(P17-058-25) A Target Trial Emulation of the PREDIMED Trial Protocol in Two US Cohorts

Objectives: The PREDIMED trial concluded that a Mediterranean diet (MedDiet) supplemented with olive oil or nuts reduced type 2 diabetes (T2D) incidence compared to a low-fat diet among high-risk older adults in Spain. To emulate the PREDIMED trial protocol for prevention of T2D in the Nurses Health Study and the Health Professionals' follow-up Study. However, is it unclear if these results are generalizable to US populations with differences in risk factors or cultural food ways. We set out to estimate the 4-year risk of incident type 2 diabetes (T2D) had all participants been told to follow a Mediterranean diet versus a low-fat pattern.

Methods: Using the TTF, we defined an analysis plan for the Nurses’ Health Study and Health Professional’s Follow-up Study cohorts based on the published PREDIMED RCT protocol. This included similar enrollment criteria and the exclusion of prevalent T2D cases. We assessed cohort participants’ exposure to MedDiet or Low-Fat pattern interventions using usual diet from food frequency questionnaires. We estimated the cumulative incidence, risk differences, and risk ratios over 4 year follow-up for risk of incident T2D, comparing scenarios of all participants intervened to follow a MedDiet vs. low-fat pattern.

Results: Our eligible sample included 10,671 participants at baseline in 2010 (77% female; mean age XX years). The US cohorts had lower prevalence of overweight/obesity (84.5% vs 94.6%), smoking (7.9% v 15.6%), and higher dyslipidemia (92.3% v 51.4%) vs. PREDIMED. Sofrito was a MedDiet item not captured in the US dataset, and we excluded this from scoring. (MedDiet: 9/13 vs 10/14; low-fat: 5/8 vs. 6/9). PREDIMED participants had higher intakes of olive oil, fruits, fish and fewer processed meats and sugary bverages than the US cohorts. We estimated the 4-year cumulative T2D risk was 3.8 (95% CI: 2.8 – 5.1) with a MedDiet diet vs. 2.7 (1.9 – 3.6) for the low-fat pattern (RR of 1.4 [1.0 – 2.3]).

Conclusions: Adapting PREDIMED protocol to two US cohorts resulted in a sample with higher prevalence of dyslipidemia, lower proportion of smokers and distinct background diets. Understanding these differences might offer insights into conflicting results between trial and observational emulations.

Funding Sources: American Diabetes Association (ADA)