ObjectiveThe essence of syndrome manifestation recognition in traditional Chinese medicine (TCM) is to infer the body’s latent pathogenesis state from clinical observational information, rather than to perform simple label matching. However, previous studies have largely modeled this task as syndrome pattern classification within a fixed label space, which does not adequately reflect the cognition process of TCM syndrome differentiation centered on pathogenesis reasoning, and is also insufficient to capture the openness, semantic variability, and cross-disease reusability of syndrome manifestation expression. This study aimed to investigate whether introducing pathogenesis reasoning chain-of-thought (PR-CoT) supervision into large language models (LLMs) could improve the quality and cognitive consistency of syndrome manifestation recognition and support cross-disease transfer. MethodsSyndrome manifestation recognition was formulated as a conditional generation task under the framework of clinical observational information (X)→pathogenesis structure (Z)→syndrome pattern output (Y), where Z serves as an explicit intermediate structural variable linking the clinical evidence and syndrome judgment. Within this framework, a PR-CoT-supervised dataset for syndrome manifestation recognition was constructed based on medical case records of spleen-stomach disorders. After preprocessing, information extraction, manual proofreading, and data cleaning, the dataset comprised 4 800 training cases, 400 development cases, and 400 test cases. Each sample was annotated with a structured PR-CoT consisting of three progressive levels: clinical information summarization, comprehensive pathogenesis analysis, and syndrome pattern output. Supervised fine-tuning was conducted on open-source LLMs, with an end-to-end model serving as the baseline. Qwen3-32B was used as the primary experimental model, and Qwen3-14B as the scale comparison model. A progressive multidimensional evaluation framework was further established, comprising a structural parsing level, a semantic similarity level, and an expert blind review level. At the structural parsing level, syndrome pattern expressions were decomposed into structural elements and evaluated using Precision, Recall, F1 score, and Jaccard similarity. At the semantic similarity level, independent LLMs scored the theoretical proximity between predicted and reference syndrome patterns. At the expert blind review level, three TCM experts independently evaluated model outputs on two dimensions: syndrome differentiation consistency and terminology standardization of syndrome patterns. In addition, zero-shot cross-disease transfer evaluation was conducted on gynecological and heart-system disorder test sets. ResultsAt the structural parsing level, PR-CoT supervision did not lead to a stable improvement in the element-wise overlap of syndrome pattern structural components. Compared with the corresponding baselines, neither Qwen3-32B nor Qwen3-14B showed consistent advantages in structural matching metrics after the introduction of PR-CoT supervision. In contrast, at the semantic similarity level, PR-CoT supervision produced stable positive gains across different model scales and evaluation systems. The average semantic score of Qwen3-32B increased from 6.425 8 in the baseline model to 6.585 0 after PR-CoT supervision, and that of Qwen3-14B increased from 5.870 0 to 5.964 2. At the expert blind review level, the overall score of Qwen3-32B (PR-CoT) was 7.026 0±0.107 7, higher than 6.416 3±0.288 9 for its baseline. In zero-shot cross-disease testing, the PR-CoT model still showed advantages in semantic evaluation and expert evaluation on both gynecological and heart-system disorder test sets, indicating a certain degree of transferability. ConclusionThe benefits of PR-CoT supervision are mainly reflected in TCM semantic consistency and clinical plausibility, rather than in improved hard matching of structural elements. These findings support understanding syndrome manifestation recognition as a process of generating and expressing latent pathogenesis structures, rather than as a classification task within a traditional fixed label space. By introducing pathogenesis reasoning as an explicit intermediate structure into the modeling process and combining it with a progressive multidimensional evaluation framework, this study provides a methodological pathway for intelligent TCM syndrome differentiation that integrates theoretical alignment, interpretability, and multi-level evaluation.

ObjectiveThe essence of syndrome manifestation recognition in traditional Chinese medicine (TCM) is to infer the body’s latent pathogenesis state from clinical observational information, rather than to perform simple label matching. However, previous studies have largely modeled this task as syndrome pattern classification within a fixed label space, which does not adequately reflect the cognition process of TCM syndrome differentiation centered on pathogenesis reasoning, and is also insufficient to capture the openness, semantic variability, and cross-disease reusability of syndrome manifestation expression. This study aimed to investigate whether introducing pathogenesis reasoning chain-of-thought (PR-CoT) supervision into large language models (LLMs) could improve the quality and cognitive consistency of syndrome manifestation recognition and support cross-disease transfer. MethodsSyndrome manifestation recognition was formulated as a conditional generation task under the framework of clinical observational information (X)→pathogenesis structure (Z)→syndrome pattern output (Y), where Z serves as an explicit intermediate structural variable linking the clinical evidence and syndrome judgment. Within this framework, a PR-CoT-supervised dataset for syndrome manifestation recognition was constructed based on medical case records of spleen-stomach disorders. After preprocessing, information extraction, manual proofreading, and data cleaning, the dataset comprised 4 800 training cases, 400 development cases, and 400 test cases. Each sample was annotated with a structured PR-CoT consisting of three progressive levels: clinical information summarization, comprehensive pathogenesis analysis, and syndrome pattern output. Supervised fine-tuning was conducted on open-source LLMs, with an end-to-end model serving as the baseline. Qwen3-32B was used as the primary experimental model, and Qwen3-14B as the scale comparison model. A progressive multidimensional evaluation framework was further established, comprising a structural parsing level, a semantic similarity level, and an expert blind review level. At the structural parsing level, syndrome pattern expressions were decomposed into structural elements and evaluated using Precision, Recall, F1 score, and Jaccard similarity. At the semantic similarity level, independent LLMs scored the theoretical proximity between predicted and reference syndrome patterns. At the expert blind review level, three TCM experts independently evaluated model outputs on two dimensions: syndrome differentiation consistency and terminology standardization of syndrome patterns. In addition, zero-shot cross-disease transfer evaluation was conducted on gynecological and heart-system disorder test sets. ResultsAt the structural parsing level, PR-CoT supervision did not lead to a stable improvement in the element-wise overlap of syndrome pattern structural components. Compared with the corresponding baselines, neither Qwen3-32B nor Qwen3-14B showed consistent advantages in structural matching metrics after the introduction of PR-CoT supervision. In contrast, at the semantic similarity level, PR-CoT supervision produced stable positive gains across different model scales and evaluation systems. The average semantic score of Qwen3-32B increased from 6.425 8 in the baseline model to 6.585 0 after PR-CoT supervision, and that of Qwen3-14B increased from 5.870 0 to 5.964 2. At the expert blind review level, the overall score of Qwen3-32B (PR-CoT) was 7.026 0±0.107 7, higher than 6.416 3±0.288 9 for its baseline. In zero-shot cross-disease testing, the PR-CoT model still showed advantages in semantic evaluation and expert evaluation on both gynecological and heart-system disorder test sets, indicating a certain degree of transferability. ConclusionThe benefits of PR-CoT supervision are mainly reflected in TCM semantic consistency and clinical plausibility, rather than in improved hard matching of structural elements. These findings support understanding syndrome manifestation recognition as a process of generating and expressing latent pathogenesis structures, rather than as a classification task within a traditional fixed label space. By introducing pathogenesis reasoning as an explicit intermediate structure into the modeling process and combining it with a progressive multidimensional evaluation framework, this study provides a methodological pathway for intelligent TCM syndrome differentiation that integrates theoretical alignment, interpretability, and multi-level evaluation.

ObjectiveTo investigate the clinical effects and safety of Shuhe Granules （舒和颗粒） in patients with rheumatoid arthritis （RA） comorbid with insomnia. MethodsA single-arm prospective exploratory design was adopted. Thirty RA patients comorbid with insomnia and diagnosed with the traditional Chinese medicine （TCM） syndrome of kidney-heart deficiency and qi-blood disharmony were enrolled. On the basis of maintaining their original RA treatment regimens， all patients received oral Shuhe Granules， one dose per day， for 12 consecutive weeks. The outcomes were assessed before treatment and at weeks 4， 8， and 12 after treatment. Sleep and related psychosomatic function were evaluated using the insomnia severity index （ISI）， Pittsburgh sleep quality index （PSQI）， fatigue severity scale （FSS）， patient health questionnaire-9 （PHQ-9）， generalized anxiety disorder-7 （GAD-7）， and Montreal cognitive assessment （MoCA） scores. RA-related assessments included the 28-joint disease activity score based on C-reactive protein （DAS28-CRP）， pain visual analog scale （VAS）， tender joint count （TJC28）， swollen joint count （SJC28）， morning stiffness duration， and health assessment questionnaire （HAQ） score. Laboratory parameters included erythrocyte sedimentation rate （ESR）， serum C-reactive protein （CRP）， anti-cyclic citrullinated peptide antibody （anti-CCP）， and rheumatoid factor （RF） levels. A literature-based historical control group was established through literature retrieval and meta-analysis. The ISI， PSQI， DAS28-CRP， and pain VAS scores were compared between the treatment group in the present study and the historical control group. Safety assessments were also performed. ResultsCompared with baseline before treatment， ISI and PSQI scores decreased at weeks 4， 8， and 12， while MoCA scores increased at weeks 4， 8， and 12. FSS scores decreased at weeks 4 and 12. At weeks 8 and 12， GAD-7 scores， DAS28-CRP scores， duration of morning stiffness， TJC28， and SJC28 were reduced. At week 12， PHQ-9， pain VAS， and HAQ scores significantly decreased （P＜0.05）. FSS and DAS28-CRP scores at week 12 were lower than those at week 4， and CRP levels at week 8 were lower than those at week 4 （P＜0.05）. Compared with the historical control group， the treatment group showed superior improvement in ISI and PSQI scores （P＜0.001）； however， no statistically significant differences were observed between the two groups in DAS28-CRP or pain VAS scores （P＞0.05）. No serious adverse events occurred during the study， and the 4 reported mild adverse events were all considered unrelated to the study medication. ConclusionShuhe Granules significantly improve sleep quality， fatigue， emotional status， cognitive function， RA-related symptoms， and joint function in patients with RA comorbid insomnia， with good safety and tolerability.

Objective: To examine the association of physical activity with cognitive function and mental health among primary school students based on latent subgroup identification and dose response modeling, and to clarify cognitive-psychological profiles and the protective role of physical exercise, so as to provide evidence based support for physical exercise promoting mental health among primary school students. Methods: A study was conducted among 546 primary school students aged 8-12 years old from a primary school in Shanghai from March to June 2025. Physical fitness indicators, including vital capacity, 50 m sprint, sit and reach, and 1 min rope skipping were measured. Physical exercise was assessed by using the Physical Activity Rating Scale-3 (PARS-3), and mental health was assessed by using the Revised Child Anxiety and Depression Scale, 47-item version (RCADS-47). Cognitive function was assessed by using Stroop, N-back, and task-switching tasks. Latent profile analysis (LPA) was performed for cognitive-psychological health of primary school students based on seven cognitive indicators and six psychological indicators. Multivariable Logistic regression, trend test, and restricted cubic spline (RCS) analysis were used to examine associations between physical exercise and psychological highrisk status. Results: Five cognitive psychological health subgroups were identified:high symptom and impaired cognition group ( n =40, 7.3%), low cognition and low symptom group ( n =108, 19.8% ), and high cognition and high symptom group ( n =110, 20.2%), cognitive advantage and psychologically healthy group ( n = 272, 49.8%), and extremely high symptom and cognitively vulnerable group ( n =16, 2.9%). Among them, the high symptom and impaired-cognition group and extremely high symptom and cognitively vulnerable group were classified as psychologically high risk types. Significant differences were observed across subgroups in psychological symptoms, cognitive function, some physical fitness indicators and physical exercise level, except for body weight, body mass index, and sit and reach ( F =2.64-297.53, all P < 0.05). The results of the multivariate Logistic regression analysis indicated that high level sports, including high total volume of physical exercise, single session duration >30 min, high intensity, and frequency ≥3 times/week, was associated with lower psychological high risk status ( OR =0.20, 0.36, 0.33, 0.40, all P <0.05). Trend analysis showed that total volume of physical exercise and single session exercise duration were negatively associated with psychological high risk status among primary school students ( OR =0.98, 0.79, both P trend <0.05). RCS analysis showed no statistically significant overall or nonlinear association between total volume of physical exercise and psychological high risk status ( χ 2=2.92, 0.01, both P >0.05). Conclusions Cognitive-psychological health among primary school students is heterogeneous. Physical exercise is associated with psychological health improvement and may show a threshold effect, with total volume of physical exercise and single session exercise duration being the key dimensions.

Depression is a prevalent mental and emotional disor-der that often results in significant emotional disturbances,cog-nitive dysfunction,and memory impairments.It is characterized by a high incidence rate,a substantial disability burden,and limited therapeutic efficacy.Currently,the long-term use of medications for the treatment of depression can result in a range of adverse reactions,highlighting the urgent need to explore no-vel approaches that can effectively alleviate depressive symptoms while minimizing side effects.Curcumin,a natural polyphenolic compound derived from the rhizome of turmeric,demonstrates considerable potential in the prevention and treatment of depres-sion,owing to its diverse array of biological activities.In recent years,numerous studies have investigated the use of curcumin for the treatment of depression.This article aims to provide a comprehensive review of the mechanisms of action underlying curcumin's efficacy in treating depression.Specifically,it focu-ses on its ability to improve neurotransmitter imbalances,restore neural plasticity,alleviate neural damage,mitigate dysfunction of the hypothalamic-pituitary-adrenal(HPA)axis,regulate in-flammatory factors and neuroinflammatory signaling pathways,and inhibit oxidative stress.This review is intended to offer in-sights and methodological references for basic research on curcu-min,as well as for the development of novel therapeutic agents for the treatment of depression.

Objective To investigate the differences in BMD,T-scores,lumbar HU values,and modified VBQ scores in patients with thoracolumbar fractures,analyze their correlations,and evaluate their predictive efficacy for thoracolumbar fractures.Methods Patients with thoracolumbar fragility fractures and hospitalized in the Department of Orthopedics at the Second Affiliated Hospital of Kunming Medical University from January 2022 to March 2025 were retrospectively enrolled.Among them,132 eligible patients were divided into the fracture group(n=63)and the non-fracture group(n=69).BMD and T-scores of L1～L4 vertebrae were measured via DEXA,lumbar HU values were obtained from CT scans,and modified VBQ scores were calculated using lumbar MRI.Intergroup comparisons were performed using independent samples t-tests.Pearson's correlation analysis was used to assess the relationships among BMD,T-scores,HU values,and modified VBQ scores.Receiver operating characteristic(ROC)curve analysis was conducted to evaluate their predictive performance for thoracolumbar fractures.Results The fracture group exhibited significantly lower mean BMD,T-scores,and HU values in L1～L4 compared to the non-fracture group,while modified VBQ scores were significantly higher(P<0.001).Correlation analysis revealed a significant positive correlation between BMD,T-scores,and HU values in L1～L4,whereas all three parameters showed a significant negative correlation with modified VBQ scores(P<0.001).The AUC values for predicting thoracolumbar fractures were 0.826(T-score),0.836(BMD),0.759(HU value),and 0.875(modified VBQ score),with optimal cutoff thresholds of-1.65(T-score),0.836 g/cm2(BMD),68.4(HU value),and 3.01(modified VBQ score),respectively.Conclusion BMD,T-scores,HU values,and modified VBQ scores in L1～L4 vertebrae are significantly correlated and can serve as the predictive indicators for thoracolumbar fractures.Among them,the modified VBQ score demonstrates the best predictive performance,making it a valuable auxiliary tool for assessing the vertebral bone quality.

OBJECTIVE: To assess the value of whole exome sequencing (WES) for the diagnosis of early-onset genetic diseases among infants aged 0 to 6 month in Ningbo region. METHODS: 268 infants presented at the Women and Children's Hospital Affiliated to Ningbo University from January 2022 to June 2024 undergoing WES-based genetic testing were enrolled. Peripheral blood samples were collected from the infants and their parents and subjected to WES. Pathogenic variants were identified by clinical manifestations. This study has been approved by the Medical Ethics Committee of the Hospital (Ethics No. EC2023-017). RESULTS: Among the 268 infants, 124 (46.3%) had phenotype-explaining genetic variants. For 42 family-based WES tests, 20 (47.62%) were abnormal, whilst in 226 single-person WES tests, 104 (46.02%) had abnormalities, with 76 (33.63%) verified by parental testing. In 96 fully family-verified cases, 31 were de novo, 40 were parent-inherited, 25 were single-parent-inherited. These included 35 inborn metabolic errors, 28 rare syndromes, 9 neurodevelopmental disorders, 4 musculoskeletal diseases, 5 congenital deafness, 2 mitochondrial diseases, 4 endocrine diseases, and 9 others. Among these, there were 7 pathogenic copy number variations (all deletions), 3 chromosomal abnormalities, and 85 single-nucleotide variations. One case of Beckwith-Wiedemann syndrome was detected by methylation MLPA. Among the single-nucleotide variants, 114 pathogenic/likely pathogenic variants were identified in 61 genes, with common ones including missense variants (64.04%), frameshifting variants (20.18%) and splicing variants (4.39%). CONCLUSION WES can offer effective diagnosis for hereditary diseases with specific/non-specific manifestations. For early-age infants, higher detection rates may be attained for inborn metabolic errors, rare syndromes, neurodevelopmental disorders, congenital deafness, and musculoskeletal diseases. Compared with single-person WES, family-based WES can attain a higher diagnostic efficiency.
Humans ; Exome Sequencing/methods* ; Infant ; Female ; Male ; Infant, Newborn ; Genetic Diseases, Inborn/diagnosis* ; Genetic Testing/methods*

Objective To investigate the clinical and imaging features of alveolar soft part sarcoma(ASPS).Methods A retro-spective analysis was conducted on the clinical and imaging data of 41 patients with pathologically confirmed ASPS.Results The primary clinical manifestation of 41 patients with ASPS was progressive tumour enlargement.The tumors were located in the but-tocks(14 cases),thighs(11 cases),lower legs(5 cases),forearms(3 cases),upper arms(2 cases),psoas major muscles(2 cases),and one case each in the paracervical region,shoulder,axilla,and plantar regions.Pulmonary metastasis were observed in 30 patients,and 7 patients had metastases to other sites,including 2 cases of intracranial metastasis,1 case of lymph node metastasis,and 4 cases of bone and soft tissue metastasis.Five patients experienced postoperative local recurrence or metastasis.MR was performed in 34 patients,and CT was performed in 32 patients.The tumors were oval or irregular in shape.On CT,the tumors appeared as soft tissue density shadow,with calcification in 8 cases and 68％(17/25)showed marked enhancement with peripheral vascular plexus.On MR,the tumors exhibited isointense T1 and long T2 signals,67％(23/34)showed flow-void vessels within the tumor and 94％(32/34)showed tortuous vascular plexus at the tumor margins.Conclusion ASPS predominantly occurs in the muscles of buttocks and thigh and is prone to lung metastasis.Enhanced CT and MR can reveal vessels within and at the margins of the tumor,which is diagnosti-cally significant for ASPS.

Objective To establish a stable liquid chromatography-tandem mass spectrometry(LC-MS/MS)method for detecting gamma-aminobutyric acid(GABA)in plasma,and to evaluate the value of GABA detection in the diagnosis of sleep disorders.Methods GABA was detected using a UPLC Xevo TQs system.The method was pre-validated and its performance was verified to establish a reference range for healthy individuals.The difference in plasma GABA levels between apparently healthy individuals and patients with sleep disorders was compared.Results We employed deuterated compounds as isotopic internal standards and utilized an Amide chromatographic column for separation.The mobile phase was 0.050%formic acid in water and 90%acetonitrile in water containing 0.175%formic acid and 5 mmol/L ammonium acetate with gradient elution in the column temperature of 35℃.The linear range for the detection of GABA by LC-MS/MS was 0.05-10.00 μmol/L,with a lower limit of quantification of 0.02 μmol/L,the inter-day CV<3.00%and intra assay CV<4.00%,respectively,and the recovery rate was 101.06%-109.02%.The reference ranges for plasma GABA were established by analyzing 300 healthy controls stratified by age:18-34 years(0.08-0.15 μmol/L),35-49 years(0.10-0.20 μmol/L),and≥50 years(0.12-0.23 μmol/L).Then plasma GABA was used as a biomarker for auxiliary diagnosis of sleep disorders in analyzing 221 patients and 300 healthy controls,which revealed that AUC values were 0.510(P=0.850),0.686(P=0.002),and 0.890(P<0.001)in the groups of 18-34 years,35-49 years,and≥50 years,respectively,with optimal cut-off values of 0.09,0.10 and 0.11 μmol/L.Conclusion A reliable LC-MS/MS method for detecting GABA has been established,which can detect plasma GABA levels sensitively and accurately and can be used in assisting the clinical diagnosis of sleep disorders.

Objective:To investigate the effects of metformin on esophageal squamous cell carcinoma cell growth, migration and angiogenesis by regulating the expression of ALKBH3.Methods:Human esophageal cancer TE-1 cells were treated with different concentrations (0, 0.5, 1.0, 2.0, 4.0, 8.0 mmol/L) of metformin, and they were divided into a blank control group, low- (0.5 mmol/L), medium- (1.0 mmol/L), and high- (2.0 mmol/L) concentration metformin groups, a metformin (2.0 mmol/L) +pcDNA-NC group, and a metformin (2.0 mmol/L) +pcDNA-ALKBH3 group. The cell viability was determined by the CCK-8 method. The cell proliferation ability was detected by the clone formation assay. The cell migration and invasion abilities were examined by the Transwell assay. The cell apoptosis was detected by flow cytometry. The tube formation ability of cells was detected by the angiogenesis assay. A xenograft tumor model was constructed using 4- to 6-week-old male BALB/c thymus-less nude mice, which were divided into a model control group, a metformin group, a metformin+pcDNA-NC group, and a metformin+pcDNA-ALKBH3 group using a random number table method, and with six in each group. And the volume and weight of the tumor were measured. The protein expression levels of apoptosis-related proteins Bcl-2, Bax, ALKBH3 and vascular endothelial growth factor A (VEGF-A) were detected by Western blotting. The expression of CD31 protein was detected by immunohistochemistry.Results:After treating TE-1 cells with 0, 0.5, 1.0, 2.0, 4.0, and 8.0 mmol/L metformin for 48 hours, the cell viability was (100.00±0.00) %, (90.31±5.23) %, (81.25±8.65) %, (63.52±6.80) %, (54.64±5.35) %, and (31.48±4.21) %, respectively, with a statistically significant difference ( F=98.11, P<0.001). There were statistically significant differences in cell viability between 0.5, 1.0, 2.0, 4.0, 8.0 mmol/L and 0 mmol/L (all P<0.05). The IC 50 of metformin for TE-1 cells was 4.46 mmol/L. The numbers of colony formations of TE-1 cells in the blank control group, low-, medium-, and high-concentration metformin groups, metformin+pcDNA-NC group, and metformin+pcDNA-ALKBH3 group were 153.15±13.55, 134.80±11.62, 116.24±10.43, 93.17±8.85, 89.39±8.46, 110.26±7.21, respectively, with a statistically significant difference ( F=34.28, P<0.001); the numbers of colony formations of TE-1 cells in the metformin groups at different concentrations decreased significantly with the increase in metformin concentration (both P<0.05); compared with the metformin+pcDNA-NC group, the number of colony formations of cells in the metformin+pcDNA-ALKBH3 group increased ( P<0.05). The numbers of migration of TE-1 cells of 6 groups were 152.13±13.40, 133.85±10.72, 115.28±8.64, 91.16±7.89, 85.39±7.23, 116.85±8.36, the numbers of invasion were 135.22±10.77, 112.07±9.53, 86.30±7.45, 69.53±6.74, 65.81±5.65, 79.80±6.32, respectively, with statistically significant differences ( F=41.35, P<0.001; F=69.06, P<0.001); the numbers of migrated and invaded cells in the metformin groups at different concentrations decreased significantly with the increase in metformin concentration (all P<0.05); compared with the metformin+pcDNA-NC group, the numbers of migrated and invaded cells in the metformin+pcDNA-ALKBH3 group increased significantly (both P<0.05). The apoptosis rates of TE-1 cells in 6 groups were (3.22±1.13) %, (13.82±1.90) %, (22.67±2.53) %, (29.18±3.24) %, (26.84±2.75) %, and (16.36±1.63) %, respectively, with a statistically significant difference ( F=103.66, P<0.001); the apoptosis rates of cells in the metformin groups at different concentrations gradually increased with the increase in metformin concentration (both P<0.05); compared with the metformin+pcDNA-NC group, the apoptosis rate of cells in the metformin+pcDNA-ALKBH3 group was relatively lower ( P<0.05). The tubular structure of cells in blank control group was intact, and there were different degrees of damage to the tubular structure of cells in the low-, medium-, high- concentration metformin groups, the degree of damage to the tubular structure of cells in the metformin+pcDNA-ALKBH3 group was reduced. The numbers of cellular tubular structures of TE-1 cells in the 6 groups were 38.35±3.20, 27.15±2.64, 15.92±3.14, 7.39±1.50, 8.61±1.37, and 29.33±4.20, respectively, with a statistically significant difference ( F=113.92, P<0.001); the number of cellular tubular structures in the low-, medium-, and high- concentration metformin groups gradually decreased (both P<0.05); the number of cellular tubular structures in the metformin+pcDNA-ALKBH3 group was more than that in the metformin+pcDNA-NC group ( P<0.05). There were statistically significant differences in the protein expressions of Bcl-2, Bax, ALKBH3, and VEGF-A in TE-1 cells among 6 groups ( F=56.36, P<0.001; F=57.26, P<0.001; F=159.30, P<0.001; F=132.89, P<0.001); compared with the blank control group, the protein expressions of Bcl-2, ALKBH3, and VEGF-A in the metformin groups at different concentrations decreased, while the protein expression of Bax increased (all P<0.05); compared with the metformin+pcDNA-NC group, the protein expressions of Bcl-2, ALKBH3, and VEGF-A in the metformin+pcDNA-ALKBH3 group increased, and the expression level of Bax decreased (all P<0.05). The weights of tumors in the model control group, metformin group, metformin+pcDNA-NC group, and metformin+pcDNA-ALKBH3 group were (1.16±0.12), (0.46±0.05), (0.50±0.06), (1.19±0.14) g, the volumes of tumors were (878.36±108.93), (413.59±50.23), (439.78±51.39), (793.75±96.98) mm 3, with statistically significant differences ( F=96.61, P<0.001; F=51.90, P<0.001); the weight of tumors were lower and the volume of tumors were smaller in the metformin group than those in the model control group (both P<0.05), the weight of tumors were higher and the volume of tumors were bigger in the metformin+pcDNA-ALKBH3 group than those in the metformin group and the metformin+pcDNA-NC group (all P<0.05). CD31 was mainly distributed in the cytoplasm and cell membrane of tumor cells. There were statistically significant differences in the positive rates of CD31 and the protein expression levels of VEGF-A in transplanted tumor tissues among 4 groups ( F=7.12, P=0.002; F=48.81, P<0.001); the positive rate of CD31 and the protein expression level of VEGF-A in the metformin group were lower than those in the model control group; the positive rate of CD31 and the protein expression level of VEGF-A in the metformin+pcDNA-ALKBH3 group were higher than those in the metformin group and the metformin+pcDNA-NC group (all P<0.05) . Conclusions:Metformin may inhibit the proliferation, migration, and tumor angiogenesis of esophageal squamous cell carcinoma by reducing ALKBH3 expression.