The association of dietary total antioxidant capacity and gestational diabetes: a prospective cohort study from the Mothers and their children’s health (MATCH)

In this cohort study, we investigated the association between DTAC and GDM using propensity score–based inverse probability weighting (IPW). We found that a diet rich in TAC was associated with a reduction in the risk of GDM, and this effect was clinically significant. In fully adjusted models, the adjusted RRs (95% CIs) for GDM from the lowest to highest quartiles of the DTAC score were 1 (reference), 0.32 (95% CI: 0.14, 0.73), 0.26 (95% CI: 0.11, 0.60), and 0.29 (95% CI: 0.12, 0.68), respectively (p for trend< 0.001). Overall, we observed a downward trend in GDM with higher DTAC among participants. This trend remained consistent even after adjusting for confounders, indicating that the protective association of DTAC with GDM is independent of body mass index (BMI) (kg/m2), occupation, age, hypertension, diabetes, education, and working rotating shift (aRR: 0.29, 95% CI: 0.12, 0.68). These findings strongly support the hypothesis that there is an association between DTAC and GDM using propensity score–based inverse probability weighting (IPW). Consistent with our study, another case-control study found that a significant association between higher FRAP and a reduced risk of GDM, with an 85% lower risk in the highest FRAP tertile (OR: 0.15; 95% CI: 0.08–0.29). However, they did not find significant associations for TRAP (OR: 1.62; 95% CI: 0.94–2.79) or TEAC (OR: 1.56; 95% CI: 0.89–2.72) [29]. Our study is a prospective cohort study that included 1856 pregnant women in their first trimester from the MATCH cohort. In contrast, Daneshzad et al. conducted a hospital-based case-control study with 463 pregnant women, comprising 263 healthy individuals and 200 with GDM. This difference in study design and population size is significant, with our cohort being larger and longitudinal, potentially providing a more comprehensive overview of dietary impacts over time. In comparing our prospective cohort study with the case-control study conducted by Daneshzad et al., several strengths of the cohort study design become evident. A key strength of cohort studies is the ability to establish a clear temporal relationship between exposure and outcome [30]. In our study, DTAC was assessed during early pregnancy, before the development of GDM. This prospective design ensures that the exposure (DTAC) precedes the outcome (GDM), which strengthens the argument for a potential causal relationship. Cohort studies allow for more comprehensive data collection on a wide range of potential confounders. In our study, detailed information on prepregnancy dietary intake, as well as other lifestyle and demographic factors, was gathered prospectively. This allows for better adjustment and control of confounding variables in the analysis, which enhances the validity of the findings [31]. While case-control studies can adjust for confounders, they are generally more limited in the data they can collect retrospectively. Cohort studies enable direct calculation of RRs, providing a more intuitive measure of the association between exposure and outcome [32]. In our study, we calculated the incidence of GDM and the RR reduction associated with higher DTAC scores. Case-control studies, on the other hand, typically provide odds ratios, which can be less straightforward to interpret, especially when the outcome is not rare. Odds ratios tend to overestimate the strength of the association compared to RRs, particularly when the outcome of interest is not rare. This exaggeration occurs because the odds of an event are inherently more extreme than the probability of an event [33]. When case-control studies are necessary and ORs are used, it is important to acknowledge the potential for exaggeration and, where feasible, to provide context by estimating the corresponding RR. This approach helps ensure that the reported associations are interpreted correctly and applied appropriately in public health and clinical settings [33, 34].

Several studies have investigated the association between DTAC and the glycemic index in prediabetes and diabetic individuals. Sotoudeh et al. conducted a case-control study to investigate the association between DTAC and prediabetes morbidity. The study included 300 participants from a Diabetes Screening Center in Iran, comprising 150 individuals with prediabetes and 150 controls. The findings indicated that higher DTAC was associated with a lower likelihood of prediabetes, with participants in the highest quartile of DTAC having an 82% reduced odds of experiencing prediabetes compared to those in the lowest quartile (OR: 0.18, 95% CI: 0.07, 0.49) [35]. Cyuńczyk et al. conducted a study to investigate the relationship between DTAC and the occurrence of prediabetes, diabetes, and insulin resistance in the Bialystok PLUS (Polish Longitudinal University Study) population. They assessed daily food consumption using 3-days 24-hour dietary recalls and calculated DTAC using the FRAP method. The study included measurements such as fasting glucose, 2-hour postprandial glucose, fasting insulin, and glycated hemoglobin to identify prediabetes, diabetes, and insulin resistance (Homeostatic Model Assessment for Insulin Resistance) (HOMA-IR). Their findings, analyzed using logistic regression and multivariate linear regression models, indicated that higher quartiles of DTAC were significantly associated with reduced odds ratios for the prevalence of prediabetes in individuals aged 35-65 years from the Bialystok PLUS population. DTAC was also inversely associated with HOMA-IR, suggesting a beneficial role in reducing insulin resistance. Additionally, DTAC showed positive associations with individual dietary antioxidants such as polyphenols, antioxidant vitamins, and minerals [36].

Li’s study analyzed data from 12,467 participants enrolled in the Natural Population Cohort of Northwest China: Ningxia Project to explore the associations of dietary antioxidant quality scores (DAQS), DTAC, and T2DM risk. Dietary intake was assessed using a validated semi-quantitative food frequency questionnaire. DAQS were calculated based on intake levels of vitamins A, C, and E, zinc (Zn), and selenium (Se), while DTAC was estimated using the ferric-reducing ability of plasma assay. Among the participants, 1,238 (9.9%) were diagnosed with T2DM. After adjusting for confounding factors, higher DAQS were associated with a reduced risk of T2DM, particularly in the highest tertile for vitamins A, E, and Se. Specifically, compared to the lowest tertile, the ORs for T2DM were 0.78 (95% CI 0.67–0.91, P-trend = 0.008) for vitamin A, 1.34 (95% CI 1.15–1.56, P-trend < 0.001) for vitamin E, 0.83 (95% CI 0.71–0.97, P-trend = 0.007) for Se, and 0.86 (95% CI 0.74–1.01, P-trend = 0.033) for Zn in the highest tertile [37].

The mechanism of DTAC in reducing GDM can be categorized into two groups. The first mechanism is through controlling blood sugar which can act in two different ways. First, antioxidants like flavonoid and polyphenols in the diet can disturb carbohydrate digestion by inhibiting alpha-amylase in the mouth and glucosidase in the gut, reducing postprandial glucose. The second way is by the binding ability of polyphenols to the GLUT2 protein, which take up glucose from the blood and transfers it to hepatocytes for glycolysis and gluconeogenesis, thereby facilitating a lower glycemic effect. Additionally, thioredoxin reductase-1, glutathione peroxidases, heme oxigenase-1, and glutathione-S-transferases are involved in the antioxidant-response elements controlled by polyphenols when they activate the nuclear factor-2 erythroid related factor-2 signaling pathway [38]. Due to the lack of enzymatic antioxidants, pancreatic beta cells would be susceptible to the detrimental impact of oxidative stress. This condition impairs the mitochondria, reduces insulin secretion, and elevates glucose levels in the bloodstream [39, 40]. Insufficient antioxidant defense and the increase in reactive oxygen species lead to damage to organelles and enzymes, intensification of protein and lipid peroxidation, and development of insulin resistance. [41, 42]. Hence, it is crucial to consume adequate amounts of antioxidants to maintain glucose homeostasis [29].

There are several strengths in the current study. First, our data are from a prospective cohort study and food intake was assessed with a validated FFQ. Second, we analyzed the data considering known confounders, including BMI (kg/m²), occupation, age, hypertension, diabetes, education, and working rotating shift of the participants which allowed analysis of the relationship between DTAC and GDM. DAGs are a powerful tool for identifying and managing confounding variables in epidemiological research. In our study, we employed DAGs to discern the minimal sufficient set of confounders necessary to estimate the association between DTAC and the risk of GDM. This methodological choice offers several advantages over traditional statistical approaches. DAGs are graphical representations that illustrate the association between variables [43, 44]. Each node represents a variable, and directed edges (arrows) indicate the direction of causality. The acyclic nature of these graphs ensures that there are no feedback loops, thereby facilitating a clear depiction of the temporal sequence of events and dependencies among variables. By mapping out these relationships, DAGs help researchers identify which variables need to be controlled for to obtain an unbiased estimate of the causal effect of an exposure on an outcome. One of the primary strengths of using DAGs is their ability to visually and analytically identify confounders—variables that are related to both the exposure (DTAC) and the outcome (GDM) and that, if not properly controlled, can bias the estimated effect. Traditional statistical methods often rely on automated variable selection procedures or researcher judgment, which can lead to either over-adjustment (controlling for variables that are not true confounders) or under-adjustment (failing to control for necessary confounders). Over-adjustment can reduce statistical power and introduce bias, while under-adjustment can leave residual confounding [45].

DAGs provide a structured approach to confounder identification by explicitly modeling the causal pathways and allowing researchers to determine the minimal sufficient adjustment set. This set includes only those variables that block all back-door paths (non-causal paths) from the exposure to the outcome, thus ensuring that the estimated effect is not confounded. In our study, we used the web tool dagitty.net to construct and analyze the DAG, which facilitated the identification of the minimal sufficient set of variables for adjustment [46, 47]. Our study also has some limitation. One limitation of our study is the potential for selection bias due to the utilization of a specialized gynecological hospital in eastern Tehran. While this approach ensured that our sample comprised pregnant women who were at risk for the outcomes of interest, it may limit the generalizability of our findings to the broader population of pregnant women. Selecting participants from a single specialized hospital could introduce a degree of selection bias, as the patient population might differ from those receiving care in other settings or geographic locations.

However, we intentionally chose a population at risk to ensure that our study outcomes, such as GDM, were relevant to the sample. If we had selected our sample from the general community, we risked including individuals who were not pregnant or did not intend to become pregnant, thereby compromising the applicability of our findings. Additionally, our inclusion and exclusion criteria were not overly restrictive, allowing for a diverse sample that enhances the generalizability of our results. Despite this limitation, we believe our findings are applicable to a broad population of pregnant women. Future studies should consider including multiple centers and diverse geographic locations to further enhance the generalizability and external validity of the results. Another notable limitation of our study is the absence of data on insulin resistance indices, which precludes a comprehensive investigation into the effects of DTAC on insulin levels. Insulin resistance is a well-established major risk factor for the development of type 2 diabetes and GDM. Understanding the association between DTAC and insulin resistance could provide deeper insights into the potential protective mechanisms of dietary antioxidants. However, due to the lack of direct measurements of insulin resistance, such as HOMA-IR or other relevant biomarkers, we are unable to evaluate how DTAC influences insulin sensitivity and beta-cell function. Future research should include these indices to elucidate the pathways through which dietary antioxidants may impact glucose metabolism and insulin action. Despite this limitation, our findings contribute valuable knowledge to the field, suggesting an inverse association between DTAC and the risk of GDM. Nonetheless, further studies incorporating insulin resistance measurements are essential to fully understand the underlying mechanisms and to validate our results.

Performance bias, or information bias, can occur when knowledge of the subject’s exposure status influences the ascertainment of outcomes, such as gestational diabetes mellitus GDM. This bias can potentially skew study results if not addressed effectively. In our study, we implemented comprehensive quality assurance and quality control measures to mitigate the impact of performance bias. These measures included thorough participant-tracking techniques from the outset of the study, ensuring that contact information was up-to-date and accessible. To maintain data integrity, we regularly monitored data quality, with weekly checks for interviewer bias and monthly independent reviews of a subset of data. Our interviewers underwent rigorous training sessions and adhered to standardized protocols outlined in a training handbook, promoting consistency in data collection procedures. These efforts were crucial in minimizing the risk of bias by ensuring that the ascertainment of GDM was conducted objectively and uniformly across all study participants. By detailing these measures in our published protocol, we aimed to enhance transparency and demonstrate the reliability of our study findings despite the inherent challenges of information bias in observational research. These quality assurance and control techniques underscore our commitment to producing robust and credible evidence in assessing the association between DTAC and GDM risk.