Risk Factors for Low Birth Weight Infants in Japan

Prevalence of low birth weight infants in Japan has been increasing, although infant mortality is the lowest in the world. We conducted a population based case-control study to clarify risk factors for low birth weight infants in Japan. Information was obtained by questionnaire. The response rate was 78.9% from 402 cases and 804 controls. The majority of low birth weight infants belongs to full-term case group. This is considered to be an important factor for low infant mortality in Japan. We examined potential risk factors by a multiple logistic regression model. Lower maternal academic career and toxemia were identified as significant risk factors for pre-term low birth weight infants, while maternal shorter height, lower prepregnancy weight, maternal smoking and employment during pregnancy and hypertension were identified as significant risk factors for full-term low birth weight infants. The Japanese small physique and prevalent hypertension are considered to be causes for the high prevalence of low birth weight infants. J Epidemiol, 1994; 4: 91-98.

Prevalence of low birth weight infants in Japan has been increasing, although infant mortality is the lowest in the world. We conducted a population based case-control study to clarify risk factors for low birth weight infants in Japan.
Information was obtained by questionnaire. The response rate was 78.9% from 402 cases and 804 controls.
The majority of low birth weight infants belongs to full-term case group. This is considered to be an important factor for low infant mortality in Japan.
We examined potential risk factors by a multiple logistic regression model. Lower maternal academic career and toxemia were identified as significant risk factors for pre-term low birth weight infants, while maternal shorter height, lower prepregnancy weight, maternal smoking and employment during pregnancy and hypertension were identified as significant risk factors for full-term low birth weight infants. The Japanese small physique and prevalent hypertension are considered to be causes for the high prevalence of low birth weight infants. J Epidemiol, 1994; 4: 91-98.
case-control study, japan, logistic regression model, low birth weight infant Infant mortality in Japan has remarkably improved since the end of the Second World War1,2). However, the prevalence of low birth weight infants (hereinafter "LBW infants") has continued to increase since 1976, although it had decreased until 1975. The prevalence of low birth weight infants in Japan is not low among developed counties (Table 1), although the infant mortality is lowest in the world3). This is notable because low birth weight is an important factor for both post-neonatal mortality and morbidity during infancy and early childhood4-9). There are few studies to introduce risk factors for LBW infants in Japan to other countries. Therefore, it is very important to identify risk factors for LBW infants and to clarify causes of the increasing prevalence of LBW infants in Japan.
There are many factors related to a higher prevalence of LBW infants10). Epidemiological studies have demonstrated a close relationship between low birth weight and socio-economic conditions11-14). Multivariate analyses have been employed to estimate the magnitude of the effect of risk factors associated with LBW infants15-21).  important to differentiate risk factors between the two. In Japan, we identify risk factors for pre-term and fullterm LBW infants, respectively, in order to clarify causes of the increasing prevalence of low birth infants in Japan, although the infant mortality rate is the lowest in the world.

MATERIALS AND METHODS
The studied area was the N district and the T district of Kobe city in Japan.
From birth certificates, we determined that cases were all single live LBW infants born in the area from January 1, 1986 to December 31, 1987. The number of cases was 402. Controls were selected from single live infants whose birth weight ranged from 3,100 g to 3,400 g, which represents Japanese typical birth weights.
For each case, we selected two controls whose birth district and gender were the same, and whose birthdays were closest to one another.   3. maternal life conditions during pregnancy: maternal smoking, paternal smoking, alcohol consumption, sleeping hours, stress factors (factors having persistently bothered a mother during pregnancy), emesis, employment, standing work, lifting work (> 5kg), VDT (Visual Display Terminal) work, night work; 4. maternal pregnancy history: number of pregnancies, previous abortions, previous premature births ; 5. maternal complicating diseases during pregnancy : toxemia, hypertension, nephritis, anemia, diabetes mellitus, heart disease, collagen disease and thyroid disease.
The questionnaire was sent to the mothers of 402 cases and 804 controls. It was answered by 308 (76.6%) cases and 644 (80.1%) controls (Table 2). One case and one control were excluded from the analysis because their birth weights proved to be unsatisfactory with the criteria by the data re-examination. As a result, 307 cases and 643 controls were analyzed. Although we asked the mothers to answer regarding pregnancy history including the recent pregnancy, many mothers answered excluded the recent pregnancy. Because of the confusion, we excluded pregnancy history from our analysis.
The cases were divided into pre-term and full-term case groups. In order to clarify the characteristics of the two case groups and the control group, we calculated their means and standard deviations for continuous variables, while we described distribution for categorical variables.
We could not directly compare the distributions of risk factors among the groups because we sampled controls matched by district and gender. Therefore, we separately employed a simple logistic regression model of each variable for detecting potential risk factor for each case group, using the same control group. The two matching factors, district and gender, were included in the model only for controlling their effects. Then, we did not interpret the regression coefficients of the matching factors because we did not have any knowledge of the sampling fraction of the control group from the mother population. However, the regression coefficients of potential risk factors are guaranteed as their log odds ratios because of the independence between the matching factors and the potential risk factors. Variables of which the total number of people having risk factors less than five were excluded from further analysis because of small sample size. These were prenatal care, maternal complicating disease of cardiac disease, collagen disease and thyroid disease.
Finally, we applied a multiple logistic regression model of all variables which were statistically significant (p<0.05) by the primary analysis, in addition to district and gender. Cases and controls who did not answer all significant risk factors were excluded from the analysis. As significant risk factors were different between pre-term and full-term case groups by the primary analysis, the numbers of controls were not exactly the same between two analyses ; for pre-term cases versus controls and for full-term cases and controls. Thus, 106 pre-term cases and 621 controls were analyzed for the former, while 182 full-term cases and 623 controls were analyzed for the latter. et al.

RESULTS
The proportion of the pre-term case group was 37.8%, while that of the full-term case group was 62.2%. LBW infants whose birth weights were less than 1,500 g were only 5.9% among LBW infants and all of them were preterm cases (Table 3). Among 643 controls, only 2 babies were pre-term.
Means and standard deviations of continuous variables are shown in Table 4, while distributions of categorical variables are shown in Table 5. Odds ratios, adjusted by district and gender, for pre-term and full-term LBW infants are shown in Table 6. Parental ages of pre-term LBW infants were older than control group. Parental heights of  Table 6. Results of simple logistic regression analysis (adjusted by district and gender). * toxemia were identified as significant risk factors for preterm LBW infants, while maternal shorter height, maternal lower pre-pregnancy weight, maternal smoking and employment during pregnancy and maternal complicating disease of hypertension were identified as significant risk factors for full-term LBW infants (Table 7).

DISCUSSION
In the study, the response rates ranged from 73.8% to 84.6% among the groups. They were high for a population based study and there were not many differences among the groups. Although it is likely that mothers whose babies had died responded differently from mothers whose babies were alive, this would not change the results because the infant mortality rate is lower than 0.5 (per hundred births).
The percentage of the full-term LBW group, whose prognosis is much better than that of the pre-term LBW group, was 62.2% among LBW infants. It was not so different from 68.9% in total Japanese population 25). In the United States, only about one third of LBW infants have a gestational age of 37 weeks or more10). It is considered to be a substantial factor of low infant mortality in Japan that the majority of LBW infants belongs to fullterm case group. In addition, our health care system has contributed to it, which was verified by the fact that only one person among 950 mothers in this study had not had prenatal care.
Lower maternal academic career and maternal complicating disease of toxemia were significant factors for preterm LBW infants by the multivariate logistic regression model. As academic career is an indicator for socioeconomic status, our study is consistent with the study showing that socioeconomic status is related to pre-term birth26). Toxemia has been established as a cause of preterm birth27,28) Although ten risk factors for pre-term LBW infants were detected by the simple logistic regression model, only two factors were significant by the multivariate logistic regression model. It suggests that most risk factors conditions of working women are less nutritional than those of housewives because the former spend less time on meals than the latter, and the former skip breakfast more frequently than the latter39,40). Further investigation is important to make clear why maternal employment has an effect on birth weight, especially as the number of working women is increasing. As hypertension was reported to have negative effects on birth weight by impairing placental blood circulation, it is understandable that hypertension is a risk factor for full-term LBW infants"). Japan has been reported to have a higher prevalence of hypertension because of its high salt diet42.43). It would influence the prevalence of full-term LBW infants in Japan. However, there are several limitations in this study. First, this study was a questionnaire survey. Information bias is not avoidable. Second, we excluded pregnancy history from our analysis. Although we cannot deny the possibility that it was a confounding factor, the risk factors which we identified were consistent with other studies. Third, some factors were considered not being significant because the number of mothers having risk factors was too small. However, it means that they are not so important in the population. Finally, this study was conducted only in one of big cities in Japan, although Japan has a homogeneous population racially and socio-economically.
The majority of LBW infants belongs to the full-term case group in Japan. This is considered to be an important reason why Japan has the lowest infant mortality rate, although LBW infants are still prevalent.
The small physique and prevalent hypertension of Japanese pregnant mothers might explain the high prevalence of LBW infants.
Also increasing maternal smoking and increasing number of working women should be considered as influencing the increasing low birth weight infants in Japan.