ABSTRACT 

Background: Spermidine administration is linked to increased survival in several animal models. Objective: The aim of this study was to test the potential association between spermidine content in diet and mortality in humans. Design: This prospective community-based cohort study included 829 participants aged 45–84 y, 49.9% of whom were male. Diet was assessed by repeated dietitian-administered validated foodfrequency questionnaires (2540 assessments) in 1995, 2000, 2005, and 2010. During follow-up between 1995 and 2015, 341 deaths occurred. Results: All-cause mortality (deaths per 1000 person-years) decreased across thirds of increasing spermidine intake from 40.5 (95% CI: 36.1, 44.7) to 23.7 (95% CI: 20.0, 27.0) and 15.1 (95% CI: 12.6, 17.8), corresponding to an age-, sex- and caloric intake–adjusted 20-y cumulative mortality incidence of 0.48 (95% CI: 0.45, 0.51), 0.41 (95% CI: 0.38, 0.45), and 0.38 (95% CI: 0.34, 0.41), respectively. The age-, sex- and caloric ratio–adjusted HR for all-cause death per 1-SD higher spermidine intake was 0.74 (95% CI: 0.66, 0.83; P < 0.001). Further adjustment for lifestyle factors, established predictors of mortality, and other dietary features yielded an HR of 0.76 (95% CI: 0.67, 0.86; P < 0.001). The association was consistent in subgroups, robust against unmeasured confounding, and independently validated in the Salzburg Atherosclerosis Prevention Program in Subjects at High Individual Risk (SAPHIR) Study (age-, sex-, and caloric ratio– adjusted HR per 1-SD higher spermidine intake: 0.71; 95% CI: 0.53, 0.95; P = 0.019). The difference in mortality risk between the top and bottom third of spermidine intakes was similar to that associated with a 5.7-y (95% CI: 3.6, 8.1 y) younger age.

INTRODUCTION 

Spermidine is the most abundant polyamine in a majority of different human tissues, with intracellular spermidine concentrations declining during the natural course of organismal aging (1–3). Conversely, the administration of spermidine is linked to increased survival of yeast, worms, flies, and human immune cells and reduces age-associated mortality in mice (Supplemental This study is part of the excellence initiative [Competence Centers for Excellent Technologies (COMET)] of the Austrian Research Promotion Agency FFG: “Research Center of Excellence in Vascular Ageing—Tyrol, VASCage” (K-Project 843536) funded by the Federal Ministry for Transport, Innovation and Technology (BMVIT), Federal Ministry of Science, Research and Economy (BMWFW), Wirtschaftsagentur Wien, and Standortagentur Tirol. PW is the recipient of an Erwin Schrödinger Fellowship (Austrian Science Fund, J3679-B13). GK is supported by Cancéropôle Ile-de-France, Institut National du Cancer (INCa), the European Research Council (ERC), the LabEx Immuno-Oncology, and the Paris Alliance of Cancer Research Institutes (PACRI). FM has received support from to the Austrian Science Fund (grants P23490-B20, P29262, P24381, P 29203, P 27893, and “SFB Lipotox”) as well as from BMWFW (grants “Unkonventionelle Forschung” and “Flysleep”) and for the BioTechMed-Graz flagship project EPIAge. TE is the recipient of an Austrian Programme for advanced Research and Technology (APART) fellowship of the Austrian Academy of Sciences.

Study participants 

The Bruneck Study is a prospective, population-based cohort study. Its study population was enrolled in 1990 as an ageand sex-stratified random sample of all inhabitants of Bruneck (Bolzano Province, Italy) aged 40–79 y (125 women and 125 men in each of the fifth to eighth decades of life; n = 1000) and re-evaluations were scheduled every 5 y since 1990 (20– 24). The population is exclusively white and is unique for its participation and in-person follow-up rates >90% (20–24), facilitated by annual population mobility proportions as low as 0.2%. Moreover, for all participants, full medical records from general practitioners and Bruneck Hospital, the only hospital in the region, were available for review. For this analysis, we defined the baseline as the year of the first detailed dietary assessment (1995) involving 829 women and men aged 45–84 y with a follow-up of 20 y (1995–2015;). The study protocol conformed to the Declaration of Helsinki and was approved by the local ethics committees (Bolzano and Verona). Participants gave their written informed consent and did not receive financial compensation. Participant characteristics were assessed by standard procedures (20–24) detailed in the online supporting material .The Salzburg Atherosclerosis Prevention Program in Subjects at High Individual Risk (SAPHIR) Study served as a prospective replication cohort. A total of 1770 healthy unrelated participants (663 women and 1107 men aged 39–67 y) were recruited by health-screening programs in large companies in and around the city of Salzburg. The cohort was examined in the years 1999– 2002, with follow-up for deaths until September 2013 (median follow-up: 12.8 y) (25, 26). The current trial was registered at www.clinicaltrials.gov as NCT03378843.

Dietary assessment 

Dietary intake was evaluated by quinquennial (1995, 2000, 2005, and 2010) dietitian-administered 118-item food-frequency questionnaires (FFQs) on the basis of the gold-standard FFQ by Willett et al. (27) and adapted to the dietary habits in the survey area (for details and validation, see Supplemental Tables 2–4). FFQs were available from 829 individuals who provided 2540 individual assessments. For each item in the FFQ, a common unit or portion size was specified, and we instructed participants to customize how often, on average, they had consumed that amount in the past years. The 9 response categories ranged from “never” to “≥6 times/d.” We calculated nutritional intake by assigning a weight proportional to the frequency of use for each food (1 time/d equals a weight of 1), multiplying this weight by the nutrient value for the specified size, and summing the contribution of all foods. Nutrient-composition data for foods were based on the USDA Nutrient Database (release 23) (28). We compiled a special nutrient database for polyamines (Supplemental Table 4). Dietitians made use of illustrative photos of foods when exploring aphasic patients and of information provided by spouses, caregivers, and nursing homes. We dissected complex foods into component foods utilizing common recipes. An open-ended section inquired about foods not considered in the FFQ that were consumed ≥1 time/wk and various kinds of nutritional supplements.

The reproducibility and validity of the original FFQ are well documented (27) and extend to its application in the Bruneck Study. A comparison with 9-d diet records is depicted in Supplemental Table 2 and shows a reasonable level of agreement. As expected, dietary patterns remained stable over time (Supplemental Table 3). Accordingly, under most circumstances, point estimates adequately reflect medium-term dietary intake. The average correlation of spermidine intake in the 4 quinquennial assessments was 0.54. Dietary assessment in the SAPHIR study relied on a 69-item FFQ and the Nutrient Database for Germany but otherwise used the same methodologic standards.

Ascertainment of causes of death 

In the Bruneck Study, we collected detailed information on the date, causes, and circumstances of death for all study participants who did not survive the entire follow-up period by consulting death certificates, all medical records ever compiled on study participants, and autopsy reports in the rare event of unexpected death. We were able to ascertain 100% of deaths and reliably classify them as vascular deaths, cancer deaths, or deaths from other causes. Classification details are summarized in Supplemental Table 5. The experienced researcher who categorized all deaths was unaware of the dietary data. Our focus was on the primary cause of death. In SAPHIR, ascertainment of death relied on the review of hospital records and information from the Austrian Death Registry (Statistics Austria). Follow-up was complete except for a presumably low proportion of expatriates.

Statistical analysis 

We estimated the conditional 20-y cumulative incidence of all-cause death in thirds of spermidine intake with adjustment for age, sex, and caloric intake with the use of parametric survival analysis with inverse probability weighting and present parametric survival curves based on flexible Parmar-Royston spline-based models and nonparametric Kaplan-Meier curves (31, 32). Cox proportional hazards models with time-varying covariates were used to estimate HRs and 95% CIs for the association between spermidine intake and death. No departure from the proportional hazards assumption was detected when inspecting Schoenfeld residuals and checking the parallelity of log-log survival plots. Our primary endpoint was death from any cause. In analyses focusing on cause-specific deaths, participant data were censored if the participant died from other causes. This approach produces HRs for each cause of death that are etiologically interpretable (33, 34). In addition to these causespecific hazard models, Fine-Gray subdistribution hazard models were fitted (35). Both approaches account for competing risks (36). We modeled spermidine intake as a categorical or continuous variable (after loge-transformation) and used restricted cubic splines to detect potential nonlinearities. All of the analyses were adjusted for age, sex, and “caloric ratio.” Multivariable analyses additionally included well-established risk factors and determinants of death such as socioeconomic status, BMI, smoking, diabetes, hypertension, and aspirin medication (all with updates every 5 y) and features of lifestyle and diet such as physical activity and alcohol consumption (both with cumulative updates every 5 y).