Effect of short-term abdominal breathing on heart rate variability as an indicator of emotional regulation in college student with internet gaming disorder

Abstract

Online games are becoming increasingly popular, but a corresponding problem has emerged: Internet gaming disorder (IGD). IGD refers to problematic game use, where gaming results in problems with cognition and emotional regulation. Abdominal breathing (AB) is one method of psychophysiological reaction regulation. However, few studies have explored the effect of AB on gamers with IGD. In this study, we investigated the effects of short-term AB training (10 min) on the heart rate variability (HRV) of college students with IGD as they watched positive and negative online gaming videos. In total, 21 and 19 participants with low-risk IGD and high-risk IGD (HIGD), respectively, were included in the analysis. The results revealed that AB training was associated with increases in the natural logarithms of the total power and low-frequency HRV of the HIGD group during both video stimuli (p < .01). The difference in the natural logarithm of total power and low-frequency HRV between before and after AB was a predictor of IGD risk (area under the curve = 0.63 and 0.64, respectively). We find that short-term AB affects the HRV responses of college students with HIGD during game-related stimuli, particularly for negative games. These findings highlight the potential benefits of adding AB training to psychotherapies to improve the psychophysiological regulation of college students with IGD. Further studies should investigate the effect of long-term AB on the psychophysiological responses of those with IGD during gameplay.

Keywords:

• Abdominal breathing
• emotional regulation
• heart rate variability
• Internet gaming disorder

In 2021, the number of gamers worldwide was approximately 3.24 billion (41% of the world population) (Clement, 2021b). Due to the COVID−19 pandemic, the number of gamers has grown rapidly (Ko & Yen, 2020). Some gamers are addicted to gaming, and such problematic gaming can adversely affect their interpersonal relationships, educational opportunities, or job performance (King et al., 2013). After observing the clinically significant harm resulting from problematic game use, members of the American Psychiatric Association made a preliminary inclusion of IGD into section III of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (American Psychiatric Association & American Psychiatric Association Press Inc, 2013). The World Health Organization has also included gaming disorder as a type of addictive behavior disorder in the International Classification of Diseases, 11th Revision (Jo et al., 2019). The majority of gamers are aged 18–34 years in the United States (Clement, 2021a). College students are particularly at high-risk for Internet addiction (Chang et al., 2015) and IGD (Yao et al., 2015) because they can freely and easily access the Internet. Thus, the psychophysiological responses of college students with IGD, particularly in the current COVID−19 era, merit careful examination.

In terms of psychological responses, studies have noted that gamers with IGD exhibit greater emotional dysregulation and dysfunctional cognitions, including impulse control, distressed personality, anxiety, depression, and aggression, compared with healthy gamers (Hollett & Harris, 2020; Kim et al., 2016; Laier et al., 2018; Mehroof & Griffiths, 2010; Park et al., 2019; Yen et al., 2018). Individuals with IGD also have difficulty dealing with stress and negative emotions (Petry et al., 2014; Young & Brand, 2017). For psychophysiological responses, the emotional regulation and IGD risk have been widely investigated using the time- and frequency-domain heart rate variability (HRV). HRV is the variance in time between the heartbeats, and HRV-related indicators include standard deviation of all normal RR intervals (SDNN), low-frequency (LF), high-frequency (HF), total power (TP), and LF/HF ratio (Shaffer & Ginsberg, 2017). The RR interval is the time interval between two R waves of the ECG signal. Both SDNN and LF represent sympathetic nervous system (SNS) and parasympathetic nervous system (PNS) activities. HF and TP represent PNS and overall autonomic nervous system activities, respectively. The LF/HF ratio reflects the balance between SNS and PNS activities. In gamers, a significant negative association was observed between distressed personality and TP or LF; these gamers’ IGD symptoms were also negatively associated with the SDNN, TP, and LF (Kim et al., 2016). Gamers with problematic Internet use exhibited decreased HRV while playing the game Bomb Defender (Chang et al., 2015). In two other studies, gamers with IGD (Hong et al., 2018 Lee et al., 2018) exhibited reduced HF while playing the game League of Legends, suggesting that those with IGD lacked executive control. Compared with healthy gamers, gamers with IGD exhibited lower HF at baseline (Park et al., 2019). Data from several studies exhibited that individuals with IGD tend to show negative emotions and have lower PNS activity. However, the relationship between emotions and HRV for college students with IGD is still unclear.

For psychophysiological regulation, slow breathing has been used to modify emotions and HRV responses for individuals with mental disorders. Individuals with anxiety who performed diaphragmatic breathing over 8 weeks exhibited lower anxiety symptoms and heart rates than individuals with anxiety without diaphragmatic breathing (Y.-F. Chen et al., 2017). The diaphragmatic breathing is inhalation and exhalation by the relaxation and contraction of diaphragm muscle (Y.-F. Chen et al., 2017). In another study, patients with depression exhibited decreased depression and anxiety symptoms and increased TP HRV after 2 weeks of HRV biofeedback (Siepmann et al., 2008). In a study by Caldwell and Steffen (2018), patients with depression exhibited greater decreases in depression symptoms and increases in SDNN after 6 weeks of HRV biofeedback combined with psychotherapy in comparison with patients with depression who only received psychotherapy. Similarly, patients with depression exhibited increases in SDNN, LF, and TP after HRV biofeedback (Lin et al., 2019). Tripathi (2019) suggested that the breathing exercises of yoga can be used to relax the mind and improve cardiorespiratory function in individuals with Internet addiction.

Although slow breathing is considered an effective method for easing negative emotions and increasing PNS activity for those with mental disorders, few studies explore the effect of slow breathing on gamers with IGD during emotional stimuli. For emotion-eliciting material, videos have been widely used to arouse negative and positive emotions (Gross & Levenson, 1995). We thus investigated emotional responses and HRV of college students with low-risk IGD (LIGD) and high-risk IGD (HIGD) during online game video stimuli with and without abdominal breathing (AB). AB, a slow breathing, involves contracting the diaphragm and abdominal muscles to move the abdominal wall to increase ventilation and speed gas exchange (Fox, 2011). In the present study, we proposed two hypotheses. First, the AB training would increase the HRV of college students with HIGD during online game-video stimuli. Second, college students with HIGD would exhibit different HRV from college students with LIGD, particularly during negative emotional video stimuli.

1. Materials and methods

1.1. Participants

A total of 50 college students (23.20 ± 3.65 years old; 36 males) with no history of bipolar or panic disorder or cardiovascular diseases were recruited from a public university, and they were gamers. A small sample was used because of the difficulty in obtaining psychophysiological data for college students. The sample size of the aforementioned empirical studies is from 41 to 48 (Chang et al., 2015; Hong et al., 2018; Lee et al., 2018). For comparison between two groups, the sample size needs to be at least 23 for each group by setting a power value of 0.8, the effect size (Cohen’s d) of 0.85, and a confidence level of 0.95 (Faul et al., 2009).

1.2. Questionnaire

To classify participants into the HIGD and LIGD groups, the IGD questionnaire (IGDQ) (Petry et al., 2014) and Chen Internet Addiction Scale (CIAS) (S.-H. Chen et al., 2003) were applied. The IGDQ was used to assess the participants’ online gaming experience. The IGDQ contains nine items that are each scored on a dichotomous scale as 0 (false) or 1 (true). The CIAS is the most well-known questionnaire for assessing Internet addiction risk. The CIAS contains 26 items that are rated on a 4-point Likert scale from 1 (does not match my experience at all) to 4 (definitely matches my experience). On the basis of the cut-off proposed for the IGDQ and CIAS (S.-H. Chen et al., 2003; Petry et al., 2014), the participants were classified into the LIGD (IGDQ score <5 and CIAS score <64) and HIGD (IGDQ score ≥5 and CIAS score ≥64) groups. The reliabilities of the CIAS and IGDQ were calculated using Cronbach’s alpha. In this study, the Cronbach’s alpha values for the IGDQ and CIAS were 0.81 and 0.94, respectively.

To assess the emotional type and emotional intensity of participants, the self-assessment manikin (SAM) (Lang, 1980) was applied in this study. The SAM is the most common tool for rating emotional valence and arousal, both of which are rated on a 9-point Likert scale. An emotional valence score of 5 indicates a neutral emotional type. A higher valence score indicates positive emotions in the participant. Emotional arousal ranged from 1 (calm) to 9 (excited).

1.3. Game video

Playing the massively multiplayer online role-playing and first-person shooter games can improve the different cardiovascular responses of gamers with IGD (Metcalf & Pammer, 2013). The League of Legends (a massively multiplayer online role-playing game, Riot Games, Los Angeles, US) can elicit both negative (frustration, anger, and sadness) and positive emotions (happiness and enjoyment) (Kou & Gui, 2020). The Resident Evil game (a massively multiplayer online third-person shooter, Capcom Co., Ltd., Osaka, Japan) was considered a horror and violent game (Krahé & Möller, 2004). MapleStory (a massively multiplayer online role-playing game, Wizet, Seoul, South Korea) can attract gamers to elicit their desire and concentration (En & Lan, 2010). Therefore, the gameplay videos of League of Legends, Resident Evil, and MapleStory were used to arouse gamers’ emotions. Six videos (2 min per video) were included in this study, and each game type had two videos, and the content of these videos were the record of gamer playing games.

1.4. Experimental procedures

Gas exchange is most efficient at a respiratory frequency of approximately 0.1 Hz, and this respiratory frequency induces the greatest oscillation in respiratory sinus arrhythmia (Lehrer & Gevirtz, 2014; Shaffer et al., 2014). Respiratory sinus arrhythmia is connected with PNS activity. Therefore, the respiratory frequency of 0.1 Hz is commonly adopted to execute slow breathing training (Caldwell & Steffen, 2018; Lin et al., 2019). We used short-term AB at six cycles per minute for 10 min for the slow breathing training. During AB training, participants placed their hands on the abdomen and then concentrated on abdominal muscle contractions and movements; the force of the abdominal muscles pushed the abdominal wall in an inward or outward direction.

To maintain steady-state physiological conditions, participants were instructed not to consume caffeine or alcohol before the experiment. Participants completed the experiment on their own in a laboratory environment. Figure 1 presents the experimental procedure. At the beginning of the experiment, the participant received an explanation of the procedures and provided signed informed consent. Subsequently, the participant was asked to provide their demographic information and complete the CIAS. Thereafter, the participant executed protocol I (representing the state before AB training), AB training, and protocol II (representing the state after AB training) sequentially (Ji & Hsiao, 2019). Each protocol contained three trials. During each trial, the participant was tested under baseline, stimuli, and recovery states sequentially, and each state was tested for 2 min. When stimulated, the participant watched one video of the aforementioned three games in a random order. In protocols I and II, participants watched three types of gameplay videos. Following the stimuli, the participants completed the SAM. In the final stage of the procedure, the participant completed the IGDQ.

Figure 1. The experimental procedure.

In the experimental procedure, the participants executed protocol I, abdominal breathing training of 10 min, and protocol II sequentially. Each protocol contained three trials, and each trial contained the baseline, watching one video game, filling out the questionnaire, and recovery states. Each state was tested for 2 min except for filling out the questionnaire.

1.5. Data analysis

To measure cardiovascular responses, an electrocardiogram (ECG, BEST-C−04056, BioSenseTek Corp., Taiwan) was used to detect the electrical activity changes of the heartbeat through three Ag/AgCl electrodes fixed over the surfaces of the right subclavian, left subclavian, and left lumbar regions. We also measured the respiratory wall movement of the thorax and abdomen through respiratory inductive plethysmography (RIPmate Inductance Belt Abdomen Kit, Adult, Alice 5, Ambu, Denmark). ECG and respiratory signals were acquired using DAQCard (USB 6218, NI Corp., Austin, TX, USA) at 1 kHz. The RR intervals were extracted from the ECG signal using the peak detector method, and those intervals were interpolated at 1 kHz. The following frequency-domain HRV values were obtained from the interpolated RR intervals by using the auto power spectrum method: LF (0.04‒0.15 Hz), HF (0.15‒0.4 Hz), TP (0.04‒0.4 Hz), and LF/HF ratio. These HRV parameters were analyzed during the baseline, stimuli, and recovery states in protocols I and II. We also calculated the differences between protocols I and II for lnLF, lnHF, lnTP, and the LF/HF ratio, referred to as ΔlnLF, ΔlnHF, ΔlnTP, and ΔLF/HF ratio, respectively. All signal collection and analysis were performed using LabVIEW software (V. 2019, NI Corp., Austin, TX, USA).

Statistical analyses were performed using SPSS version 22 (IBM Corp, Armonk, NY, USA). Between-protocol differences during the baseline, stimuli, and recovery states for the SAM score and HRV parameters (lnLF, lnHF, lnHF, and LF/HF ratio) of both the HIGD and LIGD groups were tested using paired-sample t tests. To compare the difference between participants with LIGD and HIGD in the CIAS, IGDQ, and SAM scores and HRV parameters, two-sample t tests were employed. A group (HIGD and LIGD, between) × state (baseline, stimuli, and recovery, within) × game type (League of Legends, Resident Evil, and MapleStory, within) × protocol (I and II, within) analysis of variance (ANOVA) was conducted using the HRV parameters. Moreover, a group (HIGD and LIGD, between) × state (baseline, stimuli, and recovery, within) × game type (League of Legends, Resident Evil, and MapleStory, within) analysis of variance (ANOVA) was also examined using the ΔlnLF, ΔlnHF, ΔlnTP, and ΔLF/HF ratio. Post-hoc analysis was evaluated using Bonferroni corrections. To investigate whether the ΔlnLF, ΔlnHF, ΔlnTP, and ΔLF/HF ratio can be used to differentiate between the LIGD and HIGD groups, a receiver operating characteristic (ROC) curve was employed; the area under the curve (AUC) was used to determine the cutoff score, sensitivity, and specificity. The p-value of AUC was estimated using the Mann—Whitney U test (Hajian-Tilaki, 2013). Statistical significance was set at the 5% level.

2. Results

One participant had a CIAS score lower than 64 and a IGDQ score higher than 4. Nine participants had CIAS scores higher than 63 and IGDQ scores lower than 5. These 10 participants were excluded for further analysis. In total, 21 participants (22.62 ± 1.79 years old; 12 men and 9 women) and 19 participants (23.67 ± 5.27 years old; 15 men and 4 women) were classified into the LIGD and HIGD groups, respectively. Age and sex were no significant differences between the two groups. The LIGD and HIGD groups differed significantly in terms of their IGDQ (IGDQ_LIGD = 1.48 ± 1.22, IGDQ_HIGD = 6.58 ± 1.46, p < .001) and CIAS (CIAS_LIGD = 52.10 ± 7.84, CIAS_HIGD = 79.26 ± 8.40, p < .001) scores.

For online game videos, one participant with HIGD and one participant with LIGD who never heard of Resident Evil and one participant with LIGD who never heard of MapleStory. Table 1 presents the emotional valence and arousal of the two groups for the three games tested under protocols I (before AB) and II (after AB). The results revealed that the League of Legends and MapleStory trials induced positive emotions, whereas the Resident Evil trial induced negative emotions. The emotional arousal levels of both groups during the three trials decreased after AB training, albeit nonsignificantly so.

Table 1. Mean ± standard deviation of emotional valence and arousal at three trials

trial	emotion	HIGD			LIGD
		protocol I M ± SD	protocol II M ± SD	p	protocol I M ± SD	protocol II M ± SD	p
LOL	valence	6.63 ± 1.30	6.26 ± 1.33	0.247	5.95 ± 1.32	5.95 ± 1.16	1.000
	arousal	5.26 ± 1.48	4.47 ± 2.04	0.069	4.33 ± 2.01	3.86 ± 2.20	0.353
RE	valence	4.37 ± 1.57	3.68 ± 1.20	0.067	3.90 ± 1.87	3.81 ± 1.97	0.605
	arousal	5.63 ± 1.71	5.37 ± 1.92	0.583	6.14 ± 1.80	5.67 ± 2.13	0.125
MS	valence	5.58 ± 1.43	5.47 ± 1.78	0.786	5.76 ± 1.79	5.48 ± 1.25	0.561
	arousal	4.16 ± 1.71	4.05 ± 2.01	0.794	4.33 ± 1.88	3.10 ± 1.87	0.025

Note: MS: MapleStory, LOL: League of Legends, RE: Resident Evil

Figure 2 illustrates an overview of the means (bar) and standard deviations (error bar) of the HRV parameters for the HIGD and LIGD groups during the baseline, stimuli, and recovery states in the three trials in protocols I and II. No significant differences between the HIGD and LIGD groups were found for all the HRV parameters during the baseline. We observed increases in lnTP, lnLF, lnHF, and LF/HF ratio for both groups after AB training. For the HIGD group, lnTP and lnLF during the stimuli state for the three trials in protocol II were significantly higher than in protocol I (p < .01). For the LIGD group, a significant increase in lnTP and lnLF during the baseline state for the three trials after AB training was observed (p < .01). However, the difference in HRV parameters during three states in protocols I and II between the LIGD and HIGD groups was nonsignificant.

Figure 2. The HRV parameters for the HIGD and LIGD groups at three trials. IB: baseline in protocol I, IS: stimuli in protocol I, IR: recovery in protocol I, IIB: baseline in protocol II, IIS: stimuli in protocol II, IIR: recovery in protocol II, †: p value < 0.05 for LIGD group, *: p value < 0.05 for HIGD group.

The bar graph with error bar represents lnLF, lnHF, lnTP, and LF/HF values for the HIGD and LIGD groups at the baseline, stimuli, and recovery states in protocols I and II. lnTP and lnLF of the HIGD group during the stimuli state in protocol II are significantly higher than in protocol I. The LIGD group shows a significant increase in lnTP and lnLF during the baseline state in protocol II.

Table 2 displays ANOVA for lnLF, lnHF, lnHF, and LF/HF ratio. The results show a significant effect from the group in lnHF (F(1,682) = 10.280, p = .001) and LF/HF ratio (F(1,682) = 8.769, p = .003), and the LIGD group had higher lnHF but lower LF/HF ratio than did the HIGD group. A significant effect from the protocol was observed in lnTP (F(1,682) = 47.887, p  < .001), lnLF (F(1,682) = 81.219, p  < .001), and LF/HF ratio (F(1,682) = 92.443, p  < .001), with lower lnTP, lnLF, and LF/HF ratio during protocol I than during protocol II. A significant effect from the state was revealed in LF/HF ratio (F(2, 682) = 5.557, p = .004), indicating that LF/HF ratio during the baseline was higher than during video stimuli. However, the game type presented a nonsignificant effect. The group–protocol interaction (F(1, 682) = 9.487, p = .003) and state–protocol interaction (F(1, 682) = 3.479, p = .031) effects were observed in LF/HF ratio.

Table 2. ANOVA for lnTP, lnLF, lnHF, and LF/HF ratio

Source	lnTP			lnLF			lnHF			LF/HF ratio
	F	p	ges	F	p	ges	F	p	ges	F	p	ges
Group	3.411	0.065	0.005	1.178	0.278	0.002	10.280	0.001	0.015	8.769	0.003	0.013
State	0.979	0.376	0.003	2.267	0.104	0.007	0.126	0.881	0.000	5.557	0.004	0.016
Type	0.850	0.428	0.002	0.875	0.417	0.003	0.781	0.459	0.002	0.121	0.886	0.000
Protocol	47.887	<0.001	0.066	81.219	<0.001	0.106	2.668	0.103	0.004	92.443	<0.001	0.119
Group × State	0.112	0.894	0.000	0.130	0.878	0.000	0.373	0.689	0.001	0.647	0.524	0.002
Group × Type	0.056	0.946	0.000	0.017	0.983	0.000	0.100	0.905	0.000	0.399	0.671	0.001
Group × Protocol	2.962	0.086	0.004	2.710	0.100	0.004	0.646	0.422	0.001	9.487	0.003	0.013
State × Type	0.114	0.978	0.001	0.258	0.905	0.002	0.057	0.994	0.000	0.094	0.984	0.001
State × Protocol	0.746	0.475	0.002	0.953	0.386	0.003	0.597	0.551	0.002	3.479	0.031	0.010
Type × Protocol	1.168	0.312	0.003	1.463	0.232	0.004	0.603	0.548	0.001	0.628	0.534	0.002
Group × State × Type	0.142	0.967	0.001	0.101	0.982	0.001	0.140	0.968	0.001	0.069	0.991	0.000
Group × State × Protocol	0.114	0.892	0.000	0.061	0.940	0.000	0.106	0.900	0.000	0.793	0.453	0.000
Group × Type × Protocol	0.048	0.953	0.000	0.251	0.778	0.001	0.159	0.853	0.000	0.558	0.573	0.002
State × Type × Protocol	0.115	0.977	0.001	0.117	0.976	0.001	0.166	0.956	0.001	0.197	0.940	0.001
Group × State × Type × Protocol	0.097	0.983	0.001	0.148	0.964	0.001	0.098	0.983	0.001	0.148	0.964	0.001

Note: ges: generalized eta squared,

Table 3 provides the results obtained from the ANOVA analysis of ΔlnTP, ΔlnLF, and ΔlnHF. A significant group effect was observed in ΔlnTP (F(1,340) = 9.791, p = .002), ΔlnLF (F(1,340) = 7.504, p = .006), and ΔLF/HF ratio (F(1,340) = 9.500, p = .002), indicating that the HIGD group showed higher ΔlnTP, ΔlnLF, and ΔLF/HF ratio than the LIGD group. A significant game type effect was also showed in ΔlnTP (F(2,340) = 3.774, p = .024) and ΔlnLF (F(2,340) = 3.992, p = .019). A significant state effect was presented in ΔLF/HF ratio (F(2,340) = 3.611, p = .028) with higher ΔLF/HF ratio during the baseline than during game video stimuli. The ΔlnTP (p = .024) and ΔlnLF (p = .019) during video stimuli of League of Legends was significantly lower than during video stimuli of Resident Evil. The ROC curve analysis results are presented in Table 4. We noted that ΔlnTP and ΔlnLF in the Resident Evil trial were the predictors of IGD risk.

Table 3. A ANOVA for the ΔlnTP, ΔlnLF, ΔlnHF, and ΔLF/HF ratio

Source	ΔlnTP			ΔlnLF			ΔlnHF			ΔLF/HF ratio
	F	p	ges	F	p	ges	F	p	ges	F	p	ges
Group	9.791	0.002	0.028	7.504	0.006	0.022	1.805	0.180	0.005	9.500	0.002	0.027
State	2.408	0.092	0.014	2.582	0.077	0.015	1.918	0.149	0.011	3.611	0.028	0.021
Type	3.774	0.024	0.022	3.992	0.019	0.023	1.652	0.193	0.010	0.657	0.519	0.004
Group × State	0.433	0.649	0.003	0.178	0.837	0.001	0.448	0.639	0.003	0.809	0.446	0.005
Group × Type	0.132	0.876	0.001	0.658	0.519	0.004	0.578	0.562	0.003	0.593	0.553	0.003
State × Type	0.357	0.839	0.004	0.309	0.872	0.004	0.438	0.781	0.005	0.182	0.948	0.002
Group × State × Type	0.298	0.879	0.003	0.394	0.813	0.005	0.315	0.868	0.004	0.163	0.957	0.002

Note: ges: generalized eta squared, ΔlnHF: difference in natural logarithm of high-frequency between protocols I and II, ΔlnLF: difference in natural logarithm of low-frequency between protocols I and II, ΔLF/HF ratio: difference in high-frequency/low-frequency ratio between protocols I and II, ΔlnTP: difference in natural logarithm of total power between protocols I and II

Table 4. ROC curve analysis of HRV parameters at three trials

trial	HRV	AUC	p	95 % confidence interval		cutoff score	sensitivity	specificity
				lower bound	upper bound
LOL	ΔlnTP	0.56	0.230	0.46	0.67	0.66	0.73	0.44
	ΔlnLF	0.54	0.442	0.44	0.65	0.16	0.43	0.73
	ΔlnHF	0.51	0.798	0.41	0.62	0.47	0.73	0.33
	ΔLF/HF ratio	0.57	0.178	0.47	0.68	1.88	0.44	0.70
RE	ΔlnTP	0.63	0.017	0.53	0.73	0.48	0.56	0.71
	ΔlnLF	0.64	0.009	0.54	0.74	1.05	0.75	0.55
	ΔlnHF	0.51	0.794	0.41	0.62	0.58	0.71	0.38
	ΔLF/HF ratio	0.55	0.375	0.44	0.65	3.77	0.20	0.92
SM	ΔlnTP	0.58	0.128	0.48	0.69	0.85	0.75	0.44
	ΔlnLF	0.56	0.241	0.46	0.67	0.53	0.51	0.67
	ΔlnHF	0.59	0.106	0.48	0.69	−0.06	0.56	0.69
	ΔLF/HF ratio	0.59	0.089	0.49	0.69	0.33	0.71	0.49

Note: MS: MapleStory, ΔlnHF: difference in natural logarithm of high-frequency between protocols I and II, ΔlnLF: difference in natural logarithm of low-frequency between protocols I and II, ΔLF/HF ratio: difference in high-frequency/low-frequency ratio between protocols I and II, ΔlnTP: difference in natural logarithm of total power between protocols I and II, LOL: League of Legends, RE: Resident Evil

3. Discussion

In this study, we investigated (1) the effect of short-term AB on the HRV of college students with HIGD and LIGD while they watched positive and negative online game videos and (2) whether those with HIGD would exhibit different HRV values from those with LIGD after AB training.

In both groups, we noted significant increases in lnTP, lnLF, and LF/HF ratio after AB training by ANOVA. In Figure 2, the HIGD group’s lnTP and lnLF increased significantly when they watched both positive and negative videos after AB training. However, the LIGD group’s lnTP and lnLF increased significantly during baseline after AB training. Although, these findings did not support our first hypothesis that performing AB would increase the HRV values of college students with HIGD, our finding is consistent with those of related empirical studies. Individuals with mental disorder exhibited increases in TP and LF after slow breathing (Lehrer et al., 2003; Lin et al., 2014, 2019; Siepmann et al., 2008). A possible explanation for this might be that the slow breathing affects neurophysiological mechanisms. Slow breathing increases vagal activity and promotes the neurotransmission of information to the limbic system, which is related to emotional processing (Fox, 2011), via vagal afferents; the information is then projected to the medulla oblongata for the modulation of autonomic functions, such as breathing rate and heart rate (Brown & Gerbarg, 2005). The increase in vagal or baroreflex activities would increase heartbeat oscillations and shift the overall HRV to the LF band (Lin et al., 2014; Siepmann et al., 2008), which explains why the TP and LF HRV values increase during or after slow breathing. Another explanation of our findings is that participants were not familiar with AB, and doing AB training may interfere with their respiratory control to increase SNS activity. We need to consider whether familiarity and unfamiliarity with AB training affect the psychophysiological responses of college students with IGD.

In contrast to the LIGD group, the HIGD group exhibited a lower lnHF and a higher LF/HF ratio. The ΔlnTP, ΔlnLF, and ΔLF/HF ratio from before to after AB training were significantly higher in those with HIGD than in those with LIGD. Moreover, ΔlnTP and ΔlnLF during video stimuli of Resident Evil were indexes to predict IGD risk. These findings support our second hypothesis that college students with HIGD would exhibit different HRV responses than college students with LIGD after AB training, particularly during negative emotional stimuli. Our findings may be explained by the relationship between emotional dysregulation and autonomic nervous system responses. Gamers with HIGD were prone to difficulties in negative emotional regulation (Hollett & Harris, 2020; Kim et al., 2016; Laier et al., 2018; Mehroof & Griffiths, 2010; Park et al., 2019; Petry et al., 2014; Yen et al., 2018; Young & Brand, 2017). Depression and anxiety symptoms caused the change of vagal modulation to inhibit the PNS activity that exhibited the decrease in HF value (Bleil et al., 2008; Henry et al., 2010). Autonomic dysregulation (Chang et al., 2015) and lower PNS activity (Park et al., 2019) were found in individuals with HIGD. Moreover, the long-term problematic Internet use (Hsieh & Hsiao, 2016; Lu et al., 2010) or excessive gaming behavior (Kim et al., 2016) resulted in high SNS activity. Slow breathing exercise was used to alleviate anxiety and depression symptoms in patients (Caldwell & Steffen, 2018; Y.-F. Chen et al., 2017; Siepmann et al., 2008). After the short-term slow breathing exercise, both the SNS and PNS were activated for patients with mental disorder, particularly SNS activity (Lehrer et al., 2003; Lin et al., 2014, 2019; Siepmann et al., 2008). Therefore, we infer that short-term AB training may generate greater effect on HRV responses and improvement in negative emotional regulation for college students with HIGD during negative emotional stimuli relative to those with LIGD.

The present findings suggest that short-term AB training can influence HRV responses in college students with HIGD and LIGD, particularly higher sympathetic activity of the HIGD group during negative emotional stimuli. Breathing training was used to assist in Mindfulness or cognitive behaviour therapy for enhancing the attention of those with HIGD to calm, improve self-cognitive control, and diminish their desires for gaming (Li et al., 2018; Tripathi, 2019; Young & Brand, 2017), and the increase in HRV may indicate improvements in emotional regulation (Caldwell & Steffen, 2018; Y.-F. Chen et al., 2017; Siepmann et al., 2008). Observed HRV responses during negative emotional stimuli with AB training may be used to classify gamers into the LIGD and HIGD groups. Therefore, AB was a valuable way to regulate emotions and physiological responses for college students with IGD.

Our study has six limitations. First, we employed AB training for 10 min to calm the psychophysiological responses of college students. Although the results indicated that the college students with HIGD had significantly higher lnTP and lnLF values than those with LIGD, the effect of long-term AB training on HRV should be investigated. Second, men and women may exhibit distinct psychophysiological responses. Although no significant difference in questionnaire scores was noted between the male and female participants in this study, sex difference in cardiovascular responses should be investigated. Third, our small sample size may have affected our results for the cardiovascular responses of the HIGD and LIGD groups. Further studies with larger samples should be conducted. Fourth, we did not use active stimuli (e.g., playing games) and were thus unable to observe the effects of AB training on students’ psychophysiological responses during gameplay. Fifth, we did not assess participants’ emotional regulation; thus, no conclusion can be drawn on the effect of AB on emotional regulation and the relationship between HRV and emotional regulation. Sixth, the kind of games participants mainly played may affect their psychophysiological responses. The study is limited by the lack of information on the record of the mainly played game. These limitations should be addressed in future research. Studies should also investigate the effect of long-term AB exercises on the psychophysiological responses of those with IGD during gameplay.

4. Conclusion

This is the first study to observe the effect of short-term AB training on the HRV responses of college students with IGD, while they watch positive and negative game videos. The results revealed that after training, individuals with HIGD had significantly increased lnTP and lnLF values during both film stimuli. Moreover, the ROC curve indicated that ΔlnTP and ΔlnLF (the difference in lnTP and lnLF values between before and after AB, respectively) was a strong predictor of IGD risk. These findings suggest that short-term AB training may affect the HRV responses of college students with HIGD when they are subjected to online game stimuli, particularly negative emotional game stimuli. These findings may also help clarify the effect of AB on the psychophysiological regulations of college students with HIGD. Further studies should explore the long-term effect of AB on the psychophysiological responses of those with HIGD.

Acknowledgements

The authors thank all participants who assisted in this study. This manuscript was edited by Wallace Academic Editing.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The entire ECG-IGD database is now freely available in the Open Science Framework. https://osf.io/5xcgq/?view_only=5c9f51a828a344669c8087a2fee76a2c

Additional information

Funding

This work was supported by the Taiwan Ministry of Science and Technology (grant numbers: MOST 110-2222-E-A49-010 and MOST 109-2221-E-A49-001-MY3).

Notes on contributors

Hung-Ming Chi

Hung-Ming Chi, Ph.D. is an assistant professor of the Department of Computer Science at National Yang Ming Chiao Tung University. Her research interests concern psychophysiological responses and brain images of individuals with Internet addiction and Internet gaming disorder.

Tzu-Chien Hsiao

Tzu-Chien Hsiao, Ph.D. is an associate professor of the Department of Computer Science at National Yang Ming Chiao Tung University. He is engaged in research that explores sports sciences and physiological mechanisms. His research interests include the research and development of machine learning algorithms and instantaneous frequency analysis.