Objective <p>Neurosurgical modulation of reward circuitry has been explored for severe, treatment-refractory substance use disorders (SUDs), yet the human literature remains fragmented across targets, modalities, and outcomes, limiting target selection and future trial design. We systematically reviewed human clinical evidence for deep brain stimulation (DBS) and lesion approaches targeting reward network nodes for SUDs, and summarized convergent preclinical findings to identify where the strongest clinical signal exists and to provide a framework for mechanism-informed future trials.</p> Methods <p>We systematically searched PubMed, PubMed Central, Embase, and Scopus (inception-September 9, 2025) for human DBS/ablation studies in primary SUDs targeting addiction-relevant nodes (e.g., nucleus accumbens [NAc], ventral capsule/ventral striatum/anterior limb of internal capsule [ALIC], subthalamic nucleus) and reporting extractable outcomes; in vivo addiction-model animal studies were included for mechanistic context. Findings were synthesized narratively.</p> Results <p>Forty-seven studies met inclusion criteria. The clearest controlled human signal came from bilateral NAc DBS in severe alcohol use disorder, with additional promising but less mature findings in heroin/opioid DBS. In the only randomized delayed-activation trial of bilateral NAc DBS for severe alcohol use disorder (<i>n</i> = 12), abstinent days increased from 27.6% to 56.0% at 6&#xa0;months (<i>p</i> = 0.014) and 74.2% at 18&#xa0;months (<i>p</i> = 0.004), and heavy drinking days decreased from 69.6% to 47.4% (<i>p</i> = 0.020) and 35.5% (<i>p</i> = 0.031). Craving improved (OCDS 21.0 to 10.5 at 6&#xa0;months, <i>p</i> = 0.014; 5.9 at 18&#xa0;months, <i>p</i> = 0.008), with blinded-phase separation favoring earlier stimulation for abstinent/heavy drinking days (<i>p</i> = 0.048). An open-label alcohol DBS pilot reported drinks/day decreasing from 10.4 to 2.7 at 12&#xa0;months (<i>p</i> &lt; 0.05) and OCDS from 28.7 to 8.3 (<i>p</i> &lt; 0.05). In heroin use disorder DBS targeting NAc ± ALIC (<i>n</i> = 8), 5/8 remained abstinent for &gt; 3&#xa0;years; craving VAS decreased from 7.6 to 1.4 at 3&#xa0;months and 0.8 at 24&#xa0;months (<i>p</i> &lt; 0.001), with depressive symptoms improving (HDRS 12.4 to 3.4; <i>p</i> = 0.001). In a polysubstance feasibility study (<i>n</i> = 4), two participants maintained abstinence for &gt; 1150 and &gt; 520&#xa0;days; craving decreased in individuals (opioid craving <i>p</i> = 0.036; benzodiazepine <i>p</i> = 0.001). Biomarker-guided opioid DBS reduced cue-induced craving (6.93 vs 3.09; <i>p</i> &lt; 0.001). Ablation cohorts showed short-term improvements (alcohol NAc ablation: 3/12 relapsed; dependence severity 39.08 to 9.25, <i>p</i> &lt; 0.01), but longer-term relapse and neuropsychological tradeoffs were reported (4-year abstinence 53.8% overall). Preclinical studies highlighted strong target-and state-dependence, including paradoxical relapse augmentation in alcohol-dependent models.</p> Conclusion <p>DBS and lesion approaches should be interpreted as distinct neurosurgical modalities rather than a single therapeutic class. The strongest controlled human signal currently comes from bilateral NAc DBS in severe alcohol use disorder, with additional but lower-certainty DBS evidence in opioid/heroin and polysubstance use disorders. Lesion approaches provide historical and mechanistic support for the relevance of ventral striatal circuitry, but their irreversibility, heterogeneous follow-up, and neuropsychological tradeoffs require a higher evidentiary and ethical threshold. Overall, current data support clinical plausibility and trial-design rationale, but not definitive efficacy across SUD subtypes</p>