We conducted this study at 5 diverse family medicine practices located in central North Carolina. Our primary intent was to determine the comparative effectiveness of the two intervention formats on reducing CHD risk as assessed by the Framingham Risk Score (FRS).¹⁷ Participants were randomized to receive interventions similar in contact time, educational content, and individually tailored counseling, but different in format (Figure 1). Study outcomes were assessed at 4 and 12 months. Details of the study design, study practices, participant enrollment, and intervention components are described elsewhere.¹⁷ The University of North Carolina at Chapel Hill’s Institutional Review Board (IRB) approved and monitored this study, with data collected between January 31, 2011 and November 26, 2012.

Participants were established patients (i.e., had at least one office visit in the last 2 years), age 35–79, with no known cardiovascular disease (CVD), who were at moderate to high risk for CHD (≥ 10% 10-year risk of angina, myocardial infarction, or CHD death) based on their FRS. Participants were identified by chart reviews of patients scheduled for routine office visits, supplemented by referrals from clinicians and self-referrals based on word-of-mouth or in response to waiting room flyers. As an initial eligibility screen, the FRS was calculated using risk factors assessed by chart review (age, blood pressure, total cholesterol, high density lipoprotein cholesterol (HDL-C), diabetes, smoking, aspirin use, and left ventricular hypertrophy).¹⁷ Diabetes was included in the FRS and was not considered a CVD equivalent. Because aspirin was not accounted for by the Framingham risk equation, we modelled its effect on CHD risk using a 23% risk reduction for men and 0% reduction for women.³ Those with a FRS ≥ 10% were further evaluated by their primary care clinicians who 1) determined if the patient should be excluded for a variety of previously described¹⁷ medical conditions and 2) approved participation in the overall and physical activity component of the study.

Patients screened as eligible attended an enrollment visit, during which study staff obtained written informed consent, confirmed inclusion criteria, screened participants for potential bleeding risk associated with aspirin, re-assessed smoking status, assessed blood pressure using a standard protocol, and obtained a blood sample for study lab assessments. Participants’ FRS were re-calculated based on this standardized assessment and, if ≥ 10%, they were contacted for the baseline telephone survey. Those completing this survey were invited to the first intervention visit, where they were randomized, as previously described.¹⁷

Both intervention formats began with a web-based decision aid, followed by the counseling program. As described elsewhere,¹⁷ the intervention was based on previously developed and tested lifestyle and medication interventions revised to be consistent with the latest evidence on CHD risk reduction.

Study measures addressed effectiveness, acceptability, and cost-effectiveness and were assessed by trained research staff at participating practices and by phone. The primary effectiveness measure was within group change in FRS at 4 month follow-up. The FRS was calculated using a well-validated Framingham risk equation²⁵ with input of relevant risk factor data measured in a standardized fashion and baseline age used for follow-up assessments. Pre-specified secondary effectiveness outcomes included between group changes in FRS and change in dietary intake, physical activity, smoking, medication adherence, blood pressure, blood lipids, and health related quality of life. In addition, an analysis of moderators of outcomes was also planned.

Weight, blood pressure, total cholesterol, HDL-C, directly measured low density lipoprotein cholesterol (LDL-C), hemoglobin A1c (A1c), high sensitivity C reactive protein (hsCRP), alanine aminotransferase (ALT), creatinine, and plasma carotenoids²⁶ were assessed at baseline, 4 and 12 months as previously described.¹⁷ At the first counseling visit, numeracy,²⁷ literacy,²⁸ and medication adherence²⁹ were assessed using validated instruments. At follow-up visits, aspirin use was assessed by serum thromboxane level and smoking by NicAlert urine test, as previously described.¹⁷

The following measures were assessed by telephone at baseline and in-person at 4 and 12 month follow-up: medication use, fruit and vegetable intake,³⁰ dietary fat quality,³¹ physical activity,^32,33 and quality of life (SF-12, Quality Metric, Inc., Lincoln, RI). Medication adherence²⁹ and acceptability of the interventions were assessed in-person at 4 and 12-month follow-up.

Process measures were collected at intervention sessions, by counselor or the web-program. Participants were advised to wear an Omron HJ-720ITC pedometer (Omron Healthcare, Bannockburn, IL) during the week before study measurement visits. Steps were assessed by averaging at least 3 days of 500 or more steps/day during the week prior to the visit. Assessment of costs for the cost-effective analysis are described in the Appendix.

Sample size was based on the hypothesis that both interventions would reduce the FRS by at least 1.5 percentage points (absolute risk reduction of 1.5%). Using a one-sided test, a standard deviation of 3.1 units,²⁴ an α = 0.05, and an expected 10% attrition, a sample of 225 participants in each arm would provide > 99% power to detect a within group reduction in FRS of 1.5 percentage points. This sample size would additionally provide 85% power to detect a 0.9 percentage point difference in FRS between the counselor and web arms (two-sided test).

We summarized baseline sample characteristics using descriptive statistics and compared groups using chi-square and t-tests. The primary outcome analysis was conducted using an intention-to-treat approach with a paired t-test (1-sided) for changes in FRS within each intervention arm. Additionally, for the primary outcome, we used multiple approaches for imputing missing data including last observations carried forward and multiple imputation methods.¹⁷

Secondary outcomes were examined using paired t-tests or McNemar’s tests for within group comparisons (2-sided tests). Additional analyses were conducted to compare the mean changes in FRS and other outcomes between arms using a simple t-test and a multivariable analysis of covariance model (ANCOVA) adjusting for the baseline value of the outcome, practice, and additional variables deemed relevant to behavior change a priori (age, race, educational achievement, and BMI) or that differed between intervention groups at baseline (p < 0.10). In addition, we conducted longitudinal analyses with FRS data from all 3 time points using generalized linear mixed models that included time, study groups, and time by study group interaction as fixed and participants as random effects along with site and the full set of covariates as fixed effects. To assess potential moderators of change in FRS, we used linear regression models that included the baseline FRS, the potential moderator of interest, and study arm by potential moderator interaction term.

For cost-effectiveness, we assessed the incremental cost effectiveness ratio (ICER) of each intervention from the payer, participant, and societal perspectives, as described in the Appendix. We calculate the ICER per 1 absolute percentage point reduction in CHD risk and per quality adjusted life year (QALY) gained at 12 months. We calculate QALY gained in one year by converting SF-12 scores into a health related quality of life weight using a well-defined algorithm.³⁴ Because our analysis considers only a one year time horizon, this weight is equivalent to QALYs saved over this time period. We then report incremental cost-effectiveness per QALY gained and compare these ratios to common thresholds of cost-effectiveness. All analyses were conducted using SAS version 9.3 (SAS Institute, Cary, NC) and Stata version 12 (StataCorp, College Station, TX) with p ≤ .05 considered significant.

As depicted in Figure 1, of 2274 patients eligible to be screened for the study, 633 agreed to participate. Of these, 114 were ineligible because their FRS calculated using standardized measures was less than 10%, 111 took part in another intervention for those with known CVD as described elsewhere,¹⁷ 23 were lost to follow-up or declined participation and 385 participants took part in this study.

Table 1 reports baseline characteristics of study participants. The mean age was 62 years, 24% were African American, 32% were employed full time, and 88% had health insurance. Overall, the sample was at high risk for CHD: 86% had current or previous high blood pressure, 85% had current or previous high blood cholesterol, 61% had diabetes, and the mean FRS was 16.9%. Also, two-thirds of participants reported they were comfortable or very comfortable using a computer.

As noted in Figure 1, after viewing the decision aid, 366 (95%) participants elected to work on improving their diet, 256 (66%) chose to work on increasing their physical activity, 71 (18%) decided to work on smoking cessation, and 142 (37%), chose to start or increase blood pressure or cholesterol medication or start aspirin.. Follow-up rates at 4 and 12 months were 91% and 87%, respectively Those who did not return for follow-up at 4 months were more likely to be white, younger, walk fewer minutes each week and at 12 months, consume less fruit and vegetables and be less adherent to medications (p <0.05 for comparisons).

Change in study outcomes from baseline to follow-up, by treatment arm, are shown in Table 2. For the FRS, there was a statistically significant and sustained reduction at both 4 (primary outcome) and 12 month follow-up for participants in both study groups. For the counselor group, the change was −2.3% and −1.9% at 4 and 12 months, respectively. For the web group, it was −1.5% and −1.7%, respectively. When values of no change and multiple imputations methods were used to impute missing FRS scores, results did not change appreciably.

In both groups, all components of the FRS changed in the direction of decreased risk and the majority of changes were statistically significant and maintained from 4 to 12 month follow-up. Likewise, most changes in diet and physical activity were in the direction of decreased risk and sustained over time. Moreover, there were substantial increases in appropriate use of and adherence with medication to reduce CHD risk. Other statistically significant outcomes of note include slight weight loss at 12 months, a reduction in A1c in the counselor group, and a sustained improvement in the physical component measure of quality of life in both groups.

Self-reported results for tobacco cessation and aspirin use at follow-up were confirmed by biomarkers. Of 23 smokers who reported cessation, 18 (78%) were confirmed by urine cotinine testing and of 425 participants who reported aspirin use, 319 (75%) had serum thromboxane levels consistent with aspirin use.

The difference in study outcomes between treatment arms are shown in Table 3. At 4 month follow-up, the adjusted change (SE) in FRS was −2.4% (0.3) for counselor and −1.4% (0.3) for web, difference −1.0% (95% CI −1.8% to −0.1%, p = 0.03). At 12 month follow-up, the adjusted change (SE) in FRS −2.1% (0.4) for counselor and −1.5% (0.4) for web, difference −0.6% (95% CI, −1.7% to 0.5%, p = 0.30). When change in FRS was assessed by longitudinal analysis, there was no significant time by group interaction (p = 0.27) and within and between group comparisons were similar to analyses at each time point.

Figure 2 shows the change in FRS at 4 and 12 month follow-up stratified on selected baseline variables. Assessing change in FRS by subgroups, without regard to treatment arm, the intervention was significantly more effective at 4 and 12 months among younger participants (P = .05 and <.001). In addition, at 4 month follow-up, the intervention was more effective among males (P = .04), those without diabetes (P = .02), and those choosing lifestyle and medication (P = .01). We noted little difference in the effectiveness of the counselor-delivered vs. web-based interventions when change in FRS was assessed by treatment arm and subgroups. At 4 month follow-up, there were a larger improvement in FRS among participants with diabetes in the counselor group (P for interaction = 0.03).

There were no reported adverse side effects related to dietary change or increased physical activity. Deaths due to CHD and newly diagnosed CHD are noted in Figure 1. There were no other deaths during follow-up. In addition, there was no material change in ALT or creatinine from baseline to follow-up.

Both counselor and web formats were well received. At 4 month follow-up, among 177 counselor participants completing the acceptability survey, 137 (77%) strongly agreed and 36 (20%) agreed that they would recommend this program to others. Similarly, among 173 web participants, 128 (74%) strongly agreed and 42 24%) agreed with this statement. At 12 month follow-up, among 170 counselor and 166 web participants completing the survey, 166 (98%) counselor and 161 (97%) web participants would recommend or strongly recommend this program to others.

At 12 months, the costs per participant from the payer perspective were $207 (SE 3.4) and $110 (SE: 3.5) for the Counselor and Web interventions respectively (p<0.001). From the payer perspective, the incremental cost-effectiveness ratio for the less expensive Web intervention, compared to no intervention, was $73 per percentage point reduction in CHD risk and $2,973 per QALY gained, which is considered very cost-effective based on common benchmarks³⁵. Additional results are reported in the Appendix.

A limited sensitivity analysis was conducted (Table 4) to assess change in 10-year risk for CHD as calculated with the Adult Treatment Panel (ATP) III risk calculator³⁶ (which calculates MI and CHD death) and the Framingham risk calculator used for this study²⁵ without including a term for aspirin. Overall, results were similar, with significant reductions in estimated CHD risk in both groups at 4 and 12 month follow-up.