The cohort study of mother-child pairs was conducted at Women and Children’s Hospital, School of Medicine, Xiamen University. This hospital is the leading provincial medical center for obstetrics referral center and has an average of 15,000 deliveries every year. Registered women with single pregnancy in first trimester were incorporated into the study cohort on an informed consent basis from 1 January 2020 to 30 June 2021. Pregnant women who did not participate in the whole process or have placenta previa or placental abruption were removed from the cohort study. These pregnant women were divided into two subgroups according to the existence of high-risk factors of preterm birth. All women were followed up until the end of pregnancy, including miscarriage, induced abortion, and delivery. In this study, the high risk group of PTB was defined as history of PTB and late abortion in previous pregnancies, a short cervical length (less than 25 mm, measured between 16 and 24 weeks of gestation), preterm birth at 12 months inter-pregnancy interval, and a history of cervical surgery (surgical treatment of dysplasia with cold-knife conization, loop electrosurgical excision), which are undisputed high risk factors for PTB according to the 2021 ACOG Guidelines [4]. The remaining the pregnant women were classified as the low-risk group for PTB. We included all mothers from the low-risk group of the cohort study who delivered a live singleton after more than 28 gestational weeks. This present study was conducted in accordance with the China approved guidelines.

The incidence of PTB, defined as delivery occurring before 37 weeks of gestation, was the primary outcome. The week of gestational was estimated by the last menstrual period. Early pregnancy ultrasound was used to verify gestational age.

GWG recommendations updated by the IOM 2009 are based on PP-BMI from the World Health Organization (WHO) and are independent of parity, age, race, ethnicity, or smoking history [17]. The measure of GWG used in the study was average rate of GWG in second and third trimester, calculated by dividing the difference of weight between the last and first prenatal care visit in the second and third trimester by the corresponding difference of gestational weeks, respectively.

According to 2009 IOM standards, PP-BMI was classified into four subgroups: underweight (18.5 kg/m${}^2$), normal weight (18.5–24.9 kg/m${}^2$), overweight (25–29.9 kg/m${}^2$), and obese ($\ge$30 kg/m${}^2$). Each category of PP-BMI has a recommendation for the range of GWG: underweight (0.44 to 0.58 kg/week); normal weight (0.35 to 0.50 kg/week); overweight (0.23 to 0.33 kg/week), and obese (0.17 to 0.27 kg/week). According to the range, the rate of GWG was categorized as: insufficient, adequate, and excessive.

Maternal and neonatal demographic and clinical features are reported as frequency (%) or means ($\pm$ Standard deviation (SD)). To compare the maternal sociodemographic characteristics, pregnancy history, and GWG between the preterm birth and normal group, univariate logistic regression analysis was utilized. Logistic regression models were performed to estimate the odds ratios (ORs) and 95% confidence interval (CI) for preterm birth and its subgroups. Maternal pre-pregnancy BMI, maternal age, education level, type of resident, parity, gravidity, and infant sex were adjusted in our analyses. All statistical tests were conducted by the two-tailed test, and p 0.05 was considered significant. All statistical analyses were done with SPSS 21.0 software (IBM Corp., Chicago, IL, USA).

Contradictory results between PP-BMI, GWG, and PTB have been reported previously. We used a cohort study of mother-child pairs with a one-and-a-half-year follow-up to assess the relationship between PP-BMI, GWG, and PTB in women at low risk for preterm birth. The results demonstrated that pregnant women with a low rate of GWG had an increased risk of PTB compared to women with an adequate rate of GWG. These associations were more obvious in pregnant women classified as underweight or normal weight. We did not find an association between excessive rate of GWG and PTB risk, which is inconsistent with previous studies [14, 18]; however, the risk increased if those women had a low PP-BMI, which indicated that the association between GWG and PTB varies according to PP-BMI. Our analysis identified low GWG and excess GWG with low PP-BMI as important and modifiable risk factors for PTB.

Although the biological mechanisms linking GWG to PTB are not fully understood, some pathogenic mechanisms have been postulated. Maternal weight change may be a marker of nutritional status and is also considered to be a marker of many physiological processes. It has been suggested that low GWG may indirectly stimulate the production of corticotropin-releasing hormone and prostaglandins, which increases the sensitivity to uterine contraction [19]. Insufficient GWG may also reflect nutrient deficiency, which may be one of the pathways leading to PTB. Anorexia is associated with zinc deficiency, which leads to low GWG. Alternatively, low GWG may be due to maternal anemia during pregnancy. Both factors can lead to an increased risk of PTB [20]. A low GWG may also be associated with increased inflammation and poor plasma volume expansion, both of which are potential mechanisms of PTB [20].

Twenty years ago, a few studies reported that low GWG increased the risk of PTB. However, some previous studies have shown a conflicting correlation between excessive GWG and PTB [1, 21]. Some scholars found the risk of PTB increased among pregnant women with excessive and low GWG [22]. Recent studies have also reported a positive correlation between excessive GWG and PTB [23, 24]. Weight gain above target guidelines was associated with a reduced risk of PTB in the United States and Europe [13]. This controversial result may be due to different definitions of excessive and insufficient GWG and different analytical methods.

As in most previous studies, we found that insufficient GWG significantly increased the risk of PTB, especially among pregnant women classified as underweight or normal weight. We did not find an association between excessive rate of GWG and PTB risk; however, the risk increased if women had a low PP-BMI, which is consistent with some studies [11, 25]. Although the biological mechanism is not clear, it can be considered that a reasonable GWG during pregnancy can effectively reduce the risk of PTB, especially for pregnant women with low PP-BMI. Moreover, low or excessive GWG may lead to an increased risk of PTB.

Our study had some advantages. First, given the potential confounding bias, our study used a cohort of mother-child pairs with a one-and-a-half-year follow-up, and only women lacking undisputed high-risk factors for PTB were included in the study. Several studies used retrospective designs to investigate the association between GWG and PTB, which maybe a source of bias [25, 26, 27]. While most previous studies used statistical methods to correct for research bias caused by other high-risk factors of PTB, we excluded the PTB high-risk group from the beginning of the cohort study to accurately describe the relationship between GWG and PTB. Second, due to the different length of pregnancy in the study population, we used the GWG rate to correct the study bias. A key element cited in previous studies [21, 23] on the relationship between GWG and PTB was total gestational weight gain, but women who delivered preterm had less time to gain weight, which may lead to less GWG. Finally, our data were robust. The research was carried out at a tertiary hospital that is the leading provincial medical center for obstetrics referral and has an average of 15,000 deliveries every year. Our hospital strictly implements common guidelines concerning pregnancy and delivery management and the cohort study was performed according to strict standards.

Our study had some limitations. First, it was a single-center design and the study population size may not be sufficient. Therefore, the findings may not be generalizable for all medical institutions. Hence, a continuous cohort study and a larger study population are necessary for future studies. Second, a recent study [28] proposed a generalized methodology with a hierarchical procedure (including multivariable machine learining statistics) assessing relative importance of risk factors and capable of selecting those which are essential for risk prediction vs those that are pleonastic and may be eliminated. This is the suggested method required to assess in future studies the importance of novel proposed risk factor. Finally, the uncertainty caused by COVID-19 pandemic contribute to the anxiety of pregnant women [29]. We did not consider the impact of the pandemic even though there were a few COVID-19 patients in the city of the study.