Download presentation
Presentation is loading. Please wait.
Published byLeon Williams Modified over 9 years ago
1
Advances and Controversies in Cardiovascular Risk Prediction Peter Brindle General Practitioner R&D lead Bristol, N.Somerset and S.Glouc PCTs Promises, Pitfalls and Progress
2
Outline Promises –Why do CVD risk estimation? –Background – Framingham Pitfalls –How well to current methods perform? –Two studies Progress –Four new risk scores –Where to next?
3
PROMISES
4
Why do CVD risk estimation? To identify high risk individuals Prioritise treatment - for individuals - for policy Patient education
5
Background Guidelines recommend preventive treatment in high risk patients Population screening Lifelong treatment. Or not.
6
The Framingham Heart Study Data collection started in 1948 Bi-annual follow up First CVD risk equation: Truett et al. 1967 > 20 groups of regression equations between 1967 and 2008 Modified Anderson et al 1991 used in UK
7
Framingham - Anderson Data collected 1968-75 5573 men and women followed up for 12 years Six regression equations published in 1991
8
Risk factors used to calculate the Anderson Framingham risk score -age and sex -diastolic and systolic BP -total:HDL cholesterol ratio -diabetes (Y/N) -cigarette smoking (Y/N) -LVH (Y/N) Absolute CVD Risk over 10 years
9
Different versions and coloured charts New Zealand cardiovascular risk prediction charts
10
Sheffield Tables
11
Joint British Societies 1998
12
Joint British Societies(2) 2004
13
PITFALLS
14
Possible problems with Framingham Secular trends Decline in CHD since the 1970’s Geographical variation Accuracy depends on background risk of target population Uncertain generalisability
15
Getting it wrong People with little to gain may become patients, and the benefit to risk ratio of treatment becomes too small People with much to gain may not be offered preventive treatment Over-prediction means... Under-prediction means…
16
How well does Framingham perform? Single UK population One systematic review
17
6643 men from 24 towns aged 40-59 years no evidence of CVD at entry (1978-80) baseline risk factor assessment 10 year follow-up for fatal and non-fatal CHD Framingham in the British Regional Heart study
18
PREDICTED AND OBSERVED 10-YEAR CHD EVENT RATES IN MEN 10-year CHD risk (%) predicted HIGH RISK
19
PREDICTED AND OBSERVED 10-YEAR CHD EVENT RATES IN MEN 10-year CHD risk (%) predicted observed HIGH RISK
20
12,300 men and women, aged 45-64 and no evidence of cardiovascular disease at entry (1972-76) 10-year follow up for cardiovascular disease mortality Stratified by individual social class and area deprivation Framingham in the Renfrew/Paisley study
21
Social deprivation Social class (Pred/Obs) Deprivation (Pred/Obs) Non-Manual0.69 p= 0.0005 Affluent0.64 p= 0.0017 for trend Manual0.52Intermediate0.56 Deprived0.47 10-year predicted versus observed CVD death rates by area deprivation and social class The Framingham risk score does not reflect the increased risk of people from deprived backgrounds relative to affluent people
22
Predicted over observed ratios ordered by background risk of test population
23
Issues with Framingham BP treatment Family History Deprivation Ethnicity Generalisability Statistical validity Face validity Improvements are needed
24
PROGRESS
25
SCORE Systematic Coronary Risk Evaluation 2003 205,178 men and women from 12 European cohort studies Used by “European guidelines on cardiovascular disease prevention in clinical practice”
27
SCORE – better than Framingham? SCORE BP treatment No Family History No Deprivation No Ethnicity No Generalisability ? Statistical validity Yes Face validity No
28
ASSIGN - ASSessing cardiovascular risk, using SIGN guidelines Scottish Heart and Health Extended Cohort (SHHEC) 6540 men, 6757 women Classic risk factors plus –Deprivation –Family history Shifts treatment towards the socially deprived compared to Framingham
29
ASSIGN – better than Framingham? SCOREASSIGN BP treatment No Family History NoYes Deprivation NoYes Ethnicity No Generalisability ?? Statistical validity Yes Face validity No
30
QRISK1 and QRISK2 Electronic patient record Cohort analysis based on large validated GP database (QResearch) Contains individual patient level data 15 year study period 1993 to 2008 First diagnosis of CVD (including CVD death) QRISK1 –Deprivation –Family History –BMI –On BP treatment NO Ethnicity
31
QRISK1 - better than Framingham? SCOREASSIGNQRISK1 BP treatment No Yes Family History NoYes Deprivation NoYes Ethnicity No Generalisability ??Yes Statistical validity Yes Face validity No Yes
32
QRISK2 Included ONS deaths linkage Included additional variables 2.3 million people (>16 million person yrs) Self-assigned ethnicity Derivation (1.5 million) and test cohorts
33
Self Assigned Ethnicity Initial analysis NHS 16+1 categories Categories used in derivation cohort –White1.5 million –Indian 7328 –Pakistani 4068 –Bangladeshi 2482 –Other Asian 3224 –Black Caribbean 7037 –Black African 6971 –Chinese 1987 –Other7086
34
Model performance QRISK2 vs Modified Framingham QRISK2Framingham Females R squared43.4%38.9% D statistic1.7931.632 Males R squared38.4%34.8% D statistic1.6161.495
35
Model performance QRISK2 vs Modified Framingham QRISK2Framingham Females R squared43.4%38.9% D statistic1.7931.632 ROC statistic0.8170.8 Males R squared38.4%34.8% D statistic1.6161.495 ROC statistic0.7920.779
36
Age-standardised incidence of CVD by deprivation
37
Age-standardised incidence of CVD by Ethnicity per 1000 pyrs %
38
Adjusted Hazard Ratios for CVD
39
Proportion women CVD risk >20% QRISK2 vs Framingham
40
Reclassification 40% of those at high risk on Framingham will be reclassified by QRISK2 Patients high on QRISK2 & low on Framingham have higher risks Patients at low risk on QRISK2 & high risk on Framingham have lower risks Reclassification affects women, ethnic groups, those from deprived areas Choice of score effects health inequalities
41
QRISK2 – better than Framingham? SCOREASSIGNQRISK1QRISK2 BP treatment No Yes Family History NoYes Deprivation NoYes Ethnicity No Yes Reproducibility Yes Generalisability ??Yes Statistical validity Yes Face validity No Yes
42
Where to next? Generalisability? Linkage –Census –Hospital data Improved ethnicity recording
43
Summary Promises –Why do CVD risk estimation? –Background – Framingham Pitfalls –How well to current methods perform? –Two studies Progress –Four new risk scores –Where to next? – linkage and statistics
44
CONCLUSION The idea of risk assessment is well established Existing methods flawed – but better than nothing Electronic patient record + improving data sources = exciting prospects
45
Acknowledgements British Regional Heart study team Renfrew/Paisley study team Shah Ebrahim Tom Fahey Andy Beswick Julia Hippisley-Cox John Robson Carol Coupland Yana Vinogradova Aziz Sheikh Rubin Minhas
46
CONCLUSION The idea of risk assessment is well established Existing methods flawed – but better than nothing Electronic patient record + improving data sources = exciting prospects
47
95% Confidence intervals
48
Changing hazard ratios with age for CVD
49
Patient characteristics
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.