Results

Differences between the characteristics of participants in the LBW and NBW groups are summarized in Table 1. Predictably, LBW participants had lower gestational age and birth weight, higher post-delivery hospital admission rates, and longer duration of hospital stays immediately after birth than their NBW counterparts. However, no significant differences in sex, age, or educational achievement at follow-up were observed between the two groups. There were 801 participants for whom risk scale and birth weight data were available, but who did not complete the adult follow-up of the study. Although the average birth weight was approximately 50 g less in participants who were lost to follow-up (W = 725850.0, P = 0.04), LBW participants were no more likely than their NBW counterparts to drop out of the study (OR = 1.22, 95% CI 0.89, 1.65). In either birth weight group, participants who withdrew from the study before the next follow-up (in 2000), but who provided data for inclusion into the cumulative risk scale at the study’s inception (1983) did not have a different mean cumulative risk score (W = 977996.0, P = 0.08). Overall, males were more likely to withdraw from the study than females (52.8% vs. 47.2%, χ ² = 4.95, P = 0.03). Likewise, no differences were observed in average cumulative youth risk scale score between birth weight groups, nor did they differ in the prevalence of any of the individual variables that comprised the youth risk scale. Moreover, the average level of depressive symptoms, and the prevalence of major depressive disorder diagnoses was not different in LBW and NBW groups. Finally, exploratory analyses assessing birth weight for a given gestational age were conducted by creating z-scores of birth weight at a given week of gestation (split by infant sex). However, this variable was not found to be a statistically significant predictor of either outcome measure either as a main effect or as moderated by the cumulative risk scale.

Linear regression analyses revealed a statistically significant interaction between birth weight status and youth risk scale score (P = 0.003), whereby among NBW participants, levels of depressive symptoms in adulthood increased with increasing youth risk exposure (β = 0.16, P = 0.005). However, among LBW participants, a negative association between youth risk scale score and levels of depressive symptoms was noted (β = −0.42, P = 0.025) (Table 2).

Table 2.

Cumulative risk and birth weight group predicting depression symptom score

A statistically significant moderation effect between birth weight status and youth risk scale score was also noted for MDD diagnosis (OR = 0.32, 95% CI 0.12, 0.83, P = 0.02). Among NBW participants, increasing youth risk scale score was associated with increased odds of a diagnosis of major depressive disorder. Indeed, the presence of one additional risk factor increased the odds of a depression diagnosis by 34% (OR = 1.34, 95% CI 1.10, 1.65). A trend toward decreased odds of MDD was noted with increased youth risk scale scores in those born at LBW, though this did not reach statistical significance (P = 0.07) (Table 3, Fig. 1).

Table 3.

Cumulative risk and birth weight group predicting odds of major depressive episode

Because sex differences exist in the prevalence rate of depression, and sex differences may plausibly influence the developmental trajectory and nature of risk exposures that study participants experience, sex was investigated both as a covariate, and as a potential moderator of the previously established interactions. While a main effect of sex was observed to predict depressive symptom scores in adulthood (β = 0.29, P = 0.004) (such that women had greater scores than men), it did not significantly influence the magnitude or direction of the interaction between birth weight group and cumulative risk scale (β = -0.56, P = 0.005). Likewise, while women had a greater prevalence of major depression diagnosis (OR = 1.77, 95% CI 1.19, 2.65), this effect did not substantially diminish the observed risk by birth weight interaction (OR = 0.33, 95% CI 0.12, 0.86). When sex was considered as an additional moderator (including all nested interactions in the model), there was no significant interaction between birth weight, sex, and cumulative risk to predict either depressive symptoms (β = -0.37, P = 0.35) or prevalence of MDD diagnosis (OR = 0.54, 95% CI 0.07, 4.07).

Discussion

Using data from a large sample of youth that was representative of the population of Canada’s most populous province (Ontario), exposure to an increasing number of adverse events during childhood and adolescence was associated with an increased risk of major depression in NBW adults, but not in those born at LBW. In fact, exposure to increasing levels of adversity in childhood/adolescence among LBW individuals was associated with lower levels of depression as adults. These findings support a difference in vulnerability to stressors in childhood/adolescence based on prenatal stress exposure, such that those born at LBW may develop an increased resilience to stress, while those exposed to less prenatal stress (born at NBW) may higher levels of depressive symptoms in response to later stressors.

These findings are consistent with the match–mismatch hypothesis, whereby prenatal stress (indexed by LBW) provided some aspect of adaptive benefit upon exposure to subsequent environmental adversity in childhood/adolescence,^{Reference Santarelli, Lesuis and Wang16} at least with respect to depression in adulthood. The mismatch phenomenon was initially investigated in the context of cardiovascular health,^{Reference Cleal, Poore and Boullin47} body composition,^{Reference Yajnik, Fall and Coyaji48,Reference Gluckman, Hanson, Cooper and Thornburg49} and related indices of somatic health. More recently match–mismatch relationships have been explored in the context of stress appraisal/management.^{Reference Santarelli, Lesuis and Wang16,Reference Santarelli, Zimmermann and Kalideris18} It has been proposed that this early exposure to stress may influence hypothalamic–pituitary–adrenal (HPA) axis development such that subsequent exposure to stress fails to elicit the chronic overactivation commonly associated with postnatal stress exposures, a phenomenon that may be mediated by changes in hippocampal volume and/or glucocorticoid receptor expression.^{Reference Santarelli, Zimmermann and Kalideris18,Reference Lajud, Roque, Cajero, Gutiérrez-Ospina and Torner50} Indeed, epigenetic alteration of genes encoding crucial stress-response proteins has been observed in animal models of early-life stress exposure,^{Reference Murgatroyd and Spengler51,Reference Morley-Fletcher, Mairesse and Van Camp52} as well as in children who were exposed to significant childhood stress.^{Reference Romens, Mcdonald, Svaren and Pollak53} It may therefore be the case that stress-induced adaptations during the perinatal period can moderate, influence, or limit subsequent stress effects in later periods of developmental plasticity.^{Reference Gröger, Matas and Gos54} However in nonadverse environments, people born at low birth weight may simply have a constitutively higher risk for depression compared to their normal birth weight peers,^{Reference Wojcik, Lee, Colman, Hardy and Hotopf14} who are more able to thrive in an enriched environment due to their lack of early-life stress exposure, although this phenomenon is not conclusively established. There may be a developmental cost associated with stress-induced adaptation exhibited by LBW infants, and in ideal environments, these changes may prove to be deleterious relative to NBW infants. It may be the case that behavioral or emotive strategies for instance, introversion and accentuated risk aversion, are useful for mitigating the adverse effects of high stress environments, but may not be beneficial otherwise.

As our sample of low birth weight participants included very few survivors born at the very low (<1500 g; 0.5% of our sample) and extremely low birth weight (<1000 g; no participants) categories, it is likely that our LBW group was exposed to relatively moderate (rather than severe) prenatal adversity. For this reason, the findings of this study may be attributed to the severity of the prenatal stressor. It may be the case that whether prenatal stress ‘programs’ later depression consistent with the match–mismatch or cumulative risk hypothesis may depend on the degree of prenatal stress they are exposed to. For example, the data from this study are more consistent with the match–mismatch hypothesis whereas the results of a previous study of ELBW survivors (almost certainly exposed to higher levels of prenatal stress) born in the same era and geographic region were consistent with the cumulative risk hypothesis of psychopathology (Van Lieshout et al., 2018). In other words, being born at LBW may confer an adaptive benefit that is eliminated in cases of exposure to more severe perinatal adversity.^{Reference Van Lieshout, Boyle, Saigal, Morrison and Schmidt55–Reference Lindström, Lindblad and Hjern57}

Taken together, the findings of this study and previous investigations suggest that there may be a gradient effect of prenatal stress exposure, whereby more moderate stressors can confer some relative adaptive benefit to subsequent environmental adversity, while more adverse prenatal exposures are so developmentally damaging that the opportunity for adaptation is lost and any subsequent stress exposures may combine in a cumulative manner.^{Reference Nederhof and Schmidt20}

Nonetheless, the findings of this study must be considered in the context of its limitations. This study utilized a historical cohort born several decades ago. Although the prevalence of depression in adulthood is likely to remain stable over time, perinatal care has evolved and must be considered when generalizing these findings to more contemporary cohorts. The sample sizes of the two birth weight groups are quite different, a likely consequence of population-representative sampling and the relative infrequency of low birth weight. Because of this, the frequency of the dichotomous MDD outcome was relatively low in the low-birth weight group (<10 positive cases). Studies that specifically recruit LBW participants may be more adequately powered to precisely estimate prevalence and strength of association between LBW and later stress exposures. Additionally, there was some participant attrition over the course of the study. Although males were more likely than females to leave the study, there were no significant sex differences within each birth weight group, comparing those participants who were included in the analyses and those who were lost to follow-up. Moreover, although our cumulative risk scale attempted to quantify the most common environmental stressors faced by youth, more rare traumatic experiences may not be well represented by these findings. Data that comprised the cumulative risk scale was also gathered at the 1983 visit, and as such we were not able to index changes in exposure in participants longitudinally over time. Although unmeasured covariates may contribute to the relationship observed in this study, it is thought that they would be collinear with the risks included in the cumulative risk scale, given their ubiquity. While this may obscure the role that individual, specific risks play in the development of mental health problems, it serves to quantify risk exposure holistically, offering a more generalizable estimate of the construct of “childhood risk exposure”. It may also be the case that the observed findings are due to unmeasured specific prenatal exposures not available in this data set. Likewise, prenatal risks such as gestational age classifications may also partially contribute to the observed association, but are highly collinear with low-birth weight status in this study and so could not be adjusted for statistically. Nonetheless, models exploring the relationship between z-scores of birth weight at a given gestational age (stratified by sex) and the adult depression outcomes were conducted and statistically significant interactions were note noted with the cumulative risk scale nor did they independently predict the outcomes of interest. While it is unclear if this is due to sample size/statistical power or another explanation remains unclear. However, this type of measure might index the “growth rate” of an infant during gestation, assessing differences in body size at time of birth while controlling for gestational age, such that an infant that is born full-term but in the 25th percentile of birth weight could be considered as having received differential early-life adversity than a premature infant at 30 weeks gestation, who is nonetheless at the 50^th percentile for their younger gestational age.^{Reference Stock, Schmid and Griesmaier58} Reduced birth weight at a given gestational age is frequently associated with intrauterine growth restriction,^{Reference Lang, Lieberman and Cohen59} although it is not solely diagnostic of the condition. The lack of association of this measure with the depressive symptom outcomes could suggest that absolute birth weight (which is intrinsically linked to gestational age) might index a different type of early-life adversity (and predictive of the consequences of that adversity) than relative differences in birth weight at a given gestational age. That is to say, an infant born at low birth weight at a younger gestational age (ie; 50th percentile for age) may encounter a perinatal stress of approximately equal severity to that of an infant born at low birth weight but at an older gestational age (25th percentile for age).

Future research may seek to more precisely elucidate mechanistic differences between these prenatal risks. Finally, although this study utilized well-validated measures of adult depression, a comprehensive clinician-administered diagnostic interview (that may have identified other diagnoses responsible for depressive symptoms) was not available.

Despite these limitations, this study provides new evidence in support of the match–mismatch hypothesis for developmental adversity in a large, representative, and longitudinally followed cohort. It also suggests that the mechanisms underlying the interaction of early-life stress, later adversity, and adult mental health may be more nuanced than previously described, and that a complex interplay could exist between different levels of prenatal adversity and later chronic diseases like depression. Further mechanistic studies in nonhuman animal models is required to examine the prenatal circumstances affecting how later postnatal stressors influence brain development and mental disorder risk. Future epidemiologic work in humans should attempt to replicate these findings so that we can develop a more complete understanding of how prenatal stress exposures affect later mental health, as well as which of these individuals require more intensive monitoring and early intervention, and who does not.