Limited sample size and generalisability:
The sample size (n=21) is very limited, which significantly reduces the generalisability of the findings to broader populations.

Lack of follow-up assessments:
The study did not include follow-up measurements weeks or months after the intervention. Therefore, the durability and persistence of observed effects remain unknown.

Absence of clinical outcome correlations:
Although improved effective connectivity is demonstrated, clinical outcomes, such as pain reduction or improvements in functional performance, are not presented. Including clinical measures would substantially enhance the clinical relevance of the connectivity findings.

Limitations inherent to secondary analyses:
Being a secondary analysis, the study might carry risks of selection bias or incomplete control over data quality and consistency.

Insufficient detail about the sham intervention:
Further clarification is required to verify the effectiveness and adequacy of the sham neurofeedback, ensuring participants’ blindness.

Lack of technical EEG and Granger causality methodological details:
Detailed procedures regarding systematic quality control of EEG data acquisition and validation steps for Granger causality calculations are not thoroughly reported.

Absence of participant expectation data:
Information about participants' expectations and perceived treatment efficacy was not addressed, though these factors can significantly influence outcomes.

Additionally, consider integrating recent relevant literature, such as:

"Effectiveness of tele-rehabilitation in patients with knee osteoarthritis: A randomized controlled trial" (DOI: 10.1177/20552076241286186), to discuss alternative remote-based interventions.
"Transcranial direct current stimulation in knee osteoarthritis: A Scoping Review" (DOI: 10.3390/app14167100), to discuss comparative neuromodulatory techniques.