Dementia is the leading cause of disability among people aged over 65 years worldwide. From 1990 to 2016, the global number of dementia patients more than doubled, from 20.2 to 43.8 million (GBD 2016 Dementia Collaborators, 2019). People with dementia in China account for approximately 25% of the dementia patients (Jia et al., 2020) worldwide. This scenario has imposed a heavy socioeconomic burden, impacting their quality of life. Although China has established and implemented a series of programs to manage this disease in the past decade, patients in rural areas often have difficulty accessing health services compared with patients in urban areas (Xue et al., 2018). Meanwhile, no specific therapy is available for the treatment of dementia. Therefore, early identification of high-risk populations and improvement of prognostic factors are critical, especially for people living in rural areas.

Mild cognitive impairment (MCI), the stage between normal cognition (NC) and dementia, is associated with a higher risk of dementia. Previous studies have shown that approximately 16% of patients diagnosed with MCI return to normal or near-normal cognition after a year of treatment (Koepsell and Monsell, 2012). Thus, identifying more sensitive or specific biomarkers for MCI is essential to reduce the burden of dementia.

Spermidine (SPD) is a natural polyamine present in all living organisms, and play a key role in maintaining cellular homeostasis. This chemical contributes to various critical cellular functions, including cell growth and proliferation, tissue regeneration, DNA and RNA stabilization, enzymatic modulation, and regulation of translation (Madeo et al., 2019). The association between SPD and cognition has been explored using animal experiments. Gupta et al. (2013) demonstrated that fruit flies fed SPD showed enhanced autophagy and that SPD can effectively inhibit aging-associated memory impairment via autophagy. A recent study revealed that dietary SPD passes the blood–brain barrier and improves spatial and temporal memory in aged mice (Schroeder et al., 2021).

Several small randomized controlled clinical trials have been implemented to examine the relationship between dietary SPD supplements and cognition in humans; however, the conclusions have been inconsistent (Wirth et al., 2018; Pekar et al., 2021). Additionally, a prospective cohort study involving 815 participants found that higher nutritional SPD intake is associated with a lower risk for cognitive impairment and decline (Schroeder et al., 2021). However, self-reported dietary information obtained from food frequency questionnaires is subject to measurement error that could be attributed to recall bias and social desirability. Dietary SPD intake may not entirely reflect the actual SPD concentration in the body. The present study examines SPD levels in serum to objectively reflect the concentration of SPD in the body and provide a more objective basis for subsequent promotion and application.

Therefore, we explore the association between serum SPD levels and MCI based on the mutual verification of cross-sectional and longitudinal studies conducted on a Chinese population.

Table 1 shows the baseline characteristics of the 3,291 participants stratified by the SPD tertile. In total, the mean (SD) age was 57.4 (9.8) years. Sex, alcohol consumption, smoking, SBP, DBP, and BMI at baseline significantly differed among the three groups. Figure 2 shows the prevalence (95% CI) of MCI at baseline. The prevalence (95% CIs) of MCI for participants in the first, second, and third tertile groups was 40.7% (37.8, 43.6%), 35.5% (32.6, 38.8%), and 37.4% (34.5, 40.2%), respectively.

In cross-sectional analyses, high concentrations of serum SPD were associated with low MCI prevalence (Table 2). After adjusting for age, sex, ethnicity, education levels, smoking, drinking, BMI, history of hypertension, diabetes, and CHD, compared with the first tertile, the multivariate-adjusted ORs (95% CIs) for the second and third tertiles were 0.78 (0.65, 0.93) and 0.80 (0.67, 0.96), respectively.

In the subgroup analysis, we only observed an relationship between SPD and MCI among males and the subgrouping of participants into under 60 and over 60 years. After adjusting for age, ethnicity, education levels, smoking, drinking, BMI, history of hypertension, diabetes, and CHD in males, compared with the first tertile, the multivariate-adjusted ORs (95% CIs) for the second and third tertiles were 0.56 (0.41, 0.76) and 0.64 (0.47, 0.87), respectively. In people aged <60, after adjusting for confounding factors, compared with the first tertile, the multivariate-adjusted ORs (95% CIs) for the second and third tertiles were 0.86 (0.67, 1.12) and 0.93 (0.72, 1.20), respectively. In people aged >60, after adjusting for confounding factors, compared with the first tertile, the multivariate-adjusted ORs (95% CIs) for the second and third tertiles were 0.76 (0.58, 0.98) and 0.76 (0.59, 0.98), respectively.

Restricted cubic spline models were used to evaluate the relationship between SPD and MCI (Figure 3). In multi-adjusted models, the relationships between SPD and MCI were non-linear (P for non-linear <0.001).

The ROC curves for the risk of MCI prevalence from multivariable logistic regression models, with and without SPD, were analyzed. The end points were characterized by MCI; therefore, the first model to incorporate age, sex, ethnicity, education level, smoking, drinking, BMI, history of hypertension, diabetes, and CHD had an AUC of 0.666 (95% CI: 0.648–0.685). Ln (SPD) was supplemented to Model 2 with an AUC of 0.668 (95% CI: 0.649–0.687). The difference in AUC was statistically insignificant, suggesting that the discriminative abilities of the two models were similar. Adding Ln (SPD) improved risk reclassification for MCI events, measured by continuous NRI (0.079; 95% CI, 0.009–0.149; P = 0.016) and by IDI (0.002; 95% CI, 0.000–0.003; P = 0.028). The NRI and IDI results showed that the power of the reclassification of the new model was significantly improved.

Additionally, we explored the relationship between SPD and different cognitive domains (Supplementary Table 2). In total, compared with the first tertile, the third tertile group had a decreased risk of fluency, orientation, calculation, abstract, and visual perception dysfunction, with multivariate-adjusted ORs (95% CIs) of 0.78 (0.63, 0.97), 0.78 (0.65, 0.94), 0.80 (0.67, 0.96), 0.79 (0.65, 0.96), and 0.76 (0.59, 0.96), respectively.

To further validate the above results, we analyzed the relationship between serum SPD levels and MCI prevalence using a longitudinal study (Table 3). After adjusting for age, sex, ethnicity, education levels, smoking, drinking, BMI, history of hypertension, CHD, and diabetes, and baseline MoCA-BC score, compared with the first tertile, ORs (95% CIs) for the second and third tertiles were 0.76 (0.49, 1.18) and 0.62 (0.39, 0.99), respectively.

We explored the role of serum SPD levels in MCI among 3,774 participants based on a large-scale epidemiological survey. We found a non-linear correlation between serum SPD levels and prevalent MCI events, which also showed potential of SPD to be a clinically useful indicator of MCI by improving risk reclassification. Similarly, in the longitudinal study, higher baseline SPD levels may indicate a lower risk of MCI events. To the best of our knowledge, this is the first study revealing the association of serum SPD levels with MCI in a large-scale population.

The relationship between SPD and cognition has been proven in animal experiments (Gupta et al., 2013; Hofer et al., 2021; Schroeder et al., 2021). Hofer et al. (2021) found that SPD is involved in preserving mitochondria and cognitive function. In aged fruit flies, simple SPD feeding suppressed age-induced memory impairment (Gupta et al., 2013). Most previous population studies on SPD and cognition examine dietary SPD (Wirth et al., 2018; Schwarz et al., 2020; Schroeder et al., 2021). A study involving 56 participants having cognitive declines and 47 healthy control participants showed that higher SPD intake might be a promising dietary approach to preserve brain health among older adults (Schwarz et al., 2020). The Bruneck Study shows that higher nutritional SPD intake is related to enhanced cognition in humans (Schroeder et al., 2021). A randomized controlled trial involving 30 participants with SCD showed that nutritional SPD was associated with a positive effect on memory performance (Wirth et al., 2018). Contrastingly, in another 12-month randomized clinical trial, long-term SPD supplementation in participants with SCD did not modify memory performance and biomarkers compared with placebo (Wirth et al., 2019). Additionally, Joaquim et al. (2019) monitored their study participants for 4 years and found that SPD levels were lower in both patient groups Alzheimer’s disease (AD; MCI) than in healthy controls. Graham et al. (2015) showed, through untargeted metabolomic analysis of human plasma, that polyamine metabolism was disrupted in patients with MCI.

Our results confirm that higher baseline serum SPD levels are associated with lower MCI prevalence and incidence, especially in males, and that subgrouping of participants into under 60 and over 60 years. The possible mechanism underpinning this association could be that SPD induces autophagy and exerts additional and potentially autophagy-independent metabolic and transcriptional effects in vivo (Madeo et al., 2018). However, cross-sectional analyses revealed a non-linear relationship between SPD and the risk of MCI, i.e., at very high SPD levels, the risk of MCI is elevated. We speculate that the elevated serum SPD levels in the MCI population are a compensatory effect in a pathological state. The geroprotective effects of SPD influence lifespan extension in a hormetic dose-dependent manner (Calabrese et al., 2020). In this hormesis scenario, a high dose of natural compounds is detrimental, because it inhibits antioxidant stress response pathways. The longitudinal study did not reveal a significant non-linear relationship of SPD with MCI. This finding is attributable to the small sample size (N = 888) of the longitudinal study and the high failure rate in conducting follow-up. There may be a certain loss to follow-up bias, thereby emphasizing the need for further examination using larger sample sizes.

Compared with previous studies, this study has several strengths. First, unlike previous clinical studies that included dozens of patients, we explore the relationship between serum SPD levels and MCI in a relatively large sample size (N = 3,774) of the general population for the first time. Second, there are three main sources of SPD in humans: cellular metabolism; oral absorption from dietary sources; or commensal gut bacteria (Madeo et al., 2018). Instead of assessing dietary SPD intake alone, we assessed overall SPD levels in the body to account for all SPD sources. Simultaneously, the measurement of SPD in serum can avoid the recall bias caused by the dietary assessment scale.

This study, however, also has inevitable limitations. First, we lacked an assessment of dietary SPD intake, which may affect the in vivo SPD level. Therefore, we could not establish a correspondence between serum SPD levels and dietary SPD intake. Second, we only measured the level of SPD once. Dynamically observing the changes in SPD in the blood may help to better identify its relationship with MCI over a time-course. Third, we exclusively examined the Chinese rural population, which may limit generalizability. The dietary culture of different populations differs and SPD is affected by diet (Madeo et al., 2020) and intestinal flora (Ramos-Molina et al., 2019). Therefore, the verification of the diverse population is meaningful.

The original contributions presented in this study are included in the article/Supplementary material, further inquiries can be directed to the corresponding authors.

The studies involving human participants were reviewed and approved by the procedures followed under the ethical standards of the responsible committee on human experimentation of China Medical University ([2018]083). The patients/participants provided their written informed consent to participate in this study.

LZ and ZS designed the research. JX wrote the manuscript. QZ and FX provided essential reagents or provided essential materials. RZ, RL, and ZY conducted the research. YM performed statistical analysis. LZ was responsible for formulating all the sections of the manuscript. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.