620 likes | 637 Views
Are Cholesterol Measures of LDL and HDL Falling Short? Insights from Recent Clinical Outcomes Data W. Virgil Brown, M.D. Total Cholesterol (+). fractionation. VLDL Cholesterol (+). LDL Cholesterol (+). HDL Cholesterol (-). Non-HDL Cholesterol (+).
E N D
Are Cholesterol Measures of LDL and HDL Falling Short? Insights from Recent Clinical Outcomes Data W. Virgil Brown, M.D.
Total Cholesterol (+) fractionation VLDL Cholesterol (+) LDL Cholesterol (+) HDL Cholesterol (-) Non-HDL Cholesterol (+) Cholesterol Fractions are the Traditional Measures of Lipoprotein-Related CVD Risk
This is LDL This is LDL Cholesterol (the cholesterol in LDL) LDL Cholesterol Is Not LDL!! (and the same is true for HDL!!) (measured by NMR or estimated by apoB) A convenient analytic surrogate of LDL since 1972 (Friedewald eq.)
LDL HDL VLDL NMR Spectroscopy Measures Lipoprotein Particles, Not The Cholesterol They Contain “The whole is the sum of its parts” (<1 min) Derived amplitudes of the subclass signals that produce the measured plasma signal (obtained by “deconvolution”) give the subclass concentrations Clin Lab Med 2006;26:847-70
VLDL Particle Number (VLDL-P) VLDL Size LDL Particle Number (LDL-P) LDL Size HDL Particle Number (HDL-P) HDL Size NMR LipoProfile Analysis Overview Lipoprotein Subclass Particle Numbers are Directly Measured Large VLDL Medium VLDL Small VLDL IDL Large LDL Small LDL Large HDL Med HDL Small HDL units of nmol/L µmol/L
Lipoprotein Subclass Particle Numbers: Large VLDL Medium VLDL Small VLDL IDL Large LDL Small LDL Large HDL Med HDL Small HDL VLDL Particle Number (VLDL-P) VLDL Size LDL Particle Number (LDL-P) LDL Size HDL Particle Number (HDL-P) HDL Size NMR LipoProfile Analysis Overview Cardiovascular Applications
NMR LipoProfile Analysis Overview Diabetes - Insulin Resistance Applications Lipoprotein Subclass Particle Numbers: Large VLDL Medium VLDL Small VLDL IDL Large LDL Small LDL Large HDL Med HDL Small HDL VLDL Particle Number (VLDL-P) VLDL Size LDL Particle Number (LDL-P) LDL Size HDL Particle Number (HDL-P) HDL Size
= Why LDL Cholesterol LDL Cholesterol Content of Particles Varies Widely This is LDL This is LDL Cholesterol (the cholesterol in LDL)
100 mg/dL 100 mg/dL Cholesterol Balance LDL-C Under-Represents LDL When Particles Are Small Large LDL Small LDL (less cholesterol per particle) More Particles Clin Cardiol 1999;22(6 Suppl):1121-27
100 mg/dL 100 mg/dL Cholesterol Balance Cholesterol Triglycerides LDL-C Under-Represents LDL WhenParticles Contain Less Cholesterol than Usual Normal Cholesterol Content Lower Cholesterol Content Clin Cardiol 1999;22(6 Suppl):1121-27
Cholesterol- poor LDL Cholesterol- rich LDL LDL-C and LDL-P in the Multi-Ethnic Study of Atherosclerosis (n=6,697) r = 0.74 r = 0.75 LDL-P (nmol/L) LDL-C (mg/dL) LipoScience unpublished data
Clinical Implications of Differences Between LDL-P vs LDL-C New Studies Contributing to the Evidence
Multi-Ethnic Study of Atherosclerosis (MESA) • Large NHLBI observational study of the pathogenesis and progression of subclinical atherosclerosis (6,814 multiethnic participants). • Baseline NMR measurements of entire cohort. Outcomes • Carotid atherosclerosis (intima-media thickness, IMT) of 5,361 participants not exposed to lipid medications (cross-sectional). • 217 CVD events during 3-yr follow-up. Publication: Mora S et al., Atherosclerosis 2007;192:211-217.
Outcome • 431 CVD events (MI, stroke, CHD death, angina, congestive heart failure) during 15-yr follow-up. Framingham Offspring Study • Long-running NHLBI observational study. • Blood samples (n=3,066) from 1988-91 (exam 4). Publications:Freedman et al., Clin Chem 2004;50:1189-1200; Kathiresan et al., Circulation 2006;113:20-29; Cromwell et al., J Clin Lipidol 2007;1:57-64.
Outcome • 3022 CHD events (MI, coronary death, revascularization) during 5-yr follow-up. Heart Protection Study (preliminary results) • Randomized trial of simvastatin vs placebo in >20,000 high-risk adults. • NMR analyses of stored plasma samples: 10,069 on statin; 10,034 on placebo.
Evaluating LDL-P vs LDL-C LDL-C Has 4 Clinical Applications, Not Just 1 • Risk Predictor: 1 factor among several used in multivariable risk stratification • Treatment Goal: Guides aggressiveness of LDL- lowering therapy and determines treatment success • CVD epidemiology: Represents LDL-related risk in multivariable models to assess which other markers contribute to risk “beyond LDL” • Surrogate endpoint: LDL-C decrease taken as evidence for clinical benefit of an intervention
LDL-P vs LDL-C In Risk Prediction
Atherosclerotic Endpoint LDL-P Associations Stronger? Study CHD Status Women’s Health Study Primary Incident MI, CHD YES Prevention death, CVA - Circulation 2009; 119:931-9 Secondary Non - fatal MI or VA - H IT YES Prevention CHD Death Circulation 2006;113:1556 - 63 Primary MESA YES Carotid IMT Prevention Atherosclerosis 2007;192:211-17. Framingham Heart Study Primary Incident CVD YES Prevention Events J Clin Lipidology 2007;1:583-92. Primary Incident C A D EPIC - Norfolk YES Prevention Events Atherosclerosis 2007;49:547-53. Cardiovascular Health Primary Incident MI or YES Study Prevention Angina 2002; 22:1175 - 1180 ATVB Secondary Angiographic PLAC - I YES Prevention MLD Am J Cardiol 2002;90:89-94. Primary EBCT Coronary Health y Women Study YES Prevention Calcium Score Am J Cardiol 2002;90(suppl):71-77i. Published Studies Comparing LDL-C and LDL-P
What is the Clinical Significance of the Stronger LDL-P Prediction? The Importance of Discordance • Individuals with concordant levels of LDL-C and LDL-P will be equally well served by either LDL measure. • Those individuals with discordant levels will be better served by LDL-P. • The clinical significance of the observed stronger overall prediction by LDL-P thus depends on: • 1) how many individuals have discordant values • 2) how different prediction is in these individuals.
LDL-C and LDL-P Discordance in MESA (discordance defined arbitrarily as ≥ 25 percentile difference) LipoScience unpublished data
LDL-C and LDL-P Discordance in MESA LipoScience unpublished data
LDL-C and LDL-P Discordance in MESA Clinical Relevance Implied by Carotid IMT Relations Data are from linear regression models adjusted for age, gender, race, hypertension, smoking, body mass index, and diabetes. LipoScience unpublished data
LDL-C and LDL-P Discordance in MESA Clinical Relevance Implied by Carotid IMT Relations Data are from linear regression models adjusted for age, gender, race, hypertension, smoking, body mass index, and diabetes. LipoScience unpublished data
LDL-C and LDL-P Discordance in MESA Clinical Relevance Implied by Carotid IMT Relations Data are from linear regression models adjusted for age, gender, race, hypertension, smoking, body mass index, and diabetes. LipoScience unpublished data
LDL-C and LDL-P Discordance in MESA Clinical Relevance Assessed by CVD Prediction Data are from logistic regression models adjusted for age, gender, race, hypertension, and smoking. LipoScience unpublished data
LDL-C and LDL-P Discordance in MESA Clinical Relevance Assessed by CVD Prediction Data are from logistic regression models adjusted for age, gender, race, hypertension, and smoking. LipoScience unpublished data
LDL-C and LDL-P Discordance in MESA Clinical Relevance Assessed by CVD Prediction Data are from logistic regression models adjusted for age, gender, race, hypertension, and smoking. LipoScience unpublished data
Event-Free Survival Years of Follow-up Concordant LDL-C and LDL-P in Framingham ( Cromwell et al. J Clin Lipidol 2007;1:583-92) Low Risk Group Low LDL-C Low LDL-P (n=1,249) High Risk Group High LDL-C High LDL-P (n=1,251)
Event-Free Survival Years of Follow-up Discordant LDL-C and LDL-P in Framingham ( Cromwell et al. J Clin Lipidol 2007;1:583-92) Low Risk Group Low LDL-P High LDL-C (n=284) High Risk Group High LDL-P Low LDL-C (n=282)
New Results: LDL-P Risk Prediction in HPS (statin treatment arm;1312 events/10085 subjects) LDL-C Apo B LDL-P Improvement in Model Fit (2) LipoScience unpublished data From logistic regression models predicting major coronary events (MI, coronary death, revascularization), adjusted for age, gender, smoking, SBP, hypertension, previous disease, vitamin treatment, and eGFR.
LDL-P vs LDL-C As Treatment Target
Discordant (34%) Carotid Atherosclerosis in MESA Subgroup with LDL-C <100 mg/dL (Q1; n=1,374) Implies “residual risk” n=19 n=126 Carotid IMT (µm) n=321 Concordant n=908 Q1 Q2 Q3 Q4 LDL-P Quartile LipoScience unpublished data Adjusted for age, sex, race, smoking, hypertension, BMI, diabetes, TG, and HDL-C.
Carotid Atherosclerosis in MESA Subgroup with LDL-C = 100-118 mg/dL (Q2; n=1,320) n=127 n=340 Concordant Implies LDL-C lowering unneeded n=514 Carotid IMT (µm) n=339 Q1 Q2 Q3 Q4 LDL-P Quartile LipoScience unpublished data Adjusted for age, sex, race, smoking, hypertension, BMI, diabetes, TG, and HDL-C.
CVD Event Rates in MESA Participants with Equivalently Low LDL-C or LDL-P (<25th percentile) n=1643 n=1649 37 Events per 1000* 29 LDL-C <100 mg/dL LDL-P <1100 nmol/L Less “residual risk” LipoScience unpublished data *unadjusted event rate during 3-yr follow-up
37 Events per 1000 29 LDL-C lowering not needed LDL-C <100 LDL-P <1100 n=1643 n=1649 LDL-P lowering needed n=1117 concordant n=526 discordant n=1105 concordant n=544 discordant 43 33 33 Events per 1000 Events per 1000 “residual risk” 20 LDL-P <1100 LDL-P ≥1100 LDL-C <100 LDL-C ≥100 CVD Event Rates in Low LDL Subgroups with Concordant or Discordant LDL-C and LDL-P LipoScience unpublished data
LDL-P vs LDL-C As Surrogate Endpoint
Cholesterol per particle decreases with: • statins • statin + ezetimibe • estrogen replacement therapy • anti-retrovirals (some) • low fat, high carb diet • Cholesterol per particle increases with: • fibrates • niacin • glitazones • omega 3 FAs • CETP inhibitors • exercise • low carb diet LDL-C more than LDL-P LDL-P more than LDL-C LDL-C and LDL-P are Affected Differently by Most (All?) Therapeutic Interventions
Examples of Differential Treatment Effects 1Hsia et al. Atheroscler Thromb Vasc Biol 2008 2Goldberg et al. Diabetes Care 2005;28:1547-54
LDL-P vs LDL-C In CVD Epidemiology Determining Risk Contributors “Beyond LDL”
Contributors to CVD Risk “Beyond LDL” The Issue of Confounding by LDL Definition: A confounder is associated with the risk marker of interest and is causally related to the outcome*. Risk marker of interest CVD Confounder (LDL) *Katz MH, Multivariable Analysis: A Practical Guide for Clinicians
Issue: LDL-C is traditionally the representative of “LDL” in multivariable models. To the extent LDL-C is an inadequate measure of LDL, confounding will still exist, just a different kind! Contributors to CVD Risk “Beyond LDL” Confounding Addressed by Multivariable Analysis Risk Marker (e.g., TG, HDL-C, etc.) CVD ? LDL
Contributors to Carotid IMT in MESA, Independent of LDL-C LDL-C (p<0.0001) + TG (p=0.02) + HOMA-IR (p=0.01) LDL-C plus 2nd variable + Glucose (p=0.01) + LDL Size (p=0.006) + HDL-C (p=0.001) Multivariable Model Fit (R2) LipoScience unpublished data From linear regression models adjusted for age, gender, race, hypertension, smoking, BMI, and diabetes.
Contributors to Carotid IMT in MESA, Independent of LDL-C LDL-C (p<0.0001) + TG (p=0.02) + HOMA-IR (p=0.01) LDL-C plus 2nd variable + Glucose (p=0.01) + LDL Size (p=0.006) + HDL-C (p=0.001) LDL-P Multivariable Model Fit (R2) LipoScience unpublished data From linear regression models adjusted for age, gender, race, hypertension, smoking, BMI, and diabetes.
Contributors to Carotid IMT in MESA, Independent of LDL-C LDL-P (p<0.0001) + TG (p=0.89) + HOMA-IR (p=0.19) LDL-P plus 2nd variable + Glucose (p=0.03) + LDL Size (p=0.43) + HDL-C (p=0.25) Multivariable Model Fit (R2) LipoScience unpublished data From linear regression models adjusted for age, gender, race, hypertension, smoking, BMI, and diabetes.
Contributors to CVD Event Prediction in MESA LDL-C (p<0.0001) + HOMA-IR (p=0.17) LDL-C plus 2nd variable + Glucose (p=0.02) + LDL Size (p=0.02) + HDL-C (p=0.07) LDL-P Multivariable Model Fit (2) LipoScience unpublished data From logistic regression models adjusted for age, gender, race, hypertension, smoking, BMI, and diabetes.
Contributors to CVD Event Prediction in MESA LDL-P (p<0.0001) + HOMA-IR (p=0.26) LDL-P plus 2nd variable + Glucose (p=0.03) + LDL Size (p=0.57) + HDL-C (p=0.53) Multivariable Model Fit (2) LipoScience unpublished data From logistic regression models adjusted for age, gender, race, hypertension, smoking, BMI, and diabetes.
LDL-P vs LDL-C Conclusions • In many patients (≥25%), LDL-C significantly overestimates or underestimates LDL levels. • When LDL-C and LDL-P values are discordant, risk clearly tracks with LDL-P, not LDL-C. • Treating to LDL-P goals will likely reduce the “residual risk” of patients and result in more cost-effective LDL management. • Use of LDL-C in multivariable analyses to assess the “independent” contributions of other risk markers may have produced incorrect conclusions. Reevaluation using LDL-P must be undertaken.
HDL-C and HDL-P in MESA (n=6,697) r = 0.73 HDL-P (µmol/L) HDL-C (mg/dL) LipoScience unpublished data
1400 1200 1000 Relations of HDL-C with CVD are Potentially Confounded by LDL-P (but not LDL-C) Framingham Offspring Study 180 1800 LDL-P 160 1600 LDL Particles (nmol/L) LDL Cholesterol (mg/dL) 140 LDL-C 120 100 20 40 60 80 100 HDL Cholesterol (mg/dL) Amer JCardiol 2002;90:22i-29i