Predictive models for clinical features associated with pediatric neurofibromatosis type 1

Machine learning models that include clinical and demographic features predicted accurate clinical features of neurofibromatosis type 1 (NF1), according to study results published in Neurology Clinical Practice.

The study researchers extracted electronic health data and clinical registry information from data under Washington University’s NF Clinical Program (collected 2002-2016). They included a total of 27 unique clinical features and longitudinal data on NF1 development in the models. The researchers used a machine learning approach to predict the evolution of NF1-associated clinical features.

The NF1 clinical registry compared patients (n = 798) who were 47.7% girls or women with a mean age of 13.0 (standard deviation [SD], 11.0) years. 81.1% of the patients were white.

Continue reading

Among the pediatric population (n = 578), more boys had attention-deficit hyperactivity disorder (40.3% vs. 26.3%; P <0.001), girls with scoliosis (25.2% vs. white patients with skin folds (97 , 2% vs. 90.9%; P = 0.04) and white patients with Lisch nodules (63.6% vs. 50.0%; P = 0.009), visual pathway glioma (20.9% vs. 10, 6%; P =. 02) and T2 hyperintenism basal ganglia (25.3% vs 14.2%; P = 0.02) and cerebellum (25.7% vs 13.3%; P = 0.01) .

Children in the NF1 cohort were more likely to have a maternal family history of NF1 (28.3%) than a paternal history (17.3%; X2, 15.5; P <0.001).

Using the clinical and demographic characteristics and electronic health record data, study researchers best said the onset of visual pathway glioma with an area under the recipient’s characteristic curve (AUC) of 0.82 (SD, ± 0.06) and a sensitivity of 0 .78 (SD, ± 0.06), specificity of 0.78 (SD, ± 0.07) and positive predictive value (PPV) of 0.78 (SD, ± 0.05).

The model was with an AUC of 0.74 (SD, ± 0.05), a sensitivity of 0.67 (SD, ± 0.05), a specificity of 0.68 (SD, ± 0.08) and a PPV from. second best in predicting the occurrence of attention deficit hyperactivity disorder 0.68 (SD, ± 0.06). The model performed worst in predicting plexiform neurofibromas (AUC 0.69 ± 0.08; sensitivity 0.62 ± 0.07; specificity 0.66 ± 0.09; PPV 0.65 ± 0.1).

This study was based on electronic health records and may contain bias and inaccurate or missing data.

These data indicated that clinical and demographic features could be used in models to predict the occurrence of 3 clinical features typically diagnosed in patients with NF1. The study’s researchers concluded, “Naive machine learning techniques can potentially be used to develop and validate predictive phenotype complexes applicable to risk stratification and disease management in NF1.”

reference

Morris SM, Gupta A, Kim S, Foraker RE, Gutmann DH, Payne PRO. Predictive modeling of clinical features associated with neurofibromatosis type 1. Neurol Clin Pract. Published online April 14, 2021. doi: 10.1212 / CPJ.0000000000001089