Maternal Adherence to the Mediterranean Diet during Pregnancy: A Review of Commonly Used a priori Indexes

Currently, many a priori indexes are being used to assess maternal adherence to the Mediterranean diet (MD) during pregnancy but each with different components, cut-off points, and scoring systems. This narrative review aimed to identify all observational studies utilizing a priori indexes to assess maternal adherence to the MD during pregnancy. A systematic search was conducted in Pubmed until 1 July 2020. Among the 27 studies included, eight different a priori indexes were identified. Studies included a range of 5 to 13 dietary components in their indexes. Only three dietary components—vegetables, fruits, and fish—were common among all indexes. Dairy and alcohol were the only two components modified for pregnancy. All but one study either excluded alcohol from their index or reversed its scoring to contribute to decreased adherence to the MD. Approximately half of the studies established cut-off points based on the distribution of the study population; the others utilized fixed criteria. This review emphasizes the incongruent definitions of the MD impairing effective comparison among studies relating to maternal or offspring health outcomes. Future research should carefully consider the heterogeneous definitions of the MD in a priori indexes and the relevance of incorporating pregnancy-specific nutritional requirements.


Introduction
The Mediterranean diet (MD) has long been heralded as one of the healthiest dietary patterns worldwide [1]. Traditionally, the MD reflects the common dietary characteristics of populations native to the Mediterranean basin, including countries such as Greece and Italy [2]. The predominant elements of the MD consist of high intakes of vegetables, fruits, legumes, nuts, seeds, whole grains, and olive oil; moderate intakes of fish and alcohol; low to moderate intakes of dairy; and low intakes of meat and poultry [3]. As such, the MD is largely plant-based abundant in fiber and antioxidants as well as monounsaturated (MUFA) and polyunsaturated fatty acids [2,4]. Over the past decades, numerous epidemiological studies have investigated the influence of adherence to the MD on health outcomes, ultimately showing protective associations with a host of non-communicable diseases ranging from cardiovascular disease (CVD) to cancer [5]. Recently, systematic reviews have also reported protective associations between maternal adherence to the MD during pregnancy and various maternal and offspring outcomes, including gestational diabetes mellitus (GDM) [6] and offspring adiposity [7,8].
To date, several a priori indexes have been developed to assess adherence to the MD using a scoring system based on predefined features of the MD [9][10][11]. In fact, a narrative review identified 22 different indexes used to assess adherence to the MD in adult and elderly populations [10]. In these indexes, beneficial foods contribute to increased adherence to the MD, whereas detrimental foods contribute to decreased adherence to the MD. Yet, each index features vastly different components, cut-off points, and scoring systems [10]. To illustrate, the original Mediterranean diet score (MDS) was developed by Trichopoulou et al. (1995) and consists of six beneficial components (i.e., vegetables, fruits and nuts, cereals, legumes, alcohol, and a ratio of MUFA to saturated fatty acids (MUFA:SFA)) as well as two detrimental components (i.e., meat and dairy), each of which are scored 0-1 based on the median intake levels of the study population [12]. In 2003, the original MDS was modified by the same authors to produce the MDS-2003, which additionally include fish as a beneficial component [13]. Conversely, the MDS-2006 utilizes eleven components, adding separate categories for potatoes and poultry and replacing MUFA:SFA with olive oil, each of which are scored 0-5 based on fixed servings of consumption [14]. Alternatively, the Prevención con Dieta MEDiterránea (PREDIMED) score uses the Mediterranean Diet Adherence Screener (MEDAS), a 14-item questionnaire consisting of twelve food frequency (e.g., daily amount of fruit consumed) and two food habit questions (e.g., use of olive oil during cooking), for rapid assessment of adherence to the MD during trials or clinical settings [15]. Although the PREDIMED score includes many components traditional to the original MDS and the MDS-2003 [12,13], it also incorporates components non-traditional to these indexes, including sweetened or carbonated beverages and commercial pastries [15].
Beyond the inherent differences among a priori indexes, nutritional requirements during pregnancy further complicate the application of a priori indexes in pregnant populations [16]. Although moderate alcohol consumption is considered a beneficial component in the traditional MD [2,4], abstinence from alcohol is widely recommended during pregnancy to avoid adverse fetal development [17,18]. Furthermore, several micronutrient requirements (e.g., folic acid, iron, and calcium) increase during pregnancy [19]. Consequently, many epidemiological studies evaluating maternal adherence to the MD during pregnancy on maternal and offspring health outcomes have altered the original designs of a priori indexes in order to address these nutritional requirements during pregnancy [20][21][22][23]. Such modifications include scoring alcohol as a detrimental component [22] or removing it entirely from the index [20,21,23] as well as scoring dairy as a beneficial component to meet increased calcium requirements [20]. In 2009, Mariscal-Arcas et al. (2009) proposed an a priori index specifically for use during pregnancy called the MDS for pregnancy, which incorporated folic acid, iron, and calcium but maintained dairy as a detrimental component in the scoring system [16].
Whether due to inherent differences or subsequent modifications in light of pregnancy, the heterogeneity in the inclusion and discrimination of components in a priori indexes greatly challenges comparison among studies [9,11], thus, impeding the ability to definitively draw conclusions on maternal and offspring health outcomes. Therefore, the goal of this narrative review was to identify all observational studies utilizing a priori indexes to assess maternal adherence to the MD during pregnancy with a particular emphasis on evaluating the food and nutrient components, namely the choice and discrimination of dietary components in these indexes, in addition to the cut-off values and scoring systems.

Materials and Methods
A comprehensive search was conducted in Pubmed in order to identify all observational studies published evaluating maternal adherence to the MD during pregnancy without time limits through 1 July 2020. The search string comprised keywords and Medical Subject Headings (MESH) terms relating to maternal dietary intake during pregnancy and adherence to the MD as described in Table 1. The initial database search was also supplemented by a manual search of reference lists of relevant studies in order to identify studies not retrieved by the initial search in Pubmed. Only observational studies (i.e., cohort studies, case-control studies, and cross-sectional studies) that utilized an a priori index to assess the exposure of maternal adherence to the MD during pregnancy were included in this review. Reasons for excluding articles from the review included: (1) Irrelevant population, (2) irrelevant exposure, (3) incomplete information on the a priori index in methods, (4) non-observational study design, (5) review, meta-analyses, editorials, or conference proceedings, and (6) no English translation available. Figure 1 describes the literature search and selection process in more detail. From the studies selected for inclusion in this review, we extracted data on the authors and year of publication; study design; population characteristics; method of dietary assessment; the definition, cut-off values, and scoring system of the a priori index; as well as the main outcomes.  Figure 1. Flowchart of identification, screening, and inclusion of studies using a priori indexes to assess maternal adherence to the MD during pregnancy. * Exclusion criteria for review: (1) Irrelevant population (e.g., non-pregnant, pre-pregnant, or peri-conceptional study population), (2) irrelevant exposure (i.e., exposure not pertaining to adherence to MD), (3) incomplete information on the a priori index in methods, (4) non-observational study design study (e.g., RCT), (5) reviews, metaanalyses, editorials, or conference proceedings, and (6) no English translation available. Abbreviation: MD, Mediterranean diet; RCT, randomized controlled trial.

Study Characteristics
In total, 27 observational studies were included in the review: 17 (63.0%) cohort studies [20][21][22][24][25][26][27][28][29][30][31][32][33][34][35][36][37]; 7 (25.9%) cross-sectional studies [38][39][40][41][42][43][44]; 2 (7.4%) case-control studies [23,45]; and, 1 (3.7%) nested case-control study [46]. Table 2 provides a summary of the key characteristics of these 27 studies. The studies were published from 2008 to 2020. 20 studies were conducted in Europe of which 17 were conducted in the Mediterranean countries of Spain [21][22][23][25][26][27][28][29][30][40][41][42]44,46], Greece [20,28,29,38], and Italy [43], whereas the other three were conducted in Norway [31], Denmark [35], and the United Kingdom [24]. Seven studies were conducted in North America including the United States [20,[32][33][34]36], Mexico [39], and the Caribbean [37]. One study was conducted in Iran [45]. Notably, three studies utilized cohorts from two geographical locations in their studies, including Greece and Spain [28,29] and Greece and the USA [20].      * Dietary components with (+) indicates contributed to increased adherence to the MD; (−) indicates contributed to decreased adherence to the MD; (.) indicates contribution was unspecified within the study's methods. † Lowest number in range indicates minimal adherence to the MD and highest number indicates maximal adherence. § Cut-off points based on fixed servings per day or week. ¶ Women with higher intakes of beneficial foods greater than the median intake received +1; women with lower intakes of beneficial foods less than the median received 0. ** The number of pregnant women or mothers was not specified within the study. † † Unspecified time period during pregnancy in which dietary data was collected. § § Women with higher intakes of beneficial foods greater than or equal to the median intake received +1; women with lower intakes of beneficial foods less than the median received 0. ¶ ¶ Women with upper half of intake of beneficial foods received +1; women with lower half of intake of detrimental foods received +1. *** No information provided on components, cut-off points, or scoring system of aMed. † † † PREDIMED score includes two questions on olive oil and one question on every other component. § § § Sofrito was described as a sauce consisting of tomatoes, garlic, onion, and peppers or leeks sautéed in olive oil and served with dishes of vegetables, rice, or pasta. ¶ ¶ ¶ Cut-off points based on combination of fixed servings per week or month and food habits. **** Alcohol was scored dichotomously: 0 for any consumption and 2 for no consumption. † † † † KIDMED questionnaire includes two questions each on vegetables, fruits, and dairy products and one question on every other component.

Type of Index
Eight a priori indexes were identified among the included studies for the assessment of maternal adherence to the MD during pregnancy. The most commonly used index was the MDS-2003 (n = 12) [20,23,24,[27][28][29][32][33][34]37,39,45] followed by the PREDIMED score (n = 5) [23,40,41,43,44] [48] but with additional food components (i.e., potatoes, pasta, rice, and fast food) included at the discretion of the study group [25,26]. Others used the MDS-2006 (n = 2) [23,38]; a modified version of the Mediterranean diet quality index in children and adolescents (KIDMED) index (n = 2) [42,46]; as well as Khoury's criteria (n = 2) [31,35] based on principles established in a randomized controlled trial on pregnant women [51]. Only one study used the alternative MD (aMed)-another variation of the MDS-2003 [59] adapted in a US cohort-in their main analysis [36]. Almost all studies used one index in their assessments of maternal adherence to the MD during pregnancy with the exception of Martínez-Galiano et al. (2018) in which three indexes were utilized: The MDS-2003, PREDIMED score, and MDS-2006 [23]. Notably, Fernández-Barrés et al. (2016) reported using rMed in the main analysis and aMed in the sensitivity analysis. However, no information was provided in the methods on the specific components, cut-off values, and scoring system of aMed; for this reason, only their use of rMed was evaluated in this review [21].

Food and Nutrient Components
Overall, the number of dietary components included in the indexes ranged from 5 [31,35] to 13 [42,44,46] depending upon the chosen index and any subsequent modifications to that index. However, in total, 17 different dietary components were identified among the indexes of the twenty-seven studies. 10 components were considered traditional to the components included in the original MDS and MDS-2003 [12,13]; seven components were considered non-traditional to these original indexes [12,13]. Beneficial components are defined as dietary components contributing to increased adherence to the MD, whereas detrimental components are defined as dietary components contributing to decreased adherence to the MD. Table 3 provides a comparison of the food and nutrient components incorporated into the a priori indexes of the 27 included studies. Of note, Martínez-Galiano et al. (2018) utilizes three different a priori indexes; therefore, the number of studies does not always add up to 27 when the inclusion, discrimination, or scoring of components differed among the three indexes. Any discrepancies among the three indexes in this study is indicated throughout this review. Additionally, the authors of this study did not specify the scoring for any components in the MDS-2003, but did specify the scoring of components in the PREDIMED score and MDS-2006 [23]. Table 3. Compa ison of food and nutrient components *, cut-off points, and range of scores in a priori indexes assessing maternal adherence to the MD during pregnancy.

Other Food Components
Besides the aforementioned food components traditional to the original designs of the MD [12,13], several studies also included additional food components in their indexes.
(2008) added a junk food and fat component to the MDS-2003, which collectively grouped fast food hamburgers with desserts (i.e., pastries, candies, and ice cream), snacks (i.e., chips and popcorn), as well as butter and margarine [39]. Contrastingly, the two studies employing the modified KIDMED index incorporated the food habit component of visiting a fast food ("hamburger") restaurant each week [42,46].

Sweets, Candies, and Pastries
Seven studies included sweets, candies, and pastries as detrimental components in their indexes [23,[40][41][42][43][44]46]. Notably, the two studies utilizing the modified KIDMED index included two separate categories: One for commercial baked goods and pastries and the other for sweets and candy [42,46]. Interestingly, both categories were completely divorced from the fast food component [42,46], unlike the three studies previously discussed which included sweets, pastries, and desserts in the fast food component [25,26,39]. The five studies using the PREDIMED score incorporated a component for commercial pastries, which did not include sweets or candies [23,40,41,43,44].

Sweetened or Carbonated Beverages or Coffee
Five studies included sweetened or carbonated beverages [23,40,41,43,44] and two studies included coffee [31,35] in alignment with their respective indexes [15,51]. The five studies including sweetened, carbonated beverages in their indexes all employed the PREDIMED score and designated this component as detrimental [23,40,41,43,44]. The two studies including coffee used Khoury's criteria and similarly scored drinking more than two servings of coffee a day as a detrimental component [31,35].

Skipping Breakfast
The two studies using the modified KIDMED index included a food habit component on skipping breakfast as a detrimental component in their index [42,46].
Interestingly, four studies deviated from their respective indexes to utilize fixed consumption of components as cut-off points [20,25,26,33]. In their modified version of the MDS-2004, Castro-Rodriguez et al. (2010) and (2016) utilized fixed servings and scored components 0, +1, or +2 with higher intakes of beneficial foods receiving higher scoring and higher intakes of detrimental foods receiving lower scoring [25,26].

Range of Scores
Although all studies reported a higher final score indicating higher adherence to MD and a lower score indicating less adherence, the range of scores varied greatly depending upon the chosen index and subsequent modifications to components . Overall, the range of scores ranged from 0-5 [31,35] to 0-55 [38]. A total of eight studies maintained the original range of scores designated in their respective indexes [22,31,35,37,38,42,44,46] of which five also maintained the original scoring of all components [31,35,42,44,46] and three modified alcohol to be scored as a detrimental component in their indexes [22,37,38].
Beyond the number of components, the degree of discrimination among food components was a major source of heterogeneity. First, nine studies grouped fruits and nuts together into a single category [21,22,24,[27][28][29][30]37,39] in line with the initial categorization of their indexes [13,49]. Although minimally processed fruits and nuts are both constituents in many healthful dietary patterns [60], nutritionists and dietitians generally regard fruits and nuts as two distinct foods composed of different nutritional properties as evidenced by separate categorizations and recommendations in countries' dietary guidelines [17,61]. While separating fruits and nuts is therefore in agreement with such dietary guidelines, it could affect the scoring system by perhaps disproportionately weighing the importance of these foods over the other dietary components in the indexes. Second, four studies separated vegetables and potatoes into separate components in their indexes [23,25,26,38]. Only one study explicitly reported the inclusion of root vegetables in the vegetable component [24], whereas for the other studies this was not clearly stated [20][21][22][23][27][28][29][30][31][32][33][34][35][36][37][39][40][41][42][43][44][45][46]. Although some countries such as the United States group vegetables and potatoes together in their dietary guidelines [17], a distinction between these foods in a priori indexes is likely relevant for pregnancy given results showing increased risk of GDM in women with higher pre-pregnancy consumption of potatoes [62]. Third, eight studies distinguished among the cereal component to include only non-refined or whole grains as a beneficial component in their indexes [20,23,[32][33][34]36,38,45]. In general adult populations, it was shown that consumption of non-refined, rather than refined, grains was associated with reduced risk of coronary heart disease [63] and type 2 diabetes mellitus [64]. Furthermore, recent research has shown that consumption of refined grains during pregnancy by women with GDM was associated with risk of increased body mass index z-score and overweight and obesity in offspring at 7 years (yr) [65]. This complicates the precedent of utilizing one all-encompassing component for cereals in many a priori indexes [12,13]. Additionally, five studies included red and processed meat as a detrimental component in their index as opposed to one all-encompassing component for meat [20,[32][33][34]36]. In general adult populations, red and processed meat consumption increases the risk of CVD and all-cause mortality [66,67], particularly in comparison to white-meat (e.g., poultry) [60,66]. Although comparatively research during pregnancy is limited, research has shown the higher pre-pregnancy consumption of a Western dietary pattern, explained largely by red and processed meat products, was associated with increased risk of GDM in comparison to a more prudent dietary pattern, featuring high intakes of vegetables and fruit as well as poultry and fish [68].
Aside from discrimination among components, considerable discrepancies also arose from amendments to dairy and alcohol-the only components to be modified in light of pregnancy. According to the initial scoring of most indexes included in this review, dairy stands as a detrimental component contributing to decreased adherence to the MD [13][14][15]47,49]. This likely stems from the low to moderate consumption of dairy products observed in the traditional MD [12,13] as well as the controversy surrounding reducing SFA intake for CVD prevention [69]. Nevertheless, nine studies deviated from this precedent to score dairy as a beneficial component [20,24,[27][28][29][32][33][34]37], likely due to increased calcium requirements during pregnancy [19]. Although no clear consensus has been reached, recent research has also reported a neutral association between total dairy consumption and CVD and an inverse association with type 2 diabetes mellitus [70,71]. This provides a further reason that the scoring of dairy products should be reconsidered in pregnant and non-pregnant populations alike. Moreover, given the widespread recommendation to abstain from alcohol during pregnancy [17,18], it remains unsurprising that 19 studies deviated from their indexes to exclude alcohol [20,21,[23][24][25][26][27][28][29][30][32][33][34]36,[39][40][41]43,45]. Nevertheless, future studies should be cautious about reflexively excluding alcohol from their index as this could misclassify some pregnant participants who do consume alcohol during pregnancy as non-consumers. If alcohol consumption is reported by pregnant participants in a cohort, reversing the scoring of alcohol to be considered a detrimental component could also serve as a viable option as performed in three studies in this review [22,37,38]. Although even in this approach, future studies should take into account that then very low alcohol consumption could inordinately influence the scoring system as much as low intakes of vegetables or high intakes of red and processed meat given these indexes allot equal weight to each individual dietary component.
Beyond the components themselves, differences in cut-off points further challenged the comparability of indexes. Although the original designs of the MD utilize the median intake values of the study population as cut-off points [12,13], the use of median intake levels ignores the extremes of intake within a population [72]. As a result, it ignores variance of intake within a population and erroneously assumes homogeneity of risk among all intake levels [73]. Furthermore, establishing cut-off points based on the distribution of intake within a study population limits comparability between studies, which proves particularly cumbersome for comparisons between Mediterranean and non-Mediterranean countries due to vastly different dietary habits [74]. While fixed criteria could serve as a possible solution to comparability among studies, it should be noted that the use of fixed criteria is limited in its ability to distinguish among intake levels when components are rarely consumed in a population (e.g., alcohol consumption during pregnancy). Furthermore, it is important to note that FFQs are a common dietary assessment method used to measure dietary intake in large epidemiological studies [74]. Instead of capturing absolute intake, FFQs operate by accurately ranking individuals based on habitual intake levels [75]. Therefore, basing cut-off points on the distribution remains consistent with the use of FFQs [74]. If tertiles of intake are utilized as employed in rMed, it would allow for the additional advantage of better assessing the variance of intake within a study population by resulting in a wider range of scores than possible using median intake levels alone [49].
In regard to dietary assessment methods, variability was evident in both the choice and the time period in which dietary data was collected. Most studies employed an FFQ [20][21][22][23][24][25][26][27][28][29][30][31][32][34][35][36][37][38][39][40][41]43]; two studies utilized repeated 24-h recalls [33,45]; and three used index-specific questionnaires [42,44,46]. Given that each of these methods are self-reported, information bias (e.g., recall bias) remains an ever-present possibility obscuring the accurate assessment of dietary intake by participants [76]. However, even more so, only two studies measured dietary intake during each trimester of pregnancy [22,33]. While it has been reported that dietary intake patterns do not drastically change during pregnancy [77], physiological changes during pregnancy (e.g., nausea) may alter normal food consumption over short periods of time [78], which may not be as readily captured within the timeframe of an FFQ designed to capture habitual food intake [75]. Therefore, if feasible within study constraints, studies should consider utilizing repeated 24-h recalls, an FFQ multiple times during each trimester, or a combination of the two in order to better ensure a truly accurate estimation of dietary intake during pregnancy.
Lastly, it is worth emphasizing that only one index in this review-Khoury's criteriawas specifically developed for use during pregnancy [51]. Conversely, most of the other indexes were validated in the general adult population [13,14,49,59]. Specifically, the PRED-IMED score was developed for use in older adults (55-80 years) at high-risk for CVD [15]. Alternatively, the KIDMED index was validated in younger individuals (2-24 years), therefore, explaining the exclusion of alcohol from its index [55]. Besides the incongruity of some traditional components of the MD to pregnancy (e.g., alcohol), adherence to the MD does not necessitate many important nutritional recommendations during pregnancy essential to fetal growth and development, including adequate micronutrient intake and sufficient hydration. Importantly, inadequate intake of micronutrients during pregnancy results in adverse offspring health outcomes, including neural tube defects from inadequate intake of folate [79]. To date, however, only the MDS for pregnancy has incorporated folic acid, iron, and calcium into their a priori index; yet, this index and these micronutrient considerations were not represented in any of the studies in this review [16]. Similarly, insufficient hydration during pregnancy has also been shown to be associated with adverse offspring health outcomes ranging from neural tube defects to musculoskeletal and congenital heart defects [80]. While untraditional to the original indexes of the MD [12,13], the aforementioned nutritional recommendations during pregnancy should be considered for incorporation into a priori indexes assessing maternal adherence to the MD during pregnancy when studying the MD as a proxy for a healthy diet during pregnancy due to their immense importance in fetal growth and development [79,80].
However, ultimately it must be emphasized that MD is a descriptive diet reflecting the common dietary characteristics traditional to populations in the Mediterranean basin [1,2]. As such, it is not a standard diet quality index designed to ensure optimal health based on the most current nutrition knowledge [1,81]. Similar to arguments concerning the incorporation of modern food components (e.g., fast food and desserts) [39,55], the addition of components specific to nutritional requirements during pregnancy would amplify the relevancy of a priori indexes assessing adherence to the MD in today's society. Yet, these additions would consequently transition these a priori indexes away from actually measuring the MD. Therefore, future studies must carefully weigh these advantages and disadvantages of modifying a priori indexes for use in pregnant populations evaluating whether they are truly interested in measuring adherence to the MD or simply adherence to a healthy diet quality index for pregnancy.

Conclusions
In conclusion, this narrative review emphasizes the vast heterogeneity and subjectivity present among the components, cut-off points, and scoring systems of a priori indexes used to assess maternal adherence to the MD during pregnancy. As a dietary pattern, the MD maintains the key advantage of encapsulating the synergistic effects of nutrients, which would otherwise remain undetectable when investigating single nutrients alone [9,60]. Although a priori indexes provide a useful means to assess maternal adherence to the MD during pregnancy, future studies should carefully examine the importance of addressing pregnancy-specific nutritional recommendations in a priori indexes, including altering the scoring of dairy and alcohol and incorporating micronutrients (e.g., folate) and water. While such modifications may be untraditional to the original indexes of the MDS [12,13], it would allow studies investigating the MD as a proxy for a healthy diet in pregnancy a better approach to address pregnancy-specific nutritional requirements crucial for optimal maternal and offspring health.