high dissociative experiences were recorded in 38.1% (24/63) of participants, but after six treatment sessions, this level dropped dramatically to 11.1% (6/54) at W7 and to 10.2% (5/49) at W18.

3.5. Association between severity of dissociative experiences and severity of depressive symptoms and PTSD symptoms

Before treatment, in a model adjusted for age and gender, we found a significant positive association between dissociative experiences severity (DES W1) and depressive symptom severity (BDI W1) [β = 0.447; p = .002] but not with PTSD symptom severity (PCL-S W1) [β = 0.246; p = .082].

This time, one week after treatment, we found a significant positive association between dissociative experience severity (DES W7) and PTSD symptom severity (PCL-S W7) [β = 0.473; p = .006] but not with depressive symptom severity (BDI W7) [β = 0.119; p = .473]. We found the same results three months post-treatment, i.e. a significant positive association between dissociative experience severity (DES W18) and PTSD symptom severity (PCL-S W18) [β = 0.418; p = .045] but not with depressive symptom severity (BDI W18) [β = 0.270; p = .188].

We found a very significant association between dissociative experience severity before treatment (DES W1) and PTSD symptom severity one week after treatment (PCL-S W7) [β = 0.303; p = .028] and depressive symptom severity (BDI W7) [β = 0.383; p = .005]. This predictive association of dissociative experience severity before treatment (DES W1) with PCL and BDI scores was found again three months after treatment [PCL-S W18: β = 0.306; p = .042; BDI W18: β = 0.310; p = .037]. However, no significant association was found between dissociative symptoms severity before treatment (DES W1) and the evolution of PTSD and depressive symptoms severity during treatment (W7–W1). During the 18 weeks of the study (W18–W1), there was no association focusing on PTSD symptoms nevertheless, an association was found for depressive symptoms (BDI score) [BDI: W18–W1: β = 0.318; p = .043].

4. Discussion

In our study, traumatic memory reactivation led to a global decrease of PTSD, depressive, and dissociative symptoms in patients with PTSD diagnosis independently of the adjunction of propranolol or a placebo. Interestingly, at the end of treatment, the number of participants that met the criteria necessary to be considered highly dissociated (e.g. DES > 30) dropped from 38.1% (24/63) to 11.1% (6/54) on all subjects. In addition, in these highly dissociated participants, we observed a very significant decrease in the DES score of 14 points at the end of treatment and of 19.7 points 3 months post-treatment. Interestingly, there was no more statistically significant difference of the intensity of dissociation between the two groups one week and three months after the treatment. More important, the decrease of dissociative symptoms seems to go along with both depressive and post-traumatic ones.

Even if PTSD hugely overlaps with both anxiety and depressive-related disorders with some of its symptoms being present in both conditions (i.e. anhedonia, sleep disturbance), it also integrates in its diagnostic criteria dissociative symptoms that are far less commons among the nosography. Even if dissociation is considered by some authors as a physiological defense mechanism and an attempt to gain a psychological distance from the traumatic event (Dalenberg & Carlson, 2012), some authors alleged that peritraumatic dissociation is one for the best predictors of PTSD and its intensity (Ozer et al., 2003) even though this remains debated (Lensvelt-Mulders et al., 2008). By considering it as a coping strategy, previous authors theorized that dissociation might mitigate the effects of trauma-focused treatments, because of its interference with the habituation process (Foa & Kozak, 1986). Even now, therapists often consider these dissociative symptoms as a contraindication for exposure-based treatments in PTSD, or at least believe that they would lead to poorer results (Becker et al., 2004). However, recent studies found no evidence for an attenuating effect of dissociation on psychotherapy outcomes in PTSD (Hoeboer et al., 2020) and concluded that patients with PTSD and high dissociation could benefit of trauma-focused therapies, which often decrease significantly both posttraumatic and dissociative symptoms (Atchley & Bedford, 2021). Nevertheless, even though some psychotherapeutic approaches have shown encouraging results, their overall methodological flaws constrain their generalizability (Brand et al., 2009). Our results are in lines with the most recent reviews and add one more building block to the management of patients suffering from PTSD with associated dissociative symptoms. They suggest that dissociation might not be an active coping strategy as previous authors described it, but a reaction to the traumatic event that is encoded with other features (such as emotions, behavioural response) that finally becomes a core component of the traumatic memory. This might explain the larger improvement in highly dissociate patient-subgroups in some studies (Brand et al., 2019). Trauma-focus interventions are known to act on the phenomenological features of traumatic memories in patients with PTSD (Bisson et al., 2019); these data make us hypothesize that dissociation might be one of them.

Despite all this growing body of literature, our study failed to show a correlation between the initial severity of the dissociative symptoms and the overall improvement of both posttraumatic and depressive symptoms even though there was a significant correlation between DES and BDI-II at W18.

In our study, subjects in the higher dissociation group had greater depressive and post-traumatic symptoms than those in the lower dissociation group, which is consistent with the literature. Despite this, we observed an overall reduction in BDI-II and PCL-S scores over time, a change that was even more marked in the high dissociation group. We believe this suggests that even with higher severity, traumatic memory reactivation procedures might be a promising therapeutic option. Finally, with an overall reduction in traumatic symptoms and a 12.7% patient drop rate during follow-up, our results not only showed the efficacy of traumatic memory reactivation over dissociative symptoms but also demonstrated that this procedure seems to be safe and acceptable for patients.

Nevertheless, our study suffers from certain flaws. First, it was not specifically designed to assess the efficacy of traumatic memory reactivation procedures on traumatic dissociative symptoms and thus, these results should be considered as preliminary and require replication in future controlled experiments. Previous reports also reported a link between exposure to childhood trauma and the occurrence of PTSD and dissociative symptoms throughout the lifetime (Dorahy et al., 2015; Evren et al., 2011). This might be a confounding factor that future studies need to control for.

5. Conclusion

This study suggests that traumatic memory reactivation procedures are an effective way to treat dissociative symptoms in patients with PTSD. Moreover, the improvement of these dissociative symptoms was associated with a decrease in both PTSD and depression severity. Interestingly, we observed a very low drop rate concerning the patients included, which suggests a high acceptability of this treatment. Future studies with a specific controlled design are needed to confirm these promising results for patients who suffer from PTSD with the dissociative subtype.

Funding Statement

This work was supported by a grant from the Ministère des Affaires Sociales et de la Santé [grant number PHRC0910601]. The funder of the study had no role in the study design, data collection, data analysis, data interpretation, or report compilation. The corresponding author had full access to all study data and was responsible for deciding whether or not to submit this paper for publication.

Author contributions

Profs. Roullet and Birmes had full access to all study data and assume responsibility for the integrity and accuracy of the data analysis. Study concept and design (Roullet, Thalamas, Birmes), Statistical analysis (Roullet, Yrondi, Thalamas, Birmes), article compilation (Roullet, Birmes, Yrondi, Taïb), critical revision of the article for key intellectual content: All authors. Final approval of the version to be published: all authors.

Ethics statement

Participants received detailed information about the study and gave their written informed consent to participate. All procedures were reviewed and approved by institutional review boards (ClinicalTrials.gov identifier: NCT01713556).

Disclosure statement

No potential conflict of interest was reported by the author(s).