Data and participants
Figure 1 demonstrates the inclusion criteria and workflow for this study. There were a total of 29,400 individuals who participated in NHANES from the 2013–2014, 2015–2016, and 2017–2018 annual cycles. NHANES was a continuous cross-sectional survey designed to assess the health and nutritional status of adults and children in the United States over time. Each year approximately 5,000 people were surveyed as a nationally representative sample group. At least 10% of each interviewer’s work was randomly selected and confirmed by field supervisors by telephone or site visit shortly after data collection was completed. Data were collected from the three-year cycles, which included information on infertility in female patients. Sample size was estimated based on the Event Per Variable (EPV) criterion [28]. In particular, an EPV ≥ 10 was used to determine the required minimum sample size and the maximum number of candidate predictors that could be investigated [29]. According to the purpose of the study, 14,452 men, 5,630 minor women, and 3,887 adult women without complete diagnostic information on infertility were excluded. Ultimately, the study sample consisted of 5,431 older women for whom diagnostic infertility data was available. Among them, 596 women were diagnosed as infertile.
Infertility data
Infertility status was the outcome variable of interest. All female participants aged 18 years or older were surveyed on the reproductive health questionnaire using computer-assisted personal interviews with trained interviewers. They were asked if they had “ever tried to get pregnant for at least a year without getting pregnant” and “ever seen a doctor or other health care provider because they couldn’t get pregnant.” Women were classified as having self-reported infertility if they answered yes to the first question. Likewise, if they answered yes to the second question, they were classified as having sought reproductive health care for infertility. These two types of participants became the main objects of our research.
Body Shape Index (ABSI)
ABSI, the exposure variable, has recently been recommended as a parameter that indicates body shape using waist circumference (WC), weight and height [16]. Data on height (cm), weight (kg) and height (cm) of the included individuals were obtained from the 2013–2018 NHANES. WC was measured by placing a flexible tape around the uppermost lateral border of the ilium in the upright position. Weight and height were measured to the first decimal place. Body mass index (BMI) is defined as weight (kg) divided by the square of height (m). The ABSI value is calculated as a continuous variable using the following equation:
$$text{ABSI = }frac{text{WC}}{{text{BMI}}^{text{2/3}}text{-}{text{height}}^{ text{1/2}}}$$
Patient Health Depression Questionnaire-9 (PHQ-9) score.
Depressive symptoms in the past two weeks were assessed by the PHQ-9 during face-to-face interviews with the MEC for all participants. The PHQ-9 has been widely developed and used as a screening for depression in adults in the accompanying editorial [30]. In addition to screening for depression, the PHQ-9 has been demonstrated as a continuous measurement and diagnostic method of depressive symptom severity. Respondents indicated on a scale of 0 to 3 the frequency with which they experienced the following symptoms: (1) anhedonia, (2) depressed mood, (3) sleep disturbance, (4) fatigue, (5) changes in appetite, (6) ) low self-esteem, (7) concentration problems, (8) psychomotor disturbances, and (9) suicidal thoughts. Total scores range from 0 to 27, with scores ≥ 10 indicating clinically significant depressive symptoms [23]. Subsequently, the severity of depressive symptoms was divided into 5 levels according to the total score: 0–4, minimal depression; 5–9, mild depression; 10–14, moderate depression; 15–19, moderately severe depression; and 20-27, major depression [23].
Covariates
We examined several potential covariates derived from computer-assisted personal interviews or medical examinations by highly trained medical personnel. Covariates for the 5431 participants in this study consisted of sociodemographic variables, lifestyle behaviors, and systemic chronic diseases. Sociodemographic variables included age (continuous variable), race/ethnicity (non-Hispanic white, non-Hispanic black, and other), family income (0-1.30, 1.31–3.50, and ≥ 3.51), and education (more than low from high school, high school or equivalent and college or higher). Recreational activity included none, moderate, and vigorous levels. BMI was used as a measure of obesity according to WHO guidelines: (1) normal weight (18.4 kg/m2< BMI ≤ 24.9 kg/m2); (2) underweight (BMI ≤ 18.4 kg/m2); (3) overweight (24.9 kg/m2< BMI ≤ 30.0 kg/m2); (4) and obesity (BMI > 30.0 kg/m2). Smoking status (never smoker, former smoker, and current smoker) and hyperlipidemia and hypertension were also obtained from the interviews.
Statistical analysis
To accommodate the complex design of the NHANES survey, we incorporated sampling weights, clustering, and stratification to generate nationally representative population estimates. Overall participant characteristics were shown as weighted mean (standard error (SE)) of continuous variables and number (weighted percentage (%)) of categorical variables. Weight variables were used for statistical analysis to increase representativeness of the population. Distributional differences in sociodemographic and behavioral lifestyle characteristics were first examined by infertility status using an independent t-test or chi-square test to compare sets of continuous variables and categorical variables, respectively.
Study-weighted multivariable logistic regression models were constructed to examine the associations of infertility (outcome) with ABSI and other individual components (exposure) by odds ratios (OR) and corresponding 95% confidence intervals (CI). The univariable model was not adjusted for confounders, while the multivariable model was adjusted for age, race/ethnicity, education, and family income.
A mediation analysis was performed to examine the mediating role of PHQ-9 score on ABSI-infertility in the NHANES population using the R package (version 4.5.0) ‘mediation’. As shown in fig. 2, ABSI was defined as exposure, PHQ-9 depression score as mediator, and infertility as outcome. In the mediation analysis, three steps should be performed, including the total effect (TE), direct effect (DE), and indirect effect (IE). TE refers to the sum of DE and IE. Typically, the goal of mediation analysis (Fig. 2) is to identify the TE of the exposure (eg, ABSI) on the outcome (eg, infertility), the IE of the exposure that acts through a given set of mediators (eg, PHQ-9 score) of interest and DE of exposure unexplained by these same mediators [31]. In our study, the indirect effect of ABSI on infertility refers to the effect that ABSI acts through the PHQ-9 score on infertility, which can necessarily help us distinguish the pathways that link ABSI to infertility. Finally, we calculated the proportion mediated by PHQ-9 score using the following formula [32]:
$$text{Mediated }text{proportion}text{ of PHQ-9 score}text{ = }frac{text{total effect-direct effect}}{text{total effect}}times 100 %$$
Statistical significance was established two-sided p< 0.05. All analyzes in the present study were conducted in R software (version 4.2.0) and relevant packages.