Tuesday, June 18, 2024

High Bias in Diabetic Foot Amputation Prediction Models Calls for Improvement

Similar articles

In the evolving realm of diabetic foot (DF) management, accurately predicting the likelihood of amputation remains a pivotal challenge. While numerous studies have developed and validated prediction models for such estimations, their quality and real-world applicability continue to provoke skepticism. A recent systematic review delves into the biases and practical utility of these prediction models, underscoring the need for more robust methodologies in future research endeavors.

Comprehensive Database Search

A thorough search across multiple databases, including PubMed, Web of Science, and Embase, among others, was carried out from their inception to December 24, 2023. This exhaustive search aimed to identify relevant studies that focused on prediction models for amputation in DF patients. Two investigators independently screened the literature and extracted data, ensuring a meticulous evaluation process. The Prediction Model Risk of Bias Assessment Tool (PROBAST) checklist was employed to scrutinize both the risk of bias and the applicability of these models.

Results and Findings

The analysis incorporated 20 studies, which included 17 development studies and three validation studies. These studies collectively encompassed 20 prediction models and 11 classification systems. The incidence of amputation among DF patients varied widely, from 5.9% to 58.5%. Notably, machine learning-based methods were utilized in over half of the studies, with the area under the curve (AUC) reported between 0.560 and 0.939. Commonly identified independent predictors included age, gender, HbA1c, hemoglobin, white blood cell count, low-density lipoprotein cholesterol, diabetes duration, and Wagner’s Classification.

Despite these insights, all the studies were found to exhibit a high risk of bias. The primary sources of this bias were the inadequate handling of outcome events and missing data, lack of comprehensive model performance assessments, and overfitting.

Inference Analysis

Key Takeaways for Clinical Practice

  • Machine learning methods hold promise but require rigorous evaluation to ensure reliability.
  • Consistent predictors like HbA1c and Wagner’s Classification should be prioritized in future model developments.
  • Addressing overfitting and missing data is crucial to enhance model robustness.
  • Comprehensive performance assessments must be integral to future studies.

The findings underscore a significant risk of bias in the existing prediction models for amputation in DF patients. To improve clinical outcomes, future research must focus on refining current models or developing new ones with more stringent methodological frameworks.

Original Article: Diabetol Metab Syndr. 2024 Jun 10;16(1):126. doi: 10.1186/s13098-024-01360-6.

You can follow our news on our Telegram and LinkedIn accounts.

Subscribe to our newsletter

To be updated with all the latest news, offers and special announcements.

Latest article