Scholarlink

Background <p>Understanding how SGLT2 inhibitors (SGLT2is) and GLP-1 receptor agonists (GLP-1RAs) are prescribed in relation to cardiorenal risk is crucial for assessing adherence to guidelines and optimize outcomes in type 2 diabetes (T2D). This study aimed to determine whether prescription patterns across different cardiorenal phenotypic subgroups reflect clinical risk or are affected by demographic factors. Explainable artificial intelligence (XAI) was used to identify the key factors influencing therapeutic decisions.</p> Methods <p>We analyzed 139,202 adults with T2D from the Italian AMD Annals registry (2023), stratified into four cardiorenal phenotypic subgroups based on ADA criteria: low risk, chronic kidney disease (CKD) without cardiovascular disease (CVD), atherosclerotic CVD without heart failure, and heart failure. Predictive models were developed using a Logic Learning Machine (XAI algorithm), with variable importance ranked by normalized relevance scores. Business intelligence tools and statistical analysis were used for validation.</p> Results <p>SGLT2i prescriptions were strongly associated with cardiorenal markers, including reduced eGFR (31–59 mL/min), intermediate HbA1c (5.5–8.1%), and lower BMI, with model accuracies ranging from 63.7% to 83.1%. Women were consistently less likely to receive SGLT2is across subgroups. GLP-1RA prescriptions were predominantly driven by higher BMI (&gt; 30&#xa0;kg/m²), younger age, and glycemic extremes (&lt; 5.5% or &gt; 8.1%), with lower model performance (31.6%–54.1%). Individuals with lower BMI were less frequently prescribed GLP-1RAs, even in the presence of renal or cardiovascular risk.</p> Conclusions <p>While SGLT2i prescribing generally aligned with cardiorenal risk, sex-based disparities persist. GLP-1RA use was less consistently linked to clinical indications and more heavily influenced by BMI. These findings highlight missed opportunities to deliver proven cardiorenal protective therapies to high-risk individuals, emphasising the need for more equitable, phenotype-driven prescribing strategies in T2D.</p> Graphical Abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Real-world prescriptions of GLP-1RAs and SGLT2is in type 2 diabetes prioritise BMI and age over cardiorenal risk: a machine learning-based large cohort analysis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月10日

查看全文
  • Dario Tuccinardi,
  • Rita Zilich,
  • Paola Ponzani,
  • Nicoletta Musacchio,
  • Valeria Manicardi,
  • Alberto Rocca,
  • Roberto Pontremoli,
  • Paola Fioretto,
  • Andrea Muscarà,
  • Graziano Di Cianni,
  • Riccardo Candido,
  • Salvatore A. De Cosmo,
  • Giuseppina Russo

摘要

Background

Understanding how SGLT2 inhibitors (SGLT2is) and GLP-1 receptor agonists (GLP-1RAs) are prescribed in relation to cardiorenal risk is crucial for assessing adherence to guidelines and optimize outcomes in type 2 diabetes (T2D). This study aimed to determine whether prescription patterns across different cardiorenal phenotypic subgroups reflect clinical risk or are affected by demographic factors. Explainable artificial intelligence (XAI) was used to identify the key factors influencing therapeutic decisions.

Methods

We analyzed 139,202 adults with T2D from the Italian AMD Annals registry (2023), stratified into four cardiorenal phenotypic subgroups based on ADA criteria: low risk, chronic kidney disease (CKD) without cardiovascular disease (CVD), atherosclerotic CVD without heart failure, and heart failure. Predictive models were developed using a Logic Learning Machine (XAI algorithm), with variable importance ranked by normalized relevance scores. Business intelligence tools and statistical analysis were used for validation.

Results

SGLT2i prescriptions were strongly associated with cardiorenal markers, including reduced eGFR (31–59 mL/min), intermediate HbA1c (5.5–8.1%), and lower BMI, with model accuracies ranging from 63.7% to 83.1%. Women were consistently less likely to receive SGLT2is across subgroups. GLP-1RA prescriptions were predominantly driven by higher BMI (> 30 kg/m²), younger age, and glycemic extremes (< 5.5% or > 8.1%), with lower model performance (31.6%–54.1%). Individuals with lower BMI were less frequently prescribed GLP-1RAs, even in the presence of renal or cardiovascular risk.

Conclusions

While SGLT2i prescribing generally aligned with cardiorenal risk, sex-based disparities persist. GLP-1RA use was less consistently linked to clinical indications and more heavily influenced by BMI. These findings highlight missed opportunities to deliver proven cardiorenal protective therapies to high-risk individuals, emphasising the need for more equitable, phenotype-driven prescribing strategies in T2D.

Graphical Abstract