The ASCOT (Anglo-Scandinavian Cardiac Outcomes Trial) study was used to evaluate the relationship between administering CRP prior to, and on-treatment with statins and their association with cardiovascular (CV) events. http://www.cardiovascularbusiness.com/topics/prevention/aha-crp-screening-may-not-improve-cv-risk-assessments-ascot-shows

1. Evaluation of C-reactive protein, prior
to and on-treatment, as a predictor of
benefit from atorvastatin:
Observations from ASCOT
Peter S. Sever*, Neil R Poulter, Choon L Chang,
Aroon Hingorani, Simon McG Thom, Alun D Hughes,
Paul Welsh, Naveed Sattar,
on behalf of the ASCOT Investigators
*Imperial College London
.

2. ASCOT CRP Analysis
Background
• Lowering C-reactive protein (CRP) by statins has
been shown to independently predict cardiovascular
(CV) outcomes
• The ASCOT database was used to explore the
relationship between circulating CRP prior to, and on-
treatment, with statins and their association with CV
events

3. ASCOT Study Design
19342 hypertensive patients randomized
to antihypertensive treatment
ASCOT-BPLA
stopped after 5.5 yrs
Atenolol ± Amlodipine ±
bendrofluazide perindopril
10305 patients eligible and randomized
in lipid-lowering arm ASCOT-LLA
TC ≤ 6.5 mmol/L (250 mg/dL) stopped after 3.3 yrs
Atorvastatin 10 mg Placebo
Investigator-led, multinational randomised controlled trial
conducted in hypertensive patients, 40 -79 yrs, with no prior history
of CHD, but with 3 additional cardiovascular risk factors (male sex,
> 55 yrs, smoking etc )

4. Nested Case-control Trial Profile
ASCOT – Subjects in the
UK & Ireland On-treatment Analysis Population
14 cases & 37 controls (ASCOT-LLA)
(n = 9098)
were excluded because
•LDL-C: 156 cases & 498 controls
event occurred before
•CRP: 166 cases & 522 controls
blood samples were
•LDL-C & CRP: 155 cases & 488
obtained.
White European controls
(n = 8217) 66 cases & 252 controls
did not have both
baseline and on-
treatment CRP & LDL-C
6549 subjects met entry or with missing values in 235 cases & 777 controls
criteria as potential covariates participating in ASCOT-LLA
cases/controls
485 cases matched with 1367 Baseline CRP Analysis
490 cases & 5750 controls
controls (ASCOT-BPLA case- Population (ASCOT-BPLA)
with valid lab data
control population) 452 cases & 1269 controls
No controls available 33 cases & 98 controls with
for 5 cases missing values in covariates

5. Study Subjects & Blood Samples
• 485 cases (355 CHD & 130 stroke) matched with
1367 controls
• Up to 3 controls from the same risk-set were matched
to each case by age±1 year, sex and study entry
time±90 days.
• Fasting serum HDL-C, triglycerides and total
cholesterol
• Routinely measured at study visits
• CRP at baseline and 6 months
• Measured at the same time using stored serum
samples
• By a high sensitivity method (Dade-Behring, the lower
limit of sensitivity was 0.1 mg/L)

6. Statistical Methods
• t-test or Χ2 test for baseline characteristics comparison
between cases and controls
• Odd ratio obtained from a conditional logistic regression
model for the association between CRP and the risk of
CV events (CHD or stroke)
• Loge CRP (baseline) as a continuous variable
• Categorizing CRP into tertiles with the lowest as reference
• Conditional logistic regression model
• Model 1: unadjusted
• Model 2 (Framingham CV risk factors + randomised treatment )
• Model 3 (extended CV risk factors): as in Model 2 plus
• Body mass index (BMI) ,Loge transformed fasting glucose, Family history of coronary
heart disease (FHCHD) ,Creatinine and Educational attainment

7. Baseline Characteristics
Cases Controls
Variable P value
(n = 485) (n = 1367)
Male (%) 409 (84.3) 1159 (84.8) 0.81
Age, years 64.80 (7.83) 64.65 (7.66) 0.71
Current Smokers (%) 127 (26.2) 297 (21.7) 0.04
Diabetes (%) 150 (30.9) 356 (26.0) 0.038
SBP, mmHg 164.61 (17.81) 161.70 (17.48) 0.002
DBP, mmHg 92.95 (10.37) 92.13 (9.83) 0.12
Total cholesterol, mmol/L 5.99 (1.07) 5.92 (1.07) 0.21
LDL- C, mmol/L 3.92 (0.97) 3.79 (0.96) 0.02
HDL-C, mmol/L 1.26 (0.33) 1.30 (0.34) 0.02
Loge triglyceride, mmol/L 0.52 (0.46) 0.50 (0.49) 0.44
Loge CRP, mg/L 1.13 (0.98) 0.95 (0.98) 0.0006
Loge glucose, mmol/L 1.82 (0.31) 1.78 (0.28) 0.03
Creatinine, mmol/L 102.34 (19.31) 99.23 (16.57) 0.0008
Amlodipine (%) 227 (46.8) 694 (50.8) 0.13
Atorvastatin* (%) 101 (43.0) 407(52.4) 0.01
Values are mean (SD) or n (%)
Missing: LDL-C:121, triglycerides: 79, fasting glucose: 84, CRP: 4, creatinine: 45
*235 cases and 777 controls participated in ASCOT-LLA

8. Baseline CRP and Risk of CV Events (CHD or Stroke)
Cases Control P value Trend
Per 1 SD increase in loge CRP 452 1269 0.0004
Tertile 1 CRP: <1.74 mg/L 131 448 0.005 Model 1
Tertile 2 CRP: 1.74-4.09 mg/L 153 417 0.08 Unadjusted
Tertile 3 CRP: >4.09 mg/L 168 404 0.005
Per 1 SD increase in loge CRP 452 1269 0.009
Tertile 1 CRP: <1.74 mg/L 131 448 0.057 Model 2*
Tertile 2 CRP: 1.74-4.09 mg/L 153 417 0.29
Tertile 3 CRP: >4.09 mg/L 168 404 0.06
Per 1 SD increase in loge CRP 452 1269 0.006
Tertile 1 CRP: <1.74 mg/L 131 448 0.05 Model 3
Tertile 2 CRP: 1.74-4.09 mg/L 153 417 0.14 Adjusted as for in Model
2 plus BMI, loge-glucose,
Tertile 3 CRP: >4.09 mg/L 168 404 0.05 family history of CHD,
creatinine and
educational attainment
0.5 1.0 1.5 2.0
Odd ratio (95% CI)
*Model 2: Adjusted for current smoking status, diabetes mellitus, randomized BP treatment (atenolol/amlodipine),
randomized atorvastatin/placebo/not in LLA, left ventricular hypertrophy, baseline SBP, total cholesterol and HDL-C

9. Predictive Ability of Baseline CRP in CVD
Prediction Beyond Classical Risk Factors - 1
Modified Framingham Model Full Model
(Model 2) (Model 3)
Performance Measure Without CRP With CRP Without CRP With CRP
Discrimination
0.592 0.600 0.620 0.627
Area under ROC (95% CI)
(0.562, 0.621) (0.571, 0.630) (0.591, 0.650) (0.598, 0.656)
P-value 0.20 0.18
Conditional Regression
AIC 1127.41 1121.25 1117.42 1111.76
BIC 1176.00 1175.24 1203.81 1203.55
LR chi-square (df) 35.55 (9) 43.71 (10) 59.53 (16) 67.20 (17)
Calibration
Hosmer-Lemeshow, χ² ,deciles 5.94 10.88 2.83 2.97
P-value 0.65 0.21 0.94 0.94
AIC, Akaike’s information criterion; BIC, Bayes information criterion; ROC, receiver operator curve
Framingham model: Adjusted for age, sex, smoking status, diabetic mellitus, baseline SBP, total cholesterol, HDL-C, randomized BP
treatment, randomized statin/placebo/not in LLA and LVH
Full model: as for reduced model plus loge glucose, family history of CHD, educational attainment, creatinine and BMI

10. Predictive Ability of Baseline CRP in CVD
Prediction Beyond Classical Risk Factors - 2
• Net Reclassification Improvement (NRI) for model including CRP
over model without CRP
• In Model 2, NRI=2.1% (p=0.32)
• In Model 3, NRI=3.0% (p=0.17)
• Estimation of Integrated Discrimination Improvement (IDI)
• When CRP was included
• In Model 2, IDI increased 0.38% (P=0.015)
• In Model 3, the increase in IDI was 0.49% (P=0.013)
• It suggests that the addition of CRP to the model with
established risk factors very modestly improved the
discriminatory property of the model for the prediction of risk

11. Effect of Atorvastatin by Tertile of Baseline CRP
Odd Ratios (95% CI) by Tertile of Baseline CRP
Low Middle High Interaction*
CVD 0.60 0.77 0.94
P=0.54
(0.33, 1.10) (0.41, 1.46) (0.51, 1.78)
Adjusted for current smoking status, diabetes mellitus, randomized BP
treatment (atenolol/amlodipine), randomized atorvastatin/placebo/not in
ASCOT-LLA. Left ventricular hypertrophy, baseline SBP, total cholesterol,
HDL-C, BMI, loge-glucose, family history of CHD, creatinine and educational
attainment and baseline LDL-C.
* Interaction between statin treatment and tertile baseline CRP

12. Effects of Atorvastatin on LDL-C and CRP
5 6 Baseline
On-treatment (6 month)
4
40.3% reduction
Median LDL-C (mmol/L)
4
Median CRP (mg/L)
3
27.4% reduction
2
2
1
0 0
Placebo Atorvastatin Placebo Atorvastatin
• On-treatment change in CRP was independent of baseline classical risk factors, baseline CRP and
change in LDL-C (P=0.02)
• In ASCOT-LLA, in those assigned atorvastatin, age-adjusted Spearman’s correlation between
percentage change in CRP and percentage change in LDL-C was modest (r=0.19 P=0.0006)

13. Baseline Profiles by 6 Month Median LDL-C and CRP
On-treatment On-treatment On-treatment On-treatment
Variable LDL-C ≥2.1mmol/L LDL-C <2.1mmol/L CRP ≥1.83mg/L CRP <1.83mg/L
(n = 190) (n = 178) (n = 195) (n = 193)
Male (%) 169 (88.9) 162 (91.0) 171 (87.7) 175 (90.7)
Age (Years) 64.15 (7.60) 65.70 (6.88)* 64.69 (7.67) 65.44 (6.86)
Current Smokers (%) 48 (25.3) 31 (17.4) 53 (27.2) 31 (16.1)**
Alcohol (%):
37 (19.5) 38 (21.3) 45 (23.1) 35 (18.1)
Never
<=14/21 units/week 114 (60.0) 109 (61.2) 109 (55.9) 125 (64.8)
>14/21 units/week 39 (20.5) 31 (17.4) 41 (21.0) 33 (17.1)
SBP (mmHg) 162.45 (17.55) 162.47 (18.48) 161.59 (16.95) 163.63 (19.04)
BMI (Kg/m2) 29.16 (4.17) 28.56 (4.53) 91.70 (9.17) 92.47 (9.52)***
Total Cholesterol (mmol/L) 5.87 (0.62) 5.15 (0.70)*** 29.76 (4.43) 27.95 (4.07)
LDL-C (mmol/L) 3.87 (0.59) 3.14 (0.61)*** 5.53 (0.71) 5.51 (0.75)
HDL-C (mmol/L) 1.25 (0.29) 1.27 (0.37) 3.53 (0.66) 3.50 (0.71)*
Loge Triglyceride (mmol/L) 0.44 (0.42) 0.37 (0.50) 1.23 (0.30) 1.30 (0.36)*
Loge CRP (mg/L) 0.99 (1.00) 0.94 (1.02) 0.46 (0.44) 0.35 (0.47)***
Loge glucose (mmol/L) 1.74 (0.26) 1.78 (0.27) 1.77 (0.31) 1.78 (0.25)
Creatinine (mmol/L) 100.16 (15.98) 102.03 (18.41) 1.41 (0.91) 0.49 (0.89)
Diabetes (%) 41 (21.6) 50 (28.1) 101.29 (19.43) 99.99 (14.62)*
Family history CHD (%) 30 (15.8) 19 (10.7) 61 (31.3) 41 (21.2)**
Amlodipine (%) 83 (43.7) 90 (50.6) 87 (44.6) 61 (31.6)
Values are mean (SD) or n (%)
* p<0.05; ** p<0.01; *** p<0.0001
Missing: fasting glucose: 25; creatinine: 13; triglyceride: 22; CRP: 1
Excluded events occurred before blood sampling: LDL-c≥2.1 mmol/L: 29; LDL-c <2.1 mmol/l: 38

14. On-treatment LDL-C (6 Month in Trial) and
Risk of CV Events (CHD or Stroke)
Cases Control P value
Placebo 89 232
Atorvastatin
LDL-C ≥2.1 mmol/L 44 126 0.87 Model 1
Unadjusted
LDL-C <2.1 mmol/L 23 140 0.001
LDL-C <2.1 vs. ≥2.1 mmol/L 0.002
Placebo 89 232
Atorvastatin
LDL-C ≥2.1 mmol/L 44 126 0.71 Model 2
LDL-C <2.1 mmol/L 23 140 0.003 Adjusted for
baseline LDL-C and
LDL-C <2.1 vs. ≥2.1 mmol/L 0.003 loge baseline CRP
Placebo 89 232
Atorvastatin
LDL-C ≥2.1 mmol/L 44 126 0.68 Model 3*
LDL-C <2.1 mmol/L 23 140 0.003
LDL-C <2.1 vs. ≥2.1 mmol/L 0.004
0.0 0.5 1.0 1.5 2.0
Odd ratio (95% CI)
*Model 3: Current smoking status, diabetes mellitus, randomized BP treatment (atenolol/amlodipine), left ventricular
hypertrophy, baseline SBP, HDL-C, BMI, loge-glucose, family history of CHD, creatinine, educational attainment, baseline
LDL-C or total cholesterol and loge baseline CRP

15. On-treatment CRP (6 Month in Trial) and
Risk of CV Events (CHD or Stroke)
Cases Control P value
Placebo 93 245
Atorvastatin
CRP ≥1.83 mg/L 41 137 0.47 Model 1
Unadjusted
CRP <1.83 mg/L 32 140 0.03
CRP <1.83 vs. ≥1.83 mg/L 0.17
Placebo 93 245
Atorvastatin
CRP ≥1.83 mg/L 41 137 0.39 Model 2
CRP <1.83 mg/L 32 140 0.14 Adjusted for
baseline LDL-C and
CRP <1.83 vs. ≥1.83 mg/L 0.56 loge baseline CRP
Placebo 93 245
Atorvastatin
CRP ≥1.83 mg/L 41 137 0.38 Model 3*
CRP <1.83 mg/L 32 140 0.15
CRP <1.83 vs. ≥1.83 mg/L 0.60
0.0 0.5 1.0 1.5 2.0
Odd ratio (95% CI)
*Model 3: Current smoking status, diabetes mellitus, randomized BP treatment (atenolol/amlodipine), left ventricular
hypertrophy, baseline SBP, HDL-C, BMI, loge-glucose, family history of CHD, creatinine, educational attainment, baseline
LDL-C or total cholesterol and loge baseline CRP

16. Risk of CV Events (CHD or Stroke) by On-
treatment (6 Month in Trial) LDL-C and CRP*
Cases Control P value
Placebo 88 230
LDL-C ≥2.1 & CRP ≥1.83 27 65 0.40
LDL-C ≥2.1 & CRP <1.83 17 55 0.98 ASCOT Medians
LDL-C <2.1 & CRP ≥1.83 11 62 0.02
LDL-C <2.1 & CRP <1.83 12 76 0.05
Placebo 88 230
LDL-C ≥1.8 & CRP ≥2 30 81 0.85
LDL-C ≥1.8 & CRP <2 27 94 0.80 JUPITER Cut-offs
LDL-C <1.8 & CRP ≥2 3 36 0.01
LDL-C <1.8 & CRP <2 7 47 0.06
Placebo 88 230
LDL-C ≥1.8 & CRP ≥1 46 126 0.75
LDL-C ≥1.8 & CRP <1 11 49 0.48 JUPITER Cut-offs
LDL-C <1.8 & CRP ≥1 5 53 0.003
LDL-C <1.8 & CRP <1 5 30 0.23
*LDL-C in mmol/L and CRP in mg/L 0.0 0.5 1.0 1.5 2.0 2.5
Odd ratio (95% CI)
Adjusted for current smoking status, diabetes mellitus, randomised BP treatment (atenolol/amlodipine), , left ventricular
hypertrophy, baseline SBP, total cholesterol, HDL-cholesterol, BMI, loge-glucose, family history of CHD, creatinine,
educational attainment, and baseline LDL or total cholesterol and loge baseline CRP

17. Summary
• Baseline LDL-C and loge transformed CRP were correlated and
predicted CV events respectively
• Inclusion of baseline CRP into a modified Framingham risk model in
the whole cohort very modestly improved risk prediction
• Baseline CRP was not an indicator of the magnitude of the effect of
atorvastatin on CV outcome of those assigned atorvastatin
• At 6 months, atorvastatin reduced median LDL-C by 40.3% and
median CRP by 27.4%
• In those randomized to atorvastatin
• Lower on-treatment LDL-C at 6 months was associated with a highly
significant reduction in subsequent CV events
• By contrast, lower CRP at 6 months was not associated with CV events
• Consequently, addition of on-treatment CRP to on-treatment LDL-C did not
improve prediction of statin efficacy

18. Conclusion
• In ASCOT-LLA, neither baseline nor on-treatment
CRP provide useful information about the efficacy of
statin treatment to reduce CV events beyond LDL-C
reduction
• The results do not support current proposals to
measure CRP in the clinical setting either to assign
statins to individuals on the basis of an elevated CRP
alone, or to monitor CRP levels as an indicator of the
efficacy of statin treatment

19. Back up

20. Baseline CRP and Risk of CHD Only
Cases Control P value Trend
Per 1 SD increase in loge CRP 331 939 0.0003
Tertile 1 CRP: <1.74 mg/L 87 330 0.01 Model 1
Tertile 2 CRP: 1.74-4.09 mg/L 119 310 0.024 Unadjusted
Tertile 3 CRP: >4.09 mg/L 125 299 0.01
Per 1 SD increase in loge CRP 331 939 0.009
Tertile 1 CRP: <1.74 mg/L 87 330 0.16 Model 2*
Tertile 2 CRP: 1.74-4.09 mg/L 119 310 0.13
Tertile 3 CRP: >4.09 mg/L 125 299 0.15
Per 1 SD increase in loge CRP 331 939 0.005
Tertile 1 CRP: <1.74 mg/L 87 330 0.10 Model 3
Tertile 2 CRP: 1.74-4.09 mg/L 119 310 0.09 Adjusted as for in
Model 2 plus BMI,
Tertile 3 CRP: >4.09 mg/L 125 299 0.09 loge-glucose, family
history of CHD,
creatinine and
educational
0.5 1.0 1.5 2.0 2.5 attainment
Odd ratio (95% CI)
*Model 2: Adjusted for current smoking status, diabetes mellitus, randomized BP treatment (atenolol/amlodipine),
randomized atorvastatin/placebo/not in LLA, left ventricular hypertrophy, baseline SBP, total cholesterol and HDL-C

21. Baseline CRP and Risk of Stroke Only
Cases Control P value Trend
Per 1 SD increase in loge CRP 121 330 0.44
Tertile 1 CRP: <1.74 mg/L 44 118 0.22 Model 1
Unadjusted
Tertile 2 CRP: 1.74-4.09 mg/L 34 107 0.70
Tertile 3 CRP: >4.09 mg/L 43 105 0.20
Per 1 SD increase in loge CRP 121 330 0.50
Tertile 1 CRP: <1.74 mg/L 44 118 0.24 Model 2*
Tertile 2 CRP: 1.74-4.09 mg/L 34 107 0.63
Tertile 3 CRP: >4.09 mg/L 43 105 0.25
Per 1 SD increase in loge CRP 121 330 0.64
Tertile 1 CRP: <1.74 mg/L 44 118 0.32 Model 3
Adjusted as for in
Tertile 2 CRP: 1.74-4.09 mg/L 34 107 0.91
Model 2 plus BMI,
Tertile 3 CRP: >4.09 mg/L 43 105 0.32 loge-glucose,
family history of
CHD, creatinine
and educational
0.5 1.0 1.5 2.0 2.5 attainment
Odd ratio (95% CI)
*Model 2: Adjusted for current smoking status, diabetes mellitus, randomized BP treatment (atenolol/amlodipine),
randomized atorvastatin/placebo/not in LLA, left ventricular hypertrophy, baseline SBP, total cholesterol and HDL-C

22. Predictive Ability of Baseline CRP in CVD
Prediction Beyond Classical Risk Factors - 1
Modified Framingham Model Full Model
(Model 2) (Model 3)
Performance Measure Without CRP With CRP Without CRP With CRP
Discrimination
0.592 0.600 0.620 0.627
Area under ROC (95% CI)
(0.562, 0.621) (0.571, 0.630) (0.591, 0.650) (0.598, 0.656)
P-value 0.20 0.18
Conditional Regression
AIC 1127.41 1121.25 1117.42 1111.76
BIC 1176.00 1175.24 1203.81 1203.55
LR chi-square (df) 35.55 (9) 43.71 (10) 59.53 (16) 67.20 (17)
Calibration
Hosmer-Lemeshow, χ² ,deciles 5.94 10.88 2.83 2.97
P-value 0.65 0.21 0.94 0.94
AIC, Akaike’s information criterion; BIC, Bayes information criterion; ROC, receiver operator curve
Framingham model: Adjusted for age, sex, smoking status, diabetic mellitus, baseline SBP, total cholesterol, HDL-C, randomized BP
treatment, randomized statin/placebo/not in LLA and LVH
Full model: as for reduced model plus loge glucose, family history of CHD, educational attainment, creatinine and BMI

23. Study Population
• 485 cases (355 CHD & 130 stroke) matched with
1367 controls
• Controls were selected from the UK & Ireland ASCOT
study population who were alive at the time the case
was diagnosed and free from CV disease in the study
period.
• Up to 3 controls from the same risk-set were matched
to each case by age±1 year, sex and study entry
time±90 days.

24. Laboratory Methods
• Fasting serum HDL-C, triglycerides and total cholesterol
• Routinely measured at study visits
• CRP at baseline and 6 months
• Measured at the same time using stored serum samples
• By a high sensitivity method (Dade-Behring, the lower
limit of sensitivity was 0.1 mg/L)
• At Glasgow Royal Infirmary (CPA accredited)
• On an Abbott Architect
• By technicians blinded to the case-control status of the
samples

25. Statistical Methods -1
• t-test or Χ2 test for baseline characteristics comparison
between cases and controls
• Age-adjusted partial correlation coefficients for the
correlation between loge CRP (baseline) and baseline
clinical characteristics
• Odd ratio obtained from a conditional logistic regression
model for the association between CRP and the risk of
CV events (CHD or stroke)
• Loge CRP (baseline) as a continuous variable
• Categorizing CRP into tertiles with the lowest as reference

26. Statistical Methods - 2
Conditional logistic regression model
• Model 1: unadjusted
• Model 2 (modified Framingham CV risk factors): adjusted for
• Current smoking status
• Diabetes mellitus
• Left ventricular hypertrophy
• Baseline systolic blood pressure (SBP)
• Total cholesterol and HDL
• Randomized atorvastatin/placebo/not in ASCOT-LLA
• Randomized atenolol/amlodipine
• Model 3 (extended CV risk factors): adjusted for Model 2 plus
• Body mass index (BMI)
• Loge transformed fasting glucose
• Family history of coronary heart disease (FHCHD)
• Creatinine
• Educational attainment

27. Baseline Characteristics – 2
Cases Controls
Variable P value
(n=485) (n=1367)
Alcohol (%)
Never 124 (25.6) 298 (21.8)
≤14/21 units/week 270 (55.7) 803 (58.7)
>14/21 units/week 91 (18.8) 266 (19.5) 0.23
Family history CHD (%) 85 (17.5) 233 (17.0) 0.81
BMI, kg/m2 28.71 (4.05) 29.00 (4.50) 0.21
Heart Rate, bpm 69.65 (12.91) 70.35 (11.85) 0.28
Completed Education (%)
Age: ≤12 182 (37.5) 452 (33.1)
Age: ≤15 233 (48.0) 632 (46.2)
Age: ≤18 46 (9.5) 154 (11.3)
Age: 19+ 24 (4.9) 129 (9.4) 0.007
Values are mean (SD) or n (%)

28. Predictive Ability of Baseline CRP in CVD
Prediction Beyond Classical Risk Factors – 1
Modified Framingham Model Full Model
(Model 2) (Model 3)
Performance Measure Without CRP With CRP Without CRP With CRP
Discrimination
0.592 0.600 0.620 0.627
Area under ROC (95% CI)
(0.562, 0.621) (0.571, 0.630) (0.591, 0.650) (0.598, 0.656)
P-value 0.20 0.18
Conditional Regression
AIC 1127.41 1121.25 1117.42 1111.76
BIC 1176.00 1175.24 1203.81 1203.55
LR chi-square (df) 35.55 (9) 43.71 (10) 59.53 (16) 67.20 (17)
Calibration
Hosmer-Lemeshow, χ² ,deciles 5.94 10.88 2.83 2.97
P-value 0.65 0.21 0.94 0.94
AIC, Akaike’s information criterion; BIC, Bayes information criterion; ROC, receiver operator curve
Framingham model: Adjusted for age, sex, smoking status, diabetic mellitus, baseline SBP, total cholesterol, HDL-C, randomized BP
treatment, randomized statin/placebo/not in LLA and LVH
Full model: as for reduced model plus loge glucose, family history of CHD, educational attainment, creatinine and BMI

29. Predictive Ability of Baseline CRP in CVD
Prediction Beyond Classical Risk Factors – 2
With CRP Reclassification
Without CRP
Number of Subjects Improvement %
Framingham Covariate Model (Model 2)
10-<20% 20-<30% 30-<40% 40-<50% 50%
Cases
10-<20% 38 12 0 0 0
20-<30% 13 203 29 0 0 3.32
30-<40% 0 24 109 11 0
40-<50% 0 0 2 9 2
Controls
10-<20% 213 50 0 0 0
20-<30% 67 555 69 0 0 1.18
30-<40% 0 47 229 12 0
40-<50% 0 0 2 23 0
50% 0 0 0 0 1
Net Reclassification Improvement, % 2.14
P=0.32
Framingham covariate model: Adjusted for age, sex, smoking status, diabetic mellitus, baseline SBP, total
cholesterol, HDL-C, randomized BP treatment, randomized statin/placebo/not in LLA and LVH

30. Predictive Ability of Baseline CRP in CVD
Prediction Beyond Classical Risk Factors – 3
With CRP Reclassification
Without CRP
Number of Subjects Improvement %
Full Model (Model 3)
10-<20% 20-<30% 30-<40% 40-<50% 50%
Cases
10-<20% 52 6 0 0 0
20-<30% 14 166 28 0 0 1.77
30-<40% 0 18 109 12 1
40-<50% 0 0 10 25 4
50-60% 0 0 0 1 5
Controls
10-<20% 282 46 0 0 0
20-<30% 62 473 41 0 0 -1.21
30-<40% 0 54 217 18 0
40-<50% 0 0 10 29 7
50%+ 0 0 0 1 5
Net Reclassification Improvement, % 2.98
P=0.17
Full model: as for Framingham covariate model plus loge glucose, family history CHD, educational attainment,
creatinine and BMI

31. Predictive Ability of Baseline CRP in CVD
Prediction Beyond Classical Risk Factors – 4
• Estimation of Integrated Discrimination Improvement (IDI)
• When CRP was included
• In Model 2, IDI increased 0.38% (P=0.015)
• In Model 3, the increase in IDI was 0.49% (P=0.013)
• It suggests that the addition of CRP to the model with
established risk factors very modestly improved the
discriminatory property of the model for the prediction of risk

32. Baseline Profiles by 6 Month Median LDL-C
On-treatment On-treatment
Variable LDL-C ≥2.1mmol/L LDL-C <2.1mmol/L P value
(n = 190) (n = 178)
Male (%) 169 (88.9) 162 (91.0) 0.5
Age (Years) 64.15 (7.60) 65.70 (6.88) 0.04
Current Smokers (%) 48 (25.3) 31 (17.4) 0.07
Alcohol (%): Never 37 (19.5) 38 (21.3)
<=14/21 units/week 114 (60.0) 109 (61.2)
>14/21 units/week 39 (20.5) 31 (17.4) 0.72
SBP (mmHg) 162.45 (17.55) 162.47 (18.48) 0.99
BMI (Kg/m2) 29.16 (4.17) 28.56 (4.53) 0.19
Total Cholesterol (mmol/L) 5.87 (0.62) 5.15 (0.70) <0.0001
LDL-C (mmol/L) 3.87 (0.59) 3.14 (0.61) <0.0001
HDL-C (mmol/L) 1.25 (0.29) 1.27 (0.37) 0.41
Loge Triglyceride (mmol/L) 0.44 (0.42) 0.37 (0.50) 0.14
Loge CRP (mg/L) 0.99 (1.00) 0.94 (1.02) 0.61
Loge glucose (mmol/L) 1.74 (0.26) 1.78 (0.27) 0.21
Creatinine (mmol/L) 100.16 (15.98) 102.03 (18.41) 0.31
Diabetes (%) 41 (21.6) 50 (28.1) 0.15
Family history CHD (%) 30 (15.8) 19 (10.7) 0.15
Amlodipine (%) 83 (43.7) 90 (50.6) 0.19
Values are mean (SD) or n (%)
Missing: fasting glucose: 25; creatinine: 13; triglyceride: 22; CRP: 1
Excluded events occurred before blood sampling: LDL-c≥2.1 mmol/L: 29; LDL-c
<2.1 mmol/l: 38

33. Baseline Profiles by 6 Month Median CRP
On-treatment On-treatment
Variable CRP ≥1.83mg/L CRP <1.83mg/L P value
(n = 195) (n = 193)
Male (%) 171 (87.7) 175 (90.7) 0.34
Age (Years) 64.69 (7.67) 65.44 (6.86) 0.31
Current Smokers (%) 53 (27.2) 31 (16.1) 0.008
Alcohol (%): Never 45 (23.1) 35 (18.1)
<=14/21 units/week 109 (55.9) 125 (64.8)
>14/21 units/week 41 (21.0) 33 (17.1) 0.20
SBP (mmHg) 161.59 (16.95) 163.63 (19.04) 0.27
DBP (mmHg) 91.70 (9.17) 92.47 (9.52) 0.42
BMI (Kg/m2) 29.76 (4.43) 27.95 (4.07) <0.0001
Total Cholesterol (mmol/L) 5.53 (0.71) 5.51 (0.75) 0.83
LDL-C (mmol/L) 3.53 (0.66) 3.50 (0.71) 0.69
HDL-C (mmol/L) 1.23 (0.30) 1.30 (0.36) 0.04
Loge Triglyceride (mmol/L) 0.46 (0.44) 0.35 (0.47) 0.03
Loge glucose (mmol/L) 1.77 (0.31) 1.78 (0.25) 0.73
Loge CRP (mg/L) 1.41 (0.91) 0.49 (0.89) <0.0001
Creatinine (mmol/L) 101.29 (19.43) 99.99 (14.62) 0.47
Diabetes (%) 61 (31.3) 41 (21.2) 0.02
Family history CHD (%) 87 (44.6) 61 (31.6) 0.008
Amlodipine (%) 31 (15.9) 23 (11.9) 0.26
Values are mean (SD) or n (%)
Missing: fasting glucose: 24; creatinine: 12; LDL-C: 27; triglyceride: 22

34. Statistical Methods - 3
• To examine the predictive power of loge CRP on CV events
• Unconditional logistic regression models with adjustment for age and sex on
Framingham-based (Model 2) with and without loge CRP
• The fully adjusted model (Model 3) with and without loge CRP
• To assess global fit of Models 2 and 3, and the improvement after the
addition of loge CRP
• Akaike’s information criterion (AIC)
• Bayes information criterion (BIC)
• Likelihood-ratio tests
• To measure the discrimination of the model
• Area under the Receiver Operator Characteristic (ROC) curve
• Model 2 ± loge CRP
• Model 3 ± loge CRP
• Net reclassification Improvement (NRI)
• Integrated Discrimination Improvement (IDI)
• To assess model calibration
• Hosmer-Lemeshow goodness-of-fit test

35. Statistical Methods - 4
• To determine the impact of statin on CV events by on-treatment
LDL-C and CRP
• The cases and controls on atorvastatin were divided into 2 groups based on
• LDL-C below or above the ASCOT median LDL-C at 6-month (1.83 mmol/L)
OR
• CRP below or above the ASCOT median CRP at 6-month (2.1 mg/L)
• Model 3 (multiple conditional logistic regression model ) was used to estimate odd ratios for
CV events in these groups compared with the placebo group
• To investigate the effect of on-treatment CRP across the
categories of on-treatment LDL-C and total cholesterol values
• The cases and controls on atorvastatin were divided into 4 groups based on
• LDL-C below or above 1.83 mmol/L and CRP below or above 2.1 mg/L
OR
• Cut-off values used in the JUPITER trial (LDL-C =1.8 mmol/L, CRP= 1 and 2 mg/L)
• Model 3 was used to estimate odd ratios for CV events in these groups compared with the
placebo group