Background <p>Ultrasound (US)-guided cannulation has been proposed to improve success rates and safety compared with conventional palpation-guided techniques in haemodialysis patients. This systematic review with quantitative synthesis evaluated the efficacy of US-guided versus palpation-guided cannulation for arteriovenous access.</p> Methods <p>Randomized controlled trials (RCTs) comparing US-guided with conventional cannulation were included. Searches were conducted across five electronic databases from inception to May 2025. Quantitative syntheses were performed where appropriate using Review Manager 5.4, calculating pooled risk ratios (RRs) and mean differences (MDs) with 95% confidence intervals (CIs). To address potential unit-of-analysis errors, data were stratified by reporting unit, namely patient-level versus cannulation-event-level reporting. A random-effects model was employed to provide conservative estimates given the high heterogeneity.</p> Results <p>Five RCTs involving 142 unique patients, totaling 2,894 cannulation events, met the inclusion criteria. US-guided cannulation showed a marginal improvement in the first-attempt success rate (RR = 1.08; 95% CI, 1.00–1.16; <i>p</i> = 0.04). However, after excluding the largest event-based reporting study, Wu et al. [24], the leave-one-out sensitivity analysis showed that the effect was no longer statistically significant (RR = 1.07; 95% CI, 0.98–1.16; <i>p</i> = 0.14). The observed overall effect was predominantly driven by trials using cannulation events as the unit of analysis, whereas studies reporting per-patient outcomes showed no significant differences. No definitive clinical or geographical conclusions could be drawn, as apparent variations across trials were identified as statistical artifacts resulting from these unit-of-analysis errors. No statistically significant differences were found in complication rates. Due to substantial methodological heterogeneity in procedure duration definitions, these data were synthesized qualitatively and showed no consistent benefit. High heterogeneity was observed for first-attempt success (<i>I</i>² = 91%) and complications (<i>I</i>² = 89%).</p> Conclusions <p>Although US-guided cannulation shows potential for increasing first-attempt success, the current evidence is limited by significant unit-of-analysis errors and clinical heterogeneity related to the operator learning curve. When accounting for these statistical artifacts, no definitive evidence supports superior efficacy in specific clinical or geographical settings. Future large-scale RCTs should adopt standardized, per-patient reporting to confirm these findings.</p> Protocol registration <p>PROSPERO Registration Number: CRD42023424434.</p>