Machine Learning Developed Tool Predicts Disease Progression

Researchers have unveiled a new prediction model that takes into account factors that may lead to the onset of advanced age-related macular degeneration (AMD), with the goal of offering a clearer path to precise care for, and education of, patients.

Researchers used the findings of two population-based cohort studies to inform a machine learning (ML) algorithm. The model took into account standard data such as genetic, phenotypic and lifestyle variables, as well as metrics never before incorporated into such a tool. These unique factors include pulse pressure and Mediterranean diet score—predictors that have previously been reported as associated with AMD and are modifiable, which could lead to preventive interventions, the study shows. Genotypic, phenotypic, and lifestyle risk factors for advanced AMD were examined in The Rotterdam Study I (RS-I), and in the Antioxydants, Lipides Essentiels, Nutrition et Maladies Oculaires (ALIENOR) Study. Mediterranean diet score was calculated using a 2013 study that found a reduction in overall mortality for subjects whose diet adhered to three different categories of consumption for each food group composing the Mediterranean diet.

RS-I comprised a training set of 3838 participants, 55 years of age or older, across a median follow-up period of close to 11 years, with 108 incident cases of advanced AMD, either atrophic or neovascular. ALIENOR’s validation test set included 362 participants 73 years of age or older, during a median follow-up span of 6.5 years, with 33 incident cases. 

“This prediction model reached high discrimination abilities, paving the way towards making precision medicine for AMD patients a reality in the near future,” the study authors wrote. 

Nine variables were found to be the most predictive of incident advanced AMD:

1. Intermediate drusen
2. Genetic risk score; based on 49 single nucleotide polymorphisms 
3. Age-Related Eye Disease Study simplified scale
4. Age
5. Smoking status
6. Pulse pressure (difference between systolic and diastolic blood pressure)
7. Hyperpigmentation
8. Education
9. Mediterranean diet score

They added that future studies with additional genetic cohorts would further improve this new tool. The online tool can be accessed by either physicians or patients at www.macutest.net. Patients are able to assess their AMD risk based on the most frequently found predisposing characteristics. The site also explains the two primary stages of AMD, presents ways to prevent onset and progression, and lists elements of a recommended diet. 

Reference

Ajana S, Cougnard-Grégoire A, Colijn J, Merle G, Predicting progression to advanced age-related macular degeneration from clinical, genetic and lifestyle factors using machine learning. Ophthalmol. 2020. doi: https://doi.org/10.1016/j.ophtha.2020.08.031.