Introduction

Preterm birth (< 37 weeks; PTB) and/or small for gestational age (sex-specific birth weight and/or length < 2 SD for the gestational age; SGA) are the most common risk factors for neonatal morbidity and mortality [1, 2]. Both preterm and SGA births are at higher risk for neurodevelopmental disorders, such as attention-deficit/hyperactivity disorders (ADHD) during childhood and adolescence [3,4,5,6,7,8,9,10]. Moreover, these risks increase with a higher degree of severity of PTB or SGA [11, 12]. However, studies investigating the association between PTB or SGA and psychiatric disorders with onset commonly in late adolescence and early adulthood, such as depressive disorders, and anxieties, are limited [13, 14]. Furthermore, there is convincing evidence stating that PTB has a higher frequency of in-utero growth failures compared to term-born peers (> 37 weeks) [15], and PTB in combination with SGA, may pose more adverse sequelae. It is estimated that the global incidence of both preterm and SGA births is ~ 1.5 million [16], still, the magnitude of combined effects of PTB and SGA on neurodevelopmental and psychiatric disorders is under-explored.

Studies so far have shown a higher risk for a range of neurodevelopmental and psychiatric disorders in offspring born by caesarean delivery [17, 18]. However, there is limited information on the risk of psychiatric outcomes among those born preterm and SGA in spontaneous vaginal births.

Being born large for gestational age (LGA) has also, albeit less studied and less consistent than for SGA, been associated with a higher risk for the development of psychopathology [19, 20].

The risk of neurodevelopmental disorders, such as ADHD, is more prevalent in boys than girls [21, 22]. However, studies investigating the effect of sex on the relationship between PTB/SGA/LGA, and neurodevelopmental disorders are limited, especially in spontaneous births [23,24,25].

In addition, it remains unclear whether the observed association between the degree of prematurity or birth size status and the likelihood of developing neurodevelopmental or psychiatric disorders, as reported by earlier studies, might be confounded by unknown factors, such unmeasured shared familial factors, including genetics, household-level factors and lifestyle. Previous studies have utilized quasi-experimental designs, such as discordant sibling pairs, or maternal polygenic risk scores in attempts to control for these unmeasured familial factors [26,27,28].

We investigated the associations between birth outcomes, that is, PTB/SGA/LGA, separately and as combined risk, and a wide spectrum of neurodevelopmental and psychiatric disorders, and their psychotropic medication, in singleton individuals born of spontaneous births. We also conducted sibling pair analyses to investigate whether these associations could be explained by unmeasured familial confounding. Moreover, we investigated whether the association between birth outcomes and neurodevelopmental and psychiatric disorders differed by sex.

Methods

Study population and data sources

All live singleton spontaneous delivery births in Finland from 1996 to 2014 were included in this population-based registry cohort study (819 764 births including 299 331 sibling pairs), and follow-up until December 2018. All data were retrieved from nationwide registers: The Medical Birth Register (MBR), The Finnish Register on Reimbursement Drugs (RRD), and The Finnish Care Registers for Health Care (HILMO). Information from the different registers was linked using personal identification numbers (PIN) assigned to all Finnish citizens and permanent residents.

The complete study was approved by the relevant data protection authorities and ethical review committees in Finland and Sweden. According to Finnish regulations, informed consent by participants was not required, so the individuals included in this study were not contacted. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline and data analysis was conducted from March 2021 to June 2023.

Main exposures

Data on main exposures included gestational age in weeks and size for gestational age from MBR. Gestational age was assessed based on ultrasonography-based estimates at 11–13 weeks. The MBR collects information on the last menstrual period, which is cross-referenced in cases of suspected error in gestational age.

Gestational age was categorized as extremely preterm (< 28 weeks), very preterm (28–31 completed weeks), moderate-late preterm (32–36 completed weeks), term (37–41 completed weeks), and post-term (≥ 42 weeks). SGA/LGA was defined as birth weight and/or birth length, below/above two SDs from the gestation and sex-specific mean in the Finnish population [29], based on the International Societies of Pediatric Endocrinology and the Growth Hormone Research Society [30]. Children who were neither SGA nor LGA were considered appropriate for gestational age (AGA).

Outcomes

We used ICD-10 codes to ascertain neurodevelopmental and psychiatric disorders from HILMO in individuals between 1996 and 2018 (Table 1), and dispensation of psychotropic drugs (ATC codes N05 and N06) from RRD between 1996 and 2014.

Table 1 List of included neurodevelopmental and psychiatric disorders along with ICD-10 codes, number of diagnosed children, and the corresponding proportions that were identified (until 2018) out of the estimated number of cases that would have received a diagnosis before 23 years of age (requiring the youngest cases to be followed up until 2036)
Full size table

Covariates

Based on a directed acyclic graph (Figure S1) the covariates were: birth year of the child, sex of the child (boy/girl), maternal age at child birth, parity (0 or ≥ 1), maternal cohabitation status at child birth (yes/no), maternal country of origin (Finland or other), maternal occupation (upper white-collar worker, lower white-collar worker, blue-collar worker, other status), smoking during pregnancy (yes/no) from MBR, and maternal obesity (ICD-10 codes; E65-66, yes/no), maternal in-patient (from 1987) and out-patient (from 1998) psychiatric history (yes/no), maternal systemic inflammatory disease (ICD-10 codes; M30-M36; yes/no) from MBR and HILMO, and maternal use of psychotropic medication during pregnancy (ATC codes; N05 or N06, yes/no) from RRD.

Statistical analysis

Child and maternal characteristics were summarised according to birth outcomes (PTB/SGA/LGA). Baseline characteristics were compared between gestational groups and between birth size status groups using the χ2 test for categorical variables.

We used Cox proportional hazards models to investigate the relationship between birth outcomes and neurodevelopmental and psychiatric disorders in individuals, unadjusted (crude) and after adjusting for confounders, listed above. The proportional hazards assumptions were tested mainly by evaluating the cumulative incidence of outcome in the different exposure groups across age of the child (Fig. 1). The single exposure analyses (PTB or SGA/LGA) disregarded information about a possible second exposure (PTB and SGA/LGA). Potential interactions between PTB and SGA/LGA were examined on multiplicative scales in a full-factor model with the aforementioned covariates. Due to the number of analyses conducted, and to control for Type 1 errors, a statistical significance threshold was set at p = 0.001 in the main covariate-adjusted analysis (= 0.05/50; Bonferroni-correction for 50 tests, based on 11 outcomes and 6 exposures that in-part are dependent on each other).

Fig. 1
figure 1

Cumulative incidence of autism spectrum disorder (ASD), ADHD, specific developmental disorder (SDD), intellectual disability (ID) across age of the born children stratified by exposure to preterm birth categories, or born small for gestational age (SGA)

Full size image

Sensitivity analyses were conducted to investigate (i) whether birth outcomes are associated with the individuals’ use of prescribed psychotropic drugs, (ii) sex-specific effect sizes, (iii) association with mood and anxiety diagnoses set from 10 years of age studying those followed 10 years or longer, and (iv) effect sizes of F98 subgroups. Here, two-sided P < 0.05 was considered statistically significant.

To account for possible confounding by shared familial factors, we conducted a sensitivity analysis comprising sibling pair analyses. The exposures were, due to sample sizes and effect sizes in the full cohort, limited to PTB and SGA. We estimated the risk of specific F-diagnosis in the second sibling, given the exposure or not exposure for the first and second siblings to birth outcomes. Unexposed first and second sibling pairs were used as references. We used two models of adjustments. In Model 1, we adjusted for all covariates as discussed above plus intra-pregnancy interval. In Model 2, we additionally adjusted for the presence of the corresponding F-diagnosis in the first sibling (yes/no) along with those covariates adjusted in Model 1.

Hazard risk ratios (HRs) with 95% confidence intervals (CIs) were reported as measures of effect size.

All statistical analyses were performed using SAS version 9.4 (SAS Institute Cary, USA).

Results

Of the 819 764 singleton births (48.9% girls and 51.1% boys), 35 259 (4.3%) were born preterm, 21 977 (2.7%) were born post-term, 22 969 (2.8%) were born SGA, and 18 258 (2.2%) were born LGA. In our included cohort, 137 163 (16%) were diagnosed with a neurodevelopmental or psychiatric disorder between 1996 and 2018 (Table S1–S4).

In comparison to those born at term, individuals born preterm were more likely to have mothers who conceived at a younger age, were more likely to be obese before pregnancy, continued smoking during the whole pregnancy, had a history of psychiatric illness, and had lower socioeconomic status (occupation level) (Table S1). In addition, those born preterm were born smaller for gestational age compared to their term-born peers.

Gestational age

Compared to individuals born term, those born extremely preterm (HR = 3.39 [95%CI, 3.13–3.68], very preterm (HR = 2.04 [95%CI, 1.91–2.16], and moderate-late preterm (HR = 1.23, 95%CI, 1.19–1.26) but not post-term (HR = 0.98, 95%CI, 0.95–1.02) had higher risk for any neurodevelopmental or psychiatric disorder (Table 2). When we investigated the individual F-diagnoses separately, statistically significant effect sizes for all gestational age categories but post-term were found for anxiety disorders (HRs ranging from 1.20 to 2.35 across the gestational age categories), intellectual disabilities (HRs ranging from 2.23 to 10.7), SDD (HRs ranging from 1.62 to 8.91), autism spectrum disorders (ASD) (HRs ranging from 1.25 to 4.89), ADHD (HRs ranging from 1.47 to 4.72), and other behavioural and emotional disorders (HRs ranging from 1.48 to 4.69). Extremely PTB was associated also with mood disorders, personality disorders and conduct disorders at statistically significant effect sizes (HRs ranging from 1.98 to 7.50) (Table 2, Fig. 2).

Table 2 Risk for neurodevelopmental and psychiatric disorders in individuals as a function of gestational age and size for gestational age (All singleton spontaneous delivery births between 1996 and 2014 in Finland followed until 2018, N = 819 764)
Full size table
Fig. 2
figure 2

Adjusted hazard ratios (HRs) for neurodevelopmental and psychiatric disorders in relation to gestational age and size for gestational age (All live spontaneous singleton pregnancies born between 1996 and 2014 in Finland followed until 2018). Hazard ratios are represented on the x-axis, while gestational age and/or size for gestational age are represented on the y-axis. To achieve a more appropriate alignment of hazard ratios concerning intellectual disabilities, SDD, and personality disorders, we have plotted them on a separate x-axis range. AGA appropriate for gestational age; SGA small for gestational age; LGA large for gestational age; SDD specific developmental disorders; ASD autism spectrum disorders; ADHD attention-deficit/hyperactivity disorders. Extremely preterm is defined as < 28 weeks; very preterm is defined as 28 to 31 completed weeks; moderate to late preterm is defined as 32–36 completed weeks; term is defined as 37–41 completed weeks; and post-term is defined as ≥ 42 weeks. p-values that survive multiple comparison correction (p < 0.001) are marked with an asterisk (*)

Full size image

Size for gestational age

Compared to individuals born AGA, those born SGA (HR = 1.29 [95%CI, 1.26–1.33]) had a statistically significant higher risk for any neurodevelopmental or psychiatric disorder (Table 2). When we investigated the individual F-diagnoses separately, individuals born with SGA had higher risks for all disorders studied here except for psychotic, eating, sleeping, or conduct disorders, with the largest effect sizes being for intellectual disabilities, SDD, and other emotional and developmental disorders (HRs ranging from 1.58 to 3.62). Conversely, individuals born with LGA had a mildly higher risk for SDD (HR = 1.17 [95%CI, 1.10–1.24]) (Table 2, Fig. 2).

Combined birth outcomes

Within each PTB category, at least 74% of the newborns were AGA (Table S1). Compared to individuals born at term and AGA, being exposed to both preterm and SGA/LGA births was associated with a higher risk for any neurodevelopmental or psychiatric disorder (Table 2, Fig. 2). Statistically significant effect size by combined exposures was found for individuals born extremely/very preterm and being SGA (HR = 3.16, 95%CI, 2.82–3.54), and this effect size was larger than that of exposure to extremely/very preterm (HR = 2.38 [95%CI, 2.26–2.50]) or SGA (HR = 1.29 [95%CI, 1.26–1.33]). Concerning individual F-diagnoses, births being both extremely/very preterm and SGA had higher risks for SDD (HR = 7.55 [95%CI, 6.61–8.62]) compared to births being extremely/very preterm or SGA. These two exposures did not modify each other’s effect on SDD on a multiplicative scale (χ2SDD = 0.64, p = 0.43). Likewise, individuals born moderate-late preterm and being SGA implied risk for SDD (HR = 2.43 [95%CI, 2.18–2.71]) higher than that of either of the two exposures, again without effect modification (χ2 = 0.78, p = 0.38). LGA, on the other hand, showed a protective effect in births being extremely/very preterm such that combined exposures implied a lower risk for SDD (HR = 3.80 [95%CI, 3.01–4.80]) and other emotional and behavioural disorders (F98, HR = 2.04 [95%CI, 1.72–2.43]) compared to births being preterm or LGA. Here, the two exposures did modify each other’s effect on SDD antagonistically (interaction χ2SDD = 8.89, p = 0.0029) but there was no effect modification on F98 (χ2F98 = 2.0, p = 0.16).

The unadjusted HRs for associations between birth outcomes and neurodevelopmental and psychiatric disorders are presented in Table S5.

Sensitivity analyses

Psychotropic medications

In support of the associations between birth outcomes and F-diagnoses, the risk of dispensation of psychotropic medications was, for each individual drug category studied, statistically significantly higher in individuals who were born extremely preterm (HRs ranging from 2.49 to 3.68) and very preterm (HRs ranging from 1.51 to 2.42) and also in individuals who were born SGA (HRs ranging from 1.14 to 1.51), where effect sizes were higher for anxiolytics and sedatives (N05) and psychostimulants (N06B) than for antidepressants (N06A). Further, the risks of dispensation of N05 and N06B were higher in individuals who were exposed to both moderate-late preterm and SGA births compared to those exposed to only one of these two birth outcomes. LGA, however, had no detectable effect on medication, neither alone nor in combination with PTB (Table 3).

Table 3 Hazard ratios for individuals’ psychotropic medication purchase in relation to gestational age and size for gestational age (All singleton spontaneous delivery births between 1996 and 2014 in Finland followed until 2018, N = 819 764)
Full size table

Sibling pair analysis

To control for confounding by unmeasured familial risk factors in the associations between birth outcomes and neuropsychiatric disorders, sibling pair analyses were performed focusing on the risk for the second sibling in the pair. Exposure to preterm and SGA births was studied for all F-diagnoses, except psychotic disorders, eating disorders, sleeping disorders, personality disorders, and ASD due to small sample sizes. LGA was not studied due to small main effect sizes. The risk of any neurodevelopmental and psychiatric disorder for the second siblings was higher when these second siblings were exposed to PTB compared to when only the corresponding first siblings were exposed (HR = 1.50 [95%CI, 1.42–1.58] vs HR = 1.07 [95%CI, 1.01–1.13]). This showed that the association between premature birth and any F-diagnosis was not confounded by familial factors. When we investigated the individual F-diagnoses, complete confounding of association was excluded for the PTB associations with intellectual disabilities, SDD, ADHD, and other emotional and developmental disorders (Fig. 3, Table S6).

Fig. 3
figure 3

Adjusted hazard ratios (HRs) for diagnosis in the second-born child after exposure to premature birth (< 37 weeks) or small for gestational age (SGA), as estimated by matched sibling pair analysis. Hazard ratios are represented on the x-axis. To better align hazard ratios related to intellectual disabilities, we have plotted them on a separate x-axis range. Abbreviations: both, both siblings in the pair were exposed to preterm birth or SGA; Second but not first sib exp, the second sibling but not the first sibling was exposed to preterm birth or SGA; First but not second sib exp, exposure to the first but not the second sibling; None, none of the siblings in the pair was exposed. The None group was used as a reference

Full size image

The association between SGA and any neurodevelopmental and psychiatric disorder was not confounded by familial factors. The risk for the second sibling to develop any F-diagnosis was higher when being exposed to SGA compared to when only first sibling was exposed to SGA (HR = 1.62 [95%CI, 1.48–1.77] vs HR = 1.07 [95%CI, 0.98–1.16]). The individual F-diagnoses associated with SGA without complete familial confounding were mood disorders, SDD, and other emotional and developmental disorders. The effect size was higher also for anxieties, intellectual disabilities and ADHD in second siblings when both siblings were exposed (Fig. 3, Table S7).

Sex, age and F98 sub-diagnoses

The effect sizes of sex-stratified associations indicated higher risk among girls than boys for intellectual disabilities, SSD, and other disorders (F98), and higher risk among boys for eating disorders and ASD, by extremely or very PTB (Table S8–S12).

The associations between birth outcomes and anxiety diagnoses remained when restricting age-at-onset to 10 years or older studying those followed at least 10 years, but most associations with mood diagnoses did not remain (Table S13).

In the F98 sub-diagnoses, only the F98.2–F98.3 subgroup showed a statistically significant association with birth outcomes (result S1 of supplementary information, Figure S2).

Discussion

Using a nationwide cohort of spontaneous births in Finland, we report that PTB across gestational ages and SGA, individually, are associated with higher risks of neurodevelopmental and psychiatric disorders commonly having an early childhood onset, except for sleeping disorders. However, the associations of PTB or SGA with ASD were not assessed for unmeasured shared familial confounding. The effect sizes were higher at lower gestational ages. The largest risks were seen for intellectual disabilities and SDD, with more than five-fold risks for those born extremely or very preterm, compared to term-born peers. Importantly, individuals born preterm being SGA had a higher risk for SDD than those exposed to only one of these birth outcomes. Such elevated combined risk effects were detected for SDD after exposure to extremely/very PTB or moderate-late PTB combined with SGA. Notably, LGA had a protective effect against the risk of extremely/very PTB on SDD and other behavioural and emotional disorders, and LGA and extremely/very PTB modified each other’s effects on SDD antagonistically. In support, PTB and SGA were associated also with a higher dispensation of psychotropic medication. We detected no clear effect on the disorders of post-term birth, or LGA alone. Effect sizes of extremely or very PTB on intellectual disabilities, SDD, and other disorders (F98) were higher in girls than boys, and higher among boys for eating disorders and ASD.

Extremely/very PTB was also associated with personality disorders, mood disorders and anxiety disorders (HRs ranging between 1.98 and 7.50), although familial confounding was not excluded due to smaller sample size. In accordance, extremely/very PTB was associated also with a higher dispensation of anxiolytics and sedatives. SDD, other behavioural and emotional disorders, anxiety, eating, sleeping and personality disorders have rarely been studied as effects of birth outcomes, but a few studies reported a higher risk for mood disorders [32,33,34]. While effect sizes of PTB combined with SGA on neurodevelopmental disorders are quite unreported, especially for spontaneous births, our findings of the association of PTB or SGA, individually, with risk for intellectual disabilities, ASD and ADHD and conduct disorders, were generally consistent with findings from other cohort-based studies, not stratifying for spontaneous delivery, from Denmark [35], Sweden [6], Norway [36], Finland [11], and USA [37].

Previous population-based studies, not stratifying for mode of delivery, have reported associations between low birth weight (< 2500 g) and ASD (HR = 2.44 [95%CI, 1.99–2.97]) and ADHD (HR = 1.65 [95%CI, 1.40–1.93]) [38]. In a meta-analysis, Gardener et al., found that SGA, but not PTB, was associated with ASD (odds ratio, OR = 1.35 [95%CI, 1.14–1.61]; OR = 1.16 [95%CI, 0.83–1.62], respectively) [39]. In another meta-analysis, Franz et al., found that both very preterm/very low birth weight (OR = 2.25 [95%CI, 1.56–3.26]) and extremely preterm/extremely low birth weight (OR = 4.05 [95%CI, 2.38–6.87]) were associated the risk of ADHD [3]. Deficits in general cognitive abilities were reported for 19-year-olds born extremely preterm [40]. Further, a meta-analysis showed lower intelligence for those born extremely or very preterm compared to full-term peers [41]. In addition, multiple studies have among male, compared to female, children shown a stronger association between individuals born extremely preterm or very preterm and ADHD and ASD diagnosis [24, 42, 43]. Interestingly, we observed stronger association between very PTB and ASD among males, but could not detect any sex-specific association between extremely or very preterm with ADHD, which is consistent with a Swedish cohort showing a similar risk of ADHD among males and females born extremely or moderately preterm [44].

There are proposed mechanisms for associations between birth outcomes and psychiatric morbidity in offspring. In PTB, the cerebral cortex is often underdeveloped, and more susceptible to acute injury and disrupted development during later ages [45, 46]. Less in-utero fetal growth in SGA, with common contributors being poor nutrition and placental insufficiency [47], has in human and animal studies been associated with an overrepresentation of altered brain structures [8]. Both SGA and LGA may be a consequence of a metabolically stressed intrauterine environment, with higher plasma levels of insulin, glucose, leptin and inflammatory markers which can influence the placenta and fetus with the potential to influence brain development [48]. However, it is conceivable that LGA may for babies born preterm reflect a more favourable state for postnatal development. Based on our results, we propose that incomplete brain maturation due to prematurity in combination with less in-utero fetal growth, contributes to higher risk of certain neurodevelopmental disorders than that of single adverse birth outcome, and that being LGA may compensate developmentally for some risk of psychopathology associated with premature birth. Our results could benefit risk prediction for neurodevelopmental disorders among those born spontaneously preterm, emphasizing that also birth size status should be considered.

Limitations

First, the follow-up time did not allow detection of all cases with late-onset psychiatric disorders (Table 1) [49]. Second, in this study, we categorized birth size status solely at the 5th percentile adjusted for gestational age and sex. Third, while this study adjusted for several potential confounders and performed sibling analysis, unknown paternal history and other unmeasured confounders like genetics, feeding patterns or maternal lifestyles, remain potential limitations. Maternal or paternal genotyping data were unavailable to compute polygenic risk scores, which could account for genetic predisposition. Fourth, while we identified all diagnoses of psychiatric disorders over time, comorbidities and changes in diagnoses were not considered in this study. Fifth, due to the smaller sample size of eating disorders, sleeping disorders, ASD and conduct disorder, sibling analysis to investigate familial confounders was not feasible. Lastly, this study is based on the Finnish population; therefore, the generalizability of our findings to non-European populations such as Asian or African populations may be limited.

Conclusions

Preterm and SGA births were associated with increased risks for childhood-onset neurodevelopmental disorders in individuals with spontaneous births. These two exposures combined implied a higher risk for SDD than one exposure alone, while, being born LGA lowered the risks for SDD and other emotionally or behavioural disorders in individuals born very preterm. Furthermore, notable sex-specific effects were observed in extremely or very PTB, on eating disorders, intellectual disabilities, SDD, ASD and other behavioural and emotional disorders.