Collecting dietary intake data using dietary assessments, such as food frequency questionnaires (FFQ) and diet scores derived from nutrition and lifestyle questionnaires, provides researchers with a general understanding of dietary habits. However, due to the self-reported nature of these methods, data obtained through these subjective assessments are prone to bias [1
]. Subjective assessments are sensitive to human error, including misreported information and recall bias that may potentially skew research findings [2
]. To overcome some of the limitations of the use of subjective nutritional assessment, other objective methods are often used to compliment data collection. Objective methods are quantifiable and less prone to error when compared to subjective assessments and, therefore, are able to provide more accurate information. In nutrition assessment, biomarkers are often used in conjunction with subjective methods to assess the consumption of specific foods, food groups or nutrients of interest and to help assess compliance in dietary intervention studies. Additionally, biomarkers are used to determine the reliability and validity of subjective dietary assessments and can help document compliance in dietary intervention studies [3
]. In contrast some drawbacks to the used of biomarkers are cost, participant discomfort and time.
Several studies have assessed Mediterranean Diet (MD) compliance with the use of diet scores, however, the majority of these scores were validated in older, Mediterranean populations [6
]. Limited research exists regarding the MD’s effects on middle aged, occupationally active, non-Mediterranean populations and there are currently no validated methods to assess MD compliance in this population.
Feeding America’s Bravest is a cluster-randomized-controlled trial that aimed to assess the efficacy of an MD intervention among the 44 fire stations of the Indianapolis Fire Department [7
]. To that end, a modified Mediterranean Diet Score (mMDS) to measure MD adherence [6
] had been developed in a similar Midwestern firefighter population. A previous cross-sectional study in these firefighters using the mMDS found that participants in the highest quartile of mMDS compared to the lowest, had a 35% lower risk for the presence of an additional metabolic syndrome component. Additionally, higher HDL-cholesterol and lower LDL-cholesterol were observed in those with higher mMDS adherence [6
]. However, while the cross-sectional results in the above-mentioned study among firefighters indirectly support the validity of the mMDS, there were two limitations: (1) the validity of the mMDS questionnaire had not yet been assessed against validated questionnaires and (2) the score is unable to biologically estimate the consumption of specific key MD dietary components. Mono and polyunsaturated fats from sources such as olive oil, nuts and fish, including oleic acid, omega-3 and omega-6, have been observed to change in response to MD consumption [5
]. Additionally, elevated consumption of olive oil has been previously associated with urinary tyrosol and hydroxytyrosol [16
]. These bioactive compounds, consumed as part of the MD, have been shown to positively affect the inflammatory response markers, such as Tumor Necrosis Factor (TNF-α), Interleukine-6 (IL-6) and C-Reactive Protein (CRP) [18
]. By validating the mMDS within the current intervention is important because a successful MD intervention would suggest that our approach could be disseminated among other public safety professions or workplaces in non-Mediterranean populations. Therefore, the aim of the current pilot study was (1) to assess the validity of the mMDS questionnaire with a validated FFQ and the use of biomarkers and (2) to use a panel of plasma biomarkers of fatty acids (40 fatty acids, including oleic acid, omega 3 and omega 6) and systemic inflammation (CRP, IL-6 and TNF-α), in addition to urine biomarkers (tyrosol and hydroxytyrosol) to assess compliance with an MD intervention among a random sub-sample of firefighters participating in the parent trial, “Feeding America’s Bravest.”
In the current study, we evaluated the validity of the 13-domain mMDS questionnaire to assess adherence to an MD pattern among firefighters against two high standard instruments: a previously validated, 131-item FFQ and a battery of dietary biomarkers. Regardless of either high or low adherence to the MD, the agreement between the mMDS questionnaire and the mMDS-derived from FFQ showed excellent concordance (>0.75) [32
] among the overall population and in the pilot follow-up study. Additionally, more than 80% of participants’ MD pattern (high or low adherence to MD) were correctly classified by both questionnaires. These findings are similar to previous research that has found comparable results between the ability of full-length FFQs and shorter questionnaires to assess, identify and classify MD dietary adherence [34
By food item, strong correlations were found for olive oil as type of oil used, consumption of fried foods and the type of alcohol ingested. The lowest correlations were found with fruit intake, the consumption of bread and starches with meals and the type of beverage consumed with meals. Similar correlations between food items have been reported in other studies [35
]. While the mMDS questionnaire measures food habits directly with straightforward responses (e.g., how many servings of nuts do you consume per week?), the FFQ assesses dietary consumption more generally and with at least 9 different responses (e.g., how often on average have you consumed 1 oz. of walnuts during the past year?). In addition, in the FFQ, food items are evaluated individually, using standardized portion sizes and are then calculated to estimate the overall consumption of specific food groups over the previous 12-months. Due to differences in specificity and format, this could mean that the mMDS questionnaire is more sensitive for evaluating MD adherence than the FFQ. It has been shown that dietary patterns are reported with more accuracy when asked in multiple, individual questions (like in the mMDS), as opposed to single, grouped questions [38
]. Additionally, accuracy has been shown to increase when questionnaires address the consumption of different types of foods (e.g., milk; whole, skim, 1%, 2%, etc.) using nesting methods as opposed to multiple separate questions (as in in the FFQ) [38
To further assess if adherence to MD could be measured by the mMDS, the association between key components of the MD and select biomarkers was analyzed. Traditionally, the MD is known for its high consumption of fat (>40%), specifically mono and polyunsaturated fatty acids (MUFA and PUFA) from primarily olive oil, nuts and fish sources. We found that choosing olive oil (over other sources of fat) and greater consumption of olive oil was associated with higher levels of plasma omega-3, a PUFA. Although olive oil is characterized by its high content of oleic acid (omega-9), a MUFA, previous research has shown that higher consumption of olive oil has been associated with changes in concentrations of omega-3 [36
] despite changes in concentrations of omega-9 [39
]. Similar to previous research, we did not find an association between choosing or consuming olive oil and omega-9 levels. However, levels of omega 3 improved with MD intervention in the self-sustained continuation group compared to baseline, supporting compliance to the intervention.
Consumption of olive oil was also associated with lower TNF-α at baseline. TNF-α is an inflammatory cytokine that has been linked with obesity, T2DM, hypertriglyceridemia, decreased HDL-cholesterol and CVD, among other inflammatory and infectious diseases. When elevated, TNF-α has been shown to correlate to the severity of disease and also acts as a predictor of CVD and T2DM. However, scientific literature has observed that only olive oil supplemented within the context of the MD has been shown to influence TNF-α levels [40
]. Olive oil is an antioxidant rich and MUFA dense and intake of only 50 mL per day, in addition to the consumption of an MD, has shown significant reductions of inflammatory biomarkers, including TNF-α. But when supplemented in the habitual non-MD diets of overweight, obese and diabetic subjects TNF-α production remained unaffected [40
]. This suggests that the addition of olive oil, outside of an MD context, may not elicit the same anti-inflammatory effects. High in fruits, vegetables, nuts and legumes, the effects of the MD on inflammation may be enhanced by olive oil consumption as it is well known that the intake of oils increases the bioavailability of polyphenols and carotenoids [40
]. Additionally, factors such as BMI, body composition, disease status, physical activity level, microbiome, behavioral and genetic factors all influence the effects of olive oil consumption in reducing TNF-α [42
]. As can be seen in our data (Table 1
), our participants had a high prevalence of obesity.
In regard to tyrosol and hydroxytyrosol in response to olive oil intake, no statistically significant associations were found. Immediate and extended olive oil consumption has previously been associated with strong dose responses to both phenolic compounds [16
]; however, in these studies participants were required to consume at least 25mL of olive oil per day; whereas Feeding America’s Bravest participants were only encouraged to increase olive oil consumption. It is also possible that the consumption of alcoholic beverages like beer or wine that contain also significant quantities of tyrosol may have masked the effect of olive oil [43
Additionally, our study found that the consumption of red and processed meat was associated with lower omega-3 levels. Contrary to previous research, we did not find a statistical association between meat consumption and omega-6 [44
]. Exogenous in nature, omega-3 and omega-6 fatty acids are not produced in the body and can only be obtained through dietary consumption, therefore, circulating concentrations of PUFA act as reliable indicators of their consumption [5
]. Low-fat dietary patterns are associated with decreased plasma concentrations of both omega-3 and omega-6 [45
]. However, meat consumption correlates with elevated plasma levels of arachidonic acid [44
], an omega-6 PUFA found predominantly in grain-fed animals, dairy and eggs [46
]. After 6-months of follow-up, there was a 4–5% decrease in SFA plasma biomarkers in both intervention groups. Because SFAs can be altered endogenously, plasma SFA concentrations are not considered reliable indicators of saturated fat intake [11
]. It has been observed that in carbohydrate restricted diets, where the primary source of fat came from unsaturated fatty acid sources, plasma SFA decreased while carb restricted diets, where the primary source of fat came from SFA sources, saw no changes in plasma SFA [48
In agreement with other studies [36
], we found that the consumption of at least 3 servings of fish per week was associated with higher plasma omega-3 levels, with no changes in omega-6 levels. Fish consumption was also associated with lower IL-6 levels at baseline. IL-6, another inflammatory cytokine, has previously been inversely associated with MD consumption. In their study, Mena et al. found that when assessed in response to a low-fat dietary intervention, IL-6 levels remained static or even increased, while IL-6 levels in the MD group showed consistent, significant decreases [50
]. Fish consumption has also been shown to decrease biomarkers of low grade inflammation (including IL-6) even after being adjusted for sex, energy intake, BMI, physical activity, alcohol consumption, smoking behavior and other food groups (including fruits, vegetables and dairy) [51
Omega-3 was associated with, individual components of the MD as well as with the overall mMDS. Diet quality scores such as the Alternate Healthy Eating Index [52
], Brazilian Healthy Eating Index [54
] and the Diet Quality Index [55
] have also been positively associated with omega-3 levels. This raises the question of whether omega-3 could potentially be validated as a biomarker for diet quality.
In addition, in our study specific self-reported nutrients evaluated from the FFQ showed a strong correlation with the corresponding biomarkers in plasma both at baseline and after 6-months of the intervention. Specifically, this was true for omega-3, EPA and DHA. In contrast, palmitoleic acid, a known SFA, showed a negative correlation supporting the validity of the self-reported diet. The additional correlation between PREDIMED score and mMDS supports that the mMDS is capturing the MedDiet pattern.
This study is limited by its small sample size and the possibility of some volunteer bias among the randomly selected firefighters who declined or accepted to participate in the biomarker study. In addition, the results of this study may not be generalizable to other working populations or females, as the study population for the biomarkers pilot study was majority male. Furthermore, while many biomarkers were analyzed in this study, we did not measure all relevant biomarkers, only a selection. Lastly, the Feeding America’s Bravest parent study did not initially call for the use of biomarkers for validation and, therefore, true baseline data (typically collected before the start of a dietary intervention) could not be assessed. Future research should aim to conduct a larger, more comprehensive biomarker study.