Abstract 4365024: Not All LDLs are Created Equal: Discordance Between Calculated LDL-C Estimates and Incident ASCVD in the Multi-Ethnic Study of Atherosclerosis

Introduction: The Friedewald equation (F-LDL-C) is the most widely used estimate of LDL-cholesterol (LDL-C), but it can be inaccurate at high TG and low LDL-C. The Martin-Hopkins (MH-LDL-C) and the Sampson (S-LDL-C) equations more accurately estimate LDL-C. Individuals with discordant LDL-C estimates may be undertreated for their ASCVD risk, depending on the equation used. The association of discordance in LDL-C estimates with ASCVD risk is not well established. Hypothesis: Individuals with greater discordance in LDL-C estimates are at higher risk for incident ASCVD Methods: We estimated F-LDL-C, MH-LDL-C, and S-LDL-C in 6636 patients (mean 61.6 years, 47% male) with TG < 400 mg/dL in the Multi-Ethnic Study of Atherosclerosis. We divided the cohort into quintiles (Q1-Q5) of LDL-C discordance, measured by the absolute difference of LDL-C values between equations (MH-LDL-C minus F-LDL-C, S-LDL-C minus F-LDL-C, MH-LDL-C minus S-LDL-C). We examined the association of characteristics (sex, age, race, BMI, diabetes, hypertension, tobacco use) with quintile of discordance using the Jonckheere-Terpstra test. We used multivariable adjusted Cox regression models to assess the hazard associated with quintiles of discordance for ASCVD events (MI, stroke, CV death, or revascularizations). Results: Greater LDL-C discordance (when MH-LDL-C and S-LDL-C were higher than F-LDL-C) was significantly associated with male sex, higher BMI, diabetes, hypertension, and tobacco use in unadjusted models. There were 1,275 ASCVD events over a median follow up of 18.4 years. The distribution of intra-quintile LDL-C values grew exponentially in Q5 across all groups, with the MH – F cohort demonstrating the greatest LDL-C range (absolute difference of 31.2 mg/dL). In fully adjusted Cox models, those in Q4 (HR 1.23, 95% CI 1.03-1.48) and Q5 (HR 1.28, 95% CI 1.06-1.55) of LDL-C discordance (where MH-LDL-C was higher than F-LDL-C) had a higher hazard for ASCVD events. Greater discordance between S-LDL-C and F-LDL-C trended towards increased ASCVD risk, but the results were not statistically significant. Conclusions: Greater LDL-C discordance where MH-LDL-C was higher than F-LDL-C is independently associated with greater ASCVD, after adjustments for variables associated with higher discordance. Our findings favor using newer equations to estimate LDL-C and suggest that this high-risk group is at risk for being undertreated if targeting the widely used F-LDL-C estimates.