Abnormal brain activity in nonsuicidal self-injury: a coordinate-based activation likelihood meta-analysis of functional neuroimaging studies

Abstract

Background

The high prevalence of nonsuicidal self-injury (NSSI) in youths demonstrates a substantial population-level burden on society. NSSI is often associated with emotional and social skill deficits. To date, several studies have aimed to identify the underlying neural mechanism of those deficits in NSSI by using functional magnetic resonance imaging (fMRI). However, their conclusions display poor consistency.

Objective

We aimed to conduct a meta-analysis using activation likelihood estimation (ALE) for fMRI data based on emotional and cognitive tasks to clarify the underlying neural processing deficits of NSSI.

Methods

We searched for MRI studies of NSSI in the PubMed, Cochrane, and Embase databases. We identified significant foci for the included studies and conducted two ALE meta-analyses as follows: (i) activation for the NSSI contrast healthy control group and (ii) deactivation for the NSSI contrast healthy controls. Considering the diverse sex composition of study participants and possible bias from one large sample study, we conducted sensitivity analyses for the meta-analysis.

Results

Nine studies comprising 359 participants were included, and the results demonstrated substantial activation in NSSI patients compared with healthy controls in two clusters located in the right medial frontal gyrus extending to the rostral anterior cingulate and the left inferior frontal gyrus extending to the insula.

Conclusions

The results suggest that individuals with NSSI show brain activity alterations that underpin their core symptoms, including poor emotional regulation and reward processing deficits. Our findings provide new insights into the neural mechanism of NSSI, which may serve as functional biomarkers for developing effective diagnosis and therapeutic interventions for these patients.

Introduction

Nonsuicidal self-injury (NSSI) refers to the behavior of deliberately destroying or changing body tissues without any suicidal intention (Whitlock et al., 2013). There are several forms of NSSI behavior, including scraping, picking, burning, and bruising, of which skin cutting is the most common (Nock, 2009). NSSI behaviors often co-occur and share common vulnerability factors with suicide, but the behavioral features, motivation, and prevalence rates are different (Auerbach et al., 2021). The prevalence of NSSI in nonclinical samples was estimated to be 17.2% among adolescents, 13.4% among young adults (aged 18–24 years), and 5.5% among adults (aged ≥25 years) (Swannell et al., 2014). NSSI is recommended as a separate diagnosis in the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders and is often associated with emotional and social impairment (Turner et al., 2012; Stanford and Jones, 2009). In contrast to body-focused repetitive behaviors, patients with NSSI are usually characterized by stress, anxiety, and depression (Mathew et al., 2020). NSSI is often used as a way to regulate stress, improve interpersonal difficulties, and regulate negative emotions (Zetterqvist et al., 2013).

In addition to emotional and social deficiencies, NSSI patients may also possess addictive features (Blasco-Fontecilla et al., 2016). For example, they usually cannot control NSSI urges and even feel excited while performing or viewing NSSI behavior (Zetterqvist et al., 2013). Previous studies used the Ottawa Self-Injury Inventory, a commonly used NSSI evaluation scale, to measure the functions and addictive features of NSSI patients, and found that more salient addictive features were linked with more severe NSSI behaviors (Guérin-Marion et al., 2018; Nixon et al., 2015). Moreover, evidence exists to show mutual associations between NSSI and other substance or nonsubstance addictions, such as smartphone and food addictions (Mancinelli et al., 2021; Carlson et al., 2018). However, the neuro-mechanism underlying the clinical features of NSSI is still unknown.

Recent neuroimaging studies using functional magnetic resonance imaging (fMRI) with cognitive and emotional tasks provide some clues for brain functional deficiencies in these patients. For example, studies using the Cyberball paradigm reported that NSSI patients showed hyperactivation of the amygdala and nucleus accumbens as well as the medial prefrontal cortex (PFC) and ventrolateral PFC in the social exclusion condition (Olié et al, 2018; Groschwitz et al., 2016), and enhanced activation of the dorsolateral PFC, anterior cingulate cortex (ACC), and insula in the social inclusion condition (Brown et al., 2017). Concerning the emotional deficiencies, NSSI patients showed altered neural activation patterns for emotional pictures and pictures with self-injury references, mainly located in the amygdala, middle orbitofrontal cortex, inferior and middle frontal regions, and ACC (Mayo et al., 2021; Hooley et al., 2020; Plener et al., 2012). Using gambling tasks, increased activation in the orbitofrontal cortex was observed after unexpected reward in NSSI patients (Vega et al., 2018). However, these neuroimaging studies could not come to a unanimous conclusion on the brain deficiencies related to the clinical features of NSSI.

Based on this background, we conducted a coordinate-based meta-analysis of fMRI data to clarify consistent neural dysfunctions in patients with NSSI and to explore the clinical feature-related brain regions of NSSI that can be potential targets for intervention.

Materials and Methods

Literature search

We conducted systematic searches for original research articles in the PubMed, Cochrane, and Embase databases published until July 2021, using a combination of subject words and free words for two keywords “self-injurious behavior” and “functional magnetic resonance imaging”. The subject words in PubMed included “self-injurious behavior” and “magnetic resonance imaging” and the entry terms defined by subject words were retrieved in the scope of title and abstract, respectively. Searches in Cochrane and Embase were conducted in a similar manner (as shown in the Supplemental Methods).

Inclusion and exclusion criteria

The inclusion criteria were as follows: (i) patients who exhibited current or lifetime self-injury behavior diagnosed as NSSI; (ii) fMRI data available in both the NSSI group and control group; (iii) a healthy control (HC) group comprising demographic characteristics consistent with the NSSI group; (iv) reports on neural activation or deactivation coordinates in the Montreal Neurological Institute or Talairach space; (v) reports published in a peer-reviewed journal article; and (vi) whole-brain level analysis to obtain the activation results.

The exclusion criteria were as follows: (i) indistinguishable NSSI behavior and suicidal attempts that were generally summarized as self-injury behaviors; (ii) unevaluated characteristics of self-injury or participants with suicidal attempts who were allocated to the NSSI group; (iii) NSSI or control groups that did not complete the fMRI evaluation; (iv) self-injury behavior explained by other diseases, such as skin picking disorders, autism spectrum disorders, or other physical diseases (such as encephalitis); (v) research design lacking authenticity and incomplete research results; (vi) publication in languages other than English; (vii) full text unable to be accessed; and (viii) studies reporting only regions of interest or seed-based analysis.

Selected studies

A total of 1021 articles were identified from the PubMed, Cochrane, and Embase databases. Following the removal of 45 duplicates, we excluded 867 articles unrelated to fMRI and NSSI and selected 109 articles after screening their titles and abstracts. Nine eligible studies were included in the meta-analysis, and the other 100 studies failed to provide necessary information after full-text review (Supplemental Fig. S1).

Quality assessment

The quality of the included studies was assessed by a 10-point checklist as in the previous studies (Li et al., 2020; Chen et al., 2015; Strakowski et al., 2000). The checklist focused on clinical and demographic characteristics of individual study samples and on the imaging methodology (Supplemental Table S1).

Activation Likelihood Estimation (ALE) Meta-analysis

ALE is a commonly used coordinate-based meta-analysis method to compare activity between processes or populations. ALE analysis models the uncertainty in the localization of activation foci via Gaussian probability density distributions. The voxel-wise union of these distributions generates the ALE value,  an estimate of the likelihood that at least one foci in the dataset was located within a given voxel. Two researchers (Lai and Jiang) extracted the Talairach or Montreal Neurological Institute coordinates (foci) from the included studies and edited the text files, followed by listing the study information, the number of subjects, and foci information (Montreal Neurological Institute coordinates) associated with neural activation. We used BrainMap GingerALE v.3.0.2 software (Turkeltaub et al., 2012; Eickhoff et al., 2009) to conduct the meta-analysis. We assessed the statistical significance and used cluster-level inference for the threshold method to correct for multiple comparisons. We set P < 0.001 as the cluster-forming threshold, P < 0.05 as the cluster-corrected familywise error and N = 5000 as threshold permutations. We used an anatomical image overlay program called Mango (http://ric.uthscsa.edu/mango), to display the results.

Specifically, we conducted two separate ALE meta-analyses as follows: (i) activation for NSSI contrast HC and (ii) deactivation for NSSI contrast HC. Sensitivity analyses were conducted to explore any bias from gender differences or one single study. Sensitivity analysis 1 excluded three studies involving both men and women, whereas sensitivity analysis 2 excluded a single study with a large sample size (Quevedo et al., 2016) (N = 87), thus contributing ~24% of the total sample size in the primary meta-analysis. In addition, we also conducted eight “leave one out” analyses by rerunning the primary analysis and excluding another study to test the stability of each significant cluster. We also conducted two subgroup analyses to explore whether any significant findings varied by age group or history of suicidal attempts.

Results

The meta-analyses included nine eligible studies comprising 359 participants and 53 foci (coordinates). Table 1 illustrates the characteristics of the studies included in this meta-analysis with their quality assessment scores. Five studies included only adults, whereas three studies included only adolescents. The remaining study included both adolescents and adults. Six studies included only females, whereas the remaining three studies principally included females and different proportions of males. The Cyberball task and picture-viewing task were the most frequently used tasks in fMRI. The HC group in all studies did not receive medication. Only one study controlled the influence of medication on the participants, compared with three studies that did not report psychotropic medication information. The remaining five studies reported on different extents of drug use. Eight studies reported on different comorbidity conditions in the NSSI group, of which depression, anxiety, eating disorders, attention deficit hyperactivity disorder, posttraumatic stress disorder, and bipolar disorder were the most common.

Table 1:

Characteristics of the studies included in the meta-analysis.

				Control groups
First author	Task type	Mean age (age range)	NSSI group (size/gender)	Healthy control	Clinical control	Imaging method	Activation type	Number of foci	Quality scores*
Hooley et al.(2020)	Picture-viewing task	21.27 ± 3.67  (range 18–31)	15F	15 HC	-	ROI + WB	Both	19	9
Malejko et al. (2019)	Cyberball task	23.3 ± 4.13 (all adults)	15 F	17 HC	16 NSSI a	WB	Activation	2	9.5
Vega et al. (2018)	Gambling task	29.55 ± 5.85 (range 18–45)	20F	20 HC	20 BPD b	WB	Activation	5	9.5
Dahlgren M.S et al. (2018)	Multi-Source Interference Task	21.27 ± 3.67 (range 18–31)	15F	15 HC	-	ROI + WB	Both	4	9
Brown et al. (2017)	Cyberball task	15.5 ± 2.0 (all adolescents)	10F/3M	32 HC	14 BPD c	WB	Activation	1	9.5
Quevedo et al. (2016)	Interpersonal self-processing task	14.94 ± 1.54 (all adolescents)	32F/18M	37 HC	36 DEP d	WB	Both	8	9.5
Groschwitz et al. (2016)	Cyberball task	15.4 ± 1.9 (all adolescents)	11F/3M	15 HC	14 DEP d	WB	Activation	1	9.5
Plener et al. (2012)	Picture-viewing task	15.2 ± 1.5  (range 14–18)	9F	9 HC	-	WB	Both	12	8.5
Schmahl et al. (2006)	Heat or pain stimulus	28.67 ± 5.88 (all adults)	12F	12 HC	-	WB	Both	1	9.5

F: female; M: male; HC: healthy group; DEP: depression; BPD: borderline personality disorder; WB: whole brain; ROI: regions of interest; Both: both activation and deactivation.

^aNSSI without BPD.

^bBPD without NSSI.

^cNSSI and BPD.

^dDEP without NSSI; * Quality scores out of 10

Meta-analysis 1: activation of “NSSI contrast HC”

Significant ALE clusters

We identified two clusters from 53 foci, 359 participants, and nine separate studies that survived the cluster-level inference correction threshold (Table 2). Cluster one displayed one peak in the right hemisphere and was primarily located in the medial frontal gyrus (MeFG) (77.4% of all studies) and rostral ACC (rACC) (22.6%) (Fig. 1A). Cluster two displayed one peak in the left hemisphere and was primarily located in the inferior frontal gyrus (IFG) (60% of all studies), extending to the extranuclear (24%), insula (14%), and subgyral (2%) regions (Fig. 1B).

Table 2:

Locations of MNI peak coordinates with significant ALE values: meta-analysis and sensitivity analyses.

			Peak voxel coordinates					Contributing experiments
Cluster	Anatomical region	Peak	x	y	z	Cluster size (mm3)	ALE value (×10− 2)	N	%
Meta analysis
1	Right Medial Frontal Gyrus/Anterior Cingulate	Peak 1	12	50	8	928	0.0177	4	44
2	Left Inferior Frontal Gyrus/Insula	Peak 1	−32	20	−12	824	0.0139	3	33
Sensitivity analysis 1
1	Left Inferior Frontal Gyrus/Insula	Peak 1	−32	20	−12	1032	0.0139	3	50
2	Left Anterior Cingulate	Peak 1	−10	36	−4	528	0.0173	2	33
Sensitivity analysis 2
1	Right Medial Frontal Gyrus/Anterior Cingulate	Peak 1	12	50	8	928	0.0177	4	44
2	Left Inferior Frontal Gyrus/Insula	Peak 1	−32	20	−12	880	0.0138	2	40
3	Left Anterior Cingulate	Peak 1	−10	36	−4	528	0.0173	2	40

MNI: Montreal Neurological Institute.

Figure 1:

Activated clusters in the meta-analysis comparing NSSI patients and HCs. MFG.R: right medial frontal gyrus; rACC.R: right rostral anterior cingulate; IFG.L: left inferior frontal gyrus; INS.L: left insula.

Sensitivity analysis

Sensitivity analysis 1 included 43 foci, 184 participants, and six studies. The coordinates of cluster one, identified in the left IFG extending to the insula, were included in the primary analysis. Cluster two was identified with one peak in the left hemisphere and was completely located in the rACC (100% of all studies) (Table 2 and Supplemental Fig. S2). Sensitivity analysis 2 included 45 foci, 272 participants, and eight studies. The coordinates of the two clusters, identified in the right MeFG extending to the rACC and the left IFG extending to the insula, were the same as those in the primary analysis. Cluster three displayed one peak in the left hemisphere and was completely located in the rACC (100% of all studies). The additional eight “leave one out” analyses identified four clusters. The right MeFG extending to the rACC and the left IFG extending to the insula showed great stability in the primary meta-analysis, with each reported in six separate “leave one out” analyses. However, the right parahippocampal gyrus (PHG) and left rACC showed poor stability (Supplemental Table S2). Subgroup analyses in adult patients included only five studies and found that right MeFG extending to the rACC was no longer a significant cluster. Subgroup analyses in patients with a history of suicidal attempts included five studies and identified only right PHG (Supplemental Table S3 and Fig. S2).

Meta-analysis 2: deactivation of “NSSI contrast HC”

No significant clusters were reported by the ALE analysis.

Discussion

Findings from the meta-analyses of nine fMRI studies conparing NSSI patients and HCs emphasized significant clusters of activation during various tasks in the right MeFG extending to the rACC and the left IFG extending to the insula for NSSI patients. The aforementioned regions were associated with emotional and reward processing, which is consistent with previous studies. Conversely, we did not identify significant clusters of deactivation in the analyses. We confirmed the stable significance of these regions through sensitivity and “leave one out” analyses.

The IFG and insula are related to emotional processing and regulation, which are important motivations to initiate the NSSI behavior. The IFG is considered an important brain region for emotional categorization, simulation, regulation, and social interaction. Moreover, the left IFG is reportedly involved in the initial detection of an affective arousal of emotional stimuli by continuously applying theta burst stimulation (Urgesi et al., 2016). It plays an important role in the perception of emotional content (Belyk et al., 2017) and underlies emotional interference resolution (Levens and Phelps, 2010). A neuroimaging study suggests a specific role of the left IFG in the explicit evaluation of emotional prosody, which is essential for social interaction (Bach et al., 2008). On the other hand, the activity of the insula covaries with subjective feelings, inclining us to approach or avoid the stimulus and representing emotional experiences, such as empathy (Suzuki, 2012). Researchers have reported the interaction of positive emotion and pain stimuli that leads to the activation of the left insula (Orenius et al., 2017). An electrophysiological study suggests that the anterior and posterior insula play different roles and transform from sensory to affective representations along the posterior-to-anterior axis (Zhang et al., 2019).

In addition, the MeFG and rACC are involved in the reward processing, which may be associated with the repetition of NSSI behavior. A previous fMRI study suggested that MeFG could bias subsequent risky decision-making (Huang et al., 2014). Another fMRI study separates decision uncertainty into behavioral and reward risks and observes that behavioral risk trials evoke greater activation than reward risk and no risk conditions in the MeFG, ACC, IFG, and insula (Yaxley et al., 2011). The rACC is considered the key structure of the reward system, receives information from the orbitofrontal cortex about reward and nonreward outcomes, and connects rewards to actions. Furthermore, the rACC is related to reward prediction error and modulated by uncertainty (Wang et al., 2017). Neurons in the rACC encode differences between the expected reward and the actual outcome and encode information about reward proximity and amount (Hill et al., 2016; Toda et al., 2012). An event-related brain potential study during gambling tasks also highlights the importance of the rACC in learning the reward values of task contexts to guide action selection (Umemoto et al., 2017).

Previous studies have identified similar regions activated in substance or nonsubstance addictions, which may implicate potential addictive features in NSSI patients. A meta-analysis demonstrated that patients with internet gaming disorder showed significant activations in the MeFG and ACC compared with HCs (Meng et al., 2015). Contrasting choices of larger delayed reward versus smaller but immediately available monetary reward revealed MeFG and ACC activation in pathological gamblers (Miedl et al., 2015). Compared with the control group, cocaine users demonstrate significantly more activity during exclusion versus inclusion in the right MeFG (Hanlon et al., 2019). Alcohol dependence is also associated with increased activation in the ACC and insula (Maurage et al., 2012). At the transmitter level, endogenous opioids, considered addictive substances, are suggested to be involved in the pathogenesis of NSSI. Compared with the control group, adolescents with NSSI display lower beta-endorphin levels in the plasma and cerebrospinal fluid, with increased pain thresholds and decreased pain intensity (Van et al., 2021; Stanley et al., 2010). In a national survey of Chinese adolescents, the nonmedical use of opioids and sedatives was positively associated with NSSI (Xie et al., 2021). However, two crucial characteristics of addiction, tolerance and withdrawal, have not yet been widely studied in the NSSI behavior, and the mechanism underlying the association between the improvement of NSSI behavior and antidepressant treatment is not clear enough. Nevertheless, our study provides new insight into the potential association between NSSI and addiction from a neuroimaging perspective, but more clinical and neuroimaging studies from different aspects are needed to further explore the patients’ characteristics.

Limitation

The meta-analyses had several limitations. First, considering the small sample size of included studies, any definitive consensus on neural activation differences in emotional regulation and reward processing between NSSI patients and HCs is premature. Second, most studies included only females, considering the low prevalence rates of NSSI in males and the inadequate sample size to evaluate sex effects. Thus, the sample representation was insufficient, and the applicability of the conclusions was limited. We found that sex may influence the activated brain regions in NSSI patients via sensitivity analyses, but further studies are needed considering the small sample of included studies. Studies with larger sample sizes and greater quality are warranted to explore the neural mechanism of NSSI behavior for populations with different characteristics in the future. Third, the included studies displayed poor consistency in the inclusion criteria for the NSSI group, particularly NSSI behavior frequency owing to controversial international definitions. The evaluation of self-injury behavior accompanied by suicidal thoughts and attempts was insufficient. Some studies have reported activation changes in the insula and MeFG in patients with suicidal ideation or attempts (Dir et al., 2020; Courtet and Olié, 2019; Olié et al., 2017). Fourth, despite the included studies comprising control groups, several characteristics between the NSSI group and control group were heterogeneous, particularly psychiatric comorbidity and psychotropic-using conditions, which may influence activation changes in the brain. Further studies may disentangle this aspect particularly comparing the NSSI behavioral and neuroimaging characteristics among pure NSSI and NSSI comorbid with depression, anxiety, or borderline personality disorders. The last one is the publication bias. Coordinate-based meta-analyses must rely on published findings, thus publication biases are entailed by favoring positive findings and regions hypothesized in previous studies. We have sought to minimize these biases in the same way as most ALE studies by only including studies with whole-brain analysis.

Conclusion

In this study, we conducted a coordinate-based meta-analysis to examine differential neural processes associated with the clinical features of NSSI. Compared with HCs, we observed substantial activation in the right MeFG extending to the rACC and the left IFG extending to the insula in patients with NSSI, which is related to emotion and reward processing; however, there were no deactivation clusters. The results indicate that the aforementioned regions may serve as functional biomarkers for developing effective diagnoses and future directions for clinical interventions on NSSI behavior.

Author contributions

M.F.L., P.J. and J.J.X. conceptualized and designed the study. M.F.L., P.J. and D.L. performed the literature search and screening. M.F.L. conducted the data extraction and coding, calculated the effect sizes, and performed the statistical analyses. M.F.L. and P.J. drafted the initial version of the manuscript. P.J. contributed to the interpretation of the data and made critical revisions to the paper. All authors contributed to revising the manuscript and gave final approval of the version to be published.

Conflict of interest statement

None.