From 1980 to 2004, mean birthweight decreased in Japan, especially between 1985 and 1999. The decline was not fully explained by adjustment for variables obtained from birth certificates.
Contrary to the end of the 20th century, birthweight began to decrease in USA , Sweden, and South Korea[5, 22] .
Some studies are examining changes in birthweight among subgroups, for example, fill-term homogenous subset of low-risk mothers [2, 19, 21] . As expected, the decline in BW in present study was mainly caused by gestational age in the multivariable regression analysis. Through adjustment by all variables used in the multiple regression analysis, decrease in birthweight was 5.13 gr/year compared to crude decrease of 8.07g/year.
However, the observed decreases in gestational age did not entirely explain the declines in birthweight, which suggests that decline in birthweight occurs within specified weeks of gestation. Morisaki et al  pointed out that birthweight is decreasing regardless of gestational age in USA.
Strengths of this study is that the large nationally representative data are used, the form of which does not change for decades. We had information on potential factors associated with trends in birthweight.
One of the limitations of the present study is the estimation of GA. GA was filled in birth certificate by doctors or midwives in the hospitals or clinics. Some were based on the date of the last normal menstrual period and others were on early fetal ultrasound. This may impair the validity of our gestational age estimation.
Kramer  points out that gestational age calculated by the last menstrual period was shown to be fairly accurate among term births. Considering obstetrical practice, gestational length is affected by mode of delivery. Induction of labor also steadily increased by years in Japan  . Rates of cesarean deliveries based on national growth survey increased after 2000 . These factors may cause the decline in gestational length. In US, gestational length decreased in neonates born both by vaginal delivery and by cesarean section. It also decreased both in induced labor and non-induced  .
Another limitation of the present study is the lack of data which may be associated with birthweight, such as maternal smoking status, pre-pregnancy weight and maternal diet during pregnancy. Maternal smoking restricts fetal growth restriction, and increases obstetrical complications, preterm births and stillbirths .
In Japan, proportion of pregnancy smoking was 5.0% in 1990, 10.0% in 2000 and 5.0% in 2010. Moreover, in a recent survey on mothers and children aged 3 to 4 months, it was reported that LBW was significantly associated with maternal smoking.  Decrease in proportion of pregnancy smoking could be one of the causes for slowing of decrease in BW after 2000. In Japan, Health Promotion Law was enforced in 2002 and people were encouraged to quit smoking afterwards.
Information on maternal pre-pregnancy weight was not included in our data. BMI distribution among woman within reproductive age could be the proxy for pre-pregnant BMI. The prevalence of underweight has been increasing through decades in Japan in contrast to other countries  . In fact, the increase in birthweight in Sweden between 1992 and 2001 was explained by increases in maternal BMI in the same period along with decreases in maternal smoking . In a hospital-based study in Canada, increase in birthweight was explained by increase in pre-pregnancy BMI, increase in gestational weight gain and decrease of prevalence of smoking during pregnancy . In Sweden, from 1978 to 1992, birthweight increased within all maternal BMI categories after adjustment for gestational length, age, smoking habits, parity, and employment. After that, however, birthweight decreased among normal-weight women  . Our data did not include pre-pregnancy BMI. Besides, according to National Health and Nutrition Survey  , the proportion of women with BMI less than 18.5 whose age is between 20 and 39 years is increasing concomitantly with the decline in mean birthweight (Figure2). For the discussion of causality between maternal BMI and birthweight, further investigation is needed.
Our data also did not include maternal diet which may affect fetus growth. Rather, in Japan, through the results of National Health and Nutrition Survey  time trends of per capita calorie intake look like synchronized (Figure 3), however the causality is still unclear.
Other factors suggested for increasing low birth weight in Japan is the declining in adult height  . Increase in maternal age  is shown as a factor for decline in birthweight in Korea. In our study, maternal age did not show linear correlation to birthweight and could not be included as explanatory variable.
In this study of neonate born in Japan from 1980 to 2004, decreases in birthweight were not fully explained by factors included in the birth certificates. Decrease in gestational age only partially explained the decreasing birthweight. Birth size may influence not only short-time conditions but long-time prognosis  . Factors lowering birthweight, though not analyzed in the present study, could cause various health problems among children while they grow up. So follow-up study would be necessary to investigate what sequalae would derive from small birthweight neonates.
In conclusion, our study based on birth certificates shows that over the decades, birthweight of infants has been getting smaller. These findings may partially be explained by the decline in gestational age, considered to be derived from change in mode of delivery. Their clinical and social significance has yet to be determined.