Adding EEG neurofeedback to a therapy practice is a purchasing decision, a clinical decision, and a legal one, and most guides cover only the first. The equipment is commercially available, the training pathway is defined, and a CPT code exists. The condition-specific evidence is where it gets uncomfortable, and that is the part you need to settle before you tell a client what the sessions might do.
This checklist covers five things in order: what the device is cleared for, what the trials show for the conditions people ask about, what training and credentialing involve, how billing works, and what to record once you start. The short version is that neurofeedback is a real intervention with a thin controlled evidence base, and the gap between those two facts is your informed-consent conversation.
Start with what the device is cleared to do
Most clinical EEG neurofeedback systems sit under the FDA’s biofeedback device regulation, 21 CFR 882.5050, as Class II devices. The wording of the clearance is narrower than the marketing. A representative example is the Autogenics Model A620-3 clearance letter from 2001, which lists general relaxation training as the indication for its EEG biofeedback instrument. Other devices under the same regulation carry relaxation and muscle reeducation indications.
That means “FDA-cleared” is accurate and also says nothing about ADHD, PTSD, or depression. The clearances I reviewed do not name those conditions. Before you buy, pull the 510(k) summary for the exact system and keep a copy in your practice file. If a vendor’s sales material describes a treatment claim that the clearance does not, ask for the document that supports it. A ladder note for your records: the device is on the market for relaxation training, and everything beyond that is clinical judgment backed by research-stage evidence.
Match the evidence to the condition
ADHD is the most-studied use, and the best data are unflattering. The 2016 meta-analysis from Samuele Cortese and the European ADHD Guidelines Group pooled 13 trials with 520 participants. It found a significant effect on ratings from the people closest to treatment, an SMD of 0.35 for total symptoms. The effect disappeared when ratings came from probably blinded assessors or when trials used active or sham controls. A later JAMA Psychiatry meta-analysis led by Samuel Westwood restricted itself to probably blinded outcomes: 20 trials, 1,214 participants, an SMD of 0.04 (95% CI -0.10 to 0.18). The authors concluded that neurofeedback did not meaningfully benefit people with ADHD at the group level.
PTSD is less settled. A PROSPERO-registered meta-analysis in Frontiers in Neuroscience published in December 2025 separated sham-controlled trials from waitlist and treatment-as-usual comparisons. Two sham-controlled fMRI neurofeedback RCTs showed no symptom improvement, and the authors argue that better controls are needed to know whether neurofeedback or something else drives the gains. fMRI is not what a private practice runs, but the control-design problem carries over to EEG work. For depression, a February 2026 narrative review of 15 trials reports promising alpha asymmetry and peak alpha protocols. A narrative review pools nothing, so treat it as a list of leads rather than an effect size.
If you follow this field professionally, NeurotechMag Pro gives you the structured intelligence feed to go deeper than any single article can. In the meantime, look at your own caseload. Which of your clients’ goals rest on blinded evidence, and which rest on the unblinded kind that shrinks under sham control?
Not everyone reads this evidence the same way. BrainFutures’ provider page argues that neurofeedback should be considered first-line for ADHD. That position conflicts with the blinded meta-analyses above, and a client searching online will meet both. Be ready to explain why you side with one.
Get trained, and keep your license in front
Neurofeedback training sits on top of your existing professional license and does not replace it. The main credential is BCIA board certification, which BioSource Software’s overview notes is not a substitute for a state license. BCIA’s neurofeedback pathway requires an approved health-care degree, 36 hours of didactic education, 25 contact hours with an approved mentor that include 10 sessions of your own neurofeedback, and an exam. Published filing and certification fees were $150 and $300 when I checked, so confirm current numbers with BCIA.
Equipment is the larger cost. A Trauma Research Foundation page from 2021 listed a base equipment price of $7,500 alongside a $1,600 course and practicum. Treat that as a dated reference point rather than a quote. Ask your malpractice carrier whether neurofeedback is covered under your current policy, and get the answer in writing.
Sort out billing before the first session
There is no neurofeedback-specific CPT code. Practices generally bill CPT 90901, biofeedback training by any modality, and a NeuroRegulation article on ethical billing covers the history of the codes, Medicare’s position, incident-to rules, and recent fraud cases. Read it before you hire a biller. Documentation has to support each claim, and technician-delivered sessions raise supervision questions.
Coverage is patchy. Cigna’s biofeedback policy lists EEG biofeedback and neurofeedback as experimental, investigational, or unproven for any indication. BrainFutures says at least 12 states have plans that cover it, though its page is an advocacy source. A Delaware clinic described in a Technical.ly profile charged about $100 per session and did not bill insurance. If you have ever submitted a 90901 claim for neurofeedback, pull one denial and read the stated reason. It will tell you more about your payer mix than any vendor brochure.
The starter checklist
Researchers have already written the reporting standard you can borrow. The CRED-nf consensus checklist in Brain asks for control conditions, blinded outcome raters, a record of whether participants received a strategy and what they used, and a report of whether people actually learned to regulate the targeted signal. Most of that translates into a practice setting with modest effort.
Device file: the 510(k) number, the cleared indication, and the vendor’s support documents.
Credentials: license scope confirmed, BCIA status or a training plan, and malpractice coverage in writing.
Consent: a plain statement of the evidence level for the client’s target condition, including the blinded-outcome results.
Outcome plan: a standardized measure at baseline and at a pre-agreed review point, ideally scored by someone other than the treating clinician.
Session record: protocol, electrode site, frequency band, session count, and whether the client reached regulation of the feedback signal.
Billing check: payer policy confirmed in writing, 90901 documentation reviewed, and a plan for self-pay clients.
If you are building a clinic around this, the expensive mistake is a business plan that assumes the ADHD claim holds. Whether neurofeedback ever earns that claim will depend on larger preregistered, double-blind trials with sham arms, and the field has not produced them yet. Until then, which of your current clients’ goals would still stand if you counted only the blinded evidence?