**4. Discussion**

In this sub-study of firefighters in Indianapolis participating in a cluster-randomized MedDiet intervention trial, we found that the MedDiet intervention was associated with favorable changes in markers of cardiovascular risk, specifically those related to the lipid metabolism (cholesterol, lipid composition, or cholesterol esters in the VLDL, IDL, and LDL lipoprotein subclasses, and ApoB/ApoA ratio) that were non-significant after correcting for multiple testing (except for a decrease in M-VLDL-CE). When the adherence to the MedDiet was measured with a self-reported scale (mMDS), the direction of the association with metabolites was similar at both time points (baseline and 6 months after the follow up).

Our results highlighting the changes in plasma metabolites related to lipid metabolism are in line with other studies [20,26]. A recent investigation that identified a metabolic signature of adherence to the MedDiet showed that out of 67 metabolites, 45 were lipids followed by 19 amino acids, 2 vitamins and 1 xenobiotic [26]. Although we used a different methodology and a different set of biomarkers, and thus we could not replicate this metabolic signature, our results support that the MedDiet may induce changes in relevant lipid species and subclasses related to atherogenic risk. In fact, the MedDiet is high in healthy fats (>35–40% of the total energy) mostly from monounsaturated fatty acids (MUFAs) (olive oil mostly) and PUFAs (from nuts and fish), and therefore the results are not surprising. In the firefighter population, we previously reported good correlation between nutrient intake from the

food frequency questionnaire and the corresponding plasma biomarkers (omega-3, Eicosapentaenoic acid (EPA), and DHA) [36]. In line with these results, we found that changes in the MedDiet scores showed some tendency to be associated with fatty acids (an increase in PUFA% specifically DHA% in the expense of MUFA%). Although olive oil is a main component of the MedDiet, previous research found that higher consumption of this oil was linked to changes in omega 3, but not MUFA concentrations [36,41,42]. In addition, the average olive oil consumption in the firefighters is only approximately 0.5 tbsp/day, which is similar to other US cohorts [43] but much lower than in a Spanish cohort (4 tbsp/day) [44]. Nonetheless, it looks like changes in omega 3 to fatty acid ratio, PUFA to FA ratio and DHA to FA ratio increase with changes in the adherence to the MedDiet. This is in line with other studies [20,36], and suggests that those biomarkers could serve as indicators of adherence to the MedDiet.

We found that the MedDiet intervention induced a decrease in total cholesterol, remnant-C, VLDL-C and LDL-C and an increase in HDL-C and HDL2-C. Many studies have already demonstrated the e ffect of the MedDiet on total lipid metabolism, especially reducing total cholesterol and increasing HDL-C [45]. For example, the PREDIMED study, a randomized control trial, found that those in the MedDiet intervention (with olive oil or nuts) over 6 months had an increase in HDL-C but not a reduction in LDL-C [45]. Other studies support that replacing dietary saturated fatty acids (SFAs) with PUFA reduces the plasma LDL-C and subsequently the risk of cardiovascular disease [46–48].

In our study, we also observed a decrease in large, medium and small LDL fractions such as total lipids, cholesterol, particle concentration or cholesterol esters. Similarly, the MedDiet intervention decreased total cholesterol and cholesterol esters in the large, medium and small VLDL. Literature shows that VLDL concentrations are related directly or indirectly in the development of atherosclerosis [49]; for example the fatty acid composition of VLDL is critical for the activity of lipoprotein lipase and the formation of proatherogenic LDL and VLDL remnants [50]. In the FINRISK cohort, increased risk of cardiovascular disease was associated with all VLDL, IDL, and LDL subclasses, while the L- and M-HDL subclasses were associated with lower risk [51]. Despite the evidence of the role of these metabolites in CVD development, few studies have studied the e ffect of a MedDiet intervention in di fferent lipids composition of lipoproteins or its subfractions. Interestingly, our results on lipid subfractions agree with a recent publication using the same metabolomic approach, where 47 participants were randomized to a SFA-rich diet, a MUFA-rich diet or a MED diet for 8 weeks. Additionally, in another study, compared to the control group, those participants that replaced SFAs with PUFAs reduced the lipoprotein particle concentration [52]. Finally, olive oil consumption modifies the lipid composition of VLDL [53] as well as the lipoprotein subfractions [54].

In our study, the e ffect was consistently shown in both groups, usually being stronger in the group undergoing a longer MedDiet intervention/exposure (MedDiet intervention + a self-sustained continuation phase), suggesting that the MedDiet induces favorable changes in metabolites related to CVD disease while the adherence to the MedDiet is sustained. For example, in our study, the ApoB/ApoA1 ratio was decreased in both groups after the intervention and also by adherence to the mMDS, which agrees with other short-term randomized trial with the MedDiet supplemented with olive oil, suggesting that these ratios may predict CVD beyond conventional lipid measures [55]. Finally, we found a significant association with lactate, a metabolite that was previously shown to increase the diabetes risk in the PREDIMED study [56]. However, we found that it occurs in the opposite direction.
