Different Risk Factors for Very Low Birth Weight, Term-Small-for-Gestational-Age, or Preterm Birth in Japan

From 1985 to 2013, the mean birth weight of infants in Japan decreased from 3120 g to 3000 g, and the low-birth-weight rate among live births increased from 6.3% to 9.6%. No prospective study has elucidated the risk factors for poor fetal growth and preterm birth in recent Japanese parents, such as increased parental age, maternal body figure, assisted reproductive technology (ART), and socioeconomic status. Participants were mother–infant pairs (n = 18,059) enrolled in a prospective birth cohort in Hokkaido, Japan from 2002 to 2013. Parental characteristics were obtained via self-reported questionnaires during pregnancy. Medical records helped identify very-low-birth-weight (VLBW; <1500 g), term-small-for-gestational-age (term-SGA), and preterm-birth (PTB; <37 weeks) infants. We calculated relative risks (RRs) for PTB, VLBW, and term-SGA birth based on parental characteristics. The prevalence of PTB, VLBW, and term-SGA was 4.5%, 0.4%, and 6.5%, respectively. Aged parents and ART were risk factors for PTB and VLBW. Maternal alcohol drinking during pregnancy increased the risk; a parental educational level of ≥16 years reduced risk of term-SGA. Maternal pre-pregnancy BMI of <18.5 kg/m2 increased the risk of PTB and term-SGA. The RR for low BMI was highest among mothers who have low educational level. Among various factors, appropriate nutritional education to maintain normal BMI is important to prevent PTB and term-SGA in Japan.


Introduction
Poor fetal growth, such as low birth weight (LBW), small-for-gestational age (SGA), and preterm birth (PTB) have serious health effects not only during the neonatal period and infancy, but also The Hokkaido Study on Environment and Children's Health is an ongoing cohort study that began in 2002.The study's aims and methods were described in three previous profile papers and are only briefly discussed in this study [18][19][20].From February 2003 to March 2012, the Hokkaido (large-scale) cohort enrolled women during early pregnancy (13 weeks of gestational age) who visited the maternity unit in one of the 37 associated hospitals and clinics in the Hokkaido Prefecture for prenatal health.The 37 associated hospitals and clinics cover the whole Hokkaido area.The cohort consists of 20,926 pregnant women.Among them, 1347 were lost to follow-up before giving birth (Figure 1).As this study focused on the outcomes of VLBW, term-SGA, and PTB, we excluded women who had miscarriages, stillbirths, multiple births, pregnancy-induced hypertension, and gestational diabetes (n = 1176).Thus, we eliminated the pathological causes of VLBW, term-SGA, and PTB, which could have masked and underestimated the risk factors of parental characteristics.Participants lacking information on the three outcomes of interest were also excluded (n = 344).Thus, a total of 18,059 participants were included in the statistical analysis that assessed the associations between parental factors and VLBW, term-SGA, and PTB.

Outcomes of Poor Fetal Growth and Preterm Birt
Information on sex of infant, gestational age (days), and birth weight (g) were obtained at delivery from the medical records.We adopted three outcomes for poor fetal growth and preterm birth: VLBW, term-SGA, and PTB to evaluate "birth weight" and "preterm birth" as separate variables [23].PTB was defined as live birth at <37 completed gestational weeks; VLBW was defined as a birth weight <1500 g; and term-SGA was defined as a birth weight lower than the 10th percentile of the normative reference birth weight, according to gestational age, sex, and parity, in infants live born at >37 gestational weeks.To calculate term-SGA, we used the database for birth weight published by the Japan Pediatric Society as a reference [24], because Asian people are smaller than Caucasian people.

Outcomes of Poor Fetal Growth and Preterm Birt
Information on sex of infant, gestational age (days), and birth weight (g) were obtained at delivery from the medical records.We adopted three outcomes for poor fetal growth and preterm birth: VLBW, term-SGA, and PTB to evaluate "birth weight" and "preterm birth" as separate variables [23].PTB was defined as live birth at <37 completed gestational weeks; VLBW was defined as a birth weight <1500 g; and term-SGA was defined as a birth weight lower than the 10th percentile of the normative reference birth weight, according to gestational age, sex, and parity, in infants live born at >37 gestational weeks.To calculate term-SGA, we used the database for birth weight published by the Japan Pediatric Society as a reference [24], because Asian people are smaller than Caucasian people.

Data Analysis
Continuous data are presented as the mean and standard deviation (SD).Categorical data are presented as frequency and percentage.The Chi-square test was used to assess associations between VLBW, term-SGA, and PTB, and parental factors.The relative risks (RR) of VLBW, term-SGA, and PTB according to parental characteristics were estimated using multiple Generalized Linear Models (distribution: binominal, link function: logarithm).The models with each factor were adjusted according to maternal age and educational level.A directed acyclic graph (DAG) was constructed to identify a minimum set of confounding adjustment (Figure 2) [25,26].We selected the set of covariates for each factor that were regarded as the main exposure to effect on RRs of outcome such as VLBW, term-SGA, and PTB.The confounding factor(s) 1 (F1) that directly connected to both outcome and main exposure was (were) included.In addition, the confounding factor(s) 2 (F2) that directly connected to both outcome and F1 was (were) included.The mediating factor(s) that was (were) between outcome and main exposure was (were) excluded.The collider(s) that was (were) affected by both main exposure and F1 or F2 or outcome was (were) excluded.This shows the hypothesis of relationships between maternal and paternal, and socio-economic characteristics and outcome.We selected the set of covariates for each factor that was regarded as the main exposure that affected RRs of VLBW, term-SGA, and PTB.We excluded mediators and colliders from the covariates.We examined two-way interactions between each parental risk factor as a main exposure and covariations.When P interaction was less than 0.05, then each covariate was stratified for groups and a risk factor analysis of the main exposure was conducted.When we calculated RRs, we used the majority group (for parental age, maternal BMI, parity, maternal regular use of any supplement, parental education level, parental occupation, and household income) and the lowest risk group (for maternal active and passive smoking, paternal smoking habit, parental drinking habit, parental previous medical history, and maternal regular use of any medicine) as the reference categories.To estimate the RRs of term-SGA, only participants who delivered term infants were included in the analysis.

Data Analysis
Continuous data are presented as the mean and standard deviation (SD).Categorical data are presented as frequency and percentage.The Chi-square test was used to assess associations between VLBW, term-SGA, and PTB, and parental factors.The relative risks (RR) of VLBW, term-SGA, and PTB according to parental characteristics were estimated using multiple Generalized Linear Models (distribution: binominal, link function: logarithm).The models with each factor were adjusted according to maternal age and educational level.A directed acyclic graph (DAG) was constructed to identify a minimum set of confounding adjustment (Figure 2.) [25,26].We selected the set of covariates for each factor that were regarded as the main exposure to effect on RRs of outcome such as VLBW, term-SGA, and PTB.The confounding factor(s) 1 (F1) that directly connected to both outcome and main exposure was (were) included.In addition, the confounding factor(s) 2 (F2) that directly connected to both outcome and F1 was (were) included.The mediating factor(s) that was (were) between outcome and main exposure was (were) excluded.The collider(s) that was (were) affected by both main exposure and F1 or F2 or outcome was (were) excluded.This shows the hypothesis of relationships between maternal and paternal, and socio-economic characteristics and outcome.We selected the set of covariates for each factor that was regarded as the main exposure that affected RRs of VLBW, term-SGA, and PTB.We excluded mediators and colliders from the covariates.We examined two-way interactions between each parental risk factor as a main exposure and covariations.When Pinteraction was less than 0.05, then each covariate was stratified for groups and a risk factor analysis of the main exposure was conducted.When we calculated RRs, we used the majority group (for parental age, maternal BMI, parity, maternal regular use of any supplement, parental education level, parental occupation, and household income) and the lowest risk group (for maternal active and passive smoking, paternal smoking habit, parental drinking habit, parental previous medical history, and maternal regular use of any medicine) as the reference categories.To estimate the RRs of term-SGA, only participants who delivered term infants were included in the analysis.Since there were many missing values for parental factors, we imputed missing values using partial least squares regression.Two-sided values of p < 0.05 were considered statistically significant.All statistical estimates were calculated using JMP Clinical 5 statistical software (SAS Institute Inc., Cary, NC, USA).

Ethical Approval
All participating mothers provided written informed consent before participation in the Hokkaido Study.The study protocol was approved by the ethics review board for epidemiological studies at Hokkaido University Graduate School of Medicine (March 31, 2003) and the Hokkaido University Center for Environmental and Health Sciences (reference no.14, March 22, 2012), in accordance of with principles of the Declaration of Helsinki.
Table 3 presents the RRs of term-SGA and parental characteristics.In the adjusted model of term-SGA, compared to the infants whose mothers were in the standard BMI (18.5-25 kg/m 2 ) group, the RR was significantly higher for infants whose mothers were in the lowest BMI (<18.5 kg/m 2 ) category (RR = 1.77; 95% CI, 1.55-2.03)and was significantly lower for infants whose mothers were in the BMI (25.0-29.9kg/m 2 ) category (RR = 0.70, 95% CI, 0.53-0.93).The RR of being born term-SGA was significantly higher in infants whose mothers continued to drink during the first trimester (RR = 1.57; 95% CI, 1.33-1.85)compared with those whose mothers never drank alcohol, and whose mothers regularly used any supplement (RR = 1.16; 95% CI, 1.03-1.30).The RR of being born term-SGA was significantly lower in infants whose mothers had >16 years (vs.10-12 years) of education (RR = 0.76; 95% CI, 0.61-0.94).The RR of being born term-SGA was significantly lower in infants whose fathers had >16 years (vs.10-12 years) of education (RR = 0.86; 95% CI, 0.75-1.00).Maternal passive smoking was not significantly associated after adjustment.
We found significant interaction effects between pre-pregnancy BMI and maternal educational level (P interaction = 0.01) using ART and maternal age at entry (P interaction = 0.02) for PTB, and using maternal educational level and maternal age at entry (P interaction = 0.03) for VLBW.Table 5 presents the result of stratified analysis by interaction covariates to examine parental risk factors.After stratification, RRs of maternal educational levels were not significant for stratified analysis of maternal age at entry for VLBW.Similarly, the RR of low (<18.5 kg/m 2 vs. 18.5-25 kg/m 2 ) to PTB was significantly associated among mothers' educational level ≤9 years, 9-12, and >16 years, and among them, RR was the highest in the group of ≤9 years (RR = 2.31; 95% CI, 1.15-4.65).After stratification by maternal age at entry, the RR of PTB was significantly associated only among infants whose mothers used ART (RR = 2.06; 95% CI, 1.45-2.93) in the 25-35 years old mothers group.
Supplemental tables S1 to S3 present the results of the analysis, including substituted values of parental characteristics.The distribution of the parental characteristics in Supplemental tables S1 and S2 are comparable to those presented in Tables 1-4, respectively.As shown in Supplemental table S3, paternal smoking increased the RR of PTB, whereas lower household income (<3 million yen vs. 3-5 million yen) reduced the RR significantly.For VLBW and term-SGA, the results were comparable between models with and without imputed values.

Discussion
The mean birth weight and gestational age in this study were comparable to the data obtained from recent vital statistics of Japan [3].We evaluated the non-pathological maternal and paternal factors with three proxy indicators for poor fetal growth and preterm birth: VLBW, term-SGA, and PTB.The results showed that various parental factors were associated with each of these three outcomes and suggested that the life style and socioeconomic conditions in young Japanese women affected VLBW, term-SGA, and PTB in different ways.In short, higher maternal and paternal age and using ART were the main risk factors for VLBW and PTB, whereas life styles such as maternal alcohol drinking habits during the 1st trimester increased, but maternal and paternal educational level of ≥16 years decreased the risk for term-SGA.In addition, maternal pre-pregnancy BMI of <18.5 kg/m 2 was a risk factor for both term-SGA and PTB.Maternal and paternal factors were significantly correlated with each other, so that minimum and exact covariate factors should be selected for the adjustment model [14].Thus, DAG model was used to determine the effects of parental factors on VLBW, term-SGA, and PTB in this study.
Higher maternal and paternal age and using ART were the main risk factors for VLBW and PTB.Only 0.4% of infants were born with VLBW.Although the sample size was small, maternal and paternal ages of >35 years were significantly associated with VLBW.Advanced maternal age (≥35 years) has been previously reported as a significant risk factor for VLBW [27].In this study, advanced paternal age (≥35 years vs. 25-34 years) was associated with PTB and VLBW.Advanced paternal age has been reported as a risk factor for PTB-which is related to VLBW-even if maternal age is <35 years [15].Further studies that measure paternal involvement are needed to better assess the role of fathers in enhancing prenatal health behaviors and pregnancy outcomes.Because of the lifestyle of modern Japanese people, birth to advanced aged parents and the accompanying use of ART will continue to increase.For aged parents and when using ART, advanced knowledge on PTB and VLBW are needed, even if no visible pathological cause was observed.
In this study, the higher BMI, the higher the RR of VLBW (>30 kg/m 2 vs. 18.5-25 kg/m 2 ) in the crude and adjusted models, although the negative effect was insignificant in the DAG model.Studies conducted in the US and European countries have reported that a high pre-pregnancy BMI has a disadvantageous effect on fetal growth [28,29].The US and European countries considered a BMI of >30 kg/m 2 as the standard criterion for high BMI.Previous studies reported that Asians have a lower BMI, but a higher percentage of body fat than Caucasians [30,31].However, only 2.0% of 18,059 mothers had a BMI of >30 kg/m 2 in the present study, and the proportion of VLBW was only 0.4%.Hence, we were unable to detect the negative effect of >30 kg/m 2 BMI.Notably, Tables 2 and 4 present that low maternal BMI before pregnancy significantly increased the risk of PTB and term-SGA.Moreover, the results of interaction effects between pre-pregnancy BMI and maternal educational level for PTB presents that the RR of low BMI (<18.5 kg/m 2 vs. 18.5-25 kg/m 2 ) was highest in the group with educational level of ≤9 years (RR = 2.31; 95% CI, 1.15-4.65).Han et al. reported in a meta-analysis that a low BMI in pregnant women significantly increased the risk of VLBW, PTB, and intrauterine growth restriction [32].Moreover, nutritional deficiency during pregnancy should be considered among Japanese women.The Ministry of Health, Labour, and Welfare recommends that pregnant women consume 1800-2200 kcal/day.However, in 2011, the National Health and Nutrition Survey showed that the average intake among pregnant women was only 1665 kcal/day [33].Adequate knowledge on taking essential nutrition during pre-pregnancy and pregnancy should be provided.
Previous studies suggested that smoking during pregnancy decreased newborn birth weight and gestational age [34].However, in this study, active smoking was insignificantly associated with VLBW, term-SGA, and PTB, which was examined during early pregnancy (13 weeks of gestational age), so that quitting smoking during pregnancy reduced the risk of VLBW, term-SGA, and PTB.The effect of maternal smoking during the 1st trimester is unclear and has not been extensively studied [35][36][37].If mothers continue to actively smoke until the third trimester, then the negative impact on newborn birth weight and gestational age is inevitable.Indeed, we have reported that birth weight reduction showed a dose-dependent decreasing relationship with maternal prenatal cotinine levels during the third trimester in the same cohort [38].
In this study, maternal and paternal education (≥16 years vs. 10-12 years) significantly reduced the RR of term-SGA.As mentioned, not only maternal but also paternal educational level could be an important factor in avoiding the risks associated with term-SGA.A previous study in Japan reported that parental educational level was significantly associated with SGA [12].Education represents knowledge-related assets and indicates both economic resources and status.Socioeconomic factors may affect term-SGA via smoking and alcohol consumption [21], and indeed, alcohol drinking habit increased the risk of SGA in this study.Moreover, maternal education was significantly associated with smoking and alcohol consumption during pregnancy (data not shown).A smaller proportion of mothers in the highest educational categories were active smokers or alcohol drinkers during pregnancy (p < 0.01, data not shown).Education could be an important factor to avoid the risk factors associated with VLBW, term-SGA, and PTB.
We excluded women who had stillbirths, multiple births, pregnancy-induced hypertension, and gestational diabetes.Most stillbirth and multiple-birth infants show VLBW, term-SGA, or preterm characteristics.Pregnancy-induced hypertension and gestational diabetes have already been reported to have a decreased or increased effect on gestational age and birth weight [39].Furthermore, maternal chronic hypertension and pregnancy-induced hypertension have been associated with pre-pregnancy diabetes mellitus and gestational diabetes, respectively [40,41].These pathological factors associated with VLBW, term-SGA, and PTB could mask and underestimate the parental characteristics.Thus, in this study, we excluded mothers with hypertension and gestational diabetes, so that we could determine the impact of parental characteristics as a risk factor for VLBW, term-SGA, and PTB even without pathological basis.
The strengths of this study are as follows: first, it was a prospective birth cohort study design.Participants were recruited in a general hospital setting, such as local obstetric clinics.Second, the loss-to-follow-up rate was only 5.9%.Third, we initiated a DAG model to identify a minimum set of confounding adjustment, to avoid over-adjustment of our multiple analysis model.Limitations of this study included the following: first, the amount of missing data was relatively large.For example, 17.9% of household income data were missing.However, to estimate the effects of missing values, we imputed values using partial least square regression.The distributions of parental characteristics were comparable, and RRs were not different between raw and imputed data.Second, the participants of this cohort study were pregnant women who had visited hospitals or clinics within the Hokkaido Prefecture only.However, the participating hospitals and clinics were local medical institutions and distributed throughout the prefecture, accounting for approximately 40% of the institutes with delivery units in this prefecture [20].Moreover, the distribution of participant characteristics was close to that of the overall Japanese population [42,43], suggesting the results are generalizable.Third, possible residual confounding factors may also exist.Besides parental factors obtained using the questionnaire, residual confounding factors may have had an effect on VLBW, term-SGA, and PTB.

Conclusions
This study showed that different parental factors were associated with three proxy indicators of poor fetal growth and preterm birth: VLBW, term-SGA, and PTB in Japan.These results suggest that both maternal and paternal advanced age and using ART are predictors of VLBW and PTB.Maternal alcohol drinking habit increased the risk of term-SGA, whereas both maternal and paternal high educational levels were protective to term-SGA infants.In addition, maternal pre-pregnancy BMI of <18.5 kg/m 2 was a risk factor for both term-SGA and PTB.Moreover, the results of interaction effects between pre-pregnancy BMI and maternal educational level for PTB presents that the RR of low BMI was highest in the group with educational level of ≤9 years.

Figure 1 .
Figure 1.Flow chart of participants included in the statistical analysis.

Figure 1 .
Figure 1.Flow chart of participants included in the statistical analysis.

Table 1 .
Parental characteristics of participants.

Table 2 .
The prevalence and the relative risk of very low birth weight (n = 74) stratified by parental characteristics5.

Table 2 .
Cont.Maternal passive smoking at 1st trimester was adjusted by paternal active smoking during 1st trimester and parental educational level; Maternal drinking habit at 1st trimester was adjusted by maternal educational level; Maternal previous medical history was adjusted by maternal age, and maternal educational level; Maternal regular use of any supplement was adjusted by maternal age, maternal previous medical history, and maternal educational level; Using ART was adjusted by maternal age, maternal educational level, and household income.Maternal educational level was not adjusted by anything.Paternal age was adjusted by paternal educational level.Paternal active smoking at 1st trimester was adjusted by maternal educational level.Paternal previous medical history was adjusted by paternal age and paternal educational level.Paternal educational level was not adjusted by anything.Household Income was adjusted by parental age and parental educational level. 5: Term-small for gestational age (SGA) case group was compared with a control group of infants born at 37-41 weeks' gestational age.ART: assisted reproductive technology; BMI: body mass index; CI: Confidence Interval; DAG: directed acyclic graph; RRs: Relative Risks; VLBW: Very Low Birth Weight.
1 : Calculated by Chi-square test.2:Calculated by generalized liner regression models.3:Adjustmentmodelwasadjustedby maternal age, and maternal education.4:BasedonDAG model was as Figure2as follows: Maternal age was adjusted by maternal educational level; Maternal BMI was adjusted by maternal age, maternal active smoking, and maternal educational level; Maternal active smoking at 1st trimester was adjusted by maternal educational level, and maternal drinking habit during 1st trimester;

Table 3 .
The prevalence and the relative risk of term-small for gestational age (n = 1192) stratified by parental characteristics5.

Table 3 .
Cont.Maternal active smoking at 1st trimester was adjusted by maternal educational level, and maternal drinking habit during 1st trimester; Maternal passive smoking at 1st trimester was adjusted by paternal active smoking during 1st trimester and parental educational level; Maternal drinking habit at 1st trimester was adjusted by maternal educational level; Maternal previous medical history was adjusted by maternal age, and maternal educational level; Maternal regular use of any supplement was adjusted by maternal age, maternal previous medical history, and maternal educational level; Using ART was adjusted by maternal age, maternal educational level, and household income.Maternal educational level was not adjusted by anything.Paternal age was adjusted by paternal educational level.Paternal active smoking at 1st trimester was adjusted by maternal educational level.Paternal previous medical history was adjusted by paternal age and paternal educational level.Paternal educational level was not adjusted by anything.Household Income was adjusted by parental age and parental educational level.5:Term-small for gestational age (SGA) case group was compared with a control group of infants born at 37-41 weeks' gestational age.ART: assisted reproductive technology; BMI: body mass index; CI: Confidence Interval; DAG: directed acyclic graph; RRs: Relative Risks; term-SGA: term-Small for Gestational Age.
1 : Calculated by Chi-square test.2:Calculated by generalized liner regression models.3:Adjustmentmodelwasadjustedby maternal age, and maternal education.4:BasedonDAG model was as Figure2as follows: Maternal age was adjusted by maternal educational level; Maternal BMI was adjusted by maternal age, maternal active smoking, and maternal educational level;

Table 4 .
The prevalence and the relative risk of preterm birth (n = 805) stratified by parental characteristics.

Table 5 .
Stratified analysis by interaction covariates to examine parental risk factors for preterm birth and very low birth weight.