Association of maternal endometriosis with autism spectrum disorder in offspring: a retrospective cohort study

Abstract

Background

With the increasing incidence of autism spectrum disorder (ASD), various risk factors for ASD have been identified. Among them, maternal immune activation (MIA), which can be triggered by inflammation via elevated proinflammatory cytokines, has recently been reported as a risk factor for ASD in offspring.

Objectives

We assessed the association of maternal endometriosis, a chronic gynecological inflammation that may induce MIA, with ASD in offspring.

Study design

In this retrospective cohort study, we used the data from a large-scale public health insurance database across Japan to identify dyads of mothers and their firstborn children, delivered between January 2005 and May 2022. Mothers with multiple pregnancies were excluded. The proportions of children diagnosed with ASD at 3 years old were compared between eligible mothers with endometriosis and those without a diagnosis of endometriosis prior to delivery. Multivariable logistic regression analysis was performed to evaluate the association of maternal endometriosis with ASD in children, after adjusting for maternal and peripartum characteristics.

Results

We identified 30,020 eligible dyads of mothers and their firstborn children. Among them, 3,496 mothers had endometriosis. The proportions of ASD in the offspring of the group of mothers without and with endometriosis were 2.7% (720/26,524) and 2.9% (102/3,496), respectively. Logistic regression analysis revealed no significant association of maternal endometriosis with ASD in their children (adjusted odds ratio, 1.04; 95% confidence interval, 0.84–1.29).

Conclusion

No significant association of maternal endometriosis with ASD in offspring was detected in this study. These findings suggest children whose mothers have an antepartum history of endometriosis alone may not need to be closely monitored for ASD, and health care providers or people involved with children may need to consider other reported backgrounds related to ASD.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12884-025-08454-7.

Background

Autism spectrum disorder (ASD) is a neurodevelopmental disorder characterized by two key behavioral impairments: (i) deficits in communication and social interactions, and (ii) restricted, repetitive behaviors, interests, activities, and sensory difficulties [1, 2]. Additionally, ASD can affect a person throughout their life [1]. Owing to changes and improvements in the diagnosis of ASD as well as increased public awareness, the prevalence of ASD has increased in recent years [1, 3]. Thus, the actual number of individuals with ASD may be higher than previously inferred. Early diagnosis and optimal intervention improve outcomes of cognitive ability, language, and adaptive behaviors in individuals with ASD [4, 5, 6].

Hereditary factors may account for 74–93% of ASD cases [3, 7]. Furthermore, the non-genetic factors that are reportedly associated with ASD include advanced parental age, hypertension or abnormal glucose metabolism during pregnancy, preterm birth, low birth weight, comorbidities such as polycystic ovary syndrome (PCOS), and the use of certain medications during pregnancy [3, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17]. Recently, maternal immune activation (MIA), triggered by maternal acute and chronic inflammation, has been highlighted as a potential mechanism in the development of neurodevelopmental disorders, such as ASD [18, 19, 20, 21, 22]. MIA possibly affects the fetal neurodevelopmental aspects associated with the ASD phenotype through the direct action of enhanced maternal proinflammatory cytokines by MIA, such as interleukin-6 (IL-6), or the effect of inflammatory cytokines that are released in the placenta and fetus secondary to the enhanced levels of maternal inflammatory cytokines [20, 23, 24].

Endometriosis is a gynecological, chronic inflammatory pelvic disease that affects 5%–10% of women of reproductive age and may cause pain or infertility [25, 26]. Endometriosis-induced inflammation may cause obstetric and neonatal complications, such as preeclampsia and premature birth [27, 28, 29]. Supporting evidence of this association is the finding that high IL-6 levels in the peritoneal fluid or serum of patients with endometriosis have been reported [30, 31, 32]. Given this background, maternal endometriosis may be associated with neurodevelopmental disorders, such as ASD, via MIA. However, studies of the association of maternal endometriosis with ASD in offspring have shown inconsistent results. One study found a positive association of maternal endometriosis with ASD [33], whereas another did not find a significant association [34]. These studies primarily focused on infertility and its treatment and treated endometriosis as a contributing factor to infertility. Furthermore, both were case–control studies, and the number of participants with endometriosis was small.

Therefore, in this study, we aimed to investigate the association of maternal endometriosis with ASD in offspring using data from a large-scale health insurance claims database in Japan.

Materials and Methods

Data Source

This retrospective cohort study used the JMDC Health Insurance Claims Database (JMDC Inc., Tokyo, Japan), which comprises public health insurance claims from multiple health insurers across Japan since 2005 and included data on approximately 17 million individuals as of February 2024 [35]. The insured are the employees of Japanese companies and their family members. Furthermore, it is possible to track individuals even if they are transferred between hospitals or are visiting multiple clinics. The database includes anonymized inpatient and outpatient data on diagnoses, medications, procedures, and findings of medical examinations. The International Classification of Diseases, 10th Revision (ICD-10) codes and original Japanese disease codes were used to record diagnoses in the database, which also includes an anonymous family identification number that enables the linkage of the information of each family member.

Participants

We included firstborns linked to their mothers using family identification numbers from January 2005 to May 2022. We speculated that birth order could be a factor in the timing of ASD diagnosis for the following reasons. First, ongoing clinical management for ASD in a firstborn may be associated with increased medical consultation for subsequent siblings. Second, if a firstborn child shows typical development, any clearly different developmental trajectory in a second or later-born child could lead to earlier medical consultations. Therefore, this study was limited to firstborn children. We excluded children whose (i) data did not start from the year and month of birth, (ii) mothers whose data did not cover their child’s year and month of birth, (iii) mothers with multiple pregnancies (ICD-10 code: O30), and (iv) mothers with data for < 2 years before childbirth. Eligible children were divided into two groups: those whose mothers had been diagnosed with endometriosis (ICD-10 code: N80) within 2 years prior to childbirth (endometriosis group) and those whose mothers were not diagnosed with endometriosis (non-endometriosis group).

Outcomes

We defined the primary outcome as a diagnosis of ASD by the age of 3 years. We used the ICD-10 code for pervasive developmental disorders (F84) to diagnose ASD, except for Rett syndrome (F84.2) [2, 36, 37].

We defined the endpoint as an ASD diagnosis at 3 years of age for the following reasons. In Japan, infants undergo medical checkups at 18 months and at 3 years [38]. A previous study reported that ASD traits typically appear by 24 months [39]. Standardized screening for ASD at 18 and 24 months of age is widely recommended, given that an accurate diagnosis can be reliably made by skilled professionals as early as 18 months of age [40, 41]. Furthermore, high diagnostic stability has been consistently reported for diagnoses rendered between 18 and 36 months of age [40, 42]. Therefore, detection at the 3-year-old medical check-up would facilitate timely care [4, 5, 6].

The secondary outcome was a diagnosis of attention-deficit/hyperactivity disorder (ADHD; F90.0) at 3 years of age. Similar to ASD, ADHD is a neurodevelopmental disorder that has a possible association with MIA during pregnancy [19, 22, 43, 44].

Variables

In this study, we selected the following variables for adjusting: maternal and infant demographic factors (maternal age at the year and month of birth, child’s sex, and birth year), perinatal complications, delivery methods, low birth weight, and PCOS [14, 15, 16]. Perinatal complications included (i) hypertensive disorders of pregnancy (preeclampsia, gestational hypertension, superimposed preeclampsia, and chronic hypertension), (ii) eclampsia before delivery (antepartum eclampsia and intrapartum eclampsia), (iii) preterm premature rupture of the membranes (preterm PROM), (iv) placenta previa, (v) placental abruption, (vi) threatened preterm delivery, (vii) intrauterine infection, (viii) hyperglycemic disorders of pregnancy (pregestational diabetes mellitus and overt diabetes in pregnancy, and gestational diabetes mellitus), and (ix) preterm delivery (spontaneous preterm delivery, artificial preterm delivery, preterm birth before 28 weeks of gestation, and preterm birth after 28 weeks of gestation). The delivery methods included forced delivery using forceps or vacuum, elective cesarean section, and emergency cesarean section. Low birth weight was categorized as extremely low (< 1000 g), very low (1000–1500 g), or low (1500–2500 g). Some of these variables could be considered intermediate factors between endometriosis and ASD. However, the development of ASD from these factors, such as preeclampsia or preterm birth, may be influenced by a combination of hypoxia, oxidative stress and prematurity as well as inflammation of endometriosis [45, 46]. To focus on the potential association of chronic endometriosis-induced maternal inflammation with an ASD diagnosis in offspring, we adjusted those factors in the present study. Details of the specific conditions and corresponding ICD-10 codes or original Japanese disease codes are presented in Supplemental Table 1.

The diagnosis of ASD might depend on the birth year because of the changes in the diagnostic criteria for ASD. Preterm delivery was defined when either the mother or her child had any diagnosis of preterm delivery or birth, respectively. Prematurity due to preterm delivery is associated with neurodevelopmental disorders [9, 13]; however, few cases were reported for preterm delivery before 28 weeks of gestation. Therefore, we did not adjust for gestational age in the present study. Preterm PROM was defined using the ICD-10 codes for both PROM and preterm delivery.

Statistical analysis

Continuous variables are presented as means with standard deviations. Categorical variables are presented as frequencies and proportions. Absolute standardized differences were used to assess the intergroup differences. If the absolute standardized mean difference is less than 0.1, it means well balanced between the groups [47]. We performed multivariable logistic regression analyses to evaluate the association of maternal endometriosis with the offspring’s outcomes. We excluded intrapartum eclampsia as an independent variable for ASD because of their low prevalence. Moreover, we did not include superimposed preeclampsia, chronic hypertension, antepartum eclampsia, intrapartum eclampsia, and extremely low birth weight as independent variables from the analysis for ADHD for the same reason. All statistical analyses were conducted using STATA/SE version 17 (STATA, College Station, TX, USA).

Sensitivity analysis

We performed two sensitivity analyses. First, survival analysis was performed. Children were followed up until a diagnosis of ASD or ADHD at 96 months or until they were lost to follow-up, whichever occurred first. If the observation period was 0 months (owing to the year and month of birth and the timing of change in insurance in the same year and month), we excluded these cases. We compared the time of ASD and ADHD diagnosis between the two groups using Kaplan–Meier curves and log-rank tests. We performed Cox regression analyses, adjusting for the same variables as in the multivariable logistic regression, after excluding several variables as follows because the model did not converge: intrapartum eclampsia for ASD, and chronic hypertension, antepartum eclampsia, intrapartum eclampsia and extremely low birth weight for ADHD.

Second, we stratified the children whose mothers had ≧5 years of data prior to childbirth to confirm the validity of the results and thus divided the eligible children into two groups: those whose mothers had been diagnosed with endometriosis within 5 years prior to the birth month and mothers who were not diagnosed with endometriosis by that timepoint. As in the main analysis, multivariable logistic regression analyses were performed with a diagnosis of ASD and ADHD up to 3 years of age as the outcome, and survival analyses were performed with an observation period of up to 72 months.

Ethics

This study was approved by the Research Ethics Committee of the Faculty of Medicine of the University of Tokyo (approval number: 10862-(1), approved on 13/06/2018). Informed consent of the participants was deemed unnecessary because data were de-identified. This study was performed in accordance with the Declaration of Helsinki.

Results

Main analysis

We identified 1,572,036 mother–firstborn dyads in the JMDC claims database from January 2005 to May 2022. After applying the inclusion and exclusion criteria, 30,020 dyads were included in the main analysis. Of these, 3,496 mothers were diagnosed with endometriosis within 2 years prior to childbirth (Fig. 1).

Fig. 1.

Flow diagram of the mother–child dyads included in the study. This diagram illustrates the selection process for the mother–child dyads included in the main analysis (n = 30,020). The exclusions due to insufficient data availability of the mothers and children, or multiple pregnancies are shown. The final dyads were divided into two groups based on whether the mothers were diagnosed with endometriosis within 24 months prior to delivery

Table 1 summarizes the baseline characteristics of the patients. The mean age of the mothers at birth in the endometriosis group was higher than that in the non-endometriosis group. No intergroup difference was observed in the distribution of the child’s sex and birth year. Diagnoses of placenta previa, threatened preterm delivery, cesarean section (a combination of elective and emergency), and PCOS were more prevalent in the endometriosis group than in the non-endometriosis group.

Table 1.

Baseline maternal and child characteristics

	No endometriosis		Endometriosis		Absolute SMD
	(n = 26,524)		(n = 3,496)
Maternal age at birth, years, mean (SD)	32.7	4.7	33.9	4.5	0.25
Hypertensive disorders of pregnancy, n (%)	2687	10	398	11	0.04
Preeclampsia	2337	8.8	351	10	0.04
Gestational hypertension	521	2.0	62	1.8	0.01
Superimposed preeclampsia	33	0.1	4	0.1	0.00
Chronic hypertension	27	0.1	4	0.1	0.00
Eclampsia before delivery, n (%)	18	0.1	1	0.0	0.02
Antepartum eclampsia	6	0.0	0	0.0	0.02
Intrapartum eclampsia	12	0.0	1	0.0	0.01
PROM, n (%)	6044	23	824	24	0.02
Preterm PROM	366	1.4	58	1.7	0.02
Placenta previa, n (%)	1100	4.1	220	6.3	0.10
Placental abruption, n (%)	532	2.0	92	2.6	0.04
Threatened preterm delivery, n (%)	10469	40	1604	46	0.13
Intrauterine infection, n (%)	4752	18	757	22	0.09
PCOS, n (%)	2210	8.3	582	17	0.25
Hyperglycemic disorders in pregnancy, n (%)	5106	19	771	22	0.07
Pregestational diabetes mellitus and overt diabetes in pregnancy	144	0.5	27	0.8	0.03
Gestational diabetes mellitus	4886	18	738	21	0.07
Preterm delivery, n (%)	884	3.3	153	4.4	0.05
Spontaneous preterm delivery	153	0.6	17	0.5	0.01
Artificial preterm delivery	18	0.1	7	0.2	0.04
Preterm birth before 28 weeks of gestation	7	0.0	3	0.1	0.03
Preterm birth after 28 weeks of gestation	831	3.1	144	4.1	0.05
Forced delivery (forceps and vacuum), n (%)	989	3.7	148	4.2	0.03
Cesarean section, n (%)	1849	7.0	360	10	0.12
Elective cesarean section	451	1.7	96	2.7	0.07
Emergency cesarean section	774	2.9	146	4.2	0.07
Birth weight, n (%)
Extremely low birth weight (< 1000 g)	85	0.3	14	0.4	0.01
Very low birth weight (1000–1499 g)	104	0.4	20	0.6	0.03
Low birth weight (1500–2499 g)	1539	5.8	230	6.6	0.03
Child sex (female), n (%)	12791	48	1734	50	0.03
Birth year of the child, n (%)
2007	148	0.6	20	0.6	0.00
2008	147	0.6	21	0.6	0.01
2009	149	0.6	17	0.5	0.01
2010	199	0.8	20	0.6	0.02
2011	370	1.4	35	1.0	0.04
2012	692	2.6	74	2.1	0.03
2013	941	3.5	113	3.2	0.02
2014	1258	4.7	163	4.7	0.00
2015	1827	6.9	241	6.9	0.00
2016	2032	7.7	272	7.8	0.00
2017	2553	9.6	329	9.4	0.01
2018	3061	12	413	12	0.01
2019	3646	14	467	13	0.01
2020	3910	15	557	16	0.03
2021	4017	15	523	15	0.01
2022	1574	5.9	231	6.6	0.03

PROM premature rupture of membranes, PCOS polycystic ovary syndrome, SD standard deviation, SMD standardized mean difference

Table 2 shows the results of the logistic regression analysis. After adjustment for baseline characteristics, no significant association of maternal endometriosis with ASD (odds ratio [OR], 1.04; 95% confidence interval [CI], 0.84–1.29) or ADHD (OR, 0.83; 95% CI, 0.43–1.61) in offspring was observed.

Table 2.

Results of logistic regression analysis

Outcome	No endometriosis (n = 26,524)	Endometriosis (n = 3,496)	Crude		Adjusted
			OR (95% CI)	p	OR (95% CI)	p
ASD	720 (2.7%)	102 (2.9%)	1.08 (0.87–1.33)	0.49	1.04 (0.84–1.29)	0.74
ADHD	86 (0.3%)	10 (0.3%)	0.88 (0.46–1.70)	0.71	0.83 (0.43–1.61)	0.58

We did not include intrapartum eclampsia as independent variables for ASD because the number of variables was small and the model did not converge. Moreover, we excluded superimposed preeclampsia, chronic hypertension, antepartum eclampsia, intrapartum eclampsia, and extremely low birth weight as independent variables from the analysis for ADHD

ASD Autism spectrum disorder, ADHD Attention-deficit/hyperactivity disorder, OR Odds ratio, CI Confidence interval

Sensitivity analyses

During the observation period of 96 months, 1,032 (4.0%) children in the non-endometriosis group and 147 (4.3%) children in the endometriosis group were diagnosed with ASD. The crude incidence rate per 10,000 person-years of ASD was 118 in the non-endometriosis group and 127 in the endometriosis group (Supplemental Table 2). Supplemental Fig.1 shows the Kaplan–Meier curves for ASD. After adjustment for baseline characteristics, maternal endometriosis was not found to be associated with ASD in offspring (hazard ratio [HR], 1.03; 95% CI, 0.87–1.23; Supplemental Table 2).

Regarding ADHD, 242 (1.0%) children in the non-endometriosis group and 27 (0.8%) children in the endometriosis group were diagnosed with ADHD. The crude incidence rates per 10,000 person-years for ADHD were 27 and 23 in the non-endometriosis and endometriosis groups, respectively. Supplemental Fig. 2 shows the Kaplan–Meier curve for ADHD. After adjustment for baseline characteristics, maternal endometriosis was not associated with ADHD in offspring (HR 0.82; 95% CI, 0.55–1.23; Supplemental Table2 ).

We stratified the inclusion of children to those whose mothers had ≧ 5 years of data before childbirth. Thus, a total of 4,402 dyads were included, with 3,546 and 856 dyads in the non-endometriosis and endometriosis groups, respectively (Supplemental Fig. 3). These results were consistent with those of the main analyses (Supplemental Tables 3, 4, and 5 and Supplemental Figs. 4 and 5).

Discussion

Principal findings

In this retrospective cohort study, we used a large-scale health insurance claims database to investigate the association of maternal endometriosis with ASD and ADHD in children. Maternal endometriosis was not associated with ASD or ADHD in offspring.

Results in the Context of What is Known

A few previous studies reported the association of maternal endometriosis with ASD in offspring [33, 34], and these analyses were adjusted for confounders such as maternal race, ethnicity, and education level, which could not be reconciled in the database we used. However, both were case–control studies. One study included 629 ASD cases and 909 controls [33], and the other included 537 ASD cases and 381 controls [34]. The number of participants with endometriosis was small, with 27 ASD cases and 19 controls in one study [33], and 3 ASD cases and 1 control in the other [34]. Also, recall bias may have influenced the findings [33, 34]. In the present study, we primarily focused on the association of maternal endometriosis with ASD in the offspring by using a large-scale database. A report describing an association between maternal endometriosis and ASD in children found that the association disappeared when limited to the first child [33], which is consistent with the results of the present study that included mother–firstborn dyads.

Clinical and research implications

It has been hypothesized that the pathology of ASD is initiated via MIA by enhanced circulating levels of proinflammatory cytokines, such as IL-6, in the mother, placenta, and fetus. These cytokines may cause fetal neuroinflammation through oxidative stress or mitochondrial dysfunction, which may subsequently result in ASD–related behaviors [20, 24]. Elevated serum levels of IL-6 were observed in women with endometriosis [30, 32, 48, 49]. However, several studies have challenged this hypothesis [31, 50]. Regarding the direct effect of maternal IL-6 on the fetus, our study suggests that elevated IL-6 levels in pregnant women with only endometriosis may not be sufficient to induce significant neuroinflammation in the offspring. Furthermore, the levels of IL-6 in pregnant women were higher than those observed in non-pregnant women [51]. Despite the absence of data regarding IL-6 levels in pregnant women with endometriosis, pregnancy itself may have a greater contribution to elevating IL-6 levels than endometriosis alone. On the other hand, indirect effects of uteroplacental inflammation caused by MIA have been suggested [23, 24]. An endometriosis–associated inflammatory response in the placenta or fetus has been suggested to contribute to perinatal complications, such as preeclampsia and preterm birth [29, 52–55]. To isolate the effect of endometriosis, we adjusted for baseline characteristics, including these perinatal complications, which might have underestimated the indirect effects in the present study.

Concerning ADHD, the reasons for the lack of association between endometriosis and MIA could be the same as those for ASD.

Strengths and limitations

The strengths of the present study are the use of a large-scale database and adjustment for many baseline characteristics. To the best of our knowledge, this is the first study to focus on the association of maternal endometriosis with ASD and ADHD in children. Nonetheless, this study has several limitations that must be considered. First, no validation studies have been conducted on the diagnosis of endometriosis, ASD, or ADHD. In the main analysis, the prevalence was 13.3% for endometriosis, 2.9% for ASD, and 0.3% for ADHD. In reproductive age women, the prevalence of endometriosis is 5%–10% [25]. The prevalence of ASD and ADHD is 1.0%–1.5% [56–58] and approximately 5% [59], respectively. In our study, the prevalence may not have been consistent with that reported in previous studies because we used the outcome as having a diagnosis at least once, which may cause overestimation. Diagnosing ADHD prior to age 3 may be challenging; a previous study reported that the median age of ADHD diagnosis was 7 years [60]. Therefore, ADHD diagnosis at age 3 is likely underestimated. Second, despite using a large-scale database, our study may have had limited statistical power to detect an extremely small effect. However, the minimal difference in the prevalence of ASD between the groups (2.7% vs. 2.9%) and the wide 95% CI (0.84 to 1.29) suggest that even if a true association exists, it is unlikely to be clinically meaningful. Third, the details of endometriosis were not analyzed in this study. Further, the severity of endometriosis was not described in the database. The serum and peritoneal fluid levels of IL-6 in women with endometriosis correlate with the stage of endometriosis [61]. In the present study, there may have been many cases of low-stage endometriosis and, thus, the IL-6 levels were too low to influence neural changes in children. There are different types of endometriosis, including endometrioma and adenomyosis. However, because of missing data, we could not obtain data on the types of endometriosis in all patients. Fourth, we did not include paternal age, for which an association with the diagnosis of ASD in offspring was previously reported [8]. Fifth, data on the method of conception were not included in the database. The association between infertility treatment and assisted reproductive technology with neurodevelopmental disorders remains controversial [34, 62–64]. Sixth, we only included the firstborn children. A difference in the incidence of ASD diagnosis between the first and subsequent births has been reported [33]. Lastly, prematurity due to preterm birth is associated with ASD [13]. However, we could not accurately adjust for birth weight and gestational age because of the lack of information in the database and few cases of preterm birth, especially extremely low birth weight.

Conclusions

This retrospective cohort study using a large-scale health insurance database found no significant associations of maternal endometriosis with ASD or ADHD. Thus, these findings suggest that intensive monitoring for ASD or ADHD may not be warranted based solely on the presence of maternal endometriosis, and health care providers or people involved with children may need to consider other reported backgrounds related to ASD or ADHD. However, considering the limitations of this study, further research is necessary to account for birth order and other backgrounds that could not be included.

Abbreviations

ASD

Autism spectrum disorder

MIA

Maternal immune activation

PCOS

Polycystic ovary syndrome

IL-6

Interleukin-6

ADHD

Attention-deficit/hyperactivity disorder

PROM

Premature rupture of the membranes

Authors’ contributions

MF: Conceptualization, Data curation, Formal analysis, Methodology, Visualization, Writing – original draft, Writing – review and editing. OWH: Conceptualization, Data curation, Supervision, Writing – review and editing. YE and ST: Conceptualization, Writing – review and editing. TI, YH, and YO: Conceptualization, Supervision, Writing – review and editing. YS: Conceptualization, Data curation, Formal analysis, Methodology, Supervision, Writing – review and editing. HY: Conceptualization, Data curation, Methodology, Supervision, Writing – review and editing.

Funding

This research was funded by the Ministry of Health, Labour and Welfare, Japan (grant number 24FB1001: O.WH., 23FB0101: Y. O., and grant number 23AA2003: H. Y.).

Data availability

The data supporting the findings of this study are anonymized public health insurance data from across Japan provided by a private database company (JMDC Inc.: https://www.jmdc.co.jp/en/inquiry/). Restrictions apply to the availability of these data, which were used under license for this study and are not publicly available. Researchers interested in using these data may contact the corresponding author and JMDC Inc. to inquire about data access.

Declarations

Ethics approval and consent to participate

This study was approved by the Research Ethics Committee of the Faculty of Medicine of the University of Tokyo (approval number: 10862-(1), approved on 13/06/2018). Informed consent was not required because data were de-identified. Our study adhered to the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.