Association of age at first birth and risk of non-alcoholic fatty liver disease in women: evidence from the NHANES

Numerous studies have suggested that age at first birth (AFB) is inversely associated with metabolic diseases, but positively associated with liver cancer in women. Non-alcoholic fatty liver disease (NAFLD) is a canonical example of metabolic dysfunction and inflammation-based liver disease, while the association between AFB and the risk of NAFLD remains unclear. We aimed to investigate the association between AFB and the odds of NAFLD in women. Women older than 20 years at the time of the survey were analyzed using National Health and Nutrition Examination Survey (NHANES) data from 1999 to 2018 in the US. AFB was obtained with self-administered questionnaires. NAFLD was diagnosed as fatty liver index (FLI) ≥ 60. Odds ratios (ORs) and 95% confidence intervals (CI) were estimated using logistic regression models. Of the 12,188 women included in this study, 5670 (46.5%) had NAFLD. Compared to individuals with AFB of 30–32 years old (reference group), the fully adjusted ORs and 95% CI in women with AFB < 18, 18–20, 21–23, and 24–26 years were 1.52 (95% CI 1.14, 2.03), 1.60 (95% CI 1.21, 2.11), 1.40 (95% CI 1.06, 1.84), and 1.33 (95% CI 1.01–1.76), respectively. Yet there was no significant difference between AFB of 27–29, 33–35, or > 35 years compared to the reference group. Women with younger AFB have higher odds of NAFLD in later life. Policymakers should consider focusing on those with earlier AFB for screening and prevention of NAFLD.


Background
Globally, the prevalence of non-alcoholic fatty liver disease (NAFLD) is reported to range from 25 to 45% [1]. The prevalence is increasing at an alarming pace as the rates of obesity continue to rise. NAFLD is a syndrome characterized by excessive triglyceride (TG) accumulation within hepatocytes due to a cause other than alcohol consumption [2], and it has become the most common chronic liver disease. Furthermore, a major challenge in global has been the large number of patients with NAFLD and its comorbidities, which create a huge economic burden. Clarify the risk factors for NAFLD would contribute to health policymaking and therefore assist to address this challenge.
Epidemiological evidence shows that NAFLD prevalence in women increases significantly after menopause [3]. Women's health and wellbeing are profoundly affected by estrogen levels, which can vary dramatically with age and decrease rapidly after menopause [4]. In addition to the putative protective effect of estrogen, other reproductive factors may also play a key role in the development of NAFLD.
Pregnancy is a fundamental factor affecting women's future health status [5]. During pregnancy, to support fetal development, the maternal body undergoes changes in hormones, immunity, and metabolism [6]. There are a number of studies showing conflicting results on the association between age at first birth (AFB) and women's health in later life. Some studies have suggested that women with an early AFB are at higher risk for later-life metabolic diseases and mortality [7][8][9][10][11], while others found that an early AFB is associated with a decreased liver cancer risk [12]. NAFLD as a canonical example of metabolic dysfunction and inflammation-based liver disease is also an important risk factor for liver cancer. However, the association between AFB and the long-term consequences of NAFLD remains unclear. Thus, the present study aims to investigate the relationship between AFB and the risk of NAFLD using the National Health and Nutrition Examination Survey (NHANES) data from 1999 to 2018 in the US.

Study population
The NHANES is a complex, stratified, multistage, probability-cluster program designed to assess Americans' health and nutritional status. Females older than 20 years at the time of the survey from NHANES 1999-2018, a total of 10 cycles, were included in this analysis.

Reproductive factors
Reproductive factors were obtained from self-administered questionnaires. Participants recalled their AFB, age at menarche, and age at menopause at the time of the survey through the reproductive health questionnaires. Investigators also gathered their parity, personal medication (including birth control and hormone pills), use of female hormones (not for birth control or infertility), and gynecologic surgical (e.g., oophorectomy or hysterectomy) history at the same time. Fertile lifespan was calculated by age at menopause minus age at menarche. Note that AFB refers to the first live birth in this study, pregnancy loss (prior stillbirth, miscarriage, or ectopic pregnancy) or stillbirth were not included.

Other covariates
Demographic information of each participant was collected with a questionnaire at the time of enrollment. The race was dichotomized as white or non-white; the former refers to non-Hispanic white; the latter includes Mexican American, non-Hispanic black, Asian, and other races. Education level was categorized into "less than 12th grade", "high school or equivalent", and "college graduate or above". Poverty-income ratio (PIR), a ratio of family income to the poverty threshold, reflects the annual family income level. Well-trained health technologists measured weight and height according to the anthropometry procedure manual. BMI was calculated as weight (kg) divided by height squared (m 2 ). Lifestyle factors including smoking status, alcohol intake, and leisure activity levels were assessed by separate questionnaires, respectively. In NHANES, diet was assessed using 24-h dietary recalls, which collected information on everything the participants consumed in the preceding 24 h. The healthy eating index (HEI) was calculated based on these data. One of our previous works described this assay in more detail [17], with minor modifications.

Statistical analysis
Analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, North Carolina), accounting for the Mobile Examination Center (MEC) exam sampling weights and the complex survey design of NHANES; p values of less than 0.05 were considered to be statistically significant. The missing variables were imputed as the most common value for categorical variables, and median imputation was used for missing continuous covariates. The preliminary analyses were descriptive statistics; continuous variables were shown as survey-weighted means with SE, whereas categorical variables were expressed as cases (n) and percentages (%).
For primary analyses, we performed multivariate logistic regression models to analyze the association between AFB and NAFLD. We assumed a lower risk of NAFLD among females with AFB of 30-32 years than in others, since the lowest survey-weighted prevalence of NAFLD, as well as the lowest level of BMI, WC, ALT, AST, FLI, and NFS in this group. Thus, the AFB group of 30-32 years was used as the reference group. Odds ratio (OR) with 95% confidence interval (CI) was used to determine the degree of association. We did not adjust for any covariate in Model 0, and adjusted for age (years), race (Mexican American, other Hispanic, non-Hispanic white, non-Hispanic black, other race), education level (less than 12th grade, high school or equivalent, college graduate or above), and family PIR in Model 1; Model 1 plus smoking status (never, ever, current), alcohol intake (drinks/week), leisure activity (MET-min/day), total energy intake (kcal/day), and HEI-2015 constituted Model 2; we additionally adjusted for parity (times), menopause status (yes/no), use of female hormones (never/ever), oral contraceptive use (never/ever), hysterectomy (yes/no), both ovaries removed (yes/no), age at menarche (≤ 11, 12-13, ≥ 14 years), and fertile lifespan (< 28, 28-35, > 35 years) in Model 3, which was the fully adjusted model. Within the fully adjusted model, a restricted cubic spline in a random-effects dose-response model was used to illustrate the nonlinear association between AFB and NAFLD. The spline curve was drawn with Stata software version 15.1.
We further carried out stratified analysis according to race (white or non-white), age at the survey (< 45 or ≥ 45 years), menopause status (yes or no), menopause causes (natural, hysterectomy, hysterectomy& bilateral oophorectomy), ever used female hormone therapy (never, ever), age at menarche (≤ 11, 12-13, ≥ 14 years), and fertile lifespan (< 28, 28-35, > 35 years) to explore potential sources of heterogeneity. Then five sets of sensitivity analyses were performed to test the robustness of the findings by excluding individuals who had undergone gynecologic surgery (had hysterectomy or ovaries removed), non-normal age at menarche (< 9 or > 18 years), or age at menopause (< 40 or > 55 years or premenopausal), or with parity more than 5, respectively.

Results
Of the 12,188 women included in this study, 5,670 (46.5%) had NAFLD. The baseline characteristics of the participants meeting inclusion criteria by categories of AFB are summarized in Table 1. Participants were grouped into AFB < 18, 18-20, 21-23, 24-26, 27-29, 30-32, 33-35, and > 35 years. Family PIR, HEI-2015 levels, and proportion with higher education increased progressively with an increase in AFB. Women who gave birth first between 30 and 32 years of age had the lowest level of BMI, WC, ALT, AST, FLI, and NFS, and were less likely to be current smokers, postmenopausal, or to have ever used female hormones, and also the lowest proportion of having had a hysterectomy and both ovaries removed; meanwhile, they had the highest leisure activity, age at menarche, age at menopause, and fertile lifespan. On the other hand, compared to non-NAFLD participants, individuals with NAFLD were older, with higher BMI, more parity, lower family PIR level, lower leisure activity level, lower dietary quality (HEI-2015), younger age at menarche and menopause, and younger AFB (Supplemental Table 1). Also, individuals with NAFLD were more likely to be non-white, have lower education levels, have postmenopausal status, and have had a hysterectomy and both ovaries removed.
The survey-weighted percentages of NAFLD were 50.27%, 48.59%, 43.21%, 38.87%, 35.01%, 27.99%, 34.77%, and 28.11% for the eight groups, respectively ( Table 2). Compared to participants with AFB between 30 and 32 years old, those with AFB ≤ 29 years showed higher ORs of NAFLD (from 1.39 to 2.60, p < 0.05). Further adjusting for confounding factors substantially reduced the identified associations, especially in participants with AFB of 27-29 years. The association significance became weaker but persisted for the groups of AFB ≤ 26 years after adjustment of sociodemographic characteristics, lifestyle, and reproductive factors (Model 3); the fully adjusted ORs ranged from 1.33 to 1.52 with p < 0.05. However, there was no significant difference between AFB of 33-35 or AFB > 35 years compared with the reference group. We examined the nonlinear association of AFB with NAFLD, and a U-shaped curve tendency was observed (Fig. 2).
Results varied between prespecified subgroups of race, current age, menopause status, menopause causes, ever used female hormone therapy, age at menarche, and fertile lifespan. A stronger association between AFB and NAFLD was observed in participants who were white, aged 45 years or older, with natural menopause, never used female hormone, with older age at menarche, and with a fertile lifespan of 28-35 years, respectively (Table 3). Furthermore, there was no statistically significant interaction for the majority of subgroups (p-interaction > 0.05), whereas there was a significant interaction for current age (p-interaction = 0.031).
We conducted five sets of sensitivity analyses and present the results in Table 4. The overall ORs were not significantly affected by the exclusion of people who had had a hysterectomy, both ovaries removed, age at menarche < 9 years or > 18 years, age at menopause < 40 years or > 55 years or premenopausal, or had given birth more than five times, indicating the robustness of the results in our research.
In addition, FLI and NFS reflected the severity of liver injury. Similar to the primary analysis, the risk of elevated FLI and NFS was significantly higher for groups with younger AFB (Supplemental Table 2). We also evaluated the association of AFB with MAFLD, a newly defined metabolic dysfunction-associated fatty liver disease, and the specific complications of NAFLD, including overweight, obesity, diabetes, hypertension, and dyslipidemia. Results were similar to those of the primary analysis network and details are provided in Supplemental Tables 3 and 4.

Discussion
In this large population-based study, we found that younger AFB (≤ 26 years) was associated with a higher risk of NAFLD in later life of women. An increasing trend of NAFLD risk was observed for higher AFB (> 32) compared with AFB of 30-32 years, but this did not reach statistical significance. Therefore, we can expect U-shaped associations between AFB and NAFLD, as shown by the spline.
Those who were white, aged 45 years or older, with natural menopause, never used female hormone, with older age at menarche, or with a fertile lifespan of 28-35 years had a significantly stronger association between AFB and NAFLD than the other respective comparison groups. These results add to the accumulating evidence of age and menopause status differences in NAFLD. That is possibly due to the dramatic drop in estrogen levels with increasing age and menopause, and estrogen has the potential to inhibit the activation of hepatic stellate cells [18,19].
The risk of NAFLD associated with AFB was higher in white women than in black, suggesting a modified effect of genetic background. In consistence with the meta-analysis, notable racial disparities in NAFLD prevalence exist worldwide [1]. Nonetheless, this finding should be interpreted with caution. Moreover, we also found similar results between AFB and NAFLD-related complications with the primary analyses, demonstrating the robustness of the synthetic strategy.
Numerous early studies have focused on AFB in women's health. Most shreds of evidence support that early AFB is related to an elevated risk of metabolic diseases but decreased risk of liver cancer in women [7][8][9][10][11][12]. Hepatitis is most closely related to both metabolic diseases and liver cancer. However, evidence of the relationship between AFB and hepatitis is scarce and inconsistent. Most relevant to this study is the one which suggested no significant association between AFB and hepatic steatosis among 331 women in Michigan, but there was limited statistical power due to the small sample size [20]. Besides that, there was another one conducted by Wang et al. [21] Regrettably, they compared the risk of NAFLD between women who had given birth and nulliparous women only, and did not illustrate the differences in NAFLD risk between the different AFB categories. NAFLD is not only the hepatic component of metabolic syndrome but also an emergent risk factor for liver cancer. Understanding the potential association between AFB and NAFLD is critical for the prevention and management of NAFLD, especially for older or postmenopausal women [22]. The potential mechanisms are not fully understood due to the complex physiological changes during pregnancy, but several factors may account for these responses. First, adolescent pregnancy was independently associated with an increased risk of obesity, hypertension, and elevated serum lipids [8], which are strongly related to NAFLD. Then, pregnancy and childbirth are marked by increasing adipose tissue and lipolysis, insulin resistance, and inflammation [23], which may persist after giving birth, younger mothers tend to have a higher risk of NAFLD owing to a longer exposure period. Finally, there are substantial interactions between metabolism and reproduction. From an evolutionary standpoint, specific metabolic changes promote fetal growth and development at the expense of maternal benefits [24]. For example, there is temporary remodeling in the cardiovascular system, such as increased left ventricular wall thickness and mass, myocardial angiogenesis, vascular distensibility, and increased aortic stiffness [24,25]. Adolescent mothers can adapt to cardiac remodeling better than those women who give first birth at older ages, because the adolescent mothers are ongoing growth and development for a considerable period after delivery. Thus, we speculate that they are driven to more far-reaching changes in myocardial and vessels eventually, which may irreversibly influence NAFLD risk in later life. Table 2 Survey weighted odds ratios (95% CI) for the association between age at first birth and the presence of NAFLD Model 0, without adjustment; Model 1, adjusted for age (years), race (Mexican American, other Hispanic, non-Hispanic white, non-Hispanic black, other race), education level (less than 12th grade, high school or equivalent, college graduate or above), and family poverty income ratio; Model 2, adjusted for model 1 + smoking status (never, ever, current), alcohol intake (drinks/week), leisure activity (MET-min/day), total energy intake (kcal/d), and HEI-2015; Model 3, adjusted for model 2 + parity (times), menopause status (yes/no), age at menarche (≤ 11, 12-13, ≥ 14 years), fertile lifespan (< 28, 28-35, > 35 years), use of female hormones (never/ever), oral contraceptive use (never/ever), hysterectomy (yes/no), and both ovaries removed (yes/no)

Fig. 2 Nonlinear association between age at first birth and NAFLD.
Pooled dose-response association between age at first birth and odds of NAFLD. Age at first birth was modeled with restricted cubic splines in a random-effects dose-response model (gray line). Solid lines represent the odds ratio, dotted lines represent the 95% confidence interval for the spline model. The value of 30 years served as a reference, and model 3 was adjusted Apart from purely physiological reasons, the findings of the present study can be primarily interpreted on socioeconomic levels. As an illustration, women who were younger at the time of first birth were more likely to have lower educational attainment and lower family PIR, which were often accompanied by riskier health behaviors, including smoking, poor physical performance, and higher alcohol consumption. And those unhealthy lifestyles in turn lead to higher odds of NAFLD in later life. Second, those with younger AFB were also more likely to have had unintended pregnancies, which were associated with unhealthy behaviors or delays in getting health care during the pregnancy.
The prevalence of NAFLD is rapidly increasing paralleled by the rapid increment of obesity, diabetes, and other metabolic diseases [26,27]. With the increased prevalence, management practices are focused on screening and prevention. Therefore, the identification of individuals at high risk of developing NAFLD becomes particularly important. However, there is no satisfying screening strategy for targeting people at high risk of NAFLD. We found that women with AFB ≤ 26 years had higher odds of NAFLD, especially when they were older than 45 years or postmenopausal. These findings have important public health policy implications to prevent NAFLD development and reduce the disease burden. Furthermore, our results are ethnically diverse, which provide clues to suggest that different policies may be required for populations with different ethnic/racial backgrounds.
There are several strengths to this study. Firstly, this is one of the rare studies to examine the relationship between AFB and NAFLD. This finding was robust to different assessments of NAFLD risk factors and related comorbidities. Secondly, the present analysis was based on the NHANES database, a population-based design with a large sample size, ethnically diverse, and well-characterized study population, including detailed information on reproductive/hormone-related factors and other important covariates. The large sample size was powerful enough to support the stratified analysis of different ages, menopause status, and races.
However, one needs to be alert to some limitations when interpreting this study. A large number of self-reported data included in this study, such as reproductive health, physical activity, and history of liver disease, may contain sources of recall bias and reduced reliability of the results. Then, NAFLD status was inferred based upon a previously validated noninvasive diagnostic index, but it is not a perfect proxy for NAFLD diagnosis based upon liver biopsy or other imaging examination methods. Data on lifestyle earlier in life, like BMI around pregnancy, alcohol drinking, and smoking status, which may affect the NAFLD risk later, were not available. Third, cause-and-effect relationships were not elucidated due to the cross-sectional design. Finally, because the data were from multiple crosssectional studies, the design of the questionnaires was not uniform across the cycles; therefore, non-measurable bias would be included during the data merge. Thus, more work will be required to validate these findings.
Collectively, our findings support that earlier AFB relates to unfavorable outcome of NAFLD in later life of women. Policymakers should consider focusing on women with earlier AFB for screening and prevention of NAFLD.