Defining hormone replacement therapy in longitudinal studies: impact on measures of effect

Data from a nested case-control study, designed to examine the effect of hormone replacement therapy (HRT) on colorectal cancer risk, were analyzed to determine the effect of exposure definition on the estimation of risk ratios (RR). A prescription drug plan database was used to ascertain HRT prescriptions dispensed prior to index dates to cases (n = 3059) and age-matched controls (n = 12,116). HRT exposure was defined as 'prescription' and 'tablet' counts, 'conjugated estrogen only' and a method based on proportions of minimum exposure to a number of estrogens (SUM-P3 and SUM-P12). The effect of HRT was described with reference to 'ever', <5 and > or = 5 years of HRT use. Conditional logistic regression was used to estimate ORs and 95% confidence intervals (CI). Adjusted ORs for 'ever use' of HRT ranged from 0.72 (95%CI: 0.60-0.88) to 0.86 (95%CI: 0.76-0.99); for <5 years use, from 0.70 (95%CI: 0.56-0.88) to 0.89 (95%CI: 0.78-1.01) and for >5 year of HRT use, from 0.74 (95%CI: 0.59-0.92) to 0.98 (95%CI: 0.42-2.26). Various methods used to define HRT exposure produce a range of estimated ORs that vary in magnitude similar to results reported in the literature from observational studies investigating the association between HRT and colorectal cancer.