Rationale: Previous studies with fMRI-neurofeedback in depression have demonstrated a good safety profile and considerable symptom reduction but a comparison between standard care and standard care plus fMRI-neurofeedback has not yet been carried out.
Objective: To assess the efficacy of fMRI-neurofeedback plus standard care in comparison with standard care in patients with depression.
Study design: Single-blind RCT. Patients will be randomly allocated to fMRI-neurofeedback plus standard care or standard care. Assessments will be conducted by assessors who are blinded to group allocation.
Patients will be randomised to the two groups using a block randomisation procedure (10 participants per block). Depression severity at baseline (IDS scores) will be used as minimisation variable to ensure equal severity in both groups. The study team will receive a code for group allocation. The study team and patient will be aware of the group allocation (by the nature of the study design), but assessors performing baseline, post-intervention and follow-up assessments will be blinded. During interim analyses, the analysts will remain blinded to the conditions.
If the stopping criterion is reached before the maximal N, any patients already randomised will complete the study and their data will be included in the secondary analyses.
Study population: Currently symptomatic adult patients with depression.
Sample size calculation: A Sequential Bayes Factor (SBF) sampling plan will be used. SBF allows accumulating evidence for (or against) an effect until a certain threshold, i.e. a Bayes Factor (BF), or max N is reached. This is one main advantage compared to frequentist statistics (which is biased against the null hypothesis). Moreover, type-I and type-II error rates do not exist, allowing for flexible stopping rules. This property renders SBF a more resource efficient (in particular for small to medium effect sizes) compared to (traditional frequentist) fixed-N sampling plans. Since type-II errors do not exist in Bayesian statistics, there is no concept of statistical power as in traditional frequentist statistics. However, sensitivity analyses can be carried out a priori to help determining the sample size required to reach a pre-declared level of evidence (given an assumed effect size and a specific prior distribution) that serves as a stopping criterion, For the current sampling plan a BF \>= 6 (for either the alternative or the null hypothesis) will be used as a stopping criterion, a value that is in line with recent recommendations. The analysis is carried out with an informed prior t-distribution (t(μ = 0.35, df = 3, r = 0.102)), which reflects a moderate effect size that the investigators would expect given the current design and is in line with suggested informed prior distributions in Psychology. The first interim analysis will be carried out with a minimum N=26 patients per group who have completed the primary endpoint. This minimum N is in line with recommendations for SBF sampling plans. If a BF of 6 for either the alternative or the null is hypothesis is not met, sequential sampling will be carried out with interim tests for every new patient who completes the primary endpoint until either the BF stopping criterion is met or a maximum N=38 is reached (max N). An effect size of Cohen's d = 0.6 was set as the smallest effect size of interest given the restricted resources and previously reported effect sizes reported by pre-registered randomised clinical trials of emotion self-regulation based fMRI neurofeedback. An N max of 38 per group yields 80% sensitivity to detect an effect size of 0.6 under flexible stopping as suggested by simulations (Monte Carlo with 10,000 iterations). Hence, such design is more resource efficient compared to a traditional frequentist fixed-N design (which requires N=45 to reach 80% sensitivity to detect an effect size Cohen's d = 0.6). Further, the false negative rate (i.e. the probability to accumulate erroneously sufficient evidence against the effect and stop the trial early) is only 0.1%, the false positive rate is 12%. The investigators accept this higher false positive rate (compared to the nominal 5%) given that the technology is safe to use. If the BF reaches a value BF\>=3 (but smaller than 6) for either the alter-native or null hypothesis, the investigators will revaluate available resources for a potential third sampling phase with N max of 60 per group. The sampling plan thus follows recent recommendations of best practice.
Intervention: FMRI-neurofeedback will be administered in 5 sessions of about 1 hour each, each including a visual stimulation protocol for the localisation of areas responsive to positive emotions and appr. 30 minutes of upregulation training of these areas by means of neurofeedback, using real-time fMRI signals.
Main study parameters/endpoints:
Primary outcome: symptom severity (measured by the clinician-administered Inventory of Depressive Symptomatology) after the intervention.
Secondary outcomes: measures of depression, anxiety and general mental health (Beck Depression Inventory, Self-Efficacy Scale, Hospital Anxiety and Depression Scale, Quality Of Life scale and EuroQol research foundation questionnaire), predictive value of several trait measures (moderator analysis), changes in brain activation patterns measured by fMRI for the neurofeedback group.
In the analysis, outcomes will be controlled for the measurements at baseline.
