Table 4.
Summary of included articles (perinatal).
| PERINATAL | ||||
|---|---|---|---|---|
| First Author | Purpose/Aims | Design# | Factor Domain | Summary of significant findings (values) |
| ®Sha (2022) (N=114) US | To determine whether cytokines and kynurenine metabolites can predict the development of depression in pregnancy. | Prospective4 | B | ↑ IL-1β, IL-6, and QUIN were associated with ↑ depression severity and/or ↑ odds of having depression (Percent change in OR(CI): 32.3% (7.0, 63.6), 58.4% (22.1, 111.7), 91.6% (15.0, 232.0) |
| IL-6 performed best in predicting depressive symptoms; however, KYN, QUIN, KYN/TRP ratio (rKT) also produced good predictions (AUC = 0.79 and 0.8 by Bayesian ordinal and logistic regression, respectively; ROC AUC >0.7). Precision recall analyses confirmed predictive value of model. | ||||
| The leave-one-out cross validation method indicated the predictability of the model would be optimal from mid- to late pregnancy (2nd to 3rd trimester). The full model nominally outperformed individual markers for predicting risk of significant depressive symptoms. Ordinal and logistic regression full models had ROC AUC = 0.83, PR AUC = 0.41. | ||||
| ®Kimmel (2022) (N = 30) US | Analyze trajectories of serotonin and tryptophan-related metabolites, bile acid metabolites, and microbial composition related to psychiatric history and current symptoms across the perinatal period. | Pilot3 | B | Fiber consumption was slightly ↓ in cases compared to controls (determined too small a sample to calculate p-values). |
| Mean serotonin level ↑ from pregnancy to postpartum (p = 0.0002 for 3rd trimester (V2) to 5–10 weeks postpartum (V3); p = 0.002 for 1st or 2nd trimester (V1) to V3). NEOP level trajectories followed a different pattern than serotonin by ↑ from V1 to V2 (p < 0.0001) and then ↓ postpartum (p = 0.005). Mean KYN ↑ from V1 to V2 (p = 0.003) and ↑ again from V2 to V3 (p = 0.004). The KYN/TRP ratio was ↑ at V2 and V3 compared to V1 (p < 0.0001; p < 0.0001). KA was ↑ at V3 compared to both V2 (p = 0.003) and V1 (p = 0.0004). | ||||
| Primary bile acids: Chenodexycholic acid (CDCA) ↑ from V2 to V3 (p < 0.00011) with an overall ↑ from earlier V1 to V3 (p = 0.0003); Glycochenodeoxycholic acid (GCDCA) ↑ from V2 to V3 (p < 0.0001) and remained ↑ at V3 compared to V1 (p < 0.0001); Taurochenodeoxyccholate (TCDCA) ↓ from V2 to V3 (p = 0.001); Glycocholic acid (GCA) ↑ from V1 to V2 (p = 0.003) and ↑ from V1 to V3 (p = 0.005); Taurocholic acid (TCCA) ↑ from V1 to V2 (p < 0.0001), and ↓ from V2 to V3 (p < 0.0001). | ||||
|
Secondary bile acids: Glycoursodeoxchoilic acid (GUDCA) ↓ from V2 to V3 (p < 0.0001) whereas GUDCA and Ursodeoxycholicc acid (UDCA) ↑ from V2 to V3 (p < 0.0001; p = 0.0003) and GUDCA remained ↑ from V1 to V3 (p < 0.0001); Glycolithocholic acid (GLCA) ↑ from V2 to V3 (p < 0.0001) and levels of Glychoyocholic acid (GHCA) and GLCA were ↑ compared to the initial value in pregnancy (p = 0.0001; p < 0.0001; p = 0.0005) Tauro alpha-murcholic acid (TaMCA), Taurohyocholic acid (THCA), and tarodeoxycholate hydrate (TDCA) ↓ from V2 to V3 (p < 0.0001; p = 0.0003; p < 0.0001) and V3 were ↓ than earlier in pregnancy (p < 0.0001; p = 0.0003; p = 0.002). TUDCA, TDCA, and TCA were associated with change in NEOP from V1 to V2 (q = 0.011; q = 0.021; q = 0.021). TUDCA was also associated with change in TRP (q = 0.004), KYN (q = 0.001), and KA/KYN ratio (q = 0.002). These findings became stronger when excluding those in the first trimester. | ||||
| Metabolites and microbiome: Alpha diversity did not significantly change across the perinatal period. ↑ bile acid GUDCA and UDCA levels were associated with ↓ alpha-diversity across all 4 indices (evenness, Faith’s phylogenetic diversity, count of observed OTUs, Shannon entropy). ↑ CDCA was associated with ↓ alpha diversity for the evenness index and Shannon index only, and also only when first trimester participants were included. Certain bacterial genera were associated with UDCA and TUDCA, primarily in the order Clostridiales and family Cachnospiraceae. THcA was also associated with Riseburia. UDCA was the only metabolite associated with psychiatric history (q = 0.033). | ||||
| ®Tebeka2 (2021) (N = 3,252) France | Assess the relationship between childhood trauma (CT) and perinatal depression, considering types of CT | Case control3 | S | Those reporting childhood trauma (CT) were ↑ likely to be < 26 years old (8.1% vs. 4.5%; OR = 1.8; 95% CI: 1.2–2.6) > 39 years old (11% vs. 7%; OR = 1.9; 95% CI: 1.2–2.9), single (6.7% vs. 2.7%; OR = 2.6; 95% CI: 1.5–4.2), have a lower level of education (18.1% vs. 6.8%; OR = 3.0; 95% CI: 1.8–3.6), and ↑ likely to have been unemployed (14.1% vs. 6.1%; OR = 2.5; 95% CI: 1.8–3.6). |
| Those with CT had a ↑ risk of either depression, anxiety, or suicide attempts compared those without (61.6% vs. 40.8%; OR = 2.3; 95% CI: 1.8–2.9), and a personal history of depression, anxiety, or suicide attempts were ↑ frequent in those with CT (depression: OR = 2.2; 95% CI: 1.7–2.7; anxiety: OR = 2.3; 95% CI: 1.7–3.0; suicide attempt: OR = 5.4; 95% CI: 3.5–8.4) | ||||
| Depression was ↑ common in those with a CT regardless of type of CT, and the difference was significant for emotional, physical, and sexual abuse as well as emotional neglect (p < 0.05 for each). The types of CT demonstrated specific associations with different timing of depression onset. Emotional neglect was associated with depression during pregnancy (aOR = 2.1; 95% CI: 1.2–3.8, p = 0.012); sexual abuse with both early and late onset PPD (aOR = 2.3; 95% CI: 1.2–4.6; aOR = 2.4; 95% CI: 1.2–4.9, respectively); emotional abuse was associated only with late PPD (aOR = 2.7; 95% CI: 1.4–5.1). | ||||
| A dose effect was present between CT types and risk of depression. When 1 type of CT was present there was a ↑ risk of depression (aOR = 1.6; 95% CI: 1.1–2.3, p = 0.015), whereas, when 2+ types of CT were present the risk further ↑ (aOR = 2.1; 95% CI: 1.3–3.3) even after adjusting for history of depression and sociodemographic covariates. | ||||
| Nazzari (2020) (N = 97) Italy | 1) Describe the cross-sectional and longitudinal association between tryptophan, kynurenine, and kynurenine/tryptophan ratio and depression symptoms in late pregnancy through the first year postpartum 2) examine the role of inflammatory (IL-6) and stress (cortisol) markers in moderating any associations 3) determine if specific to depressive symptoms or can be replicated with anxiety given high concurrence of these disorders | Prospective4 | B | ↑ prenatal Kyn levels were associated with ↓ depressive symptoms in late pregnancy (estimate = − 0.002, SE = 0.001, p = 0.03) after adjusting for maternal age. |
| Pre-pregnancy BMI was mildly associated with IL-6 levels (r = 0.23, p = 0.03) in preliminary analysis but adjusting models for BMI did not alter the direction or significance of findings. | ||||
| Model 2: There was a three-way interaction among prenatal Trp levels, IL-6, and slopes of time on depression scores (ps < 0.05). ↓ levels of prenatal Trp and ↑ IL-6 were associated with ↑ depressive symptoms in late pregnancy (p = 0.04) and with the change in depressive symptoms from pregnancy to three postpartum time points (ps = 0.04). | ||||
| Model 3: A three-way interaction among the KYN/TRP ratio, IL-6, and the depression scores trajectory from pregnancy to 12 months postpartum. ↓ levels of prenatal KYN/TRP ratio and ↑ levels of IL-6 were associated with ↑ depressive scores at delivery (p = 0.05) and 12 months postpartum (p = 0.004) and with a flatter trajectory of change in depressive symptoms from pregnancy to 12 months postpartum (p = 0.048). Conversely, at ↑ levels of KYN/TRP ratio and ↑ IL-6 levels were associated with a ↓ in depressive scores from pregnancy to 3 (p = 0.03) and 12 months (p = 0.014) postpartum. | ||||
| Garman2 (2019) (N = 384) South Africa | Identify trajectories of perinatal depressive symptoms and their predictors among low-income South African women who were already at risk of depression during pregnancy. | Prospective4 | Bh, S, E | Food insecurity predicted classification of either prenatal only depression or prenatal and postpartum depression. The odds of being classified in the prenatal and postpartum depression trajectory was 2.5 greater (95% CI: 1.21–5.15; p = 0.013) among participants who reported being severely food insecure. |
| Overall levels of social support at baseline ↓ the odds of belonging to the prenatal and postpartum depression class (OR = 0.97, 95% CI: 0.95–0.99; p = 0.011). When looking at specific types of support, only a ↑ level of family support (OR = 0.91, 95% CI: 0.86–0.96; p = 0.001) or ↑ level of support from a significant other (OR = 0.94, 95% CI: 0.88–1.00; p = 0.046) ↓ the odds of being classified into the prenatal and postpartum depression class. | ||||
| Those who reported IPV at baseline were 2.8 times ↑ likely (95% CI: 1.23–6.52; p = 0.014) to belong to the prenatal and postpartum depression class. | ||||
| Odds of belonging to the prenatal and postpartum depression class were ↑ among those who reported greater functional impairment (OR = 1.03, 95% CI: 1.02–1.06; p = 0.002), heavy drinking during pregnancy (OR = 2.12, 95% CI: 0.03–4.37; p = 0.042), had current (OR = 2.77, 95% CI: 1-32–5.80; p = 0.007) or lifetime diagnosis of depression (OR = 2.85, 95% CI: 1.38–5.87; p = 0.004), and high risk of suicide (OR = 2.58, 95% CI: 1.19– 5.61; p = 0.017). | ||||
| Teshigawara (2019) (N = 132) Japan | To determine whether cytokines and kynurenine metabolites can predict the development of depression in pregnancy. | Prospective3 | B | In the non-depressed group: TRP, KYN, 3HK, and KA were ↑ postpartum compared to pregnancy (two-way repeated ANOVA, Trp: Fgroup (3, 128) = 1.44, p = 0.234, Fperiod (1, 128) = 64.3, p < 0.0001, Fgroup × period (3, 128) = 0.376, p = 0.771; Kyn: Fgroup (3, 128) = 0.927, p = 0.430, Fperiod (1, 128) = 96.4, p < 0.01, Fgroup × period (3, 128) = 6.09, p < 0.01; 3HK: Fgroup (3, 128) = 0.0662, p = 0.978, Fperiod (1, 128) = 6.09, p < 0.05, Fgroup × period (3, 128) = 1.98, p = 0.120; KA: Fgroup (3, 128) = 1.52, p = 0.213, Fperiod (1, 128) = 2.11, p = 0.149, Fgroup × period (3, 128) = 5.32, p < 0.01). |
| In the postpartum depressed group: KYN and KA were ↑ during pregnancy, but 3HAA during the postpartum period was ↓ than that of the non-depressed group. No differences were noted in TRP or its metabolites between the temporary gestational depressive group or the continuous depressive group and the non-depressive group. | ||||
| The ratio of KYN in the postpartum period compared to that during pregnancy was significantly ↓ in the postpartum depressive group compared to the non-depressive group (one-way ANOVA, F(3, 128) = 5.27, p < 0.01). | ||||
| In the postpartum depressive group KYN/TRP and KA/KYN ratio during pregnancy were ↑ than those in the non-depressive group. KYN/TRP during postpartum to that during pregnancy was significantly ↓ than the non-depressive group (one-way ANOVA, F (3, 128) = 4.54, p < 0.01). | ||||
| KYN, KA, and KYN/TRP, and KA/KYN ratio during pregnancy were ↑ and 3HAA during postpartum was ↓ in the postpartum depressive group compared to non-depressive group. | ||||
| KYN, KA, and KYN/TRP during pregnancy was correlated with depression scores during the postpartum period (Pearson’s correlation: KYN: r(77) = 0.330, p < 0.01, KA: r(77) = 0.278, p < 0.05, KYN/TRP: r(77) = 0.229, p < 0.05, KA/KYN: r(77) = 0.221, p = 0.05). There was a negative relationship between 3HAA levels during postpartum period and depression scores (Pearson’s correlation: r(77) = −0.259, p < 0.05). | ||||
| Vargas-Terrones (2017) (N = 124) Spain | Analyze trajectories of serotonin and tryptophan-related metabolites, bile acid metabolites, and microbial composition related to psychiatric history and current symptoms across the perinatal period. | Randomized control trial3 | Bh | The percentage of depressed participants was ↓ in the intervention group compared to the control group at week 38 (18.6% vs. 35.6%) (χ2 = 4.190: p = 0.041) and at 6 weeks postpartum (14.5% vs 29.8%) (χ2 = 3.985: p = 0.046). |
| Significant differences were noted in the multiple imputation analysis at 38 weeks (18.6% vs. 34.4%) (χ2= 4.085; p = 0.049). | ||||
| A treatment effect was found in the per-protocol (F2, 220 = 3.798; p = 0.024) and in the simple imputation (F2,244 = 3.351; p = 0.037) analyses. Differences were also found in the group-time interaction between gestational weeks 12–16 (baseline) and 6 weeks postpartum (p = 0.014) in the per-protocol analysis. | ||||
| Differences were found in the group-time interaction between depression scores at baseline and gestational week 38 (p = 0.046), and between baseline and 6 weeks postpartum (p = 0.025), with a ↓ depression score in the intervention group than in the control group. | ||||
| The participants considered to have excessive gestational weight gain, the control group had a ↑ percentage of depression at week 38 (χ2 = 9.489: p = 0.002) and at 6 weeks postpartum (χ2 = 5.202; p = 0.023). | ||||
| The percentage of depressed women was ↓ in the intervention group compared to the control group at week 38 for those with pre-pregnancy normal-weight BMI (χ2 = 4.688; p = 0.030). | ||||
| ®Robertson Blackmore2 (2016) (N = 171) US | Examine the relationship between exposure of intimate partner violence (IPV) and proinflammatory cytokine levels, a candidate mechanism accounting for poor psychiatric and obstetric outcomes, across the perinatal period | Prospective4 | B, S | Lifetime exposure to IPV was associated with a range of psychiatric conditions, including generalized anxiety disorder, post-traumatic stress disorder, and depression. Further, IPV was associated with experiencing depression during both pregnancy and postpartum. |
| Those with a history of IPV had ↑ levels of TNF-α (z = −2.29, p < 0.05) compared to those with no IPV exposure. | ||||
| After controlling for participants characteristics, a greater change in the levels of IL-6 during pregnancy compared to the postpartum period remained (β = 0.21, p = 0.04). This trend was different according to IPV status. Those who experienced violence had smaller changes in IL-6 across the time points compared to those not exposed to violence (β = −0.36, p = 0.04). From 6 weeks to 6-month PP, those exposed to violence had a greater ↓ in IL-6 compared to those without exposure (β = 0.36, p = 0.04). | ||||
| The change in TNF-α levels at 32 weeks’ gestation to 6 weeks PP was ↑ than the change from 6 weeks to 6 months PP (β = 1.54, p < 0.01). | ||||
| Fasching (2012) (N = 361) Germany | Identify trajectories of perinatal depressive symptoms and their predictors among low-income South African women who were already at risk of depression during pregnancy. | Prospective3 | B | Haplotype block analysis showed that 10 of the 14 haplotypes of the THP2 gene were assembled in three haplotype blocks (B1-B3). SNPs rs6582071 and rs11178997 (haplotype A) were also analyzed given these SNPs are known to be of functional relevance. |
| Genotype-phenotype association in haplotype Block A: The most common haplotype was GT (63.4% homozygous for this haplotype and 31.6% had one allele for GT). The extremely rare haplotype GA (only one carrier) was excluded. The linear mixed model indicated an effect for time (p < 0.00001, F-test) as well as haplotype GT (p = 0.02, F-test) and the interaction of time and haplotype GT (p = 0.03, F-test). Pairwise comparison demonstrated ↑ depression scores at different timepoints: 1) time point 3 for those non-carriers of the GT haplotype compared to those carrying one copy of GT at time point 3 (p < 0.01). At timepoints 1 and 3, those non-carriers of the GT haplotype showed ↑ depression scores than those carrying two copies of the GT (p = 0.01; p = 0.01). ↑ depression scores were found at timepoint 1 compared to timepoint 2 in all three haplotype groups (0 GT: p < 0.001, 1 GT: p < 0.01, 2 GT: p < 0.00001). There was an ↑ in depression scores from timepoint 2 to timepoint 3 for non-carriers of a GT haplotype (p = 0.01) and for carriers of two copies of GT (p < 0.001). | ||||
| Haplotype block B1: SNPs: rs6582071, rs11178997, rs1117899; Haplotypes: CAT, CGA, CGT, TAA Results are identical to those from haplotype block A described above. | ||||
| Haplotype block B3: Block B3 resulted in four haplotypes (GAA, TAA, TA, TTG) with the most common being TTA. 33% of those carrying two copies and 51.8%^ carrying one copy. Linear mixed model: Those carrying two copies of TAA (0.6%) were joined with the carriers of one copy of TAA (15.5%). An effect for time was shown (p < 0.00001, F-test) as well as the interaction between TAA and time (p = 0.01, F-test). Differences between the patient groups at time 1 were seen for TAA, and both genotype groups were different between all three time points (p < 0.00001, p < 0.00001, p < 0.01). Pairwise comparison: Three timepoints showed ↑ depression scores at time 1 and time 2 for TAA (0 TAA: p < 0.0001, 1 + 2 TAA: p < 0.0001). At time 2 and 3, an ↑ in depression scores was seen in both groups (0 TAA: p = 0.03, 1 + 2 TAA: p = 0.02), and depression scores were ower at time 1 compared to time 3 (0 TAA: p < 0.01, 1 + 2 TAA: p < 0.01). | ||||
| SNPs outside of haplotype blocks: rs10879354 (T/T + T/C vs C/C) showed an effect for time (p < 0.00001) and SNP (p = 0.04) but not for interaction. Pairwise comparison of the three timepoints showed ↑ depression scores at time 1 compared to time 2 (p < 0.00001); time 2 compared to time 3 indicated a depression score ↑ (p < 0.001); time 3 was ↑ than time 1 (p < 0.01). | ||||
| Lin2 (2009) (N = 200) Taiwan | To determine whether cytokines and kynurenine metabolites can predict the development of depression in pregnancy. | Cross-sectional1 | B | Six SNPs (T-703G, T-473A, A90G, C2755A, C10662T, G93329A) were noted from the TPH2 gene. |
| Two SNPs were found in the cases (T-473A, p = 0.042; A90G, p = 0.038) that were not found in controls. | ||||
| Risk analysis showed that the “A” allele conferred a risk (RR = 1.73; 95% CI: 1.59–1.88) and demonstrated a dominant gene effect (A-allele carrier vs non-A allele carrier, AC vs CC; p = 0.038). | ||||
| A strong linkage disequilibrium in the 5’ region between SNPs −703A and A90G in both groups (D’ ranged from 0.87 to 1) and the D’ dropped as the distance between the pairs of markers ↑ (D’ ranged from 0.50–0.76). | ||||
| The GTAA haplotype, which contains the risk 2755A allele, was different among patients and controls (Fisher’s exact test, p = 0.044); however, the significant in distribution of the GTAA haplotypes disappeared in a rigid permutation test (p = 0.086). | ||||
| Murakami2 (2008) (N = 865) Japan | To examine the association between dietary GI and glycemic load (GL) and postpartum depression. | Prospective2 | Bh | Compared with dietary glycemic index (GI) in the first quartile, dietary GI in the third quartile, but not the fourth was associated with ↓ risk of PP depression. Multivariate ORs (95% Cis) for PP depression for each of the four quartiles were: 1.00 (reference), 0.68 (0.39–1.17), 0.56 (0.32–0.995, p = 0.048), and 0.72 (0.41–1.26), respectively (p for trend = 0.18). |
Author2 = secondary analysis; Design# = Design + Number of timepoints investigated; Domain of factors investigated in relation to depression: B = Biological, Bh = Behavioral, E = Environmental, S = Social ; Values (when provided) = statistical values respective to analysis; Factors investigated in relation to depression bold; Timeframe and/or groups investigated underlined.
®
Study reported race/ethnicity