Familial Hypercholesterolemia Identification by Machine Learning Using Lipid Profile Data Performs as Well as Clinical Diagnostic Criteria

Abstract

Background: Familial hypercholesterolemia (FH) is a common genetic disorder and, if not diagnosed and treated early, results in premature cardiovascular disease. Most individuals with FH are undiagnosed and machine learning offers a new prospect to improve FH identification. Our objective was to create a machine learning model from basic lipid profile data with better screening performance than LDL-C (low-density lipoprotein cholesterol) cutoff levels and diagnostic performance comparable to the Dutch Lipid Clinic Network criteria. Methods: The model was developed combining logistic regression, deep learning, and random forest classification and trained on a 70% split of a data set of individuals clinically suspected of having FH. Model performance, as well as that of the LDL-C cutoff and Dutch Lipid Clinic Network criteria, were assessed on the internal 30% testing data set and an external data set by comparing the area under the receiver operator characteristic (AUROC) curves. All methodologies were measured against the gold standard of FH diagnosis by mutation identification. Furthermore, the model was also tested on 2 lower prevalence data sets. Results: The machine learning model achieved an AUROC curve of 0.711 on the external data set (n=1376; FH prevalence=64%), which was superior to the LDL-C cutoff (AUROC=0.642) and comparable to the Dutch Lipid Clinic Network criteria (AUROC=0.705). The model performed even better when tested on the medium-prevalence (n=2655; FH prevalence=20%) and low-prevalence (n=1616; FH prevalence=1%) data sets, with AUROC curve values of 0.801 and 0.856, respectively. Conclusions: Despite absence of clinical information, the model better identified genetically confirmed FH in a cohort of individuals suspected of having FH than LDL-C cutoff values and was comparable to the Dutch Lipid Clinic Network criteria. The model achieved higher accuracy when tested on 2 cohorts with lower FH prevalence. The application of machine learning is, therefore, a promising tool in both the screening for, and diagnosis of, individuals with FH.

Applications of machine learning in familial hypercholesterolemia

“…In line with EHR, machine learning models built on basic lipid data training can also be applied to disease screening. Hesse et al ( 22 ) used the primary lipid profile data [Total Cholesterol (TC), High-density lipoprotein cholesterol, LDL-C and Triglycerides] from the laboratory information systems ( n = 555, 68% White individuals, 26% Indian individuals, and 3.2% Black African individuals) to create an ML model that combined logistic regression (LR), deep learning, and RF classification algorithms. In this study, patients with blood lipid levels exceeding LDL-C cutoff (4.5 mmol/L) and a model labeled probability of disease greater than 60% were identified as likely or clear FH.…”

“…This model was trained on 70% of the internal data sets, and outperformed the LDL-C threshold in both the 30% internal validation set test (AUROC 0.754 vs. 0.682) and the external validation (AUROC 0.711 vs. 0.642) of the Groote Schuur Hospital database ( n = 1,376; FH prevalence = 64%), showing better performance and accuracy. In addition, the accuracy and F score of the model were higher in the medium and low prevalence cases with AUROC curve values of 0.801 and 0.856, respectively ( 22 ). Therefore, based on simple lipid spectrometry data, the ML model still accurately identifies FH patients and has better screening performance than the LDL-C cutoff value.…”

“…There are no uniform criteria for the diagnosis of FH, and the most commonly used criteria in clinical practice are the DLCN criteria (9,32) and the Simon Broome criteria (33). However, these As demonstrated in the previous FINDFH study (22), the ML models have recognition performance comparable to DLCN criteria in internal validation (n = 166, AUROC 0.754 vs. 0.755) and external validation of high prevalence (64%) (n = 1,376, AUROC 0.711 vs. 0.705), suggesting that ML models can replace clinical criteria and provide new insights for future FH diagnostics. However, different algorithmic models exhibit performance differences in identifying FH, which reflects the importance of algorithmic model selection when developing alternative diagnostic procedures (35).…”

Inteligencia artificial. Aplicación en las imágenes cardiovasculares y la prevención cardiovascular

“…En este contexto, varios estudios han demostrado que los modelos basados en IA tienen un excelente valor predictivo para detectar casos genéticamente confirmados de HF. (48)(49)(50)(51) Inclusive, ese valor predictivo fue mejor que los métodos tradicionales basados en puntajes de riesgo, como los criterios de la Red Holandesa de las Clínicas de Lípidos.…”

LDLR c.415G > A causes familial hypercholesterolemia by weakening LDLR binding to LDL

“…At present, the diagnosis rate of FH is very low in most countries and regions [ 7 ]. For example, it is < 10% in the United States [ 8 ], 4% in Australia and New Zealand [ 9 ], 2% in South Africa [ 10 ], and even < 1% in Russia, Latin America, and other countries [ 11 ]. Only a small percentage of those diagnosed with FH have undergone genetic testing.…”