A 55-year-old male, who was healthy all along. He presented with symptoms of dizziness, weakness, blurred vision, and shortness of breath about 20 min after ingestion. Upon admission, his Glasgow Coma scale was 13/15, blood pressure 144/96 mmHg, respiratory rate of 30 breaths/min, pulse rate of 100 bpm, and blood oxygen saturation level of 100%. His dyspnea worsened progressively, accompanied by unconsciousness and cyanosis of the lips and fingernails at 20:10. Due to acute respiratory failure, the patient was intubated and transferred to the emergency intensive care unit (EICU). Arterial blood gas analysis revealed pH 7.395, PaO2 182.3 mmHg, PaCO_2:36.5 mmHg, CO2 22.3 mmol/L, BE:-2.5 mmol/L, K⁺3.7 mmol/L, Glu:7.8 mmol/L, SO2 55%, lactate 4.2 mmol/L.

A 49-year-old previous heathy female, was the wife of case 1. She developed symptoms of blurred vision, peripheral numbness with weakness, and nausea by accompanied emetic tendency about 20 min post-ingestion. Her Glasgow Coma scale was 14/15, blood pressure 133/76 mmHg, respiratory rate of 20 breaths/min, pulse rate of 98 bpm, and temperature 36.8°C. Both her pupils were 3 mm in size and were reactive to light. The limb muscle power was 3/5 and deep tendon reflexes were normal. At 20:42, the patient developed somnolence with intermittent limb convulsions accompanied by a decline in oxygen saturation (SpO₂) to 68%. Endotracheal intubation with mechanical ventilation was subsequently initiated, followed by ICU admission for continuous intensive monitoring and therapeutic management.

A 33-year-old woman with history of Sjögren’s syndrome, was the daughter in law of case 1. Her clinical manifestations resembled those observed in Case 2, with concomitant development of generalized erythema. Her Glasgow Coma scale was 13/15, blood pressure 119/86 mmHg, respiratory rate of 24 breaths/min, pulse rate of 115 bpm, and temperature 36.5°C. Both her pupils were 3 mm in size and were reactive to light. At 20:40, the patient developed a comatose state accompanied by intermittent limb convulsions and frothy oral secretions. Endotracheal intubation was emergently performed with subsequent initiation of mechanical ventilation. The patient was transferred to the intensive care unit (ICU) for advanced life support and ongoing critical care management.

Given the short poisoning interval (30 min), the medical team performed emergent gastric lavage (4) and hemoperfusion was subsequently initiated to enhance toxin elimination following admission to the EICU (5). Laboratory findings revealed leukocytosis in all patients (Case 1: 11.15 × 10⁹/L; Case 2: 11.73 × 10⁹/L; Case 3: 13.76 × 10⁹/L), accompanied by elevated serum amyloid A (SAA) levels. There were no notable abnormalities in the patients’ coagulation function, cardiac enzymes, chest X-ray, and brain CT scan. Adjunctive therapies including anti-infection, gastrointestinal protection and correction of electrolyte imbalances. All patients regained spontaneous respiration by hospital day 2 and were discharged after a five-day treatment period without any neurological complications.

On April 7, the Foshan Center for Disease Control and Prevention (CDC) confirmed the cause of poisoning as G. elegans through liquid chromatography analysis of food samples collected on-site (Figure 5).

Literature screening and data analysis were conducted in accordance with PRISMA guidelines (6). The study protocol has been resisted in the PROSPERO database (Registration number: 2025 CRD420251027380). The scope of the review was to answer the following PICO question: “In patients with acute Gelsemium elegans poisoning, does gastric lavage compared with supportive care without gastric lavage reduce mortality and incidence of long-term sequelae?.” The primary research question was formulated according to the PICO framework (P: population, I: intervention, C: comparator, O: outcome; Table 1). A comprehensive literature search, encompassing databases Web of Science, PubMed, China National Knowledge Infrastructure (CNKI), and Wanfang Data has been performed with the following Boolean Indicators: (“Gelsemium” OR “gelsemium poisoning” Table 2).

The inclusion criteria specified clinical trials, case reports or case series on patients with no follow up limitations. The exclusion criteria were limited to animal studies, systematic reviews, short communications and subjects who died prior to hospital admission were excluded from the analysis. All databases were queried for entries indexed from 2000 onward.

Following the search strategy across the designated databases, all retrieved records were exported into NoteExpress software for duplication remove. Two independent reviewers (Y.Z, B.S.) conducted an initial screening based on the titles and abstracts of all unique records. Full texts were obtained for all eligible articles. Excluded publications were categorized, specifying the reasons for exclusion (Figure 6).

The following data were recorded and collected: publication year, analytical methodology, patient cohort size, geographic provenance (province), age and sex distribution, therapeutic interventions administered, time to first medical contact, implicated food types, and outcome.

Bias risk was assessed using the software package Review Manager (RevMan) software, Version 5.4 (The Cochrane Collaboration).and the OHAT guidelines. The final evaluation considered the following parameters: randomization sequence, allocation concealment, blinding of outcome assessment, completeness of procedure description, clarity of inclusion criteria, attrition bias, reporting bias, follow-up duration, and other biases. Bias risk was categorized as low, unclear, or high.

Odds ratios (ORs) with corresponding 95% confidence intervals (CIs) were calculated for all reported outcomes using RevMan 5.4 (The Cochrane Collaboration). The Mantel–Haenszel method was applied under a random-effects model to account for anticipated cross-trial variations in study designs and intervention characteristics. Between-study heterogeneity was quantified using the I² statistic in accordance with Cochrane Handbook guidelines.

Database screening initially identified 963 publications. During the initial screening phase, 178 duplicate records were removed for duplication. Subsequently, 756 articles were excluded based on title and abstract evaluation. 27 articles proceeded to the eligibility assessment stage. Following full-text review, 14 publications were excluded from the final synthesis for the following reasons:5 articles were excluded due to off-topic,5 were review articles,2 involved animal or non-human subjects, One study was excluded because the subjects experienced prehospital mortality.

Risk of bias assessment for all included studies is presented in Figures 7, 8. All publications exhibited elevated bias risk (Figures 7, 8) attributable to extensive study design divergence, constrained observational periods, and abundant non-comparative clinical data.

A summary is depicted in the Table 3. Thirty fatalities occurred among 160 patients, yielding an in-hospital mortality rate of 18.75%. Among the thirteen case reports, six cases occurred in Guangdong, three in Guangxi, and one case each in Guizhou, Zhejiang, and Hong Kong SAR. The age distribution of cases is relatively wide, covering age groups including adolescents, middle-aged people, and the older adult. Due to the limited number of patients with clearly-defined gender data, a comprehensive gender-based analysis was not feasible. In all previous reports, no gender differences among deceased patients have been mentioned. According to summary of food type, the primary cause is accidental ingestion of soups, medicinal liquors containing liquids. Other causes include suicide and homicide (7, 8).

We used Mantel–Haenszel methods and forest plots to test if gastric lavage helps patients. Because age, toxin types, and other factors affect outcomes, we chose a random-effects model. The results surprised us: Lavage cut death rates by 61% versus controls (OR 0.39, 95% CI 0.15–0.99, p = 0.05). Studies agreed closely (I² = 0%, Figure 9). We conducted a subgroup analysis of the clinical characteristics of 27 deceased patients for whom relatively detailed data were available (3, 7, 9–21).