Anaemia in children remains epidemic worldwide(Reference Balarajan, Ramakrishnan and Ozaltin1), adversely affecting their cognitive, motor and behavioural development(Reference McCann and Ames2). In 2011, 273 million children under 5 years of age were anaemic, with a prevalence as high as 43 %(Reference Stevens, Finucane and Deregil3). The underlying reasons are complex. Maternal nutrition during pregnancy, in addition to infection and maternal anaemia, is crucial(Reference Balarajan, Ramakrishnan and Ozaltin1,Reference Rahman, Abe and Rahman4) . Gestational weight gain (GWG) is a comprehensive indicator of maternal nutritional status, reflecting maternal fat accumulation, fetal growth, increased maternal blood volume and expansion of uterus and placenta(Reference Pritchar5,Reference Widen, Whyatt and Hoepner6) . A recent meta-analysis(Reference Goldstein, Abell and Ranasinha7) with a sample of over 1·3 million pregnant women showed that the prevalence of excessive GWG was 51 %, and the prevalence did not materially differ among the USA, Europe and Asia including China(Reference Ye, Yifan and Jie8). Given the universal excessive GWG, its adverse consequences on offspring health are much concerned(Reference Goldstein, Abell and Ranasinha9,Reference Bodnar, Siega-Riz and Simhan10) .
Excessive GWG has been linked to maternal anaemia during pregnancy(Reference Tan, Qi and He11,Reference Yanoff, Menzie and Denkinger12) that probably predisposes offspring to anaemia(Reference Meinzen-Derr, Guerrero and Altaye13). Two small studies have explored the association of GWG with Fe indices in cord blood(Reference Cao, Pressman and Cooper14,Reference Phillips, Roy and Lundberg15) , but no study has directly investigated the association between GWG and offspring anaemia. In this large Chinese birth cohort analysis, we examined (1) the relationship of GWG rate in the second/third trimester with infant Hb concentrations at 6 and 12 months; (2) the association between GWG rate and the risk of infant anaemia and (3) whether the association is mediated or modified by maternal anaemia during pregnancy.
Methods
Study population
Data were retrieved from a randomised controlled trial conducted in Hebei Province, China from May 2006 through April 2009. The trial investigated the effect of prenatal micronutrient supplementation on the pregnancy outcomes among 18 775 nulliparous women. The women with an Hb level greater than 100 g/l were enrolled before twenty gestational weeks and randomised to receive daily folic acid, Fe–folic acid (containing 30 mg Fe) or multiple micronutrients (containing 30 mg Fe); they were followed monthly from enrolment to delivery, and their infants were followed until 1 year after birth. The study design and primary findings of the trial have been detailed elsewhere(Reference Liu, Mei and Ye16).
For this present analysis, we initially excluded 1027 women who had spontaneous/induced abortions (n 815), stillbirths (n 82) or multiple pregnancies (n 67), as well as those who dropped out (n 33), moving out of the study area (n 28) or died (n 2), leaving 17 748 women who delivered singleton live births. We excluded a further 3983 mother–infant pairs with an end-pregnancy weight measured over 2 weeks before delivery (n 3533), missing data in the end-pregnancy weight records (n 18), an extremely low (–0·3 kg/week) or high (>1·5 kg/week) GWG rate (n 42), missing data on maternal Hb in mid-pregnancy (n 85) or missing data on infant Hb at 5–7 (n 169) or 11–13 (n 136) months. Finally, 13 765 (73·3 %) mother–infant pairs were included in our analyses (Fig. 1). Most of maternal and infant characteristics differed between the included (n 13 765) and excluded (n 5010) mother–infant pairs due to the large sample size, such as farmer (91·7 v. 88·7 %), vaginal delivery (49·6 v. 45·1 %) and exclusively breast-feeding (31·3 v. 42·2 %) (all P < 0·05), but the micronutrient supplementations were comparable between the included and excluded (P = 0·813) (online Supplementary Table S1).
The sample size (about 2750 per quintile group) would have a statistical power of 80 % or higher to detect a significant OR of 0·74 or lower (if a protective effect existed) or a significant OR of 1·31 or higher (if an adverse effect existed) with an estimated infant anaemia prevalence of 11·6 % in the reference group(Reference Fang, Yu and Guo17), at a two-sided significance level of α = 0·0125 (Bonferroni correction used for multiple comparisons).
The trial was approved by the Institutional Review Boards of the US Centers of Disease Control and Prevention (Atlanta, GA) and Peking University (Beijing, China). Informed consent to participate in the trial was obtained from all women.
Gestational weight gain rate in the second/third trimester
Maternal weight was measured at enrolment and at the end of pregnancy (within 2 weeks before delivery) by trained staff using standard procedures; height was measured at enrolment(Reference Liu, Mei and Ye16). Weight was measured to the nearest 50 g using an electronic scale (BW 150; UWE), and height to the nearest 0·1 cm using a collapsible height board.
In the analyses, GWG rate in the second/third trimester was the exposure measure. Compared with total GWG, GWG rate is preferable as it does not rely heavily on the duration of pregnancy and is less likely to be associated with pregnancy outcomes such as preterm, which could increase the risk of anaemia in later life(Reference Stockman, Garcia and Oski18). Considering weight gain is slower in the first trimester, and increases more rapidly and linearly in the second/third trimester(Reference Siega-Riz, Adair and Hobel19), the GWG rate in the second/third trimester was calculated separately for women enrolled before 12 gestational weeks and for those enrolled at 12–20 gestational weeks. For women enrolled before 12 gestational weeks, the GWG rate in the second/third trimester was calculated as ((total GWG – estimated first trimester weight gain)/(gestational age at delivery – 12 weeks)). Here, total GWG was calculated by subtracting the weight at enrolment from the end-pregnancy weight; the estimated first trimester weight gain was calculated as follows: (12 weeks – gestational age at enrolment) × 0·11 kg/week (the GWG rate in the first trimester for Asians)(Reference Abrams, Carmichael and Selvin20). For women enrolled at 12–20 gestational weeks, the GWG rate was calculated directly by dividing the total GWG by the number of weeks between enrolment and measurement of the end-pregnancy weight. As GWG rate differs with maternal BMI(21), the GWG rate was then categorised into highest, higher, middle (reference), lower and lowest quintiles by early pregnancy BMI categories (online Supplementary Table S2).
For women enrolled before 12 gestational weeks, early pregnancy BMI (kg/m2) was calculated as the weight at enrolment in kg divided by the height in m2. For women enrolled at 12–20 gestational weeks, weight in the first trimester was first calculated as: weight at enrolment – (gestational age at enrolment – 12 weeks) × 0·56 kg/week (the GWG rate in the second trimester for Asians)(Reference Abrams, Carmichael and Selvin20); the early pregnancy BMI was then calculated. Accordingly, women were categorised into four categories: underweight (<18·5 kg/m2), normal weight (18·5–22·9 kg/m2), overweight (23·0–27·4 kg/m2) or obese (≥27·5 kg/m2), based on the WHO recommended cut-offs for Asians(22).
Infant anaemia and covariates
The outcomes of interest in our study were the infant Hb concentration (g/l) and anaemia at 6 and 12 months. Hb concentrations were measured at 5–7 (mean, 6·3) and 11–13 (mean, 12·3) months of age using a HemoCue photometric instrument (Model 201; HemoCue) by trained staff, and anaemia was defined as an Hb concentration <110 g/l, in accordance with WHO guidelines(23).
Maternal Hb concentration was measured at enrolment and mid-pregnancy (24–28 gestational weeks), and anaemia was defined as an Hb concentration <110 g/l(24). During the 6-month postnatal visit, mothers were asked by healthcare providers to classify their infant feeding modes as follows: (1) exclusively breast-feeding, representing infants received only breast milk and without other liquids or solids with the exception of drops or syrups consisting of micronutrient, or medicines; (2) mixed feeding, representing infant received a combination of breast milk and formula milk and (3) formula feeding, representing infant received exclusively formula milk. Other covariates included maternal age (<25, 25–29 or ≥30 years), education (high school or above, middle school or primary school or less), ethnicity (Han or other), occupation (farmer or other), gestational age at enrolment (calculated by subtracting the date of last menstrual period from date of enrolment, continuous variable), gestational age at delivery (calculated by subtracting the date of last menstrual period from date of delivery, <37 or ≥37 weeks), mode of delivery (vaginal or Caesarean delivery), infant sex (male or female) and birth weight (<2500, 2500–3999 or ≥4000 g). Above all information on relevant covariates was retrieved from medical records or surveillance system.
Statistical analyses
Categorical variables are presented as frequencies and percentages, and continuous variables as mean values and standard deviations or as medians and interquartile ranges. The χ 2 test was used to compare categorical variables among GWG rate quintiles; ANOVA or the Kruskal–Wallis test was used to analyse continuous variables.
We first examined the relationship between GWG rate and infant Hb concentration using fractional polynomial model with generalised linear regression(Reference Royston, Ambler and Sauerbrei25) and estimated respective regression coefficients using linear regression. We then estimated the crude OR of infant anaemia across GWG rate quintiles using univariate logistic regression, with the middle quintile serving as the reference. We then estimated the adjusted OR using multiple logistic regression with covariates of maternal age, education, ethnicity, occupation, micronutrient supplementation, early pregnancy BMI, gestational age at enrolment (as a continuous variable), anaemia during mid-pregnancy (yes or no), gestational age at Hb measurement (as a continuous variable), gestational age at delivery, mode of delivery, infant sex, birth weight, feeding mode and age of infant at Hb measurement (as a continuous variable).
To evaluate the robustness of the results, we performed sensitivity analyses, first excluding women enrolled before 12 weeks of gestation and then those with anaemia at enrolment or during mid-pregnancy; an additional sensitivity analysis was performed by using multiple imputation methods in which missing values on the quintiles of GWG rate were imputed. For comparison with the results from Western countries, we repeated the main analyses using the categories of GWG rate in the second/third trimester (inadequate, appropriate and excessive) recommended in the 2009 Institute of Medicine guidelines(21) with BMI classified using the cut-off values recommended for Western populations(26): underweight <18·5 kg/m2; normal weight 18·5–24·9 kg/m2; overweight 25·0–29·9 kg/m2 and obese ≥ 30·0 kg/m2. We performed mediation and stratified analyses to examine whether maternal anaemia during mid-pregnancy, previously linked to offspring anaemia(Reference Zhang, Jin and Liu27), was a mediator or modifier of the association.
All analyses were conducted using SPSS software (version 24.0; IBM Corp.), and a two-sided P value <0·05 indicated statistical significance.
Results
The GWG rate in the second/third trimester was 0·5 (SD 0·2) kg/week for the 13 765 enrolled mothers, of whom 13 598 (98·8 %) were ethnic Han, 12 618 (91·7 %) were farmers and 13 554 (98·5 %) had a middle school education or above. Of the infants, 6887 (50·0 %) were born by Caesarean section and 4301 (31·3 %) were exclusively breast-feeding. Table 1 presents more detailed characteristics of the mothers and infants by GWG rate quintiles.
* P < 0·01.
Gestational weight gain rate and infant Hb concentrations
The mean Hb concentration of the infants was 121·7 (sd 8·7) g/l at 6 months and 122·2 (sd 8·2) g/l at 12 months. Fractional polynomial regression showed inverse and linear relationships between the GWG rate and Hb concentrations at 6 and 12 months (Fig. 2), that is, the Hb concentrations tended to be lower in infants born to mothers who gained weight at a faster rate during the second or third trimester. After adjustment for covariates, linear regression showed that the Hb concentration of a 6-month-old infant decreased by 3·22 g/l when maternal weight increased by 1 kg/week (Table 2).
GWG, gestational weight gain.
* Adjusted covariates included maternal age, ethnicity, education, occupation, micronutrient supplementation, early pregnancy BMI, gestational age at enrolment, gestational age at delivery, mode of delivery, and infant sex, birth weight, feeding mode and age of infant at Hb measurement.
Quintiles of gestational weight gain rate and infant anaemia
Of the 13 765 infants, 949 (6·9 %) were anaemic at 6 months and 728 (5·3 %) at 12 months. The results of the crude analyses were similar to those of the multivariate analyses (Table 3). After multivariate adjustment, the highest quintile of GWG rate as compared with the middle quintile was associated with an increased risk of anaemia at both 6 months (adjusted OR 1·30, 95 % CI 1·07, 1·59) and 12 months (adjusted OR 1·74, 95 % CI 1·40, 2·17); the lowest quintile was associated with a lower risk of anaemia at 6 months (adjusted OR 0·76, 95 % CI 0·61, 0·95), but not at 12 months. Moreover, after further adjustment for maternal anaemia during mid-pregnancy, the OR were slightly attenuated but remained significant. In the sensitivity analyses excluding women enrolled before 12 weeks of gestation, and then also excluding women who were anaemic at enrolment or during mid-pregnancy, the association between the lowest quintile and anaemia risk at 6 months became insignificant, while the associations with the highest quintile persisted (Fig. 3). In the sensitivity analysis with multiple imputation methods used, the results were not materially changed (online Supplementary Table S3).
* P < 0·05.
† Rate of infant anaemia.
‡ Adjusted covariates included maternal age, ethnicity, education, occupation, micronutrient supplementation, early pregnancy BMI, gestational age at enrolment, gestational age at delivery, mode of delivery, and infant sex, birth weight, feeding mode and age of infant at Hb measurement.
§ Additionally adjusted for maternal anaemia during mid-pregnancy.
Mediation analyses showed that maternal anaemia during mid-pregnancy mediated the association between the highest quintile and anaemia at 6 months (Z = 3·861, P < 0·001) and 12 months (Z = 3·536, P < 0·001).
In repeated analyses of GWG rate in the second/third trimester classified according to the 2009 Institute of Medicine guidelines, a positive association between excessive GWG rate and anaemia at 6 and 12 months was consistently observed (online Supplementary Table S4).
Discussion
In this large Chinese birth cohort, we found negative linear relations between GWG rate in the second/third trimester and infant Hb concentrations at both 6 and 12 months of age. When the Hb concentrations were dichotomised, the highest quintile of GWG rate as compared with the middle quintile was associated with a 30 and 74 % increased risk of infant anaemia at 6 and 12 months, respectively. The results persisted in infants born to mothers who were enrolled at 12–20 gestational weeks and born to mothers who were not anaemic at enrolment or during mid-pregnancy. The results were robust, when GWG rate was categorised according to the 2009 Institute of Medicine guidelines. That excessive GWG increased offspring anaemia risk appeared, to some extent, partly independent of maternal anaemia during pregnancy, though maternal anaemia probably mediated the associations.
To date, there has been no study to investigate the association between GWG and offspring anaemia. Two studies investigated the relationship of GWG with Hb concentration or Fe indices in umbilical cord blood(Reference Cao, Pressman and Cooper14,Reference Phillips, Roy and Lundberg15) . One cross-sectional study with a sample size of 230 American women reported negative relationships of GWG with Hb concentration and Fe indices in cord blood, but not statistically significant, possibly due to small sample size(Reference Cao, Pressman and Cooper14). Another cohort study with a sample size of 316 American women showed that GWG >18 kg v. <18 kg was associated with poor Fe indices in cord blood(Reference Phillips, Roy and Lundberg15). Our study with a large sample size consistently found that excessive GWG rate increased infant anaemia risk, and that the adverse effect persisted up to 12 months after birth. Underlying mechanisms remained unclear. Excessive GWG could expand blood volume or stimulate secretion of hepcidin(Reference Pritchar5,Reference Bekri, Gual and Anty28) , a biomarker of inflammation and a regulator of Fe metabolism; both blood volume expansion and higher hepcidin level are probably leading to maternal anaemia(Reference Tan, Qi and He11) that was linked to offspring anaemia(Reference Zhang, Jin and Liu27), as indicated by our mediation analyses. On the other hand, excessive GWG was likely to play an independent role. Pregnant women with excessive GWG rates were more likely to undergo a Caesarean delivery that was associated with an increased risk of offspring anaemia(Reference Zhou, Li and Ye29,Reference Li, Trasande and Zhu30) . Excessive GWG was also associated with higher birth weight that was linked to higher hepcidin levels in newborns, consequently leading to offspring anaemia(Reference Subhan, Colman and McCargar31,Reference Sanad, Osman and Gharib32) .
Unexpectedly, a protective effect of the lowest quintile of GWG rate on the infant anaemia at 6 months was observed in the overall analysis. Previous studies reported that women with normal weight were more likely to have a low GWG, probably because of balanced diet and appropriate physical activity(Reference Warren33,Reference Mamun, Kinarivala and O’Callaghan34) , benefiting to their offspring. In addition, we cannot completely rule out the possibility of residual bias in the protective effect. For example, women in the lowest quintile group were more likely to be enrolled before 12 gestational weeks (62·7 %), which indicated that they began supplements consumption at an earlier gestational age and thus a lower risk of offspring anaemia. Consistently, in sensitivity analyses by excluding women enrolled before 12 weeks of gestation, the protective effect turned to be insignificant (as shown in Fig. 3(a)).
To the best of our knowledge, this is the first large-scale prospective study to investigate the relationships of GWG rate in the second/third trimester with infant Hb concentrations and anaemia risk at 6 and 12 months of age after birth. Key variables including maternal weight, infant Hb and major covariates were measured or collected by trained healthcare providers using standard procedures. The GWG rate rather than the total GWG was assessed as the exposure measure, assuredly minimising potential biases related to pregnancy duration. Our study had several limitations. First, the data were retrieved from a randomised controlled trial of prenatal Fe-containing micronutrient supplementation on well-nourished women whose Hb concentrations were greater than 100 g/l at enrolment, probably reducing the generalisability of the findings to other populations. For example, our analysis might have been biased by the start time of micronutrient supplementation because pregnant women were enrolled at different gestational age, such that our findings could not necessarily apply to populations where extensive supplementation was routine(Reference Vaivada, Gaffey and Bhutta35). Second, though we could not exclude a possibility of biases related to Fe-containing supplementation, if any it ought to have the estimated risk towards null. Besides, we did not observe a significant interaction between GWG rate and supplementation type (all P interaction > 0·05, online Supplementary Table S5). Third, we did not collect data on maternal and infant health status, duration of breast-feeding and complementary foods after 6 months(Reference Pham, Nguyen and Salvignol36,Reference Wang, Liu and Wan37) which might bias our results on infant anaemia at 12 months. Fourth, data on biomarkers of Fe and inflammation were not recorded. Given the above limitations, further studies in China are needed to confirm the findings.
In conclusion, this prospective cohort analysis showed negative linear relationships between GWG rate in the second/third trimester and infant Hb concentrations. The excessive GWG rate, partly independent of maternal anaemia during pregnancy, appeared to be associated with an increased risk of anaemia in infants. Maternal anaemia probably mediated the associations. These findings indicate a need to optimise GWG and control maternal anaemia during pregnancy to improve offspring health. Although optimum GWG needs to be further studied, balanced diet and appropriate physical activity during pregnancy are warranted to be advocated. Further studies are needed on whether these findings are applied to other populations, and on appropriate GWG in different populations.
Acknowledgements
We thank study participants, healthcare workers from five counties in northern China and all staff who participated in the project at the Peking University Health Science Center and Centers for Disease Control and Prevention, Atlanta.
The present study was supported by the National Key Research and Development Program of China (J. L., grant numbers 2016YFC1000401 and H. L., 2016YFC1000406-1) and the National Natural Science Foundation of China (J. L., grant number 81571517, and Y. Z., grant number 81801542).
J. L. conceived the study; S. Y., J. L. and Y. Z. wrote the manuscript; S. Y. and Z. C. performed the statistical analyses; H. L. Y. Z., L. Z., J. L. and J. L. contributed to the critical review of the manuscript; H. L., Z. C., Y. Z., L. Z. and J. L. participated in the acquisition of the data; J. L. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis and all authors read and approved the final manuscript.
The authors declare that there are no conflicts of interest.
Supplementary material
For supplementary materials referred to in this article, please visit https://doi.org/10.1017/S0007114520002354