The primary objective is to establish one-year efficacy of the A-FFIP on change in ASD-specific symptoms, assessed by a standardised interaction scene of the child with an unknown adult, which is coded in a blinded fashion.
The secondary objectives are (1) To assess A-FFIP effects on child’s cognition, language and behaviour; on parent’s competences, anxiety, depression and stress, and family quality of life; (2) To study child’s and parents’ characteristics as predictors and moderators of outcome; (3) To explore treatment mechanisms (mediators) related to parents’ and child’s competences and objectively measured behaviour.
The trial is designed as a confirmatory phase-III, prospective, randomized, multi-centre, controlled, parallel-group study with two treatment arms and six measurement time points. The trial time flow is shown in figure 1.
The study will be done in a clinical setting at four German study centres which provide a special outpatient service for individuals with ASD (University based Departments of Child and Adolescent Psychiatry, Psychosomatics, and Psychotherapy at Augsburg, Dresden, Frankfurt, and Würzburg). The Sponsor is the Goethe University Frankfurt (GU), and the principal investigator is Prof. Dr. C.M. Freitag, University Hospital Frankfurt at GU. The study related diagnostic tests for all children (A-FFIP and EIAU) are performed at the four study centres, and are coded in a blinded fashion by independent individuals who are not involved in any intervention. A-FFIP is delivered by 2 therapists / child at the four study centres, who are not involved in any study related diagnostic procedures. EIAU is delivered in any local setting outside the four study sites.
Only children meeting the following criteria are included:
- All subjects must meet DSM-5 criteria for Autism Spectrum Disorder. Diagnostic assessment is standardised according to DSM-5, by performing the semi-structured Autism Diagnostic Interview (ADI-R), toddler algorithm with the parents, and the standardised Autism Diagnostic Observation Schedule-2 (ADOS-2) with the child.
- Age range: 24–66 months old at T2.
- Written informed consent of the legal caretakers of the patient.
- Ability to regularly and reliably attend appointments.
- Parents ability and willingness to attend at least every 5th therapy session with the child.
Subjects presenting with any of the following criteria will not be included in the trial:
- non-verbal developmental quotient (DQ) / Intelligence quotient (IQ) <= 30
- non-verbal mental age <= 12 months
- diagnosed vision or hearing impairments interfering with therapy
- cerebral palsy
- chronic neurological disorder
- unstable epilepsy
- neurodegenerative disorder
- Rett / Angelman Syndrome
- (history of) severe psychosocial deprivation
- insufficient care by parents
- attachment disorder
- institutional upbringing
- parents not verbally fluent in German and/or unable to read German
The majority of the local ASD patients are diagnosed at the four participating study centres. Every ASD patient meeting inclusion criteria will be invited to participate in the study. Additionally, flyers with information about the study will be provided at other medical and public facilities to reach more interested parents for inclusion. To promote participant retention and complete follow-up in all participants, the individual baseline and endpoint results of the ADOS-2 and the results of the developmental and cognitive tests will be provided and explained to the parents after the completion of the study. The families in the control group are offered to participate in the A-FFIP intervention after completion of the study related assessments after approximately one year. In case of any unexpected and unanticipated adverse events of the intervention during the study (the assessment of serious adverse events is described under “harms”), parents will be informed about these events prior participation or during the study. Published research data will be shared with interested families through mail.
Subjects may either withdraw from the intervention, but will stay in the study, or the subjects may totally withdraw from the study. A third option is that due to serious adverse or other events, the principal investigator decides that the subject has to withdraw from the intervention. In all cases, the reason for withdrawal must be recorded in the patient’s Case Report Form (CRF) and in the subject’s medical records. In case of (full) withdrawal of a subject at the request of their legal representative, the reasons will be explored as extensively as possible. The subject will be followed up and – if possible - all examinations scheduled for the final trial day will be performed on all subjects and documented.
Intervention to be investigated
The A-FFIP intervention manual  provides the clinical therapist (behaviourally trained psychiatrist / psychologist / social worker) with information on the theoretical background, description of the diagnostic assessments necessary before start of the intervention (including parent interviews), choice of individualised treatment targets, and a detailed description of the target related specific exercises, including the toy based training material (For more detailed information in English see attachment 1 of Kitzerow et al. 2019 : https://econtent.hogrefe.com/doi/suppl/10.1024/1422-4917/a000661/suppl_file/1422-4917_a000661_esm1.pdf).
Intensive A-FIPP therapy training for all involved therapists will be done before treating children within the study. The therapy training comprises:
- Individual study of the A-FFIP manual
- 2x participation in a 3-day workshop lead by the first author of the A-FFIP manual and head of the autism intervention centre at Frankfurt (K. Teufel) OR shadowing a qualified A-FFIP therapist of the Frankfurt group for one week.
- Practising the implementation of A-FFIP (planning of treatment targets, treatment principles, etc.) with a preschool-aged child with ASD not included in the study; video-documentation of intervention sessions.
- Targeted supervision based on these videos, which are rated prior to supervision by an experienced therapist based on the A-FFIP therapy fidelity checklist.
- Based on the results of the therapy fidelity achieved by the therapist, the practising and supervision are repeated until a satisfying A-FFIP fidelity score is achieved:
- Total score (Range 0-56): 50 points
- Item scores (Range 0-2): At least 1 point in each item
Continuous training and video-based phone supervision will be provided by the Frankfurt group to ensure high treatment fidelity of therapists. The on-going supervision per centre comprises:
- Every 2nd session of the first 5 patients per centre and every 8th session of subsequent patients will be video recorded, sent to the Frankfurt supervision team and rated with the A-FFIP fidelity checklist.
- Detailed feedback and supervision is provided by telephone once a month or in case of particular serious manual violations.
An additional 2-day workshop once a year lead by first author of the A-FFIP manual with current questions by the therapists, examples from the supervisors and presentation of the current fidelity scores.
Sessions are provided in an outpatient setting at the respective study centre. All treatment rooms are furnished in a standardised way, with standardized and well-arranged play material, ensuring a structured environment providing orientation and a minimum of distraction.
The intervention comprises 2 hours of intervention/week with 2 therapists working with the child. The two-therapist concept is an essential A-FFIP component. While the main therapist interacts with the child, the co-therapist acts as a shadow and prompts the child individually and in a situation specific way, during the learning of new abilities and only if necessary. This concept is believed to support the natural learning of the child, because the child will learn the specific skill or task faster, and the interaction with the main interaction partner (therapist/parent) is independent from any prompting efforts. If the child is very advanced and shows interest in social interaction with peers, 2 therapists will work with 2 equally advanced children. At least one parent is participating in a minimum of every 5th therapy session for psychoeducation and exercises with her/his individual child, with the aim to learn to use effective strategies for practising at home. Parents are asked to practise established skills at home in many naturally occurring situations; these tasks are determined by the therapist and the implementation is discussed regularly with the parents. To ensure compliance by supporting parents without increasing pressure, no specific amount of practising hours or situations is expected. Parental adherence and competence is assessed by a weekly questionnaire (PATCS, see section “mediating mechanisms”).
If the child already attends kindergarten, one therapist visits the child’s kindergarten teachers, and (if existing) the personal assistant 3x/year for individualised psychoeducation with the aim to support generalisation of the child’s acquired skills at kindergarten. The first appointment is organized directly at the beginning of the intervention.
Individualized treatment targets
The concept of A-FFIP is based on the individualised practising of developmentally based and highly specific treatment targets, taking latest research findings on ASD-specific development into account. This has been achieved by developing specific exercises based on developmental psychological science findings. Especially this third aspect is taking on but also going beyond current low-intensity NDBIs such as JASPER , PACT , pivotal response treatment  or imitation training , which focus on one or two core developmental aspects.
A-FFIP targets six core basic abilities (attentional control, joint attention, imitation, representation, planning, self/other distinction) and five developmental domains (language & communication, interaction & play, emotion regulation, cognition and adaptive behaviour).
For each of the core basic abilities and developmental domains, exercises for beginners, intermediate and advanced learners are provided in the A-FFIP manual. In addition, for each exercise, specific advice for therapists is given how to correctly execute the exercises and how to teach parents in correct execution. Play material is used highly flexible according to the preferences of the child, increasing the child’s motivation to learn and to interact with the therapist. A-FFIP is an individualised intervention program that takes the child’s current ability level into account [44, 46]. This rationale is essential due to heterogeneous strengths and weaknesses of children with ASD in different developmental domains. A-FFIP expects the therapists to choose exercises slightly more advanced than the child’s current abilities promoting successful learning. Methods and exercises are highly standardised.
At the start of the A-FFIP intervention, a thorough assessment based on ADOS-2 and Bayley-III results, play sessions with the child and parental report on the child’s behaviour at home is done to elicit the abilities of the child in each of the basic and specific developmental domains. The collected information from the standardised instruments and from the direct observation is transferred to the detailed A-FFIP intervention targets checklist (for further information see attachment 1 of Kitzerow et al. 2019 ), which is the basis for planning the individual relevant interventions goals. Additionally the parents are asked to complete a list of rewarding activities, and a broad range of positive rewards are tested at the beginning of therapy. Based on results of this first assessment and parental priority, specific A-FFIP treatment targets are chosen together with the parents (max. 8 targets).
Progress in each area is monitored by the therapist filling in a specific documentation form adjusted to the A-FFIP intervention targets checklist after each session. Once the child has achieved a treatment target, a new target is focussed, which builds on the previous targets. Established skills will be trained in an ongoing fashion in different situations (e.g. at home with parents/siblings, at kindergarten) to assure generalization.
Implementation of specific methods
A-FFIP has been developed on a strict empirical basis, therefore effective behaviourally based learning techniques are implemented, especially techniques supporting associative, operant, imitative and social learning (such as prompting, naturally based positive reinforcement, modelling, generalization), and techniques reducing interfering aggressive and stereotyped behaviour, such as antecedent- and consequence based interventions  focussing on enhancing the child’s coping skills (e.g. by providing structure and routine). The child’s motivation to learn, play, and socially interact is increased by natural reinforcement intrinsic to the child’s interests and activities, by therapist (and parental) synchrony, and by consistent positive reinforcement of self-initiated learning.
Control intervention: Early intervention as usual (EIAU)
For EIAU, individual or group therapy intensity of 1-10 hrs/week as well as waiting time prior to any intervention onset is allowed. This is representative for the German ASD preschool population, and comparable average intervention intensity per week is expected for both groups with higher variability for EIAU than A-FFIP .
Additional treatments and medication use
In the A-FFIP as well as EIAU groups, the following additional treatments are allowed: stable psychopharmacotherapy, stable medication for chronic medical conditions not interfering with the therapy, individual speech & language, occupational or physiotherapy, personal support at kindergarten/preschool or family support. Any additional treatment will be documented exactly (kind of intervention, dose, frequency etc.) and will be compared for non-random distribution between groups in the statistical analysis.
Psychotropic medication will be started or changed at least four weeks before randomization and will remain stable (mg/kg body weight) throughout the intervention (with the exception of dose adjustment to body weight changes). The following psychotropic medication will be allowed as single or combined treatment: SSRI, other antidepressants, (a)typical neuroleptics, stimulants, atomoxetine, mood stabilisers. In addition, stable medication for the treatment of chronic conditions as allergies, asthma, epilepsy, enuresis, sleeping problems, and intermitting medication for acute upper respiratory infections and diarrhoea will be allowed. Pharmacological treatment will be documented at each time of assessment (T1-T6) and psychotropic medication effects on treatment outcome will be explored in analysing study outcomes.
Children in the Intervention group are permitted to receive any other intervention during the waiting time before the start of A-FFIP.
The following concomitant treatments are not permitted during the trial:
- Additional general or ASD-specific early intervention (in the A-FFIP group)
- Additional parent training (in the A-FFIP group)
- Treatment in a day-care facility or ward of a child psychiatric department (A-FFIP and EIAU group)
- Elimination diets or therapy (A-FFIP and EIAU group).
Parents will be informed about these study requirements prior to inclusion and randomization. If the clinical necessity arises to start any elimination diet or pharmacotherapy, this will be documented, and the child will be allowed to further participate in the study. If a child needs to be admitted for day-care or inpatient child psychiatric treatment, the child will have to stop participation in the study.
Change of ASD-specific symptoms is of strong clinical relevance due to long-term effects on adult outcome of the disorder [8, 24]. Therefore the absolute change of ASD-specific symptoms, measured by the average total score of the Brief Observation of Social Communication Change (BOSCC-AT; latest version from December, 11th 2017) between baseline (T2) and one-year follow-up (T6) is the primary outcome measure. The BOSCC [25, 45, 49] is a new, reliable (inter-rater ICC=0.98), change sensitive and valid observational measure to study change in social communication and interaction as well as repetitive and stereotyped behaviours in ASD which has been recommended in a recent Health technology assessment .
The ADI-R is a semi-structured diagnostic interview with the parents, assessing the core autism symptoms in the areas of social interaction, communication, and stereotyped behaviour [50, 51].
The ADOS-2 is an observational; semi structured and standardised assessment of communication, social interaction, play and restricted and repetitive behaviours. It is the gold standard to use for ASD diagnosis [52–54]. It comprises different modules dependent on age and verbal level.
The Bayley Scales of Infant and Toddler Development – Third Edition (Bayley- III)  is an international developmental test und will be used for children with a developmental age below or equal to 42 months at the respective time-point. The reliability for the German version lies between r=.77 and r=.89, and construct validity is comparable to the original version. In this study, the cognitive, language, and fine-motor subscales will be used.
The Wechsler Preschool and Primary Scale of Intelligence (WPPSI-III)  is an international multidimensional measure of intelligence for preschool aged children (3;0 - 7;2 years), and will be done in children with a developmental age > 42 months. Reliability is high with r=.95 for total IQ.
No valid, uniform IQ test covering the entire developmental age of the sample is available in Germany, thus the two different IQ measures with current norms had to be chosen.
- Child: ASD-specific symptoms:
The BOSCC subscales (social communication BOSCC-SC, restrictive and repetitive behaviour BOSCC-RRB) and single items are examined for a differentiated picture of change in ASD-specific symptoms [25, 45]. Mean single item scores have shown satisfying inter-rater reliability (ICC=.54-.97) and internal consistency (social communication subscale BOSCC-SC α=.83, repetitive behaviour subscale BOSCC-RRB α=.41) in a recent version. A revised and updated version of the BOSCC will be used for this study, which is expected to show even higher reliability.
The ADOS-2 comparison and domain scores will additionally be used. The comparison score allows the comparison between different time points, independent of the respective module. The ADOS-2 comparison score is based on the ADOS severity score. To compare the subscales Social Affect (SA) and Restricted and Repetitive Behaviours (RRB) calibrated domain scores will be used . The ADOS-2 shows high interrater reliability for module 1-3, with an ICC of .96 for the overall total. Internal consistence varies from α=0.87 to α=0.92 in the SA domain and from α=0.51 to α=0.66 in the RRB domain .
The parent and kindergarten teacher rated Social Responsiveness Scale (SRS)  measures social responsiveness. The internal consistence is high (α=.93-.97), and the SRS has shown moderate validity compared to other ASD-specific measures. It has been recommended for use in ASD early intervention trials . We will use the SRS-16 item short version which has been shown to have a high internal consistency and reliably measures change of ASD symptoms in one social-communication domain .
The parent and kindergarten teacher rated Repetitive Behaviour Scale-Revised (RBS-R)  measures ASD-specific repetitive behaviours. These behaviours can substantially interfere with family activities and learning processes. Internal consistency for the subscales is satisfying ranging from α=.78 to α=.91 . High correlations with the CBCL and the ADI-R were found, and factor structure and reliability have been replicated for young children with ASD [61, 62]. The German translation and analysis of German norming data of the RSB-R has replicated the factor structure of the original publication (unpublished data, Frankfurt).
Child: Cognition and language:
Depending on child’s developmental level a standardised test with current German norms is done to assess cognition and language. Different forms of early intervention in ASD have been shown to improve cognitive and language development [11, 12, 28], which is of long-term relevance, because higher IQ and language level in childhood predicted adult outcome . The Bayley- III  or the WPPSI-III  will be used in relation to the child’s developmental stage (see above, diagnostic process).
Child: Additional measures
The parent rating form Child Behaviour Checklist 1½-5 (CBCL1½-5)  and the Teacher Report Form 1½-5 (C-TRF)  are two of the most widely used valid and reliable measures in clinical research, dimensionally measuring social-emotional or behavioural problems . The CBCL has been recommended as one of 12 most valid instruments assessing outcome in ASD early intervention trials .
The Behavior Rating Inventory of Executive Function-Preschool Version (BRIEF-P)  is a rating scale assessing executive function (EF) in preschool children (2;0 to 6;11 years). Internal consistency varies between α=.75 and α=.96 depending on rater/subscale. Inter-rater reliability for the total score is r=.56. Concurrent validity was shown for CBCL subscales. In preschool children with ASD, real world EF impairments were observed, which were not related to ASD symptoms . Given the relevance of EF problems for adult outcome, change in EF by early intervention is an important outcome .
Parents and family:
The German adaptation of the Parenting Sense of Competence Scale (PSOC/FSW)  is chosen to measure parental self-efficacy. It has been developed and used in previous DFG projects (Zukunft Familie I, Ha1400/14 – 1-4). The revised questionnaire has been validated with young children (2;5 to 6;5 years old) and shown satisfying internal consistency for mother and father assessments (α=.78/.79). Convergent validity with other measures of parental stress and competences has also been confirmed in the validation study . The PSOC has been used in several ASD studies .
The Depression Anxiety and Stress Scales – short form (DASS-21)  is chosen to measure parental mental distress. Internal consistency is high, varying from α=.76 to α=.91. The validation study has shown good convergent validity to other measures of depression and anxiety (r=.59 to r=.86). Parental mental distress likely influences parent-child interaction, and may be changed by early intervention .
The Family quality of Life Survey-2006 ID/DD version (FQOLS-2006-ID/DD) was developed for main caregivers of persons with ID and has been translated into 20 languages, including German . It covers nine domains, of which four are assessed here. A validation study reported high internal consistency (α=.85) and concurrent validity (r=.63) with another quality of life measure . Improving family quality of life for families with an ASD affected member is an important goal in autism intervention .
Predictors and moderators
IQ/DQ, BOSCC-AT, gender, ADI-R and both parent‘s educational status will be investigated. The educational status will be classified according to the International Standard Classification of Education (ISCED) 2011.
An eye-tracking battery will be applied at T2, T4, and T6. This eye-tracking battery includes measures of sensory perception and visual orienting preferences. These tasks include: pupil adaption fixation stability (PAFS), smooth pursuit (SMT), visual oddball (VOT), visual search (VST), non-social versus social reward (NSRT), dynamic emotion expressing faces (DEEF), natural scenes (NST), biological motion visual preference (BMVP), joint attention task (JAT).
Spontaneous imitation is assessed by automated movement analysis via depth sensor cameras during video-based assessments (ESCS, BOSCC, DCMA, ADOS-2).
Parental treatment attendance will be documented weekly by therapists .
The Early Social Communication Scales (ESCS)  is a standardised behavioural observation measure of nonverbal communication and social interaction skills that typically emerge in children with a developmental age of 8 to 30 months. For joint attention and social interaction, inter-rater reliability is >.8. The ESCS have been used as change sensitive outcome measure of early intervention trials [43, 77].
Bayley-III scales: see above. The cognitive and fine-motor scales will be studied as mediators.
The Dyadic Communication Measure for Autism (DCMA)  systematically rates quality of dyadic parent-child communication. Inter-rater reliability was ICC=.8 for parent synchrony and ICC=.59 for child initiations. The DCMA has shown good sensitivity to change, and parent synchrony mediated the child’s ASD-specific outcome in a large RCT on parent mediated early intervention [14, 24, 29].
Parental treatment adherence and competence in employing techniques at home will be studied by the A-FFIP adjusted Parent Adherence to Treatment and Competence Scale (PATCS), which has been used in previous early intervention trials . It is a self-report with 4 items related to adherence and 2 items to competence; all rated by a 5-point scale (low to high); Cronbach’s alpha =.82.
As the RCT is a behavioural therapy intervention study, blinding of participants, therapists, parents, and kindergarten teachers is not possible. To minimize observation and detection bias, blinded trained observers will collect data for all direct observation measures (BOSCC, play session, ESCS, cognition, language) especially for the primary endpoint (BOSCC-AT). The rating will be done by independent coders blind to treatment and randomization status.
After written informed consent and baseline testing for eligibility patients will be allocated by randomization in a 1:1 manner to the interventions groups using a centralized web based tool (randomizer.at). Block randomization will be performed for each centre and gender to achieve equal group sizes within these strata.
All findings including clinical data will be documented in the subject's medical record and in the CRF.
The BOSCC will be done by an independent local tester who plays with the child. The BOSCC scores will be rated from video by independent raters located at Frankfurt. BOSCC video data will be stored until ~ year 2 of data collection. Regular coding will then be done by 3-4 raters, trained to high inter-rater reliability. Every 5th video will be coded by all coders for measuring ongoing inter-rater reliability. The BOSCC is coded based on a 12-min semi-standardised videotaped play situation. 2 x 6-min segments are watched twice taking notes, and are coded immediately by a standardised scheme. 15 items are rated from 0-5 according to item specific decision trees. Mean values from both segments are calculated for each item, and the summary scores are derived: BOSCC-AT (total score, items 1-13; [without item 9]), BOSCC-SC (items 1-8), BOSCC-RRB (items 10-13) and BOSCC single item scores.
In each centre, every 2nd therapy session of the first five patients and every 5th session of the following patients are videotaped for quality assurance.
The Institute of Medical Biometry and Informatics (IMBI) is responsible for the data management using a validated system. In order to ensure that the database reproduces the CRFs correctly, a double entry of data is performed by two different persons. A query process based on beforehand specified data validation plan is established. Any entry and correction in the study database will be reported automatically in an audit file. All data management activities will be done according to the current Standard Operating Procedures (SOPs) of the IMBI.
Sample Size calculation
Sample size calculation refers to the primary outcome measure absolute change of BOSCC-AT between baseline (T2) and one-year follow-up (T6). An age and DQ matched, but not randomised, observer blind case-control study of 2 x N=20 children with ASD aged 3;2-7;9 years and an IQ/DQ of 37-134 at start of intervention resulted in an A-FFIP one-year effect size of 0.61 for the ADOS-severity score . In a pre-post sample (N=21) aged 3;8-5;8 years with a lower IQ/DQ 37-108 one year pre-post effect size of the BOSCC-AT was 0.63 . ADOS items are the basis for the ADOS-severity score and the BOSCC . The BOSCC shows a higher number of items coded on a 6-point scale compared to the 2-point scale of the ADOS, resulting in higher variability capturing more subtle changes (range BOSCC-AT 0-60 versus range ADOS 1-10). For the BOSCC-AT, a more moderate, clinically meaningful, effect size of 0.55 is expected. With a significant level of α=5% (two-sided) and a power of 1-β=80%, a sample size of 106 (2x53) is required to detect an effect size of 0.55 with the two-sample t-test (ADDPLAN, version 6.1.1). Considering a drop-out rate of 20%, 134 (2x67) patients will be randomized to the treatment groups. It can be expected that including covariates in the confirmatory analysis will increase power compared to the t-test.
Definition of Analysis Sets
Patients will be allocated to the different population sets (per protocol set, full analysis set, according to the intention-to-treat (ITT) principle and safety set). A final definition of the analysis sets is in given the statistical analysis, which is finalized prior to the analysis.
The primary analysis will be conducted based on the ITT population. Let µ denote the unknown true mean change of the BOSCC-AT between T2 and T6. The null hypotheses: H0: µA-FFIP = µEIAU is tested against the alternative: H1: µA-FFIP ≠ µEIAU.
The confirmatory test for treatment group difference with respect to the primary efficacy endpoint will be done applying a mixed model for repeated measures (MMRM) approach  modelling the difference to baseline including the fixed effects baseline BOSCC-AT, chronological age, treatment group, time, and treatment-group-by-time interaction; centre will be included as random effect. The time points of measurement are T2 (baseline), T4 (week 26-27) and T6 (week 52-54). The significance test for treatment group difference will be based on least-squares means using a significant level of α=5% (two-sided).
No missing data are expected for the covariates. Missing values for the primary outcome measure are dealt with as follows in the MMRM approach: The likelihood-based approach models jointly all actual observations without imputing missing data but using the within-patient correlation structure to provide information about the unobserved post-baseline primary outcomes and gives reliable results under the missing data at random (MAR) assumption. The MMRM approach shows favourable characteristics in terms of type I error rate, power, and bias of estimates as compared to alternative methods dealing with missing values, such as last-observation-carried-forward (LOCF) [80–82], even in the presence of a drop-out rate of 20% .
Data quality and homogeneity of intervention groups at baseline will be evaluated. All secondary endpoints will be analysed descriptively, using appropriate statistical methods.
As sensitivity analysis the primary endpoint will be evaluated based on the per-protocol set of patients without major protocol violations. Additional sensitivity analyses will be done using LOCF and complete case analysis via ANCOVA for repeated measurements. Safety analysis includes calculation and comparison of frequencies and rates of (serious) adverse events.
Analysis of predictors and moderators: In an exploratory fashion, implementing the above mentioned MMRM approach, including the pre-specified variables as main or interaction effects in the respective models. For each pre-specified variable an own model will be calculated.
Treatment mechanism on the primary efficacy endpoint will be investigated following the mediation analysis approach described in Pickles et al. 2015 .
All analyses will be done using SAS® version 9.4, except for utilizing Mplus software for the investigation of treatment mechanism.