Outcomes and risk factors for failure after trabeculectomy in Taiwanese patients: medical chart reviews from 2006 to 2017

Predicting Glaucoma Surgical Outcomes Using Neural Networks and Machine Learning on Electronic Health Records

Purpose

To develop machine learning (ML) and deep learning (DL) models to predict glaucoma surgical outcomes, including postoperative intraocular pressure, use of ocular antihypertensive medications, and need for repeat surgery.

Methods

We identified glaucoma surgeries performed at Stanford from 2013-2024, with two or more postoperative visits with intraocular pressure (IOP) measurement. Patient features were identified from the electronic health record (EHR), including demographics, prior diagnosis and procedure codes, medications and eye exam findings. Classical ML and DL models were developed to predict which glaucoma surgeries would result in surgical failure, defined as (1) IOP not reduced by more than 20% of preoperative baseline on two consecutive postoperative visits, (2) increased classes of glaucoma medications, and (3) need for additional glaucoma surgery or revision of original surgery.

Results

A total of 2398 glaucoma surgeries of 1571 patients were included, of which 1677 surgeries met failure criteria. Random forest performed best for prediction of overall surgical failure, with accuracy of 75.5% and area under the receiver operator curve (AUROC) of 76.7%, similar to the deep learning model (accuracy 75.5%, AUROC 76.6%). Across all models, prediction performance was better for IOP outcomes (AUROC 86%) than need for an additional surgery (AUROC 76%) or need for additional glaucoma medication (AUC 70%).

Conclusions

ML and DL algorithms can predict glaucoma surgery outcomes using structured data inputs from EHRs.

Translational Relevance

Models that predict outcomes of glaucoma surgery may one day provide the basis for clinical decision support tools supporting surgeons in personalizing glaucoma treatment plans.

Bleb-related infection after primary trabeculectomy: medical chart reviews from 1993 to 2021

BACKGROUND

To investigate the incidence of and risk factors for bleb-related infection (BRI) in patients who underwent mitomycin C-augmented primary trabeculectomy.

METHODS

We reviewed the medical charts of consecutive patients who had received primary trabeculectomy in Taipei Veterans General Hospital. We recorded the demographic and clinical characteristics of patients before, during and after surgery. Furthermore, we recorded the time interval between surgery and infection onset, clinical manifestations and visual outcomes of patients with BRI. The cumulative incidence of BRI was estimated using the Kaplan-Meier method. A Cox proportional hazards model was used to explore factors associated with BRI.

RESULTS

In total, 1663 eyes were postoperatively followed up for 94.57±65.23 months. The cumulative incidence of BRI was 1.86 per 1000 person-years during the 28-year follow-up period: 24 (1.44%) patients developed BRI and 6 (0.36%) patients additionally developed endophthalmitis. A multivariate analysis revealed a significant association of BRI with wound manipulation, high myopia and hyperlipidaemia. Patients younger than 60 years were more likely to receive wound manipulation than their elderly counterparts (<0.001). One year after BRI, the best corrected visual acuity of the eyes with blebitis did not change significantly, whereas that of the eyes with endophthalmitis worsened significantly.

CONCLUSION

Risk factors for BRI after trabeculectomy include wound manipulation, high myopia and hyperlipidaemia. Considering myopia is highly prevalent throughout the world and is a risk factor for glaucoma, the lifelong risk of BRI after trabeculectomy in eyes with high myopia warrants the attention of ophthalmologists.

Efficacy and Safety of the Intraoperative Application of Mitomycin in Glaucoma Patients with Trabeculectomy: A Systematic Review and Meta-Analysis

What Is Known and Objective. The primary cause of trabeculectomy failure in glaucoma surgery is the imperfect formation of the filter bubble, which blocks the filtration effect. This systematic review and meta-analysis aimed to compare the effects and safety of mitomycin C (MMC) or no antimetabolite in trabeculectomies cases that require needling revision. Methods. We searched PubMed, Cochrane, and EMBASE to identify randomized trials published between the time the databases were built and May 31, 2022. To compare the effectiveness and safety of mitomycin with or without mitomycin in trabeculectomy, intraocular pressure (IOP), the surgical failure rate, and functional follicle formation were used as efficacy indicators, and the occurrence of postoperative complications was used as a safety indicator. Meta-analyses were performed for comparisons. Results and Discussion. After trabeculectomy, MMC was significantly better than a placebo at lowering postoperative IOP (MD = −11.31 mmHg, 95% CI −19.73 to −2.88 mmHg). MMC outperformed the control group in terms of filtering blebs formation (relative risk (RR) = 1.18, 95% CI: 1.09 to 1.27). The surgical failure rate was significantly lower with MMC compared to placebo (RR = 0.35, 95% CI: 0.22 to 0.58). No significant difference was observed between MMC and placebo in terms of hypotony, anterior chamber bleeding, filter bubble leakage, and endophthalmitis, apart from the shallow anterior chamber (RR: 1.51, 95% CI: 1.02 to 2.25). What Is New and Conclusion. The use of MMC in glaucoma trabeculectomy can increase the success rate of the procedure and provide significant patient benefits. However, it is important to be extremely careful and aware of complications such as shallow anterior chambers.

Long-term success after trabeculectomy in open-angle glaucoma: results of a retrospective cohort study

OBJECTIVE

To evaluate the long-term outcomes of trabeculectomy (TE) surgery in a large cohort with a minimum follow-up of 3 years.

DESIGN

Retrospective cohort study.

SETTING

University Eye Hospital, Germany.

PARTICIPANTS

Three hundred and seventy-nine patients with open-angle glaucoma underwent TE with mitomycin C (MMC) between January 2013 and February 2017 with a minimal follow-up of 3 years. Eligible patients were identified via an electronic surgical case register.

INTERVENTIONS

All patients had undergone TE with MMC following a set surgical protocol. To assess the influence of cataract surgery following TE, eyes which underwent cataract surgery at least 6 months after TE were matched 1:3 by sex and age to eyes who did not undergo cataract surgery during the follow-up period.

MAIN OUTCOME MEASURES

Primary outcome was the proportion of surgical success based on intraocular pressure (IOP), surgical complications, the need for revision surgery, loss of light perception and the need for additional pressure-lowering medication.

RESULTS

The mean follow-up time was 6 (±0.8, IQR: 5.4-6.5) years. Seventy-three per cent of eyes achieved qualified surgical success at the last follow-up (IOP≥5 mm Hg and ≤18 mm Hg, without surgical complications or complete loss of vision) but necessitated additional medical therapy, complete surgical success with no additional medical therapy was achieved in 69% of eyes. There was no significant difference in the success probability between eyes that had undergone cataract surgery after TE and those that had not (p=0.45).

CONCLUSIONS

The results demonstrate a high and stable success rate of TE after a mean follow-up time of approximately 6 years, that is, not affected by later cataract surgery.

Evaluation of machine learning algorithms for trabeculectomy outcome prediction in patients with glaucoma

The purpose of this study was to evaluate the performance of machine learning algorithms to predict trabeculectomy surgical outcomes. Preoperative systemic, demographic and ocular data from consecutive trabeculectomy surgeries from a single academic institution between January 2014 and December 2018 were incorporated into models using random forest, support vector machine, artificial neural networks and multivariable logistic regression. Mean area under the receiver operating characteristic curve (AUC) and accuracy were used to evaluate the discrimination of each model to predict complete success of trabeculectomy surgery at 1 year. The top performing model was optimized using recursive feature selection and hyperparameter tuning. Calibration and net benefit of the final models were assessed. Among the 230 trabeculectomy surgeries performed on 184 patients, 104 (45.2%) were classified as complete success. Random forest was found to be the top performing model with an accuracy of 0.68 and AUC of 0.74 using 5-fold cross-validation to evaluate the final optimized model. These results provide evidence that machine learning models offer value in predicting trabeculectomy outcomes in patients with refractory glaucoma.