Who Responds to HRV Biofeedback for Depression?

Depression is a heterogeneous condition, and one of the more honest lessons of the past two decades of treatment research is that no single intervention works for everyone. We have antidepressants that help many but not all, psychotherapies with broadly similar success rates and substantially different mechanisms, and a growing collection of physiological interventions — HRV biofeedback, neurofeedback, mindfulness-based therapies, exercise — each with a real but modest average effect size. Active neuromodulation techniques like rTMS and tCDS also show some promise in this domain. The clinical question, more and more, is not does this intervention work? but who is this intervention likely to help?

A new study published in the AAPB journal, Applied Psychophysiology and Biofeedback, in 2026 by Gao and colleagues at Kaohsiung Medical University and the University of Southern California takes that question seriously for HRV biofeedback in young adults with depressive disorders. Forty-eight patients aged 18 to 35, all referred by a psychiatrist with a DSM-5-TR depressive disorder diagnosis and BDI-II scores in the moderate-to-severe range, were assigned to either ten sessions of HRVB or an active progressive-muscle-relaxation (RT) control over five weeks. (Biofeedback, broadly, is a training method that gives people real-time information about a physiological signal — heart rate, respiration, muscle tension, skin conductance, brainwaves — so they can learn to regulate it. Heart rate variability biofeedback, the variant studied here, trains paced breathing at an individual's resonance frequency — typically between 4.5 and 6.5 breaths per minute — to maximize the amplitude of respiratory sinus arrhythmia and engage the baroreflex, the closed-loop control system that links blood pressure, heart rate, and respiration.)

The study set up two distinct contrasts. The first asked whether HRVB outperformed an active control on depression and on autonomic measures — the standard efficacy question. The second, more interesting one, was nested inside the HRVB arm: which baseline characteristics distinguished the participants who showed measurable parasympathetic gains from those who did not? This is a predictor question, and predictor questions are where personalized clinical reasoning starts.

Why should clinicians care about predictors? Three reasons. First, HRVB is a time-intensive intervention — ten in-clinic sessions plus daily home practice across five weeks is a meaningful ask of any client, especially one whose energy and motivation are already depressed. Knowing in advance who is likely to respond changes the cost-benefit conversation. Second, the field's responder/non-responder ratios for any biofeedback or neurofeedback protocol routinely sit somewhere between 30 and 50 percent — non-trivial, and worth taking seriously rather than averaging away. Third, the predictors that emerge from a study like this one are often the bridges between research and the kind of pre-treatment assessment a thoughtful practitioner is already doing intuitively. They give clinical intuition a measurable shape.

Methods

Sample and design. Of 101 referred patients, 59 completed pre-test and 48 completed both pre- and post-test (HRVB n = 24; RT n = 24). Inclusion required psychiatrist referral with a DSM-5-TR depressive disorder diagnosis, a Beck Depression Inventory-II (BDI-II) score of at least 14, age 18 to 35, and right-handedness. Exclusions covered bipolar disorder, schizophrenia, PTSD, OCD, autism, ADHD, substance use disorders, pregnancy, prior long-term breath training (yoga, qigong), and rotating or night-shift work — a sensible filter for a study examining short-term autonomic plasticity. Allocation followed a sequential alternating procedure rather than strict randomization, which reduces selection bias without quite meeting full RCT standards. Mean age was 25.13 (HRVB) and 24.54 (RT), 75% and 62.5% female respectively, with comparable BDI-II totals (28.83 vs 32.04 — moderate-to-severe range) and roughly half of each group meeting criteria for comorbid anxiety. Notably, 83% of HRVB participants and 50% of RT participants were on beta-blockers during the study, which matters for autonomic measurement and which we will return to.

Procedure. Both groups received 60-minute sessions twice weekly for ten sessions across five weeks. Each session followed a standardized structure: discussion of homework (5 min), pre-training rest with ECG (5 min), psychoeducation (5 min), four 5-minute training runs interleaved with 2.5-minute discussion periods, a 5-minute self-guided transfer run, post-training rest (5 min), and review (5 min). Pre-test and post-test included the BDI-II, Beck Anxiety Inventory (BAI), demographic data, and a 25- to 30-minute multi-stage ECG protocol. The post-test stages were rest, paced breathing at 12 breaths per minute (a standardized comparison condition), and a self-guided transfer stage in which participants applied their trained skill — resonance-frequency breathing for HRVB, self-guided muscle relaxation for RT — without any real-time feedback. The transfer stage matters here, and we will return to why.

HRVB protocol. Adapted from Lehrer and Gevirtz's resonance-frequency tradition (Lehrer et al., 2010; Lin et al., 2019; Yoo et al., 2022). Each participant's individual resonance frequency was identified within the 4.5–6.5 breaths-per-minute range. ECG and respiration sensors fed a ProComp Infiniti system displaying real-time heart rate and respiration waveforms, plus HRV indices — LF power, LF percentage, HR max-min — on a 28-inch monitor. Training combined paced diaphragmatic and pursed-lip breathing at resonance frequency with progressive muscle relaxation. Daily home practice was conducted twice for 5 minutes (10 min/day total) at least 5 days per week using a Eureka wearable device, with an accompanying mobile application providing LF percentage feedback and emailing training data to the research team for review.

Active control. Progressive muscle relaxation based on Lin (2018), targeting different muscle regions across sessions (hands and forehead in sessions 3–4; shoulders, neck, and back in 5–6; legs and buttocks in 7–8; whole body in 9–10), with surface EMG feedback aimed at maintaining trapezius EMG below 5 µV and stable heart rate. Same Eureka home practice with audio guidance.

Outcomes. BDI-II and BAI for symptoms; ECG-derived HRV indices computed in Kubios Premium 3.5 — SDNN, RMSSD, LF (0.04–0.15 Hz), HF (0.15–0.4 Hz), LF/HF ratio, total power. All HRV indices were natural-log-transformed. Two-way mixed-design ANOVAs tested Group × Time interactions across two time contrasts: pre-test rest vs post-test rest, and pre-test rest vs post-test transfer. Bonferroni correction across six HRV indices set significance at p<.0083 (.05/6), p<.0017 (.01/6), and p<.00017 (.001/6), with partial eta squared interpreted via Cohen's conventions. Within HRVB, participants were classified as responders (n = 10) or non-responders (n = 14) based on whether RMSSD increased or decreased from pre-test rest to post-test rest. Multivariate analyses of variance compared baseline characteristics, and univariate logistic regressions — forced by perfect separation in the multivariate model — estimated odds ratios for response per one-point decrease in symptom score.

Results

The straightforward result first. Both groups showed substantial reductions in depression and anxiety from pre- to post-test. BDI-II totals fell from 28.83 to 18.25 in the HRVB group and from 32.04 to 22.13 in the RT group. The same pattern held for somatic depression, cognitive depression, BAI total, and cognitive anxiety. The main effect of Time was robust and large (BDI-II total F = 64.55, p < .001; BAI total F = 19.73). The Group × Time interactions on psychological variables were not significant, which means HRVB and RT reduced symptoms comparably. This is consistent with the broader meta-analytic literature showing moderate effects for both HRVB and relaxation training on depressive outcomes (Goessl et al., 2017; Manzoni et al., 2008), and it argues against any reading of HRVB as categorically superior to a credible behavioral comparator on the symptom outcome itself.

The autonomic story is more interesting — and it lives in the transfer stage. When HRV indices were compared from pre-test rest to post-test rest, no Group × Time interaction survived Bonferroni correction. The modest signals on LF (η²p = .119) and HR max-min (η²p = .090) did not reach the corrected threshold. But when the contrast shifted to pre-test rest vs post-test transfer — the stage in which participants applied their trained skill without real-time feedback — the picture changed substantially. The HRVB group showed significantly larger gains than RT on lnSDNN (η²p = .267, large), lnLF (η²p = .465, very large), lnTP (η²p = .271, large), and HR max-min (η²p = .489, very large). lnHF showed a moderate-to-large interaction (η²p = .168) that survived a less stringent threshold. RMSSD did not — the HRVB-RT difference at transfer was small for that index, which is expected physiology rather than a paradox. We will say more about this below.

The mechanistic link to depression came through LF power. Across HRVB participants, the increase in LF power during transfer was negatively correlated with the reduction in BDI-II scores (r = −.46, p = .03). Participants who built more LF amplitude during their self-guided transfer practice — that is, who better engaged the resonance-frequency mechanism HRVB was trying to train — saw bigger reductions in depression. This is the cleanest piece of mechanism the study offers, and it points to baroreflex engagement and cardiorespiratory synchronization as the part of HRVB that does the work, rather than generic relaxation.

Who responded? Within the HRVB group, the 10 responders and 14 non-responders looked surprisingly different at baseline — though not on the global BDI-II total (24.40 vs 32.00, p = .084, trend only). The differences sat at the item level, in a specific symptom cluster. Responders had lower scores on loss of pleasure (0.70 vs 1.57, p = .028), loss of interest (0.90 vs 1.64, p = .015), worthlessness (1.10 vs 2.14, p = .003), and loss of energy (1.40 vs 1.93, p = .017). The logistic regression estimates were striking: each one-point decrease in worthlessness was associated with roughly 6.3-fold higher odds of being a responder; each one-point decrease in loss of energy with 8.7-fold; loss of interest, 5.6-fold; loss of pleasure, 3.3-fold. The effect sizes are large enough to take seriously and small-sample enough to hold loosely.

The autonomic baseline added another piece. Responders had lower pre-test parasympathetic activity (lnHF 5.25 vs 6.26, p = .047) and a higher LF/HF ratio (1.38 vs 0.68, p = .032) than non-responders. The responders' LF/HF (1.38) sat closer to Shaffer and Ginsberg's (2017) normative range of 1.5–2.0; the non-responders' (0.68) sat well below. In a depressive sample, in other words, the participants who improved had baseline autonomic profiles that were paradoxically closer to healthy on this index than the non-responders, despite having lower absolute parasympathetic activity. The authors interpret this as preserved physiological flexibility: room to grow, plus the cortical and motivational resources to actually engage with the practice.

Discussion

This study does something I find genuinely useful: it treats the question of who responds as a first-class scientific question, not a footnote to the efficacy comparison. Most HRV biofeedback efficacy literature has been organized around group means — does HRVB beat a control? — and group means routinely obscure the more clinically useful pattern of strong response in a subset of patients and minimal response in another. The Gao et al. design accepts that obscuration is a problem worth solving and uses the responder/non-responder split to start solving it.

The headline efficacy result — that HRVB and RT reduced depression comparably — is consistent with a broader pattern in the psychophysiological intervention literature. Active controls work. Progressive muscle relaxation has its own established mechanism (Manzoni et al., 2008), and asking whether HRVB outperforms it on a five-week BDI-II reduction is a high bar that, in this sample, was not cleared. That null finding should not be read as HRVB does not work for depression — both groups improved meaningfully, and prior literature using waitlist or sham comparators does support HRVB efficacy (Lin et al., 2019; Goessl et al., 2017). It should be read as HRVB and a credible behavioral comparator reduce short-term depressive symptoms similarly. The clinical implication is that the case for HRVB cannot rest on symptom-superiority alone; it has to lean on the autonomic mechanism, the durability story, and the responder profile.

The cleanest empirical contribution sits in the contrast between rest-state and transfer-state autonomic outcomes. At rest — when HRVB participants were just being measured, not deploying their trained skill — there was no significant group difference. At transfer — when they were actively breathing at their resonance frequency without feedback — the differences were large and consistent across LF, SDNN, total power, and HR max-min. That distinction tells us something important about what HRVB actually trains. It is not a passive shift in autonomic tone that the body adopts whether the person is engaged or not. It is a learned skill — a deployable capacity for voluntary autonomic regulation that the person activates when they choose to. The study's authors put this clearly: HRVB enhanced the capacity for voluntary autonomic regulation rather than producing sustained resting-state changes. That framing has clinical legs. Skills generalize when practiced across contexts; passive shifts do not.

The HF-vs-LF dissociation during transfer is worth one more sentence. RMSSD and HF did not show large group × time interactions during transfer, while LF did — dramatically. This is not paradoxical. Resonance-frequency breathing centers heart-rate oscillation in the LF band (around 0.1 Hz) and reduces relative oscillation power in the HF band, by definition. The right metric for grading transfer-stage HRVB performance is LF power, total power, and HR max-min, not HF or RMSSD. Studies that grade HRVB transfer using HF and RMSSD will systematically underestimate the effect.

The responder profile is the part of the paper most likely to change clinical practice if it replicates. Responders had milder anhedonic symptoms — pleasure, interest, worthlessness, energy — and lower baseline parasympathetic activity with healthier sympathovagal balance. The interpretation that ties these together is not complicated. Anhedonia and energy depletion are the symptoms most likely to predict difficulty engaging with a multi-week, twice-daily practice. A patient who cannot summon the energy to do the home practice, and who derives no pleasure from anything including practice itself, is at risk of failing to acquire the skill regardless of how good the protocol is. Meanwhile, very low parasympathetic activity at baseline with depressed sympathetic activity (the non-responder profile in this sample) may reflect a more globally dysregulated autonomic state that needs preparation before HRVB is asked to do its work. The responder pattern — milder symptoms in the engagement-relevant cluster plus relatively preserved sympathovagal balance — is what room to learn looks like physiologically.

Limitations are substantive. The sample size is small (24 per arm; 10 vs 14 in the responder analysis), and the perfect separation in the multivariate logistic regression forced single-predictor models — which means the striking odds ratios should be read as suggestive rather than definitive. Eighty-three percent of HRVB participants were on beta-blockers, which act directly on cardiac chronotropy and likely attenuate the LF and HR max-min gains the protocol can produce. Sequential alternating allocation reduces but does not eliminate selection bias. The intervention was short — five weeks — and the literature on baroreflex plasticity (Lehrer and Gevirtz, 2014) suggests three months or more of consistent practice for sustained autonomic restructuring. The post-test transfer assessment captured the immediate skill window and is silent on whether the gains were retained one or three months later. The sample is also young and Taiwanese, which limits generalization to older adults, adolescents, and culturally different populations.

For clients considering HRVB for depression, the practical signal is mixed and worth being honest about. The intervention reliably teaches a deployable autonomic skill, and that skill is associated with depression improvement on average — but in this sample HRVB did not outperform a credible relaxation alternative on symptom reduction. Clients with milder anhedonic and energy-related symptoms, and with autonomic baselines that suggest physiological flexibility, are the most likely candidates for substantial response. For referring physicians and psychotherapists, the actionable takeaway is the responder profile: BDI-II item-level review, and where possible a brief baseline HRV measurement, can help stratify the conversation about whether HRVB is the right next step or whether other foundational work — psychotherapeutic engagement with motivation, sleep stabilization, antidepressant optimization — should come first. For biofeedback practitioners, the message is methodological as much as clinical: the autonomic work HRVB does shows up in the transfer stage, not at rest, and protocols should be designed and graded accordingly.

A clinically provocative question to end on: should some patients with severe anhedonia and energy depletion be offered HRVB only after a shorter foundational phase — pharmacological optimization, behavioral activation, or a brief resonance-frequency primer with very low practice load — has restored enough capacity to engage with the protocol? The study cannot answer that. But the responder data suggest the question should be investigated.

Brendan's Perspective

What this study actually moves

Most of what we tell clients and referring colleagues about HRV biofeedback rests on a few well-cited lines. It enhances vagal tone. It reduces depressive and anxiety symptoms. It is a low-risk, learnable, generalizable skill. All of that is true on average, and it is also where the conversation usually stops. What this paper adds — and what makes it worth a careful read — is a structural reminder that averages are how we lose patients. The Gao et al. design is built around the question I find most useful in clinical practice: not does this work? but who is going to walk out with a real gain after ten sessions and twice-daily home practice for five weeks?

The answer it sketches is preliminary, sample-limited, and contingent on replication. But the shape of the answer is consistent with what a careful intake interview tries to capture, and that consistency is exactly what gives early-stage predictor research its clinical value. When a study tells you something you would have guessed at the level of intuition, but now with numbers attached, the value is not novelty. The value is calibration.

Reading the responder profile clinically

Responders were not the patients with mild depression. Responders were the patients whose depressive presentation was tilted away from the anhedonic-energy cluster — less loss of pleasure, less loss of interest, less worthlessness, less loss of energy. They could still be moderately depressed in aggregate. But the symptoms most likely to interfere with the work itself were the ones running lower in their profiles. That is not a coincidence. HRVB is, in a sense, a high-effort low-stimulation intervention. It asks the patient to sit, attend to their breath, follow a paced display, do the boring twice-daily home practice, and trust that something is shifting at a timescale that is not always immediately rewarding. The patient who cannot summon interest, cannot feel pleasure, and is exhausted before they even sit down is the patient most likely to drop out cognitively even if they show up physically.

This should change how I frame intake. When a depressed client presents with a BDI-II in the moderate-to-severe range and the highest-loaded items are exactly Q4, Q12, Q14, and Q15 — pleasure, interest, worthlessness, energy — I want to be more thoughtful about whether HRVB as a primary intervention is the right move, or whether the foundational work has to come first. Foundational work, in this context, can mean a number of things. Pharmacological optimization with the prescribing physician. Behavioral activation. A brief course of psychotherapy targeting motivation and engagement. Sleep stabilization, since chronically poor sleep depletes the same engagement reserves the protocol asks for. A very-low-dose primer of HRVB itself — three sessions of psychoeducation and resonance-frequency identification with minimal home practice load — to test whether the patient can engage at all before committing to the full protocol. The point is not that HRVB is contraindicated for severely anhedonic patients. The point is that the protocol's success depends on a substrate of engagement that some depressive presentations have, and others do not.

The autonomic baseline finding adds a complementary piece. Responders had lower baseline parasympathetic activity but a sympathovagal ratio closer to normative range. Non-responders had higher resting HF — which sounds like a good thing — but a notably depressed LF/HF ratio that the authors read as autonomic dysregulation rather than autonomic health. That distinction matters. Healthy HF is not the same thing as healthy autonomic function in a depressed sample. The non-responders' high HF combined with very low LF may reflect parasympathetic dominance without sympathetic counter-regulation — a flaccid autonomic state rather than a flexible one. HRVB asks the autonomic system to oscillate at resonance frequency, which means it asks for both sympathetic and parasympathetic engagement in coordinated alternation. A flaccid system does not produce that alternation easily.

This is not a universal claim. Other interpretations are possible, and a 24-person sample is not where you build clinical certainty. But it lines up with what experienced biofeedback practitioners have been saying about autonomic readiness for years — that some patients are ready to learn voluntary regulation and others need more foundational work to get there.

The transfer stage is the right outcome — and most studies miss it

The cleanest methodological contribution of this paper, in my reading, is also the one most likely to be overlooked: the post-test transfer stage. Most HRVB studies grade the protocol on resting-state HRV change. That is the wrong outcome. HRVB does not, primarily, train the resting state. It trains a deployable skill. The clinically relevant question is not whether the patient's HF goes up while sitting passively in a chair after five weeks. It is whether the patient can, on demand, engage their resonance-frequency breathing and produce measurable autonomic flexibility — increased LF, increased SDNN, increased HR max-min — in the moments when they need it.

That is what the transfer stage measures, and that is what HRVB participants in this study could do reliably, while the relaxation control could not. In clinical practice this is exactly the framing I want clients to internalize. We are not training your resting heart rate variability. We are training your ability to deliberately engage a regulatory state when stress, intrusive thoughts, panic prodromes, or depressive ruminations show up. A patient who can do that — even if their resting HF looks unchanged on a Sunday afternoon measurement — has acquired the clinically meaningful skill.

Practically, this means I want the assessment battery for any HRVB course to include a brief transfer-stage measurement, not just a resting baseline. Five minutes of self-guided resonance breathing without feedback at the start and end of a course tells me whether the skill is being acquired. Resting HF tells me considerably less.

Some tips for running this in real practice

Translating the study into a clinical protocol means more than scheduling ten sessions. Six lanes worth thinking through:

Threshold management — but for breath, not brainwaves. The HRVB analog of EEG threshold management is resonance-frequency identification fidelity. The protocol is only as effective as the resonance-frequency targeting that anchors it. A breath rate set generically at six breaths per minute, without testing the individual's specific resonance frequency in the 4.5–6.5 range, is a protocol running at a generic threshold. The Gao et al. study identified resonance frequency individually, and that individual targeting is part of why the LF gains were as large as they were. In real practice, this is not a methodological footnote — it is the difference between a patient who acquires the skill and a patient who is doing roughly-paced breathing for ten weeks.

Transfer-task design. The transfer stage in the lab is the prototype, not the endpoint. The clinically meaningful transfer task is whatever moment in the patient's life the regulatory skill needs to deploy — pre-sleep when ruminations spike, mid-workday when panic prodromes start, during a difficult conversation, before a high-stakes performance. The home practice should evolve across the course from quiet morning practice in optimal conditions to deliberate practice in moments where regulation matters. If the practice never leaves the morning chair, the skill never reaches the contexts where it changes the patient's life.

Sequencing logic. Where does HRVB sit in a depression treatment plan? If the patient is severely anhedonic and energy-depleted — the non-responder profile in this study — HRVB is probably not the first move. Pharmacological optimization, behavioral activation, sleep work, and a low-dose foundational psychotherapy may need to come first. If the patient is moderately depressed with relatively preserved engagement and an autonomic baseline showing flexibility, HRVB makes a reasonable early-course intervention, possibly combined with psychotherapy. If the patient is in remission or partial remission and the goal is relapse prevention, HRVB sits well as a maintenance skill. The protocol is the same. The clinical context changes which sequence makes sense.

Adaptive logic. If a patient is four to six sessions into HRVB and not increasing LF amplitude during transfer, the right move is not to keep going. The right move is to ask whether this is the wrong target, the wrong moment in their treatment arc, or a sign that medication or sleep is undermining the autonomic substrate the protocol needs. Alternatives sit nearby: resonance-frequency identification re-checked, a step back to slower breath-pacing fundamentals, a temporary shift to EMG biofeedback if muscle tension is dominating, or a sequencing reset to address motivation and engagement before resuming.

Multi-modal integration. The study trained one modality. Real clinical work rarely does. For depressed patients with cognitive slowing alongside autonomic dysregulation, HRVB plus alpha-frequency neurofeedback addresses both timing and autonomic readiness. For patients with significant somatic anxiety, HRVB plus surface EMG biofeedback accelerates the muscle-tension piece while the cardiorespiratory piece is being trained. For patients with sleep architecture disruption, HRVB pairs naturally with sleep-focused CBT-I and behavioral sleep work, since vagal tone and sleep regulation reinforce each other. The dimensions interact; the protocol should reflect that.

Learning-tracking. Exposure is not acquisition. The responder/non-responder split in this study was anchored on RMSSD change — which, for the reasons discussed, is not the ideal index for transfer. In clinical practice, I want to track LF amplitude during transfer-stage measurement at session 4, session 8, and post-test. If LF is not moving by session 4, the patient is not learning the protocol. That signal should change the conversation in the session, not just the post-test report.

If I had to crystallize the lesson: HRV biofeedback is not a calmness intervention. It is a self-regulation skill, and skills require a substrate of engagement and a substrate of physiological flexibility. The patients with both substrates are the ones the data say will benefit. The clinician's job is to know which substrate is missing, and to address it before — or alongside — the protocol itself.

That framing keeps the protocol where it belongs: as one tool in a broader clinical formulation, applied when the conditions for learning are present, and held in reserve when the foundational work has to come first.

Conclusion

This is a careful, predictor-focused study in a population — young adults with depressive disorders — that has been underrepresented in the HRV biofeedback literature. Its headline efficacy finding is honest and important: HRVB and an active progressive-muscle-relaxation control reduced depression and anxiety to a similar degree across five weeks. That null is not a defeat for HRVB; it is a methodological maturation, the kind that happens when the comparison group stops being a waitlist and starts being a real intervention.

The autonomic story is where HRVB earns its distinctive clinical place. The skill HRVB trains shows up in the transfer stage — when patients deploy resonance-frequency breathing without real-time feedback — not in passive resting measurements. Within that transfer-stage gain, the increase in LF amplitude is the part statistically tied to depression reduction (r = −.46). The mechanism is the resonance breathing itself, not generic relaxation, and the outcome is a deployable self-regulation capacity rather than a passive autonomic shift.

The responder profile is the most actionable finding for clinicians. Patients with milder anhedonia, less interest loss, less worthlessness, less energy depletion, and autonomic baselines closer to normative sympathovagal balance are the ones most likely to see substantial response. The non-responder profile suggests a foundational-work-first approach: pharmacological optimization, behavioral activation, motivation work, sleep stabilization, and only then the demanding learning curve HRVB asks for.

HRV biofeedback is not magic for depression, and this study does not claim it is. But it is a learnable, deployable, multi-context regulatory skill, and for the patients with the engagement and the autonomic flexibility to acquire it, the gains are clinically real. The work ahead is not better protocols. It is better stratification — knowing who walks in ready, and what foundational work the others need before the protocol is the right next step.

That is a beautiful clinical question, and increasingly an answerable one.

References

• Gao, C., Mather, M., Liao, H.-Y., Lee, K.-L., Yeh, Y.-C., Ke, C. L. K., Yen, C.-F., & Lin, I.-M. (2026). Baseline depressive symptoms and heart rate variability indices predict HRV biofeedback outcomes in young adults with depression. Applied Psychophysiology and Biofeedback. https://doi.org/10.1007/s10484-026-09785-7

• Goessl, V. C., Curtiss, J. E., & Hofmann, S. G. (2017). The effect of heart rate variability biofeedback training on stress and anxiety: A meta-analysis. Psychological Medicine, 47(15), 2578–2586. https://doi.org/10.1017/S0033291717001003

• Lehrer, P. M., & Gevirtz, R. (2014). Heart rate variability biofeedback: How and why does it work? Frontiers in Psychology, 5, Article 756. https://doi.org/10.3389/fpsyg.2014.00756

• Lin, I.-M., Fan, S.-Y., Yen, C.-F., Yeh, Y.-C., Tang, T.-C., Huang, M.-F., Liu, T.-L., Wang, P.-W., Lin, H.-C., Tsai, H.-Y., & Tsai, Y. C. (2019). Heart rate variability biofeedback increased autonomic activation and improved symptoms of depression and insomnia among patients with major depressive disorder. Clinical Psychopharmacology and Neuroscience, 17(2), 222–232. https://doi.org/10.9758/cpn.2019.17.2.222

• Manzoni, G. M., Pagnini, F., Castelnuovo, G., & Molinari, E. (2008). Relaxation training for anxiety: A ten-years systematic review with meta-analysis. BMC Psychiatry, 8(1), 41. https://doi.org/10.1186/1471-244X-8-41

• Shaffer, F., & Ginsberg, J. P. (2017). An overview of heart rate variability metrics and norms. Frontiers in Public Health, 5, Article 258. https://doi.org/10.3389/fpubh.2017.00258