Unusual Cholesterol Levels in Very Old People from a Sardinian Longevity Area

This retrospective longitudinal study was conducted on a representative sample of individuals aged 90 years and older from the LBZ area in central Sardinia, recruited from two waves in 2018 [35] and 2023 [36]. The local ethical committee (Comitato Etico ASL no. 1 di Sassari, Italy) approved the study protocol (2101/CE). Written informed consent was obtained from all participants or their caregivers in cases of severe cognitive impairment.

Figure 1 illustrates the recruitment process. Of the 200 subjects recruited in previous studies [35,36], 13 were excluded due to the unavailability of laboratory tests, including lipid profiles. An additional 19 subjects were excluded because they were taking statins or loop diuretics that could have altered the baseline lipid profile. The exclusion criteria included severe dementia or severe comorbidity, being bedridden, and the unavailability of lipid profiles.

The inclusion criteria included being of Sardinian origin, being 90 years or older, and having four grandparents born within the Blue Zone.

All participants underwent a baseline home interview performed by trained personnel according to standard procedures [35]. Data were collected using validated questionnaires during a geriatric multidimensional evaluation (see details below). The vital status of participants was ascertained by directly asking the municipality of birth about their survival. Participants’ self-reported health was evaluated according to Idler et al. [37], and classified as excellent, very good, good, fair, or poor.

Covariates gathered from participants’ baseline questionnaires included sex, age at recruitment, smoking, physical activity, and comorbidity. Information on dietary habits, anthropometric measurements, self-reported physical health, psychological health, living conditions, health behaviors, health care utilization, social support, and previous occupation was also collected. Arterial hypertension was defined as systolic blood pressure ≥ 140 mm Hg, diastolic blood pressure ≥ 90 mm Hg, or the assumption of antihypertensive medications. Diabetes mellitus was determined by clinical history (as evaluated by a specialist) or the assumption of insulin or glucose-lowering oral medications. Anthropometric parameters were gathered using standard procedures [38]. These parameters included waist circumference, the average circumference of the calf (in correspondence with the gastrocnemius muscle), and the brachial circumference (in correspondence with the biceps muscle). Body mass index (BMI) was calculated by dividing body weight (in kilograms) by height (in meters) squared (kg/m²).

Plasma lipids were measured by standard methods. After an overnight fast, blood samples were collected by venipuncture using silica-coated (plain) tubes (Vacutainer^® Serum Tubes, Becton Dickinson Italia S.p.A, Milan, Italy). Serum was separated at 2500 rpm in a centrifuge at 4 °C within 1–2 h after collection. The separated plasma was subdivided into 500 µL aliquots and frozen at −80 °C for later determination of lipid levels. Total cholesterol, HDL-C, and TG were measured with an ADVIA^® Chemistry XPT (Siemens Healthcare Diagnostics s.r.l., Milan, Italy) according to the manufacturer’s protocol. The LDL-C was calculated according to the Friedewald formula [39], while VLDL cholesterol (VLDL-C) was calculated as the difference between total cholesterol and HDL-C + LDL-C. Patients were divided into two subgroups based on whether their baseline LDL-C levels were higher or lower than 130 mg/dL according to the National Cholesterol Education Program guidelines [40]. Moreover, additional calculated lipid parameters such as non-high-density (non-HDL) cholesterol and lipid ratios (TG/HDL, LDL/HDL) were analyzed, as they have been recognized as better risk indicators in patients with coronary artery diseases [41,42,43].

Dietary habits were evaluated according to a previously reported procedure [33,36,44]. A short qualitative food frequency questionnaire used in the past for nutritional surveys in Sardinia [45] was adapted for the present study. Briefly, the average frequency of intake of 15 essential foods (non-fish meat, fish and seafood, legumes, leafy greens, cereals, potato, fruit, sweets, olive oil, dairy food, and beverages, including wine and coffee) over the past 1 year was estimated. The frequency of food consumption was ranked into five categories from never to every day, i.e., never/less than once a month, 2–3 servings/month, 1–2 servings/week, 3–5 servings/week, every day, without specifying the serving size.

Marital status was stratified into the following categories: single, married, widowed, or separated/divorced. Educational attainment was expressed as the number of years spent in school and was analyzed as a continuous variable. Household composition was categorized into living alone, married without children, married with children, widowed with children, and others, while smoking status was stratified into never smoker, former smoker (having smoked at least 200 cigarettes), and current smoker. To assess for physical activity, respondents were asked how many times per week they engaged in physical activity (categorized as <3 times/week or ≥3 times/week). To estimate comorbidity, the score obtained using the Cumulative Illness Rating Scale (CIRS) was calculated [46].

Data analysis was performed using the SPSS Package for Windows version 22.0. Continuous data were expressed as the mean ± standard deviation or median (range) after checking for normality via visual inspection of Q–Q plots or the Kolmogorov–Smirnov test. Categorical data were expressed as numbers (percentages). Differences between groups were evaluated using the Mann–Whitney U test, while correlation analysis was performed by calculating the Spearman correlation coefficient.

Survival analysis was conducted by building Kaplan–Meier (KM) curves and using Cox proportional hazards regression. The lipid levels of the participants were stratified into binary subgroups using a median split to maximize between-group variation. The covariates included smoking habits, self-assessed health, education, marital status, and comorbidity. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated for the association of lipid profiles and survival. Two-sided p values < 0.05 were considered statistically significant.

The baseline characteristics of study participants selected from the original cohort of 200 LBZ nonagenarians [35,36] are illustrated in Table 1. The age range was 90–107 among men and 90–106 among women. The years spent in school were 4.1 ± 1.4 years for males and 3.3 ± 1.4 years for females. Regarding marital status, 14% and 18% of males and females, respectively, were single, 55% of males and 14% of females were currently married, and 31% of males and 68% of females were widowed. Interestingly, all male and female widowers lived with their children, except for one male. Among the study participants, none were separated or divorced. There were 6% and 11% of males and females, respectively, who lived alone, although their relatives or neighbors visited them daily; more than 50% of males lived with their wives, with or without children, whereas this occurred only in 14% of females. Nine percent and 7% of males and females lived with grandchildren or other caregivers, respectively.

There were no current smokers among the participants, only ex-smokers; thus, the two categories were collapsed into one. The percentage of current or former smokers was 33.3% and 4.6% in males and females, respectively. The distribution of BMI categories showed that females were less frequently obese compared to their male counterparts.

Most (85.2%) of the males and 69.0% of the females were engaged in physical activity three or more times per week. One hundred forty-eight deaths occurred among study participants during the follow-up period, with a mean survival time of 3.17 ± 1.81 years among males and 2.86 ± 1.45 years among females. Among the studied cohort, 20 individuals (12 males and eight females) (11.9%) were still alive in December 2024 (Table 1). Considering the small number of participants still alive, these data nearly represent the final lifespan curves in this study.

Table 2 shows the baseline lipid levels of the study participants, stratified by sex. Total cholesterol levels ranged from 89 mg/dL to 324 mg/dL, while LDL-C ranged from 31 mg/dL to 254 mg/dL. No statistically significant differences were observed between males and females for any lipid parameters considered.

Figure 2 illustrates the baseline profile of lipid parameters subdivided by sex. Two nonagenarians, one male and one female, exhibited cholesterol levels compatible with a diagnosis of familial hypercholesterolemia (LDL-C 254 mg/dL and 245 mg/dL, respectively).

Table 3 reports the survival duration in years for participants, stratified into subgroups based on lipid parameter levels.

Survival was significantly longer among participants of both sexes with LDL-C ≥ 130 mg/dL (moderate hypercholesterolemia), whereas participants in the normal/hypocholesterolemic range showed the shortest survival. A similar difference was observed for non-HDL-C levels above 158 mg/dL (p = 0.010), as well as for total cholesterol values (p < 0.0001) only among females.

Figure 3 reports survival analysis performed with KM curves, according to LDL-C and HDL-C levels. Participants were stratified according to their lipid parameters at recruitment. Survival was significantly longer among participants with LDL-C levels above 130 mg/dL (mild hypercholesterolemia) (log rank test, p < 0.0001), whereas there were no statistically significant differences according to HDL-C levels. Supplementary Figure S1 shows KM curves for all the remaining lipid parameters, revealing that participants with non-HDL-C above 158 mg/dL and total cholesterol between 200 and 249 mg/dL category also exhibited significantly longer survival. The log-rank test confirmed a statistically significant difference in survival between the mild hypercholesterolemic category and the other categories (Figure 3). The other lipid parameters behaved differently. For instance, when stratified into high and low values based on the median, HDL-C and TG did not show any association with survival. Non-HDL-C was associated with survival similar to that of the LDL-C but with more modest statistical significance. The HDL/LDL and HDL/TG ratios did not show any association with survival (Supplementary MaterialsFigure S1).

The results of the multivariable Cox proportional hazard models are reported in Table 4.

After adjusting for the covariates sex, age at recruitment, body mass index, smoking, self-rated health, and comorbidity, the analysis revealed a significant reduction in mortality in participants with mild hypercholesterolemia (LDL-C ≥ 130 mg/dL) compared to those with normal cholesterol (OR 0.606, 95% CI 0.403–0.912). Among the covariates, age at recruitment significantly impacted survival, while mild overweight but not overt obesity played a protective role.

Supplementary Table S1 reports cholesterol levels stratified according to anthropometric parameters. No statistically significant differences were observed, although the percentage of obese participants was higher among those with total cholesterol ≥ 250 mg/dL compared with those with cholesterol between 200 and 250 mg/dL (p = 0.368).

illustrates the illnesses identified through the CIRS questionnaire. The participants with diabetes had a higher frequency of the highest cholesterol levels, although the small number of participants did not allow for any definitive conclusion. The frequency of the other illnesses did not differ across the other categories. Overall, the correlation between lipid parameters, CIRS comorbidity score, and self-rated health, reported in, did not show statistically significant differences from zero. Supplemental Table S2 Supplementary Table S3

Table 5 illustrates the bivariate correlation between baseline lipid levels and food frequency consumption among study participants. Overall, the results are quite varied and not easy to interpret. The most consistent finding is the significant positive correlation between total cholesterol, LDL-C, and non-HDL-C and the frequency of cereals consumption (r = 0.194, 0.199, and 0.241 for TC, LDL-C, and non-HDL-C, respectively). Furthermore, a significant negative correlation was observed between the frequency of olive oil consumption and levels of non-HDL-C, as well as a positive correlation between coffee consumption and HDL-C levels. The Spearman correlation coefficients for the remaining foods and the other lipid parameters were not statistically different from zero.

According to some studies, the cholesterol paradox would be the consequence of the higher mortality recorded in subjects with low cholesterol levels, who might be frailer and affected by critical illnesses, including cancer, which would be the direct cause of a shortened lifespan [70]. It has also been hypothesized that in old age, the activity of acetyl-CoA acetyltransferase 2 (ACAT2) may decrease, thus reducing the conversion of free cholesterol to esterified cholesterol. This may lead to a reduced production of VLDL-C and, in turn, a concomitant drop in LDL-C [71].

In the systematic review by Ravnskov, which considered 19 studies on the association between LDL-C levels and all-cause mortality risk in subjects over 60 years [20], the possibility of reverse causality was unlikely, as at least five epidemiological studies included in this review had excluded patients with terminal diseases and even those who died within the first year of observation. Yet, this study could be considered unbalanced because the inverse correlation data appeared in a single study [72] that examined 43,666 patients out of 68,094 individuals (64.1%). Otherwise, some participants with moderate hypercholesterolemia could have started dietary treatment or even the assumption of statins during the study; as such, the impact of therapy on increased survival cannot be ruled out [72]. In the present study, reverse causality was also unlikely since subjects with the most serious pathologies did not show lower cholesterol levels. More specifically, cholesterol levels were higher in nonagenarians with type 2 diabetes, who represented 16% [34], and also in patients with low self-rated health. However, excluding both these categories, the cholesterol paradox remained statistically significant, suggesting that reverse causality cannot adequately explain the cholesterol paradox phenomenon. We also considered the possibility that higher LDL-C values were paralleled by higher HDL-C values, which are generally associated with a reduced risk of heart disease and, therefore, may help extend the lifespan by preventing atherosclerosis [73]. However, when considering the broader lipid profile (i.e., also including HDL-C, VLDL-C, and TG), it turned out that LDL-C levels did not correlate with HDL-C levels (r = −0.043, p = 0.603) and that high HDL-C levels did not predict longevity to the same effect size as high LDL-C (log-rank test, p = 0.396). Finally, from the present study, statin treatment cannot be considered a confounding factor since nonagenarians taking these drugs were excluded from the analysis.

Cholesterol is a molecule with multiple functions, and in specific contexts or population subgroups, higher cholesterol levels may be less harmful or even beneficial. In older adults, cholesterol might protect against illness by providing energy and supporting cellular repair. Some situations in which hypercholesterolemia may be beneficial can be summarized as follows: (i) being an integral component of every cell, cholesterol maintains the integrity and fluidity of cell membranes, which is crucial for proper cellular functioning. In particular, cholesterol is a component of lipid rafts, the membrane microdomains involved in downstream cell signaling [74]. For this reason, increased cholesterol levels in senescent cells may act as stabilizers of lipid raft structure, thus improving the capacity of some ligands to bind receptors [75]. (ii) Cholesterol is a precursor of steroid hormones such as cortisol, estrogen, progesterone, and testosterone. These hormones are critical for several functions in the elderly, including metabolism and immune response. Higher cholesterol levels may sustain hormone synthesis, whereas low cholesterol, also induced by cholesterol-lowering drugs such as statins and PCSK9 inhibitors, may exacerbate age-related hormonal deficiency [76]. (iii) Cholesterol is a precursor in the synthesis of vitamin D in sunlight-exposed skin. Vitamin D plays a crucial role in immune system function, bone health, and overall well-being in elderly individuals [77], and higher levels have been reported to promote longevity [78]. (iv) Higher cholesterol levels may increase the production of specific molecules that are part of the immune system’s function [79]. It has been reported that high cholesterol content in the cell membranes of cytotoxic T lymphocytes and natural killer cells can protect them from accidental perforin lysis, which causes apoptosis of target cells [80]. Some evidence indicates that higher cholesterol levels in later life might imply a stronger immune system and improved resilience to infections and diseases [81,82], which could contribute to greater longevity. (v) Cholesterol is vital for brain health as it is involved in forming synapses and supporting neurotransmission. In particular, low cholesterol levels have been associated with cognitive decline and conditions such as Alzheimer’s disease [83].

Hypercholesterolemia may result from hypothyroidism. A higher prevalence of hypothyroidism has been previously reported among the Sardinia LBZ [84]. Therefore, it could be hypothesized that this was the cause of the increased frequency of dyslipidemia in the sample population in the present study. However, only four participants (2.4%) showed signs of hypothyroidism, and only one of them fell into the highest tertile of cholesterol.

The cholesterol paradox is not an isolated phenomenon. It often occurs in conjunction with the obesity [85] and hypertension paradox [64], which are indicated by the expression “reverse epidemiology” [29]. Indeed, among long-livers from Sardinia, the phenomenon of the obesity paradox was recently reported [34]. However, in the present study, it was impossible to test the hypothesis that high blood pressure among the nonagenarians had a life-extending effect because most of them were receiving antihypertensive treatment, and the basal blood pressure values were unknown.

Since evidence shows that older individuals with high cholesterol levels do not exhibit higher-than-expected mortality rates, this suggests that other factors such as genetics or lifestyle, may play a role in these mortality rates. It is known that in the past, the population of Sardinia was characterized by a high prevalence of malaria [86]. The ability of P. falciparum to enter erythrocytes depends on the cholesterol content of the plasma membrane. Hence, having a higher cholesterol level tends to make the membrane less fluid, potentially disfavoring the entry of parasites into cells [87]. At a speculative level, it can therefore be hypothesized that during the centuries in which malaria was endemic in Sardinia, specific genetic factors were selected that increased the cholesterol set point, allowing the population to maintain higher cholesterol levels despite remaining relatively exempt from the typical adverse cardiovascular effects, by creating a condition that discouraged the development of infectious diseases. Apart from malaria, it has also been hypothesized that moderately increased cholesterol levels are protective against infectious diseases [82]; however, this hypothesis was not tested in our study due to a lack of specific information.

Additionally, lifestyle factors such as diet, exercise, and stress management significantly influence the impact of cholesterol on longevity. In this regard, the Sardinian long-lived community leads a healthy lifestyle that delays the onset of diseases [32]. The diet of the community living in the Sardinia LBZ showed a relatively high consumption of saturated fats [33,44], which could explain the moderate hypercholesterolemia detected in the present study. The inverse correlation between cholesterol levels and low consumption of monounsaturated fats (olive oil) observed in our study is intriguing and deserves further investigation. Other aspects of lifestyle, such as occupation and daily activity, could also act as mediators between cholesterol levels and survival. However, the relationship between physical activity and cholesterol levels was beyond the scope of this study and will be the subject of future research.

Our study has several limitations. The relatively small number of elderly individuals examined may have affected the results. Additionally, lipid profiles were determined only at the time of recruitment, meaning there was no information on cholesterol and other lipid levels at younger ages. Consequently, we could only assess survival from the age of 90, leaving the cumulative impact of lipids over the entire life cycle unknown. Furthermore, most participants were on antihypertensive therapy at enrollment, preventing us from including this variable in the survival analysis to determine whether increased blood pressure levels were as protective as cholesterol levels. Finally, the relationships between dietary habits and cholesterol levels are complex and not adequately addressed by our correlation analysis. Therefore, investigating the impact of diet on the cholesterol paradox will require further research with a more suitable study design.