Scholarlink

Background <p>Severe hypoglycemia (SH) in adults with type 1 diabetes mellitus (T1DM) is associated with significant morbidity and mortality; however, its underlying causes are often complex and multifactorial. Improved tools to identify individuals at a high risk of SH are critically needed. In this study, machine learning techniques were applied to continuous glucose monitoring (CGM) data to identify distinguishing features between individuals with and without SH episodes.</p> Methods <p>We analyzed data from the real-world study of adults with T1DM enrolled in the FGM-Japan study. Eleven machine learning algorithms using continuous glucose monitoring (CGM) metrics were applied to identify SH and assess the relative importance of the contributing features. The CGM metrics included mean glucose/GMI, time above range (TAR &gt; 250 and &gt; 180&#xa0;mg/dL), time in range (TIR 70–180&#xa0;mg/dL), time below range (TBR &lt; 70 and &lt; 54&#xa0;mg/dL), coefficient of variation (%CV), and glycemic risk index (GRI). The model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, and F1 score.</p> Results <p>Data from 264 adults with T1DM were analyzed. Across the models, XGBoost showed the highest AUC, significantly outperforming logistic regression, k-NN, and SVM but performed marginally below Naive Bayes. The F1-score analysis showed that logistic regression and neural networks provided a better balance between precision and recall. The model using four CGM variables (TBR &lt; 70, %CV, GMI, and GRI) achieved the highest AUC of 0.794.</p> Conclusions <p>XGBoost offers strong overall discrimination; however, simpler models exhibit better F1 performance. Features like ‘TBR’, ‘%CV’, ‘GMI,’ and ‘GRI’ were key features, suggesting their usefulness in identifying individuals at risk for adverse glycemic events.</p> Trial registration <p>Clinical Trial Registry No. UMIN000039376.</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

Identification of severe hypoglycemia in adults with type 1 diabetes using CGM-based machine learning: evidence from the FGM-Japan study

  • 期刊论文

  • 发表日期：

    2026年2月1日

查看全文
  • Naoki Sakane,
  • Yushi Hirota,
  • Akane Yamamoto,
  • Junnosuke Miura,
  • Hiroko Takaike,
  • Sari Hoshina,
  • Masao Toyoda,
  • Nobumichi Saito,
  • Kiminori Hosoda,
  • Masaki Matsubara,
  • Atsuhito Tone,
  • Satoshi Kawashima,
  • Hideaki Sawaki,
  • Tomokazu Matsuda,
  • Masayuki Domichi,
  • Akiko Suganuma,
  • Seiko Sakane,
  • Takashi Murata

摘要

Background

Severe hypoglycemia (SH) in adults with type 1 diabetes mellitus (T1DM) is associated with significant morbidity and mortality; however, its underlying causes are often complex and multifactorial. Improved tools to identify individuals at a high risk of SH are critically needed. In this study, machine learning techniques were applied to continuous glucose monitoring (CGM) data to identify distinguishing features between individuals with and without SH episodes.

Methods

We analyzed data from the real-world study of adults with T1DM enrolled in the FGM-Japan study. Eleven machine learning algorithms using continuous glucose monitoring (CGM) metrics were applied to identify SH and assess the relative importance of the contributing features. The CGM metrics included mean glucose/GMI, time above range (TAR > 250 and > 180 mg/dL), time in range (TIR 70–180 mg/dL), time below range (TBR < 70 and < 54 mg/dL), coefficient of variation (%CV), and glycemic risk index (GRI). The model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, and F1 score.

Results

Data from 264 adults with T1DM were analyzed. Across the models, XGBoost showed the highest AUC, significantly outperforming logistic regression, k-NN, and SVM but performed marginally below Naive Bayes. The F1-score analysis showed that logistic regression and neural networks provided a better balance between precision and recall. The model using four CGM variables (TBR < 70, %CV, GMI, and GRI) achieved the highest AUC of 0.794.

Conclusions

XGBoost offers strong overall discrimination; however, simpler models exhibit better F1 performance. Features like ‘TBR’, ‘%CV’, ‘GMI,’ and ‘GRI’ were key features, suggesting their usefulness in identifying individuals at risk for adverse glycemic events.

Trial registration

Clinical Trial Registry No. UMIN000039376.