The impact of early-life antibiotics and probiotics on gut microbial ecology and infant health outcomes: a Pregnancy and Birth Cohort in Northwest China (PBCC) study protocol

The PBCC is a prospective, ongoing pregnancy and birth cohort study conducted in rural, suburban, and urban areas, which are representative of the different economic levels, geographic coverage, and medical services in Shaanxi Province, Northwest China from January 2018 to December 2024. The rural site is the People’s Hospital in Binzhou city, as well as six town health centres of primary health facilities. The suburban site is Lintong District Maternal and Child Hospital, Xi’an city, which serves a mixed urban and rural population, and the urban site is the Second Affiliated Hospital of Xi’an Jiaotong University. Specifically, the birth numbers at the three sites varied, and the average numbers per year are 1731, 2040, and 3000 at the rural, suburban, and urban sites, respectively. Almost all pregnant women will deliver in hospitals, while some may deliver at home in rural areas (average of 2 per year). The caesarean rate in the three site hospitals varies between 25 and 50%, and the average rate of NICU admission is 10% but varies with hospital level. Critically ill pregnant women will be transferred to higher-level hospitals in advance. After delivery, the exclusive breastfeeding rates declined from 58.3% among newborns to 29.1% at three or four months and declined further to 13.6% among those aged five to six months [31], and the exclusive breastfeeding rates in rural areas, small urban areas and megacities at 6 months were 28.3%, 23.3%, and 35.6%, respectively. After weaning, infants usually receive formula. Child health care centres are usually located in the family’s district within a 1-h walk, and the proportion of children receiving health care is 50%-70%, but this proportion is lower in rural areas because a greater proportion of parents work outside the town.

From the perspective of antibiotic and probiotic use in early life, we considered using probiotics to prevent antibiotic-associated diarrhoea (AAD) in infants to calculate the sample size. The rate of AAD after broad-spectrum antibiotic treatment is reported to be 25%-53% in children [32]. A meta-analysis of probiotic uses to prevent AAD and Clostridium difficile infection among hospitalized patients revealed a statistically significant reduction in the risks of AAD (RR = 0.61, 95% CI 0.47 to 0.79) and a lower number of CDI at the end (RR = 0.37, 95% CI 0.22 to 0.61) [33].

Therefore, with a statistical power of 90% (β = 0.1) and a two-tailed significance level of 5% (α = 0.05), for each district, the sample size was calculated using the following formula:

where p₀ denotes the risk of AAD in the control group, p₁ denotes the risk of AAD in the treatment group, and \(\overline{q }\)=1-\(\overline{p }\).

Then, we estimated 354 as the minimum number of participants for each group (AAD or no AAD). Considering a 15% foetus attrition rate and 5% loss to follow-up rate, we expanded the number by 20%. Hence, for the three districts, we need to enrol 2,550 participants. Finally, after discussion with the site managers, 3,000 eligible women and their healthy, singleton infants will be recruited for this study.

We collaborate with local doctors, nurses, and medical staff. The pregnant women are enrolled during their first antenatal care (ANC) visits (any trimester) by obstetricians and gynaecologists in the collaborating units. Written informed consent is obtained from the women after enrolment.

The inclusion criteria are as follows: (1) pregnant women who planned to deliver in the collaborating hospital; (2) singleton foetus; and (3) women agreeing to completing at least 2 years of postnatal follow-up. The exclusion criteria are as follows: (1) planning to move out of the town; (2) rejecting or not adhering to the 2-year follow-up; (3) participating in other research projects; and (4) chronic disease or chronic medication use (diagnosed pre-pregnancy). As a result, we will enrol term and preterm infants of any gestation, including neonates who require intensive care, and the medical records will be carefully collected.

Data collection entails three stages including pregnancy, delivery and after birth, as shown by the timeline in Fig. 1. The ANC, delivery record, and follow-up data from the hospitals are independently managed by the hospital and collected by hospital staff using their routine hospital information system (HIS).

Faecal samples will be collected in three stages (shown in Table 1). The sample collection and laboratory experiments are as follows:

We use a one-time sterile collection tube to collect fresh faecal samples with a 75% alcohol-disinfected stencil-sealed cover, which will be transferred quickly within two hours (using an icebox for temporary storage) to the hospital’s clinical laboratory and preserved in a -20 °C refrigerator and frozen after marking. Then, every two weeks, these samples will be temporarily transferred to a mobile fridge, with an average travel time of 60 min, ranging from 30 to 120 min to the lab located at Xi'an Jiaotong University, School of Public Health Centre. All the samples will be stored in a -80 °C refrigerator.

Bacterial DNA is extracted from faecal samples with a QIAamp Fast DNA Stool Mini Kit (Qiagen, Cat. No./ID: 51,604), and PCR amplification is conducted with barcode-specific bacterial primers targeting the variable regions V3-V4 of the 16S rRNA gene: the forward primer 341F for 5′-CCTAYGGGRBGCASCAG-3’ and 806R for 5′-GGACTACNNGGGTATCTAAT-3’ [34]. Sequencing libraries and paired-end reads are generated on an Illumina HiSeq 2500 platform from Biomarker Technologies Co.Ltd. (Beijing, China) according to a standard protocol.

We then use quantitative PCR (qPCR) to detect the expression of ARGs in faecal samples. After a careful literature review, we selected 6 ARGs of four kinds of antibiotics: beta-lactam, tetracycline, fluoroquinolone, and macrolide, which are commonly used during pregnancy and early life [12, 35, 36]. Each sample includes three replicates for data accuracy.

Data will be collected and managed using REDCap (Research Electronic Data Capture) tools by investigators, and questionnaire data entry will be conducted as soon as possible at each timepoint. Moreover, the external data including the records from the HIS, will be inputted regularly. All pregnant women enrolled in the study will be included in the final analysis.

The characteristics of the women or infants are described as the means ± SDs or medians for quantitative variables and percentages with numbers for categorical variables. One-way ANOVA is used to test the difference in means for continuous variables, and chi-square tests or Fisher’s exact test are used to summarize the associations of categorical variables.

To analyse the 16S rRNA gene sequence data, we will not only perform cross-sectional comparisons across individuals at the same time but also perform longitudinal follow-up within the same individual at different time points. All the data will be analysed using QIIME tools. (1) For community diversity and dissimilarity, alpha-diversity metrics including diversity (Shannon, Simpson) and richness (Chao, ACE) are used to measure diversity within species, and beta-diversity metrics are employed using the Bray‒Curtis or binary Jaccard distance matrix to measure the differential abundance among species based on antibiotic exposure, probiotic consumption, delivery mode and feeding patterns. (2) Then, principal coordinate analysis (PCoA), nonmetric multidimensional scaling (NMDS), the unweighted pair-group method with arithmetic means (UPGMA) and heatmap of species clusters are used to conduct cluster and dimensionality reduction. (3) Furthermore, linear discriminant analysis (LDA) effect size (LEfSe), and Metastats methods are used to identify significant differences from the phylum to genus levels among groups. (4) Multivariable analysis is used to adjust for confounding factors and analyse the associations between clinical and microbiota data and identify the influencing factors of abundance and diversity of the mother-infant microbiota. (5) For longitudinal data analysis of gut microbiota changes over the first two years, a linear mixed effects model or multilevel mixed effects model are used. (6) A Dirichlet multinomial mixtures (DMM) model [40] is used for gut microbiota community type determination.

All analyses are performed using Stata v.13.1 (StataCorp, College Station, TX), and visualization is performed using R statistical software v3.6.1 (http://​www.​r-project.​org/​) with the ggplot2 package. Statistical significance is evaluated using a 95% confidence interval and a two-tailed p value of < 0.05.

There are several strengths in our study. First, the study is conducted in underdeveloped districts containing rural, suburban, and urban areas in Northwest China, which could be representative and provide evidence for a similar study in other less-developed districts worldwide. Second, not only exposure data during pregnancy, but also delivery and early life data of health outcomes are collected. Third, to ensure a high follow-up response rate, a detailed investigation plan had been formulated. Furthermore, for antibiotic and probiotic assessment, a questionnaire of knowledge and behaviour related to antibiotics for parents is conducted, and the Bayley Scales of Infant and Toddler Development (Bayley-IV) are used, to make the evaluation of the results more multidimensional. However, the study has some limitations. The registration of the study is relatively late due to the COVID-19 pandemic, but the normality of the cohort is not affected by late registrations. A possible selection bias exists in the study population because investigators recruited individuals who cooperated rather than a random sample, and the rate of antibiotic use varies during pregnancy, so a large sample size is needed. Moreover, due to limitations in medical knowledge and women’s migration for work, many participants will fail to complete the follow-up of the two-year study. However, we will keep close contact with the participants to limit loss to follow-up. In addition to the time points aligned with scheduled immunization and health services, it seems urgent that we combine the information with health service systems of different institutions for the integrity of the data. In the future, if the study budget and time permit, the follow-up time will be extended.

The PBCC provides a multidimensional platform for studying antibiotic resistance and health outcomes during early life through a longitudinal perspective with a focus on the alteration of the gut ecology. The rich dataset from three representative areas will inform antibiotic and probiotic use for pregnant women and infants and contribute to establishing rational strategies on the use of antibiotics and probiotics for paediatricians, health practitioners, and drug administration policy-makers. In the study, we will extend the analysis of biological samples to breast milk, vaginal swabs and oral swabs from the same cohort in further studies. Furthermore, we are considering collaborating with microbiologists to create a biobank of probiotic strains to serve as stable selection samples for human use. Finally, we intend to combine data from different institutions (including primary health care units and the Centre for Disease Control) and multiple biological samples (including breastmilk, vaginal and oral samples) to explore new evidence between antibiotic/probiotic exposure during the perinatal period and health outcomes in early life.