Validation study for the QRISK cardiovascular risk assessment tool
An independent validation study of the QRISK score found that it was better at identifying cardiovascular risk in a UK population than the established scoring system in current use and based on the Framingham equations.
The authors note some of the problems with risk prediction models, commenting that many of those published are of poor methodological quality. An essential step often missed is independent external validation, and in this paper they describe an independent validation study on the QRISK cardiovascular risk scoring tool that has been developed using UK data. QRISK has been validated by its original authors, revised following criticism, and re-validated by its original authors.
The study population came from the THIN database, which comprises anonymised patient data from UK GP practices running one of the standard GP software packages in use. Exclusion criteria were pre-existing cardiovascular disease, invalid data, age under 35 years, or over 75 years, were missing Townsend scores, had a diagnosis of pre-existing diabetes, or were prescribed statins at baseline. The authors then calculated the QRISK score for each patient in the cohort, using age-sex reference values for missing risk-factor data.
Predictive performance of QRISK was assessed against observed cardiovascular risk by examining measures of calibration and discrimination. Calibration measures how closely predicted 10 year cardiovascular disease risk agrees with observed 10 year cardiovascular disease risk; discrimination is the ability of the risk prediction model to differentiate between patients who experience a cardiovascular disease event during the study and those who do not. QRISK performance was then compared with scores using the Anderson Framingham equation described in current UK guidelines.
There were data on 1,787,169 patients in the THIN database. After exclusions, 1,072,800 patients were eligible for analysis: median follow-up for this cohort was 4.9 years (range 0 to 12 years), and 36,483 patients were followed for at least 10 years. One of the three main risk factors (total : HDL serum cholesterol ratio, systolic blood pressure, and BMI) was missing for almost two-thirds of patients (63%) and all were missing for 9%: standard values were used for these patients. Overall 10-year observed risk of a cardiovascular event in men aged 35-74 years was 9.87% (95% CI, 9.71% to 10.03%) and in women was 6.55% (95% CI, 6.43% to 6.68%). Observed 10-year risk was significantly higher for patients with all risk factors recorded (19.9% for men, 12.8% for women) compared to those with at least one risk factor missing (5.0% and 3.4% respectively).
In comparison to the Anderson Framingham equation, QRISK gives a more accurate estimate of predicted 10-year cardiovascular risk for both men and women. QRISK tended to under-predict risk (by 13% for men, 10% for women), whereas Anderson Framingham tended to over-predict risk (32% for men and 10% for women).
The authors conclude that the QRISK cardiovascular risk equation offers an improvement over the Anderson Framingham equation in terms of identifying a high risk population for cardiovascular disease in the United Kingdom. While QRISK underestimates 10 year cardiovascular disease risk, the magnitude of under-prediction is smaller than the over-prediction with Anderson Framingham. QRISK also identified a group of high risk patients who will go on to experience more cardiovascular events over the next 10 years than a similar high risk group identified by Framingham. The authors note potential weaknesses of their study, and discuss other cardiovascular risk scoring systems; they also note areas for future research.
The authors note some of the problems with risk prediction models, commenting that many of those published are of poor methodological quality. An essential step often missed is independent external validation, and in this paper they describe an independent validation study on the QRISK cardiovascular risk scoring tool that has been developed using UK data. QRISK has been validated by its original authors, revised following criticism, and re-validated by its original authors.
The study population came from the THIN database, which comprises anonymised patient data from UK GP practices running one of the standard GP software packages in use. Exclusion criteria were pre-existing cardiovascular disease, invalid data, age under 35 years, or over 75 years, were missing Townsend scores, had a diagnosis of pre-existing diabetes, or were prescribed statins at baseline. The authors then calculated the QRISK score for each patient in the cohort, using age-sex reference values for missing risk-factor data.
Predictive performance of QRISK was assessed against observed cardiovascular risk by examining measures of calibration and discrimination. Calibration measures how closely predicted 10 year cardiovascular disease risk agrees with observed 10 year cardiovascular disease risk; discrimination is the ability of the risk prediction model to differentiate between patients who experience a cardiovascular disease event during the study and those who do not. QRISK performance was then compared with scores using the Anderson Framingham equation described in current UK guidelines.
There were data on 1,787,169 patients in the THIN database. After exclusions, 1,072,800 patients were eligible for analysis: median follow-up for this cohort was 4.9 years (range 0 to 12 years), and 36,483 patients were followed for at least 10 years. One of the three main risk factors (total : HDL serum cholesterol ratio, systolic blood pressure, and BMI) was missing for almost two-thirds of patients (63%) and all were missing for 9%: standard values were used for these patients. Overall 10-year observed risk of a cardiovascular event in men aged 35-74 years was 9.87% (95% CI, 9.71% to 10.03%) and in women was 6.55% (95% CI, 6.43% to 6.68%). Observed 10-year risk was significantly higher for patients with all risk factors recorded (19.9% for men, 12.8% for women) compared to those with at least one risk factor missing (5.0% and 3.4% respectively).
In comparison to the Anderson Framingham equation, QRISK gives a more accurate estimate of predicted 10-year cardiovascular risk for both men and women. QRISK tended to under-predict risk (by 13% for men, 10% for women), whereas Anderson Framingham tended to over-predict risk (32% for men and 10% for women).
The authors conclude that the QRISK cardiovascular risk equation offers an improvement over the Anderson Framingham equation in terms of identifying a high risk population for cardiovascular disease in the United Kingdom. While QRISK underestimates 10 year cardiovascular disease risk, the magnitude of under-prediction is smaller than the over-prediction with Anderson Framingham. QRISK also identified a group of high risk patients who will go on to experience more cardiovascular events over the next 10 years than a similar high risk group identified by Framingham. The authors note potential weaknesses of their study, and discuss other cardiovascular risk scoring systems; they also note areas for future research.
BMJ 2009; 339: b2584 (link to abstract, full text freely available at time of posting)
1 comment:
tadalis sx 20 mg is a group of medicines called phosphodiesterase type 5 inhibitors. is used to treat adult men with erectile dysfunction and symptoms of benign prostatic hyperplasia also to manage pulmonary arterial hypertension.
Post a Comment