Our primary analyses used the detailed physical-activity assessment
at base line. Person-time for each woman was calculated from
the date of return of the base-line questionnaire to the date of a confirmed
cardiovascular event, death from any cause, or August 27,
2000, whichever came first. Age-adjusted relative risks were computed
as the incidence rate in a specific category of activity divided
by the incidence rate in the lowest quintile, with adjustment
for one-year age categories. We conducted tests of linear trend by
treating the categories as a continuous variable and assigning the
median score for each category.
11
All tests of statistical significance
were two-sided.