Using MKDA and ES-SDM allowed the current meta-analysis to characterize the neural correlates and age-related effects on different subcomponents of inhibitory control. We observed brain areas including the inferior frontal gyrus, insula, middle cingulate cortex and the inferior parietal gyrus are activated across two subcomponents. Contrast analyses to elucidate the distinct neural substrates for each subcomponent revealed that relative to response inhibition, cognitive inhibition produced stronger activation in the left superior parietal lobule, while response inhibition primarily recruited the right inferior frontal gyrus, insula, middle temporal gyrus and angular gyrus. Importantly, by performing a meta-regression analysis with age as a continuous variable, we found distinct age-related activation patterns in different subcomponents of inhibitory control in brain regions including the left anterior cingulate cortex, the left inferior frontal gyrus, the left superior parietal lobule and bilateral insula. Overall, our results indicate common and distinct neural correlates and distinct age-related activation patterns in two subcomponents of inhibitory control.
4.1 Common and distinct neural activation in two subcomponents of inhibitory control
The MKDA results showed that brain regions including the inferior frontal gyrus, insula, middle cingulate cortex, and the superior parietal lobule were activated by both inhibition subcomponents. Therefore, it suggested that the inferior frontal gyrus, insula, middle cingulate cortex, and inferior parietal lobule played core roles in inhibitory control(Choi EY et al., 2012; Yeo BT et al., 2011), which is in line with previous studies(Cieslik EC et al., 2015; Lemire-Rodger et al., 2019; Zhang F & S., 2019). Moreover, Hobeika et al.(Hobeika L et al., 2016) reported activation of domain-oriented regions within the inferior frontal gyrus and conflict-detecting regions within the middle cingulate cortex in both inhibition subcomponents, which can be interpreted as that either cognitive inhibition process or response inhibition process involves the process of spatial orienting and conflict detecting(Hung et al., 2018).
The main clusters of activation between two subcomponents of inhibitory control were observed in (1) the IFG extending to the insula and (2) the middle cingulate cortex (MCC) and the superior parietal lobule. The IFG is known to engage in the process of inhibiting automatic but irrelevant actions while activating task relevant responses at the same time(Sharp et al., 2010; Wang et al., 2019). Moreover, activation of the IFG during detecting changes in the stimulus features is also observed(Dodds CM et al., 2010). The anterior insula has been considered as the center that controls brain activity across different tasks and stimulus modalities and regulates inhibitory control mechanisms(Cai et al., 2019). Previous study from Wager et al(Wager et al., 2005) has reported a positive correlation between neural activity in the anterior insula and task performance in different inhibitory control tasks. One explanation for this positive correlations is that regions including the anterior insula implement a regulating processes that increases with greater input conflict(Miller & Cohen, 2001). Regarding the middle cingulate cortex, studies have revealed that the MCC is the key region for conflict detection in information processing, reallocation of attention resources, and the formation of corresponding actions(Badzakova-Trajkov G et al., 2009). When participants were required to perform a dual task, such as the Stroop task, stronger MCC activation can be observed(Hoffstaedter F et al., 2014; Hoffstaedter F et al., 2013; Palomero-Gallagher N et al., 2019). Based on previous findings and the results on the common regions engaged in different inhibitory control tasks, we propose that the inferior frontal gyrus, the insula, the superior parietal lobule and MCC may comprise the core neural network of the inhibitory control system.
In this meta-analysis, the inhibitory control paradigms classified as cognitive inhibition required conflict resolution and inhibition of response tendencies for successful responding(Nee DE et al., 2007). When performing the cognitive inhibition tasks (i.e. Stroop, Simon, Flanker tasks), participants need to actively reorient attention away from task-relevant stimulus location or feature and then select and initiate an adequate response. Reorienting of attention mainly involved the pre-supplementary motor area and the superior parietal lobule. The superior parietal lobule is showed to play an essential role in facilitating attention re-allocating to characteristics of stimuli and then re-directing attention. Therefore, significantly stronger activation in the left superior parietal lobule observed in cognitive inhibition than response inhibition in current contrast analysis indicated more attentional reallocation load or requirement when performing cognitive inhibition tasks. It thus proved that cognitive inhibition depends largely on inhibition processes of predominant mental set regulated by goal and conflicts(Nee DE et al., 2007).
Whereas for response inhibition, Go/NoGo and stop signals tasks encompass future action selection and inhibition of a predominant response tendency or an ongoing response respectively. As mentioned above, the inferior frontal gyrus plays an inhibitory role in resolving conflicts during response execution and the anterior insula involves in the regulating process of response inhibition. Thus, activated regions in response inhibition were greater than cognitive inhibition primarily located in the inferior frontal gyrus and the anterior insula. The distinctiveness between response and cognitive inhibition, we suggest, may partly due to the difference of cognitive load in these inhibitory control tasks. Participants are required to resolve conflicts and involved more sensory or stimulus-related neural activity in cognitive inhibition tasks, while inhibitory load may further increase in the response inhibition tasks, which require inhibiting a predominant tendency or stopping of already initiated actions. Furthermore, these tasks differ in terms of task-related complexity. Suppressing a response tendency or canceling an ongoing action might increase the inhibitory demand as compared to suppressing interference due to irrelevant information or resolving conflicts, as is the case in cognitive inhibition tasks(Sebastian, Baldermann, et al., 2013). As the engagement of the IFG, MFG and insula plays a core role in the process of inhibitory control, activation in these regions were observed increase with the demands of inhibitory control tasks increase in response inhibition. Overall, these results provide further support for the distinctiveness between response and cognitive inhibition.
4.2 Age-related changes in activation on subcomponents of inhibitory control
In this meta-analysis, we observed neither a completely coherent increase nor a decrease in the inhibition network between two subcomponents. In the cognitive inhibition tasks, activation showed positive association with age in the anterior cingulate cortex, the insula, the superior parietal lobule and the inferior frontal gyrus. These age-related changes fit with the existing literature that prefrontal regions, including the IFG and the MFG became more active with aging(Sebastian, Pohl, et al., 2013). Older adults increasingly recruit additional prefrontal regions to compensate for age-related declining brain structure and function in cognitive inhibition tasks(Sebastian, Baldermann, et al., 2013).Meanwhile, Nielson et al(Kristy A. Nielson et al., 2002) has revealed compensational activation in the left prefrontal cortex during cognitive inhibition. These results may support our assumption that, a simple task in cognitive inhibition required enough functional compensation in prefrontal regions recruited with aging.
A different pattern of functional age-related changes was found in the response inhibition tasks. We found activation of the response inhibition network including the left anterior cingulate cortex, bilateral inferior frontal gyrus and insula, left superior parietal lobule and the right superior frontal gyrus was negatively correlated with age. These seemingly differential results might also be explained by differences in inhibitory load. Based on study from Reuter-Lorenz and Cappell(Reuter-Lorenz & Cappell, 2008), the current findings suggest that the aging brain fails to recruit additional inhibitory regions with inhibition load increasing and a resource ceiling is reached. With task demand increasing, relative hypoactivation is associated with aging in both core and expand inhibition networks, which may further represent a limitation of abilities for flexibly recruiting additional inhibition networks in older adults(Cappell KA et al., 2010; Schneider-Garces NJ et al., 2010). Prakash et al(Prakash RS et al., 2009) has pointed out the flexibility of the cortical regions becomes limited in older adults with number of conflicts increasing. It is important to note that the above-mentioned theories have partly been based on researches about age-related differences in working memory. Turner and Spreng(Turner & Spreng, 2012) reported differential changes in activation patterns for working memory with different cognitive load and inhibition with age. In addition, results from Sebastian(Sebastian, Baldermann, et al., 2013) in contrast to those from Turner and Spreng(Turner & Spreng, 2012) indicated different activation patterns in prefrontal regions during inhibition with medium inhibitory load between low inhibitory load. Our results and these studies indicate that high inhibition task load might result in limited allocation of cognitive resources in older adults, which can be reflected in declined performance associated with lower activation of inhibition networks(Billig AR et al., 2020; Bloemendaal et al., 2018; Pasion R et al., 2019).
Several neuroimaging studies have contributed a lot in enhancing our knowledge of neural correlates of subcomponents of inhibitory control or age-related change in activation in two subcomponents(Hung et al., 2018; Simmonds et al., 2008; Swick et al., 2011; Wright et al., 2014; Zhang et al., 2017). However, these studies are limited for that they focused on a restricted age range(Kristy A. Nielson et al., 2002; K. A. Nielson et al., 2004), included a single subcomponent of inhibitory control(S. Hu et al., 2018; Simmonds et al., 2008; Wright et al., 2014), or used a small sample size(Simmonds et al., 2008; Tsvetanov et al., 2018). For example, Simmonds et al.(Simmonds et al., 2008) included only 11 studies in their meta-analysis. It has been argued that to keep the replicability of a meta-analysis, which should include at least 20 studies(Eickhoff SB et al., 2016), otherwise the conclusions may be questionable. Moreover, Tsvetanov et al.(Tsvetanov et al., 2018) conducted a study on activity and connectivity differences underlying inhibitory control across the adult life span only using response inhibition tasks including Go/NoGo and stop signal tasks. Hung et al.(Hung et al., 2018) reported that unique neural activity was associated with different inhibitory control tasks, but the age-related effects on different types of inhibitory control tasks was unknown. Our meta-analysis addressed these limitations and provided an updated review; thus, our understanding on changes of neural correlates underlying inhibitory control with aging was further advanced.
Through synthesizing data from different subcomponents, we found that brain regions including the inferior frontal gyrus and anterior insula, as well as regions including the middle cingulate cortex and supplementary motor cortex are consistently activated across all inhibition tasks. This finding may suggest that these brain areas are core inhibitory control regions. Meanwhile, different age-related changes in activation between subcomponents of inhibitory control can be observed. Functional reorganization of the aging brain in different inhibitory control tasks showed a complex pattern of increase and decline: the corresponding cognitive inhibition tasks require the older adults to increasingly recruit the core inhibition network and additional inhibitory regions, such as frontal regions and bilateral insula. However, a contrary pattern of an age-related decline in the inhibitory network including prefrontal areas and MCC were showed during the process of response inhibition. Current results suggest that these differences might result from the increasing demands on inhibitory function from cognitive inhibition to response inhibition. Furthermore, age-related increased activation of additional inhibitory networks is limited. When the tasks demand exceeds the older adults’ capacity, the activation in inhibition network decreased evidently. These findings are of significance for the understanding of the neuro-developmental mechanisms of inhibitory control and may provide insights into inhibitory control deficits in clinical settings.
The current study still has some limitations. We note the potential limitation in meta-analysis methods in general is that any meta-analysis method is prone to publication bias, since we only consider results available in the published literature and original studies which report coordinates. Moreover, we cannot control the statistical method used in original articles for thresholding the data. A trend to store unthresholded statistical maps is growing up, which allows to perform image-based meta-analyses in the future studies(Gorgolewski et al., 2015).
Another unavoidable limitation – given that the age computed in meta-regression analysis as a continuous variable was determined by the mean age of each sample in the original articles – was that the mean age was affected by extreme values, which cannot well represent the age distribution of all subjects in each original literature. As mentioned above, to more completely explore the age-related changes in two subcomponents of inhibitory control within individuals of different ages, we performed additional MKDA analyses as validation analyses. To be specific, we divided dataset from all articles included in the current meta-analysis into four age-groups: age ranges between 0 and 18 years of age for underaged, 18–35 years of age for young adults, 35–55 years of age for middle-aged adults and 55–80 years of age for older adults. Then we performed contrast analyses and computed differences among all age-groups in two subcomponents of inhibitory control: underaged vs. young adults, young adults vs. middle-aged adults, middle-aged adults vs. older adults, underaged vs. middle-aged adults, young adults vs. older adults and underaged vs. older adults. The results (see Table S4-S6 and Figure S3-S5of the Supplementary Materials) are basically consistent with the current research, which may confirm the reliability and stability of the current research to a certain extent. However, more research reporting results for narrower age-ranges is still critical for future work.