Letter to the Editor: From glycemia to mood: Exploration of a postpartum depression prediction model in gestational diabetes

Abstract

This letter discusses a recent study by Wu JX and Wu FF which published in World Journal of Psychiatry that examined factors influencing postpartum depression (PPD) in patients with gestational diabetes mellitus (GDM), and constructed a risk prediction model. The study enrolled 204 GDM patients, dividing them into PPD (n = 52) and non-PPD (n = 152) groups based on Edinburgh Postnatal Depression Scale scores at 6 weeks postpartum. Multivariate logistic regression analysis revealed that elevated 2-hour postprandial glucose at diagnosis, poor glycemic control during pregnancy, postpartum mother-infant separation, low family care, and low social support were independent risk factors for PPD. The prediction model, defined as Logit (P) = 0.508 × 2-hour postprandial blood glucose + 0.687 × gestational blood glucose control + 1.092 × postpartum mother-infant separation + 0.745 × low family care + 0.289 × low social support - 4.766, demonstrated robust performance with an area under the curve of 0.840, sensitivity of 0.839, and specificity of 0.825. The significance of the study lies in identifying GDM-specific PPD risk factors and constructing a practical predictive tool for clinical use. It establishes a critical link between glycemic indicators and postpartum mental health, expanding our understanding of psychological issues in GDM patients. This model enables early identification of high-risk individuals, facilitating targeted interventions such as enhanced glycemic management, promotion of mother-infant contact, and mobilization of family and social support resources. By shifting from passive treatment to active prevention, this approach has important implications for improving maternal mental health outcomes, promoting mother-infant wellbeing, and reducing family and societal burden. Although this was a single-center study with a small sample, these findings provide a foundation for future multicenter research and the development of digital risk assessment tools and personalized intervention strategies.

Key Words: Gestational diabetes; Postpartum depression; Prediction model; Blood glucose; 2-hour postprandial glucose; Maternal mental health; Perinatal care

Core Tip: This study investigated key factors influencing postpartum depression (PPD) in women with gestational diabetes mellitus (GDM) and developed a risk prediction model. PPD in GDM results from metabolic, obstetric, and psychosocial factors. Independent risk factors included elevated 2-hour postprandial glucose, suboptimal glycemic control, mother-infant separation, low family care, and low perceived social support. A predictive model based on these five factors demonstrated good discrimination and calibration, offering quantitative early screening and stratified management. Findings underscore the importance of integrated biopsychosocial interventions in perinatal care. Future multicenter studies should externally validate the model and incorporate additional biomarkers and psychological measures.

TO THE EDITOR

We read with interest the article titled “Analysis of influencing factors of postpartum depression in patients with gestational diabetes mellitus and construction of prediction model” published in World Journal of Psychiatry, and we have prepared a flowchart of the study (Figure 1) to facilitate readers' understanding[1]. This cross-sectional study systematically explored factors associated with postpartum depression (PPD) in women with gestational diabetes mellitus (GDM) and developed a risk prediction model, offering valuable insights for the early identification of high-risk individuals in clinical settings. We commend the authors for their rigorous design and the clinical relevance of their findings. Nevertheless, after a detailed review, we believe several issues warrant further discussion to advance research in this field.

Figure 1 Flowchart of this study. GDM: Gestational diabetes mellitus; PPD: Postpartum depression.

The study included 2-hour postprandial glucose (2hPG) at GDM diagnosis as an independent risk factor in the predictive model, but did not incorporate fasting plasma glucose (FPG) or glycated hemoglobin (HbA1c). Accumulating evidence suggests that both FPG and HbA1c are associated with perinatal mood disorders[2-4]. The superior predictive value of 2hPG over FPG and HbA1c may be attributed to its association with postprandial glycemic fluctuations influencing neuroendocrine responses, or to psychological distress related to meal management. Future studies incorporating continuous glucose monitoring to assess glycemic variability parameters may help elucidate the relationship between glucose metabolism and postpartum mental health. Additionally, in clinical practice, GDM is frequently diagnosed based on elevated 2hPG, which may also contribute to its representativeness in the model.

Another noteworthy issue is the identification of postpartum mother-infant separation as a strong predictor, although the reasons for separation (e.g., neonatal complications, preterm birth, and maternal conditions) and its duration were not specified. Different causes and lengths of separation may exert heterogeneous effects on maternal psychological outcomes. Prolonged separation is associated with higher PPD risk[5]. Stratified analyses based on etiology and duration of separation would deepen our understanding and facilitate the development of targeted interventions.

Regarding the methodological aspects of model development, although the proposed model demonstrated satisfactory discrimination and calibration, its single-center design and small sample size (52 cases in the PPD group) may limit its generalizability. Internal validation was conducted using cross-validation; however, external validation across diverse populations is necessary before broader clinical implementation[6,7]. Furthermore, incorporating additional predictors, such as inflammatory cytokines or validated psychological assessment instruments (e.g., pregnancy-related anxiety scales), may improve model performance[8-11].

In summary, this study makes a meaningful contribution by identifying key risk factors for PPD in GDM patients and developing a clinically useful predictive model. We commend the authors for their rigorous work and hope our comments stimulate further research to validate, refine, and translate these findings into practice, ultimately improving maternal mental health outcomes.

Main findings and implications of this study

The primary findings and significance of this study lie in the comprehensive integration of clinical biochemical indicators, obstetric factors, and psychosocial determinants to construct a risk prediction model for PPD specifically tailored to patients with GDM. The results identify 2hPG at GDM diagnosis, glycemic control during pregnancy, postpartum mother-infant separation, family care, and social support as independent influencing factors for PPD. These findings carry multifaceted clinical and theoretical implications.

From a metabolic perspective, this study provides evidence supporting a significant association between glucose levels, particularly postprandial glycemic fluctuations, and perinatal mental health. Postprandial hyperglycemia reflects impaired pancreatic β-cell function and the degree of insulin resistance. It also contributes to increased susceptibility to depression through mechanisms involving oxidative stress, activation of inflammatory responses (e.g., release of proinflammatory cytokines such as interleukin-6 and tumor necrosis factor-α), and dysregulation of the hypothalamic-pituitary-adrenal, ultimately interfering with neurotransmitter metabolism (e.g., serotonin)[12,13]. These findings offer epidemiological support for the metabolism-emotion linkage in the perinatal context. It should be noted that current evidence focuses primarily on the associations of FPG levels and HbA1c with the risk of depression, whereas studies examining the predictive value of 2hPG for depressive risk remain scarce[14,15]. Therefore, from a metabolic perspective, whether 2hPG offers superior predictive utility for PPD compared with established glycemic indicators such as FPG and HbA1c warrants further investigation in future studies.

The study underscores the buffering and protective role of psychosocial factors in this specific population. Patients with GDM face the physiological changes of pregnancy as well as the psychological burden of strict dietary management, self-monitoring of blood glucose, and concerns regarding long-term fetal health—a dual burden that challenges psychological resilience. Adequate family care (e.g., emotional support and shared caregiving responsibilities) and high levels of social support (e.g., informational and instrumental support) may mitigate PPD risk by alleviating perceived stress, enhancing disease adaptation, and improving self-efficacy.

Most importantly, the development of this predictive model offers clinicians a practical and quantifiable tool for early risk assessment. By applying this model, obstetric healthcare providers can identify individuals at high risk for PPD during the early postpartum period (e.g., within 6 weeks after delivery) and implement stratified management strategies—such as increasing the frequency of postpartum visits, providing psychological counseling, or facilitating timely referral to mental health specialists. The satisfactory discrimination and calibration of the model underscore its clinical utility and support a paradigm shift in perinatal care from a traditional “treatment of established depression” approach toward a “prevention of potential risk” model. Ultimately, this may contribute to reducing the incidence of PPD among GDM patients, improving long-term maternal and neonatal outcomes, and alleviating the disease burden on families and society.

Strengths and limitations of the study

The strengths of this study are primarily reflected in the following aspects. Regarding the selection of the study population, the focus on GDM as a specific high-risk subgroup offered greater specificity and clinical relevance compared to previous generalized studies on perinatal depression. As a distinct population, patients with GDM experience postpartum mental health outcomes shaped by the interplay between metabolic and psychosocial factors. This study precisely identified the unique risk factors pertinent to this group, providing evidence-based support for developing stratified intervention strategies.

In terms of variable selection and analytical framework, this study transcended the limitations of a unidimensional approach by innovatively constructing a comprehensive model that integrates objective biochemical indicators (2hPG), obstetric clinical parameters (postpartum mother-infant separation), and modifiable psychosocial factors (family care and social support). This biopsychosocial integrated perspective significantly enhanced the predictive performance of the model and identified clear targets for subsequent intervention research. For instance, among postpartum women with low levels of family care, support systems could be strengthened by involving partners in childcare activities and implementing family health education. For those with low social support, linkage to community resources or the establishment of peer support groups may have been beneficial.

With respect to clinical translation and application, all predictive factors included in the model were either routinely collected in clinical practice (e.g., blood glucose values and delivery details) or were assessed using brief validated scales (e.g., Family Care Questionnaire, and Social Support Rating Scale), offering favorable feasibility and potential for widespread adoption. In future clinical implementation, this predictive equation could be further developed into a nomogram, a risk scoring card, or integrated as an automated calculation tool within electronic medical record systems, enabling obstetricians, midwives, and community healthcare providers to perform rapid risk assessments amidst busy clinical workflows.

However, although this study demonstrates a certain degree of methodological rigor, several limitations warrant discussion, which also provide directions for improvement in future research. Regarding the study design, this investigation used a cross-sectional survey, collecting data and assessing PPD only at 6 weeks postpartum. This approach inherently limits the ability to establish causal relationships and can only reveal associations between variables. For example, although the study identified an association between low family care and the occurrence of PPD, it is difficult to determine whether low family care contributed to the development of PPD, or whether depressive mood influenced postpartum women’s perception and evaluation of family care. Future research should adopt prospective cohort designs with longitudinal follow-up, to track the trajectory of psychological status in GDM patients from early pregnancy onward, with repeated measurements of key variables at multiple time points (e.g., second trimester, third trimester, 6 weeks postpartum, and 6 months postpartum).

Regarding sample representativeness, this study was conducted at a single center, with all participants recruited from one hospital and a small sample (only 52 cases in the PPD group), which may have introduced selection bias. Patients with GDM from different regions and hospital levels may vary in terms of diagnostic and treatment patterns, socioeconomic status, cultural backgrounds, and social support resources; factors that could modify the strength of associations between risk factors and PPD. The small number of positive events (PPD cases) constrained the number of predictors that could be included in the model, potentially omitting some important variables.

In addition, the confounding factors in this study were not sufficiently evaluated. For instance, although a diagnosis of antenatal depression was taken into account and used as an exclusion criterion, a history of other psychiatric conditions, such as previous anxiety disorders, and a family history of mental illness were not considered. Furthermore, factors directly related to the risk of PPD, including sleep quality and the degree of postpartum fatigue, were also not included in the analysis.

In summary, future research should conduct multicenter, large prospective cohort studies covering diverse geographical areas and healthcare facilities at various levels, with adequate assessment of confounding factors and rigorous sample size estimation during the design phase to ensure adequate statistical power for detecting clinically meaningful effect sizes. Additionally, the existing model should be validated in independent external populations to assess its discrimination and calibration across different populations.

In terms of variable measurement, there were some limitations regarding the assessment tools and operational definitions of some key variables. Both family care and social support were evaluated using self-report scales. Although these scales have been widely validated in Chinese populations, self-assessment methods are susceptible to influence from respondents’ current emotional states and social desirability bias, potentially introducing reporting bias. Future studies could incorporate multisource assessments combining observer ratings (e.g., spousal evaluations and family member interviews) or objective indicators to enhance the objectivity and comprehensiveness of measurement. Furthermore, glycemic control during pregnancy was simply dichotomized as “good control” vs “poor control”; a coarse categorization that may have masked gradient effects of glycemic levels. In reality, subtle differences in glycemic control (e.g., mild vs severe abnormalities) may exert differential impacts on psychological outcomes. Future research should collect continuous glycemic monitoring data, such as HbA1c and glycemic variability parameters (e.g., mean amplitude of glycemic excursions, and time in range), to more precisely characterize the dose-response relationship between glucose metabolism and postpartum mental health. The classification of glycemic management approaches during pregnancy (e.g., dietary control, oral hypoglycemic agents, or insulin therapy) was not further refined. Adherence to different interventions, their side effects, and impacts on quality of life may vary. Subsequent studies should explore the heterogeneous effects of different treatment strategies on maternal psychological outcomes in greater depth. Regarding the operational definition of the key variable “mother-infant separation”, the study simply defined it as “newborn admission to the neonatal unit after delivery”, without further distinguishing specific reasons for separation (e.g., preterm birth, neonatal asphyxia, infectious diseases, or congenital anomalies) or duration (e.g., length of hospitalization). Different causes and durations of mother-infant separation may have markedly different effects on maternal psychology. For instance, prolonged separation due to preterm birth vs short-term separation for brief observation would clearly differ in terms of psychological stress intensity and recovery trajectories. Future research could collect detailed clinical information on mother-infant separation, including reasons for separation, neonatal disease severity, separation duration, and visitation frequency, and conduct stratified analyses or construct composite indicators to more accurately assess the independent contribution of mother-infant separation to PPD.

Finally, regarding statistical modeling, although the study used multivariate logistic regression and conducted cross-validation, the model was developed based on the available data without utilizing more advanced variable selection methods (such as LASSO regression or random forest) to address potential multicollinearity issues. Sensitivity analyses to assess dependence of the model on specific variables or samples were also not performed. For example, we believe that among the indicators included in the regression analysis, family care and the family support component of social support may exhibit collinearity.

Although this study provides a valuable predictive model for the early identification of PPD among GDM patients, the aforementioned limitations suggest that caution is warranted when interpreting and applying the findings. Future research should pursue continuous optimization across multiple dimensions, including study design, measurement, analysis, and translational application, to advance this field toward higher quality and greater precision.

Conclusion

Patients with GDM exhibit a high incidence of PPD, which arises from the interplay of metabolic factors, obstetric conditions, and psychosocial determinants. In this study, a risk prediction model based on key indicators from these dimensions demonstrated satisfactory discrimination and calibration, with good stability confirmed through cross-validation. This model may serve as a quantitative reference for the early clinical identification of individuals at high risk for PPD within the GDM population. These findings suggest that perinatal management for GDM patients should focus on glycemic control and prioritize psychosocial support systems, with particular attention to those experiencing mother-infant separation. Future multicenter prospective studies are warranted to externally validate the model and to further explore the potential contributions of glycemic variability indices and biomarkers in risk prediction, thereby advancing the development of individualized prevention strategies.