Introduction:
While LDL-C concentration remains a key indicator in lipid profile assessment and cardiovascular risk prediction, non-HDL cholesterol (non-HDL-C) was acknowledged as a superior marker, particularly among people with elevated levels of triglycerides (TG). However, non-HDL-C is not always calculated or reported in laboratory panels.

Aim:
The main purpose of this study is to evaluate the effectiveness of artificial neural networks (ANNs) in predicting non-HDL-C levels based on standard lipid parameters, i.e. LDL-C and TG.

Material and methods:
Based on 26,914 anonymized lipid profiles (total cholesterol [TC], TG, LDL-C, non-HDL-C), an MLP 2-4-1 neural network was trained to estimate non-HDL-C values.

Results and discussion:
The non-HDL-C values predicted by the model showed a strong correlation with laboratory results (R² = 0.944 overall). Correlation was slightly higher for TG < 400 mg/dL (R² = 0.945) and remained relatively high in the hypertriglyceridemic subgroup (R² = 0.896).

Conclusions:
This study demonstrates that artificial neural networks are able to accurately estimate non-HDL-C values using the results of lipid parameters survey, even in the cases of severe hypertriglyceridemia. Given its high prediction accuracy and its simplicity, the model is promising for the future as a practical decision-support tool in cardiovascular risk assessment, especially where non-HDL-C is unavailable or unreported.