First-trimester diet quality in association with maternal subcutaneous and visceral adipose tissue thicknesses and glucose homeostasis during pregnancy

Abstract We aimed to characterise the associations between first-trimester diet quality, adiposity, and glucose homeostasis measurements throughout pregnancy in a sample of 104 healthy pregnant women. Three Web-based 24-h recalls were completed, from which the Alternate Healthy Eating Index (AHEI) was calculated. At each trimester (12.5 ± 0.7, 22.8 ± 1.0, and 33.6 ± 1.3 weeks of gestation), fasting glucose and insulin were measured to compute an insulin resistance index (HOMA-IR). Subcutaneous and visceral adipose tissue thicknesses were estimated by ultrasound at the end of the first trimester. Inverse associations were observed between the first-trimester AHEI and first-trimester fasting insulin (r = 0.24; p < 0.05), and HOMA-IR (r = −0.22; p < 0.05), as well as third-trimester fasting insulin (r = −0.20; p < 0.05). A trend was also observed between first-trimester AHEI and first-trimester SAT thickness (r = −0.17; p < 0.1). Pre- and early-pregnancy adiposity measurements were identified as high predictors fasting insulin concentrations throughout pregnancy. Higher early-pregnancy diet quality is associated with more favourable metabolic measurements during pregnancy.


Introduction
Gestational diabetes mellitus (GDM) is the most common disease in pregnancy (Poblete and Olmos 2021). This is partly due to the fact that more pregnant women have excessive weight and adipose fat storage. Indeed, ultrasound-estimated visceral (VAT) adipose tissue thickness in early pregnancy is strongly associated with insulin resistance and higher risks of developing GDM, independently from other known risk factors associated with GDM, such as age and prepregnancy body mass index (ppBMI) (De Souza et al. 2014D'Ambrosi et al. 2018;Alves et al. 2020).
Over the last decades, a growing body of evidence has demonstrated that a healthy diet during pregnancy can improve the overall health of the mother and foetus, notably by supporting adequate maternal glucose homeostasis (Mijatovic-Vukas et al. 2018;Raghavan et al. 2019;Abdollahi et al. 2021). Recent systematic reviews and meta-analyses showed that high-quality diets during pregnancy, characterised by high consumption of vegetables, fruit, whole grains, nuts, legumes, and fish, and low consumption of red and processed meat, are associated with lower risks of obesity and GDM (Mijatovic-Vukas et al. 2018;Raghavan et al. 2019;Zaragoza-Mart ı et al. 2022). The Alternate Healthy Eating Index (AHEI) is a diet quality score that is highly predictive of the risks of metabolic diseases in the general population (Hodge et al. 2021;Li et al. 2021;Patel et al. 2021;Pestoni et al. 2022), and has already been used in pregnant populations in which it has been associated with the risks of GDM and other complications (Tobias et al. 2012;Li et al. 2021;Reyes-L opez et al. 2021). Although several prospective cohort studies showed that a healthy diet is associated with reduced risks of GDM (Tryggvadottir 2016;Mak et al. 2018;Pajunen et al. 2022), it is important to understand if healthier diets have a positive effect on glucose homeostasis in a non-GDM population. In this context, only a few studies have examined the associations between diet quality and maternal metabolic measurements and suggested that adherence to healthy dietary patterns was inversely associated with maternal insulin concentrations and insulin resistance (Martin CL et al. 2016;Lindsay et al. 2020). However, these studies do not include repeated measures of glucose homeostasis. In addition, no study has examined the associations between diet quality and adipose tissue measurements during pregnancy.
Thus, current literature offers data that remains to be confirmed regarding the associations between early-pregnancy diet quality and maternal metabolic measurements throughout pregnancy using a prospective design. Therefore, this study aims to characterise the associations between first-trimester diet quality and 1) first-trimester measurements of adiposity (subcutaneous (SAT), and visceral (VAT) adipose tissue thicknesses) as well as 2) first-, second-, and third-trimester measurements of glucose homeostasis (fasting glucose and insulin concentrations, and an insulin resistance index (HOMA-IR)). We hypothesise that a higher first-trimester diet quality is associated with lower adiposity measurements and better glucose homeostasis throughout pregnancy.

Participants
This is a secondary analysis of data from a merged sample of 104 pregnant women from the ANGE (Apports Nutritionnels Durant la GrossessE) and PAGG (Prise alimentaire, Apports nutritionnels et Gain de poids Durant la Grossesse) prospective cohort studies. The ANGE cohort consists of 86 pregnant women followed from their first trimester of pregnancy until delivery at the Centre de recherche du CHU de Qu ebec-Universit e Laval and was described more thoroughly elsewhere (Savard et al. 2018). Our final analysis includes 79 participants from this cohort. The PAGG cohort consists of 33 pregnant women who were recruited as part of a pilot project aiming at characterising food-related hormones and biochemical markers adaptation throughout pregnancy and examining their associations with food intake and gestational weight gain, between November 2018 and September 2020, also at the Centre de recherche du CHU de Qu ebec-Universit e Laval. Our final analysis includes 25 participants from this cohort. The flow of participants is presented in Figure 1. Both study protocols were similar, as they included on-site visits at each trimester, with metabolic testing, using the same methods and self-administered Web questionnaires, as previously detailed elsewhere (Savard et al. 2018). The selection criteria were identical for both projects. Participants were included if they were aged at least 18 years, were in their first trimester of pregnancy and were able to communicate in French. Exclusion criteria included multiple pregnancies and previously diagnosed severe medical conditions affecting their metabolic profile or intakes (i.e. type 1 and type 2 diabetes, renal and liver disease, as well as inflammatory and autoimmune disorders). Both cohort studies followed the Declaration of Helsinki guidelines and were approved by the Ethics Committee of the CHU de Qu ebec-Universit e Laval Research Centre (ANGE reference number: 2016-2866; PAGG reference number: 2019-4342). All participants gave their informed written consent at their first on-site visit.

Assessment of first-trimester dietary intakes
First-trimester dietary intakes were assessed by three automated and self-administered Web-based 24-hour dietary recalls (R24W), through the Rappel de 24 h Web virtual platform described elsewhere and validated among pregnant women Savard et al. 2018). Briefly, the participants received an automatically generated e-mail on three random days (two weekdays and one weekend day) inviting them to report all foods and drinks consumed the day before (24-hour period). All food items available on the platform are coded using the 2015 version of the Canadian Nutrient File, therefore enabling the automatic extraction of nutritional information. Dietary data were used in the present analysis if the participant completed at least two of the three administered R24W.

First-trimester diet quality
First-trimester diet quality was measured using a modified version of the Alternate Eating Healthy Index 2010 (AHEI) (Chiuve et al. 2012). Briefly, this index allows scoring of an individual's diet quality based on his consumption of foods and nutrients predictive of chronic disease risks. The purpose of this index, linking diet with metabolic diseases, was directly related to the aim of the present study. Mean daily intakes for each food group, expressed in servings, milligrams (mg), grams (g), or percentages of energy, were directly obtained from the R24W. They were then compared to the AHEI scoring criteria (Table S1) to compute the total diet quality score and its subscores: (1) Vegetables, (2) Fruit, (3) Whole grains, (4) Sugar-sweetened beverages (SSB) and fruit juices, (5) Nuts and legumes, (6) Red/processed meat, (7) Trans fats, (8) Long-chain (n-3) fats, (9) Polyunsaturated fatty acids (PUFA), (10) Sodium, and (11) Alcohol. Since the safest choice for pregnant women is not to drink alcohol, we decided to omit the alcohol subscore, as did other authors (Badon et al. 2018;Emond et al. 2018;Gonzalez-Nahm et al. 2019). A score between the minimum (0) and maximum (10) was attributed to every subscore, in proportion to the consumption of each food group or nutrient, except for the "Sugar-sweetened beverages and fruit juices", "Red/processed meat", "Trans fat", and "Sodium" subscores which were scored inversely to the consumption of those foods or nutrients that should be consumed in moderation. Scores for each component were then summed to obtain the total score ranging from 0 to 100, with a higher score representing a higher diet quality.

Measurements of adiposity
A total of 97 participants underwent ultrasound measurement of subcutaneous (SAT) and visceral (VAT) adipose tissue thicknesses between 10 and 14 weeks of gestation on a Voluson E8 Expert system (GE Healthcare Inc., Milwaukee, IL) ultrasound machine using a RAB4-8-D/OB probe. Transabdominal ultrasound was performed by qualified ultrasound technicians using a previously described technique (Martin et al. 2009). All measurements were done in the supine position, perpendicularly to the aorta, at the level of the Linea alba, 2.5 cm above the umbilicus. Improper pressure of the probe, which could compress tissues of interest, was avoided. Images were captured at the end of an expiration when the aorta appeared closest to the surface to avoid the influence of the respiratory status or abdominal wall tension. SAT thickness was measured from the subcutaneous fat layer to the outer border of the rectus abdominus muscle at the level of the Linea alba. VAT thickness was measured from the inner border of the rectus abdominus muscle at the level of the Linea alba to the anterior wall of the abdominal aorta. This technique can produce accurate, reliable, repeatable, and acceptable assessments of subcutaneous and visceral adipose tissue thickness and can allow the detection of individuals with unfavourable metabolic profiles both in the general population (Ribeiro-Filho et al. 2001) and the gestational context (Cremona et al. 2019). It is also validated against the use of computed tomography scans (Stolk et al. 2001;Berker et al. 2010;Bazzocchi et al. 2011) which are not recommended in pregnancy (Menias et al. 2007;Bazzocchi et al. 2011;Flanagan and Bell 2020). Finally, weeks of pregnancy were confirmed by ultrasound in the first trimester.

Statistical analyses
Means, standard deviations (SD), and proportions were used to characterise the study sample. Repeated measures ANOVA with Tukey's honest significant difference (HSD) post hoc tests (with all non-normally distributed variables normalised by BoxCox) were conducted to examine the evolution of circulating fasting concentrations of glucose and insulin, and HOMA-IR index across trimesters. Spearman's nonparametric correlation analyses were conducted on non-normal distributions to examine the associations between first-trimester diet quality scores, and metabolic measurements. Correlation coefficients are presented both raw and adjusted for ppBMI. Multivariate analyses were also conducted to identify predictors of fasting insulin concentrations throughout pregnancy, transformed by root square. Based on the literature and clinical relevance, variables were identified as potential predictors of insulin concentrations. The identified variables (n ¼ 7) were used in the multivariate linear regression using the stepwise procedure (mixed approach) at a probability to enter of p < 0.25 and a probability to leave of p < 0.1 and were tested for collinearity. Finally, four variables were entered in the final explicative models using one variable of adiposity (ppBMI, SAT or VAT thicknesses), first-trimester diet quality, age, and parity, with the three latter forced into the model when applicable. Statistical analyses were performed with JMP Pro version 15.2 (SAS Institute Inc., Cary, NC, USA). The threshold of statistical significance was defined at p 0.05, and 0.05 < p 0.1 were considered as trends.

First-trimester diet quality
Ninety-eight (94.2%) participants filled three R24W, whereas six (5.8%) completed two. First-trimester means of AHEI total score and subscores are presented in Table 2 and Table S2, respectively. The mean total diet quality score was 50.8 ± 12.2, ranging from 25.2 to 77.6. 14.9 ± 4.9 BMI: Body Mass Index; CAD: Canadian dollars. a Other ethnicities include Middle-Eastern (n ¼ 1), Venezuelan (n ¼ 1), and African American (n ¼ 1). b In Quebec, college refers to a degree obtained after high school and before university. Total gestational weight gain, n ¼ 82.

First-trimester diet quality in association with first-trimester adiposity measurements
First-trimester adiposity measurements are presented in Table 2. Mean first-trimester SAT and VAT thicknesses were 17.4 ± 7.6 mm and 27.0 ± 14.6 mm, respectively. As demonstrated in Table 3, before adjustments for ppBMI, the first-trimester AHEI total score tended to be inversely associated with SAT thickness. Yet, this trend did not remain after adjustment for ppBMI. No significant association was observed between VAT thickness and overall diet quality. Participants with a first-trimester AHEI total score greater than or equal to the median score (49.8) had significantly lower first-trimester SAT thickness compared to the participants with a diet score below the median, but no difference was observed regarding VAT thickness (Table S3). The "Fruit" and "Red/processed meat" subscores (the latter being scored inversely to the consumption) were inversely associated with first-trimester SAT thickness, while the "Whole grains" subscore correlated negatively with VAT thickness (Table 4). None of these findings remained significant after adjustments for ppBMI. Additional exploratory analyses have shown no significant association between first-trimester diet quality and total gestational weight gain, with and without adjustment for ppBMI (data not shown).

First-trimester diet quality in association with trimester-specific glucose homeostasis measurements
Trimester-specific glucose homeostasis measurements and evolution throughout pregnancy are presented in Table 2. As shown in Table 3, first-trimester AHEI total score correlated inversely with first-and thirdtrimester fasting insulin, as well as first-trimester HOMA-IR, but those associations did not remain significant after adjustment for ppBMI. Yet, trends remained for the first-trimester fasting insulin and HOMA-IR. No association was found to be significant between the first-trimester AHEI total score and fasting glucose for any trimester. Participants with an AHEI total score greater than or equal to the median score (49.8) in the first trimester, had lower fasting insulin concentrations both in their first and third trimesters, as well as a lower HOMA-IR in the third trimester, compared to the ones below the median (Table S3). The "Fruit", "Whole grains", and "Red/ processed meat" subscores (the latter being scored inversely to the consumption) were all negatively associated with first-trimester fasting insulin and HOMA-IR, and some associations remained in the third trimester (Table 4). An inverse correlation was found between the "Long-chain (n-3) fats" subscore and  first-trimester fasting glycaemia, whereas a positive one was observed between the 'Sodium' subscore (scored inversely to the consumption) and first-trimester fasting glycaemia. However, after adjusting for ppBMI, the majority of the observed associations lost statistical significance. Few other associations emerged through covariate adjustment, as the 'SSB and fruit juices' subscore was then inversely associated with first-trimester insulin and HOMA-IR. Few other inverse associations and trends were observed later in pregnancy.

Predictors of first-trimester fasting insulin concentrations
Among the health-related characteristics included in the stepwise procedure, four of them were chosen to form the linear regression models of participants' fasting insulin concentrations throughout pregnancy ( Table 5). Because of the high collinearity between all adiposity variables (data not shown), three separate models were performed, including either ppBMI, firsttrimester SAT thickness, or VAT thickness, in combination with first-trimester AHEI total score, age, and parity. The final models revealed that a higher ppBMI, SAT and VAT thickness in early pregnancy were highly associated with first-trimester fasting insulin concentrations, with a small contribution from first-trimester diet quality (9.5%), but only in the model including VAT thickness. Altogether, the models explained 33 to 49% of first-trimester fasting insulin concentrations variance (Model 1, 49%; Model 2, 39%; Model 3, 33%), with pre-pregnancy or earlypregnancy adiposity measurements contributing to approximately 90% of the total percentage (Model 1, 96%; Model 2, 97%; Model 3, 88%). Moreover, neither age nor parity was a statistically significant predictor of first-trimester fasting insulin concentrations. Similar results were observed when examining predictors of second-and third-trimester fasting insulin concentrations (data not shown).

Discussion
Overall diet quality in early pregnancy is associated with adiposity measurements and glucose homeostasis during pregnancy. Pregnant women who consumed more fruit and less red or processed meat in the first trimester had significantly lower SAT thickness, whereas those who consume more whole-grain products had significantly lower VAT thickness. Higher consumption of fruit, whole-grain products, and long- Table 4. Associations between first-trimester diet quality subscores and first-trimester adiposity and trimester-specific glucose homeostasis measurements (n chain (n-3) fatty acids, as well as lower consumption of red or processed meat, was significantly associated with better glucose homeostasis measures in the first trimester. Nevertheless, because most associations lost significance and power with covariate adjustment, these associations could be partially influenced by ppBMI. However, independently of ppBMI, lower consumption of sweet beverages and fruit juices was significantly associated with better glucose homeostasis in the first trimester. These findings suggest that pregnant women who adhere to a healthier diet, have reduced adiposity in the first trimester. This is the first study to examine the associations between diet quality and ultrasoundestimated SAT and VAT thicknesses in pregnancy. Yet, a Brazilian group analysing the impact of modifiable predictors related to abnormal amounts of maternal visceral fat during pregnancy found that neither the quality nor quantity of food consumed during pregnancy were identified as visceral fat predictors (Bernardi et al. 2021). Few studies have evaluated the relationship between diet and other markers of adiposity, such as fat mass and fat-free mass. On one hand, a study investigating the association of maternal diet with body composition in a sample of 110 overweight and obese pregnant women demonstrated those with a higher diet quality score had more fatfree mass, compared to women with a lower score, both in early and late pregnancy (Pellonpera et al. 2019). On the other hand, an American research group recently concluded that the Diet Inflammatory Index score was not associated with fat mass during pregnancy (Lindsay et al. 2021). Although these results cannot be directly compared to those of pregnant populations, systematic reviews and meta-analyses conducted in the general population demonstrated that higher consumption of fruit and whole grains products, as well as lower consumption of meat, was associated with reduced risk of obesity (Schwingshackl et al. 2015;Albertson et al. 2016;DeClercq et al. 2017;Hebden et al. 2017), thus corroborating our findings.
Some associations were also observed between overall diet quality and glucose homeostasis during pregnancy, mainly between the consumption of fruit, whole grains, red/processed meat and fasting insulin concentrations, and insulin resistance. However, because only trends remained after adjustments for the covariate, our results suggest that those associations could be partially influenced by ppBMI. However, independently of ppBMI, a higher diet quality in early pregnancy was significantly associated with better glucose homeostasis throughout pregnancy. These results confirm those from other studies that demonstrated that higher diet quality during pregnancy was associated with lower maternal fasting glucose and insulin concentrations (Martin CL et al. 2016;Lindsay et al. 2020). Nevertheless, other groups have found contrasting results (Radesky et al. 2008;Sauder 2016;Siregar et al. 2020), which could be explained by the low number of dietary recalls per participant (one or two), dietary or glycemic measurements performed in the first or second trimester only, and the choice of dichotomous glycemic markers instead of continuous ones. Using multivariable linear regression analysis, we identified that pre-pregnancy and early-pregnancy adiposity measurements, (ppBMI, as well as first-trimester SAT and VAT thicknesses) are important predictors of fasting insulin concentrations throughout pregnancy. Those results underline the importance of a better understanding of the participant's fat distribution profile in the preconception period and the influence of the latter during pregnancy. According to the linear regression model, diet quality, as measured with the AHEI total score contributed modestly to the variability of first-trimester fasting insulin concentrations, within the model using VAT thickness as an adiposity variable. Those results are in line with past and current literature suggesting that ppBMI and ultrasoundestimated SAT and VAT thickness can predict insulin concentrations (Pontual et al. 2016), insulin resistance (De Souza et al. 2014;Basraon et al. 2016;Pontual et al. 2016) and the risks of developing GDM during pregnancy (Basraon et al. 2016;Budak et al. 2019;Alves et al. 2020;D'Ambrosi et al. 2020;Li et al. 2022).
Our results suggest that low diet quality could influence glucose metabolism during gestation, especially in early pregnancy. Knowing that a higher ppBMI is associated with lower diet quality during pregnancy (Savard et al. 2019), as proposed by others, pre-pregnancy dietary habits, as reflected by ppBMI or early-pregnancy abdominal adipose tissue thicknesses, could have a considerable impact on the development of glucose intolerance, insulin resistance during pregnancy (Radesky et al. 2008), supporting the loss of statistical significance in most of our associations with the adjustment for this covariate. Although this is only speculation and cannot be confirmed by our results, consuming a lower quality diet, could place one at a higher risk of consuming a less satiating diet and higher energy density foods, which may lead to an increased risk of positive energy balance. This excess of nutrients, before or in early pregnancy, could induce fat accumulation resulting in a higher ppBMI and thicker abdominal adipose tissues, which, in turn, may result in adipose tissue dysfunction and adiposity-related systemic inflammation and insulin resistance (Tchernof and Despres 2013;Martinez-Santibañez and Lumeng 2014). The herein proposed hypothetical effect of abdominal adiposity on the associations between diet quality and glucose homeostasis as well as inflammation has previously been demonstrated in the general population (Mazidi et al. 2018;Mazidi et al. 2021). In the gestational context, numerous publications have demonstrated strong associations between SAT and VAT thickness measurements and glucose metabolism with ultrasoundestimated abdominal adipose tissue thickness being a potential predictor of GDM Kansu-Celik et al. 2018;Alves et al. 2020;Rocha et al. 2020).
Still, it is to be noted that independently of ppBMI, some associations between first-trimester diet quality and metabolic measurements remained, mainly in the second and third trimesters of pregnancy. This could be explained by natural metabolic changes (namely the development of low-grade inflammation and insulin resistance due to hormonal changes) that occur during pregnancy. The influence of ppBMI on the glycemic profile could thus lose its power with the advancing of pregnancy when other physiologic factors may come into play. Hence, improving diet quality in early pregnancy and even before, may provide better metabolic health during pregnancy, which is in line with the current recommendations of the 2019 Canada Food Guide. On that note, we can think that interventions started early in pregnancy aiming to improve diet quality when pregnant women are motivated to make favourable changes in their eating habits (Poon et al. 2018), could benefit a large number of pregnant women, regardless of their weight status and adipose profile at conception. To the best of our knowledge, this is the first study to examine the associations between first-trimester diet quality and ultrasound-estimated SAT and VAT thicknesses. It is also one of the few studies to investigate the associations between diet quality and maternal glucose homeostasis markers, other than GDM risk or incidence. The current study is also unique since we recruited a pregnant population as soon as the tenth week of gestation, allowing us to measure both diet quality and metabolic markers early in pregnancy and to perform repeated measurements of the latter until delivery. The use of validated Web-based 24-h recalls generated detailed information on dietary intakes during pregnancy. Furthermore, the use of a validated diet quality score that has already been strongly associated with metabolic diseases in some pregnant cohorts adds to our results. Because of the small sample size and the high homogeneity of the participants recruited regarding their ethnicity and socioeconomic standing, this study will need to be replicated in other more diversified populations. In addition, it is to be noted that no information on gender identification was available, making it impossible to assume that all study subjects identify as women. Thus, the term "women" was used in this manuscript solely regarding the biological sex of participants and defined as one's ability to gestate and birth a child. Recognising that individuals who do not identify as female also have perinatal needs and experiences that may be similar to, but also unique from those of cisgender women (Moseson et al. 2020), the recruitment of participants of diverse gender identities should be considered in future research.
Our findings suggest that higher diet quality in the first trimester of pregnancy, characterised by higher consumption of fruits, whole grains, and long-chain fatty acids, as well as lower consumption of red or processed meat and sweetened beverages or fruit juices, is associated with more favourable metabolic markers in pregnancy. Those results appear to be dependent on BMI in our sample, which remains to be confirmed in larger populations. Nevertheless, this does not take away the importance of promoting, through intervention studies or public health measures, the importance of healthy eating in early pregnancy, which is consistent with Canadian recommendations.

Disclosure statement
A.T. receives funding from Johnson & Johnson, Medtronic, and GI Windows for research unrelated to the present study. A.T. has acted as consultant for Bausch Health, Novo Nordisk, and Biotwin. Other authors declare no conflict of interest.

Funding
The ANGE project is funded by the Danone Institute of Canada (grant number: FO115961) and by start-up funds from the Fonds de recherche du Qu ebec-Sant e and the Fondation du CHU de Qu ebec, Universit e Laval). The PAGG project is funded by the Fondation du CHU de Qu ebec, Universit e Laval, and the Institut sur la nutrition et les aliments fonctionnels (INAF), Universit e Laval. All funding allowed the collection, analysis, and interpretation of data, but played no role in the writing of this manuscript. E.B. received graduate scholarship awards from the Canadian Institutes of Health Research and the Fonds de recherche du Qu ebec-Sant e.

Data availability statement
The datasets generated and/or analysed during the present study are available from the corresponding author upon reasonable request.