This pilot RCT study examined whether HRVB training could increase the level of cardiac vagal modulation in adolescents with autism. In addition, the effect on other physiological outcome measures (heart rate and breathing frequency), cortisol levels and outcomes related to psychosocial functioning and self-perceived stress was examined. Further, the feasibility of a home-based, non-supervised HRVB training was determined after the supervised HRVB training. The study findings partially support the first hypothesis, stating that HRVB would result in an increase in cardiac vagal modulation given that an increase in RMSSD was indeed present in the HRVB group with a small effect size in contrast to the sham group. However, this increase was only observed as a late-emerging effect at the T2 follow-up session, five weeks after finalizing the supervised HRVB training period. Regarding the second hypothesis, only heart rate and cortisol levels changed significantly, with a small effect size, immediately following the supervised HRVB training (increase and decrease, respectively), but none of these effects remained significant after five weeks (T2). None of the clinical-behavioral outcome measures on psychosocial functioning and self-perceived stress revealed a significant change following the supervised HRVB training in contrast to the sham group. The last hypothesis stated that a home-based, non-supervised HRVB training would be feasible in adolescents with ASD. The adolescents in this study were able to perform this home-based training, but the compliance rate during this training period dropped significantly. In addition, although most of the outcome measures did not change depending on this additional home-based training, adolescents in the home-based training group did report significantly lower symptoms of stress with a medium effect size (as assessed by the stress scale of the DASS) at T2, regardless of the training that was performed in advance (HRVB or sham).
A recent meta-analysis of Laborde et al. (2022) has focused on the influence of HRVB on the parasympathetic nervous system. They demonstrated that increased vagally mediated HRV, expressed as RMSSD, was present after multiple sessions of HRVB training (or slow-paced breathing). Although the average increase in RMSSD was small after multiple HRVB sessions, the authors suggested that functional changes may indeed occur in vagal nerve efferents (leading to increased HRV). This could have been caused by the frequent stimulation of the vagal afferents during HRVB, thereby confirming the contributing role of the vagus nerve in HRVB (Lehrer & Gevirtz, 2014). The findings of this study are partially in accordance with the ones from Laborde et al. (2022) given the significant increase in cardiac vagal modulation following the supervised HRVB training. However, this effect emerged later since it was only present at five weeks after the training period (T2). As far as the authors know, only two previous studies examined the feasibility and efficacy of HRVB in children and adolescents with autism (Coulter et al., 2022; Goodman et al., 2018). The results of the current study replicate the ones from Goodman et al. (2018) in not finding immediate effects on cardiac vagal modulation following HRVB training. Nevertheless, they did not include a follow-up session, thereby hampering the verification of the proposed hypothesis regarding a late increase of cardiac vagal modulation in adolescents with autism. With regard to other populations, Laborde et al. (2022) reported that 45% of the studies concerning HRVB did not report the influence on cardiac vagal modulation (RMSSD). However, out of the 22 studies that did, three were relevant with respect to the current study as they did not reveal significant changes in RMSSD following multiple sessions of HRVB (Raaijmakers et al., 2013; Winstead, 2019) or breathing at a frequency that maximized HRV amplitude (Tatschl et al., 2020). In the study of Winstead (2019), six weeks of HRVB training (weekly guided sessions and daily home-practice) were performed by veterans (54 ± 11 years). No differences in cardiac vagal modulation were reported at 8-, 12- and 16-weeks post-baseline. Participants in the study of Raaijmakers et al. (2013) were healthy male volunteers (19–27 years) who were asked to engage in a biofeedback game during 7 sessions, spread across 16 days. During this game, both skin conductance and HRV were used as control variables while no home-practice was performed. The authors suggested that HRVB might lead to an overall increase in RMSSD if the feedback on HRV would be used in daily living situations instead of only using it as a control variable during games as in their study. Finally, the study of Tatschl et al. (2020) used a similar training dose as the current study in adults (48.7 ± 9.4 years) who were admitted at an inpatient psychiatric rehabilitation clinic. The authors suggested that the advanced age in their study sample and the density of pharmacological interventions (f.i. antidepressants and antipsychotics) contributed to the lack of change in RMSSD since both factors are known to negatively influence changes in HRV (Alvares et al., 2016; Lehrer et al., 2006; Linder et al., 2014). Given the young sample in this study, the argument of the potential influence of advanced age was not considered. However, the use of psychotropic medication (such as antipsychotics, methylphenidate, and antidepressants) may have moderated the efficacy of the HRVB training since 50% of the adolescents in the HRVB group reported this type of pharmacological intervention. The current study sample was deemed too small to perform moderation analysis on the possible role of psychotropics.
As far as the authors know, this is the first study to suggest a late-emerging effect on cardiac vagal modulation following five weeks of HRVB training. A possible explanation for this later increase might be that adolescents who received the HRVB training implicitly continued to use the learned breathing technique in daily life between T1 and T2, also outside the explicit training context. Although transfer of practice to daily life constitutes an important goal of biofeedback training in general, this study did not include any measurement to verify the extent of transfer. It should be noted that while half of the HRVB group was instructed to proceed with the home-based HRVB training (without supervision), thereby prolonging their explicit training period, this did not result in an additional effect on the level of cardiac vagal modulation at T2. Given that several questions remain regarding the exact working mechanisms of HRVB and the most efficient training protocol (Lehrer, 2022; Lehrer et al., 2020; Pizzoli et al., 2021; Sevoz-Couche & Laborde, 2022), more information on these topics is needed to further explain the late-emerging effects found in this study.
Contrary to the late influence on cardiac vagal modulation, an immediate effect was found on heart rate and cortisol levels in adolescents with autism following HRVB training. While no immediate effect on cardiac vagal modulation was present, the current study showed an increased mean heart rate at T1 in the HRVB group in comparison to the sham group. This was somewhat surprising since the hypothesized increase of cardiac vagal modulation would intuitively result in a lower heart rate. A possible explanation may be that the slow breathing frequency at which the adolescents in the HRVB group had to practice was stress inducing. This may have increased the activity of the sympathetic nervous system in the HRVB group at T1, resulting in an increased heart rate. Exploratory within-group analyses revealed a similar pattern of increased heart rate at T2 for the adolescents who received home-based HRVB after sham training, while this was not present in those who continued HRVB training after supervised training. The inclusion of physiological data on sympathetic nervous system activity in future research is needed to explore this hypothesis of increased sympathetic activity. Five weeks after the HRVB training (T2), the increased heart rate in the HRVB group was no longer apparent, probably due to the increased cardiac vagal modulation with respect to the sham group. Furthermore, the immediate decrease of mean salivary cortisol after HRVB training in comparison to the sham training suggests that HRVB may indeed exert an influence on the activity of the HPA axis. This is similar to what has been found in other studies in adult samples (Bouchard et al., 2012; Brinkmann et al., 2020). This effect, however, was not related to an increase in cardiac vagal modulation, as was hypothesized in the current study. Moreover, at T2, when the effect on cardiac vagal modulation was significant, the difference in cortisol levels between the two groups was only borderline significant. In sum, the current study results suggest that HRVB has a distinct impact on the autonomic nervous system and the HPA axis in adolescents with autism.
The mixed model analyses did not reveal a change in the adolescents’ breathing frequency following HRVB training, despite the significant decrease in breathing frequency for the HRVB group at T1 demonstrated by the exploratory within-group analyses. One explanation may be that the sample size was not large enough; another could be that the decrease in breathing frequency during the HRVB training was not generalized after the training sessions. Of note, a decrease in breathing frequency was only expected for the resting periods, not during the stress-provoking tasks given their verbal character.
The interest in HRVB for various clinical and non-clinical populations across the lifespan has increased exponentially, given the positive effects on physical symptoms, such as asthma and hypertension, and on mental health symptoms such as depression, anxiety and stress (Blase et al., 2021; Dormal et al., 2021; Fernández-Alvarez et al., 2022; Lehrer et al., 2020; Pizzoli et al., 2021; Tatschl et al., 2020; Yetwin et al., 2022). Porges’ Polyvagal Theory (Porges, 2011) and the Neurovisceral Integration Theory (Thayer & Lane, 2000) are often cited in support of these research findings. The Polyvagal Theory proposes that lower levels of cardiac vagal modulation are associated with difficulties in social behavior. The Neurovisceral Integration Theory builds upon the latter theory by describing that cardiac vagal modulation is associated with self-regulatory behavior due to the bidirectional connection between the heart and several related brain regions. Lower levels of cardiac vagal modulation have been found in adolescents with autism in comparison to typically developing peers (Cheng et al., 2020; Makris et al., 2022; Thoen et al., 2023). Thus, it was hypothesized that HRVB would upregulate the level of cardiac vagal modulation in adolescents with autism and would, additionally, result in a decrease of the severity of associated psychosocial and behavioral problems. Of note, the complementary character of HRVB to other well-established and effective psychological treatment modalities has been suggested in the meta-analyses of Lehrer et al. (2020). Therefore, HRVB was considered as a possible support strategy for adolescents with autism to facilitate stress- and self-regulation. However, non-significant changes on psychosocial and behavioral outcome measures were revealed following the HRVB training. This contrasts with the findings of both Goodman et al. (2018) and Coulter et al. (2022), who did report beneficial effects on social behavior, emotion regulation and self-reported anxiety in children and adolescents with autism. It has been suggested that behavioral changes may only become apparent when training duration is increased (Sevoz-Couche & Laborde, 2022). The findings in this study partially confirm this as less stress symptoms, assessed with a number of items of the DASS, were indeed present after the second round of HRVB training (cf. home-based HRVB) but not after the first round (cf. supervised HRVB). It has also been hypothesized by Deschodt-Arsac et al. (2018) that the time course in physiological and behavioral changes resulting from HRVB may be different in a way that physiological changes may precede any changes in behavior. The current study results do not support this hypothesis since the increase in cardiac vagal modulation and the decrease in symptoms of stress were present during the same assessment. In addition, the increase in cardiac vagal modulation was the result of the supervised HRVB training whereas the decrease in stress symptoms was caused by the home-based, non-supervised HRVB training. In line with this, there are studies in which changes in behavior or clinical symptoms were present regardless of any physiological changes (Lehrer, 2022). It is important to mention that half of the adolescents received the sham training prior to the home-based HRVB training. This implies that the effect on stress symptoms in this study cannot be explained entirely by the increased training load of HRVB as not all adolescents received ten weeks of HRVB. It could be argued that the sham training in this study was a type of slow-paced breathing given the slower breathing frequency of the adolescents during the training (10.90 ± 1.77 breaths per minute) with respect to their rather high breathing frequency during the baseline measurement at T0 (17.70 ± 3.31 breaths per minute). Therefore, it could be argued that adolescents in the sham group may have benefited from this training as well. However, the exploratory within-group analyses did not reveal any improvements in self-reported stress immediately after the sham training whereas these effects were present for the HRVB group. None of these immediate effects reached statistical significance in the mixed models (primary analyses), probably due to the small sample size of the groups. Furthermore, no changes were apparent in outcome measures reflecting the impact of autism characteristics or behavioral symptoms in the current study. More research on how HRVB exerts its effects is warranted to explain these inconsistent and seemingly contradictory effects. Finally, at T1, a significant decrease with a medium effect size in sensory hypersensitivity was present in both groups (HRVB and sham training). Since this outcome measure did not assess sensory hypersensitivity across different sensory modes, it would be interesting to explore this valuable effect more in detail in future research.
Strengths and limitations
This pilot RCT has several strengths such as the inclusion of both boys and girls with autism; a rigorously described assessment protocol that includes the assessment of both cardiac vagal modulation and salivary cortisol levels as well as parent and self-reported outcome measures and the combination of a supervised and home-based HRVB training period. However, some limitations need to be acknowledged as well. One limitation is the small study sample, proper to a pilot study. Replication in a larger RCT is necessary. Furthermore, caution is warranted regarding the interpretation of the exploratory within-group differences since no main effect of group has been found in the mixed-effect models (primary analyses). However, these exploratory analyses were included for completeness and to determine whether possible trends were present within the groups that could not be detected in the primary analyses due to the small group sizes. Possible moderating effects such as sex, presence of co-occurring disorders or psychosocial symptoms, use of psychotropic medication and severity of autism characteristics were not included in the analyses since the study sample was deemed too small to perform moderation analyses. Of note, a previous study using a subgroup of the current sample of adolescents with autism revealed a non-significant effect of psychotropic medication on the level of cardiac vagal modulation during the stress-provoking protocol (Thoen et al., 2023). However, more information regarding possible moderating factors is valuable to further define who would benefit most from HRVB and to determine whether personalized HRVB protocols are needed to maximize training-related benefits (Lehrer, 2022; Pizzoli et al., 2021). Finally, this study only included adolescents with autism without intellectual disability (ID) despite the preliminary evidence for the feasibility of slow-paced breathing in adolescents with ID (15–19 years), as confirmed in a pilot study of Laborde et al. (2017). Of note, their training protocol was based on breathing at a frequency of six breaths per minute, which is close to the mean resonance frequency found in adults, thereby relying on a similar working mechanism as HRVB. Short-term increases of cardiac vagal modulation during a cognitive task were present in their slow-paced breathing group in contrast to the comparison group that listened to an audiobook. However, they recommended more research given their small sample size, the preponderance of males in their study population and the exploration of short-term effects only (Laborde et al., 2017). Given the large co-occurrence of ID in individuals with autism, it would be interesting to further expand research on HRVB efficacy in this population.
Directions for future research
Suggestions for future research are in place to further examine the efficacy of HRVB in adolescents with autism and on HRVB in general.
First, regarding the study design, it might be valuable to include a follow-up assessment, such as in this study, to determine whether late influences on physiological and/or behavioral parameters are present (Goessl et al., 2017; Lehrer et al., 2020). In addition, a longer training period has been suggested to result in better clinical outcomes (Blase et al., 2021), but currently there is no consensus regarding the best training protocol to obtain long-lasting effects after HRVB training (Lehrer, 2022; Pizzoli et al., 2021).
Second, given the significant decrease in compliance rate during the home-based HRVB training period in this study, it might be valuable to include virtual follow-up sessions to motivate adolescents with autism in continuing with the training. Furthermore, it has been suggested to use different modes of feedback during HRVB (e.g., tactile, auditory, and visual feedback) in addition to a personalized interface (Fernández-Alvarez et al., 2022; Morales et al., 2022). In a recent study in children with autism (6–10 years), an adaptive biofeedback model was proposed which integrates physiological data, therapy adherence and environmental data during breathing exercises in order to provide personalized feedback (Morales et al., 2022). Despite the need for further development of this adaptive model, it might be valuable for adolescents with autism as well. Although many of the adolescents in this study did not report frustrations upon using the breathing pacer applications, some of them did mention that using more engaging interfaces could have increased their therapy motivation and compliance.
Third, collecting objective data regarding training compliance would be beneficial in future research. For instance, using devices that capture data during home-training sessions could provide researchers with accurate information regarding training duration and timing. In this study, a similar device (Polar heart rate sensor) and application (Elite HRV) were originally described to use during the home-based training period. However, technical problems and practical issues (e.g., difficult to apply the heart rate sensor across settings) prevented multiple adolescents from using them. Therefore, they continued using the previously installed breathing pacer applications (from the supervised training period) which hampered the objective data acquisition. This contrasts with the findings of Coulter et al. (2022), who reported good usability of two HRVB devices in young individuals with autism. However, these devices were based on infrared photoplethysmography (PPG) sensors that were placed around one’s finger or earlobe, thus are easier to use as opposed to the chest belt used in this study. Additionally, qualitative interviews regarding the usability of HRVB devices and applications as well as the tolerance of the HRVB training in adolescents with autism would complement the objective information (Yu et al., 2018).
Fourth, including an outcome measure that targets immediate session-related effects on self-perceived stress could provide valuable information as well. The reasoning behind this is based on the hypothesized induced sense of relaxation following HRVB training since the vagus nerve is stimulated (Lehrer et al., 2020).
Fifth, this study included a sham training group to control for the possible placebo effect of breathing awareness or distraction during HRVB (Sevoz-Couche & Laborde, 2022). However, HRVB might be beneficial in adolescents with autism in comparison to other stress-management techniques or methods (e.g., mindfulness). Therefore, it would be valuable to explore this by including additional control groups based on these techniques.
Finally, it has been recommended to further elucidate the possible working mechanisms of HRVB (Goessl et al., 2017; Lehrer, 2022; Lehrer et al., 2020; Sevoz-Couche & Laborde, 2022). The inclusion of outcome measures that reflect the functioning of both branches of the autonomic nervous system as well as measures reflecting the functioning of the HPA axis might provide additional information regarding physiological changes caused by HRVB. Recent studies also indicated that HRVB may enhance resting state oscillatory blood flow and brain connectivity in cortical areas related to emotion regulation (Mather & Thayer, 2018; Nashiro et al., 2023; Schumann et al., 2021). This might explain the various positive behavioral changes following HRVB (Lehrer et al., 2020). However, future research should incorporate outcome measures that can capture brain-related changes to further elucidate and confirm this hypothesized working mechanism.