To our knowledge, this is the first meta-analysis that aims to characterize deficits in functional brain connectivity associated with substance use problems. Using 96 studies (5757 subjects) that adopted a seed-based approach, we observed that brain regions involved in the reward system were disconnected with several brain structures crucial to our understanding of substance use problems. For instance, meta-analysis on vmPFC/OFC revealed hyperconnectivity with the ventral striatum and hypoconnectivity with the hippocampus and amygdala. Furthermore, the limbic striatum showed hyperconnectivity with the orbitofrontal cortex and hypoconnectivity with the precuneus/PCC, whereas the executive striatum was hyperconnected to the motor thalamus and BA 6/8 and hypoconnected to the rostral ACC and anterior insula. These results provide substantial insight into our understanding of the neural mechanism underlying substance use problems.

Striatal deficiencies have directed neurobiological research on substance use for the past decades. Here we showed that subjects with substance use problems showed impaired striatal connectivity with limbic and executive subregions. In healthy subjects, the VS (limbic striatum) show strong connectivity with the vmPFC/OFC ^34,35. Indeed, past researches have robustly shown that these regions are involved in reward prediction and processing of the magnitude of received reward ^36,37, defining both regions as crucial nodes of the reward system³⁸. In our study, we observed that SUD samples (compared to users) were more likely to show VS-vmPFC/OFC dysconnectivity. These are consistent with results of task-based fMRI studies on subjects with substance use problems showing reliable activations of brain reward regions in response to drug cues ^9,39,40. Taken together, these results are in line with influential neurobiological models which state that addiction is characterized by an acquired over-valuation of the motivational value of drug stimuli, stemming from the neural sensitization of dopamine-related reward systems^41,42. In addition, we observed that substance use problems were associated with a reduced connectivity between the VS and the Precuneus/PCC, which is a core region of the default-mode network and plays a well-established role in self-referential processes^43,44. Interestingly, the functional connectivity between the VS and the precuneus/PCC has also been found in other forms of addiction such as internet gaming disorder⁴⁵. Such results are consistent with newer neurobiological models emphasizing that importance of impaired self-awareness in human addiction⁴⁶.

In healthy subjects, the dorsal part of the striatum (executive) is mainly connected to the aMCC/pre-SMA, insula, thalamus and dlPFC^34,47-49. In line with these results, we found that the executive striatum was hyperconnected with the motor thalamus and the dlPFC. Moreover, we observed that the executive striatum was hypoconnected with the rACC and anterior insula, which are the two core regions of the salience network⁴⁴. Past systematic reviews and meta-analyses on functional neuroimaging studies reported that SUD subjects showed prominent hypoactivity in these particular regions during cognitive control tasks^12,50,51, but hyperactivity in responses to drug cues^12,50. As theorized by Everitt and Robbins⁵², repeated use of substances may both exaggerate the attention salience to drug cues (e.g., ACC and anterior insula) and alter executive control (e.g., dlPFC, motor thalamus) over such stimuli. Future research may seek to examine the specific role of these dysconnectivities in relation to SUD symptoms.

Meta-analytical evidence of fMRI studies in healthy subjects indicates a crucial role of the vmPFC in subjective valuation and emotional decision-making⁵³, whereas vmPFC lesion studies found increased risk appetite under hot decision conditions ^54-56. In our meta-analysis, we not only observed a hyperconnectivity between the vmPFC/OFC and VS, but also a hypoconnectivity between the former region and the hippocampus/ amygdala in subjects with substance use problems. One possible interpretation is that the vmPFC may compute subjective valuation by weighting the anticipated cost (amygdala) and benefits (VS), in order to adapt subsequent behaviors ⁵⁷. Hence, a reduced vmPFC-Amygdala connectivity in SU subjects would bias the cost–benefit analysis, resulting in a systematic over-estimation of benefits (vmPFC-VS) over the costs (vmPFC-Amygdala). This concurs with the fact that SU subjects compulsively use and seek drugs despite major negative consequences (e.g., neglecting health, work, relationships).

Limitations of the current meta-analysis need to be acknowledged. First, to achieve 80% power to detect effects that are present in approximately 1/3 of the population, a sample size of 17 experiments is required when using cluster-level FWE³¹. Therefore, results from meta-analyses with less than 17 experiments should be interpreted carefully, as it is the case here for the sub-analyses performed on specific substances. Second, regarding the meta-analyses on subtypes of substance, differences in seed selection between substances may have confounded our results. Thus, larger sample sizes and direct comparison between subtypes of substances will be needed to truly examine the specificity of (or lack of) functional dysconnectivities observed between substance types. Third, most studies in the field have selected seeds from the reward system, which has limited our ability to look at other crucial brain regions. Finally, we performed a coordinate-based meta-analysis rather than a meta-analysis of original statistical brain maps which may produce results with reduced accuracy⁵⁸.