Video Games Among People With Schizophrenia (GAME-S)

The effectiveness of the gaming intervention will be assessed using a controlled, controlled, single-blind clinical trial with a pragmatic, three-arm parallel-group design.

Interventions: Participants will be randomised into three groups: Cognitive training (intervention group, CogniFit, N=78), entertainment video gaming (active control group, SIMS4, N=78) and treatment as usual (a passive control group, N=78). The mechanism of CogniFit (intervention) lies on the evidence for computerised exercises focusing on auditory and verbal processing, which are likely to yield improved verbal learning and memory and activate reward systems of the brain that drive brain plasticity in adults with schizophrenia. SIMS4 game (active control) offers entertainment without any known cognitive or health-related outcome. The data collection, subject recruitment and training will be carried out in an outpatient psychiatric units or outpatient clinics, mental health associations, and residential homes in Hong Kong. The study is aimed for patients' diagnoses of schizophrenia.

Recruitment: An extended informed consent process will be used. The health service's medical records at the study site will be screened by the authority of the staff of the organisation. Patients will be screened to determine their eligibility for study participation. Eligible participant will first receive a short leaflet of the study from the staff or during a short information session organised for the participants to consider their availability. If an eligible patient shows interest, more detailed written and oral information will be shared.

After signing informed consent forms (two identical copies), baseline data with background information, Intelligence Quotient (IQ, if available) and MMSE will be collected to show evidence of the participants' capacity to give informed consent. Medication dosage [chlorpromazine equivalence] will be collected as some medications currently used in schizophrenia may affect the response to cognitive training strategies.

Recruitment will continue until the required sample size has been obtained.

Randomisation: After baseline data collection, the Trial Manager will be informed (by email, text message, WhatsApp) about a new participant and he/she will allocate the participant to one of the three arms based on a list of computer-generated random numbers provided by an external clinical trial randomisation service (blocks of 6 consecutive patients, a 1:1:1 ratio). Allocation will be masked to the outcome assessors, and the trial statistician, but it cannot be masked from the RAs who recruit the patients and the treatment staff working with the patients.

For neuroimaging assessment (EEG, fMRI, resting state), a sub-sample from our total sample will be randomised at baseline. Randomisation will be based on a list of computer-generated random numbers.

Power analysis and sample size: The investigators calculated the sample size based on (1) the number of actual pairwise tests to be made for the efficacy of the primary outcome and (2) the two-level modelling approach in the final data analysis, in which the type I error has been adjusted. The statistical efficiency will be ensured using the MANOVA method for multiple group comparison. First, given that video gaming is a fairly novel strategy, the investigators will base the sample size calculation (a priori) on a cognitive-efficacy meta-analysis for patients with schizophrenia showing and overall effect size (ES, Cohens' D) of 0.58 on verbal working memory (a primary outcome). Based on our hypothesis, the primary endpoints are the effects in verbal working memory at 3 and 6 months of cognitive training in the gaming group, in comparison with the other two groups: cognitive training vs. entertainment gaming, cognitive training vs. non-gaming control group. Four pairwise interactions between the contrast of the two comparisons and the two time points will be tested. For multiple comparison tests of four, for a type I error level of 5% (two-sided), an adjusted significant level should be = (1- (1- 0.05)4) /2 = 0.01274 /2 = 0.0064, and the corresponding z score for a one-sided test is 2.49. Given the effect size 0.58, assuming equal sample size of the three groups, with a statistical power of 0.8 and = 0.01274, the investigators require at least 198 subjects (66 per group) by applying the equation, 2(Z1- + Z1- /2)2 / ES2. According to a meta-analysis,40 the total sample size in previous cognitive training studies has typically been 50 (range 10-138).

Using evidence-based rationale for patient flow in this study, the investigators can assume that about 60% of patients that will be screened will not be eligible for the study due to age or lack of capacity to participate in the study. Based on the literature, about 45% of patients with schizophrenia will refuse to participate in the RCT studies, and 16% will drop-out during the course of intervention. Thus, the investigators need a total of 234 participants to be randomply allocated to three study groups, and 198 participants in follow-ups. Based on this knowledge, a total of 985 subjects need to be approached. The numbers are realistic given that the total number of patients with schizophrenia in our study sites is about 5,500, and the total number of schizophrenia patients in Hong Kong is 40,000.

Second, for neuroimaging assessment, a sub-sample of 126 participants from our total sample (N = 198) will need (63%) to be randomised at baseline. Randomisation will be based on a list of computer-generated random numbers. The investigators assume that 30% will drop-out between baseline and the 3-month follow-up assessment, which leave us with 29 patients in each group (totally 87 participants at baseline). The sample size will be appropriate for our neuroimaging assessment; the average number of participants in RCT studies assessing patient cognition or changes in brain structure has been about 20.

The investigators will also verify whether the interventions have been delivered as designed (intervention fidelity). Therefore, patient gaming logs (gaming frequency [number of gaming sessions per week], the length of each session [minutes], number of drop-outs) and patient gaming diaries including descriptions of possible strengths and limitations of the interventions will be analysed with the content analysis based on the data to be collected from patient diaries.