Introduction
Experts have raised concerns about the potential neurodevelopmental risks of perinatal synthetic oxytocin for offspring. Many women receive synthetic oxytocin for labor induction and/or augmentation. Induction is involved in approximately 25% of all term births in developed countries (World Health Organization, 2011), and nearly 50% of induced births use synthetic oxytocin (Goffinet et al., Reference Goffinet, Dreyfus, Carbonne, Magnin and Cabrol2003). Synthetic oxytocin is also the first-choice agent for labor augmentation (World Health Organization, 2014). Estimates of augmentation are as high as 55 and 57% of births in Sweden and the USA (Declercq et al., Reference Declercq, Sakala, Corry and Applebaum2007; Selin, Reference Selin2009). Overmedicalization of labor is a global issue, with coverage estimates for oxytocin-augmentation as high as 79% in some countries (Miller, Reference Miller2016). Experts on autism spectrum disorder (ASD) proposed that synthetic oxytocin for induction and augmentation could enter fetal circulation and deactivate oxytocin receptors, increasing the risk of psychiatric disorders, such as ASD (Hollander et al., Reference Hollander, Cartwright, Wong, DeCaria, DelGiudice-Asch, Buchsbaum and Aronowitz1998; Wahl, Reference Wahl2004). Neuroscientists stated this hypothesis more boldly, asserting that oxytocin used for induction causes disruption in the oxytocinergic system of infants, triggering the development of ASD in children with genetic and epigenetic vulnerabilities (Gialloreti et al., Reference Gialloreti, Benvenuto, Benassi and Curatolo2014). Oxytocin experts within obstetrics and psychiatry have also suggested that synthetic oxytocin may increase the risk of other neurodevelopmental disorders, such as attention-deficit/hyperactivity disorder (ADHD; Bell et al., Reference Bell, Erickson and Carter2014) and general problems related to social behavior and the ability to cope with stress (Buckley, Reference Buckley2015; Lefevre and Sirigu, Reference Lefevre and Sirigu2016).
These hypotheses are based on the important roles endogenous oxytocin plays in the regulation of stress, anxiety, and social cognition and behaviors in humans (Landgraf and Neumann, Reference Landgraf and Neumann2004) and findings from animal studies showing that neonatal exposure to synthetic oxytocin leads to changes in social behavior and the oxytocinergic system in adulthood (Bales and Carter, Reference Bales and Carter2003; Jia et al., Reference Jia, Tai, An, Broders and Sun2008a, Reference Jia, Tai, An, Broders, Ding, Kong, Zhao and Zhang2008b; Lefevre and Sirigu, Reference Lefevre and Sirigu2016). Disruptions in the oxytocinergic system have been implicated in a range of psychiatric disorders (Cochran et al., Reference Cochran, Fallon, Hill and Frazier2013), such as ASD (Gregory et al., Reference Gregory, Connelly, Towers, Johnson, Biscocho, Markunas, Lintas, Abramson, Wright, Ellis, Langford, Worley, Delong, Murphy, Cuccaro, Persico and Pericak-Vance2009) and ADHD (Sasaki et al., Reference Sasaki, Hashimoto, Oda, Ishima, Kurata, Takahashi, Kamata, Kimura, Niitsu, Komatsu, Ishikawa, Hasegawa, Shiina, Hashimoto, Kanahara, Shiraishi and Iyo2015). Genetic and epigenetic studies consistently show associations between the oxytocin receptor gene (OXTR) and ASD (Gregory et al., Reference Gregory, Connelly, Towers, Johnson, Biscocho, Markunas, Lintas, Abramson, Wright, Ellis, Langford, Worley, Delong, Murphy, Cuccaro, Persico and Pericak-Vance2009). Genetic studies have not supported a role of the OXTR in ADHD (Park et al., Reference Park, Willmott, Vetuz, Toye, Kirley, Hawi, Brookes, Gill and Kent2010; Kalyoncu et al., Reference Kalyoncu, Özbaran, Köse and Onay2017), but youth with ADHD have lower serum levels of oxytocin than neurotypical controls (Sasaki et al., Reference Sasaki, Hashimoto, Oda, Ishima, Kurata, Takahashi, Kamata, Kimura, Niitsu, Komatsu, Ishikawa, Hasegawa, Shiina, Hashimoto, Kanahara, Shiraishi and Iyo2015).
It is unclear whether perinatal exposure to synthetic oxytocin has deleterious effects in humans. The widespread use of synthetic oxytocin and its importance to obstetric practice makes it imperative to investigate this question by reviewing and analyzing all accessible data. No systematic reviews of the evidence linking perinatal oxytocin to neurodevelopmental outcomes in offspring have been published. Thus, we aimed to provide a comprehensive overview of research on neurodevelopmental outcomes in offspring related to the obstetric use of oxytocin that systematically and critically evaluates the currently available evidence. We sought to answer the following question: Is perinatal synthetic oxytocin associated with increased neurodevelopmental problems in human offspring? If perinatal synthetic oxytocin increases the risk of neurodevelopmental problems, this knowledge may incentivize the more judicious use of oxytocin in obstetrics. Null findings, although unable to rule out an association, are just as important to disseminate as significant findings, especially to calm concerns after the publication of several papers linking, at least putatively, perinatal synthetic oxytocin use to neurodevelopmental problems in offspring (Wahl, Reference Wahl2004; Plothe, Reference Plothe2009; Gialloreti et al., Reference Gialloreti, Benvenuto, Benassi and Curatolo2014).
Methods
We conducted a systematic literature search according to MOOSE Guidelines for Meta-Analyses and Systematic Reviews of Observational Studies (Stroup et al., Reference Stroup, Berlin, Morton, Olkin, Williamson, Rennie, Moher, Becker, Sipe and Thacker2000).
Search strategy
We searched the databases Web of Science, Medline, and EMBASE up to 25 August 2017, for studies on the administration of synthetic oxytocin in the perinatal period without language or date restrictions. For EMBASE, we used the terms ‘Pitocin®’ or ‘Syntocinon®’ or ‘synthetic oxytocin’ and ‘labour/labor’ or ‘birth’ or ‘perinatal’ or ‘prenatal’ or ‘obstetric’. In Web of Science, we combined the search terms used in the EMBASE search with ‘delivery’ and ‘augment*’ or ‘induc*’. For MEDLINE, we used the EMBASE search terms in addition to ‘antepartum’. We also manually searched the reference lists of papers that matched the eligibility criteria and one review article (Bell et al., Reference Bell, Erickson and Carter2014). We contacted authors of included studies to request any unpublished studies and results. As of 9 September 2018, two new relevant studies emerged.
Inclusion and exclusion criteria
To be included in the review, publications had to meet the following criteria: (1) describe original research, (2) report results on the association between perinatal synthetic oxytocin exposure and any childhood neurodevelopmental outcome, and (3) written in English. Neurodevelopmental outcomes include all disorders within the Diagnostic and Statistical Manual of Mental Disorders (DSM-5; American Psychiatric Association, 2013) or F-diagnoses in the International Statistical Classification of Diseases and Related Health Problems (ICD-10; World Health Organization, 1992) and supplementary somatic diagnoses for delayed development (R41.8, R62, R62.0, R62.9) and dimensional measures of psychiatric symptom severity. We included studies that examined pharmacological augmentation even if studies did not explicitly report synthetic oxytocin, as it is the primary pharmacological method for augmentation. We excluded studies that described pharmacological induction without specifying the use of oxytocin, as prostaglandins are the recommended first-choice pharmacological preparation for certain indications of induction (World Health Organization, 2011). In cases of publications using the same sample and outcomes, we retained publications with the earliest publication date and excluded later publications.
To be included in a meta-analysis, two or more studies had to have the same outcome. Equivalent ICD and DSM diagnoses with different names (e.g. ADHD and disturbances of hyperactivity and attention) were regarded as the same outcome. Included studies also had to report a risk estimate or data that would enable us to calculate a risk estimate (e.g. frequencies or percentages).
Study selection procedure
First, the first author imported search results into Endnote X8 and removed duplicates. Then, she reviewed primary titles and abstracts to select articles for further scrutiny. For the retained articles, the first author obtained the full text and imported the references into Review Manager 5.3 (The Cochrane Collaboration, 2014). Next, the first and second authors independently read full texts to determine eligibility. In cases of disagreement, the first and second authors discussed the specific articles to reach consensus. If consensus could not be reached, the last author broke the tie.
Data collection
We extracted the following information from included studies: author, year of publication, study design, country in which the study was conducted, mode of oxytocin administration, level of measurement for exposure, outcome criteria, comparison group type, sample and group size, percentage of females, children's ages at follow-up, risk estimates [i.e. odds ratios, risk ratios, rate ratios, and standard errors (s.e.) or confidence interval (CI), when s.e. not available], and control variables (Table 1).
Characteristics of studies
ADHD, attention deficit/hyperactivity disorder; admin, administration; ASD, autism spectrum disorder; aug, labor augmentation; cat., category; CBCL, Child Behavior Checklist; CC, case control; Di, dichotomous; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition; DSM-IV-TR, DSM-IV, Text Revision; Exp. Meas., exposure measurement; ICD-10, International Classification of Diseases 10th revision; ind, labor induction; OT, oxytocin; NOS, Newcastle-Ottawa Assessment Scales for case-control and cohort studies. Two sample sizes reported for one study indicate the different samples sizes for labor inductions and augmentations. Age is in years at follow-up.
* = statistically significant harmful association.
Quality assessment
We assessed study quality using the Newcastle-Ottawa Quality Assessment Scales (NOS) for case-control and cohort studies (Wells et al., Reference Wells, Shea, O'Connell, Peterson, Welch, Losos and Tugwell2010). NOS ratings informed our interpretations of the sensitivity analyses. We made modifications to the NOS based in part by issues raised about the scale (Stang, Reference Stang2010). The first and second authors independently rated studies within three categories: selection, comparability, and exposure (for case-control studies) or outcome (for cohort studies). The last author decided the final score on any discrepant scores. Total scores could range from 0 to 11. We grouped the scores into intervals of low (0–3), moderate (4–7) and high (8–11) study quality. The selection section consists of four items (Table 2), for which studies could receive one star per item.
Modified Newcastle–Ottawa Quality Assessment Scale for case-control and cohort studies
Ascert., Ascertainment; Assess., Assessment; Def., definition; Exp, exposure; FU, follow-up period; Maternal psych. Hx, maternal history of psychiatric illness; Rep., representativeness; resp., response; Select., selection; SES, socioeconomic status.
Note: Adequate follow-up of cohorts ⩾80%.
* fulfilled criteria. / did not fulfill criteria.
For groups to be judged as comparable within studies, analyses needed to be adequately adjusted: either confounding variables had to be included in the statistical models or individuals had to be matched on the confounding variables. We decided on the specific criteria for comparability after identifying the studies to include in the review, as these criteria depend on the outcomes of individual studies. For comparability, studies could receive 0–4 stars. Studies could earn two stars if analyses adjusted for maternal psychiatric history. Studies could earn one more star if they controlled for socioeconomic status (i.e. income or maternal education). Studies could receive an additional star if they controlled for two or more other potential confounders: hypertensive disorders during pregnancy, preterm premature rupture of membranes, rhesus incompatibility, maternal diabetes, size for gestational age, length of labor, maternal BMI, maternal age, and maternal race, ethnicity or immigrant status.
For the final study quality criterion, studies could earn one star for each of the three items on exposure or outcome. We deemed follow-up of cohorts adequate if at least 80% of the cohort was included in the follow-up or if attrition was independent of exposure and outcome (Hartling et al., Reference Hartling, Hamm, Milne, Vandermeer, Santaguida, Ansari, Tsertsvadze, Hempel, Shekelle and Dryden2012).
Statistical analysis
We conducted meta-analyses using Review Manager 5.3 (The Cochrane Collaboration, 2014). As ASD and ADHD are relatively rare outcomes, with 2 and 5% prevalence rates, we treated odds ratios, risk ratios (RR), and rate ratios as comparable estimates of the relative risk (Symons and Moore, Reference Symons and Moore2002). We used adjusted risk estimates if studies reported these or we could obtain them from the authors. If not, we used the raw data available in the publication. Due to under-reporting and misclassification of induction and augmentation in medical records in many countries (Lain, Reference Lain2012), we combined separate estimates for induction and augmentation from mutually exclusive samples. Otherwise, the augmentation estimates were entered first and exchanged with the induction estimates in analyses.
We used random effect models to estimate pooled RRs to account for heterogeneity (Shadish and Haddock, Reference Shadish, Haddock, Cooper, Hedges and Valetine2009). I 2 estimated the homogeneity of the RRs (Higgins et al., Reference Higgins, Thompson, Deeks and Altman2003). A low amount of heterogeneity is associated with I 2 ⩽ 25%, moderate heterogeneity begins at 50%, and a high amount starts at 75% (Higgins et al., Reference Higgins, Thompson, Deeks and Altman2003). To test the robustness of our results and handle heterogeneity, we removed studies one at a time, re-ran the analysis each time, and sequentially removed the studies responsible for the greatest decrease in I 2 until I 2 was 25% or below (Patsopoulos et al., Reference Patsopoulos, Evangelou and Ioannidis2008). When studies reported separate risk estimates for female and male offspring, we tested subgroup differences. Given the small number of studies, recommended tests of publication bias were not feasible (trim and fill, funnel plots).
Results
Our search in electronic databases yielded 2581 titles, and our manual search of bibliographies yielded 18 additional titles (Fig. 1). We excluded 152 duplicates and 2423 titles based on title or abstract. We reviewed 25 full-text articles for eligibility, of which 12 studies met inclusion criteria, with four different outcomes: cognitive functioning (n = 2), problem behavior (n = 1), ADHD (n = 3), and ASD (n = 7). Seven of the 12 authors we contacted responded. Two authors are preparing new data for publication, but no author made unpublished data available for this review. Three authors provided additional information on study characteristics (Table 1). Table 2 provides an overview of the quality assessments of the included studies.
Flowchart of study selection process.
Perinatal oxytocin and cognitive impairment
The earliest published study did not support an association between use of oxytocin and developmental delay (RR 1.46; 95% CI 0.79–2.71; Gonzalez-Valenzuela et al., Reference Gonzalez-Valenzuela, García-Fortea, Delgado-Ríos, Cazorla-Granados, Blasco-Alonso and Gonzalez-Mesa2014). Risk estimates were significant for mothers under the age of 28 and over 35 years, in infants born after non-instrumental deliveries, and infants born without anesthesia. However, CIs were even wider for these estimates. A second study, which only investigated oxytocin exposure by means of induction, found that exposed children scored 0.45 (95% CI −2.12 to 1.23; p = 0.60) standard deviation units lower on a test of receptive vocabulary (IQ scores) and 0.14 (95% CI −0.26 to −0.03; p = 0.02) standard deviation units lower on a test of problem-solving ability (z-scores) than the children not exposed to induction with oxytocin (Freedman et al., Reference Freedman, Brown, Shen and Schaefer2015).
Perinatal oxytocin and problem behavior
One study examined the relationship between oxytocin-induced and/or -augmented labor and parent-rated child behavior and emotional problems at five time points during childhood and adolescence (Guastella et al., Reference Guastella, Cooper, White, White, Pennell and Whitehouse2018). At age 10, the oxytocin-augmented group had a higher percentage of children with a Child Behavior Checklist (CBCL; Achenbach, Reference Achenbach1991) T-score of 60 or above (23.3%) than the oxytocin-induced (17.8%) and no-oxytocin groups (16.4%). A Chi-squared test indicated a significant difference between groups (p = 0.04). However, oxytocin exposure was not a significant predictor of dichotomous or continuous CBCL scores at any age. Based on a subsample of 542 children, higher oxytocin dosage was associated with a slight increased risk of a CBCL score of 60 or above (OR 1.03, 95% CI 1.01–1.06).
Perinatal oxytocin and ADHD
Three studies, with a total of 5 47 278 participants, investigated the association between perinatal oxytocin and an ADHD diagnosis. The smallest study found that oxytocin-induction or augmentation posed a slight increased risk for ADHD (Kurth and Haussmann, Reference Kurth and Haussmann2011), whereas the remaining two studies showed no increased or decreased risk (Silva et al., Reference Silva, Colvin, Hagemann and Bower2014; Henriksen et al., Reference Henriksen, Wu, Secher, Obel and Juhl2015). Pooling the estimates of the three studies suggested that oxytocin-induction and/or –augmentation is not associated with ADHD in offspring (see Fig. 2). The pooled estimates remained effectively unchanged after replacing the partially adjusted estimate with the fully adjusted estimate of one study (Silva et al., Reference Silva, Colvin, Hagemann and Bower2014), and subsequently replacing the estimate for oxytocin-induction with the oxytocin-augmentation estimate (results not shown; Silva et al., Reference Silva, Colvin, Hagemann and Bower2014). Heterogeneity among the estimates was high, possibly due to methodological differences among the studies. Risk estimates from the first study were based on unadjusted analyses (Kurth and Haussmann, Reference Kurth and Haussmann2011), whereas the second and third studies reported adjusted risk estimates (Silva et al., Reference Silva, Colvin, Hagemann and Bower2014; Henriksen et al., Reference Henriksen, Wu, Secher, Obel and Juhl2015) The studies also differed in definitions of oxytocin exposure and ADHD, follow-up length, and female/male sample composition (Table 1).
Relationship between perinatal synthetic oxytocin use and risk of attention-deficit/hyperactivity disorder (ADHD) using a meta-analysis of comparative epidemiological studies. Data are presented as risk ratios (RRs). Squares indicate the RR estimates and whiskers their 95% confidence interval. The center of the diamond represents the pooled RR and the corners the 95% confidence interval. An RR greater than 1 indicates a higher risk of ADHD in those exposed to perinatal oxytocin.
Two studies (Kurth and Haussmann, Reference Kurth and Haussmann2011; Silva et al., Reference Silva, Colvin, Hagemann and Bower2014) reported data for a subgroup analysis on sex. The separate estimates for females (RR 1.22; 95% CI 0.27–5.55) and males (RR 1.28–95% CI 0.67–2.44) and the difference between them (p = 0.96) were non-significant (see online Supplementary eFig. S1). Heterogeneity was high for both group estimates (I 2 = 91%; I 2 = 89%), which is likely due to the opposite directions of the estimates from the two studies. Replacing the estimate for oxytocin-induction with the oxytocin-augmentation estimate did not change results (results not shown).
In the sensitivity analysis, removing the first study (Kurth and Haussmann, Reference Kurth and Haussmann2011), removed the smallest study, moved the estimate closer to one (RR 0.98–95% CI 0.77–1.25), and reduced heterogeneity to a moderate amount (I 2 = 40%). Due to the small number of studies, it was not possible to reduce heterogeneity below the threshold.
Perinatal oxytocin and ASD
Seven studies, with a total of 8 87 470 participants, investigated the association between perinatal oxytocin and ASD in children. One study reported three separate risk estimates for three dosage categories (Herrera et al., Reference Herrera, Bakian, Fraser, Varner, Esplin and Clark2017). Another study reported separate risk estimates for oxytocin-induction and -augmentation, but the induction and augmentation groups were not mutually exclusive (Weisman et al., Reference Weisman, Agerbo, Carter, Harris, Uldbjerg, Henriksen, Thygesen, Mortensen, Leckman and Dalsgaard2015), which precluded including both estimates in the same meta-analysis. Using the estimate for induction (Weisman et al., Reference Weisman, Agerbo, Carter, Harris, Uldbjerg, Henriksen, Thygesen, Mortensen, Leckman and Dalsgaard2015), the pooled estimate was RR 1.08; 95% CI 1.00–1.18 (see online Supplementary eFig. S2). Heterogeneity among the estimates was significant, possibly stemming from definitions of oxytocin exposure. Using the estimate for augmentation instead of the induction estimate (Weisman et al., Reference Weisman, Agerbo, Carter, Harris, Uldbjerg, Henriksen, Thygesen, Mortensen, Leckman and Dalsgaard2015), did not change the point estimate (RR 1.10; 95% CI 1.04–1.17; see Fig. 3). Heterogeneity was reduced, but still just above threshold (I 2 = 27%). Excluding one study (Smallwood et al., Reference Smallwood, Sareen, Baker, Hannusch, Kwessi and Williams2016) eliminated heterogeneity (I 2 = 0%) and provided a less biased estimate as it reduced inconsistency and removed a study with a quality score of one out of 11 (RR 1.10; 95% CI 1.05–1.16). Heterogeneity may have been reduced as the two largest studies defined oxytocin exposure similarly (Gregory et al., Reference Gregory, Anthopolos, Osgood, Grotegut and Miranda2013; Weisman et al., Reference Weisman, Agerbo, Carter, Harris, Uldbjerg, Henriksen, Thygesen, Mortensen, Leckman and Dalsgaard2015).
Relationship between perinatal synthetic oxytocin use and risk of autism spectrum disorder (ASD) using a meta-analysis of comparative epidemiological studies. This analysis uses the estimate for labor augmentation from Weisman et al., Reference Weisman, Agerbo, Carter, Harris, Uldbjerg, Henriksen, Thygesen, Mortensen, Leckman and Dalsgaard2015. Data are presented as risk ratios (RRs). Squares indicate the RR estimates and whiskers their 95% confidence interval. The center of the diamond represents the pooled RR and the corners the 95% confidence interval. An RR greater than 1 indicates a higher risk of ASD in those exposed to perinatal oxytocin.
The subgroup analysis on sex, pooling five estimates for females and six for males from three and four studies, respectively (see online Supplementary eFig. S3; Gale et al., Reference Gale, Ozonoff and Lainhart2003; Gregory et al., Reference Gregory, Anthopolos, Osgood, Grotegut and Miranda2013; Herrera et al., Reference Herrera, Bakian, Fraser, Varner, Esplin and Clark2017; Weisman et al., Reference Weisman, Agerbo, Carter, Harris, Uldbjerg, Henriksen, Thygesen, Mortensen, Leckman and Dalsgaard2015), suggested that the magnitude of risk did not significantly differ between the sexes (p = 0.17). The pooled estimate for females was RR 1.17–95% CI 1.05–1.30, with no indication of heterogeneity (I 2 = 0%). However, the wide CI for the individual female estimates may account for the homogeneity among the estimates. The pooled estimate for male offspring was RR 1.07; 95% CI 1.00–1.15, with negligible heterogeneity (I 2 = 23%).
Discussion
Several individual studies have concluded that oxytocin used to induce and augment labor increases the risk of specific psychiatric symptoms and disorders. Our systematic search of the literature revealed that few studies have examined associations between perinatal oxytocin use and neurodevelopmental outcomes in offspring. Outcomes include cognitive impairment, problem behavior, ADHD, and ASD diagnoses.
Current findings for cognitive impairment are inconclusive, as they were based on only two studies that used different scales to measure cognitive abilities. Furthermore, the quality of the two studies on cognitive impairment was low to moderate. The significant associations between perinatal oxytocin and child problem behavior (internalizing and externalizing) were small and came from a single study of moderate quality that included many statistical tests (Guastella et al., Reference Guastella, Cooper, White, White, Pennell and Whitehouse2018). For ADHD, three studies of moderate to high quality were identified, and the meta-analyses did not support an association between perinatal oxytocin and ADHD. Our finding is consistent with the results of a recent cohort study on the association between perinatal synthetic oxytocin use and dimensionally measured ADHD symptoms (Stokholm et al., Reference Stokholm, Juhl, Lønfeldt, Mikkelsen, Obel and Strandberg-Larsen2018). This cohort study was not included in our meta-analysis as it used a subsample of an included population-based study (Henriksen et al., Reference Henriksen, Wu, Secher, Obel and Juhl2015). The lack of an association between perinatal synthetic oxytocin and ADHD is not surprising given the absence of a known link between ADHD and the oxytocinergic system. Studies have failed to find an association between oxytocin and ADHD, but they have supported oxytocin's role in symptoms marked by problems in social behavior (Gregory et al., Reference Gregory, Connelly, Towers, Johnson, Biscocho, Markunas, Lintas, Abramson, Wright, Ellis, Langford, Worley, Delong, Murphy, Cuccaro, Persico and Pericak-Vance2009; Sasaki et al., Reference Sasaki, Hashimoto, Oda, Ishima, Kurata, Takahashi, Kamata, Kimura, Niitsu, Komatsu, Ishikawa, Hasegawa, Shiina, Hashimoto, Kanahara, Shiraishi and Iyo2015).
Our second set of meta-analyses found a marginal association between perinatal oxytocin and ASD in offspring. The quality of the ASD studies varied greatly. The two most influential studies were high-quality cohort studies: NOS scores of 8 and 11 (Gregory et al., Reference Gregory, Anthopolos, Osgood, Grotegut and Miranda2013; Weisman et al., Reference Weisman, Agerbo, Carter, Harris, Uldbjerg, Henriksen, Thygesen, Mortensen, Leckman and Dalsgaard2015). The studies adjusted for some risk factors for induction and augmentation, earning them quality points for comparability, but they did not adjust for all known risks for stimulated labor. Furthermore, one of the studies was unable to account for siblings within their sample (Gregory et al., Reference Gregory, Anthopolos, Osgood, Grotegut and Miranda2013). The other study calculated robust and appropriate-for-use-tests to account for siblings in the dataset (Weisman et al., Reference Weisman, Agerbo, Carter, Harris, Uldbjerg, Henriksen, Thygesen, Mortensen, Leckman and Dalsgaard2015), yet comparing discordant sibling pairs or better yet discordant cousins matched on parity, may produce more accurate results. Our marginal finding is consistent with a study that found that the association between labor induction by any means and autism disappeared when tested within siblings and in first-born children (Oberg et al., Reference Oberg, D'Onofrio, Rickert, Hernandez-Diaz, Ecker, Almqvist, Larsson, Lichtenstein and Bateman2016). Our finding also may be rendered non-significant in sibling comparison designs. Familial comparison designs control for confounding by genetic vulnerability and/or shared environmental factors, such as maternal stress during pregnancy (Weinstock, Reference Weinstock2005). Indeed, any relationship between oxytocin-stimulated labor and ASD may be confounded by the genetic or environmental vulnerability. ASD is highly heritable, and a meta-analysis has demonstrated consistent associations between ASD and single-nucleotide polymorphisms on the oxytocin receptor gene (OXTR; Loparo and Waldman, Reference Loparo and Waldman2015). Preliminary findings indicate that genetic vulnerability may also be inherited by other means. A study observed that a mother and a sibling of a child with ASD had a genomic deletion containing the OXTR; the child with ASD did not have this deletion, but did have increased methylation of OXTR (Gregory et al., Reference Gregory, Connelly, Towers, Johnson, Biscocho, Markunas, Lintas, Abramson, Wright, Ellis, Langford, Worley, Delong, Murphy, Cuccaro, Persico and Pericak-Vance2009). Hypermethylation is associated with decreased peripheral levels of oxytocin (Dadds et al., Reference Dadds, Moul, Cauchi, Dobson-Stone, Hawes, Brennan and Ebstein2014). Therefore, women with variations in the OXTR related to ASD may also produce less oxytocin, increasing their need for labor stimulation.
The current review has major strengths, such as a comprehensive search and large, pooled sample sizes in the meta-analyses, but it also has some limitations that should be considered when interpreting our findings. First, we cannot exclude the possibility of publication bias toward significant effects. Recommended methods (i.e. funnel plots, trim-and-fill) for assessing publication bias were not feasible with the small number of studies. One study had a problem with selective outcome reporting: They reported the existence of a non-significant estimate but did not report the actual risk estimate (Smallwood et al., Reference Smallwood, Sareen, Baker, Hannusch, Kwessi and Williams2016). Second, the NOS is not perfect (Stang, Reference Stang2010). We tried to improve the scale by increasing the number of points studies could receive for controlling for specific confounders and weighting maternal psychiatric history as a more important variable to adjust for than other potential confounders. Still, given the differences in labor induction and augmentation practices across countries, and even across different hospitals within the same country, different confounders may exist for different countries. Third, we excluded non-English-language studies for practical reasons. Our search only revealed one French study citation on the association between oxytocin-induced labor and developmental delay in psychomotor abilities (Sacrez et al., Reference Sacrez, Berland, Gandar, Irrmann, Schleider and Francfort1969). We also excluded the outcomes of two studies, bipolar disorder, and autistic-like behavior, as the follow-up period extended beyond childhood (Freedman et al., Reference Freedman, Brown, Shen and Schaefer2015; Guastella et al., Reference Guastella, Cooper, White, White, Pennell and Whitehouse2018). Finally, all, except one, of the existing studies were conducted in Western countries, limiting generalizability to populations in other parts of the world.
Our systematic review indicates that perinatal synthetic oxytocin poses little or no increased risk to child neurodevelopment. Thus, previous conclusions that perinatal oxytocin poses an increased risk for neurodevelopmental problems have been premature. Any future investigation of the association between oxytocin-stimulated labor and neurodevelopmental outcomes in offspring would benefit from controlling for maternal psychopathology or including a sibling or even cousin comparison. Such designs that control for confounding by genetic vulnerability and prenatal environmental factors would likely eliminate the minute association between perinatal oxytocin and ASD. Despite the lack of evidence for the increased risk of cognitive impairment, ADHD and ASD, clinicians should continue to follow clear guidelines that reduce the ubiquitous use of synthetic oxytocin due to the known physiological risks associated with the medication (World Health Organization, 2011, 2014).
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291718003021.
Acknowledgements
We would like to thank the authors who took time out of their busy schedules to respond to our request for additional information for this review. A grant awarded by the Department of Psychiatry of the Capital Region of Denmark to NNL funded this review. Role of the Funder/Sponsor: The funder had no role in the study design, collection, analysis, or interpretation of the data, writing of the manuscript, or the decision to submit the paper for publication.
Author contributions
NNL, ERL, KJP, and KSL were involved in the design of this study. NNL conducted the literature search, screened articles for inclusion, performed the meta-analyses, and wrote the first draft of the article. NNL and FCV assessed articles for eligibility and study quality. ERL settled disagreements on study quality scores. All authors were involved in determining the criteria for study quality and read and approved the final version of the article.
Conflict of interest
Although not relevant for this review, F.C.V. is the distributor of Dutch translations of ASEBA from which he receives remuneration. The remaining authors have no conflicts of interest to report.
