Investigation of optimal weight gain during pregnancy for Japanese Women

This study aims to compare the US Institute of Medicine (IOM) and Japanese guidelines proposed by the Ministry and the Japan Society for the Study of Obesity on gestational weight gain (GWG), and to explore the optimal GWG range in Japanese women. We investigated 8,152 Japanese women who had full-term singleton babies between 2010 and 2013 at a single center in Tokyo. Logistic regression models showed that GWG below the recommendation of the IOM and Japanese guidelines was similarly associated with an increased risk of light-for-date (LFD), whereas GWG above these guidelines was similarly associated with an increased risk of heavy-for-date (HFD) in pre-pregnancy body mass index categories of underweight (<18.5 kg/m2, n = 1559), normal-weight (18.5–24.9 kg/m2, n = 4998), overweight (25.0–29.9 kg/m2, n = 270), and obese (30 ≤ kg/m2, n = 60). The receiver-operating characteristic curve demonstrated that the optimal cutoffs for LFD and HFD were 9.7 and 10.4 kg, respectively in normal-weight mothers. The IOM and Japanese guidelines identified the risk of LFD or HFD equally well. The optimal GWG range in normal-weight women observed in this study was more close to Japanese guideline (i.e., 7–12 kg) compared to the IOM guideline (i.e., 11.5–16 kg).

Odds risks for a LFD or HFD infant among underweight, normal-weight, overweight and obese mothers. The odds risks for LFD or HFD are shown in Table 2 according to pre-pregnancy BMI categories.
Among normal-weight mothers (n = 4998), 14% (n = 691) experienced GWG below the JMHLW recommendations and 26% (n = 1279) experienced above the guideline. After adjusting for age, cesarean section, and   Among overweight mothers (n = 270), 46% (n = 124) experienced GWG below the JSSO recommendations and 54% (n = 146) experienced above the guideline. After adjusting for cesarean section selected by a stepwise model, weight gain below the IOM guidelines of 7 kg (OR 3.37, 95% CI: 1.20-9.47] or the JSSO guideline of 7 kg (OR 4.00, 95% CI: 1.45-11.00) was significantly associated with an increased risk of LFD. Stepwise logistic models did not find any covariates in overweight women, and accordingly univariate logistic regression models demonstrated that weight gain above the IOM guideline of 11.5 kg (OR 1.68, 95% CI: 0.75-3.73) or the JSSO guideline of 7 kg (OR 2.30, 95% CI: 1.17-4.52) was significantly associated with an increased risk of HFD.
Among obese mothers (n = 60), 53% (n = 32) experienced GWG below the JSSO recommendations and 47% (n = 28) experienced above the guideline. Stepwise model did not select any covariates for adjustment in obese women, and accordingly univariate logistic regression models demonstrated that weight gain below the IOM guidelines of 5 kg or the JSSO guideline of 5 kg was not significantly associated with an increased risk of LFD.
Stepwise logistic models did not find any covariates and accordingly univariate logistic regression models demonstrated that weight gain above the IOM guideline of 5 kg or the JSSO guideline of 5 kg was not significantly associated with an increased risk of HFD.
Area under the ROC curve (AUC) and sensitivity, specificity at the optimal cutoff point. Table 3 shows area under the ROC curve and sensitivity, specificity at the optimal cutoff point in normal-weight mothers (n = 4998). The area under the adjusted-covariates receiver-operating characteristic (ROC) curves (AUC) was between 0.61 for LFD (Fig. 1a) and 0.62, for HFD (Fig. 1b), which is considered acceptable. The adjusted model demonstrated that optimal cutoffs for LFD and HFD among normal-weight mothers were 9.7 kg and 10.4 kg, respectively, which were more close to the GWG range supported by the JMHLW (7-12 kg) compared to the GWG suggested by the IOM guideline (i.e., 11.5-17 kg).

Discussion
This study compared the risk of LFD or HFD associated with below or above the GWG recommendation of the IOM or JMHLW guideline in underweight and normal-weight mothers, and the GWG recommendation of the IOM and JSSO guideline in overweight and obese mothers, and also determined the optimal GWG during pregnancy in normal-weight women. The IOM and the two Japanese guidelines similarly discriminated the risks of LFD or HFD in underweight, normal-weight, and overweight mothers, whereas there was no significant findings observed in obese mothers but this may be due to small sample size of this group. In normal-weight women, the GWG range found in this study (9.7-10.4 kg) was close to the current JMHLW recommendation (7-12 kg) compared to the GWG proposed by the IOM guideline (i.e., 11.5-17 kg).
In 1990, the IOM revised the guidelines for weight gain during pregnancy, and recommended optimal weight gain ranges based on pre-pregnancy BMI 3 . Due to the absence of official GWG recommendations in Asian countries, including China 9 and Taiwan 10 , the IOM guidelines are generally followed. In 2016, large-scale retrospective studies of 97,157 Japanese women 4 and 48,867 Chinese women 11 found that the IOM guidelines can be applied to Asian women in that they can statistically assess the risk of delivery complications including LFD or HFD. However, Asian women are smaller and experience lower weight gains; therefore, excessive weight gain may lead to harmful events 12, 13 , including macrosomia, preterm birth, preeclampsia, gestational diabetes, pregnancy-induced hypertension, and short-and long-term postpartum weight retention 14 . Thus, an accurate GWG range should be determined for Asian women.
Although previous studies investigated the accuracy of guidelines by estimating odds ratios, the current study also investigated GWG ranges according to optimal cutoffs using ROC curves to predict LFD or HFD in normal-weight women. Among these individuals, the optimal GWG range was 9.7-10.4 kg, which is included  Table 3. Area under the ROC curve (AUC) and Sensitivity, specificity for at the optimal cutoff point in Normal weight. 18.5 ≤ BMI < 25 kg/m 2 (n = 4998). a Adjusting for age, cesarean section, and smoking. b Adjusting for age and cesarean section.
within the JMHLW recommendation of 7-12 kg. This GWG range was also supported by a study of 956 Taiwanese women that found an optimal range of 10-14 kg in normal-weight mothers 10 .
One strength of this study is that the dataset contained smoking information, which was absent in one of the largest epidemiological study in Japan 4 . Smoking is an independent risk factor for low birth weight babies, and it is therefore necessary to include this information especially in the estimation of the risk for LFD. In our study, smoking was selected by a stepwise multivariable model to estimate the risk for LFD in underweight and normal-weight women but overweight or obese groups; this was because the overweight and obese populations were too small for covariate-adjusted analyses. One limitation of this study was that all individuals analyzed were from a single hospital in Tokyo. Therefore, they may have been more affluent, healthy, and educated compared to the general population. Additionally, patients may have had unknown risk factors or complications during pregnancy given that the hospital is an advanced prenatal center. Thus, the applicability of the data may be limited to pregnant women in a large city such as Tokyo. Third, pre-pregnancy weight was measured based on a self-reporting which may cause information bias. In this regard, we performed the additional analyses by using maternal weight measured at 20-25 weeks and confirmed the result was not changed in normal-weight women; due to a large number of missing values on this variable, the analyses were unable to be performed in underweight, overweight, and obese women though. Fourth, we used customized birth weight centile charts that provide gestational week based birth weight according to gender of an infant. However, our dataset did not have the information of infant gender and thus we used an average of birthweight between a girl and a boy, which may induce bias. Fifth, in this study, our dataset does not have the information about diet, physical activity, which may affect weight gain during pregnancy. Sixth, ideal GWG ranges would be based on multiple factors, including parity, postpartum hemorrhage, preterm birth, gestational diabetes, pregnancy-induced hypertension, preeclampsia, and postpartum weight retention 16 . Thus, there might have been unmeasured confounders that should be included in our analyses. Seventhly, adjustments based on ROC curves were limited to normal-weight mothers due to small sample sizes of underweight, overweight and obese women. In spite of large sample size of normal-weight women, the AUC values for both LFD and HFD are slightly low (i.e., 0.61 and 0.62, respectively) compared to a reference of the acceptance (i.e., 0.60-0.80), and therefore there may be limited use in applying ROC curve to estimate GWG range. Finally, a large baby is known to be associated with a family history of diabetes mellitus 12 ; however, the current study did not include family histories. Hence, the result of our study requires careful interpretations.
In conclusion, no evidence was found in this study to change the current Ministry-recommended GWG guidelines for mothers with singleton babies regardless of pre-pregnancy BMI levels. This study also found that based on the covariate adjusted ROC curve, GWG of 9.7-10.4 kg in normal-weight mothers is suitable to minimize the risk of LFD or HFD.

Methods
Study design and participants. This cross-sectional study evaluated consecutive deliveries performed at a single hospital between January 2010 and June 2013. The hospital is a Red Cross medical Center in the Tokyo Figure 1. The covariate-adjusted receiver-operating characteristics curve of gestational weight gain for a lightfor-date infant (a) and for a heavy-for-date infant (b). The receiver-operating characteristics curve of gestational weight gain in the risk for a light-for-date infant (a) and for a heavy-for-date infant (b) was drawn based on Pepe & Longton method. Adjusted avariables selected at a stepwise multivariable logistic analysis included age, cesarean section, and smoking for a light-for-date infant (a) and age and cesarean section for a heavy-for-date infnat (b).
Metropolitan Area, which experiences the second largest number of deliveries in Tokyo. A total of 9,419 mothers visited and registered at the study site during the 3.5 years, and after excluding individuals for early miscarriages (n = 144), multiple pregnancies (n = 568), and stillbirths (n = 55), the dataset comprised 8,652 individuals who gave birth to singleton babies. This study focused on full-term deliveries, so individuals were also excluded due to preterm births (birth before 37 gestational weeks; n = 743), post-term births (births at 42 weeks or later; n = 24), or unverified gestational week (n = 430). Individuals with missing pre-pregnancy BMI (n = 122) or GWG data (n = 440) were also excluded, as were individuals with gestational weight loss during pregnancy of <8 kg (n = 5), who were considered outliers. Thus, a total of 6,887 individuals were assessed in this study (underweight, n = 1559; normal-weight, n = 4998; overweight, n = 270; and obese, n = 60). All participants provided written informed consent and the study enrollment is shown in Fig. 2. The ethics committee at the Teikyo University School of Medicine, Tokyo, Japan, approved this study (TU-COI 13-1592), confirming that all methods were performed in accordance with the relevant guidelines and regulations.
Outcome. LFD and HFD were the outcomes considered in this study, which refer to an infant born with a birth weight less than the 10th percentile and more than the 90th percentile for its estimated gestational age, respectively. We used the latest version of customized birth weight centile charts for Japanese issued by Japan Pediatric Society 17 . This new charts provide gestational week based birth weight according to gender of an infant. Because our dataset did not have the information of infant gender, we estimated an average of birthweight between a girl and a boy according to gestational week.
Pre-pregnancy BMI categories, weight gain during pregnancy, and covariates. Pre-pregnancy weight was obtained by the self-reporting and maternal height was measured at outpatient clinic for the first prenatal visit. Maternal weight at delivery was measured within 2 days after delivery and thus pre-pregnancy BMI and BMI at delivery were calcul ated as mother's weight obtained at two different points in time divided by her height squared. Weight gain during pregnancy (i.e., GWG) was calculated by subtracting pre-pregnancy weight from maternal weight at delivery.

Statistical analyses.
An association between pre-pregnancy BMI and baseline characteristics was assessed using the chi-square or the Fisher's exact test for categorical variables. Analysis of variance was used to assess continuous variables. Logistic regression analysis was used to estimate the odds ratios and 95% CI for GWG and covariates selected by a stepwise selection method were put into multivariable models. The recommended GWG by the IOM was compared to that by the JMHLW guideline for underweight and normal-weight individuals 7 , and compared to that by the JSSO for overweight and obese individuals 8 . Based on logistic regression models, crude ROC curves and covariate-adjusted ROC curves were used to determine the optimal cutoffs for Youden's index 18 . CIs for the AUC and covariate-adjusted ROC curves were calculated based on the method of DeLong et al. 19 or Pepe al. 20 . AUC values between 0.6 and 0.8 were considered acceptable based on Hosmer and Lemeshow 21 . Due to the limited sample size, we only focused on normal-weight women to estimate GWG with the lower limit which was considered the optimal cut-off for LFD, and the upper limit which was considered the optimal cutoff for HFD. All data were analyzed using SAS version 9.4 for Windows (Cary, NC, USA) and Stata version 14.2 (Stata Corp., College Station, TX, USA). All CIs were estimated at the 95% level, and P-values <0.05 were considered statistically significant. Data Availability. The data that support the findings of this study might be available from Japanese Red Cross Medical Center but restrictions apply to the availability of these data based on the ethical committee decision, and so are not be publicly available. Data might be however available from the authors upon reasonable request and with permission of Japanese Red Cross Medical Center.