Short-term fasting induced changes in HRV are associated with interoceptive accuracy: Evidence from two independent within-subjects studies

Previous research suggested increased cardiac interoceptive accuracy after 24-hours food-deprivation by means of the heartbeat tracking task. The present study investigated if 16-hours of voluntary fasting shows similar effects and whether changes in interoceptive accuracy are accompanied by changes in autonomic function. In two independent within-subjects studies two measures of interoceptive accuracy, the heartbeat tracking task and the heartbeat discrimination task were applied. In study 1 (n=24) and study 2 (n=72) vagally-mediated HRV increases and heart rate decreases were observed. Stronger effects of fasting on vagally-mediated HRV went along with a higher interoceptive accuracy increase in the heartbeat tracking task. Furthermore, the fasting associated changes in interoceptive accuracy in both tasks were significantly associated, suggesting that these tasks are suitable to track changes in cardiac interoception. Taken together, fasting of 16-hours might be suitable to increase participants' parasympathetic efference, thereby facilitating interoception.


Introduction
Interoceptive accuracy is an important health promoting skill, which forms the physical self [2,11], shows associations with mental illness (e. g., [24]), and emotion regulation (e.g., [15,16]; for overviews see, e.g., [41,44]). Nevertheless, the skill to adequately perceive organismic cues is not as stable as most of us might think and there are situations when we tend to perceive bodily signals more accurate or when we struggle to perceive our inner states. However, which contexts facilitate or dampen interoception?
Complementing these autonomic nervous system (ANS) changes, it has been shown that short-term food-deprivation of 18-h could facilitate the cortical processing of heartbeats, thus suggesting an intensified interoceptive awareness [53]. Moreover, implementing a fasting period of 24-h in a highly controlled setting, Herbert et al. [20] could observe increased accuracy of the perception of own heartbeats (cardiac interoception), assessed by the heartbeat tracking task. They investigated 22 selected women (out of 50 after a detailed examination of physiological and psychological function) during 2 days. In addition to the expected increase of interoceptive accuracy, they found a food-deprivation-related increase of negative and a decrease of positive mood. However, they also reported an increase of heart rate (HR) and decrease of vagally-mediated heart rate variability (HRV), among other physiological changes. They interpreted their results in terms of a contribution of vagal activity to the self-regulation of interoceptive signals during acute food-deprivation. However, although the authors implemented rigorous experimental control conditions, the order of the 24-hour food-deprivation was not randomized, making it difficult to disentangle food-deprivation condition (i.e., yes/no) and order effects (day of investigation). The second day always being the food-deprivation day hampers the interpretation of short-term food-deprivation effects on physiology, affect, and interoceptive accuracy. Furthermore, the authors assessed interoceptive accuracy only by the heartbeat tracking task [51].
The heartbeat tracking task requires participants to count their heartbeats during random time intervals (e.g., 25 s, 30 s, 35 s). The answers are then compared with the number of actual heartbeats during these intervals. Although the heartbeat tracking task constitutes an easy to implement procedure with high face validity, it has been criticized by many authors during the last decades. In line with this, recent research seriously questioned the validity of this task [9,12,69], which is, however, still a matter of ongoing discussions (see e.g., [1,70]). Specifically, implicit knowledge about the own heartbeat could facilitate task performance [40,66] and therefore, participants might achieve high accuracy not due to heartbeat perception, but accurate knowledge of their heart rate [48]. An alternative approach involves discrimination between true and false sensory feedback of individual heartbeats [63]. Although discrimination tasks have been criticized as well [7,10], heartbeat discrimination might be considered a valid approach [22,25], since it allows to assess perceptual sensitivity separately from other non-perceptual factors by applying signal detection theory [63,69]. In order to overcome the limitations associated with each task, the present study assessed interoceptive accuracy using both heartbeat discrimination and heartbeat tracking tasks.
Following from above, the fasting induced changes in HRV and interoceptive accuracy might be associated with each other. Specifically, the neurovisceral integration theory [57,58] and the polyvagal theory [46] suggest that increased HRV, as a measure of vagal activity, indicates an adaptation of the ANS providing (cognitive) flexibility, attentional resources, and better emotion regulation [56]. This theoretical consideration is further strengthened by empirical findings connecting better emotional self-regulation with (1) increased HRV (from a tonic and phasic perspective; for review see [3], for gender effects see [32,65]) as well as (2) increased interoceptive accuracy (e.g., [16,45]; for gender effects see [30]). Pinna and Edwards [41] elaborated on this integrative perspective of HRV, interoceptive accuracy, and emotion regulation in a systematic review. This review illustrated that, although reasonable, a link between HRV and interoceptive accuracy was not often reported.
An exception is the study of Lischke et al. [31], which indicated a small to moderate association between interoceptive accuracy and increased HRV (but see [27,38]). In accordance with this cross-sectional finding, two recent transcutaneous vagus nerve stimulation (tVNS) studies coherently showed that vagal activity could enhance interoceptive performance [47,61]. Therefore, the present within-subjects studies, designed to stimulate participants HRV and interoceptive accuracy via a short-term fasting intervention of 16-h, might be well suited to investigate a potential link between autonomic changes (i.e., HR, HRV) and changes in interoceptive accuracy. First, the fasting intervention was expected to result in elevated vagally-mediated HRV and decreased HR. Second, we hypothesized that 16-h of fasting would increase interoceptive accuracy as assessed by two different interoception tasks (i.e., the heartbeat discrimination task and the heartbeat tracking task). Finally, we hypothesized that fasting induced increases in interoceptive accuracy might be more pronounced in individuals who showed fasting induced increases in HRV.

Participants
Twenty-four participants (14 women) with a mean age of 23.88 years (SD=3.00) were investigated. An a-priori calculated power analysis using the software GPower 3.1 [14] indicated that a sample size of 20 participants was required to detect a medium to large effect (f = 0.30) with a power of 0.80 and a type I error probability of 5%. Of note Herbert et al. [20] reported a large effect of short-term food-deprivation on interoceptive accuracy (f = 0.50) in a sample of 20 women. All participants were free of (self-reported) major psychiatric disorders, neurological diseases, and cardiovascular diseases. The mean BMI was 23.91 kg/m 2 (SD=3.15) and the mean waist to hip ratio (WHR) was 0.82 (SD=0.09). Participants had a mean systolic blood pressure of 123.80 mmHG (SD=16.22) and a mean diastolic blood pressure of 73.90 mmHg (SD=10.21). This study was approved by the authorized local ethics committee and all participants provided written informed consent. They were requested to refrain from caloric consumption at least 16-h prior to their lab appointment and to arrive at the laboratory well rested.

Study design
A within-subject design was applied with condition (control vs. 16hour fasting) as independent variable and affect (positive, negative), interoceptive accuracy (assessed with the heartbeat tracking task and the heartbeat discrimination task), and cardiac activation (HRV, HR) as dependent variables. Participants were randomly allocated to the condition order.

Electrocardiogram (ECG)
The ECG was recorded with a Biopac MP150 amplifier system (1000 Hz) running AcqKnowledge 4.3 (standard lead II configuration). The Rwaves were identified with the Accusync® 72 ECG Trigger Monitor, which sent R-wave contingent triggers to the computer running the heartbeat discrimination task (PsychoPy; [39]). Auditory stimuli were presented via stereo loudspeakers approximately 2 m in front of the participants, who sat in a separated quiet and light attenuated room in a comfortable chair (1 m in front of a computer screen). The instruction conformity was monitored by the investigator via two cameras (one at the front and one at the back of the participant). In contrast to Herbert et al. [20], the ECG was recorded during the two interoception tasks, which allows to assess the task-specific association between changes in interoceptive performance and physiology.

Cardiac interoception tasks
2.1.4.1. Heartbeat tracking task. Interoceptive accuracy (IA HT ) was assessed using the heartbeat tracking task [20,51], where participants are asked to mentally track their own heartbeat. We used seven different intervals of 20, 25, 35, 45, 55, 65, and 75 s (for similar time intervals see the meta-analysis of Hickman et al. [21]), which were randomly presented and separated by a self-paced answer period and a two second inter-trial interval. Participants were requested to count their own heartbeats. An acoustic start and stop signal indicated the beginning and the end of the counting phases. In accordance with previously published task instructions, participants were not allowed to take their pulse or to attempt any other manipulations that could facilitate the detection of heartbeats. Participants should type the number of heartbeats they counted via the keyboard. They were naïve about the length of the counting phases and got no performance feedback during the task. A heartbeat perception score was calculated as the mean score of all seven heartbeat perception intervals: This score theoretically varies between 0 and 1 (see Table 1 for descriptive statistics), with a score of 1 indicating absolute accuracy of heartbeat perception.

Heartbeat discrimination task.
Interoceptive accuracy (IA HD ) was measured by the heartbeat discrimination task. Auditory feedback of the participants' heartbeat was presented with either a minimal (230 ms) or prolonged (540 ms) delay [37,49,54,63]. Participants with good interoceptive skills generally judge minor delayed feedback (around 200 ms) as synchronous to their heartbeats (e.g., [64], see [22] for an overview). The task was to decide, after 10 tones (50 ms duration, see e. g., [37]) if the feedback accurately represented (was synchronous with) the own heartbeats or not. The participants gave their self-paced answer via keyboard. Interoceptive accuracy was assessed using 20 trials for the pre/post intervention, respectively. Interoceptive accuracy was indexed by D-prime [37,63]: Values around zero indicate random performance, while increasing values index more accurate performance.

ECG analysis.
The ECG signals were manually checked for artifacts by means of the software Kubios Premium version 2.2. Vagallymediated HRV was estimated in the time domain by means of the root mean square of successive RR interval differences (RMSSD), which was calculated for 3 min during both interoception tasks. RMSSD seems to be particularly suited to index vagally-mediated HRV (see, [55]). RMSSD values were transformed by the natural logarithm (ln) to account for skewness.

Short-term fasting intervention of 16-h
Participants were randomly allocated to start with one of two experimental conditions: either with (1) interoception with fasting or (2) interoception without fasting (i.e., control condition). After a detailed face-to-face instruction, participants received reminders via email one day before appointment. Participants were asked to avoid the consumption of food or caloric drinks and were only allowed to drink water during the 16-h of fasting. In the control condition participants could consume food and caloric drinks as usual. Before participants worked on one of the two interoception tasks, which were presented in random order, they were asked to rate their feelings of hunger and to indicate their affective wellbeing.

Hunger
Similar to Herbert et al. [20], perceived hunger was assessed using a four-point rating scale from 1 (not hungry at all) to 4 (extremely hungry).

PANAS
Affect was assessed by the German version of the Positive and Negative Affect Schedule (PANAS; [28,62]). Participants were requested to indicate how they feel on 20 adjectives using a five-point Likert scale (from 1=not at all to 5=very much so). Cronbach's α varied between 0.84 and 0.88 for positive affect (PA) and 0.73 and 0.95 for negative affect (NA).

Statistical analysis
In order to evaluate if (1) 16-h of fasting were associated with changes in autonomic function, two within-subjects ANOVAs were calculated comparing lnRMSSD and HR during the fasting and the control condition. Analogous ANOVAs were calculated in order to indicate if (2) fasting was associated with a change in interoceptive accuracy. One ANOVA was calculated for interoceptive accuracy in the heartbeat tracking task and the other for accuracy scores derived from the heartbeat discrimination task (i.e., IA HT , IA HD ). To evaluate if (3) the change of interoceptive accuracy from control to fasting condition was moderated by changes in HRV, an ANOVA was calculated with the change-score of HRV as the continuous between-subjects factor. Furthermore, a similar ANOVA was calculated with change of HR as a continuous between-subjects variable, separately for both interoception tasks.
In order to gain more information on the observed effects of fasting on interoceptive accuracy, we calculated three additional analyses. First, for a direct comparison of changes in interoceptive accuracy between the two interoception tasks, the z-transformed scores of each task (across control and fasting condition) were analyzed in an ANOVA with the within-subjects factors task (heartbeat tracking vs heartbeat discrimination) and condition (fasting vs. control). Second, we evaluated time order effects for all statistical comparisons, using appointment (first vs. second) instead of condition as within-subjects factor. This analysis aimed to rule out that the (experimentally controlled) order of assessment had an impact on the reported results. Third, we investigated if gender, age, BMI, and WHR had an impact on study results by adding these variables to the calculated ANOVAs.

Manipulation check
Participants rated the hunger during the period of fasting significantly higher than 0 (t (22) Table 1).

Additional sensitivity analyses.
We additionally analyzed the effects of order (first vs. second session) on vagal and autonomic activity (lnRMSSD and HR). Analyses revealed no significant order effect neither for lnRMSSD (ps ≥ 0.840) nor for HR during the heartbeat

Additional sensitivity analysis.
In order to examine if changes in interoceptive accuracy from control to the fasting condition depended on the respective interoception task, z-scores for each task were used as dependent variables and condition and task were treated as independent within-subjects' variables. This analysis indicated a significant and medium-sized main effect of condition (F(1,23)=11.71, p=.002,  η p 2 =0.34), suggesting that fasting went along with higher interoceptive accuracy across tasks. The non-significant interaction between condition and task (F(1,23)=0.00, p=.969, η p 2 =0.00) further suggests that the changes in task performance from control to fasting were comparable. In accordance with this, both tasks showed (small to) medium effects sizes ( Table 1). The non-significant main effect of task (F(1,23)=0.00, p = 1.00, η p 2 = 0.00) resulted from the z-transformation procedure, which led to a mean score of 0 and a standard deviation of 1 within each task (and across both conditions). Therefore, the non-significant main effect of task is due to the transformation procedure and allows a straightforward interpretation of results. We additionally analyzed the effects of order (first vs. second session) on interoceptive accuracy. Performance on the heartbeat tracking task tended to be better (i.e., higher accuracy) in the second as compared to the first session (t(23)=− 1.98, p=.060), which suggest some training effects for the heartbeat tracking task. No effect of order was observed for the heartbeat discrimination task (t(23)=− 0.26, p=.801).
The ANCOVAs with age, gender, WHR, and BMI as covariates indicated virtually no changes of the fasten induced performance changes (see Supplemental Table 1). 2.4.4.12. Additional sensitivity analyses. Sensitivity analyses including age, gender, BMI, and WHR as co-variables showed virtually no changes of the significant interaction effects observed for lnRMSSD and HR for the heartbeat tracking task (see Supplemental Table 1). Interoceptive accuracy was not significantly associated with HRV and HR during control and fasting condition.

2.4.4.13.
Discussion of study 1. The aim of this study was to examine the impact of a voluntary fasting period of 16-h on interoceptive accuracy and autonomic function. As hypothesized, the fasting condition went along with better interoceptive accuracy [20]. This effect was significant for the heartbeat tracking task and marginally significant for the heartbeat discrimination task and hence extends previous results of Herbert et al. in several ways. First, similar to highly controlled food-deprivation interventions, which compared no-food intake with a specific caloric intake, this study was able to show facilitative effects on interoceptive accuracy by means of a 16-h of fasting. The fasting intervention applied in the present study provides a higher ecological validity and applicability for future studies conducted in everyday life. Secondly, this study was conducted in a randomized within-subjects design, thus allowing to control for order effects. Thirdly, in contrast to Herbert et al. [20] and Schulz et al. [53] both genders were investigated, which argues for some generalizability of the findings. Thirdly, this study indicated a fasting-related increase of interoceptive accuracy by means of two distinct tasks. Importantly, additional analyses indicated that both tasks were equally sensitive to track improvements in interoceptive accuracy from the control condition to the fasting condition with effects of medium sizes. Therefore, 16-h of fasting seemed to improve the ability to adequately monitor internal bodily processes, specifically cardiac activity, which has been considered health relevant (e.g., [44]). A further health relevant effect of fasting was the increase of vagally-mediated HRV (see e.g., [59]) from the control to the fasting condition, which is in line with studies showing increased vagal activity after skipping breakfast in children ( [42,43]; but see [20] for a decrease in adults). Furthermore, the HRV-increase was associated with increases in interoceptive accuracy as assessed by the heartbeat tracking task. In line with Thayer et al. [57], this finding may indicate that increases in vagal activity facilitate the accuracy with which internal bodily signals are perceived [31,41] and are therefore in nice accordance with studies showing better interoceptive accuracy after tVNS [47,61].
Of note, this finding is in contrast to Herbert et al. [20], who reported decreases in HRV after food-deprivation on the second study day, thus suggesting vagal withdrawal (see e.g., [53] for no changes in HRV). They interpreted the improvements in cardiac interoceptive accuracy in terms of increased ANS activity, thus paralleling findings of increased interoceptive accuracy following physical activity [52]. However, Herbert et al. assessed HRV during a baseline period, while this study assessed HRV during the performance of the task. Irrespective of this difference, further studies are certainly needed in order to evaluate the relationship between interoceptive accuracy and autonomic function in more detail.
Although the sample size of study 1 was sufficient to detect relatively large effects of fasting on interoceptive accuracy and cardiac activity, it might have been too small to reliably estimate effect sizes as reflected by the large confidence intervals (see Table 1). Therefore, study 2 followed two main aims. First, we targeted to replicate the findings of study 1 with a larger sample. Second, we were interested in a more accurate estimation of the effects of fasting on autonomic and interoceptive functions.

Participants
Seventy-two participants took part in the second study. Due to technical problems, a final sample of 70 participants was available for all analyses during both the heartbeat tracking and the heartbeat discrimination task. The sample was large enough to detect a medium effect with a power of 0.80 and a type I error of 5%. The age of participants was between 18 and 30 years (M = 22.24, SD=2.36). In total 31 men and 38 women were investigated (one participant rated diverse as the most appropriate gender). The mean BMI was 23.53 kg/m 2 (SD=4.14), the mean WHR was 0.82 (SD=0.10), mean systolic blood pressure was 116.66 mmHg (SD=14.78), and mean diastolic blood pressure was 65.76 mmHg (SD=7.36).

Procedure and measures
The measures, questionnaires, and the procedure were the same as used in study 1. The only change was that participants were requested to rate the confidence in perceiving their own heartbeat on a visual analogue scale from "total guess" to "complete confidence" [17] for each trial in the heartbeat tracking task and the heartbeat discrimination task, respectively. This procedure would allow to assess participants' metacognition (i.e., the skill to perceive when they showed good interoceptive accuracy and when they showed lower accuracy). However, this was not at the main focus of the present study. All assessments took place in the same laboratory as study 1; however, with other investigators. Study 1 was conducted by the first author (CR) and study 2 by the third (AA) and fourth authors (LB).

Manipulation check.
Similar to study 1, the mean hunger rating was significantly higher as compared to 0 (t(69)=17.19, p < .001; M = 2.10, SD=1.02), which indicated that participants had a relatively strong feeling of hunger before the laboratory session following a fasting period. Both paired samples t-tests comparing affective wellbeing between conditions were not significant (PA: t(69)=− 1.34, p=.184, d=− 0.16; NA: t(69)=1.05, p=.296, d = 0.13), however, the pattern of findings was similar as observed in study 1 (see, Table 2).

Effects of fasting on changes in cardiac activity during the interoception tasks
3.1.4.15. Heartbeat tracking task. Similar to study 1, the effect of condition was significant for lnRMSSD (F(1,69)=5.85, p=.018, d = 0.29), indicating a small to medium-sized effect. As illustrated in Table 2, lnRMSSD increased from the control to the fasting condition (see, Fig. 4 for the distribution of lnRMSSD scores). A similar medium-sized effect was observed for HR (F(1,69)=16.19, p<.001, d=− 0.48), with a lower mean HR during fasting in contrast to control (see Table 2).  Table 2). Furthermore, in the fasting condition HR was lower during the heartbeat discrimination task in contrast to the control condition (F (1,69)=13.70, p<.001, d=− 0.44, descriptive statistics see, Table 2).

Additional sensitivity analyses.
The additional inference statistics investigating order effects (first vs. second session) on vagal activity (i.e., RMSSD and HR) were not significant (ps ≥ 0.580), thus suggesting that reported effects could be reliably attributed to the fasting intervention. The ANCOVAs with age, gender, WHR, and BMI as covariates indicated virtually no changes of the fasten induced performance changes (see Supplemental Table 2).

Effects of fasting on interoceptive accuracy.
The fasting condition failed to show increased interoceptive accuracy as assessed by the heartbeat tracking task in contrast to the control condition (F(1,69)= 0.72, p=.398, d = 0.10, see Fig. 5). However, similar to study 1, the heartbeat discrimination task showed a significant effect suggesting that interoceptive accuracy was higher after the fasting condition in contrast to the control condition (F(1,69)=7.01, p=.010, d = 0.32; see Fig. 5).

Additional sensitivity analyses.
In order to examine if changes in interoceptive accuracy from control to fasting depended on the task used, z-transformed scores for each task were used as dependent variables and condition served as a within-subjects factor. This ANOVA indicated a significant effect of condition (F(1,69)=5.73, p=.019, η p 2 =0.08) of small to medium size and a significant interaction between condition and task (F(1,65)=4.31, p=.042, η p 2 =0.06), documenting that the heartbeat tracking task showed a significantly weaker effect due to fasting than the heartbeat discrimination task (see, Fig. 5). The ANCO-VAs with age, gender, WHR, and BMI as covariates indicated virtually no changes of the fasten induced performance changes (see Supplemental  Table 2). Note. IA=interoceptive accuracy, HD=heartbeat discrimination task, HT=heartbeat tracking task. All effect sizes and CIs were estimated by means of the R package MOTE (version 1.0.2; [4]).
However, the fasting induced changes in interoceptive accuracy in both tasks were significantly associated (r = 0.29, p=.014; see Fig. 6), suggesting that both tasks shared approximately 8% of the variance in fasting-induced interoceptive accuracy changes. Hence, both tasks seemed to track fasting induced changes in interoceptive accuracy, although the heartbeat tracking task seemed to be less sensitive.
The additional inference statistics investigating order effects (first vs. second session) on interoceptive accuracy were not significant (ps ≥ 0.128), thus suggesting that reported effects could be reliably attributed to the fasting intervention. Furthermore, there was no significant difference between the fasting and the control condition for time of day the appointment took place (t(69)=− 1.18, p=.240).

Moderating effect of fasten induced HRV/HR changes of vagally mediated HRV on fasten induced changes in interoceptive accuracy
3.1.5.1. Heartbeat tracking task. In accordance with study 1, the ANOVA with performance change in the heartbeat tracking task as continuous independent variable indicated that the changes in lnRMSSD were associated with changes in interoceptive accuracy (F(1,68)=4.34, p=.041, η p 2 =0.06; Fig. 7, left panel). The ANOVA with performance change in the heartbeat tracking task as continuous independent variable and HR as dependent variable showed an analogous, but nonsignificant interaction effect of condition and performance change (F (1,68)=3.26, p=.076, η p 2 =0.05).

Heartbeat discrimination task.
The ANOVA with performance change in the heartbeat discrimination task as continuous betweensubjects variable showed no association with the changes in cardiac

Discussion of study 2.
The results of study 2 were in large parts comparable with the findings of study 1. The only striking difference between these two independent studies was that the effect of fasting was smaller (and not significant), when interoceptive accuracy was assessed by means of the heartbeat tracking task. Nevertheless, fasting-induced effects were quite coherent in study 1 and study 2. First, effects of both studies pointed toward the same direction and suggest that a 16-h fasting period could increase interoceptive accuracy [20], lower HR, and elevate vagally-mediated HRV [29,43]. Second, changes in vagally-mediated HRV (from control to fasting) were positively linked with changes in interoceptive accuracy (assessed by the heartbeat tracking task) in both studies (see [47] for similar effects after tVNS). Thirdly, interoceptive accuracy in the heartbeat discrimination task showed increases with a comparable effect size in both studies from control to the fasting condition (see [61] for comparable effects after tVNS).
3.1.5.5. General discussion. The present study aimed to investigate if a naturally implemented 16-h fasting intervention is associated with changes in cardiac function and interoceptive accuracy. This assumption was based on findings of Herbert et al. [20], who studied the effects of a well-controlled food deprivation on interoceptive accuracy in a women only sample. Furthermore, we investigated if induced increases in interoceptive accuracy might be more pronounced in individuals who showed fasting induced increases in HRV. This assumption based on findings showing a positive association between interoceptive accuracy and HRV [31] and studies showing improvements of heartbeat perception after experimental vagal stimulation (i.e., tVNS; [47,61]). In the present study, we assessed (cardiac) interoceptive accuracy by two established paradigms in two consecutive and independent studies. We found that short-term fasting of 16-h appears well suited to alter the autonomic regulation of the heart, reducing HR and increasing vagally-mediated HRV. Furthermore, fasting was found to benefit interoceptive accuracy.
In contrast to Herbert et al. [20], who rigorously controlled compliance of the 22 women in the food deprivation condition, we  investigated the direct effects of 16-h fasting, which might constitute a more ecological and valid approach to study effects of starving in a voluntary, self-paced context relevant for intermittent fasting. Some might argue that investigating fasting in a more voluntary and less controlled setting is not objective and reliable. However, by means of this approach, we consistently found fasting-induced effects on feeling hunger, autonomic regulation, and interoception. Importantly, the reduction of HR and increase of HRV is in line with studies reporting short-term effects of food deprivation and food intake on cardiac activity [19,23,29,42,43].
Remarkably, the fasting-induced change in vagally-mediated HRV and interoceptive accuracy were intercorrelated, when using the heartbeat tracking task. This association is in line with a study of [34] Machado et al. (2009), who applied a between-person design. They reported lower HR in good heartbeat perceivers as compared to poor perceivers. Furthermore, these findings are in accordance with the neurovisceral integration theory of Thayer and Lane [58], which suggests interconnections between vagal activity and the prefrontal cortex, the insula, and other cortical and subcortical areas mandatory for interoception and emotion regulation [41,56,57]. Grounding on this framework the fasting induced increase of vagal activity might have enhanced interoceptive skills as assessed by the heartbeat tracking task. This interpretation is nicely in line with a study of Villani et al. [61], which indicated increased interoceptive accuracy after a non-invasive transcutaneous auricular vagus nerve stimulation (for similar results see [47]).
Critically, it could be argued that the finding of an association between fasting-induced changes in vagal activity and interoceptive accuracy might constitute a methodological artifact of the heartbeat tracking task itself. HRV is inversely related with HR [18] and recent studies indicate that performance in the heartbeat tracking task is strongly influenced by underreporting of own heartbeats ( [12,69]; but see [1]). Specifically, the interoceptive accuracy score derived from the heartbeat tracking task seems to be driven by the persons' decision threshold in reporting heartbeats [9,12,69]. Since fasting increased participants' HRV and decreased HR, the degree of underreporting would be lowered, thus resulting in better interoceptive accuracy [61,66]. Hence, the mere change in cardiac activity despite unaltered threshold (to report a heartbeat), implicit beliefs about the own HR, and interoceptive skills could explain the pattern of findings for the heartbeat tracking task. However, the association between (fasten related) changes in interoceptive accuracy and cardiac changes were only significant for HRV and not for HR in study 2, which might argue for a somewhat more robust effect of vagally-mediated HRV. Additionally, lower HR might go along with a higher signal intensity (e.g., due to a higher stroke volume given a constant cardiac output). If interoception accuracy increases because the cardiac signal became stronger, the conclusion of improved interoceptive processing might be challenging due to the possibility that interoceptive accuracy is confounded with cardiac characteristics (see [9] for similar argumentation).
However, the findings of study 1 and study 2 argue against the dominant effect of cardiac signal intensity on the increase of interoceptive accuracy assessed by the heartbeat discrimination task, since the performance changes were not associated with changes in the cardiac parameters at all. This nicely illustrates that using two measures of interoceptive accuracy allowed to examine the effects of fasting in a more holistic manner. Of note, the performance changes in both tasks were moderately correlated and shared 8% of the variance in interoceptive accuracy. This finding might indicate that fasting could change participants' interoceptive skills and modulate the ability to count individual heartbeats and discriminate between synchronous and asynchronous external stimuli. This effect seems to be driven by the observed autonomic changes [27]. However, it is important to consider that fasting did not lead to interoceptive accuracy increases in every participant; some even showed decreases. The same holds true for the cardiac changes, where some participants also showed decreases in HRV and increases in HR. Hence, future studies are warranted to analyze potential moderators of these effects. However, this observation is in accordance with our assumption that HRV changes should moderate the fasten induced increase of interoceptive accuracy, and those participants showing increases in HRV from control to fasting should benefit most.
Additionally, the fasting period did not evoke strong changes in affect. Potential and probably weak effects might be restricted to PA (lower during fasting). However, this effect needs further investigations since it was not significant in either study. The fragile and weak effects on wellbeing strengthen the applicability of this approach in future studies, since it might induce specific changes in participants' autonomic function and cardiac interoception without manipulating subjective well-being. This is an advantage to more controlled studies potentially inducing negative affective states [20].
The present research is not without limitations. First, it is important to note that the present findings are only valid for short-term effects of 16-h fasting and not generalizable to longer-term fasting, which might be very different in nature. In particular, it has been found that a more extended fasting period of 48-h could result in heightened sympathetic activity (Mazurak et al. [36]; for similar results after 72-h of fasting see, [8]). Mazurak et al. found a decrease of HRV from the beginning to the end of fasting in a small sample of young women, but in accordance with our findings they reported a slight (however, non-significant) HRV-increase from the beginning to the midterm of the fasting intervention (24-h) in a resting condition. Hammoud et al. [19] found no long-term effects of Ramadan fasting (from week 1 to week 4) on HRV and HR in a gender-diverse sample but indicated significant short-term effects within a day. Specifically, they reported higher vagally-mediated HRV (i.e., RMSSD) and lower HR during fasting periods as compared to food-intake periods. This pattern of findings is well in line with the present results and further strengthens our assumptions that 16-h of fasting constitutes an objective, reliable, and powerful tool in research context to (1) investigate the short-term effects of intermittent fasting and (2) experimentally manipulate participant's autonomic state (i.e., vagal activity) associated with interoceptive accuracy [31,61].
A further limitation of the present study is the applied version of the heartbeat discrimination task. First, it consisted of 20 trials, which might have been too few to reliably assess interoceptive accuracy. Kleckner et al. [26] suggested 40 trials for the heartbeat discrimination task in order to reveal reliable accuracy scores. Second, Brener and Ring [7] criticized the two alternative forced choice discrimination task, since people largely differ in the way they perceive delayed feedback as synchronous with their heartbeats or not. Nevertheless, Wiens and Palmer [64] indicted that most individuals perceive intervals of about 200 ms as synchronous and shorter as well as longer intervals as asynchronous [6]. Grounded on this evidence, the applied heart beat discrimination task (and the heartbeat tracking task) are in accordance with available cardiac interoception research (for a meta-analysis providing an informative overview of task specifications of studies applying both the heartbeat discrimination and the tracking task see [21]).
In line with this observation, we were able to document an increase in performance from control to fasting and a significant association between these increases in the heartbeat discrimination task and the heartbeat tracking task, thus suggesting robustness of this assessment. Associated with the application of multiple interoception tasks is the need of a higher number of statistical tests, which might inflate the alpha-error. However, it should be noticed that we tested three well defined hypothesis (separately for heartbeat tracking task and heartbeat discrimination task), which were partly replicated in two independent samples. The high number of statistical analyses was due to additional sensitivity analyses, which should evaluate the robustness of findings.
To conclude, the present studies indicate that voluntary and selfpaced 16-h of fasting could be well suited to manipulate peoples' autonomic regulation (i.e., cardiac activity). Furthermore, we found changes in interoceptive accuracy by the application of two tasks. First, the interoceptive accuracy assessed by the heartbeat discrimination task increased from control to fasting condition [61]. Second, the accuracy change measured by the heartbeat tracking task was moderated by the change in vagally-mediated HRV. This replicated pattern of findings supports the notion that using different measures of interoceptive accuracy might help to assess changes of cardiac sensations more precisely and completely [49,61].