OBJECTIVE To assess the efficacy of unopposed estrogen, and three estrogen/progestin regimens on selected heart disease risk factors among adherent women and to contrast those results with efficacy among all women in the PEPI study. DESIGN A 3-year, multicenter, randomized, double-blinded, placebo-controlled clinical trial. PARTICIPANTS A total of 847 healthy postmenopausal women aged 45 to 64 years of age with no known contraindication to hormone therapy, who attended their 36 month clinical visit. INTERVENTION Participants were randomized in equal numbers to one of the following treatments: (1) placebo; (2) conjugated equine estrogen (CEE) 0.625 mg daily; (3) CEE 0.625 daily plus medroxyprogesterone acetate (MPA) 10 mg, days 1-12; (4) CEE 0.625 daily plus MPA 2.5 mg daily; or (5) CEE 0.625 daily plus micronized progesterone (MP) 200 mg, days 1-12. ANALYSIS Analyses are based on adherent women, where adherence is defined as taking at least 80% of pills at each 6-month visit. RESULTS Adherence rates were high in all groups except women with a uterus assigned to unopposed CEE. The difference in HDL-C levels resulting from the CEE vs. CEE+MP was approximately three times larger than in the intent-to-treat analyses, reaching statistical significance (P < 0.05). In each active treatment, LDL-C decreased 10-15%. Triglycerides increased 15-20% in each opposed CEE arm and over 25% in the CEE only arm; this difference was not statistically significant. Fibrinogen increased by 7% among placebo adherers, but decreased or remained fairly stable among the active arm adherers. Systolic blood pressure increased 3-5% in all treatment arms. Women adherent to the CEE+MPA arms had twice the increase of 2 h glucose levels as women adherent to CEE only, or CEE+MP (8-9% vs. 3-4%). Two-hour insulin levels decreased 3-12% for all arms. The patterns of change for fibrinogen, SBP, 2 h glucose and insulin were similar to those from the intent-to-treat analyses. CONCLUSIONS In analyses limited to adherent women, all active treatments, compared to placebo, continued to have similar and favorable effects on LDL-cholesterol and fibrinogen and no significant effects on blood pressure or insulin levels. Given the overall high adherence rates in PEPI, the results are similar to the intent-to-treat analyses, as expected. Only the trend of HDL-C to have a larger increase in the CEE only arm (in the intent-to-treat analyses) gained statistical significance in analyses restricted to adherers.