Labor Pain, Analgesia, and Postpartum Depression: Are We Asking the Right Questions?

Not everything that counts can be counted, and not everything that can be counted counts.

–William Bruce Cameron

Pregnancy and the postpartum periods are major life events that can trigger anxiety, depression, and adjustment disorders in susceptible individuals. Acute pain is also a risk factor for persistent postsurgical pain, which can be associated with depressed mood. Given these links, questions about the relationship between labor pain, epidural labor analgesia (EDA), and postpartum depression have been raised. Potential, yet unverified, mechanisms may include stress mitigation, risk reduction for persistent or chronic pain, and possibly emotion regulation. Recent studies on this topic have been primarily observational and conflicting in results. Some prospective and retrospective cohort studies have identified a protective effect of labor analgesia on postpartum depression.^1–4 In contrast, one study showed no effect of labor analgesia on postpartum depression,⁵ while another study showed a harmful effect for women who planned to avoid, but ended up using labor epidural analgesia.⁶ The relationships among pain, emotion, and cognition are complex and are influenced by culture, expectations, and the quality of social support networks. Most existing studies have overlooked these complex interrelationships among pain, social support, fear of childbirth, and antenatal depression due to the limitations of their respective data sets. Furthermore, these studies have used simple analytic strategies to tease out the relationships among pain, analgesia, and depression in the perinatal period as well as the role of key confounders. These analyses disregard the multifaceted interrelationships among pain, analgesic efficacy, social support, fear of childbirth, and antenatal depression.

To account for the factors that may explain the influence of epidural analgesia on postpartum depression risk, Eckerdal et al⁷ utilized conceptual path analysis to interpret findings from a longitudinal cohort study of 1503 pregnant women in Sweden. The primary outcome was Edinburgh Postnatal Depression Scale (postpartum depression [PPD]) score ≥12 at 6 weeks postpartum, and the primary predictor of interest was the use of EDA. The cohort included primiparous women with spontaneous labor. It excluded women with preterm labor/gestation and intrauterine fetal demise. Medical records and patient self-report were used to collect data throughout pregnancy and 6 weeks after delivery. Multivariable logistic regression models adjusted for age, fear of childbirth, and antenatal depressive symptoms.

To help choose which variables to adjust for when assessing the relationship between EDA and PPD, the authors constructed a directed acyclic graph (DAG). A DAG is a diagram that displays all variables of interest and their hypothesized relationships with the stipulation that all relationships are unidirectional and indicate a causal relationship in direction of the arrow. Lack of an arrow connecting 2 variables indicates no (either assumed or tested) relationship between them. Although a DAG can be modified as data are analyzed, the original construction is often based on previous knowledge and biological reasoning. In the Eckerdal et al⁷ article, the authors start with an a priori constructed DAG in their first figure, outlining hypothesized relationships between variables and pathways based on published literature, or by their bivariate analysis revealing factors with associations P < .05 with both EDA and PPD at 6 weeks postpartum. This cutoff of 0.05 is important in interpreting their results, as explained below. Next, the authors used generalized structural equation modeling to help deconstruct potential explanatory factors in the EDA and PPD relationship and depicted this step in their second figure of their conceptual path analysis. This step attempted to separate the direct and indirect effects of the complex relationships among the various factors in the model, namely age, prenatal psychological status variables, and childbirth-related parameters. In the final step (Figure 4 in Eckerdal et al⁷), final effect sizes of the estimated path analyses are presented.

The results from their final multivariable logistic regression showed that EDA was not associated with PPD scores at 6 weeks postpartum, after confounder adjustment (adjusted odds ratio = 1.22, 95% confidence interval, 0.87–1.72). As with any observational study, it is important to ask whether confounding was sufficiently addressed. In Eckerdal et al,⁷ it would be difficult to believe that it was, since only 3 variables (fear of childbirth, antenatal depressive symptoms, age) were adjusted for, even though they were carefully chosen. There are likely many other variables, measured or unmeasured, that are associated with both EDA and PPD that were omitted in the model. In general, one should not rely on P < .05 when deciding whether to include a variable as confounder. Even small amounts of confounding across several variables can bias results, and a 0.05 cutoff is likely too conservative for variable selection criteria.⁸ In fact, some recommend setting the noncandidate inclusion level to 0.15 to retain meaningful confounders.⁸ For example, in their cohort, the authors found that lifetime history of smoking was associated with EDA use (P = .028) and more weakly with 6-week PPD (P = .088); using the 0.15 level specified above, this variable should have been considered for inclusion in the model. Smoking status can be associated with other social determinants of health that increase risk for PPD as well as propensity to use EDA. In extreme cases, the consequences of incomplete confounder adjustment are to imply causal relationships where they do not truly exist or to conceal true effects. In a similar way, the omission of preterm birth and fetal demise (risk factors for PPD) and induced labor (a risk factor for higher pain and potential increased EDA utilization) make the results uninterpretable in the context of these high-risk populations.

Most often, researchers only assess the “total” effect of exposure on outcome, ignoring the possible explanatory factors, or intermediary variables (“mediators”). To their credit, Eckerdal et al⁷ do consider possible factors that may explain how EDA might affect PPD. In doing so, they attempted to isolate the “direct” effect of EDA on PPD, independent of any other variables as well as the “indirect” or “mediation” effects of EDA on PPD. Under some assumptions (eg, no interaction between the exposure and mediator), the sum of the direct and indirect effects would be expected to equal the “total” effect. Considering a variable to be a true mediator implies that the variable occurs after the exposure and before the outcome and that both the exposure–mediator and mediator–outcome relationships are causal.

After deciding which variables were most likely to be either confounding variables or potential mediators of the EDA versus PPD association, Eckerdal et al⁷ fit a generalized structural equation model (termed a path analysis) that enabled them to simultaneously estimate all of the relationships of interest. From this model, they identified no direct effect of EDA on PPD, consistent with their main analysis results, but some evidence of an indirect effect through childbirth experience (Figure 4 in Eckerdal et al⁷). Their chosen methodology allows a substantively more nuanced interpretation of the complex relationships between variables, even though many assumptions (mostly untestable) are being made. For example, the directionality of all relationships was assumed, potentially complex interactions were assumed not to exist, and confounding was assumed to be well controlled. Results need to be carefully interpreted in that context.

Importantly, in this Swedish cohort, women who elect to use EDA are frequently distinguished by a fear of childbirth. This cultural characteristic is distinct and limits the generalizability of the findings to other settings, particularly if attempting to reconstruct a culturally representative pathway analysis. In contrast to Scandinavia, EDA in the United States is a commonly utilized form of pain management in labor, and its use is not a reliable marker of fear of childbirth. Further underscoring the cultural underpinnings that influence pain, analgesia utilization, and childbirth experiences, in general, are consecutive studies from Chinese cohorts suggesting strong, protective relationships between epidural analgesia and postpartum depression symptoms.^1,9 However, in these studies, it is possible that the strength of the relationships is overestimated due to lack of ability to account for third variables that could not be measured (eg, the women in these studies who attended childbirth classes were more educated on their pain control options including EDA and were more likely to use it; simultaneously, because they were more educated, they may have had a better understanding about what to expect during labor, delivery, and postpartum periods, and this improved understanding could have protected against PPD). Because of these cultural distinctions, it will be important to differentiate between universal and culturally unique patterns of EDA utilization in future studies.

In addition, the study was unable to assess pain before, during, or after labor and delivery, and thus it cannot shed adequate light on the complex interrelationships between these variables and PPD. EDA should not be considered a proxy for good analgesia because not every epidural catheter works perfectly, and pain and its relief are not all-or-nothing constructs.² To this point, 1 observational study in laboring women showed high variability for pain improvement after EDA initiation, with some women experiencing better pain improvement than others.² Further, epidural catheters may be placed for other reasons besides pain control. Only by characterizing the pain experience during the perinatal period can the relationship between analgesia (ie, adequate control of labor pain) and PPD be sufficiently assessed.

It is worth taking a step back to evaluate well-established risk factors for perinatal depression, to better assess the potential role for labor pain and analgesia in these associations. Much evidence indicates that adverse early life events (eg, negative childhood events, sexual abuse, physical abuse) sensitize the central nervous system circuitry to stress and emotion. That sensitization increases susceptibility to later anxiety, depression, or chronic pain after subsequent exposures to physical or psychological injury. Given this framework, some may question the biological plausibility with which effective EDA—a brief intervention targeting a relatively episodic event in the context of a lifetime perspective—could mitigate the influence of powerful early life events on postpartum mental health. This skepticism seems particularly justified considering that perinatal life stress continues beyond the typical duration of labor pain and analgesia, and encompasses postpartum physical changes, sleep disturbances, life with a newborn, and lactation or breastfeeding adjustments—elements that are not modifiable by effective labor analgesia alone. Other well-established risk factors that place women at higher risk for postpartum depression¹⁰ are also not modifiable by labor analgesia, including history of perinatal depression, any lifetime history of depression or bipolar disorder or any psychiatric disorder, family history of depression or other mental illness, stressful life events during pregnancy or postpartum, childbirth complications such as preterm delivery or neonatal medical conditions, lack of positive feelings about the pregnancy, lack of emotional or social support, and substance abuse problems. With respect to the relative contribution of labor analgesia on risk for depression, cultural differences in labor analgesia utilization have not necessarily translated to differences in depression risk. In low-resource countries where labor analgesia utilization is low (1%–12%),¹¹ perinatal depression rates remain 11%–21%,¹² consistent with high-resource countries with higher labor analgesia utilization rates. What is clear is that the etiology of perinatal depression is multifactorial and not necessarily attributable to a single factor alone, such as labor analgesia.

On the other hand, it is possible that the event of labor and its associated pain is physically and emotionally distressing enough to serve as either the primary nidus of sensitization or as a secondary injury, after which additional stressors “tip” a susceptible individual over the edge toward depression. The pain experience requires a cognitive process that involves an assignment of emotional valence—a process that may trigger a physiologic stress response and the associated psychosocial sequelae, such as anxiety and depression. In 1 observational study, peripartum stress hormone levels of serum cortisol were reduced among women who received EDA, compared to women who did not receive any analgesia; those receiving EDA also had lower rates of postpartum depression at 6 weeks postpartum.³ However, it should be noted that stress hormone correlations for postpartum depression are not well established, and several studies have failed to establish relationships between serum and urine cortisol and postpartum depression.¹³ Moreover, for some women, prenatal systematic birth preparation by which the cognitive appraisal or valuation of labor pain can be trained is associated with reduced perception of labor pain.¹⁴ Taken together, these previous data suggest the importance of considering the physical and psychosocial correlates of labor pain in investigating the link between pain and postpartum psychiatric outcomes.

Where do we go from here? Observational research continues to be the best available approach for these questions. Indeed, a randomized trial is not feasible or practicable, particularly in high-resource settings where EDA has been established as a standard therapy for labor pain management. In such settings, randomization to no epidural analgesia or to “sham” epidural analgesia would seem unethical, and randomization to inferior analgesia such as intravenous patient-controlled analgesia would be limited by anticipated high crossover rates that limit the ability to interpret results, even with intent-to-treat analyses.

The Hill criteria¹⁵ can provide some guidance: these criteria are a set of points that help to establish epidemiological support of causality between a “cause” and “effect” of interest. The criteria include things like strength (small associations do not necessarily mean cause and effect), plausibility (a plausible biological mechanism to explain the relationship can be offered and/or tested), reproducibility (findings consistent between different investigators, across different populations), and biological gradient (greater exposure should generally lead to greater incidence—or lower incidence, if the inverse proportion is observed). With respect to strength, it is worth noting that in the existing published literature, although statistical significance for the relationship between EDA and PPD may be reached, the strength of these associations are weak, with odds ratios consistently below 4.^1,4,6 Weak associations are common in observational research and likely attributable to bias rather than causality; weak associations have the potential to engender “false alarms,” and have reduced credibility in several areas of health sciences research.^16,17 Therefore, although observational methods continue to be the most practicable for these questions about pain, EDA and PPD, observational approaches that go beyond epidemiology are necessary to draw meaningful data and conclusions.

Using the Hill criteria structure, more targeted questions on the topic of labor pain, analgesia, and PPD can be raised. What is the biological plausibility of this relationship? How can that potential biological explanation be tested and quantified? What other mediators or moderators should be evaluated? Do the literature inconsistencies noted in the relationship between EDA and PPD across cultural and geographic spectra mean there are subpopulations for whom EDA can be more protective? Conversely, are there subpopulations for whom EDA could be harmful? Do women with depression experience pain differently during labor? Does this difference drive EDA utilization? Does more labor pain lead to higher depression scores? Indeed, 1 study has noted a linear relationship between greater improvements in labor pain with EDA and lower postpartum depression scores.² How do birth plans and expectations interact with these relationships? Orbach-Zinger et al⁶ noted a protective effect for women desiring EDA and delivering with EDA for the PPD, whereas there was an increased risk for PPD among women intending to deliver without EDA but who eventually received it.

The idea that EDA could have protective benefits for PPD is appealing because compared to other PPD risk factors such as social support and preterm birth, labor pain seems readily treatable. However, a solid case of any relationship is elusive. Based on the existing evidence, it is also premature to exclude any potential benefit of EDA in preventing PPD—and perhaps, the EDA-PPD inquiries are not the best questions to be asking. Rather, questions about the role of pain itself—rather than the decision to use or not to use EDA—persist. Researchers who are interested in this matter are charged with addressing basic issues about effect size, reproducibility, biological gradient, and plausibility, among others. Mechanistic investigations, in both human and preclinical models, should be pursued, such as studies on psychophysics, neural function/connectivity, or studies at the cellular, subcellular, or molecular level. Moving forward, we suggest that a short-term strategy from the scientific community should include disseminating the results of exploratory investigations, so that research priorities can be established, and so that more informed study methodology can be formulated in the future. Clearly, these questions remain important; ongoing investigations will clarify the role that perinatal pain has in long-term sequelae associated with perceived birth trauma, reveal novel pathways underlying perinatal pain and depression, and advance knowledge in individualized risk stratification for both perinatal pain and depression.

GLOSSARY

DAG

directed acyclic graph

EDA

epidural analgesia

PPD

postpartum depression