Gestational diabetes mellitus (GDM) is defined as the alteration of the metabolism of hydrocarbons which meets established diagnostic criteria, it is first recognised during pregnancy and that it does not comply with the established diagnostic criteria for diabetes mellitus prior to gestation(1). The prevalence of GDM has increased in recent years, especially due to the advanced age of pregnant women, the sedentary lifestyle and the increasing incidence of overweight and obesity(Reference Solomon, Willett and Carey2,Reference Aune, Sen and Henriksen3) .
GDM is associated with several complications relevant to both the child (i.e. macrosomia, large for gestational age, prematurity, traumatic births and caesarean, metabolic alterations such as hypoglycaemia and hyperbilirubinemia, polyhydramnios, possibility of developing long-term alteration of the metabolism of hydrocarbons and metabolic syndrome) and the mother (i.e. pre-eclampsia and eclampsia, type 2 diabetes (DM2) and CVD)(Reference Metzger and Lowe4–Reference Carpenter15).
Well-established risk factors for developing GDM are advanced maternal age, overweight and obesity and excessive weight gain in the index pregnancy, physical inactivity, ethnicity, history of macrosomic babies (birth weight 4000 g or more) or previous unexplained perinatal loss(Reference Berger, Crane and Farine16,Reference Mirghani Dirar and Doupis17) or birth of a malformed infant(Reference Wu, Liu and Sun18), previous diagnosis of GDM or other alterations of glucose metabolism, multiple gestation, polycystic ovarian syndrome(Reference Yin, Falconer and Yin19), or family history of DM2(Reference Santos, Voerman and Amiano20–Reference Hod, Kapur and Sacks22). Moreover, genetic risk has also been described(Reference Popova, Klyushina and Vasilyeva23).
Diet has also been postulated as a modifiable risk factor for GDM(Reference Popova, Pustozero and Tkachuk24). Previous articles have reported preventive association between healthful dietary patterns, in particular the Mediterranean diet, and GDM risk(Reference Tobias, Zhang and Chavarro25–Reference Donazar-Ezcurra, Lopez-Del Burgo and Martinez-Gonzalez29). Other several nutritional factors have been found to increase the risk of developing GDM: high intake of animal fat(Reference Bowers, Tobias and Yeung30), high cholesterol and eggs consumption (≥7 eggs/week)(Reference Qiu, Frederick and Zhang31), high intake of heme iron(Reference Bowers, Yeung and Williams32) and frequent consumption of fried food, mainly out of home(Reference Bao, Tobias and Olsen33). Previous analysis within our cohort found a higher incidence of GDM associated with higher levels of red and processed meat consumption, Fe intake and high levels of fast-food consumption(Reference Marí-Sanchis, Díaz-Jurado and Basterra-Gortari34,Reference Dominguez, Martínez-González and Basterra-Gortari35) .
Important dietary factors potentially related to the risk of GDM are both carbohydrate quantity and quality(Reference Filardi, Panimolle and Crescioli36). Four previous studies found that a lower fibre intake, a higher glycaemic load, glycaemic index, or a diet low in carbohydrates but high in proteins and fats from animal products may increase the risk of developing gestational diabetes(Reference Zhang, Liu and Solomon37–Reference Bao, Bowers and Tobias40). However, none of these studies assessed a multidimensional index of carbohydrate quality. In this context, we aimed to analyse the overall quality of carbohydrates ingested before pregnancy, by means of a composite multidimensional index, previously used in other studies, considering simultaneously four characteristics: Glycaemic Index (GI), fibre content, solid or liquid form, and the degree of processing, namely the previously described carbohydrate quality index (CQI)(Reference Fernandez-Lazaro, Zazpe and Santiago41–Reference Santiago, Zazpe and Bes-Rastrollo45). We related this in association with GDM risk in a prospective cohort: the Seguimiento Universidad de Navarra (SUN) Project.
Subjects and methods
Study population
The SUN Project is a Spanish, prospective, multipurpose, dynamic cohort study composed of university graduates. The main objective was to investigate the association between diet, health conditions and multiple chronic diseases. The recruitment has been permanently opened since 1999. The design, follow-up methods and validated questionnaires have been formerly published(Reference Seguí-Gómez, de la Fuente and Vázquez46,Reference Martínez-González, Sánchez-Villegas and De Irala47) . A mailed questionnaire regarding dietary habits, lifestyles and health conditions is used to invite graduates to participate in the study, and a first voluntary response is assumed as informed consent to participate. Every 2 years, the information about diet, lifestyle, risk factors and medical conditions is updated with further follow-up questionnaires. The study protocol was performed as directed by the Declaration of Helsinki and approved by the Institutional Review Board of the University of Navarra. The SUN Project is registered at clinicaltrials.gov as NCT02669602. All questionnaires are available at https://www.unav.edu/en/web/departamento-de-medicina-preventiva-y-salud-publica/proyecto-sun/informacion-para-investigadores.
For the present analysis, we used the available database as of December 2019, accounting for 22 894 participants. The flow chart of participants is presented in Fig. 1. We excluded 341 subjects as they had been less than 2 years and 9 months in the cohort, to make sure that all the participants had been in the cohort time enough to complete at least the first follow-up questionnaire, and 8720 men. Up to the remaining 13 833 women, 12 589 women completed at least one follow-up questionnaire and 1244 were lost to follow-up. Therefore, the overall retention in the cohort was 91 %. Subsequently, other 8645 were excluded as they reported no pregnancies during follow-up. Women with a diagnosis of GDM previous to baseline were also excluded (n 18). Additionally, participants who reported a medical diagnosis of diabetes (both type 1 or 2) and/or they were being treated with insulin and/or oral antidiabetics at baseline were excluded (n 21). Women with implausible levels of total energy intake (lower than percentile 1 or higher than percentile 99) were also excluded in order to assure their correct fulfilment of the dietary information (n 78). Therefore, the final sample included 3827 pregnant women, with a total of 6869 incident pregnancies and 202 incident GDM cases.
Fig. 1. Flow chart of participants. GDM, gestational diabetes mellitus
Assessment of exposure: carbohydrate quality index
A semi-quantitative FFQ with 136 food items was used to assess the dietary exposures at baseline and after 10 years of follow-up(Reference Martin-Moreno, Boyle and Gorgojo48,Reference De La Fuente-Arrillaga, Vázquez Ruiz and Bes-Rastrollo49) . It has been repeatedly validated in Spain, and it has shown good validity for assessing carbohydrate intake(Reference Fernández-Ballart, Piñol and Zazpe50). There were nine options for the average frequency of consumption (never, 1 to 3 times/month, once/week, 2–4 times/week, 5–6/week, once daily, 2–3 times daily, 4–6 times daily, and 6 or more times daily). For this purpose, nutrient composition contained in the portion size specified for each food (typical portion size) was multiplied by the frequency of consumption.
Baseline dietary intake data were used to compute the CQI as proposed by Zazpe et al.(Reference Zazpe, Sánchez-Taínta and Santiago51). The CQI considers four criteria: (a) the GI (inversely weighted); (b) ratio of carbohydrates from whole grains to carbohydrates from total grains (whole grains, refined grains and its products); (c) dietary fibre intake (g/d) and (d) ratio of solid carbohydrates to total (solid and liquid). GI was derived for each food and beverage item in the FFQ using reference tables that used glucose as the reference food.
For each of these four components, participants were categorised into quintiles (score from 1 to 5, except for GI, that was weighted from 5 to 1), with higher values meaning better quality of each element. We added up the total score that ranged from 4 to 20. Finally, we categorised participants in three categories of the CQI: 4–9, 10–14 and 15–20 (see Supplementary Table 1).
Ascertainment of the outcome: gestational diabetes
Pregnant women who reported GDM diagnosis for the first time were asked to submit further information: medical reports, a previous diagnosis of diabetes, written confirmation of the date of the diagnosis, highest fasting glucose, glycated haemoglobin during pregnancy and oral glucose tolerance test as well as insulin use during the pregnancy. This information was verified by a medical team that was blinded to any exposure of the participants, and only the cases that this team confirmed were considered in the analysis.
There is no universal criterion for the diagnosis of GDM. Several protocols with different cut-off points are used for diagnosis after overload, which is also performed with different amounts of ingested glucose. An endocrinologist blinded to the information about dietary habits examined the information provided by women (additional questionnaires and medical reports) to confirm the diagnosis of incident gestational diabetes. The most common diagnosis criteria for GDM in Spain were those with 100 g oral glucose tolerance test with the cut-offs of Carpenter and Coustan or the cut-offs from the National Diabetes Data Group after a positive 50 g glucose challenge test.
Assessment of covariates
In the baseline questionnaire, other covariates were considered: social demographic variables (age), anthropometric measures (weight and BMI), health-related habits (physical activity, hours sitting down/d, smoking habits, total energy intake, adherence to Mediterranean diet, adherence to special diets, fast-food and snack consumption) and medical history (medication intake, parity and family history of DM2, CVD or hypertension).
A specific study with a subsample of this cohort pointed out a high validity of the self-reported weight and BMI(Reference Bes-Rastrollo, Pérez Valdivieso and Sánchez-Villegas52). Physical activity was evaluated according to a validated questionnaire and was measured in metabolic equivalent tasks (MET-h/week)(Reference Martínez-González, López-Fontana and Varo53). It was obtained by multiplying the duration of each activity in h/week by its characteristic energy expenditure. Adherence to the Mediterranean diet was computed using the nine-item score proposed by Trichopoulou et al.(Reference Trichopoulou, Costacou and Bamia54) but excluding cereals item to avoid overlapping with the main exposure, and alcohol intake, that was considered separately.
Statistical analysis
To describe baseline characteristics of the participants according to categories of CQI, we used percentages for categorical variables and means and standard deviations for continuous variables. To assess the relationship between CQI and the incidence of GDM, generalised estimating equations were fitted, assuming a binomial distribution, a logit link function, an exchangeable correlation matrix, and with a robust variance estimator, obtaining OR with 95 % CI. Each pregnancy was individually considered. For a woman who developed GDM during follow-up, no further pregnancies were considered thereafter. For those pregnancies occurring after more than 10 years of follow-up, exposure in the 10-year questionnaire was considered. Four models were fitted with increasing levels of adjustment: (a) unadjusted model; (b) adjusted for age at the first pregnancy at the cohort (continuous), BMI (continuous); (c) additionally adjusted for parity (0, 1, 2, 3 or more), family history of DM (yes/no), leisure-time physical activity (MET-h/week) (tertiles), television viewing (h/d) (continuous), smoking habit (never smokers, former smokers and current smokers), total energy intake (continuous), adherence to Mediterranean diet (three categories), fast-food consumption (three categories), snacking between meals (yes/no), following special diet at the baseline (yes/no); and (d) additionally adjusted for alcohol intake (abstainer, less than 10 g/d, 10–20 g/d and more than 20 g/d), CVD prior to enrolment (including CHD and stroke: yes or no) and hypertension prevalence (yes or no). Test for linear trend was performed using the median value of CQI for each category of CQI as a continuous variable in the model. The association between tertiles of individual components of the CQI and the incidence of GDM was also assessed. Several sensitivity analyses were performed to test the robustness of the results. Additionally, a joint analysis was conducted to jointly consider the quality (three categories of CQI) and the quantity of dietary carbohydrates (≤40 %, >40–<50 % and ≥50 %). Tests were two-sided, and P < 0·05 was considered statistically significant. The analyses were performed with STATA software version 16.1.
Results
Among 6869 incident pregnancies, 202 incident cases of new-onset GDM were identified (2·94 %). Baseline characteristics of subjects according to categories of baseline CQI are shown in Table 1. Participants in the lowest CQI category were younger and had their first pregnancy at a younger age, had a lower level of physical activity, and were more likely to be current smokers. Moreover, their total energy intake was lower, and they consumed lower quantities of total fat intake, MUFA, PUFA, SFA, and trans-fatty acid intake, and their adherence to the Mediterranean dietary pattern was lower.
Table 1. Baseline characteristics of participants according to baseline categories of carbohydrate quality index
(Mean values and standard deviations or percentages)
* Adherence to the Mediterranean dietary pattern was computed using the nine-item score proposed by Trichopoulou et al. (Reference Zazpe, Sánchez-Taínta and Santiago51) excluding cereals and alcohol intake.
In the fully adjusted model (Table 2), for the comparison between the highest CQI category v. the lowest, the OR was 0·67, suggesting an inverse association, but the 95 % CI was too wide to reach a definitive conclusion (95 % CI (0·40, 1·10)). Increasing adherence to the CQI was inversely associated with the incidence of GDM, but the results did not reach the conventional level for statistical significance: the P for trend was 0·102 and the OR (95 % CI) for a two-point increment in the CQI was 0·96 (95 % CI (0·86, 1·07)) in the fully adjusted model.
Table 2. Association between carbohydrate quality index and the incidence of gestational diabetes
(Odd ratio and 95 % confidence intervals)
Model 1: Adjusted for age at the first pregnancy at the cohort (continuous), BMI (continuous).
Model 2: Additionally adjusted for baseline parity (0, 1, 2, 3 or more), family history of DM (yes/no), leisure-time physical activity (metabolic equivalent-h/week, tertiles), television viewing (h/d, continuous), smoking habit (never smokers, former smokers and current smokers), total energy intake (continuous), adherence to Mediterranean diet (three categories), fast-food consumption (three categories), snacking between meals (yes/no) and following special diet at baseline (yes/no).
Model 3: Additionally adjusted for alcohol intake (0, >0 to 10 g/d, >10 to 20 g/d and >20 g/d), CVD (yes or no) and hypertension prevalence (yes or no).
Model 4: Excluding Mediterranean diet from model 3.
Table 3 shows the association between the incidence of GDM and each of the four components of the CQI. We found that none of the CQI components separately was significantly associated with a lower incidence of developing GDM in the fully adjusted model, additionally adjusted for all the other components. The lower OR was found for the lowest tertile of GI (0·80, 95 % CI (0·53, 1·20)), and the highest tertile of carbohydrates from whole grain to carbohydrates from total grain (0·81, 95 % CI (0·57, 1·13)).
Table 3. OR and 95 % CI of incident gestational diabetes mellitus by tertiles of glycaemic index, fibre intake, ratio of solid carbohydrates/total carbohydrates and ratio of carbohydrates from whole grains to carbohydrates from total grains
(Odd ratio and 95 % confidence intervals)
CQI, carbohydrate quality index.
Model 3: Adjusted for age at the first pregnancy at the cohort (continuous), BMI (continuous), baseline parity (0, 1, 2, 3 or more), family history of DM (yes/no), leisure-time physical activity (metabolic equivalent-h/week, tertiles), television viewing (h/d, continuous), smoking habit (never smokers, former smokers and current smokers), total energy intake (continuous), adherence to Mediterranean diet (three categories), fast-food consumption (three categories), snacking between meals (yes/no), following special diet at baseline (yes/no), alcohol intake (0, >0 to 10 g/d, >10–20 g/d and >20 g/d), CVD (yes or no) and hypertension prevalence (yes or no).
Model 4: Additionally adjusted for the other three components of the CQI.
Table 4 shows the results of the joint analysis of the quality and quantity of carbohydrates and incidence of GDM. Whereas there was no statistically significant interaction between both factors (P = 0·62), those participants with the highest CQI and with daily carbohydrate amounts > 50 % of total energy intake exhibited a significantly lower incidence of GDM: OR 0·29, 95 % CI (0·09, 0·89) as compared with those with the lowest quality and quantity (<40 %).
Table 4. Association between carbohydrate quality and quantity and the incidence of gestational diabetes
(Odd ratio and 95 % confidence intervals)
Adjusted for age at the first pregnancy at the cohort (continuous), BMI (continuous), baseline parity (0, 1, 2, 3 or more), family history of DM (yes/no), leisure-time physical activity (metabolic equivalent-h/week, tertiles), television viewing (h/d, continuous), smoking habit (never smokers, former smokers and current smokers), total energy intake (continuous), adherence to Mediterranean diet (three categories), fast-food consumption (three categories), snacking between meals (yes/no), following special diet at the baseline (yes/no), alcohol intake (0, >0 to 10 g/d, >10–20 g/d and >20 g/d), CVD (yes or no) and hypertension prevalence (yes or no).
P for interaction = 0·62.
Table 5 shows the results of the sensitivity analyses. Compared with the main analysis, results of the OR were not substantially different in any of the analysed scenarios, although in some cases the 95 % CI was narrower and more suggestive of an inverse association.
Table 5. Sensitivity analyses
(Odd ratio and 95 % confidence intervals)
Adjusted for age at the first pregnancy at the cohort (continuous), BMI (continuous), baseline parity (0, 1, 2, 3 o more), family history of DM (yes/no), leisure-time physical activity (metabolic equivalent-h/week, tertiles), television viewing (h/d, continuous), smoking habit (never smokers, former smokers and current smokers), total energy intake (continuous), adherence to Mediterranean diet (three categories), fast-food consumption (three categories), snacking between meals (yes/no), following special diet at the baseline (yes/no), alcohol intake (0, >0 to 10 g/d, >10–20 g/d and >20 g/d), CVD (yes or no) and hypertension prevalence (yes or no).
Discussion
In this prospective cohort study, pre-conceptional global quality of carbohydrate was not significantly associated with the incidence of GDM. We found an association between the incidence of GDM and a high-quality and high-quantity pre-conceptional dietary carbohydrates. Although most results were inconclusive, based on the CI and not only on the P-value(Reference Wasserstein and Lazar55,Reference Amrhein, Greenland and Mcshane56) , results of this study suggest that further studies should jointly consider the quality and the quantity of dietary carbohydrates, as the quality might be of importance, in particular in women with a higher intake of carbohydrates.
Given the complications with which GDM is related for both the mother and the fetus, the ideal is to prevent its development by previously acting on modifiable risk factors such as diet.
In GDM, there is a situation of insulin resistance with dysfunctional β-cells, pathophysiological situation very similar to that which occurs in DM2. The relationship between the intake of high-quality carbohydrates with a lower incidence of DM2 has already been reported(Reference Ley, Hamdy and Mohan57). In fact, the main scientific societies propose some similar recommendations in the dietary management of both DM2 and GDM(Reference Cheng and You58,Reference Mustad, Huynh and López-Pedrosa59) valuing the quality of ingested carbohydrates more than a fixed distribution of macronutrients that, by restricting the amount of carbohydrates in the diet, could lead to a compensatory increase in the percentage of ingested fat and thus insulin resistance(Reference Ros Pérez and Medina-Gómez60).
The Nurses’ Health Study II found that a low-carbohydrate dietary pattern with energy replacement from protein and fat from animal food sources may lead to an increased risk of GDM(Reference Bao, Bowers and Tobias40,Reference Bao, Bowers and Tobias61) . The Australian Longitudinal Study on Women’s Health (ALSWH) showed that habitual diet low in carbohydrates and high in fat and protein was associated with a higher incidence of GDM(Reference Looman, Schoenaker and Soedamah-Muthu39).
Concerning the carbohydrate quality, dietary total fibre intake, and in particular fruit fibre, seems to reduce the risk of GDM(Reference Zhang, Liu and Solomon37,Reference Looman, Schoenaker and Soedamah-Muthu39) . A high glycaemic load diet might increase the risk of GDM(Reference Zhang, Liu and Solomon37,Reference Looman, Schoenaker and Soedamah-Muthu39) . The value of dietary glycaemic load in terms of postprandial glycaemic response has been described in both healthy subjects(Reference Bao, Atkinson and Petocz62) and overweight patients with DM2(Reference Fabricatore, Ebbeling and Wadden63). However, in women with GDM, the results have been inconclusive(Reference Pustozerov, Tkachuk and Vasukova64,Reference Pustozerov, Tkachuk and Vasukova65) .
Also a higher consumption of French fries might increase the risk of GDM, since it is mainly a food with a high glycaemic index and that after the frying process it will contain advanced glycosylation products that will favour the development of insulin resistance(Reference Bao, Tobias and Hu66). Moreover, regular consumption of sugar-sweetened cola (which provides with considerable rapidly absorbable sugars) had been associated with a higher incidence of GDM(Reference Chen, Hu and Yeung67,Reference Donazar-Ezcurra, Lopez-del Burgo and Martinez-Gonzalez68) .
Considering the whole dietary pattern, several cohorts have found a lower incidence of GDM with greater adherence to Mediterranean diet(Reference Tobias, Zhang and Chavarro25–Reference Olmedo-Requena, Gómez-Fernández and Amezcua-Prieto28,Reference Karamanos, Thanopoulou and Anastasiou69–Reference Hassani Zadeh, Boffetta and Hosseinzadeh71) . Carbohydrates in the Mediterranean diet are mainly from whole grain, avoiding the consumption of those with rapid absorption and high GI like the sugar-sweetened drinks; it is rich in fibre, with the high consumption of fruits and vegetables, legumes, and nuts. Other dietary patterns such as the DASH (Dietary Approaches to Stop Hypertension), and the AHEI (Alternate Healthy Eating Index) have also been associated with a lower incidence of GDM(Reference Tobias, Zhang and Chavarro25).
Regarding intervention studies before or during pregnancy, the results are inconsistent(Reference Donazar-Ezcurra, López-del Burgo and Bes-Rastrollo72) or inconclusive(Reference Griffith, Alsweiler and Moore73,Reference Tieu, Shepherd and Middleton74) . Some interventions during pregnancy on diet and physical activity achieved lower incidences of GDM(Reference Koivusalo, Rönö and Klemetti75–Reference Bruno, Petrella and Bertarini77). Moreover, interventions during pregnancy to increase adherence to Mediterranean diet also reduced the incidence of GDM(Reference Al Wattar, Dodds and Placzek78,Reference Assaf-Balut, García de la Torre and Durán79) . Further randomised trials are needed.
Regarding biological plausibility, several potential mechanisms have been described in the literature. Dietary fibre can delay gastric emptying by reducing appetite and slowing and decreasing glucose absorption and insulin response(Reference Basu, Feng and Planinic80,Reference Chiattarella, Lombardo and Morlando81) . It can lead to a lower total daily energy intake which in turn will reduce fat and improve insulin sensitivity. In addition, it can reduce inflammation, aid in lipid control, and have antioxidant effects.
Data for whole grains alone are limited due to disparities in the definition of whole food in previous epidemiological studies. The American Nutrition Society supports the consumption of foods rich in fibre from cereals and whole grains to reduce the risk of DM2, obesity and CVD(Reference Cho, Qi and Fahey82).
The liquid form of carbohydrates favours faster absorption, less satiety, greater total energy intake and higher postprandial blood glucose elevation, with respect to carbohydrates ingested in solid form. Increased sugar consumption can also lead to deterioration of pancreatic β-cells, possibly due to the accumulation of reactive oxygen species(Reference Pan and Hu83).
Finally, the high glycaemic index of food implies a faster and greater increase in blood glucose and insulin levels after ingestion. This increase in insulin stimulates the absorption of nutrients by the cell and lipogenesis and inhibits the production of glucose in the liver(Reference Augustin, Kendall and Jenkins84).
The correlation between fibre and whole-grain intake was not high in our study (0·263). Therefore, we assumed that we were not evaluating the same question twice, but that fibre intake will include that of other foods beyond those contained in whole grains.
Our study has some limitations. For the evaluation of the diet, a self-reported FFQ was used. In our study, participants with higher BMI (and consequently higher risk of GDM) might be more likely to follow an energy-restricted diet and also to under-report their intake of unhealthy food. However, the FFQ is the most reliable and valid tool to assess the diet in large epidemiological studies. It has been previously evaluated with good correlation between the data obtained from the FFQ and those from dietary records(Reference Martin-Moreno, Boyle and Gorgojo48–Reference Fernández-Ballart, Piñol and Zazpe50). On the other hand, a particular variable to capture dieting was added as a covariate in the model, so potential differences or misreporting due to this factor were controlled for in the models.
As we have already indicated, the FFQ questionnaire was carried out at the beginning of the study and at 10 years of follow-up, but not during pregnancy. It is possible that by knowing their pregnancy status, participants might have modified their eating habits. However, the dietary pattern maintained prior to pregnancy seems to have a stronger influence on the development of GDM than the temporary changes in intake during pregnancy, which tend to affect specific foods more than the dietary pattern itself(Reference Zhang, Rawal and Chong85–Reference Gresham, Collins and Mishra87).
We have used a pre-defined index that assesses the quality of carbohydrates in a wide manner(Reference Fernandez-Lazaro, Zazpe and Santiago41–Reference Santiago, Zazpe and Bes-Rastrollo45). However, it cannot encompass all aspects related to the intake of these macronutrients. In addition, there is the controversy regarding the definition of a whole-grain or carbohydrate-restricted diet, which makes comparison between studies difficult.
Another limitation is that we do not have the sufficient power to compare the effect of the CQI on each of the subgroups that would be created according to the severity of GDM.
In addition, women diagnosed with GDM with several of the different criteria accepted for it were included as incident cases. We also do not have data on weight gain during pregnancy, since the questionnaires were completed every 2 years.
In the study, we adjusted for multiple confounding factors; however, we cannot totally rule out the possibility of residual confounding.
We only included women with a first diagnosis of GDM, without accounting for recurrence in subsequent pregnancies. Those women could have changed their diet after the first GDM diagnosis and on the other hand, they will also be more likely to have GDM recurrence. To avoid that nulliparous women were overrepresented in the GDM group, we adjusted for parity in the multivariate analysis and also carried out sensitivity analyses restricting nulliparous women, obtaining results similar to the main analysis.
On the other hand, our cohort only included women with a high level of education. This limits the representativeness of the sample, but at the same time it allows to avoid several confounding factors with the use of restriction. Our results acquire relevance given the explanation by the possible pathophysiological mechanisms described and not so much by their representativeness or external validity. The characteristics of this population with a high educational level could also condition that the women included had in general a better quality of the diet due to knowledge of the negative effects of the ingestion of lower quality carbohydrates, and therefore there would be fewer differences between them, which could justify our results. It is possible that we did not find a significant reduction in risk due to lack of statistical power as the number of incident cases was low.
However, our study has a number of strengths: it is a cohort of prospective design with large numbers of participants followed for a long time, in which it has been possible to adjust for multiple confounding factors. These are participants with a high level of education, which allows greater precision of their self-reported data and better retention in the cohort (overall retention 91 %). Moreover, our medical team confirmed the diagnoses of the GDM cases reported.
Regarding age, a prominent risk factor for GDM(Reference Li, Ren and He88), we looked for the best adjustment option given the characteristics of our cohort. For this, we used the age of the first pregnancy in the cohort, therefore subsequent to the collection of data on dietary exposure. On the other hand, if we had used the age in the last pregnancy, we could create an inverse artificial relationship with the risk of gestational diabetes. In addition, we carried out several sensitivity analyses, adjusting for the age of entry into the cohort, the age at each pregnancy when doing a repeated-measures study, and the combined adjustment for the age at entry into the cohort and the age at each pregnancy. In each of the assumptions, the results of the analysis were very similar, so we assumed the option of adjusting for the age of the first pregnancy at follow-up, since it would be the one that would have the most relationship with baseline exposure.
In conclusion, in this Mediterranean prospective cohort, quality of dietary carbohydrates was not significantly associated with the incidence of GDM. Results suggest that a higher quality of dietary carbohydrates might be associated with lower incidence of GDM, in particular in women with a higher intake of carbohydrates. Future studies should jointly consider both the quality and the quantity of dietary carbohydrates. However, these studies must overcome the limitations of the current study, with a larger sample size and a dietary assessment closer to conception, to confirm the effect of the quality and quantity of dietary carbohydrates on the development of GDM.
Acknowledgements
The authors thank other members of the SUN Group: Aguilera-Buenosvinos I, Alonso A, Álvarez-Álvarez I, Balaguer A, Barbería-Latasa M, Barrio-López MT, Battezzati A, Bazal P, Benito S, Bertoli S, Beunza JJ, Buil-Cosiales P, Carlos S, de Irala J, de la Fuente-Arrillaga C, de la O V, de la Rosa PA, Delgado-Rodríguez M, Díaz-Gutiérrez J, Díez Espino J, Domínguez L, Donat-Vargas C, Donazar M, Eguaras S, Fernández-Lázaro CI, Fernández-Montero A, Fresán U, Galbete C, García-Arellano A, Gardeazábal I, Gutiérrez-Bedmar M, Gomes-Domingos AL, Gómez-Donoso C, Gómez-Gracia E, Goñi E, Goñi L, Guillén F, Hernández-Hernández A, Hershey MS, Hidalgo-Santamaría M, Hu E, Lahortiga F, Llavero M, Llorca J, López del Burgo C, Marí A, Martí A, Martín-Calvo N, Martín-Moreno JM, Martínez JA, Mendonça R, Menéndez C, Molendijk M, Molero P, Muñoz M, Navarro AM, Pano O, Pérez de Ciriza P, Pérez-Cornago A, Pérez de Rojas J, Pimenta AM, Ramallal R, Razquin C, Rico-Campà A, Romanos-Nanclares A, Ruiz L, Ruiz-Canela M, San Julián B, Sánchez D, Sánchez-Bayona R, Sánchez-Tainta A, Sánchez-Villegas A, Sayón-Orea C, Toledo E, Vázquez Z, Zazpe I.
The SUN Project has received funding from the Spanish Government-Instituto de Salud Carlos III, the European Regional Development Fund (FEDER) (RD 06/0045, CIBER-OBN, Grants PI10/02658, PI10/02293, PI13/00615, PI14/01668, PI14/01798, PI14/01764, PI17/01795, PI20/00564 and G03/140), the Navarra Regional Government (27/2011, 45/2011, 122/2014), Delegación del Gobierno para el Plan Nacional Sobre Drogas (2020|021) and the University of Navarra.
Conceptualization, M.B.R, S.S., A.G. and M.A.M.G.; methodology, M.B.R., S.S., F.J.B.G., A.G. and M.A.M.G.; formal analysis, E.F.G, D.S.E, A.G.; resources, M.B.R., M.A.M.G.; data curation, A.G., F.J.B.G; writing—Original draft preparation, E.F.G.; writing—Review and editing, M.A.M.G., M.B.R., D.S.E., F.J.B.G., S.S., A.G.; supervision M.A.M.B, M.B.R, A.G.; funding acquisition, M.B.R., M.A.M.G., A.G. All authors have read and agreed to the published version of the manuscript.
There are no conflicts of interest.
Open access
