Trial design and participants

This study was a controlled, quasi-experimental, multi-centric, prospective, interventional, and exploratory trial conducted between November 2014 and December 2017 in six psychiatric rehabilitation centers (Grenoble, La Roche sur Foron, Valence, Lyon, Clermont-Ferrand, and Saint-Etienne), 1-day hospital (Chambéry) and three medico-social supported living services (Cotagon center, Chardon Bleu and ALPHI homes) located in the Auvergne-Rhône-Alpes region in France. Eligible participants were all 18–65 years-old patients meeting the DSM V criteria for schizophrenia spectrum disorders enough physically fit to practice a PA (determined by a physical examination by patient’s general practitioner) and willing to give informed consent. Exclusion criteria were: (a) a known neurological disorder; (b) intellectual disability; (c) inability to read or speak French; and (d) inability to practice a PA. Participation to other programs having an impact on social function (SCR and SST) was prohibited during the follow-up period. Eighty-seven clinically stable (i.e., absence of significant changes in pharmacological treatment in the past month) participants diagnosed with schizophrenia were consecutively recruited and assessed at baseline, post-treatment (4 months) and at 6 months of follow-up (for RR patients only). The control group was not assessed at follow-up to allow the participation to other interventions improving social function. Forty-seven males and 10 females participated to the RR group, 23 males, and 7 females to the TR group (homogeneity male–female in the access to the intervention, p = 0.575 using Fisher’s exact test). Participant’s mean age was 33.3 in the RR group, 33.1 in the TR group. Baseline characteristics of the participants allocated to the two arms are presented in Table 1. The withdrawal of one control center (Bourg en Bresse, reason invoked = lack of time) before the beginning of the study explains the difference in the sample sizes. Psychoeducation about the illness was provided to three RR participants and supportive group psychotherapy to one TR participant during the treatment period or follow-up, with no incidence on treatment’s effectiveness. Participants were allocated in the experimental or the control group depending on the routine clinical practice of the recruiting center. Centers practicing social cognitive remediation animated the experimental group (Grenoble, Lyon, Valence, La Roche sur Foron, and Clermont-Ferrand) and centers practicing PA under the supervision of a specialized sport educator animated the control group (Cotagon, Chardon Bleu, ALPHI, Chambéry, and Saint-Etienne). Experimental centers did not have specialized sport educators and control centers could not implement the experimental intervention (absence or lack of CR-trained facilitators). This made impossible single-blind randomization. Evaluators who did not take part into treatment and blind to the time of assessment conducted baseline and follow-up assessments. All the clinicians involved in the experimental group attended to 1-day training event in RemedRugby, led by authors J.D., F.G., M.F., and F.M. Training included discussion and role-playing exercises, with corrective feedback from the trainers. The manual was comprehensive giving the detail of each session. Regular group supervisions were organized to ensure treatment fidelity. Changes in pharmacological treatment and all the interventions received during the follow-up period were systematically collected for both groups. The trial protocol was in accordance with the Declaration of Helsinki. It was approved by the local ethics committee (CPP Sud-Est V) under the number 2014-002655-26 and registered at www.clinicaltrials.gov (NCT03775564). Written informed consent was sought for all participants.

Table 1. Baseline characteristics in the RR and the TR groups

For each of the 42 outcome variables, the table displays the sample mean and standard deviation of the variable over the Expe patients, then over the TR patients, then the results (test statistic, degrees of freedom, p value) of the Walsh two-sample two-sided T test between both. A p value smaller than 0.05 indicates that the difference of means is significant. CPZ100eq: dose equivalent to 100 mg/day of chlorpromazine calculated according to the minimum effective dose method [Reference Farreny, Aguado, Ochoa, Haro and Usall25]. The mean antipsychotic dose in the sample was equivalent to 6 mg/day chlorpromazine in the RR group and to 5.5 in the TR group.

Abbreviations: AcSo, self-assessment of social cognition; AIHQ, Ambiguous Intentions and Hostility Questionnaire (agress, aggression bias; anger, anger score; blame, blame score; host, hostility bias; resp, attribution of responsibility); BUES, Boston University Empowerment Scale; BEM-RI, BEM immediate recall; BEM-RD, BEM delayed recall; Coding, WAIS-IV Coding subscale; Education, Education level (years); ERF-CS, Social Cognition—Functional Outcomes Scale; ISMI, Internalized Stigma of Mental Illness scale (tot, total score; Alien, alienation subscale; stereo, stereotype endorsement; discri, discrimination experience; soc-with, social withdrawal; Rstig, stigma resistance); MASC, Movie for the Assessment of Social Cognition (Abs, absence of ToM; CM, correct mentalization score; Ho, hypomentalization; Hy, hypermentalization score); PANSS, Positive and Negative Syndrome Scale (positive, negative, and general psychopathology subscales); PerSo, social perception test; PSP, Personal and Social Performance Scale total score; QCAE, Questionnaire of Cognitive and Affective Empathy; SERS-SF, Self-Esteem Rating Scale-short form (positive and negative subscales); Simil NB, WAIS IV Similarities subtest; STORI, stage of recovery instrument (awar, awareness; growth, growth stage; mora, moratorium stage; prep, preparation stage; rebuil, rebuilding stage); TMT, Trail Making Test A and B; Bold values indicate the p-values significant at p<0.05.

Outcome measures

Secondary outcomes

General information on education, marital status, economic status, illness onset and trajectory and comorbidities, was recorded. Illness severity was assessed using the Positive and Negative Syndrome Scale (PANSS) [Reference Kay, Fiszbein and Opler37] with an adequate inter-rater reliability (ICC = 0.66). Self-stigma was assessed using the Internalized Stigma of Mental Illness scale (ISMI) [Reference Boyd Ritsher, Otilingam and Grajales38], a 29-item self-report measure designed to assess people’s personal experience of stigma related to mental disorders. A higher score reflects a higher level of self-stigma. This scale has shown good internal consistency (CA = 0.80–0.92) [Reference Boyd, Adler, Otilingam and Peters39] and test–retest reliability (0.92) [Reference Boyd, Adler, Otilingam and Peters39]. Self-esteem was measured with the Self-Esteem Rating Scale-Short Form (SERS-SF; CA = 0.87–0.91) [Reference Lecomte, Corbière and Laisné40], empowerment with the Boston University Empowerment Scale (BUES; CA = 0.86) [Reference Rogers, Chamberlin, Ellison and Crean41] and personal recovery using the self-reported Stage of Recovery Instrument (STORI) [Reference Andresen, Caputi and Oades42]. Neuropsychological baseline and follow-up cognitive assessments included Wechsler Adult Intelligence Scale-fourth edition (WAIS-IV) [Reference Wechsler and WAIS43] Coding and Similarities subtests respectively for speed of processing and verbal abstraction, Trail Making Test A and B (TMT-A and B) [Reference Godefroy44], respectively for speed of processing and reactive mental flexibility, shopping test-revised for planning abilities [Reference Fournet, Demazière-Pelletier, Favier, Hugonot-Diener, Thomas-Anthérion and Sellal45], BEM-144 story subtest for auditory-verbal memory [Reference Signoret46], and premorbid IQ with the French-National Adult Reading Test (f-NART) [Reference Bovet47].

Interventions

The experimental intervention

RR is a manualized 15-session group-based intervention following the principles of cognitive remediation (“a behavioral training intervention targeting cognitive deficits using scientific principles of learning, with the ultimate goal of improving functional outcomes” [Reference Bowie, Bell, Fiszdon, Johannesen, Lindenmayer and McGurk48]). Figure 1 provides an overview of the RR topics.

Figure 1. Overview of the RemedRugby topics.

RR consists of 12 weekly 2-h sessions animated in the local rugby club by two facilitators and three daylong TR tournaments regrouping participants from all the centers. Sessions are conducted in a sequential fashion. In each session, participants are encouraged to generate information about the topic of the session and to remind the content of the previous session (freely and with cuing when necessary). The major part of each session is dedicated to the learning of a technique with its practical applications. Participants are invited to read the exercises from their manual, complete the prompts and to share them with other group members and to discuss their views related to them. They are invited to identify and monitor the strategies they use during the exercises and to consider how they could apply these strategies to their daily life. At the end of a session participants are encouraged to summarize what was learnt with its potential applications to their daily life. A part of each session (15 mn) and one specific session are dedicated to the practice of TR under facilitator’s supervision. At-home practice exercises are carried out between the sessions. RR has four major foci each of which targeting a specific domain: (a) SCR and SST applied to the situations that can be lived in the field (i.e., a referee not signaling a fault or a partner missing a pass), the following emotional possible reactions and their consequences on oneself and on his/her partner; (b) SCR via a blog common to all teams where participants are invited to formulate their remarks, questions, compliments, and criticism about the rules, the duration of matches, the organization of the previous or next tournament. After each tournament, group members are encouraged to list freely the themes they would like to discuss with the other teams, to categorize them and to complete the exercise from their manual (redaction of an e-mail addressing all the selected themes). Participants are invited to take other’s perspectives when writing their messages and to ask themselves about the intention of their interlocutor or the impact of their answer on others. Participants are encouraged to share their strategies and feelings with others and to formulate the team’s common answer on the blog; (c) Problem-solving skills via the organization of a daylong TR tournament regrouping all participants and facilitators (>80 persons). Three tournaments are organized during the program (one per experimental center). Participants from the inviting team are encouraged to consider the different steps of the organization of a sport event (material needed, number of guests and matches, starting and finishing hours, site preparation, and cleaning; guest reception) and their duration. The 23 steps of the organization are presented in disorder to the participants. After reordering them in the right chronological fashion, participants are invited to estimate the duration of each step and to proceed to a cumulative calculation aiming at determining the ending time of the tournament. Participants are then encouraged to list freely the material necessary to the organization and to order it in categories. They are invited to organize to themselves in pairs, each pair having specific tasks to realize during the day. The program is then published on the blog to inform the other teams of the details of the event. Group members from the other teams take note of the program and location of the tournament, estimate with a map the time needed for transportation and set an appointed leaving time; (d) SCR and SST applied to interaction with the medias. Group members are encouraged to reflect on stigma, self-stigma and the ways to reduce stigma. Participants are encouraged to share with other group members their personal stories about coping with stigma and to discuss the potential impact of a media intervention in changing people’s opinion about mental illness. They are invited to complete an exercise from their manual in which they imagine the questions a journalist could ask on the day of the tournament and write the answers they would like to make. Participants discuss their answers with the other group members and formulate the message they would like to pass to the press and the general public. They are encouraged to use problem-solving skills to plan the situation and to realize role-playing exercises. Role-playing trainings are made on the field (close to the real situation), one player playing the journalist, one the cameraman and the others being interviewed. Group members and facilitators provide feedback on the role-play. On the final tournament, participants who are willing to be interviewed can answer to local journalists with the objective to contribute to mental illness destigmatization.

Statistical analysis

RR patients were evaluated at three visits V1, V2, and V3: V1 occurred before the treatment, V2 immediately after, V3 6 months later. The TR patients only had V1 and V2 after the same delay as RR patients. Forty-six variables were treated. The 42 score variables were evaluated at V1, V2, and V3 for RR patients, only V1 and V2 for the TR patients. A first question was to detect possible differences between the two populations of RR and TR patients over the 46 variables of interest at V1. This was done for the 45 continuous variables, using Welch two-sample T test. Results are shown on Table 1. For the discrete variable (gender) Fisher’s exact test was used, but no dependence was detected. The second question was to assess the effect of the treatment on the 57 RR patients at V2, and whether it had lasted until V3. For each of the 42 score variables, the values on the 57 RR patients at V1, V2, and V3 were compared two by two using a Welch’s two-sample T test, the side being chosen so that a small p value should correspond to improvement of the patient’s condition between successive visits. The p values are reported in Table 2. The third issue was to evaluate the effect of the treatment on the improvement of the different scores between V1 and V2. Improvement was defined as the difference between the values at V1 and V2, oriented so that a positive difference should correspond to an improvement of the patient’s condition. Each difference was adjusted by a linear regression over its confounding value at V0. The question was: is adjusted improvement higher among RR patients? Improvement in mean was tested by Welch’s two-sample T test for each of the 42 score variables. Results are shown on Table 3. A power analysis was performed using the R package “pwr” [Reference Champely50] based on Cohen [Reference Cohen51]. For sample sizes of 37 and 50, an alpha level of 0.05, and an effect size (Cohen’s d) of 0.554 (see Table 3), the expected power of a two-sample T test is 0.68. It increases with the effect size, and is above 0.99 for an effect size greater than 1. The Shapiro–Wilk normality test was applied to all the tested samples. A False Discovery Rate correction was applied to each vector of p values, using the Benjamini–Yekutieli method, which is the most stringent, and also the most robust to possible dependencies. Referring to Table 3, all p values <0.001 were still <0.01 after FDR correction (results not reported here, available upon request). For the multivariate analysis, only adjusted improvements in the four PANSS scores, and the PSP score were considered. The values at V1 of the five variables of interest were considered as covariates. Five more covariates were added: age, gender, education level, and participation rate. Thus, improvement for each of the five variables of interest was tested against a set of 11 potentially explanatory factors, including the binary predictor: RR versus TR. All covariates were included in a linear regression model. A stepwise variable selection was performed using Bayes’ information criterion (BIC). p Values for the complete linear regression model are shown on Table 4. Data was analyzed using the R software, version 3.2.3 (R Core Team) [52]. The psych package version 1.5.8, was used Revelle [Reference Revelle53]. The significance level was set at 0.05. Effect size (Cohen’s d) was calculated using package effsize [Reference Revelle53]. Size effects inferior to 0.20 were considered as negligible, from 0.20 to 0.40 as small, from 0.40 to 0.60 as moderate, and superior to 0.60 as strong [Reference Revelle53].

Table 2. Improvements in the RemedRugby group at post-treatment and follow-up

For each of the 42 outcomes variable, the table displays the p values of three Walsh paired samples one-sided T tests: values at V1 against V2, then V2 against V3, then V1 against V3. In each case, the alternative has been chosen so that a small p value should correspond to improvement of the patient’s condition between consecutive visits. CPZ100eq: dose equivalent to 100 mg/day of chlorpromazine calculated according to the minimum effective dose method [Reference Leucht, Samara, Heres and Davis49].

Abbreviations: AcSo, self-assessment of social cognition; AIHQ, Ambiguous Intentions and Hostility Questionnaire (agress, aggression bias; anger, anger score; blame, blame score; host, hostility bias; resp, attribution of responsibility); BEM-RD, BEM delayed recall; BEM-RI, BEM immediate recall; BUES, Boston University Empowerment Scale; Coding, WAIS-IV Coding subscale; Education, education level (years); ERF-CS, Social Cognition—Functional Outcomes Scale; ISMI, Internalized Stigma of Mental Illness scale (Alien, alienation subscale; discri, discrimination experience; Rstig, stigma resistance; stereo, stereotype endorsement; soc-with, social withdrawal; tot, total score); MASC, Movie for the Assessment of Social Cognition (Abs, absence of ToM; CM, correct mentalization score; Ho, hypomentalization; Hy, hypermentalization score); PANSS, Positive and Negative Syndrome Scale (positive, negative, and general psychopathology subscales); PerSo, social perception test; PSP, Personal and Social Performance Scale total score; QCAE, Questionnaire of Cognitive and Affective Empathy; SERS-SF, Self-Esteem Rating Scale-Short Form (positive and negative subscales); Simil NB, WAIS IV Similarities subtest; STORI, tage of recovery instrument (awar, awareness; growth, growth stage; mora, moratorium stage; prep, preparation stage; rebuil, rebuilding stage); TMT, Trail Making Test A and B; Bold values indicate the p-values significant at p<0.05.

Table 3. Univariate differences between V1 and V2

For each of the 42 outcome variables, improvement was defined as the difference between the value at V2 and V1, oriented so that a positive difference should correspond to an improvement of the patient’s condition. Each difference was adjusted by a linear regression over its confounding value at V1. The table displays the sample size mean and standard-deviation of the adjusted difference over the Expe patients, then over the TR patients, then the results (test statistic, degrees of freedom, p value) of the Walsh two-sample one-sided T test between both. A p value smaller than 0.05 indicates that the mean improvement over Expe patients is significantly higher. CPZ100eq: dose equivalent to 100 mg/day of chlorpromazine calculated according to the minimum effective dose method [Reference Leucht, Samara, Heres and Davis49].

Abbreviations: AcSo, self-assessment of social cognition; AIHQ, Ambiguous Intentions and Hostility Questionnaire (agress, aggression bias; anger, anger score; blame, blame score; host, hostility bias; resp, attribution of responsibility); BEM-RD, BEM delayed recall; BEM-RI, BEM immediate recall; BUES, Boston University Empowerment Scale; Coding, WAIS-IV Coding subscale; Education, education level (years); ERF-CS, Social Cognition—Functional Outcomes Scale; ISMI, Internalized Stigma of Mental Illness scale (Alien, alienation subscale; discri, discrimination experience; Rstig, stigma resistance; stereo, stereotype endorsement; soc-with, social withdrawal; tot, total score); MASC, Movie for the Assessment of Social Cognition (Abs, absence of ToM; CM, correct mentalization score; Ho, hypomentalization; Hy, hypermentalization score); PANSS, Positive and Negative Syndrome Scale (positive, negative, and general psychopathology subscales); PerSo, social perception test; PSP, Personal and Social Performance Scale total score; QCAE, Questionnaire of Cognitive and Affective Empathy; SERS-SF, Self-Esteem Rating Scale-Short Form (positive and negative subscales); Simil NB, WAIS IV Similarities subtest; STORI, tage of recovery instrument (awar, awareness; growth, growth stage; mora, moratorium stage; prep, preparation stage; rebuil, rebuilding stage); TMT, Trail Making Test A and B; Bold values indicate the p-values significant at p<0.05.

Table 4. Multivariate analysis

The table displays results of the multivariate analysis for the adjusted differences (improvement) of the four PANSS scores, and the PSP score, successively considered as response variables. Eleven potentially explanatory factors are considered. They include the exposure to the treatment, four descriptive variables, the values at V1 of the five variables of interest, and the level of treatment CPZ.EQ. The first 11 rows give the p value of the contribution to the complete linear model of each explanatory variable. On the last row, the p value of the two-way a nova of the simple linear regression against the full model is displayed.

Abbreviations: PANSS, Positive and Negative Syndrome Scale; PSP, Personal and Social Performance Scale; Bold values indicate the p-values significant at p<0.05.