Journal of Evidence‐Based Medicine最新文献

Objective

China faces a significant burden of gastrointestinal tumors driven by socioeconomic, environmental, and lifestyle factors. Using GBD2021 data, this study analyses epidemiological trends and disease burden for six major gastrointestinal tumor cancers (esophagus, gastric, colorectum, liver, pancreas, gallbladder and biliary tract) in China (1990–2021).

Methods

The GBD 2021 was used to extract the incidence, mortality, and disability-adjusted life years (DALYs) data of gastrointestinal tumors in China. Age-standardized rates (ASRs) and 95% uncertainty intervals (UIs) were calculated. Temporal trends were assessed by joinpoint regression analysis, and average annual percent change (AAPC) and annual percentage change (APC) were calculated and analyzed stratified by gender and age group.

Results

In 2021, China recorded 1.96 million new gastrointestinal cancer cases, with 1.35 million deaths and 33.07 million DALYs. Gastric cancer led in mortality, and colorectal cancer demonstrated the most rapid incidence growth (AAPC = 1.68). Significant reductions were observed in gastric cancer age-standardized mortality rates (ASMR) (AAPC = –2.44) and esophageal cancer age-standardized disability-adjusted life year rates (ASDR) (AAPC = –2.31). Gender disparities were particularly pronounced in esophageal cancer, with the male-to-female mortality ratio (M/F) escalating from 2.50 (1990) to 4.12 (2021). The age group with the highest mortality burden was 70–74, while the age group with the most significant loss of DALYs was 65–69.

Conclusion

China has significantly reduced gastrointestinal cancer burden, but gender and age disparities persist, necessitating targeted interventions. Future efforts should focus on tertiary prevention for high-risk groups, especially males and the elderly, while enhancing molecular subtyping and regional data stratification for precision cancer control.

Osteoporosis (OP) is one of the most serious health problems, causing a huge economic burden to patients, families, and society. OP rehabilitation treatment plays an important role in pain-relieving, reducing the risk of fracture, improving the ability of daily activities, and promoting the healing of OP fractures, and has been increasingly valued and recommended by domestic and foreign guidelines. This guideline was updated based on the Chinese Expert Consensus on Rehabilitation Intervention for Primary OP 2019 editions and was initiated by the Chinese Society of Physical Medicine and rehabilitation and the West China Hospital of Sichuan University. This guideline development was guided by domestic and international guideline development methods and principles and selected through clinical issue screening and deconstruction and two rounds of Delphi questionnaire consultation. The International Classification of Functioning, Disability and Health (ICF) was used as the theoretical framework, and the Grading of Recommendations Assessment, Development and Evaluation (GRADE) was used based on the best available evidence. The guideline was developed based on the best available evidence, used the GRADE method to grade the quality of evidence and recommendations, and reported according to the Reporting Items for Practice Guidelines in Healthcare. Taking into account patients' preferences and values and the needs of Chinese clinical practice, it puts forward 22 recommendations covering seven aspects (rehabilitation assessment, therapeutic modalities, occupational therapy, assistive devices, cognitive behavior and psychological therapy, traditional Chinese medicine therapy, and health education) to systematically standardize OP rehabilitation.

Background

Formulating evidene-based recommendations for practice guidelines is a complex process that requires substantial expertise. Artificial intelligence (AI) is promising in accelerating the guideline development process. This study evaluates the feasibility of leveraging five large language models (LLMs)—ChatGPT-3.5, Claude-3 sonnet, Bard, ChatGLM-4, Kimi chat—to generate recommendations based on structured evidence, assesses their concordance, and explores the potential for AI.

Methods

The general and specific prompts were drafted and validated. We searched PubMed to include evidence-based guidelines related to health and lifestyle. We randomly selected one recommendation from every included guideline as the sample and extracted the evidence base supporting the selected recommendations. The prompts and evidence were fed into five LLMs to generate structured recommendations.

Results

ChatGPT-3.5 demonstrated the highest proficiency in comprehensively extracting and synthesizing evidence to formulate novel insights. Bard consistently adhered to existing guideline principles, aligning its algorithm with these tenets. Claude generated fewer topical recommendations, focusing instead on evidence analysis and mitigating irrelevant information. ChatGLM-4 exhibited a balanced approach, combining evidence extraction with adherence to guideline principles. Kimi showed potential in generating concise and targeted recommendations. Among the six generated recommendations, average consistency ranged from 50% to 91.7%.

Conclusion

The findings of this study suggest that LLMs hold immense potential in accelerating the formulation of evidence-based recommendations. LLMs can rapidly and comprehensively extract and synthesize relevant information from structured evidence, generating recommendations that align with the available evidence.

Aim

The white-coat hypertension (WCH) detection by monitoring the out-of-office blood pressure (BP) consumes resources and time. This study aimed at developing the prediction model based on patients’ characteristics obtained from clinical data.

Methods

Individuals who participated in two large hospitals health check-up examination were screened. Participants with twice readings of elevated office blood pressure in different visits, while no history of hypertension were included. Combination with home blood pressure monitoring, participants were defined as having WCH or sustained hypertension (SH), respectively. Independent predictors were found by employing multivariate logistic regression on training set. A nomogram was built using independent predictors.

Results

In total, 383 outpatients with elevated office blood pressure were enrolled. Two hundred and thirty-three of them from one hospital were included for the development of the prediction model (training sets), and 150 patients from another independent study site were included for external validation (external validation sets). We identified six predictors including office systolic blood pressure, body mass index, sex, total cholesterol, homocysteine, and heart rate being linked to WCH diagnosis. Area under receiver operating characteristic curve (AUC) for the model was 0.792 and 0.692 regarding training and external validation sets, respectively. The calibration curve and decision curve analyses further demonstrated that the model had good performance for distinguishing WCH from SH.

Conclusions

This prediction model can help clinicians to identify WCH individuals from those with SH, providing an effective tool for guiding personalized recommendations of abnormal blood pressure management.

Systematic reviews and meta-analyses are considered the highest level of evidence, but their reliability can be undermined by publication bias. Traditional methods for assessing publication bias, such as funnel plots and p-value-based tests (e.g., Egger test), have notable limitations, including reliance on subjective interpretation and dependence on the number of studies included in a meta-analysis (k). The Doi plot and LFK index offer promising alternatives, providing improved visualization and quantification of plot asymmetry. This study revisits the application of the Doi plot and LFK index for detecting publication bias, addresses recent criticisms, and evaluates their performance compared to p-value-based methods using simulation study. Simulations included scenarios with varying study numbers (k = 5, 10, 20, 50), study sample sizes (small, large), and simulated bias level (ρ = 0, –0.3, –0.5, –0.9) generated using the Copas selection model. Diagnostic performance metrics (i.e., sensitivity and specificity) were estimated and compared for the LFK index and Egger test. The LFK index exhibited consistent higher sensitivity across varying k and simulated bias levels. In contrast, the Egger test was highly dependent on k, with sensitivity declining sharply in small meta-analyses (k < 20). Specificity of the LFK index adjusted with random error, while Egger test specificity remained fixed at ∼90%. The Doi plot and LFK index effectively address the limitations of traditional methods, offering robust k-independent performance and more reliable detection of publication bias. These findings support a transition to the Doi plot and LFK index for publication bias assessment in meta-analyses.