Table 5.
Reclassification Measures for 10-Year Cardiovascular Disease Risk Prediction Comparing Statin-Naive Risk Predictions Versus Standard Risk Predictions for Men in the Validation Data Set (in Landmark Ages at 40, 50, 60, and 70 Yearsa), Clinical Practice Research Datalink, Hospital Episode Statistics, and the Office for National Statistics, England, United Kingdom, 2004–2017
| Outcome | Categorical NRI b , c | 95% CI | IDI b | 95% CI | Continuous NRI d | 95% CI |
|---|---|---|---|---|---|---|
| Landmark Age 40 Years | ||||||
| Eventb,e | 0.0092 | 0.0002, 0.0183 | 0.0013 | 0.0011, 0.0016 | 0.8939 | 0.8394, 0.9484 |
| Noneventb,e | −0.0015 | −0.0032, 0.0002 | −0.0006 | −0.0007, −0.0006 | −0.8665 | −0.8716, −0.8614 |
| Overall | 0.0077 | −0.0015, 0.0169 | 0.0007 | 0.0005, 0.0009 | 0.0274 | −0.0294, 0.0843 |
| Landmark Age 50 Years | ||||||
| Event | 0.0423 | 0.0312, 0.0535 | 0.0043 | 0.0040, 0.0046 | 0.9168 | 0.8853, 0.9482 |
| Nonevent | −0.0325 | −0.0408, −0.0242 | −0.0027 | −0.0028, −0.0026 | −0.8734 | −0.8786, −0.8683 |
| Overall | 0.0099 | −0.0041, 0.0238 | 0.0016 | 0.0012, 0.0019 | 0.0433 | 0.0095, 0.0772 |
| Landmark Age 60 Years | ||||||
| Event | 0.0277 | 0.0204, 0.0350 | 0.0095 | 0.0092, 0.0098 | 0.9736 | 0.9623, 0.9848 |
| Nonevent | −0.0419 | −0.0523, −0.0315 | −0.0075 | −0.0077, −0.0072 | −0.9638 | −0.9672, −0.9604 |
| Overall | −0.0142 | −0.0269, −0.0014 | 0.0020 | 0.0016, 0.0024 | 0.0098 | −0.0036, 0.0231 |
| Landmark Age 70 Years | ||||||
| Event | 0.0000 | 0.0000, 0.0000 | 0.0157 | 0.0153, 0.0160 | 0.9824 | 0.9690, 0.9958 |
| Nonevent | 0.0000 | 0.0000, 0.0000 | −0.0142 | −0.0147, −0.0137 | −0.9849 | −0.9900, −0.9798 |
| Overall | 0.0000 | 0.0000, 0.0000 | 0.0015 | 0.0008, 0.0021 | −0.0025 | −0.0200, 0.0149 |
Abbreviations: CI, confidence interval; CVD, cardiovascular disease; IDI, integrated discrimination improvement; NRI, net reclassification improvement.
a The results are presented in 10-year increments in landmark age at 40, 50, 60, 70. Above landmark age 69 for men, the predicted 10-year CVD risk for all individuals in the risk set was greater than 10% for both standard risk predictions and statin-naive risk predictions; therefore, there was no movement between the 2 categories and the categorical NRIs were 0 for those older landmark age groups.
b Categorical NRI and IDI were calculated using information from individuals who were not censored at 10 years (either with CVD events within 10 years or event-free at 10-years). Events within 10 years and event-free at 10 years, for the calculation of categorical NRI and IDI, were defined using the counterfactual follow-up time assuming statin had not been initiated.
c Categorical NRI was calculated based on the 4 categories of predicted risk of <10% and ≥10%.
d Continuous NRI (the prospective form NRI) was calculated based on continuous predicted risk and used information from all individuals, including the censored subjects.
e Events and nonevents for continuous NRI (the prospective form of NRI) were the expected results estimated using the Kaplan-Meier approach with counterfactual follow-up time assuming statin had not been initiated, so this prospective form of NRI uses the whole sample and does not require the restriction to the noncensored subjects.