Increasing prevalence, time trend and seasonality of gastroschisis in São Paulo state, Brazil, 2005–2016

To estimate the gastroschisis seasonality and trend of prevalence in recent years, stratified by maternal age and geographical clusters of São Paulo state, a population–based study was designed. We used data from the Live Births Information System (SINASC) in São Paulo state, Brazil, from 2005 to 2016. Trends of prevalence were evaluated for the specific subgroups using the Prais–Winsten regression model, and the Durbin–Watson test was used, to estimate the regression coefficient, the annual percent change (APC), and 95% confidence interval (CI). We observed 1576 cases of gastroschisis among 7,317,657 live births (LB), a prevalence of 2.154 (95% CI: 2.047–2.260) per 10,000 LB which included, 50.6% males, 67.4% Caucasians, 53.4% preterm births, and 80.9% caesarean births. The prevalence of gastroschisis significantly increased by 2.6% (95% CI: 0.0–5.2) per year, and this trend was higher in mothers aged 30–34 years (APC: 10.2, 95% CI: 1.4–19.4) than in mothers of other age groups. Between 2011 and 2016, we identified the existence of seasonality based on the date of conception in the middle months of the year (p = 0.002). This is the first and largest population–based study summarizing current epidemiology and identifying trend of prevalence of gastroschisis in São Paulo state.


Discussion
To the best of our knowledge, this is the first population-based study that used the Live Births Information System in São Paulo state to describe the recent seasonality, and prevalence trend of gastroschisis in Brazil. Although an increasing trend in gastroschisis prevalence was observed in São Paulo from 2005-2016 (APC = 2.6%; 95% CI: 0.0-5.2), significant increasing trend results were observed only in São Paulo city (APC = 5.0%; 95% CI: 1.6-8.4) and São Paulo Metropolitan Region (APC = 4.7%; 95% CI: 0.2-9.1), and this is because all gastroschisis cases in São Paulo city were counted in the São Paulo Metropolitan region as well. When we excluded these cases from the São Paulo Metropolitan Region (SPMR without SPC), the result is a stationary trend (APC = 4,5%; 95% CI: -0.1-10.2) similar to other presented clusters.
In the first period (2005-2010), São Paulo state presented a significant increasing trend in gastroschisis prevalence (APC = 6.7%; 95% CI: 4.0-9.6) and again, only two clusters presented a significant increasing trend, the Central North cluster (APC = 36.1%; 95% CI: 8.4-71.4) and the Northwest cluster (APC = 4.7%; 95% CI: 0.9-8.6), and all other clusters presented a stationary trend. Considering that the Central North cluster and Northwest cluster are predominantly rural areas, it can be suggested that, as in other studies [6][7][8] , that agricultural exposures, such as the use of pesticides, may have influenced the increasing number of gastroschisis cases in these clusters. At the second period (2011-2016), São Paulo state and all clusters presented a stationary trend, and this is mainly explained by the decrease in gastroschisis prevalence observed since 2015.
The association between young maternal age and gastroschisis is a well-documented risk factor [14][15][16]19,20 . However, the increase in gastroschisis prevalence is not because of an increase in teen births, since birth rates have decreased among mothers <25 years, and the highest increase in prevalence trend occurred among mothers aged 30-34 years. Our results confirm that young maternal age is a significant risk factor for gastroschisis in São Paulo state. The highest gastroschisis prevalence was in the maternal age group <14 years (9.6 cases/10,000 LB; 95% CI: 6.59-12.62). The lowest gastroschisis prevalence was in the maternal age group >35 years (0.39 cases/10,000 LB; 95% CI: 0.26-0.51). Moreover, the number of LB decreased by 19,76% in the young maternal age groups such as, Our findings are consistent with those of other studies that showed that neonates born with gastroschisis are often Caucasian, born preterm, and with low birthweight 8,14,15,21,22 . Most mothers usually have low educational status, and they are more likely to give birth by cesarean delivery, as in USA 1,10,23 and Australia 18 www.nature.com/scientificreports www.nature.com/scientificreports/ et al. 25 meta-analysis report, no quantifiable advantage of caesarean section over a vaginal delivery was observed for infants with gastroschisis. In the Salihu et al. 26 study, the mode of delivery was not found to be associated with neonatal survival of infants with gastroschisis. Skarsgard 27 suggest that as caesarean delivery has not demonstrated any benefits, vaginal delivery should be preferred, unless obstetric factors dictate otherwise.
We observed a significant increase in gastroschisis prevalence when conceptions occurred during the winter months (June, July and August) from May 2011 to April 2016 (95% CI 0.01-0.05, p = 0.002). Waller et al. 8 found that from 1987 to 2006 in the Washington state, 805 cases of gastroschisis were detected in conceptions during the spring period. However, from 1995 to 2012 in California state, Anderson et al. 10 reported that the risk of gastroschisis did not vary by the season of conception. This is the first study to demonstrate a significant seasonal variation in the most important state of a large middle-income country such as Brazil, and this may raise some questions: Can the influence of climate change, or ambient air pollution, the use of vaccines or drugs before and/or during pregnancy influence the occurrence of gastroschisis?
There are some limitations to this study. There may have been mistakes during the registration of designations for congenital anomalies in the live birth declaration at the public database. Moreover, we did do not have data regarding stillbirths with gastroschisis.
The strength of this study are as follows. It is a population-based study describing the time trend and seasonality of gastroschisis prevalence in the most populous state in Brazil 28 with the highest birth rate 28 and with reliable public database 29,30 . We studied a relatively long time period of data up to 2016 and used, the most recent reports from the Brazilian Unified Health System (DATASUS), which allowed us to analyse both in the total period and in the 2 subperiods of 6 years each. www.nature.com/scientificreports www.nature.com/scientificreports/ This is the first and largest population-based study summarizing the current epidemiology and identifying trend of prevalence and seasonality of gastroschisis in the São Paulo state. Our findings demonstrate an increasing trend of prevalence in São Paulo state in recent years, being higher in older mothers and in São Paulo city. The highest overall prevalence was in the Central-South cluster, and the lowest was in Taubaté administrative region. Significant seasonal variation of gastroschisis prevalence was found, being higher, when conceptions occurred during the winter months of the year during 2011-2016.

Study design & settings.
This is a population-based study with time trend 31  The unit of analysis selected for this study was São Paulo state, which is the most populous state in Brazil, with a population (41,262,199 in 2010) 28 and birth rate (610,000/year) 32 comparable to many countries in Europe [34][35][36] and Latin America 37,38 , and where the completeness of public data is more reliable 29,30 than other states. Birth data correspond to the period between January 1, 2005, and December 31, 2016.  www.nature.com/scientificreports www.nature.com/scientificreports/ Participants. We included all LB of mothers residing in São Paulo state, Brazil, whose information in the field of congenital malformations of the Live Birth Certificate was completed with Gastroschisis, identified by the following International Classification of Diseases, 10 th Revision, ICD-10: Q79.3. We excluded patients who had a main or secondary diagnosis of Gastroschisis and who had Omphalocele/Exonphalia (ICD-10: Q79.2), hypoplasia/malformation of the abdominal muscles (ICD-10: Q79.5), or umbilical hernia (ICD-10: K42).

Variables.
To describe the outcome of this study, sociodemographic and clinical explanatory variables were selected: gender, race/color, gestational age, birth weight, maternal age, maternal schooling, maternal occupation, type of gestation, type of delivery and the number of prenatal consultations.  Cases  3  1  2  5  5  3  3  6  2  2  7  -39   LB  3,163  3,458  3,536  3,652  3,636  3,347  3,315  3,547  3,472  3,604  3,187  2,   Data sources. The microdata was extracted from the file transfer service provided by DATASUS. The TABNET and TABWIN programs were used to consult the data. Those tabs were developed to perform fast tabulations on.DBF files, then the files were exported to.XLS version; further, the variables selected for this study were classified in EXCEL ® . In order to minimize possible discrepancies, the data were collected by two different researchers independently. The Live Birth Information System was developed by DATASUS to gather epidemiological information on births reported throughout the country to subsidize interventions related to women's and children's health for all levels of the Unified Health System (SUS -Sistema Único de Saúde), such as actions of attention to the pregnant woman and to the newborn, as well as the monitoring of the evolution of the SINASC's historical series, that allows the identification of intervention priorities and contribution to the effective improvement of this information system. Through the Internet, DATASUS and the Health Surveillance Service (SVS -Sistema de Vigilância em Saúde) provide the main information for tabulation on the SINASC's Databases.   www.nature.com/scientificreports www.nature.com/scientificreports/ Bias. Only the population of LB was used to obtain the prevalence rates and the proportion of maternal age range did not differ in the linear regression model by clusters in São Paulo state (p > 0.05). With this, the result obtained by the analysis can be used for comparison to other studies.
Some limitations have been identified, inherent in the recording of public data, in the data collection stages and, in the quality of the information where there were fields with missing data, until the transmission of the data to the information systems.
Statistical methods. Gastroschisis prevalence rates were calculated for 10,000 LB by maternal age group and territorial clusters. For trends analysis, the Prais-Winsten regression model, following Antunes and Cardoso 31 methodological indications, were used. The dependent variable was the logarithm of the rates, and the independent variable, and the years of the historical series. The Annual Percent Change (APC) of the rates was also calculated, as suggested by Antunes and Waldman (2002) 40 .
The data modelling process includes, transforming the standardized rates into a base 10 logarithmic function using the Durbin-Watson test to measure the existence of the first-order autocorrelation of the time series composed of the annual coefficients, as well as to verify that the correlation was compatible with the random regression residuals hypothesis. Annual rates of increase or decrease (APC), according to maternal age and geographic clusters, were then calculated, with the respective confidence intervals (95% CI). This procedure makes it possible to classify gastroschisis trend, as increasing, decreasing, or stationary. The trend was, considered to be stationary when the coefficient was not significantly different from zero (p > 0.05) 31 . To facilitate the visualization of trends, the third order centred moving averages technique was performed for trends and for seasonality 40,41 , without the outliers (Figs 1-3).
To model seasonality we used monthly measurements for LB with gastroschisis. For monthly measurements, calendar month was numbered sequentially (totalling 66 months during 2011-2016). In addition, to identify seasonal variations, gastroschisis monthly prevalence rates were calculated on the basis of the date of conception. Antunes and Waldman 40 methodological indications were used for the seasonality hypothesis test. The seasonal variation was considered significant if one or more of the coefficients of the seasonal term (B3 and B5 for Seno and B2 and B4 for Cosseno) were statistically different from zero (p < 0.05) 31 . All statistical analyses, were performed using STATA 15.1 (CollegeStation, TX, 2018) conducted between February 2018, and August 2019.