Adjunctive flupentixol–melitracen for patients labelled as having refractory gastroesophageal reflux disease in China: a systematic review and meta-analysis

Persistent reflux-like symptoms during proton pump inhibitor (PPI) therapy may reflect ongoing reflux, reflux hypersensitivity, functional heartburn, or overlapping disorders of gut–brain interaction. We assessed adjunctive flupentixol–melitracen in Chinese patients labelled as having refractory gastroesophageal reflux disease (rGERD). We reviewed controlled trials explicitly described as randomized. Quasi-randomized and non-randomized allocation were ineligible. Eight databases were searched from inception to 28 July 2026. Post-treatment arm-level data were used consistently for continuous outcomes. Random-effects models, 95% prediction intervals (PIs), exploratory subgroup and sensitivity analyses, small-study-effect tests, and GRADE assessments were performed. Twenty-nine trials (3242 participants), all conducted in China, were retained after source re-audit. Adjunctive therapy was associated with higher total clinical effectiveness (26 trials; RR 1.24, 95% CI 1.20 to 1.29; I²=0%), but this was a non-standard composite assessed largely without blinding. Anxiety (23 trials; SMD − 2.38, 95% CI − 2.93 to − 1.84; I²=97.0%; 95% PI − 5.01 to 0.24), depression (24 trials; SMD − 2.36, 95% CI − 2.86 to − 1.85; I²=96.5%; 95% PI − 4.82 to 0.11), and RDQ scores (8 trials; MD − 4.01, 95% CI − 5.27 to − 2.74; I²=98.6%; 95% PI − 7.67 to − 0.35) favored adjunctive therapy on average, but effect magnitude was highly uncertain. Eight trials sparsely reported adverse events (RR 0.55, 95% CI 0.37 to 0.84); this cannot establish a protective effect or comparative safety. Small-study effects were detected for total effectiveness and the principal psychological outcomes. Certainty was very low. In Chinese trials of patients labelled as having rGERD, adjunctive flupentixol–melitracen was associated with short-term improvements in composite response and symptom scores. Extreme heterogeneity, prediction intervals crossing no effect for the psychological domains, limited objective reflux confirmation, unclear allocation concealment and blinding, small-study effects, and sparse harms reporting preclude confident estimates of benefit or safety and do not support routine use.