Details of the statistical analysis procedures and sample-size calculation are provided in the Supplementary Appendix . In brief, we estimated the probability of remaining event-free using the Kaplan-Meier method, and its complement, cumulative incidence, was used for plots. In intention-to-treat analyses, Cox proportional-hazards models were used to compare the aspirin group with the placebo group with regard to time-to-event end points and to evaluate effects in subgroups with the use of interaction terms.
Subgroups that were specified in the statistical analysis plan included sex, age (younger than the median age vs. the median age or older), country of residence (Australia vs. the United States), race or ethnic group (white in Australia, white in the United States, black, Hispanic, or other), body-mass index (the weight in kilograms divided by the square of the height in meters; <20.0 [underweight], 20.0 to 24.9 [normal weight], 25.0 to 29.9 [overweight], or ≥30.0 [obese]), previous regular use of aspirin (yes vs. no), frailty category (not frail, prefrail, or frail), personal history of cancer (yes vs. no), smoking (never smoked, former smoker, or current smoker), and the presence of diabetes, hypertension, and dyslipidemia at baseline (yes vs. no, for each condition).11 (link),21 (link) The frailty category was determined on the basis of the adapted Fried frailty criteria,21 (link) which include body weight, strength, exhaustion, walking speed, and physical activity (see theSupplementary Appendix ); the category of prefrail included participants who met one or two criteria, and the category of frail included those who met three or more criteria.
There was no plan for the imputation of missing data. Data censoring occurred at the latest time point that an end point could have been reached and was assumed to be for reasons that would not alter the prospect of the participant having an end point, as compared with participants who continued to be followed. There was no plan for adjustment for multiple comparisons of secondary end points, and only point estimates with confidence intervals that were unadjusted for multiple comparisons are reported, without P values, except for the safety end point of major hemorrhage. For safety analyses, a significance level of 0.05 was applied. An interim analysis was planned for when 1893 primary end-point events had occurred, according to a Haybittle–Peto stopping rule.
Subgroups that were specified in the statistical analysis plan included sex, age (younger than the median age vs. the median age or older), country of residence (Australia vs. the United States), race or ethnic group (white in Australia, white in the United States, black, Hispanic, or other), body-mass index (the weight in kilograms divided by the square of the height in meters; <20.0 [underweight], 20.0 to 24.9 [normal weight], 25.0 to 29.9 [overweight], or ≥30.0 [obese]), previous regular use of aspirin (yes vs. no), frailty category (not frail, prefrail, or frail), personal history of cancer (yes vs. no), smoking (never smoked, former smoker, or current smoker), and the presence of diabetes, hypertension, and dyslipidemia at baseline (yes vs. no, for each condition).11 (link),21 (link) The frailty category was determined on the basis of the adapted Fried frailty criteria,21 (link) which include body weight, strength, exhaustion, walking speed, and physical activity (see the
There was no plan for the imputation of missing data. Data censoring occurred at the latest time point that an end point could have been reached and was assumed to be for reasons that would not alter the prospect of the participant having an end point, as compared with participants who continued to be followed. There was no plan for adjustment for multiple comparisons of secondary end points, and only point estimates with confidence intervals that were unadjusted for multiple comparisons are reported, without P values, except for the safety end point of major hemorrhage. For safety analyses, a significance level of 0.05 was applied. An interim analysis was planned for when 1893 primary end-point events had occurred, according to a Haybittle–Peto stopping rule.
