Personalized nutrition and healthy aging

Abstract

The human lifespan and quality of life depend on complex interactions among genetic, environmental, and lifestyle factors. Aging research has been remarkably advanced by the development of high-throughput “omics” technologies. Differences between chronological and biological ages, and identification of factors (eg, nutrition) that modulate the rate of aging can now be assessed at the individual level on the basis of telomere length, the epigenome, and the metabolome. Nevertheless, the understanding of the different responses of people to dietary factors, which is the focus of precision nutrition research, remains incomplete. The lack of reliable dietary assessment methods constitutes a significant challenge in nutrition research, especially in elderly populations. For practical and successful personalized diet advice, big data techniques are needed to analyze and integrate the relevant omics (ie, genomic, epigenomic, metabolomics) with an objective and longitudinal capture of individual nutritional and environmental information. Application of such techniques will provide the scientific evidence and knowledge needed to offer actionable, personalized health recommendations to transform the promise of personalized nutrition into reality.

Introduction

The human lifespan depends on genetic, environmental, and lifestyle factors, which also exert a strong influence over one’s health status. The genetic lifespan influence varies with age and is estimated to be 20%–25% before the age of 60 years, increasing moderately after age 60 and more markedly after age 90 years.¹ Therefore, for the most part, healthy aging and longevity are determined by exogenous factors, including good dietary habits. Over the past century, improvements in several lifestyle and extrinsic factors, including medical advancements, have resulted in dramatic increases in life expectancy. Maximizing quality of life (the so-called health span) during the final years of life remains the biggest challenge of longer life spans.¹

Promoting healthy aging by tailoring nutritional guidance on the basis of a person’s individuality is an emerging science that has great promise.² To achieve this objective, it is vital to gain a better understanding of each of the pieces of the intricate machinery that drives the lifelong process of aging. The objective of this review is to provide an overview of several concepts related to personalized nutrition in elderly individuals, including the definition of “elderly,” the comparison between biological and chronological age, dietary assessment in the elderly, and the current status of interactions between genetics and diet.

WHO QUALIFIES AS ELDERLY?

Despite the frequent use of the term elderly in the scientific literature and the society at large, there is no clear definition of “the term.” Traditionally, those whose chronological age is ≥ 65 years are described as elderly, with additional subclassification as early elderly (ages 65–74 years) and late elderly (age ≥ 75 years). The basis for this threshold, however, is unknown. Some trace the origin back to 1889, when Germany became the first nation in the world to adopt an old-age social insurance program, which was designed by Germany's chancellor, Otto von Bismarck. The idea was that “those who are disabled from work by age and invalidity have a well-grounded claim to care from the state” (https://www.ssa.gov/history/age65.html). The existing saga is that the German program adopted age 65 years as the standard retirement age because that was Bismarck's age. Then, in 1935, when the United States designed its own Social Security plan, it adopted age 65 years as the age when one could receive retirement benefits, because this was the age adopted by Germany when that country created its program. The flaws in this story are that Germany initially set the age of 70 years as the retirement age, that Bismarck was 74 years old at the time, and it was not until 1916 that the retirement age was lowered to 65 years. It should be noted that the life expectancy in Germany at the time of the social insurance program was well below 70 years.

Another instance in history in which age 65 years was used to classify an individual as elderly is in the United Kingdom, where civil service pensions became payable from 1859 at a fixed retirement age of 65 years because this was the age at which “bodily and mental vigour begin to decline.” The argument was based on “imprecise evidence.” An alternative proposal for an individualized retirement age that would enable civil servants to retire when, on medical grounds, they became incapable of performing regular and efficient work was rejected as being “too difficult to administer and likely to cause resentment.”³ When the UK government introduced the first “old-age” pension in 1908, it was only available to men ≥ 70 years old at a time when the average life expectancy in the United Kingdom was 47 years.

Therefore, although the basis for the decision remains unclear, it is possible that the concept of “elderly at 65” comes from the Social Security Act, signed into law by President Franklin D. Roosevelt in 1935, providing a federal safety net for the elderly by paying financial benefits to retirees ≥ 65 years old. Keep in mind that life expectancy in the United States in 1935 was 59.9 years for men and 63.9 years for women.

It is evident, based on the life expectancies in Germany, the United Kingdom, the United States, and other countries that adopted similar pension systems, that the cost of these programs would not be an excessive burden to the countries’ economies, given the limited number of beneficiaries and the expected short duration of the benefit. This has dramatically changed, with a life expectancy in those countries of close to 80 years. Therefore, several authors have proposed different definitions and subclassifications of the term “elderly”. Orimo et al⁴ recommended changing the definition of elderly to mean those older than 75 years , instead of the current 65 years, and reforming the social system to match the coming aging society. Even earlier, in the 1980s, Watkin⁵ proposed a radical recategorization in which people older than 90 years would be considered “aged”; ages 80–89 years would constitute the new “elderly”; ages 72–79 years would be considered “aging”, and those aged 60–71 years would be “middle-aged.”

BIOLOGICAL VERSUS CHRONOLOGICAL AGE

When an alternative proposal for an individual and flexible retirement age was rejected in the United Kingdom back in the 1850s, a major reason was the difficulty of establishing an objective assessment of an individual's ability to perform regular and efficient work. Although chronological age is, on average, the major risk factor for functional disabilities, chronic diseases, and death, there is dramatic interindividual heterogeneity in the age of presentation of those outcomes. Thus, there is a need for reliable markers of the aging level of an individual, that is, a “biological age” that can predict physiological aging and risk of events more reliably than chronological age.

Several different approaches have been investigated and proposed during the past few decades. The field was recently energized by findings in 2 specific areas: telomere length and epigenetic clocks. In addition, promising results have been obtained using metabolomics, proteomics, and transcriptomics and their combinations.

Telomeres and aging

Telomeres, the TTAGGG repetitive DNA elements at the ends of the chromosomes that, in combination with a protein complex known as shelterin, form a protective loop structure against chromosome fusion and degradation, are responsible for maintaining genome integrity. Telomere length is maximum at birth and gradually shortens throughout life as a result of cell replication. Therefore, average telomere length, usually measured in human blood lymphocytes, is a proposed marker of biological age, because decrease in telomere length is accelerated in association with age-related diseases such as cancer, Alzheimer disease, type 2 diabetes (T2D), and hypertension.⁶ Proposed mechanisms for accelerated telomere shortening include inflammation and oxidative stress.⁶ The regulation of telomere length is complex, however, and telomerase activity can potentially counteract telomere shortening.

Previous research demonstrated that telomerase RNA component (TERC) genetic variants are associated with lymphocyte telomere length (LTL). Thus, a genome-wide association study of mean LTL in 2917 individuals, with follow-up replication in 9492 individuals, found that for the rs12696304 (C>G) single nucleotide polymorphism (SNP), the presence of 1 copy of the G allele (the less common allele in White populations) was highly significantly associated with a approximately 75-base-pair reduction in the mean telomere length, which is equivalent to approximately 3.6 years of age-related telomere-length attrition. In other words, in terms of biological aging, this implies that individuals carrying the rs12696304(G/C) or (G/G) genotypes will appear approximately 3.6 and approximately 7.2 years “older”, respectively, than those with the rs12696304(C/C) genotype.⁷ It should be noted that these findings were obtained in White cohorts, and it is not known if similar results will be obtained for other races.

On the basis of these results, Gomez-Delgado et al⁸ investigated whether variability at the TERC gene locus was associated with inflammation and glucose metabolism and whether diet modified the potential association. For these purposes, inflammation status (indicated by measures of high-sensitivity C-reactive protein [hsCRP], glucose metabolism [ie, glucose, insulin, and glycated hemoglobin values, and homeostasis model assessment of insulin resistance], LTL, and of the TERC gene SNPs [rs12696304, rs16847897, and rs3772190]) were determined in 1002 patients with coronary heart disease participating in the dietary intervention (low-fat diet vs. Mediterranean diet) Coronary Diet Intervention With Olive Oil and Cardiovascular Prevention (CORDIOPREV) study. Consistent with the results reported by the aforementioned genome-wide association study,⁷ the results from the CORDIOPREV study demonstrated that the presence of the G allele at TERC rs12696304 is associated, at baseline, with lower average LTL compared with the C/C genotype. Moreover, the investigators found statistically significant interactions of the TERC rs12696304 SNP with monounsaturated fatty acids affecting LTL and hsCRP. In brief, among individuals with baseline monounsaturated fatty acid levels above the median, individuals with the C/C genotype had higher LTL and lower hsCRP levels than did G-allele carriers.

Moreover, there was a significant interaction between habitual consumption of a Mediterranean diet and the TERC rs12696304 SNP modulating inflammation.⁸ Specifically, individuals with the C/C genotype had a more significant decrease in hsCRP concentration than did G-allele carriers. Therefore, these findings suggest that the TERC rs12696304 SNP interacts with monounsaturated fatty acids, improving inflammation status and telomere attrition related to coronary heart disease. Also, the Mediterranean diet intervention improved the inflammatory profile in individuals with the C/C genotype compared with G-allele carriers. These interactions support the notion of devising personalized nutrition advice for patients with coronary heart disease. Whether this applies to other populations requires additional investigation.

DNA methylation age

The aging process results in epigenetic changes that modulate gene expression without altering the nucleotide sequence. The nature of these changes is not fully understood, but an increasing amount of evidence indicates these epigenetic changes are involved in aging and longevity. Along these lines, authors of several studies have proposed DNA methylation age (DNAmAge), also known as the epigenetic clock or Horvath’s clock, as a predictor of biological age. Epigenetic clocks comprise a set of CpG sites whose DNA methylation levels supposedly measure a person’s biological age. These clocks have been recognized by many as accurate molecular correlates of chronological age in humans and other vertebrates.⁹ The 2 most commonly used epigenetic clocks are those developed concurrently by Horvath¹⁰ and Hannum et al,¹¹ both of whom reported correlations with chronological age greater than 0.9 in their initial reports. The original Horvath’s clock is based on methylation levels at 353 CpG sites using the Illumina 27k array.¹⁰ In contrast, the Hannum et al¹¹ clock was developed on the basis of 71 CpG sites from the Illumina 450k array. Interestingly, they only have 6 CpG sites in common, and the age correlations obtained using different clocks range from moderate (r = 0.37)¹² to strong (r = 0.76).¹³

Despite some promising and exciting results obtained for DNAmAge, a recent review and meta-analysis provide a precautionary view of the field.¹⁴ In this systematic review, the authors identified and synthesized the evidence for an association between peripherally measured DNA methylation age and longevity, age-related disease, and mortality risk. The authors identified 23 articles (n = 41 607 participants); 4 studies focused on aging and longevity, 11 on age-related diseases (namely, cardiovascular disease, cancer, and dementia), and 11 on death. Overall, the authors found evidence for associations between increased DNAmAge and the risk of age-related diseases. The meta-analysis focusing on death suggested that for each 5-year increase in DNA methylation age, there was an 8% to 15% increased risk of death, depending on the use of the age acceleration calculated with Horvath’s clock or that calculated with Hannum’s clock, respectively. A positive publication bias, however, may have existed. Overall, the authors concluded that given the paucity of studies and their heterogeneity, the association between DNAmAge and age-related disease and longevity is inconclusive.

A better understanding of the mechanisms associated with the methylation clock may provide the opportunity to prevent or revert accelerated aging. After all, it is possible that epigenetic marks, unlike DNA variants, may be reversible.¹⁵ Therefore, some studies have examined the relationship between methylation age and the environment, including behavioral (eg, diet, exercise, education), and lifestyle factors. Quach et al¹⁶ analyzed cross-sectional data from > 4000 postmenopausal women participating in the Women's Health Initiative, as well as > 400 men and women participants in the Italian cohort study, Invecchiare nel Chianti. Extrinsic epigenetic age acceleration, which incorporates intrinsic measures as well as blood cell proportions, showed significant associations with fish intake, moderate alcohol consumption, education, body mass index (BMI), and blood carotenoid levels. Another measure, intrinsic epigenetic age acceleration, which captures cellular age acceleration independently of blood-cell proportions that are known to change with age, was associated with poultry intake and BMI. Extrinsic and intrinsic epigenetic age acceleration were also associated with metabolic syndrome. Analyses of longitudinal data indicated that increases in BMI were associated with increases in both extrinsic and intrinsic epigenetic age acceleration. Therefore, these data support the benefits of eating fruits and vegetables, lean meats, and moderate alcohol consumption, as well as performing physical activity. Conversely, the findings underline the risks of obesity and metabolic syndrome for healthy aging.

Taken together, the epigenetic clock appears to be associated with a broad spectrum of aging outcomes. The positive findings may represent the tip of the iceberg of this body of research, however, and many negative results may remain unpublished. Moreover, different organs may age at different speeds, and most research in humans has been restricted to blood cells; however, the epigenetic aging rate in 1 tissue can be quite different from that of another, and for biomarker purposes, it is not realistic to obtain a combination of epigenetic age estimates in several tissues. Another remaining question is whether the clock’s rate can be modified by diet, and how. For this reason, it will be essential to define whether the methylation changes observed with age drive the phenotypes or whether they reflect the work of other genomic control mechanisms.¹⁷

Metabolomic age.

The use of metabolomics to define an individual’s biological age is an area of increasing interest. Metabolomics integrates clues from the environment as well as the individual’s genetic makeup; thus, it could provide a more complete assessment of an individual’s health status and, potentially, biological age. Using proton nuclear magnetic resonance metabolomics, and phenotypic information on > 25 000 people, van den Akker et al¹⁸ constructed a biological age prediction tool known as metaboAge (available at https://metaboage.researchlumc.nl). These investigators demonstrated that a positive difference for biological versus chronological age was associated with increased risk for cardiovascular disease, death, and functional decline.

Using the same metabolomic technique, this group previously developed metabolomic predictors of mortality using data from > 40 000 people.¹⁹ The prediction accuracy of 5- and 10-year mortality based on a model including the identified biomarkers and sex was 0.837 and 0.830, respectively (C statistics), which was improved accuracy compared with a model containing conventional mortality risk factors (C = 0.77 and 0.79, respectively).

On the basis of urine metabolomic data generated using proton nuclear magnetic resonance, Hertel et al²⁰ developed a metabolomic age score having significant associations with chronobiological age and clinical phenotypes. Moreover, the score was a good predictor of survival after 13 years of follow-up. Finally, the investigators demonstrated the value of the metabolomic age score to predict weight loss after bariatric surgery, underscoring its possible application for precision medicine. Other scores with the same purpose were developed by Rist et al²¹ using a combination of metabolomic techniques and data from urine and plasma.

Multi-omic biological age estimation

Different omics may be able to capture different aspects of the aging processes. Therefore, it is plausible that a multi-omic approach could provide more comprehensive biological age prediction and mortality models. The development of such an estimate was carried out by Earls et al²² using longitudinal deep phenotyping data (genetic, clinical laboratory, metabolomic, and proteomic) of 3558 individuals. Biological age was older than chronological age in the presence of chronic diseases. Conversely, individuals participating in a wellness program had a < 1-year rate of biological age change per year, supporting the notion that biological age could be modifiable and that a biological age younger than the chronobiological age could be an indication of healthy aging.

More recently, Robinson et al²³ developed a regression model of age based on several metabolomic platforms, including nuclear magnetic resonance and liquid chromatography-mass spectrometry, in urine and serum using a sample (n = 2,239) from the UK Occupational Airwave cohort. They investigated determinants of accelerated aging, including genetic, lifestyle, and psychological risk factors for early death. Using the metabolomic platforms, the correlation between chronological and biological age was 0.85, whereas the correlation using the DNAmAge was 0.91. Increased metabolomic age acceleration was associated with overweight and obesity, depression, high amount of alcohol use, and low income. DNAmAge acceleration was associated with high amount of alcohol consumption, anxiety, and post-traumatic stress disorder. DNAmAge was not correlated with metabolomic age acceleration, however, suggesting that metabolomic age and change in DNAmAge capture different aspects of the aging process. Among the genetic factors examined, only the APOE locus was significantly associated with both metabolomic age acceleration and DNAmAge acceleration.

Of note, there was no significant association after adjusting for multiple comparisons between any of the biological markers of age and accelerated age and dietary factors (ie, fish, fruit, red meat, vegetables, whole grain, Mediterranean diet score, and Dietary Approaches to Stop Hypertension score). Given the current knowledge of the relationship between diet and healthy aging, one can hypothesize that lack of accuracy on the assessment of the dietary variables could be responsible for the lack of positive results.

CHALLENGES AND OPPORTUNITIES FOR PRECISION NUTRITION IN ELDERLY POPULATIONS

Dietary assessment methods certainly constitute a major challenges in nutrition research and personalized nutrition applications. Traditional dietary assessment methods are frequently criticized for their subjectivity and poor accuracy. Some perform better than others, however, depending on their intrinsic characteristics, the aim of the study, and the targeted population. When the target is the elderly population, it is crucial to consider the individual characteristics. For mentally and physically healthy elderly persons, it may be appropriate to use a food frequency questionnaire or dietary recall. On the other hand, for mentally and/or physically compromised elderly persons, tailored instruments may be required.²⁴ de Vries et al²⁵ described their experience with dietary assessment in populations aged ≥ 65 years. They used dietary history in longitudinal studies; trained staff observed and recorded food consumption in nursing home studies, and food frequency questionnaires were used in trials involving healthy elderly participants. In general, there was an underestimation of dietary intake regardless of the instrument. Moreover, these authors concluded that the major challenge for appropriate dietary assessment in older adults is to distinguish between those who will be able to respond correctly to the less involved methods, such as 24-hour recalls or food frequency questionnaires, and those who will not and, therefore, will require other techniques, such as observational records. There may also be bias with dietary recalls, because of social desirability, even with elderly persons, and also because people with specific diseases such as diabetes may also under-report their intake.²⁶,²⁷

Several additional factors affect dietary assessment in the elderly. Given the frequency of chronic diseases, there is a high probability of people being on special diets (eg,, low fat). This will affect the dietary intake, but it also could increase the reporting bias because people will tend to report what they should be eating rather than what they are really eating. The opposite may also be true, however, and these people, being more aware of their diets, could report them more accurately. Moreover, the elderly often take more nutritional supplements,²⁸ complicating the dietary assessment. Therefore, the variability in functional status among the elderly suggests the need for alternative approaches to assessing dietary intake.²⁹

New technology-based dietary assessment tools were recently evaluated.³⁰ One such tool is the use of image-assisted dietary assessment.³¹ The technique yields more objective data on food selection, intake, and waste, and has been used in different settings.³² Its validity in a geriatric setting was investigated by Pouyet et al³³ in French nursing homes. Total food intake was assessed by photography and by weighing the food before and after consumption. The findings suggested the photographic method is reliable for measuring food intake by the elderly in an institutionalized setting.

The ultimate goal in the dietary assessment is to determine reliable biomarkers providing an objective assessment of food intake and nutritional exposure. Advances in metabolomics techniques and bioinformatic tools may facilitate such biomarker development. Innovations in this area will allow us to develop new dietary assessment methods that take into account not only food intake but also metabolism, reflecting the “internal dose” and nutritional status of individuals. Although several needs and gaps need to be addressed before metabolomic-based dietary biomarkers can be used in public health, clinical research, and individual dietary assessment,³⁴ these tools are expected to offer powerful alternatives to overcome well-known limitations of the currently available methods.

Longitudinal studies conducted in elderly populations share other challenges in data collection and analysis. In particular, increased attrition and healthy survivor bias in aging cohorts may compromise the validity of the findings, especially in the case of samples of advanced age with lower survival rates. Not-missing-at-random analysis models can facilitate the handling of missing data resulting from death or poor health compared with other popular strategies that make additional assumptions about missingness mechanisms.³⁵

In terms of practical applications, it is evident that precision nutrition strategies designed for elderly men and women will favor complete solutions that are able to characterize and provide diets adapted to the individual needs (eg, nutritional profile, organoleptic characteristics, format, delivery). Although personalized nutrition advice and counseling may not be as effective as in younger populations and equitable access remains a challenge, precision nutrition diets for elderly individuals have the potential to mitigate age-associated comorbidities, ensure proper nutrition, and promote quality of life at every life stage.³⁶

GENE–DIET INTERACTIONS IN THE ELDERLY

Disparities in the response of people to dietary factors have been documented for almost a century³⁷ and even earlier, based on the expression attributed to Titus Lucretius Carus in the first century BC: “quod ali cibus est aliis fuat acre venenum” (Latin, meaning “what is food for one man may be bitter poison to others”). This provides the foundation for developing personalized nutrition approaches. Initially, this field evolved with the study of gene-by-diet interactions, which gave origin to nutrigenetics, which has been defined as “the discipline that studies the different phenotypic response to diet depending on the genotype of each individual.”³⁸ Whereas most previous research has focused on age-related diseases (eg, cardiovascular diseases, obesity, T2D), there is a paucity of studies focusing on gene-by-diet interactions in the elderly and, in general, they tend to be over a decade old.^39–43 Even the Cohorts for Heart and Aging Research in Genomic Epidemiology consortium has not reported gene-by-diet interactions in the elderly, despite examining traits very relevant in elderly subjects (eg, inflammaging).⁴⁴ In 1 study, however, of genome-wide interactions with dairy intake in relation to BMI, using sensitivity analyses with cohorts divided by geographic region and mean age, researchers observed greater consistency for the interactions in European cohorts and older individuals.⁴⁵

A notable concern related to gene-by-diet interaction studies in the elderly is the potential confounding effect of medications. According to the American Society of Consultant Pharmacists, people aged 65–69 years take an average of 15 prescribed medications per year, whereas those aged 80–84 years take 18 prescribed medications per year. This is on top of the countless over-the-counter drugs and dietary supplements they may take.²⁸ Therefore, depending on the specific trait investigated, the effects of drugs may easily overwhelm the gene-by-diet interaction.

Intervention studies such as Prevención con Dieta Mediterránea (PREDIMED)⁴⁶ and CORDIOPREV⁴⁷ are rich sources of information for the discovery and replication of gene–diet interactions related to cardiovascular diseases.⁸,^48–53 The potential for interactions among the 3 factors—genetics, diet, and age—however, remains unexplored.⁵⁴

Although the PREDIMED study did not strictly qualify as an study of elderly persons at baseline (inclusion criteria were men aged 55–80 years or women aged 60–80 years without cardiovascular disease), a finding from the study provides a glimpse into the potential benefits of personalized nutrition in the elderly.⁵⁰ Along these lines, the authors investigated whether the associations between the eTCF7L2-rs7903146 (C>T) SNPs and T2D, glucose, lipids, and cardiovascular disease incidence (∼4.8-year follow-up) were modulated by a Mediterranean diet. As expected, the TCF7L2-rs7903146 SNP was significantly associated with T2D. Moreover, the Mediterranean diet interacted significantly with this SNP on fasting glucose at baseline. When adherence to the Mediterranean diet was low, TT participants (ie, those homozygous for the risk allele) had significantly higher fasting glucose concentrations than participants with the CC+CT genotype. Conversely, when the adherence was high, there were no significant differences in glucose levels between genotypes. Most importantly, during the trial, TT participants in the control (ie, low-fat diet) group had a higher stroke incidence than CC participants; however, TT participants in the Mediterranean diet group had a much lower stroke incidence, similar to that of participants with the CC genotype who consumed the Mediterranean diet. These findings are meaningful and promising because they suggest personalized dietary intervention could have a beneficial effect in persons at high risk even if the intervention is implemented at 55–60 years of age and during a relatively short period (< 5 years).

CONCLUSION: THE FUTURE OF PERSONALIZED NUTRITION

Nutrigenetics represents just the tip of the iceberg of personalized nutrition and, by itself, will not be able to provide practical and successful personalized advice to the population at large and the increasingly elderly population in particular. For practical and successful personalized diet advice, the power of big data techniques will be needed to analyze, in an integrated manner, the relevant omics (ie, genomic, epigenomic, metabolomics, and microbiomics)^55–57 with an objective and longitudinal capture of the individual environmental information (exposomics).⁵⁸ The results of these analyses will provide the scientific evidence and knowledge to offer actionable, personalized health recommendations to improve the quality of the aging process and to transform the promise of personalized nutrition into reality.

Acknowledgments

Author contributions. J.M.O. and S.B. contributed equally to the conception and writing of this manuscript.

Declaration of Interests. J.M.O. has received research funding from the US Department of Agriculture for work on personalized nutrition and from Archer Daniels Midland for research on probiotics; has served as a Scientific Advisory Board consultant for Nutrigenomix, the Predict Study, GNC, and Weight Watchers. Travel support was received from International Life Sciences Institute Japan to participate in the International Conference on Nutrition and Aging.

Funding. This work was partially supported by the US Department of Agriculture (agreement no. 8050-51000-098-00 D).

References