Association Between Household Income-Related Education and Postpartum Depression: A Two-Step Mendelian Randomization Study

Abstract

Background:

Previous research has demonstrated the associations between educational attainment, economic status, and postpartum depression (PPD). However, the associations between educational attainment, household income, and PPD based on Mendelian randomization (MR) have yet to be fully elucidated. Moreover, whether household income serves as a mediator in the association between education and PPD remains unclear.

Methods:

Using single-nucleotide polymorphisms (SNPs) extracted from genome-wide association studies (GWAS) as instrumental variables (IVs), we assessed the associations between education, household income, and PPD using MR analysis methods, such as inverse-variance weighting (IVW). The mediating effect was evaluated by examining (1) the association between education and household income and (2) the role of household income in the education—PPD relationship. Sensitivity analyses, including MR-Egger intercept analysis, leave-one-out analysis, Cochran’s Q test, and MR Pleiotropy RESidual Sum and Outlier (MR-PRESSO), were performed to assess robustness.

Results:

The MR analysis indicated that higher education (odds ratio (OR) = 0.877, 95% confidence interval (CI): 0.828–0.930; p = 1.050 × 10^-5) and greater household income (OR = 0.443, 95% CI: 0.204–0.962; p = 0.040) were both associated with a reduced risk of PPD. Furthermore, we identified a significant association between education and household income (OR = 1.095, 95% CI: 1.085–1.105; p = 2.805 × 10^-87). The mediation analysis demonstrated that household income partially mediated the relationship between education and PPD, with an indirect effect of –0.074 (95% CI: –0.146 to –0.004), accounting for 56.67% of the total effect.

Conclusions:

These findings suggest that household income significantly mediates the association between education and PPD. Moreover, higher education and increased household income can serve as important protective factors against PPD, underscoring the necessity for socioeconomic interventions aimed at reducing the risk of PPD.

Graphical abstract

Keywords

association / genetic evidence / education / household income / postpartum depression

1. Introduction

Postpartum depression (PPD), classified as a depressive disorder with peripartum onset, is a psychiatric condition affecting women after childbirth that is typically identified through standardized screening scales [1]. Most cases manifest within six weeks postpartum, and although the majority of patients experience spontaneous remission, approximately 20% may develop persistent symptoms lasting years [2]. The incidence of PPD is influenced by multiple socioeconomic and cultural factors, resulting in significant cross-national prevalence variations ranging from 0.5% to 60.0% [3]. In China, reported prevalence rates range from 6.7% to 56.7% [4]. PPD typically presents with core symptoms including depressed mood, profound sadness, persistent fatigue, anxiety, and appetite loss, frequently accompanied by diminished mother-infant bonding. In severe cases, suicidal ideation or infanticidal tendencies may occur. Substantial evidence confirms that PPD contributes to delayed cognitive and behavioral development in infants, compromising offspring development while imposing significant burdens on families and society [5, 6, 7]. Therefore, early identification of risk factors combined with timely psychological interventions and pharmacological management is essential for enhancing maternal quality of life, promoting neonatal development, and maintaining familial and societal well-being [8, 9].

PPD is a complex multifactorial disorder with an unclear etiology. Cumulative evidence from numerous studies has identified key correlates of PPD, including a history of psychiatric disorders, antenatal depressive symptoms [10], neonatal or obstetric complications [11], and inadequate social support [12]. A study has shown that income is associated with the incidence of PPD [13]. Additionally, research indicates that lower household income is linked to heightened financial strain following childbirth, thereby increasing the risk of depression [14]. In turn, education has been recognized as a critical determinant of household income. However, the interplay among these three factors—education, household income, and PPD—lacks robust empirical validation.

To further investigate and clarify the association between educational attainment and PPD, as well as the mediating role of household income in this association, we employed Mendelian randomization (MR) analysis based on genome-wide association study (GWAS) data. This approach enables the assessment of potential associations between specific phenotypic traits and disease outcomes. By minimizing confounding biases and reverse association, while also offering significant efficiency in terms of time and cost, MR has proven to be clinically valuable. Consequently, it has emerged as a widely adopted tool in research on disease mechanisms and clinical decision-making [15].

2. Materials and Methods

2.1 Study Design

In this study, we conducted a two-step, two-sample MR analysis that included three specific analyses: the relationship between education and PPD, education and household income, and household income and PPD (Fig. 1). This approach aimed to assess the associations among educational level, household income, and PPD. Details regarding the GWAS datasets can be found in the “Availability of Data and Materials” section. The educational-level dataset comprised 461,457 individuals of European origin. In terms of household income data, 397,751 participants were involved. The PPD-related dataset covered 67,205 individuals, with 7604 diagnosed with PPD and 59,601 acting as controls.

2.2 Instrumental Variables (IVs) Selection

First, a selection threshold of p 5 \times 10^-8. was established for single-nucleotide polymorphisms (SNPs). To validate the selected instrumental variables (IVs), we conducted the linkage disequilibrium (LD) analysis on the screened SNPs, with the window size parameter set to 10,000 kb and the R² threshold set to less than 0.001. Additionally, by leveraging the PhenoScanner tool (https://www.phenoscanner.medschl.cam.ac.uk), we identified genome-wide features significantly associated (p 5 \times 10^-8) with these SNPs [16]. This step aimed to exclude SNPs linked with confounding factors related to education, household income, and PPD, thereby enhancing the reliability of subsequent analyses.

Second, to ensure the validity of the MR results, three key assumptions needed to be verified throughout the entire process [17]:

(i) IVs should have a direct association with the exposure factor;

(ii) IVs should not directly influence the outcome factor;

(iii) IVs should not be affected by the potential confounders that influence both the exposure and the outcome. To implement this step, we performed two key actions: first, we searched the PhenoScanner database (https://www.phenoscanner.medschl.cam.ac.uk) to identify features associated with confounding factors; second, we applied the MR-Egger intercept method and the MR Pleiotropy RESidual Sum and Outlier (MR-PRESSO) method to analyze and remove IVs with horizontal pleiotropy.

Third, we evaluated the strength of each instrumental variable. Weakly correlated IVs were excluded by calculating the F-statistic using the formula: F = R²(N – 2)/(1 – R²). In this formula, R² represents the proportion of variance in the exposure factor explained by each IV, and N denotes the sample size of the GWAS for the exposure factor. An F-statistic exceeding 10 indicates that the IVs are sufficiently robust, ensuring adequate validity for subsequent MR analyses and eliminating bias caused by weak instruments [18].

In summary, we implemented stringent screening of IVs to ensure the robustness and reliability of the MR analysis. The screening process included: applying a strict significance threshold (p 5 \times 10^-8), removing SNPs in LD, using the PhenoScanner database to eliminate SNPs associated with confounding factors, examining and excluding SNPs showing significant horizontal pleiotropy, and filtering out SNPs with low F-statistics.

2.3 Statistical Analyses

In the present study, the inverse variance weighting (IVW) method was primarily used to infer potential associations. Additionally, four complementary approaches—MR-Egger, weighted median, simple mode, and weighted mode—were utilized alongside the IVW method to validate these associations [19]. A two-sided p-value of less than 0.05 was considered statistically significant. Cochrane’s Q statistic (Q) was applied. Furthermore, a leave-one-out analysis was performed to identify individual SNPs that significantly influenced the results. Directed pleiotropy was evaluated by examining the deviation of the MR-Egger intercept, with a p-value of less than 0.05 indicating the presence of horizontal pleiotropy [20, 21]. The MR-PRESSO method was further applied to detect SNPs with horizontal pleiotropy [22]. Both the MR-Egger intercept and MR-PRESSO methods can also verify whether all IVs satisfy the third assumption of MR. Finally, the coefficient product method was adopted: the indirect effect was calculated by multiplying the effect of education on household income by the effect of household income on PPD. All statistical analyses were conducted using the TwoSampleMR and MR-PRESSO packages in R software version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria).

3. Results

3.1 IVs Selection

A total of 101 SNPs were included as IVs when education was the exposure factor and PPD was the outcome factor. There were 147 SNPs included as IVs when education was the exposure factor and household income was the outcome. Additionally, 15 SNPs were included as IVs when household income was the exposure factor and PPD was the outcome factor. The F-statistics for all SNPs were greater than 10.

3.2 MR Analysis

IVW analysis showed a negative association between education and PPD (OR = 0.877, 95% CI: 0.828–0.930, p = 1.050 \times 10^-5); a positive association between education and household income (OR = 1.095, 95% CI: 1.085–1.105, p = 2.805 \times 10^-87); and a negative association between household income and PPD (OR = 0.443, 95% CI: 0.204–0.962, p = 0.040). The remaining four methods showed that the results of the IVW method are stable and reliable (Table 1). The results of the coefficient product method showed a mediating effect coefficient of –0.074 (95% CI: –0.146 to –0.004), which accounted for 56.67% of the total effect.

3.3 Sensitivity Analysis

Cochrane’s Q statistics demonstrated that no heterogeneity was present in the IVs incorporated in both the IVW and MR-Egger methods for the associations between education and PPD, as well as household income and PPD (both p > 0.05). However, the heterogeneity of IVs was more salient in MR analyses where education was the exposure factor and household income was the outcome factor (p 0.05, Table 2). Leave-one-out analyses showed that results were susceptible to a single IV when household income was the exposure factor and PPD was the outcome factor, whereas the other two MR analyses did not see an IV that significantly affected the results of the analyses (Fig. 2). MR-Egger intercept analysis (Fig. 3, Table 3) indicated that there was potential horizontal pleiotropy in the MR analysis between household income and PPD. However, the results of the MR-PRESSO analysis (Table 3) demonstrated that within the scope of the above three MR analyses. However, the results of the MR-PRESSO analysis (Table 3) indicate that, within the scope of the three MR analyses mentioned above, none of the IVs used exhibited significant horizontal pleiotropy. It is particularly important to note that while the MR-PRESSO method suggests the presence of horizontally pleiotropic IVs in the association between education and household income, the specific IVs with pleiotropy could not be identified.

4. Discussion

Through the application of MR analysis, we identified a significant association between education and PPD. Specifically, women who spend more time in education have a lower risk of developing PPD. Additionally, an increase in household income plays an important mediating role in the association between time spent on education and PPD.

A cohort study involving 90,194 pregnant women demonstrated that those with a brief period of education were more likely to develop PPD compared to those with a longer period of education [14]. Another retrospective study also found that a lower level of education was a significant predictor of PPD [23]. In this study, household income accounted for 56.67% of the total effect in the association between education and PPD, underscoring the importance of household income in this association. Education level is a key measure of socio-economic status, and pregnant women from families with lower socio-economic status have a significantly higher risk of PPD [24, 25]. In empirical studies, years of schooling are usually used as a proxy for educational attainment. Modern theories of human capital highlight education to be an important factor in increasing individual incomes and contributing to a country’s economic development, as confirmed by several social surveys [26, 27, 28]. Socio-economic status is typically measured by level of education, income, and occupation, with education level often serving as a more reliable indicator than income or occupation [14, 29]. After childbirth, the costs of childcare and basic living expenses increase substantially. Women with low incomes, who are concerned about their family’s financial situation, are nearly four times more susceptible to PPD than those with higher income levels [30]. Increasing the number of years of education enhances overall improvement of knowledge and skills, enabling women to cope with stressful situations more effectively and optimistically. However, further investigation is needed to determine whether this relationship is consistent across diverse contexts, including variations in countries and social health care policies.

In China, the prevalence of maternal PPD is notably high, representing a significant public health challenge. However, due to the heterogeneity of the disease and the diversity of the scales, PPD is still underdiagnosed, under-recognized, and inadequately treated [31, 32]. Access to education and the acquisition of knowledge and skills can help women to improve their baseline cognitive level, develop a healthy mindset and a positive attitude towards life, and become more competitive at work [33, 34, 35]. Such women can bring more income, exert a positive and active influence on their families, raise the next generation with both virtue and talent, and promote the healthy development of society [36]; they can also provide the market with high-quality talents, promote the optimization of the industrial structure, and contribute to overall socio-economic development [37]. Consequently, it is important to reduce the incidence of PPD by increasing the number of years of female education, which in turn increases household income and improves the state of female mental health.

In this study, the MR analysis revealed a significant negative association between years of education and PPD. Specifically, the promotion of female education and the increase in female educational attainment can effectively decrease the incidence of PPD.

Limitations

This study has several limitations. First, all participants included in this study were drawn from a European population. While this reduces population heterogeneity and enhances the credibility of the results, additional research is required to determine whether these findings generalize to other ethnicities. Second, in the mediational two-sample MR analysis we conducted, we did not account for the effects of other potential confounders, such as age, on the outcome, nor did we assess their role in the association. However, MR analysis, akin to a randomized controlled trial, is minimally influenced by confounders after strict quality control of IVs. Third, the p-value of the MR-Egger intercept test in the analysis of household income and PPD is less than 0.05, and there is significant horizontal pleiotropy in the MR-PRESSO analysis of education and household income, which raises concerns about the reliability and validity of the research conclusions. Nevertheless, we rigorously screened SNPs to ensure they served as valid IVs. Additionally, we utilized the MR-PRESSO method to examine IVs for potential horizontal pleiotropy, but no specific IVs exhibited the issue. We also employed the Weighted Median Estimator, a robust method that is less sensitive to invalid IVs, for validation. The results of the weighted median analysis indicated that the association effect remained statistically significant. Both primary analysis methods (IVW and Weighted Median) yielded statistically significant results with consistent directional outcomes. This indicates that, despite concerns about pleiotropy, the primary findings of the MR analysis remain robust. However, the interpretation and further analysis of each association should be approached with caution.

5. Conclusions

This study reveals the associations between education level, household income, and PPD from a genetic perspective. We also analyze the indirect effect of household income on the relationship between education level and PPD. The results offer a theoretical basis for the screening and prevention of PPD, indicating that special attention should be given to the education level and household income of pregnant and postpartum women at high risk of PPD. However, since all GWAS datasets used in this study were derived from European populations, the interpretation of the associations between education level, household income, and PPD is limited. Whether these conclusions can be generalized to other populations (e.g., Asian populations) requires further validation through prospective cohort studies or GWAS data from other ethnic groups in the future.

Availability of Data and Materials

The datasets used in this study are available in public databases, and the data analysis process can be obtained from the corresponding author. GWAS data on education (ebi-a-GCST90029013), household income (ukb-b-7408), and PPD (finn-b-O15_POSTPART_DEPR) used in our study were obtained from the IEU GWAS Open Project (https://gwas.mrcieu.ac.uk/).

References

[1] ACOG Committee Opinion No. 757: Screening for Perinatal Depression. Obstetrics and Gynecology. 2018; 132: e208–e212. https://doi.org/10.1097/AOG.0000000000002927.

[2] Stewart DE, Vigod SN. Postpartum Depression: Pathophysiology, Treatment, and Emerging Therapeutics. Annual Review of Medicine. 2019; 70: 183–196. https://doi.org/10.1146/annurev-med-041217-011106.

[3] Wang Z, Liu J, Shuai H, Cai Z, Fu X, Liu Y, et al. Mapping global prevalence of depression among postpartum women. Translational Psychiatry. 2021; 11: 543. https://doi.org/10.1038/s41398-021-01663-6.

[4] Peng S, Lai X, Du Y, Meng L, Gan Y, Zhang X. Prevalence and risk factors of postpartum depression in China: A hospital-based cross-sectional study. Journal of Affective Disorders. 2021; 282: 1096–1100. https://doi.org/10.1016/j.jad.2021.01.012.

[5] Meltzer-Brody S, Howard LM, Bergink V, Vigod S, Jones I, Munk-Olsen T, et al. Postpartum psychiatric disorders. Nature Reviews. Disease Primers. 2018; 4: 18022. https://doi.org/10.1038/nrdp.2018.22.

[6] Zhao XH, Zhang ZH. Risk factors for postpartum depression: An evidence-based systematic review of systematic reviews and meta-analyses. Asian Journal of Psychiatry. 2020; 53: 102353. https://doi.org/10.1016/j.ajp.2020.102353.

[7] Nguyen HTH, Hoang PA, Do TKL, Taylor-Robinson AW, Nguyen TTH. Postpartum depression in Vietnam: a scoping review of symptoms, consequences, and management. BMC Women’s Health. 2023; 23: 391. https://doi.org/10.1186/s12905-023-02519-5.

[8] Wells T. Postpartum Depression: Screening and Collaborative Management. Primary Care. 2023; 50: 127–142. https://doi.org/10.1016/j.pop.2022.10.011.

[9] Dennis CL, Singla DR, Brown HK, Savel K, Clark CT, Grigoriadis S, et al. Postpartum Depression: A Clinical Review of Impact and Current Treatment Solutions. Drugs. 2024; 84: 645–659. https://doi.org/10.1007/s40265-024-02038-z.

[10] Ertmann RK, Lyngsøe BK, Nicolaisdottir DR, Kragstrup J, Siersma V. Mental vulnerability before and depressive symptoms during pregnancy and postpartum: a prospective population-based cohort study from general practice. Nordic Journal of Psychiatry. 2022; 76: 243–249. https://doi.org/10.1080/08039488.2021.1953583.

[11] Accortt EE, Arora C, Mirocha J, Jackman S, Liang R, Karumanchi SA, et al. Low Prenatal Vitamin D Metabolite Ratio and Subsequent Postpartum Depression Risk. Journal of Women’s Health (2002). 2021; 30: 113–120. https://doi.org/10.1089/jwh.2019.8209.

[12] Cho H, Lee K, Choi E, Cho HN, Park B, Suh M, et al. Author Correction: Association between social support and postpartum depression. Scientific Reports. 2022; 12: 3899. https://doi.org/10.1038/s41598-022-08119-x.

[13] Miyake Y, Tanaka K, Arakawa M. Associations of job type, income, and education with postpartum depressive symptoms: The Kyushu Okinawa Maternal and Child Health Study. Psychiatry Research. 2020; 291: 113224. https://doi.org/10.1016/j.psychres.2020.113224.

[14] Matsumura K, Hamazaki K, Tsuchida A, Kasamatsu H, Inadera H, Japan Environment and Children’s Study (JECS) Group. Education level and risk of postpartum depression: results from the Japan Environment and Children’s Study (JECS). BMC Psychiatry. 2019; 19: 419. https://doi.org/10.1186/s12888-019-2401-3.

[15] Guo D, Diao Z, Wang K, Pang C. Causal association between rheumatoid arthritis and pregnancy loss and intrauterine growth retardation: A bidirectional two-sample Mendelian randomization study. Medicine. 2024; 103: e36873. https://doi.org/10.1097/MD.0000000000036873.

[16] Staley JR, Blackshaw J, Kamat MA, Ellis S, Surendran P, Sun BB, et al. PhenoScanner: a database of human genotype-phenotype associations. Bioinformatics (Oxford, England). 2016; 32: 3207–3209. https://doi.org/10.1093/bioinformatics/btw373.

[17] Davies NM, Holmes MV, Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians. BMJ (Clinical Research Ed.). 2018; 362: k601. https://doi.org/10.1136/bmj.k601.

[18] Burgess S, Thompson SG, CRP CHD Genetics Collaboration. Avoiding bias from weak instruments in Mendelian randomization studies. International Journal of Epidemiology. 2011; 40: 755–764. https://doi.org/10.1093/ije/dyr036.

[19] Chen Y, Lyu S, Xiao W, Yi S, Liu P, Liu J. Sleep Traits Causally Affect the Brain Cortical Structure: A Mendelian Randomization Study. Biomedicines. 2023; 11: 2296. https://doi.org/10.3390/biomedicines11082296.

[20] Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. International Journal of Epidemiology. 2015; 44: 512–525. https://doi.org/10.1093/ije/dyv080.

[21] Burgess S, Thompson SG. Erratum to: Interpreting findings from Mendelian randomization using the MR-Egger method. European Journal of Epidemiology. 2017; 32: 391–392. https://doi.org/10.1007/s10654-017-0276-5.

[22] Verbanck M, Chen CY, Neale B, Do R. Publisher Correction: Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nature Genetics. 2018; 50: 1196. https://doi.org/10.1038/s41588-018-0164-2.

[23] Tsai JM, Tsai LY, Tsay SL, Chen YH. The prevalence and risk factors of postpartum depression among women during the early postpartum period: a retrospective secondary data analysis. Taiwanese Journal of Obstetrics & Gynecology. 2023; 62: 406–411. https://doi.org/10.1016/j.tjog.2023.03.003.

[24] Agrawal I, Mehendale AM, Malhotra R. Risk Factors of Postpartum Depression. Cureus. 2022; 14: e30898. https://doi.org/10.7759/cureus.30898.

[25] Qi W, Zhao F, Liu Y, Li Q, Hu J. Psychosocial risk factors for postpartum depression in Chinese women: a meta-analysis. BMC Pregnancy and Childbirth. 2021; 21: 174. https://doi.org/10.1186/s12884-021-03657-0.

[26] Cuñado J, De Gracia FP. Does Education Affect Happiness? Evidence for Spain. Social Indicators Research. 2012; 108: 185–196. https://doi.org/10.1007/s11205-011-9874-x.

[27] Nikolaev B. Does Higher Education Increase Hedonic and Eudaimonic Happiness? Journal of Happiness Studies. 2018; 19: 483–504. https://doi.org/10.1007/s10902-016-9833-y.

[28] Cseh Papp I, Karácsony P, Juhász T. Study Preferences in Higher Education. Acta Polytechnica Hungarica. 2023; 20: 229–248. https://doi.org/10.12700/APH.20.4.2023.4.13.

[29] Padrutt ER, DeJoseph ML, Wilson S, Mills-Koonce R, Berry D. Measurement invariance of maternal depressive symptoms across the first 2 years since birth and across racial group, education, income, primiparity, and age. Psychological Assessment. 2023; 35: 646–658. https://doi.org/10.1037/pas0001242.

[30] Segre LS, O’Hara MW, Arndt S, Stuart S. The prevalence of postpartum depression: the relative significance of three social status indices. Social Psychiatry and Psychiatric Epidemiology. 2007; 42: 316–321. https://doi.org/10.1007/s00127-007-0168-1.

[31] Dominiak M, Antosik-Wojcinska AZ, Baron M, Mierzejewski P, Swiecicki L. Recommendations for the prevention and treatment of postpartum depression. Ginekologia Polska. 2021; 92: 153–164. https://doi.org/10.5603/GP.a2020.0141.

[32] Sharma P, Khera K. A novel tool for risk assessment, screening, diagnosis, assessment, and therapy in postpartum depression. International Journal of Gynaecology and Obstetrics: the Official Organ of the International Federation of Gynaecology and Obstetrics. 2024; 166: 617–625. https://doi.org/10.1002/ijgo.15421.

[33] Ritchie SJ, Tucker-Drob EM. How Much Does Education Improve Intelligence? A Meta-Analysis. Psychological Science. 2018; 29: 1358–1369. https://doi.org/10.1177/0956797618774253.

[34] Lövdén M, Fratiglioni L, Glymour MM, Lindenberger U, Tucker-Drob EM. Education and Cognitive Functioning Across the Life Span. Psychological Science in the Public Interest: a Journal of the American Psychological Society. 2020; 21: 6–41. https://doi.org/10.1177/1529100620920576.

[35] Tran DB, Pham TDN, Nguyen TT. The influence of education on women’s well-being: Evidence from Australia. PloS One. 2021; 16: e0247765. https://doi.org/10.1371/journal.pone.0247765.

[36] Davis-Kean PE. The influence of parent education and family income on child achievement: the indirect role of parental expectations and the home environment. Journal of Family Psychology: JFP: Journal of the Division of Family Psychology of the American Psychological Association (Division 43). 2005; 19: 294–304. https://doi.org/10.1037/0893-3200.19.2.294.

[37] Hong G, Kim S, Park G, Sim SG. Female Education Externality and Inclusive Growth. Sustainability. 2019; 11: 3344. https://doi.org/10.3390/su11123344.

Funding

Research Personnel Cultivation Programme of Zhongda Hospital Southeast University(CZXM-GSP-RC147)