This single-center, prospective observational study collected baseline data from January 2019 to December 2019, with follow-up currently ongoing. A total of 618 participants aged 80–98 years (81% male) were initially enrolled at the Sixth People’s Hospital Affiliated to Shanghai Jiao Tong University, Shanghai, China (clinical trial registration number: ChiCTR1800018015). Of these, 611 participants completed the Mini-Mental State Examination (MMSE) for cognitive assessment and underwent measurement of plasma reproductive hormone levels. Finally, a total of 509 individuals (440 males and 69 females) were included in this study. Participants with disorders of the hypothalamo-pituitary-gonadal axis or those receiving sex hormone replacement therapy, both of which could influence reproductive hormone levels, were excluded. The detailed inclusion and exclusion criteria have been described previously (Zhu et al., 2021; Ke et al., 2021). In addition, individuals with mild cognitive impairment were excluded to better address the specific aims of the present study. Baseline assessments included a physical examination, blood sample collection, and in-person interviews that gathered information on sociodemographic characteristics, lifestyle factors, cognitive function, and medical history. The study was approved by the Ethics Committee of Shanghai Jiao Tong University Affiliated Sixth People’s Hospital (Reference number: 2018–109) and was conducted in accordance with the principles of the Declaration of Helsinki. Written informed consent was obtained from all participants or their legal guardians.

Following an overnight fast, peripheral venous blood samples were collected from each participant. Serum and plasma were separated and stored for analysis. The measurement of serum folate (reference range: 4.20–19.80 μg/L) and vitamin B12 (reference range: 197–771 ng/L) levels were conducted using the Roche Elecsys Immunoassay on the Cobas e601 analyzer. The levels of free triiodothyronine [FT3; detection range (DR): 0.4–50.0 pM; reference range: 3.1–6.8 pM], free thyroxine (FT4; DR: 0.3–100.0 pM; reference range: 12.0–22.0 pM), and thyroid-stimulating hormone (TSH; DR: 0.06–99.0 mIU/L; reference range: 0.27–4.20 mIU/L) were determined using electrochemiluminescence immunoassays (Roche Diagnostics GmbH). Sex hormones, including total testosterone (TT; DR: 0.087–52.000 nM; reference range: 6.68–25.70 nM) and E₂ (DR: 18.35–15,781.00 pM; reference range: 41.4–159.0 pM), as well as pituitary gonadotropins—FSH (DR: 0.5–160.0 IU/L; reference range: 1.5–12.4 IU/L) and LH (DR: 0.1–200.0 IU/L; reference range: 1.7–8.6 IU/L) were measured using chemiluminescence assays (Abbott GmbH & Co. KG). All assays demonstrated inter-assay and intra-assay coefficients of variation below 10%. Four samples with values below the detection limits for LH, E₂, or TT were re-analyzed using the respective lower detection limits. APOE genotyping was conducted using polymerase chain reaction followed by direct sequencing. Participants were categorized based on APOE ε4 status (carriers vs. non-carriers).

Global cognitive function was assessed using the MMSE, a widely used tool consisting of multiple simple tasks, with a maximum score of 30 points. All-cause dementia (ACD) was diagnosed complying with the 2011 criteria established by the National Institute on Aging–Alzheimer’s Association (NIA-AA) workgroups (Mckhann et al., 2011). AD-D was further identified based on NIA-AA criteria and confirmed through cranial magnetic resonance imaging (cMRI). VD was diagnosed if cognitive impairment occurred in association with stroke or cerebrovascular lesions, provided dementia developed within 3–6 months post-event and persisted for at least 3 months.

Data on demographic characteristics (age, sex, and years of education) and comorbidities known to be strongly associated with cognitive function, such as diabetes, hypertension, and depression, were collected through self-report and verified by medical records. Lifestyle risk factors, including smoking and alcohol consumption, were also recorded. Due to the relatively low number of current or former smokers and drinkers, only the prevalence of non-smoking and non-drinking was reported. Waist circumference (WC) was measured at the midpoint between the lower rib margin and the iliac crest along the midaxillary line, with participants standing in a relaxed posture while wearing underwear.

The Shapiro-Wilk test was used to assess the normality of variable distributions. Normally distributed variables were presented as mean ± standard deviation and compared between groups using the Student’s t-test, whereas skewed variables were reported as median (25th–75th percentile) and compared between groups using Mann-Whitney U-test. Additionally, categorical variables were presented with counts and percentages, and compared using chi-square test. Linear regression models were used to examine the association between plasma gonadotropins, either as a continuous variable or as a dichotomous variable (based on a median split), and MMSE scores. Binary logistic regression models were adopted to explore the link between gonadotropins and the risk of dementia, including ACD, AD-D, and VD. Crude and adjusted odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. The linear and logistic regression models were adjusted for covariates that may influence gonadotropins and cognitive function (age, years of education, history of hypertension or depression, WC, TT, E₂, and FT3), and additionally for LH or FSH, respectively, to clarify whether the associations remained independent of the other gonadotropin. Two-way analysis of variance (ANOVA) was performed to assess the interaction between gonadotropins and APOE ε4 status on cognitive performance, adjusting for age and education. Receiver operating characteristic (ROC) curve analysis was performed to determine optimal cut-off value of FSH for diagnosing AD-D in males, and the area under the ROC curve (AUROC) with 95% CI was calculated. The optimal cutoff point was identified using the maximum Youden Index. All statistical analyses were performed using SPSS for Windows, version 23.0, with figures generated using GraphPad Prism version 8.0. Statistical significance was defined as a two-tailed p-value less than 0.05.

A total of 172 elderly participants (33.8%) were classified as having ACD, of whom 97 (56.4%) were diagnosed with AD-D and 75 (43.6%) with VD. MMSE scores were significantly lower in both the AD-D and VD groups compared to the normal cognition (NC) group (both p < 0.01). Table 1 shows baseline characteristics of the different groups. Compared to participants with NC, individuals with AD-D or VD were older, had fewer years of education, higher prevalence of depression, and lower FT3 levels (all p < 0.01). In addition, the prevalence of hypertension was significantly lower in the AD-D group in contrast with the NC group (p < 0.01). No significant differences were observed among the groups regarding sex, proportion of non-smokers and non-drinkers, APOE ε4 carrier status, prevalence of diabetes, WC, or levels of folate, vitamin B12, FT4, and TSH.

The levels of reproductive hormones across different groups are summarized in Supplementary Table 1 and illustrated in Figure 1. FSH and LH levels were significantly higher in the VD group compared to the NC group in both sexes (all p < 0.05). Among males, FSH levels were also significantly higher in the AD-D group in contrast with the NC group (p < 0.05). In females, FSH levels were higher in the VD group when compared to the AD-D group (p < 0.05). TT and E₂ levels were not different with statistical significance among the groups.

Given the observed differences in FSH and LH levels among groups with varying cognitive status, the association between gonadotropins and MMSE scores was further analyzed. After adjusting for age, years of education, history of hypertension and depression, WC, FT3, E₂, TT and APOE ε4 genotype, no significant association was found between continuous FSH levels and MMSE scores in the linear regression model. Given the skewed distribution of FSH levels, we further dichotomized it into “high” and “low” groups based on sex-specific median values (male: 19.9 IU/L; female: 67.1 IU/L); however, no significant association was observed in the multiple linear regression model. Notably, a significant association emerged after additional adjustment for LH (the total population: β = −2.126, p = 0.030; males: β = −2.719, p = 0.015). In contrast, continuous LH levels were positively associated with MMSE scores in the total population and in males regardless of FSH adjustment, whereas dichotomized LH (male: 10.1 IU/L; female: 27.4 IU/L) showed no significant association (Table 2). Collectively, these findings suggest a threshold effect of FSH and a linear inverse association of LH with cognitive function.

Considering the additive effect of FSH and APOE ε4 in the development of AD-like pathogenesis in female mice (Xiong et al., 2023), we investigated whether gonadotropins and APOE ε4 interact in cognition impairment. After adjusting for age and education, the MMSE scores were significantly influenced by the interaction between FSH and APOE ε4 in both the total cohort and male participants (Figure 2 and Table 3). Among APOE ε4 carriers, participants in the high-FSH subgroup had significantly lower MMSE scores than those in the low-FSH subgroup (p < 0.01). In male APOE ε4 carriers, a similar decreasing trend was observed (p = 0.053) (Table 3). We also analyzed the interaction between LH and APOE genotype on MMSE, and found no significant effect. Interestingly, in APOE ε4 non-carriers, the high-LH group exhibited lower MMSE scores than the low-LH group, both in the total population and in males (p < 0.01). Due to the small sample size and low number of APOE ε4 carriers among females, the results are unreliable and the power is low. Therefore, the analysis was not conducted in females.

The relationship between gonadotropins levels and dementia is presented in Table 4. The proportion of subjects with ACD, AD-D, or VD was higher in the high-FSH group in contrast with the low-FSH group. Binary logistic regression analysis revealed that neither continuous FSH nor dichotomized FSH was associated with the risk of AD-D after adjusting for age, years of education, history of hypertension and depression, WC, FT3, E₂, TT, and APOE ε4 genotype. However, after further adjusting for LH, both continuous FSH and dichotomized FSH were significantly associated with an increased risk of AD-D in the total population (continuous FSH, OR: 1.033, 95% CI: 1.004–1.063; dichotomized FSH, OR: 3.109; 95% CI: 1.276–7.576) and males (continuous FSH, OR: 1.048, 95% CI: 1.011–1.086; dichotomized FSH, OR: 3.597; 95% CI: 1.159–11.166). In the overall cohort, dichotomized FSH was also associated with an increased risk of ACD (OR: 2.092, 95% CI: 1.099–3.983). To assess the role of FSH in predicting the risk of AD-D in older males, ROC curve analysis was performed. As shown in Figure 3, FSH demonstrated an AUROC of 0.600 (95% CI: 0.528–0.671; p < 0.01), with an optimal cutoff value of 28.4 IU/L, achieving a sensitivity of 68.5% and a specificity of 56.6% for AD-D. For LH, continuous but not dichotomized LH was significantly associated with a decreased risk of AD-D in the total and males populations regardless of adjusting for FSH. No significant association was observed between gonadotropins and VD. Due to the limited number of female participants and dementia cases across subgroups, the regression models were unstable and the results unreliable; therefore, corresponding analyses were not conducted in females.