A new retrospective cohort study in the Journal of Medical Internet Research reports what its authors call the first large-scale evaluation of at-home subcutaneous ketamine therapy, drawing on 3,041 patients treated through Mindbloom’s telehealth platform across 38 states. The reported results are strong: high rates of clinical improvement across depression, anxiety, and PTSD, a low overall rate of reported side effects, and a delivery route, subcutaneous injection, that offers meaningfully better bioavailability than the sublingual tablets that dominate the existing at-home ketamine literature. All of that is worth taking seriously. It should be read alongside three things the topline framing does not lead with: a conflict of interest that runs through nearly the entire author list, an attrition rate that leaves the reported outcomes resting on a shrinking fraction of the original sample, and a meaningful gap between the specific clinical bar the headline percentages clear and the higher bar that actually matters most.
What the study is
Mindbloom is a direct-to-consumer telehealth ketamine provider. Between January 2024 and October 2025, 3,870 patients were approved for its subcutaneous protocol; after excluding those who never started treatment or missed the enrollment window, 3,041 were included in the analysis. Patients self-administered subcutaneous ketamine at home, starting at 0.5 mg per kilogram and titrated upward under remote clinician supervision, with a mandatory peer monitor present, a blood pressure check before each session, and a clinician-on-call line. Six sessions were tracked over a median of 44 days. This is not a clinical trial. There is no control group, no randomization, and no comparison arm of any kind, a limitation the authors themselves state plainly.
The conflict of interest, stated precisely
Of the seven listed authors, three are paid contractors who have received consulting fees from Mindbloom, and three more are Mindbloom employees. One author has no financial relationship with the company. This means the study evaluating whether Mindbloom’s own commercial protocol works, using Mindbloom’s own patient data, was written almost entirely by people paid by Mindbloom. That does not make the findings false. Retrospective real-world evidence studies conducted by the company whose product is being studied are common in this space, and this desk has covered several from other telehealth ketamine providers with similar structures. But it is the single most important piece of context for reading everything that follows, and it belongs in the first read of this study, not a footnote to it.
The headline numbers, and the bar they actually clear
All three are shares of the 737 depression patients still reporting at session 6, down from 1,672 at baseline. They are not percentages of everyone who started treatment.
Source: Parks et al., Journal of Medical Internet Research, July 20, 2026, Table 2 (PHQ-9, assessment wave 3). Study conducted and authored by Mindbloom-affiliated researchers.
By the sixth session, 81.8 percent of depression patients, 80 percent of anxiety patients, and 84.6 percent of PTSD patients still providing data had achieved a minimal clinically important difference, a defined threshold improvement on standard symptom scales. Those are the numbers likely to travel. They describe the easiest clinical bar in the paper. The same table reports two higher, more clinically meaningful bars directly alongside it: treatment response, a 50 percent or greater symptom reduction, and remission, a return below the clinical threshold entirely. Response rates at the final assessment were 56.6 percent for depression, 60.6 percent for anxiety, and 76.7 percent for PTSD. Remission rates, the outcome that most directly answers whether a patient’s condition actually resolved, were 27.7 percent for depression, 29.6 percent for anxiety, and 56.7 percent for PTSD. A headline built on the 80-percent-plus MCID figures describes a meaningfully rosier picture than one built on remission, and both are accurate readings of the same dataset.
The attrition the percentages don’t show
Those percentages are calculated only among patients who were still providing data at each assessment point, and the number providing data fell sharply across the six sessions. The depression-measure sample ran from 1,672 patients at the point they became eligible for analysis down to 737 by the final assessment, a drop of more than half. The anxiety measure fell from 1,465 to 594. The PTSD measure fell from 623 to 240, over 60 percent gone by the last session. The authors ran sensitivity analyses, multiple imputation and a conservative last-observation-carried-forward approach, and reported that results held up reasonably well under both. That is good, honest methodological practice, and it should be credited. It does not change the underlying fact that the reported response and remission rates describe the roughly 38 to 44 percent of each starting subsample that was both still in treatment and still reporting outcomes six sessions in, not the full group that started.
Safety, reported plainly
Reported side effects were low across all three assessment points, running from 2.8 to 3.2 percent, most commonly lower abdominal pain and, later, memory loss, and only 0.6 percent of patients switched away from subcutaneous administration, with 0.2 percent reporting side effects from the injections themselves. Three serious adverse events were reported across the 22-month study period, an overall serious-event rate of 0.08 percent. One was a psychotic episode following a patient’s first treatment, judged likely related to the ketamine. Two patients died by suicide during the treatment period, one following their first session and one following their second. The study states plainly that whether these deaths were attributable to treatment could not be determined. Three events in a sample this size is too small a number to establish causality in either direction, and the authors are appropriately cautious about drawing one. It is also not a number to look past. A population with elevated psychiatric symptom severity carries meaningful baseline suicide risk independent of any treatment received, and the honest position is that these deaths occurred, they were reported transparently by the study’s authors, and the sample cannot settle the question of whether treatment played any role.
Why the underlying comparison still matters
None of this erases the genuine scientific question the study is pointed at. Nearly all prior real-world evidence on at-home ketamine, including large prior datasets from Hull and colleagues and from Mathai and colleagues, involves sublingual delivery, which the current study’s own authors note tops out around 10 percent bioavailability and carries documented patient complaints about inconsistent session intensity. Subcutaneous delivery, at roughly 93 percent bioavailability, is a genuinely different and understudied route, and the dosing, safety infrastructure, and cost data reported here, $165 to $215 per dose against $300 to $690 for in-clinic intravenous treatment, are useful regardless of how the efficacy numbers are ultimately weighted. The question of whether subcutaneous at-home delivery is safe and feasible is a real one this study speaks to usefully. The question of how well it actually works is the one that needs the remission rate, the attrition context, and the funding source held in view at the same time.
The frame
This is genuinely the largest dataset yet assembled on at-home subcutaneous ketamine, and the delivery-route question it addresses is a real gap in the literature. It is also a study conducted almost entirely by people paid by or employed at the company whose product it evaluates, reporting outcomes that shrink considerably in a more conservative reading of the same numbers, on a population that thinned by more than half before the final assessment. Both of those things can be true simultaneously, and reading this study well means holding both rather than defaulting to either the impressive topline or a reflexive dismissal on conflict-of-interest grounds alone. The safe, careful summary is that at-home subcutaneous ketamine appears feasible and shows a real signal of benefit in a self-selected, paying population, delivered by the company with the most to gain from that conclusion, with a meaningful share of patients not sticking around to find out, and two deaths in the dataset whose relationship to treatment nobody, including the study’s own authors, can currently answer.
This piece discusses deaths by suicide in the context of clinical trial safety reporting. If you or someone you know is struggling, the 988 Suicide and Crisis Lifeline is available by call or text at 988 in the United States.