Martial Arts and Cognitive Control in Children with Attention-Deficit Hyperactivity Disorder and Children Born Very Preterm: A Combined Analysis of Two Randomized Controlled Trials

ABSTRACT

Introduction

Very preterm birth and attention-deficit hyperactivity disorder (ADHD) are associated with impairments in response inhibition that often persist beyond childhood. Athletes skilled in martial arts show a neurocognitive profile that is associated with an improved inhibition processing stream, suggesting that engagement in this kind of sport has the potential to reduce impairments in this cognitive function. We investigated the behavioral and neurocognitive effects of judo training on response inhibition in children born very preterm and children with ADHD by a combined analysis of two randomized controlled trials.

Methods

In both the CHIPMANC (n = 65) and JETPAC (n = 63) studies, participants were randomly allocated to a waitlist or a 12-wk judo training program in a 1:1 ratio. At pretest and posttest, participants completed a Go/NoGo task, the Movement Assessment Battery for Children-2 and a physical work capacity test on a bicycle ergometer. During the cognitive task, event-related potentials (N2, P3a, P3b) were recorded via electroencephalography.

Results

The effects of the judo training were moderated by the study group. In contrast to children with ADHD (JETPAC), judo training reduced the commission error rate on the Go/NoGo task and increased the P3a amplitude in children born very preterm (CHIPMANC). No treatment effects were found for N2, P3b and physical fitness outcomes.

Conclusions

The neurodevelopmental condition influences the cognitive benefits of judo training. Whereas judo may be ineffective in children with ADHD, children born very preterm can expect improved response inhibition due to a more effective engagement of focal attention to resolve the task-related response conflict.

The prevalence of preterm birth (<37 wk of gestation) increased since 1990 for most countries and the global estimate is now 10% (1). Complications related to preterm birth are the leading cause of mortality in infants younger than 5 yr (2), but the survival rate has increased over the last three decades (3). Individuals born very (≤32 wk of gestation) or extremely (≤28 wk of gestation) preterm, who survive the first years, often face a wide range of neurodevelopmental disabilities that contribute to adverse long-term outcomes (4,5). In this respect, meta-analytical findings support that children born preterm are more often affected by behavioral problems, impairments in motor skills, and poor academic performance (6). In addition, preterm birth is associated with a high risk of neurodevelopmental disorders and attention-deficit hyperactivity disorder (ADHD) in particular (7,8). The prevalence of ADHD increases with decreasing gestational age, resulting in a twofold risk in children born very preterm (9).

Attention-deficit hyperactivity disorder is a childhood disorder characterized by developmentally inappropriate and impairing levels of hyperactivity, impulsivity, and/or inattention that become apparent before 12 yr of age (10). In the cognitive domain, children with ADHD show pronounced executive function (i.e., top-down control of behavior) deficits (11), which in turn explain a unique proportion of variance in symptom severity (12). A quantitative synthesis suggests that children and adolescents born preterm are also affected by impairments in executive function (13). This negative effect of preterm birth was independent from age, indicating that children do not necessarily compensate their developmental delay during adolescence. Although (especially very) preterm birth and ADHD may cause distinct executive function profiles, poor response inhibition (i.e., failure to suppress a prepotent motor response) is a common characteristic of both conditions that even persists across development (14,15). The association of impaired response inhibition and unhealthy behaviors, e.g., substance abuse (16), internet addiction (17) and binge eating (18), renders this cognitive function an important target for intervention.

Evidence suggests cognitive benefits of exercise in healthy populations, with meta-analytical findings supporting programs with high coordinative demands to be most effective across all age groups (19). The high effectiveness of coordinative exercise types may be due to a promotion of motor skills, which in turn are related to different aspects of executive function (20). The potential of exercise to reduce response inhibition deficits is further supported on mechanistic level. Using event-related potentials (ERP) recorded via electroencephalography, evidence suggests that preterm birth and ADHD cause abnormalities in cognitive control processes. When compared with healthy peers, both conditions were related to a diminished N2 amplitude that was elicited from an inhibition task in adolescents (21). The N2 is a negative-going wave with a frontal distribution and its facilitation reflects changes in monitoring stimulus or response conflict (22). Moreover, preterm birth and ADHD have further been related to diminished amplitudes of components of the P3 family during a Go/NoGo task paradigm (23), with higher decreases in P3a and lower decreases in P3b in children with ADHD. Although P3a indexes frontal lobe activity required for stimulus evaluation, the P3b reflects temporoparietal activity associated with the attentional resource allocation (24). Consequently, the abnormalities in the N2–P3 complex indicate common problems in conflict monitoring, but more pronounced impairments in stimulus evaluation in ADHD and a preterm birth-related lack of flexible modulation of attention. Exercise acts on these cognitive control processes (25) and has much potential to reduce associated executive function deficits in children with neurodevelopmental disabilities (26). In this respect, previous findings indicate that motor skills and response inhibition are linked indirectly via P3a amplitude (27). Motor skills also have an important role for children born (very) preterm, given that they fully mediated the cognitive control deficit indicated by a decreased P3 amplitude (28).

Among different exercise types, martial arts summarize disciplines that emphasize the acquisition and refinement of motor skills, thus having much potential to improve response inhibition. This is further supported by ERP findings from athletes experienced with martial arts. The examination of their neurocognitive profile has revealed a higher negativity of the N2 amplitude and an increased P3b (29). A high P3b in particular does not only differentiate athletes practicing martial arts from inactive individuals but also from athletes engaging in other sports types (30). Neuroimaging findings further suggest that experts in martial arts also seem to be able to adapt frontal lobe recruitment better to task demands than novices, as they showed more pronounced differences in P3a source activity between trials with low and high cognitive load (31). Although evidence on the association of martial arts and the N2–P3 complex has been generated by cross-sectional studies in healthy adults, a single randomized controlled trial (RCT) provides a first indication for a causal relation. In this experiment, children allocated to a judo training program showed a greater improvement in response inhibition and increased N2 negativity than those allocated to a wait-list (32). Consequently, martial arts and judo in particular may reduce response inhibition deficits related to preterm birth and ADHD by acting on abnormalities in the N2–P3 complex that are common across both groups. Recalling that children with ADHD display higher aggression in sports (33), judo is considered appropriate as it bears the lowest risk of injury among different types of martial arts (34).

By a primary analysis of two RCTs, we examine the effects of judo training on behavioral and neurocognitive indices of response inhibition in both children born preterm and those with ADHD. Based on previous research on the neurocognitive profiles of martial arts athletes, we expected judo to improve response inhibition and to facilitate the N2–P3 complex. A secondary, explorative objective was the comparison of treatment effects between studies to examine whether potential benefits of judo generalize to different conditions, which share deficits in cognitive control.

METHODS

Participants

Two parallel studies were conducted to investigate the effects of judo training on response inhibition in children born very preterm (Study 1, Children Born Preterm, Martial Arts & Neurocognition [CHIPMANC]) and children with ADHD (Study 2, Judo Exercise Therapy and Pharmacotherapy in ADHD Children [JETPAC]). The recruitment procedures, the study design, the intervention, and (most of) the assessments were overlapping to ensure a high comparability. In both studies, participants were recruited from local clinics in Basel and Aarau, Switzerland, as well as Lörrach, Germany. Legal guardians of participants received a formal information letter that was followed up by a telephone call. Participants had to express their willingness to participate and their legal guardians had to provide written informed consent before the screening procedures. Specific eligibility criteria in CHIPMANC were birth ≤32 wk of gestation, age 8 to 14 yr, and an intelligence quotient of 85 or higher. In JETPAC, participants had to be diagnosed with ADHD (without comorbid Autism Spectrum Disorder) according to DSM-5 criteria (10), age 8 to 12 yr and treated with methylphenidate or dexamphetamine within the last 3 months (with no dosage change in the last month and no additional therapies). The treatment with stimulants was selected as inclusion criterion to reduce the variability in impairment severity among children with ADHD. Common eligibility criteria across both studies were right handedness, corrected-to or normal vision, no regular participation (once a week) in martial arts 3 months before the intervention, no structural epilepsy, no injuries and or diseases affecting the functionality of the right hand, as well as being free of any chronic disease classified as contraindication for exercise. Based on a moderate effect size and an alpha level of P = 0.05, a priori sample size calculations (Supplemental Digital Content, http://links.lww.com/MSS/C778) indicated that 56 participants were required to detect a treatment effect (using repeated-measures ANOVA) with 85% power within the CHIPMANC and JETPAC studies, respectively. The study protocols were preregistered at the German Registry of Clinical Trials (CHIPMANC: DRKS00017067; JETPAC: DRKS00020125), approved by the local ethics committee and complied with the guidelines of the Declaration of Helsinki and its amendments. The reporting of our studies followed the Consolidated Standards of Reporting Trials (see Table S1, Supplemental Digital Content, CONSORT checklist, http://links.lww.com/MSS/C778).

Study design

Recruited participants were randomly assigned (stratum: age in CHIPMANC; medication dose in JETPAC) to a judo intervention group (INT) and a wait-list control group (CON) in a 1:1 ratio. A researcher not involved in the studies generated the randomization lists (variable block sizes: 4, 6) with sealed envelope™ (London, UK). Sequentially numbered opaque envelopes were used to conceal the allocation. After the pretest, participants were notified of their group assignment via email or telephone by the principal investigator. During both pretests and posttests, they completed a computerized Go/NoGo task with simultaneous electroencephalographic recordings, a submaximal cycling ergometer test and the Movement Assessment Battery for children–2 (MABC-2). In addition, the Intelligence and Development Scales–2 (IDS-2) were administered and participants filled in the Strengths and Difficulties Questionnaire (SDQ) as well as the Family Affluence Scale (FAS). Pretests and posttests had an identical order of the assessments and were completed at the same time of the day. Laboratory personnel conducting the tests was not informed on the participants’ group allocation. The data collection period took place from 2019 to 2022.

Intervention

Participants in INT completed a 60-min judo session twice a week over a period of 12 wk. This dose constellation was chosen as previous studies prescribing a similar total volume of judo training reported improvements in executive function among healthy children (32,35). The training followed the guidelines of the Swiss Judo Confederation (https://sjv.ch) for introducing judo to beginners and their preparation for the first belt examination. The judo sessions emphasized the acquisition and refinement of techniques for attack and defense (including injury prevention). This learning of basic techniques (30–40 min) was complemented by playful and child-appropriate exercises that aimed to improve physical fitness (10–15 min). Each judo session also included Randori (10–15 min), which is judo-specific sparring while standing or on the ground. During this specific form of practice, complex techniques and physical fitness were trained in combination. To assess the intensity of the judo program, participants were asked to rate perceived exertion on the child-appropriate RPE scale immediately after the training on at least four sessions. Each Judo session included blocks of children-appropriate and playful physical fitness training, technique learning, and their application in Randori (free fighting on the ground or while standing).

Cognitive task

Response inhibition was assessed with a computerized Go/NoGo task (32), which used gray geometric shapes (presented against black background) as visual stimuli. The contour lines of these shapes were either yellow, pink, or blue (color saturation and intensity was matched) and reflected three different trial types. Participants were instructed to withhold their response to NoGo trials (30%), but press a button on a serial response box (Celeritas, PST, USA) when presented with frequent (40%) and standard Go trials (30%). Two types of Go trials were implemented to allow the comparison of ERP elicited by NoGo and Go trials that is less confounded by differences in probability. Reaction time and accuracy were emphasized equally during instructions. After an interstimulus interval that varied randomly between 1200 and 1400 ms, geometric shapes were presented focally for 150 ms. The sequence of trial types was randomized. After 30 practice trials, three blocks with 100 trials each were administered. E-Prime 3.0 (Psychological Software Tools, USA, PA) was used for stimulus presentation and response logging. For statistical analyses, average reaction time on response-correct standard Go trials as well as commission error rate (i.e., pressing a button in response to NoGo trials) and omission error rate (i.e., pressing no button in response to Go trials) were calculated.

EEG recording and processing

Sixty-four active electrodes were arranged according the 10:10 system and mounted to the participant’s head using a flexible cap. After the impedance was reduced to 10 KΩ or lower (on at least 90% of all electrodes) by applying highly conductive gel, data were amplified with the ActiCHamp (BrainProducts, Germany), band-pass filtered (0.01–100 Hz), and recorded at 500 Hz. Cz served as online reference and AFz as ground. Offline data processing was performed with BESA Research 7.1 (Brain Electric Source Analysis, Germany) and followed the steps described in a previous protocol (32). In short, automatic adaptive artifact correction was applied to blinks after their detection from virtual VEOG and HEOG channels. Subsequently, data were baseline-corrected (−200 ms to stimulus onset), submitted to low-pass filters (zero-phase shift of 30 Hz; slope 24 dB per octave) and high-pass filters (forward phase shift of 0.1 Hz; slope 6 dB per octave) and re-referenced to the average of mastoids. In segments covering the latency range from −200 to 800 ms relative to stimulus onset, uncorrected artifacts were rejected using individual gradient (75.1 ± 2.3 μV) and amplitude thresholds (118.6 ± 21.3 μV). Separate averaged segments were created for standard Go and NoGo trials with correct responses. Frequent Go trials were disregarded as their higher probability limited the comparability with NoGo trials. The selection of latency ranges and regions used to extract the ERP components of interest were guided by previous studies with children that used the same task (27,28). Within the latency range of 400 to 650 ms relative to stimulus onset, the P3a (elicited by NoGo trials) was extracted as the mean amplitude at the frontocentral (FCz, FC1, FC2, Cz, C1, C2) region and the P3b (elicited by Go trials) as the mean amplitude at the centro-parietal (CPz, CP1, CP2, Pz, P1, P2) region. The N2 (elicited by NoGo trials) was derived from the mean amplitude at the frontal region (average of AFz, AF3, AF4, Fz, F1, F2) within 250 to 350 ms after stimulus onset. The ERP components were derived using mean amplitudes, given that this method is most robust against increased noise (36), which has to be expected in EEG assessments with children born very preterm and peers with ADHD.

Physical fitness and motor skills

Participants completed the PWC170 test on a children-appropriate cycling ergometer (E200P, COSMED, Italy). From 20 W (body weight < 50 kg) or 30 W (body weight ≥ 50 kg) at the beginning of the test, the workload was increased every 2 min until the participant reached a heart rate of at least 165 bpm, was unable to maintain the target pedaling cadence or unwilling to continue. The increment was based on the heart rate within the last 10 s of each stage. The PWC170 has appropriate test–retest reliability in children, and the 2-min protocol we used shows the highest correlation with the maximal oxygen uptake (37).

The MABC-2 included tasks related to manual dexterity (three items), balance skills (three items) and aiming/catching (two items) (38). Based on the age of participants at pretest (≤11 or ≥12 yr), the appropriate version was selected. After the conversion of raw scores from each dimension into age- and gender-corrected standard scores, they were combined to yield an overall score.

Statistical analyses

SPSS Statistics 28.0 for Windows (IBM, Armonk, NY) was used for statistical analyses. Zero-order correlations were calculated to investigate the relation between potential confounders (age, sex, body mass index [BMI], IDS-2, FAS, SDQ), cognitive performance and ERP components. To increase the precision of the estimation of treatment effects, variables showing a statistically significant correlation or a correlation of at least r = (−)0.2 were entered as covariates in subsequent main analyses. Pearson correlations were also calculated to investigate the association between cognitive performance and ERP components within pretests and posttests. The treatment effects were examined by using 2 (time: pretest, posttest) × 2 (group: CON, INT) × 2 (study: CHIPMANC, JETPAC) ANOVA on cognitive performance, ERP components and physical fitness outcomes. When omission and commission error rates were analyzed, change in reaction time (posttest–pretest) was included as additional covariate to account for a speed-accuracy trade-off. Our models examined the interaction of time and group (to address the primary objective) as well as the interaction of time, group and study (to address the secondary, explorative objective). Statistically significant three-way interactions were followed-up by investigating the interaction of time and group within each study. The level of statistical significance was set to P < 0.05. Effect sizes were considered small, medium, and large at η² ≥ 0.01, η² ≥ 0.06, and η² ≥ 0.14, respectively.

RESULTS

As shown in the flow diagram (Fig. 1), data from 113 participants remained for the final analysis. The comparison of participants’ characteristics at pretest revealed lower FAS score and higher SDQ score in the JETPAC compared with the CHIPMANC sample (Table 1). The compliance to the intervention was 82.1% ± 11.6% and 81.3% ± 11.3% in CHIPMANC and in JETPAC, respectively. Children born very preterm (29.5 ± 2.0 wk of gestation) rated the intensity of the judo 13.0 ± 0.6 on the RPE scale (ratings were possible in the range from 6 to 20). An RPE of 13.3 ± 0.7 was reported by children with ADHD (receiving 28.5 ± 12.9 mg methylphenidate/dexamphetamine per day). The results of the preliminary analyses showed that age, sex, BMI and IDS-2 were associated with at least one of the outcomes (see Table S2, Supplemental Digital Content, Correlation between potential confounders, behavioral performance and ERP components, http://links.lww.com/MSS/C778), so that they were accounted for in subsequent main analyses. The findings of the repeated measures ANOVA investigating treatment effects are summarized in Table S3 (see Supplemental Digital Content, Overview of interactions of time, group, and study for cognitive performance, event-related potential components, and physical fitness, http://links.lww.com/MSS/C778).

FIGURE 1.

Flow of study participants in the JETPAC and CHIPMANC studies.

TABLE 1.

Comparison of participants’ characteristics at pretest.

	CHIPMANC (n = 28 f, 28 m)				JETPAC (n = 15 f, 42 m)
	INT (n = 29)		CON (n = 27)		INT (n = 30)		CON (n = 27)		CHIPMANC vs JETPAC
	M	SD	M	SD	M	SD	M	SD	F
Age, yr	10.7	2.0	10.4	1.8	10.0	1.2	10.8	1.2	0.24
Body mass, kg	37.3	9.0	33.6	9.6	38.3	14.1	38.8	9.1	2.25
BMI in kg·m−2	17.5	2.4	16.7	2.8	18.1	4.2	18.3	2.7	3.44
IDS-2	11.9	6.6	13.3	6.3	13.0	4.6	10.3	3.3	3.55
FAS score	6.5	1.4	6.7	1.8	5.6	1.4	5.9	1.6	8.77*
SDQ sum score	10.9	5.9	13.3	4.6	15.4	4.3	16.3	5.6	15.24*

*P < 0.05 for the comparison between CHIPMANC and JETPAC.

Behavioral performance

For Go reaction time and omission error rate, there were no interactions of time and group as well as time, group and study. Although there was also no interaction of time and group for commission error rate, a statistically significant interaction emerged for time, group, and study (F[1,104] = 6.93, P = 0.010, η² = 0.062). Decomposition of the interaction showed that time and group only interacted in CHIPMANC (F [1,52] = 5.62, P = 0.021, η² = 0.098), indicating a higher decrease of commission error rate over the intervention period in INT compared with CON (Table 2).

TABLE 2.

Pretest and posttest cognitive performance, components of ERPs and physical fitness in the intervention and control groups within the CHIPMANC and JETPAC studies.

		CHIPMANC (n = 28 f, 28 m)				JETPAC (n = 15 f, 42 m)
		INT (n = 29)		CON (n = 27)		INT (n = 29)		CON (n = 27)
		M	SD	M	SD	M	SD	M	SD
Go reaction time, ms	T1	443.2	82.3	446.9	75.7	439	74.9	412.2	85.9
	T2	445.1	69.2	427.3	84.4	463.4	70.7	417	85.1
Omission error rate, %	T1	0.06	0.07	0.06	0.07	0.08	0.10	0.09	0.09
	T2	0.05	0.07	0.04	0.06	0.05	0.09	0.08	0.13
Commission error rate, %	T1	0.35	0.23	0.32	0.16	0.37	0.20	0.40	0.18
	T2	0.24	0.16	0.30	0.18	0.31	0.22	0.30	0.22
N2 amplitude, μVa	T1	−0.66	5.06	0.56	4.42	−2.88	5.35	−2.74	6.28
	T2	1.21	6.84	0.90	6.28	−0.97	5.86	−2.80	5.42
P3a amplitude, μVa	T1	6.89	5.16	9.78	6.83	7.27	5.6	5.78	7.77
	T2	10.64	6.61	10.13	5.32	8.32	7.13	8.01	8.24
P3b amplitude, μVb	T1	9.87	6.54	10.97	5.57	11.47	6.43	9.37	5.78
	T2	9.79	5.50	9.29	5.65	10.53	6.62	9.82	6.54
MABC-2 overall score	T1	42.9	25.1	44.8	28.8	31.1	22.6	31.8	21.8
	T2	51.3	32.5	56.5	24.0	37.7	26.9	45.4	29.4
PWC170, W·kg−1	T1	2.07	0.59	1.85	0.41	1.99	0.73	1.83	0.54
	T2	2.08	0.59	1.94	0.41	1.70	0.56	1.68	0.42

^aElicited from NoGo trials.

^bElicited from Go trials.

T1, pretest; T2, posttest.

Event-related potentials

Whereas N2 was not associated with cognitive performance outcomes, higher P3a amplitude was correlated with a lower omission, −0.42 ≤ r ≤ −0.24, and commission error rate, −0.25 ≤ r ≤ −0.22, within pretests and posttests (Table 3). Similarly, a lower commission error rate was associated with a higher P3b amplitude at both time points, −0.39 ≤ r ≤ −0.23.

TABLE 3.

Correlation between behavioral performance on the Go/NoGo task and the N2, P3a as well as P3b components of ERP collapsed across intervention and control groups, as well as the CHIPMANC and JETPAC studies.

	1	2	3	4	5	6
1. N2 amplitudea	—	−0.07	0.31*	−0.10	0.08	0.12
2. P3b amplitudeb	0.10	—	0.53*	0.18	−0.14	−0.23*
3. P3a amplitudea	0.35*	0.56*	—	−0.05	−0.24*	−0.22*
4. Go RT	−0.04	0.41*	0.02	—	−0.09	−0.56*
5. Omission E	−0.11	−0.20*	−0.42*	0.07	—	0.31*
6. Commission ER	0.10	−0.39*	−0.25*	−0.62*	0.27*	—

Correlations within pretest are shown below the principal and correlations within posttest are shown above the principal.

^aElicited from NoGo trials.

^bElicited from Go trials.

*P < 0.05.

ER, error rate; RT, reaction time.

Regarding the intervention effects, no interactions of time and groups as well as time, group and study were found for N2 and P3b amplitudes (Fig. 2). Similarly, no interaction of time and group was found for P3a. However, the statistical analysis revealed an interaction of time, group, and study for this component of ERP (F[1,109] = 4.76, P = 0.031, η² = 0.042). The decomposition of the interaction revealed a time–group effect in CHIPMANC only, supporting a greater increase in P3a over the intervention period in INT compared with CON (Fig. 3; F [1,54] = 5.63, P = 0.021, η² = 0.094).

FIGURE 2.

ERP waveforms at Fz, Cz, and PZ elicited by Go and NoGo trials in the intervention and control groups across the CHIPMANC and JETPAC studies. The vertical bars indicate the latency ranges used for extracting the N2 and P3a/b components.

FIGURE 3.

Topographic distribution of the P3a amplitude at pretest and posttest in the control and intervention groups of the CHIPMANC study. Dots indicate the positions of the electrodes. The amplitude was averaged over the latency range from 400 to 650 ms relative to stimulus onset.

Physical fitness

For both PWC170 and MABC-2 overall score, no interactions of time and group as well as time, group, and study were found (Table 2).

DISCUSSION

Despite the prescription of the same exercise intervention to participants in JETPAC and CHIPMANC, the results differed between both studies. A treatment effect of the judo training on response inhibition was only supported in the CHIPMANC cohort, that is, in the group of very preterm children. This effect was moderate and indicated by a greater reduction of the commission error rate from pretest to posttest in INT compared with CON. Improved behavioral performance was accompanied by a moderate increase in P3a amplitude. In contrast, changes in behavioral and neurocognitive indices of response inhibition did not differ between INT and CON among participants of JETPAC. Consequently, benefits of judo for this cognitive domain are rather selective and do not generalize to children with ADHD.

Positive effects of exercise on inhibitory control have been reported in previous meta-analyses (39,40), but included studies mainly investigated the subdomain of interference control. Our findings add that response inhibition is also sensitive to structured exercise and that judo in particular has a normalizing effect (i.e., reduction of deficits) on this cognitive function, at least in children born very preterm. As the reduction of the commission error rate was statistically significant even after accounting for changes in reaction time, this effect was not simply due to a speed accuracy tradeoff. Our results are inconsistent with the lack of improvements in response inhibition reported by intervention studies, in which exercise programs that primarily targeted cardiorespiratory fitness were prescribed to children and adolescents without known neurodevelopmental disabilities (41,42). This discrepancy cannot only be attributed to a greater adaption reserve in children born very preterm, given that an experimental trial that examined the effects of a judo training program in healthy children also detected benefits for response inhibition (32). This suggests that response inhibition is sensitive to specific characteristics of judo that are less emphasized in exercise programs with a primary focus on cardiorespiratory demands. Recalling that motor skills have been suggested to mediate the association of very preterm birth and response inhibition (28), we expected improvements in motor skills following judo to translate into cognitive benefits. However, we observed no changes in motor skills and physical working capacity across both the CHIPMANC and JETPAC studies, meaning that improvements in this domain are no precondition for the normalization of response inhibition.

The ERP findings contribute to a better understanding of the mechanism by which judo training elicited benefits for response inhibition in children born very preterm. Based on the relation between ERP components and behavioral performance, higher amplitudes of P3a and P3b individually contributed to a lower commission error rate, whereas the N2 amplitude did not consistently explain variance in outcomes of the Go/NoGo task. The judo training increased the P3a, indicating changes in the inhibition processing stream. This change in P3a signals an increase in frontal lobe activity to engage focal attention, followed by the transmission of stimulus information to parietal locations, where the P3b occurs (24). In NoGo trials, focal attention is required for conflict resolution (by inhibiting the response) and the subsequent evaluation of inhibition processes (43). Consequently, judo training did not generally enhance the resource allocation indexed by the P3b but improved the upregulation of frontal lobe activity to cope with the conflict induced by NoGo trials. In the first place, this might be considered a compensatory effect as preterm birth has been associated with a more pronounced decrease in P3b relative to P3a amplitude (23). However, in our sample, the opposite pattern was observed at pretest, indicating that the treatment effect of judo can be better described as a normalization tendency. This benefit might have been elicited by the cognitive rather than the motor skill demands inherent in judo. The intervention promoted the acquisition of throwing and grappling techniques and their practice during Randori. In this judo-specific form free fighting, participants needed to maintain their own balance while off-balancing the opponent (44). This process constantly requires the assessment, execution, and reevaluation of the appropriate motor response while anticipating movements and techniques of the opponent. In a more abstract form, these processes are required during the Go/NoGo task (45,46). Consequently, Randori might have contributed to the observed treatment effect as it encompasses response inhibition training embedded in a sports context. This is a feature of judo that is generally shared with open-skill sports, which are performed in an unpredictable and dynamic environment, thus requiring constant adaptation to changing cognitive demands (47). In comparison to unspecific sports and closed-skill sports, which include self-paced activities in a predictable environment, engagement in open skill sports has been related to superior cognitive benefits (48), especially for the inhibitory aspect of executive function (49). The absence of improvements in MABC-2 score and PWC170, suggests that the benefits observed for response inhibition are not unique to judo, but rather shared with other cognitively demanding open-skill sports.

In contrast to the CHIPMANC cohort, participants in JETPAC showed no improvements in response inhibition despite a similar compliance rate and intensity rating. Comparing participants’ baseline characteristics between both study groups, children with ADHD recruited in JETPAC showed poor motor skills, lower socioeconomic status and increased psychopathology ratings on the SDQ. The difference in motor skills in particular might partly explain the lack of a treatment effect, given that basic motor skills need to be developed to allow full engagement in complex exercises, which refine sports-specific skills (50). Thus, learning complex judo techniques could have led to a higher perceived difficulty in children with ADHD, increasing the risk of excessive demands. This risk is further supported by neuroimaging findings. Although the initial acquisition of motor skills requires the activation of prefrontal cortex, this region shows a decreasing activation with increasing automaticity over time. (51) However, children with ADHD are characterized by a general task-related hypoactivation in networks associated with cognitive control (52). During the Go/NoGo task, this is reflected by a reduced prefrontal cortex activation (53). This region contributes to response inhibition and is involved in the generation of the P3a and P3b amplitudes (45). Based on these findings, it appears that the judo training program did not elicit benefits for behavioral and neurocognitive indices of response inhibition, because the intervention induced excessive demands in children with ADHD. These demands might have prevented the refinement of judo techniques in children with lower motor skills, especially under cognitively challenging conditions, such as Randori. Further support for this assumption is provided by findings from an experimental study that investigated acute effects of exercise. Although children with ADHD showed improved inhibitory control and an increased P3b amplitude after an aerobic exercise session, these benefits were canceled when the exercise session included coordinative and cognitive demands. In contrast, the canceling effect was not observed in healthy peers (54). Acute changes in components of the P3 family have been associated with the response to pharmacological treatment (55,56), suggesting that the lack of amplitude increases after coordinatively challenging exercise indexes a nonresponse to this type of exercise. Consequently, we assume that the complexity of judo might have prevented the elicitation of benefits for response inhibition in children with ADHD within the 12-wk intervention period, suggesting that for this population, exercise needs to be tailored to individual cognitive and motor abilities.

Despite the use of RCT design and a mechanistic approach to investigate the effects of exercise on response inhibition, some limitations need to be considered for the interpretation of our results. First, the procedures in CHIPMANC and JETPAC were almost identical, but there were slight differences in inclusion criteria and the randomization stratum. These differences can affect the baseline comparability between INT and CON within and across studies. However, we reduced potential bias by adjusting our main analyses for baseline variables that were associated with the outcomes. Second, using the PWC170 and MABC-2, we could not detect an effect of judo training on aerobic fitness and motor skill across both studies. Thus, the observed benefits for response inhibition might be independent from changes in these aspects of physical fitness. Alternatively, the cognitive benefits could rather be related to a refinement of judo-specific skills, which were not measured with the applied instruments. Third, engagement in other activities that may have the potential to improve response inhibition have not been monitored, so that it remains unclear if the observed treatment effect is solely due to the judo training. However, findings of a meta-regression suggest that the type of control group does not influence the cognitive effects of exercise (19), indicating a robust treatment effect. Fourth, the study used a Go/NoGo paradigm to elicit specific ERP components, but this type of task only allows for an abstract measurement of response inhibition. The use of or combination with a more ecologically valid assessment in future studies may help to predict how improved response inhibition translates to performance in real-life domains. Fifth, the low-to-moderate inverse correlation between P3a and commission error rate suggests that changes in the processes indexed by this ERP component do only partly explain the cognitive benefits of judo training in children born preterm. However, we only investigated cognitive control processes, so that the potential contributions of other mechanisms, such as molecular/cellular signaling and mental states (57), remain unclear. Sixth, we found a moderating effect of very preterm birth on the treatment effect of judo, underlining a condition-specific effect. As we only included children with very preterm birth and children with ADHD, we cannot conclude whether the beneficial effect of judo or nonresponse to judo generalizes to children with neurodevelopmental disabilities.

CONCLUSIONS

A judo training program that seeks to prepare participants for their first belt examination elicits different responses among children with neurodevelopmental problems. In contrast to children with ADHD, children born very preterm improve their ability to inhibit a motor response after an intervention period of 3 months. This early adaptation to judo training is detectable, even when there are no changes in motor skills and aerobic fitness. On a neurocognitive level, the effect of judo on response inhibition in children born very preterm is supported by an increase of the P3a amplitude, indicating a change toward a more effective engagement of focal attention for resolving a task-related response conflict. Other cognitive control processes, such as conflict monitoring and the more general allocation of attentional resources, do not seem to be sensitive to judo training.