NASHmap: clinical utility of a machine learning model to identify patients at risk of NASH in real-world settings.

The NASHmap model is a non-invasive tool using 14 variables (features) collected in standard clinical practice to classify patients as probable nonalcoholic steatohepatitis (NASH) or non-NASH, and here we have explored its performance and prediction accuracy. The National Institute of Diabetes and Digestive Kidney Diseases (NIDDK) NAFLD Adult Database and the Optum Electronic Health Record (EHR) were used for patient data. Model performance metrics were calculated from correct and incorrect classifications for 281 NIDDK (biopsy-confirmed NASH and non-NASH, with and without stratification by type 2 diabetes status) and 1,016 Optum (biopsy-confirmed NASH) patients. NASHmap sensitivity in NIDDK is 81%, with a slightly higher sensitivity in T2DM patients (86%) than non-T2DM patients (77%). NIDDK patients misclassified by NASHmap had mean feature values distinct from correctly predicted patients, particularly for aspartate transaminase (AST; 75.88 U/L true positive vs 34.94 U/L false negative), and alanine transaminase (ALT; 104.09 U/L vs 47.99 U/L). Sensitivity was slightly lower in Optum at 72%. In an undiagnosed Optum cohort at risk for NASH (n = 2.9 M), NASHmap predicted 31% of patients as NASH. This predicted NASH group had AST and ALT mean levels above normal range of 0-35 U/L, and 87% had HbA1C levels > 5.7%. Overall, NASHmap demonstrates good sensitivity in predicting NASH status in both datasets, and NASH patients misclassified as non-NASH by NASHmap have clinical profiles closer to non-NASH patients.

Development of a novel machine learning model to predict presence of nonalcoholic steatohepatitis.

OBJECTIVE: To develop a computer model to predict patients with nonalcoholic steatohepatitis (NASH) using machine learning (ML). MATERIALS AND METHODS: This retrospective study utilized two databases: a) the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) nonalcoholic fatty liver disease (NAFLD) adult database (2004-2009), and b) the Optum® de-identified Electronic Health Record dataset (2007-2018), a real-world dataset representative of common electronic health records in the United States. We developed an ML model to predict NASH, using confirmed NASH and non-NASH based on liver histology results in the NIDDK dataset to train the model. RESULTS: Models were trained and tested on NIDDK NAFLD data (704 patients) and the best-performing models evaluated on Optum data (~3,000,000 patients). An eXtreme Gradient Boosting model (XGBoost) consisting of 14 features exhibited high performance as measured by area under the curve (0.82), sensitivity (81%), and precision (81%) in predicting NASH. Slightly reduced performance was observed with an abbreviated feature set of 5 variables (0.79, 80%, 80%, respectively). The full model demonstrated good performance (AUC 0.76) to predict NASH in Optum data. DISCUSSION: The proposed model, named NASHmap, is the first ML model developed with confirmed NASH and non-NASH cases as determined through liver biopsy and validated on a large, real-world patient dataset. Both the 14 and 5-feature versions exhibit high performance. CONCLUSION: The NASHmap model is a convenient and high performing tool that could be used to identify patients likely to have NASH in clinical settings, allowing better patient management and optimal allocation of clinical resources.

Using Patient-Level Data to Develop Meaningful Cross-Trial Comparisons of Visual Impairment in Individuals with Diabetic Macular Edema.

INTRODUCTION: The aim of this study was to assess the impact of baseline characteristics on visual outcome of patients with diabetic macular edema and compare the results of clinical trials with different patient populations. METHODS: A model was created with patient-level data from the RESPOND/RESTORE trials to estimate the impact of baseline characteristics on increases in best-corrected visual acuity (BCVA) with anti-vascular endothelial growth factor therapies, measured by letters gained on the Early Treatment Diabetic Retinopathy Study scale from baseline to month 12. Mean BCVA gains with ranibizumab 0.5 mg pro re nata or laser photocoagulation monotherapy were predicted, assuming baseline characteristics equivalent to those in the VIVID-DME/VISTA-DME trials. These results were compared with the gain with aflibercept 2.0 mg every 8 weeks in VIVID-DME/VISTA-DME. Sensitivity analyses assessed outcome robustness. RESULTS: Baseline BCVA and central retinal thickness differed significantly between trials. In unadjusted data, patients in RESPOND/RESTORE receiving ranibizumab gained an additional 6.6 letters [95% confidence interval (CI): 4.5-8.7] compared with patients receiving laser monotherapy. After adjusting data to assume baseline characteristics equivalent to VIVID-DME/VISTA-DME, patients receiving ranibizumab were predicted to gain an additional 9.9 letters (95% CI: 7.3-12.4) compared with those receiving laser monotherapy. These results were similar (0.1-letter difference in favor of aflibercept; 95% CI: -2.9 to 3.2; P = 0.94) to the gain in BCVA in patients receiving aflibercept in VIVID-DME/VISTA-DME compared with those receiving laser monotherapy (10.0 letters, 95% CI: 8.3-11.7). CONCLUSION: After adjusting for baseline characteristics, the difference in letters gained between patients receiving ranibizumab versus aflibercept was non-significant across trials, highlighting the importance of adjusting for baseline characteristics in future comparisons. FUNDING: Novartis Pharma AG.

Review and comparison of methodologies for indirect comparison of clinical trial results: an illustration with ranibizumab and aflibercept.

AIM: To review and compare methods for indirect comparison of aflibercept and ranibizumab in patients with diabetic macular edema. METHODS: Post-stratification, inverse probability weighting based on simulated data, weight optimization, and regression model techniques were used to compare pooled individual patient-level data from the RESTORE and RESPOND (ranibizumab 0.5 mg as needed after 3 initial monthly doses) studies with summary-level data from the VIVID and VISTA (aflibercept 2.0 mg every 8 weeks after 5 initial monthly doses, 2q8) studies. The impact of adjusting for up to two baseline characteristics was assessed. RESULTS: All methods provided similar results. After adjustment for baseline best-corrected visual acuity and central retinal thickness, no statistically significant difference in average gain in baseline best-corrected visual acuity from baseline at month 12 was found between ranibizumab 0.5 mg and aflibercept 2q8. CONCLUSIONS: Weight optimization and regression methods are useful options to adjust for more than one baseline characteristic.

Comparative efficacy and safety of approved treatments for macular oedema secondary to branch retinal vein occlusion: a network meta-analysis.

OBJECTIVE: To compare the efficacy and safety of approved treatments for macular oedema secondary to branch retinal vein occlusion (BRVO). DESIGN: Randomised controlled trials (RCTs) evaluating the efficacy and safety of approved treatments for macular oedema secondary to BRVO were identified from an updated systematic review. SETTING: A Bayesian network meta-analysis of RCTs of treatments for macular oedema secondary to BRVO. INTERVENTIONS: Ranibizumab 0.5 mg pro re nata, aflibercept 2 mg monthly (2q4), dexamethasone 0.7 mg implant, laser photocoagulation, ranibizumab+laser, or sham intervention. Bevacizumab and triamcinolone were excluded. OUTCOME MEASURES: Efficacy outcomes were mean change in best corrected visual acuity (Early Treatment Diabetic Retinopathy Study scale) and the percentage of patients gaining ≥ 15 letters. Safety outcome was the percentage of patients with increased intraocular pressure (IOP)/ocular hypertension (OH). RESULTS: 8 RCTs were identified for inclusion with 1743 adult patients. The probability of being the most efficacious treatment at month 6 or 12 based on letters gained was 54% for ranibizumab monotherapy, 30% for aflibercept, 16% for ranibizumab plus laser (adjunctive or prompt), and 0% for dexamethasone implant, laser or sham. The probability of being the most efficacious treatment for patients gaining ≥ 15 letters was 39% for aflibercept, 35% for ranibizumab monotherapy, 24% for ranibizumab plus laser, 2% for dexamethasone implant, and less than 1% for laser or sham. There was no statistical difference between ranibizumab monotherapy and aflibercept for letters gained (+1.4 letters for ranibizumab vs aflibercept with 95% credible interval (CrI) of -5.2 to +8.5 letters) or the OR for gaining ≥ 15 letters: 1.06 (95% CrI 0.16 to 8.94)). Dexamethasone implant was associated with significantly higher IOP/OH than antivascular endothelial growth factor agents (OR 13.1 (95% CrI 1.7 to 116.9)). CONCLUSIONS: There was no statistically significant difference between ranibizumab and aflibercept.