The role of prepulse inhibition in predicting new-onset postpartum depression

Postpartum depression (PPD), a subtype of the clinical condition defined as peripartum depression, begins within the first 4 weeks after childbirth¹; however, clinically, depressive episodes diagnosed within the first year postpartum are often included in the categorization of PPD². Symptoms of PPD include depressed mood, lack of energy and reduced interest in daily activities. Women affected by PPD are a diverse group³; some women are at increased risk for new postpartum depressive episodes in successive pregnancies and have a higher incidence of sick leave, morbidity and suicide^4,5,6,7. Many psychosocial risk factors for PPD are known. Studies investigating self-reported PPD have identified poor socioeconomic status, pregnancy and delivery complications, and having a history of depression to be linked to increased risk of PPD^3,8,9. In a large register-based study, women with a history of depression were 20× more likely to suffer from clinically diagnosed PPD than those without a history of depression¹⁰. Previous studies have identified personality traits, such as high levels of neuroticism and anxiety and low resilience, as strong risk factors for PPD^11,12,13. Poor social support also increases the risk of PPD, with a shorter relationship with a partner, poor marital relationship, and lack of social support from family and friends noted as risk factors for PPD^14,15,16.

Furthermore, large alterations in hormonal levels occurring during the pregnancy and postpartum periods, including those in the hypothalamus–pituitary–adrenal (HPA) axis, could put individuals sensitive to endocrinological changes at increased risk for PPD^2,14,17,18. The hormonal and physical changes associated with pregnancy and childbirth constitute a stress test of the female body¹⁹. An individual’s adaptation to acute or chronic stress is predictive of mental health conditions²⁰ and poor stress adaptation has been linked to depression²¹. Pregnancy is generally associated with a decrease in the neuroendocrine response to acute stressors, whereas increased reactivity is believed to be associated with a greater likelihood of PPD²².

The gating mechanism, a process for input filtration, has a pivotal role in stress regulation by safeguarding cortical areas from the inundation of unnecessary or irrelevant information^23,24. One type of gating, referred to as sensorimotor gating, involves the ability to automatically inhibit a motor response to a sensory event. One paradigm that is commonly used to study sensorimotor gating involves measuring the inhibition of the startle response, which is the reflex triggered by an auditory, visual or tactile stimulus that causes heart rate acceleration and contraction of body and face muscles, thereby instinctively prompting the blink reflex²⁵. The acoustic startle response (ASR) is mostly used in research and is triggered by an auditory stimulus. When a low-salience auditory stimulus immediately precedes a startle stimulus, the startle motor reaction response decreases²⁶. This is referred to as prepulse inhibition (PPI) and is generally recognized as an operational measure of sensorimotor gating^27,28,29,30.

Mounting evidence points to a role for gonadal hormones, such as progesterone, in the regulation of sensorimotor gating^31,32. Changes in PPI are seen across the menstrual cycle^33,34 and women who are pregnant show lower levels of PPI compared with women who are postpartum³⁵. An increased startle response and reduced levels of PPI have also been found in women with premenstrual dysphoric disorder (PMDD)^36,37,38,39.

Gating deficiency, indicated by reduced PPI, is observed in various psychiatric disorders^{24,27,36,40,41}, including schizophrenia^42,43, antisocial personality disorder⁴⁴, obsessive–compulsive disorder^45,46, bipolar disorder^40,47 and post-traumatic stress disorder⁴⁸. Animal and human studies related to PPI and depression have been scarce and have often yielded mixed results. In animal studies, mice with separation-induced depressive symptoms showed lower PPI levels than non-isolated mice in the control group^49,50. A study⁵¹ investigating PPI in individuals with major depressive disorder found that these individuals only showed a non-significant tendency toward lower PPI than the healthy controls. In another study²⁴, individuals with only depression or only anxiety were not found to have significantly lower levels of PPI than healthy controls; individuals with comorbid depression and anxiety, however, were found to have significantly lower PPI compared with individuals with depression or anxiety alone and with healthy controls. One previous study investigating PPI in the postpartum period showed that sensorimotor gating was reduced among women with PPD³⁶.

There is a growing body of knowledge suggesting that mental disorders can be predicted with the use of biomarkers⁵². Relatively high levels of sensorimotor gating have been associated with better future treatment response to cognitive behavioral therapy in patients with schizophrenia⁵³, and a recent study investigating PPI in individuals at clinically high risk for psychosis found that deficits in PPI occur before the onset of full-scale psychosis⁵⁴. Previous research aimed at predicting PPD has suggested models based on self-reports and clinical health characteristics^11,55; however, the use of physiological measures to predict PPD has remained largely unexplored.

Despite findings supporting that PPI is reduced in depression and in PPD and that alterations of PPI are seen in pregnancy, there have not been studies investigating the potential of reduced PPI in pregnancy to predict the development of postpartum depressive symptoms. Thus this study aimed to investigate whether PPI, measured in late pregnancy, could predict depressive symptom status at 6 weeks postpartum. Furthermore, we aimed to explore the predictive value of PPI among women with and without depressive symptoms during pregnancy. We hypothesized that reduced PPI in late pregnancy would be predictive of depression in the postpartum period.

Sample characteristics

Data were drawn from a longitudinal study about perinatal depression from Uppsala, Sweden, named the Biology, Affect, Stress, Imaging and Cognition (BASIC) cohort⁵⁶. In this substudy, pregnant women participating in the BASIC study were invited between January 2010 and May 2013 during gestational weeks 35–39 to measure the ASR and the PPI⁵⁷. Depression status during pregnancy and at 6 weeks postpartum was assessed by the Edinburgh Postnatal Depression Scale (EPDS).

Out of 179 participants with complete data, 28 reported scores of 12 or more on the EPDS at 6 weeks postpartum and were categorized as having developed PPD (15.6%). Women with PPD were less likely to have attended university (P = 0.02), or to be working full- or part-time (P < 0.001) than women without PPD (Table 1). Women with PPD were also more likely to have depression during pregnancy (P < 0.001) and anxiety at the time of ASR measurements (P < 0.001) than women without PPD. Moreover, women with PPD were more likely to have had premenstrual syndrome (PMS) or PMDD when not pregnant (P = 0.003). No group differences were found in age, prepregnancy body mass index (BMI), selective serotonin reuptake inhibitor (SSRI) use during pregnancy, sleep the night before ASR measurement and baseline ASR (Table 1).

Results from testing the association between PPI and PPD

The associations between the PPI measures at 72 dB, 76 dB, 78 dB and 86 dB; a combined measure of global PPI; and PPD were investigated using logistic regression. In the crude models, inhibition following all prepulse decibel levels was not significantly associated with PPD.

In the adjusted models, including interaction terms for PPI and depression during pregnancy, significant negative associations were found between PPI at 86 dB and PPD. For every unit increase in PPI at 86 dB, the odds of developing PPD decreased by 3% (adjusted odds ratio (aOR), 0.97; 95% confidence interval (CI), 0.93–1.00; P = 0.04; Table 2). In the same model, a significant interaction was found between PPI at 86 dB and depression during pregnancy in association with PPD (aOR, 1.04; 95% CI, 1.00–1.08; P = 0.04; Table 2).

Analyses were then stratified according to depression during pregnancy. Among participants without depression during pregnancy (n = 124), baseline ASR was not significantly different between those with and those without PPD (Fig. 1a). As PPI levels were found to not be normally distributed, non-parametric tests were conducted to assess PPD versus non-PPD group differences in startle magnitude and across prepulse intervals for women who were not depressed during pregnancy. Wilcoxon rank-sum tests revealed that PPI at 78 dB, PPI at 86 dB and global PPI were significantly different between those with and those without PPD (Fig. 1b). Using multivariable logistic regression, the odds for PPD were reduced by 3% for every unit increase in PPI at 86 dB (aOR, 0.95; 95% CI, 0.91–0.99; P = 0.04; Table 3 and Supplementary Table 1). This association was not found for participants with depression during pregnancy (aOR, 1.01; 95% CI, 0.98–1.03; P = 0.62; Table 3). Among those without pregnancy depression, ASR was not significant in any of the models (Supplementary Table 1).

a,b, Boxplots for results of two-sided Wilcoxon rank-sum tests for startle magnitude at baseline (a) and percentage PPI across prepulse intervals (b) among women without pregnancy depression (n = 124). Boxplots consist of the interquartile range (IQR) and the median, and whiskers are 1.5× IQR. Outliers are defined as values >1.5× IQR and are shown as dots. The blue square corresponds to the mean. Dashed line corresponds to 0% PPI.

Full size image

Receiver operating characteristic curve analyses for prediction of PPD using PPI

We used receiver operating characteristic curve (ROC) analyses to determine the ability of PPI at 86 dB to predict PPD in women not depressed during pregnancy (Fig. 2). The area under the curve (AUC) for the crude model (only PPI at 86 dB) was 81.1%. The coordinates of the curve (sensitivity and specificity) are shown in Supplementary Table 2 along with the corresponding values of percentage reduction at each level of sensitivity and specificity. The AUC for the adjusted model (PPI at 86 dB together with covariates) was 90.5%, and the AUC for the covariates-only model (excluding PPI at 86 dB) was 87.1%. There was no significant difference in AUC between the crude and adjusted models (P = 0.07), the crude and covariates-only models P = 0.46), and the covariates-only and adjusted model (P = 0.43).

The percentage AUC for the crude model (only PPI at 86 dB), adjusted model (PPI at 86 dB together with covariates) and covariate-only model (excluding PPI at 86 dB) as an estimate of the ability of PPI at 86 dB to predict PPD for participants without depression during pregnancy.

Full size image

This study set out to investigate if sensorimotor gating, that is, the ability to inhibit ASR following a prepulse signal, measured in late pregnancy, could predict PPD. Among women without depressive symptoms during pregnancy, we found that inhibition of the ASR following a prepulse signal at 86 dB, measured during weeks gestational weeks 35–39, was associated with PPD. Although the addition of PPI did not significantly improve predictive power for PPD in existing models comprising self-report scales, we found that PPI at 86 dB alone had good predictive power for new-onset depression. We have shown the potential of PPI as an objective biological measure to be used in late pregnancy for predicting women at risk for new-onset depressive symptoms postpartum.

Previous studies used PPI measurement as a predictive marker for schizophrenia^54,58,59. Our findings expand the predictive potential of PPI in psychiatric disorders to include PPD. One previous study³⁶ investigating PPI in the postpartum period found reduced sensorimotor gating at 78 dB and 86 dB among women with concurrent PPD. However, the presence of depressive symptoms during pregnancy was not investigated and PPI was measured only during the postpartum period. Our study showed that lower PPI was already present in late pregnancy, before the onset of depressive symptoms in the puerperium. Interestingly, this marker predicted only new-onset PPD and was not predictive of the continuation or amelioration of symptoms among those who were already depressed during pregnancy. Moreover, when we compared the AUC of the covariates-only model with that of the crude model (only PPI at 86 dB), they were not significantly different. In both models, the AUC is considered good, implying the potential of PPI as an objective measure to predict PPD. Although self-report covariates are easily obtained, they can still be subjective and vulnerable to self-report bias. Previous research also suggests that self-report measures alone may be insufficient as there is a reluctance to seek help even among women who recognize that they are suffering from poor mental health⁶⁰. PPI measurement, although comparatively indirect, provides a method to objectively measure physiological changes. This study provides proof of concept for the use of non-invasive physiological measures to predict postpartum-onset depression.

Although the exact underlying pathophysiology is still unclear, sensitivity to stress and hormonal changes may be related to our findings. As previously discussed, gonadal hormones, such as estrogen and progesterone, dramatically increase during pregnancy, which can result in decreased activity in areas associated with stress regulation, such as the HPA axis²². Higher estrogen and progesterone levels have also been associated with decreased PPI in both non-pregnant³⁴ and pregnant women³⁵. Another measure of HPA axis stress system dysregulation, the cortisol-awakening response, was shown to be positively associated with PPI at 86 dB in women who are pregnant⁵⁷. Furthermore, the amount of perceived stress increases during pregnancy and childbirth¹⁹. Although all women who are pregnant are exposed to endocrine and physical challenges, there are individual differences in the capacity to regulate these stressors. Previous studies have shown that stress adaptation is predictive of mental health outcomes²⁰ and that PPI is impaired by stress in both animal and human models^61,62. Future research could investigate if individuals with lower pregnancy PPI are also at an increased risk of a future diagnosis of bipolar disorder. Individuals with early postpartum onset of depression have been shown to be at increased risk of later conversion to bipolar disorder⁶³ and, like depression, bipolar disorder has been shown to be associated with increased sensitivity to both gonadal and stress hormones⁶⁴. At present, increased sensitivity to hormonal changes and stress during pregnancy is primarily recognized when scores on psychological self-report questionnaires meet a certain clinical threshold indicating adverse symptoms. Our results may have captured an element of identification of at-risk individuals that could otherwise be overlooked via self-report methods; some women who do not report depressive symptoms during pregnancy may be showing subtle, yet measurable, physiological markers of hormonal and stress sensitivity. As objective, predictive tools are still lacking in routine care, future studies should validate laboratory findings, such as those in this study, using feasible portable devices or mobile phone applications for use in pregnant women to improve identification and early intervention for prevention of PPD.

This study has strengths and limitations. It is among very few studies that have explored PPI as a tool to determine those at risk for developing psychiatric disorders, and no previous study has investigated risk for PPD. Given the well-characterized cohort in our study, we could adjust for several relevant variables. The sample size is modest compared with other studies using PPI as a predictive marker; further studies with larger sample sizes are, therefore, needed to confirm our findings. In this study, PPI was measured only at a single time point. Collecting PPI measurements at several time points throughout pregnancy until postpartum may give a better picture of the patterns in sensorimotor gating during this vulnerable and dynamic period and may improve prediction of women at risk for developing PPD. Furthermore, although we included a global PPI measurement to take into account all prepulse intervals, we also prioritized isolating each prepulse decibel level in independent analyses to investigate their specific contributions. However, we acknowledge that prepulse levels cannot be viewed entirely independently as they were administered within the same experimental protocol. Future studies should replicate our findings using only PPI at 86 dB to verify our results and further optimize experimental conditions for use in a clinical setting. Sleep could be an important confounder when studying PPI and it might be a limitation that the sleep variable used was a single-item measure of the sleep experienced the night before the ASR test instead of a more comprehensive measure of sleep. However, PPI was measured before birth, that is, before the major changes in sleep duration and quality that arise because of a newborn. An additional limitation was that other potential conditions, such as intellectual disability, personality disorders and substance abuse, were not taken into consideration in this study.

Decreased PPI during late pregnancy was predictive of new-onset depressive symptoms postpartum. Our study encourages further investigation into the potential of PPI as a non-invasive biological measure to identify women who may develop PPD, especially among women who do not display established risk factors, such as previous depression. Further studies are warranted to develop clinically feasible tools for the use of PPI measurement in routine care.

Participants and data collection

This study was conducted as a part of the BASIC study⁵⁶ in Uppsala, Sweden. Women were invited to participate in the BASIC study in association with the routine ultrasound visit between pregnancy weeks 16 and 18. Exclusion criteria included inadequate understanding of Swedish, age less than 18 years, protected identity, bloodborne illness and an unviable pregnancy as diagnosed by routine ultrasound. Women who agreed to participate filled out online questionnaires twice during pregnancy and again at 6 weeks postpartum, including the EPDS. The EPDS is a screening tool consisting of ten self-report questions to detect depression in pregnant and postpartum women⁶⁵, which has been validated in Sweden⁶⁶. The online questionnaires also included sociodemographic information (age, education and employment status) and information related to general, mental, and pregnancy health and lifestyle (for example, smoking habits, prepregnancy weight and depression history). Premenstrual mood symptoms preceding pregnancy were also reported in the survey and classified into PMS or, if symptoms had a negative effect on social activities or relationships, PMDD. Data on SSRI use were collected at week 32. Data on gestational length were retrieved from medical records after delivery. Participants were selected for the substudy based on their reported scores on the week 32 EPDS, and women with EPDS scores of ≥13 were oversampled. The participation rate in the substudies within the whole BASIC cohort study was 48.8% for pregnancy test sessions⁵⁶. Exclusion criteria for the substudy included pregnancy-related conditions such as preeclampsia, gestational diabetes, intrauterine growth restriction and twin pregnancy. The study was approved by the Regional Ethical Review Board at Uppsala University (number 2009/171) and the study procedure was conducted in accordance with ethical standards for human experimentation. Written informed consent was obtained from all participants. Participants in the substudy were compensated with two cinema tickets after the measurements.

Experimental procedure

At approximately gestational week 38, women included in the current substudy came to the research laboratory for measurements. The eye-blink component of the ASR was measured using electromyographic measurements of the orbicularis oculi muscle, which is innervated by the facial nerve⁶⁷, applied on the orbicularis oculi muscle of the right eye. The startle pulse was delivered in earphones in both ears (TDH-39-P; Maico) and a startle system (SR-HLAB; San Diego Instruments) was used to record the startle reflex. A Quest electronics meter was used to calibrate the sound (model 210; Quest Technologies). Two electromyographic electrodes (In Vivo Metric) were used to record the blink response; one electrode was placed below the right eye in line with the pupil and the second electrode was placed 1–2 cm laterally to the first. Furthermore, an isolated ground electrode was placed in the middle of the forehead to function as an electrically inactive site.

The participants were first exposed to 5 min of background white noise of 70 dB, followed by the experiment consisting of 3 blocks of trials. Between each trial, 70 dB of background white noise would resume. Block 1 examined the baseline startle response and had 5 startle pulse trials of 115 dB and 40 ms background white noise. Blocks 2 and 3 included 25 trials in pseudo-random order; 5 trials comprised only startle pulses and in 20 trials a prepulse noise burst lasting 20 ms occurred 100 ms before the startle pulse. The prepulse noise bursts were 72 dB, 74 dB, 78 dB and 86 dB.

The ASR was measured as peak startle amplitudes within 20–150 ms from the onset of the startle pulse. If the peak startle occurred before 20 ms or after 150 ms, if the baseline shift was more than 40 arbitrary units (1 unit equaled 0.076 mV) or if the startle response was 25 arbitrary amplitude units or less, the participant was considered a non-responder and excluded from further analyses.

In conjunction with the ASR measurement, participants were asked to rate their sleep the night before the experiment. The Mini International Neuropsychiatric Interview (MINI) was conducted to investigate symptoms of depression and anxiety during pregnancy⁶⁸.

Study sample

Nine women in the substudy chose not to participate in the PPI measurement, three chose to cease participation because the task was challenging, two participants had technical issues in the measurements and six were found to be non-responders. This resulted in 214 women with complete PPI data, of which 207 completed the EPDS 6-weeks-postpartum outcome measure. An additional 28 participants had missing data on covariates from the questionnaires or interviews (sleep the previous night, prepregnancy BMI, presence of PMS or PMDD, education and/or employment) resulting in a final sample size of 179 women.

For categorization of pregnancy depression, women with major depression according to the MINI at the time of ASR measurement or who scored 12 or higher on the EPDS at week 32 of gestation were considered depressed during pregnancy. For categorization of PPD, participants scoring between 0 and 11 on the EPDS at 6 weeks postpartum were considered as not having depressive symptoms, whereas women scoring 12 or higher were considered to have depressive symptoms, hereafter referred to as non-PPD and PPD, respectively^69,70.

Statistical analysis

Calculation of the PPI is the percentage reduction in peak magnitude of the startle on pulse-alone trials and is calculated using the formula below⁵⁷:

where M_PA is the mean magnitude of pulse alone in blocks 2 and 3 and M_PP is the mean magnitude of prepulse + pulse blocks. This was done for each prepulse level. Global PPI was also calculated using the formula below:

in which M_PP72 is PPI at 72 dB, M_PP74 is PPI at 74 dB, M_PP78 is PPI at 78 dB, and M_PP86 is PPI at 86 dB. Descriptive statistics related to background and pregnancy variables and PPI were used to determine group differences between non-PPD and PPD women. Independent t-tests, Wilcoxon rank-sum, or χ² tests were conducted based on whether the variables of interest were continuous (parametric versus non-parametric) or categorical. To determine the association between PPI and PPD, logistic regressions were performed for PPI at each decibel level and global PPI with PPD as the outcome. The following types of models were included in the logistic regression analyses: a crude (or univariate) model and an adjusted model. The adjusted model controlled for background and pregnancy covariates, including initial startle response value, maternal age, prepregnancy BMI, education level (university versus non-university), employment (employed full-time or part-time versus unemployed/studying/parental leave/sick leave), anxiety at the time of ASR measurement, PMS or PMDD (yes versus no), SSRI use in pregnancy (yes versus no) and sleep the night before ASR measurement. An interaction term between pregnancy depression (yes versus no) and PPI was also included in the adjusted model. Analyses were then stratified according to depression during pregnancy to further test the interaction between pregnancy depression and PPI. Background and pregnancy covariates controlled for in the adjusted model (apart from depression during pregnancy) were also controlled for in the stratified model. Differences in the PPD versus non-PPD group in startle magnitude and across prepulse intervals for women not depressed during pregnancy were assessed using Wilcoxon rank-sum tests and were visualized using boxplots.

The area under the ROC curve was calculated to estimate the ability of PPI at 86 dB to predict PPD for participants without depression during pregnancy. AUCs derived from the crude model (only PPI at 86 dB), adjusted model (PPI at 86 dB together with covariates) and an additional model with only covariates (excluding PPI at 86 dB) were compared based on the bootstrap percentile method^71,72.

Statistical analyses were conducted using the R programming language⁷³ through RStudio⁷⁴ with packages ggplot2⁷⁵, reshape2⁷⁶, pROC⁷¹, cutpointr⁷⁷ and mice⁷². The significance level was set at P < 0.05.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.