Nomogram-based model to predict prognosis in anisometropic amblyopia

Wenyuan Xu, Xiaoman Li, Lizhong Wang, Xiaogang Xiang, Youjian Peng, Hongyi Li, Xuewen Ding, Jianing Zhang, Xiaoyue Hu, Jie Chen, Wuhe ChenFrontiers in Medicine 2025 · DOI

Abstract: The nomogram, incorporating baseline age, differences in interocular BCVA, and SE, provides individualized prognostic predictions for anisometropic amblyopia treatment, aiding clinicians in personalized treatment planning and better resource allocation. Furthermore, the nomogram could enhance shared decision-making with parents by providing objective prognostic data, thus improving treatment compliance.

Introduction

Amblyopia, commonly known as "lazy eye," affects approximately 1.36% of the global population. Anisometropia—a significant difference in refractive error between the two eyes—is a leading cause of this condition. While standard treatments such as optical correction and patching are effective, the success rate varies widely among patients. Historically, clinicians have lacked a standardized, objective method to predict which children are most likely to achieve a "cure," defined as a best-corrected visual acuity (BCVA) of 0.1 LogMAR or better with minimal interocular difference.

This research, published in Frontiers in Medicine, addresses this gap by developing a validated nomogram. By moving away from a "one-size-fits-all" approach, this tool enables personalized, data-driven prognoses that can improve clinical resource allocation and parental communication.

Methodology and Feature Selection

The study involved a retrospective analysis of 2,897 patients aged 3 to 18 years, treated over a 20-year period at the Eye Hospital of Wenzhou Medical University. This large-scale dataset, which included a minimum 12-month follow-up, was divided into a training set (70%) and a validation set (30%).

To identify the most critical prognostic factors from an initial pool of 15 clinical variables, the researchers employed Least Absolute Shrinkage and Selection Operator (LASSO) regression. This machine learning technique penalizes less significant variables, ensuring the final model is both accurate and clinically practical.

Figure 1: Feature selection using LASSO regression to identify the most significant clinical predictors.

Following feature selection, the authors utilized Restricted Cubic Splines (RCS) to model the non-linear relationships between baseline age, initial vision loss, and treatment outcomes. This approach acknowledges that the difficulty of treating amblyopia does not increase at a constant rate as a child ages.

Key Results and Model Performance

The analysis revealed that three primary factors are essential for predicting a cure: baseline age, interocular BCVA difference, and interocular Spherical Equivalence (SE) difference. Among these, age was the most powerful individual predictor (AUC 0.703), reinforcing the importance of the "critical period" in neuroplasticity.

The final nomogram demonstrated strong predictive accuracy:

  • Area Under the Curve (AUC): 0.783 in the training set and 0.782 in the validation set, indicating a "good" and clinically useful model.
  • Calibration: Calibration curves showed excellent agreement between predicted cure probabilities and actual observed outcomes.
  • Clinical Utility: Decision Curve Analysis (DCA) confirmed that the nomogram provides a higher "net benefit" compared to treating all patients uniformly or not treating them at all.

Discussion and Clinical Implications

By converting complex logistic regression formulas into a visual nomogram, the researchers have created a practical tool for daily ophthalmic practice. During a consultation, a clinician can use the nomogram to provide parents with a specific "Cure Probability" percentage. This objective data is expected to enhance shared decision-making and improve treatment compliance by setting realistic expectations.

Limitations and Future Directions

Despite its strengths, the study has limitations. As a retrospective analysis, it could not objectively measure patient compliance with patching protocols. Additionally, the model does not yet include high-level visual functions like stereopsis (3D depth perception). Future research should involve multi-center studies across diverse geographic populations to further validate the model's generalizability.

In conclusion, this nomogram provides an evidence-based bridge between statistical modeling and pediatric ophthalmic care, offering a new standard for personalized amblyopia management.

Summary updated

Abstract

The nomogram, incorporating baseline age, differences in interocular BCVA, and SE, provides individualized prognostic predictions for anisometropic amblyopia treatment, aiding clinicians in personalized treatment planning and better resource allocation. Furthermore, the nomogram could enhance shared decision-making with parents by providing objective prognostic data, thus improving treatment compliance.
BibTeX
@inproceedings{xu2025nomogram,
  author = {Wenyuan Xu and Xiaoman Li and Lizhong Wang and Xiaogang Xiang and Youjian Peng and Hongyi Li and Xuewen Ding and Jianing Zhang and Xiaoyue Hu and Jie Chen and Wuhe Chen},
  title = {{Nomogram-based model to predict prognosis in anisometropic amblyopia}},
  booktitle = {Frontiers in Medicine},
  year = {2025},
  doi = {10.3389/fmed.2025.1612845},
}
Figure