The present study utilizes data from the China Kadoorie Biobank (CKB), a prospective database, on 302,510 women aged between 30 and 79 years that were recruited between 2004 and 2008 from 10 geographically defined regions of China. From an initial cohort of 302,510 women, 2,881 women who reported never having been pregnant and 47 women who had missing data were excluded. The remaining 299,582 women were included in the final analysis. The study design, characteristics of participants, and survey methods of the CKB database have been previously described in detail elsewhere (26,27).

Briefly, data was collected through an interviewer-administered questionnaire, which included but was not limited to sociodemographic status, lifestyle factors, and personal and family medical history. Personal medical history consists of physician-diagnosed disease status. Respondents were asked: “Has a doctor ever told you that you had the following disease?” followed by a list of diseases including cardiovascular disease (coronary heart disease, hypertension, and stroke), and diabetes (27). The definitions of hypertension and diabetes included either the self-reported physician-diagnosed hypertension and diabetes or screening-detected hypertension and diabetes from the physical examination (28,29). In the present study, screen-detected hypertension was defined if participants had a measured diastolic blood pressure ≥90 mm Hg or systolic blood pressure ≥140 mm Hg. For self-reported diabetes, those diagnosed at an age below 30 years and currently being treated with insulin were considered as probable cases of type 1 diabetes, and were excluded at baseline enrollment. Screen-detected type 2 diabetes was defined as a random blood glucose level ≥11.1 mmol/L with a fasting time <8 h or a random blood glucose level ≥7.0 mmol/L with a fasting time ≥8 h or a fasting blood glucose ≥7.0 mmol/L. In addition, the median age of these participants at diabetes diagnosis were 54.00 (IQR: 48.00–60.00) years. The age of the majority of women at diabetes diagnosis was past the age of reproduction, therefore gestational diabetes may not occur. Factors relating to the reproductive history of women include age at menarche, parity, number of spontaneous abortions and stillbirths, menopausal status, and contraceptives use were also collected. Spontaneous abortion was defined as the loss of pregnancy naturally in the absence of elective medical or surgical measures to terminate the pregnancy. Stillbirth was classified as the death or loss of a fetus before or during delivery. Participants were considered to have pregnancy loss if they experienced spontaneous abortion or stillbirth. Anthropometric measurements: weight, height, and waist-hip circumference, were taken by trained technicians using standard protocol and procedures. Body mass index (BMI) was calculated as weight divided by the square of height (kg/m²). The physical activity level was measured by adding up metabolic equivalent tasks (METs) for daily work or leisure activities.

The CKB study was given ethics approval from the University of Oxford, Peking University, the China National Center for Disease Control and Prevention (CDC), and the institutional review boards of the local CDCs in the study areas. All participants have provided written informed consent according to the Declaration of Helsinki for participation (26).

Death, including the cause of death, and health outcomes was collected periodically from baseline until 31 December 2016, via linkages with hospital records, national health registries, and social health insurance databases in the study areas. To minimize loss to follow-up, active follow-up involving visitations to the local community or direct contact with participants was performed annually (26). Fatal events entered into the CKB follow-up system were coded according to the 10th International Classification of Diseases (30). All records, including scanned images of original death certificates, were reviewed centrally by study clinicians blinded to baseline information (26). The outcome in analyses of this study was death from all causes. All participants were prospectively followed up, with a median follow-up of 10.20 years.

Baseline characteristics were presented as means (SD) for continuous variables and as percentages for categorical variables, stratified by the number of spontaneous abortions and stillbirths. Continuous variables were compared using the one-way analysis of variance (ANOVA) test for variables with normal distribution and the Kruskal-Wallis test for variables with skewed distribution. Categorical variables were compared using the chi-square test. The outcome was divided into two categories by the presence of all-cause mortality (yes or no). Cox proportional hazards regression was used to obtain the hazard ratio (HR) and 95% confidence intervals (CI) for the associations between pregnancy loss (total pregnancy loss, spontaneous abortion, and stillbirth) and all-cause mortality. Covariates were selected for inclusion in the models based on prior knowledge and published literature (13,19). Models were adjusted for age, region, BMI, education, annual household income, physical activity, smoking, alcohol consumption, cardiovascular disease, diabetes, number of livebirths and, where appropriate, number of spontaneous abortions, and stillbirths.

Subgroup analyses were also performed to obtain the adjusted HR (aHR) and 95% CI for the association between pregnancy loss and all-cause mortality by the presence of cardiovascular disease and diabetes, age (<50 or ≥50 years), BMI (<25 or ≥25 kg/m²), study region (rural or urban), level of education (elementary school and below, middle and high school, or university and above), annual household income (<20,000 or ≥20,000 yuan/year), and MET (<17 or ≥17 h/day). Statistical significance was set at p < 0.05 for all statistical analyses and were performed with SAS software package version 9.4 (SAS Institute Inc.).

The characteristics of the participants stratified by the number of pregnancy loss are presented in Table 1. Among a total of 299,582 women who reported having ever been pregnant, 41,571 (13.88%) have experienced at least one pregnancy loss, of which 27,156 (9.06%) had a history of spontaneous abortion, and 17,041 (5.69%) had a history of stillbirth. Compared to women who have never had a pregnancy loss, women who have had a history of pregnancy loss were older (50.84 vs. one: 54.42, two or more: 57.86), had lower BMI (23.86 vs. one: 23.65, two or more: 23.32), were more likely to reside in rural regions (53.47% vs. one: 66.58%, two or more: 74.93%), were more likely to have educational levels elementary school and below (54.50% vs. one: 69.21%, two or more: 77.73%), have incomes of less than 2,500 yuan (2.91% vs. one: 3.88%, two or more: 5.35%), had lower MET hours (20.75 vs. one: 19.08%, two or more: 16.97%), were more likely to smoke (4.90% vs. one: 5.94%, two or more: 6.16%) but were less likely to drink alcohol (37.31% vs. one: 31.45%, two or more: 28.01%), and had more livebirth numbers (2.15 vs. one: 2.68, two or more: 3.06).

The association between pregnancy loss with all-cause mortality is presented in Table 2 . Compared to women without pregnancy loss, women with a history of two or more pregnancy loss have higher all-cause mortality: aHR 1.10, 95% CI 1.03–1.18. Similarly, women with more than one spontaneous abortion or stillbirth have higher all-cause mortality: aHR 1.10, 95% CI 1.01–1.21 for two or more spontaneous abortions, and aHR 1.14, 95% CI 1.04–1.25 for two or more stillbirths, respectively. However, the associations were not statistically significant for women with one pregnancy loss, spontaneous abortion, or stillbirth.

The associations between pregnancy loss and all-cause mortality stratified by age as well as the presence of cardiovascular disease or diabetes are presented in Table 3. Compared to women aged <50 without a history of pregnancy loss and cardiovascular disease, women aged ≥50 with a history of two or more pregnancy loss and cardiovascular disease had higher all-cause mortality: aHR 1.32, 95% CI 1.18–1.48. Women aged ≥50 with a history of two or more pregnancy loss and diabetes also had higher all-cause mortality: aHR 1.30, 95% CI 1.06–1.60.

Women aged ≥50 with a history of two or more spontaneous abortion and cardiovascular disease had higher all-cause mortality: aHR 1.27, 95% CI 1.07–1.50. Similarly, women aged ≥50 with a history of two or more stillbirth and cardiovascular disease had higher all-cause mortality: aHR 1.39, 95% CI 1.20–1.62. Women aged ≥50 with a history of stillbirth and diabetes also had higher all-cause mortality: aHR 1.24, 95% CI 1.03–1.51 and 1.41, 95% CI 1.06–1.87 for one and two or more stillbirths, respectively.

Compared to women aged ≥50 with a history of pregnancy loss (total pregnancy loss, spontaneous abortion, and stillbirth) and cardiovascular disease, women aged <50 with a history of pregnancy loss (total pregnancy loss, spontaneous abortion, and stillbirth) and without cardiovascular disease were not significantly associated with all-cause mortality. The associations were also not significant for women aged <50 with a history of pregnancy loss (total pregnancy loss, spontaneous abortion, and stillbirth) and without diabetes.

The associations between each additional pregnancy loss and all-cause mortality stratified by age, BMI, study region, educational level, annual income and MET hours are presented in Figure 1. Of statistical significance was the association between pregnancy loss and all-cause mortality in women between the ages <50 and ≥50 years old (<50 years old: aHR 1.12, 95% CI 1.03–1.21; ≥50 years old: aHR 1.15, 95% CI 1.12–1.19), whose BMI was less than 25 (aHR 1.05, 95% 1.01–1.09), who resided in rural regions (aHR 1.06, 95% CI 1.02–1.09), with educational levels elementary school and below (aHR 1.04, 95% 1.01–1.07), who had annual incomes of less than 10,000 yuan (aHR 1.04, 95% CI 1.01–1.08), and who had MET hours of less than 17 per day (aHR 1.04, 95% CI 1.01–1.08). However, the association between pregnancy loss and all-cause mortality in women whose BMI was ≥25, who resided in urban regions, who had educational levels middle and high school or university and above, who had annual income of more than 20,000 yuan, and who had MET hours ≥17 per day were not statistically significant.

The association between stillbirth with the risk of all-cause mortality in the subgroup analyses was broadly similar to that of pregnancy loss in general, while the association between spontaneous abortion with the risk of all-cause mortality in the subgroup analyses was largely non-significant.

This study is based on a large-scale prospective cohort which included women from 10 diverse regions of China, ensuring the generalizability of our findings to the Chinese population. The baseline survey also covered a comprehensive range of demographic, socioeconomic, and lifestyle information, allowing us to adjust for many confounders. In addition, data for this study was collected via standardized approaches with stringent quality control, making our results reliable and reproducible.

However, there are still several limitations that should be noted. First, the pregnancy history was self-reported, which may result in some certain recall bias. Second, although we have performed stratification analysis to identify associated factors, we were not able to include all the conditions or risk factors for pregnancy loss before or during pregnancies due to the lack of data. Third, although it is most likely that the majority of women developed diabetes after the age of reproduction, a small number of women included in our study may have gestational diabetes. Future research is needed to explore the potential effect of gestational diabetes on the association between pregnancy loss and all-cause mortality. Finally, residual confounding from other known or unknown risk factors may still exist despite a wide range of potential confounders taken into account for analysis.

In conclusion, we found that recurrent spontaneous abortion and stillbirth, was associated with an increased risk of all-cause mortality in Chinese women, particularly among older women with cardiovascular disease or diabetes at baseline. Future studies are needed to further elucidate the mechanisms underlying these relationships to prevent the onset of long-term adverse health outcomes in women.