What is aging? At first glance, the answer seems to have been known for a long time. However, over the past several decades, the biology of aging has proposed dozens of different theories and hundreds of presumed mechanisms, without coming closer to a unified understanding of what aging actually is and how it can be managed. This article offers a different view of the problem. Instead of searching for an endless number of molecular causes, it considers what can be objectively measured, verified, and used in clinical practice: the age-related increase in mortality, the age-related incidence of diseases, and the gradual decline in organ function. This approach makes it possible to move from theoretical discussions about the mechanisms of aging to a practical strategy for managing human health and lifespan.
If Age-Related Changes Did Not Increase the Risk of Death and Serious Diseases, Aging Would Not Be a Major Medical Problem
Aging is not gray hair, wrinkles, or the age listed in a passport. If age-related changes did not increase the risk of death and serious diseases, aging would not be a major medical problem. Gray hair by itself does not prevent people from enjoying life. The real tragedy of aging lies elsewhere. With each passing year, the probability of myocardial infarction, stroke, cancer, dementia, loss of independence, and death increases at an ever faster rate. This is why modern medicine regards aging primarily as the greatest risk factor for most chronic diseases [https://www.nature.com/articles/s41467-021-23014-1].
The most alarming aspect is that this risk does not increase linearly but almost exponentially. In adults, the probability of death approximately doubles every 8–11 years (according to different estimates)—a pattern known as the Gompertz law, which has been repeatedly confirmed in large demographic datasets [https://pubmed.ncbi.nlm.nih.gov/30729179/] [https://biobank.ndph.ox.ac.uk/ukb/pub.cgi?id=1828]. Similarly, the accumulation of chronic diseases and functional impairments accelerates with age. Therefore, the main problem of aging is not changes in appearance but the increasingly rapid rise in the probability of severe disease, disability, and death [https://pubmed.ncbi.nlm.nih.gov/38622100/]. This is precisely what makes combating the biological causes of aging one of the most important challenges of modern medicine.
All these facts lead to the conclusion that any theory of aging must mathematically explain the actual age-related patterns of mortality. According to aging researchers Leonid Gavrilov and Natalia Gavrilova [https://pubmed.ncbi.nlm.nih.gov/38622100/], the ability to reproduce the laws of mortality is an important criterion for the adequacy of a model of aging.
2. Since the Main Medical Problem of Aging Is the Age-Related Increase in Mortality and Disease, These Phenomena Must Be Described Quantitatively and Mathematically
In the previous section, we established that any theory of aging must mathematically explain the actual age-related patterns of mortality. Mathematically, aging is a process or a set of processes that, with increasing age, lead to an increase in the probability of dying at each subsequent moment in time [https://pubmed.ncbi.nlm.nih.gov/25750242/] [https://pubmed.ncbi.nlm.nih.gov/9074828/], or, in other words, to an increase in the force of mortality. R(t) = A × e^(G×t), where R is the force of mortality, t is age, A is the baseline force of mortality, and G is the rate of increase in the force of mortality. In aging, G is always greater than zero.
Non-aging is a state in which the force of mortality does not increase with age. Mathematically, this means that in the equation R(t) = A × e^(G×t), the rate of increase in the force of mortality, G, is always equal to zero. In this case, the force of mortality R(t) remains constant and equal to A throughout life. It is important to understand that even under non-aging, the probability of dying still increases with age because the cumulative probability of death continues to rise. The more time passes, the greater the chance that a random fatal event will eventually occur, but the instantaneous probability of dying at any given moment remains unchanged.
In the Gompertz equation, the baseline force of mortality, A, reflects the overall level of age-related risks at the initial stage, whereas G describes the "rate of aging," that is, the rate of exponential increase in these risks with age. If we eliminate one of the causes of aging that has the same rate of risk increase (the same G) as the other causes, we primarily reduce A—the overall initial burden of age-related causes—but we do not substantially affect G, because the remaining causes continue to increase at nearly the same rate. Thus, by eliminating one of the causes of aging, we may extend lifespan without changing the rate of aging itself.
But where does mortality in young people come from if accidents and infectious diseases are excluded? Does the absence of an age-related increase in mortality imply an almost endless life? No. Even between the ages of 10 and 24 years, about 27% of all deaths worldwide are caused by non-communicable diseases, including cardiovascular diseases, cancer, diabetes, and other chronic disorders. Consequently, even in the absence of pronounced aging, humans retain a baseline risk of dying from disease. Aging does not create mortality from scratch; rather, it primarily multiplies an already existing risk [https://www.thelancet.com/journals/lancet/article/piis0140-6736%2821%2901546-4/fulltext].
Imagine two jars: one with a capacity of 2 liters and another with a capacity of 5 liters. We begin pouring water into both of them. Suppose that water flows quickly into one jar and slowly into the other. Intuitively, it may seem that the pouring speed is the most important factor because the faster-filling jar will be filled sooner. However, this is not always the case. If one jar is much larger, it may overflow later despite being filled more rapidly than a smaller jar into which water is poured slowly. What matters is not how quickly the water is poured but which jar overflows first. This is precisely the essence of the difference between the rate of aging and the baseline force of mortality when described in the language of the science of life and death.
In the Gompertz law, which describes how the risk of death changes with age, there are two main parameters. The first is the baseline force of mortality—the initial level of vulnerability of the organism. This is analogous to the size of the jar—the initial reserve of resilience. The second parameter is the rate of increase in the force of mortality, that is, the rate of aging. This is analogous to the speed at which water is poured into the jar. We often focus primarily on this speed, assuming that slower aging necessarily means a longer life. However, this is not always true.
A person with high baseline vulnerability (a small jar) may live a shorter life even if they age slowly. Conversely, a person with low initial vulnerability (a large jar) may survive to an advanced age despite a relatively rapid rate of aging. What matters is not only how fast the process proceeds but also where it begins. Life is not only about speed—it is also about capacity.
This is why lifespan cannot be judged solely by the rate of aging. As in the example of the jars, the final outcome—when the water overflows or when a person dies—depends on the combination of two factors: the baseline force of mortality and the rate at which it increases. In reality, what matters is not who ages faster but whose life "overflows" first.
Let us consider an example of how it is possible to have even a "zero" rate of aging while still dying much sooner than under aging. In naked mole-rats, the baseline force of mortality, A, is already as high in early life as it is in 72-year-old humans. This means that their life expectancy is comparable to that of 72-year-old humans—that is, half of them will die within the next 27 years. If humans stopped aging at the age of 72 years, they still would not live much longer because their force of mortality would already be high. In naked mole-rats, either the increase in mortality from age-related diseases is masked by very high mortality due to fighting, or, even if they truly do not age, they still cannot live very long because of their high baseline force of mortality according to the Gompertz model. Aging increases the probability of dying at each subsequent moment in time, whereas non-aging merely means that the force of mortality remains constant, not that it disappears. In humans aged 35 years, the probability of dying within one year is approximately 0.0015, whereas in naked mole-rats this value remains approximately 0.0365 per year even after 35 years of life, corresponding to the mortality level of a 72-year-old human. If the force of mortality remains constant, the expected remaining lifespan of the mole-rat would be 1 / 0.0365 = 27 years, meaning that half of the 35-year-old individuals would die within the next 27 years, reaching an age of about 62 years. However, this conclusion is valid only if the force of mortality truly does not increase with age, which has not been definitively established. Therefore, our goal should not be to combat the rate of aging itself, but rather any process that shortens life [https://link.springer.com/article/10.1007/s11357-024-01201-4].
3. Mathematical Models Make It Possible to Classify Biological Processes According to Their Relationship with the Age-Related Increase in Mortality and Disease. They Do Not Explain the Mechanisms of Aging but Allow Their Clinical Significance to Be Evaluated.
It is important to distinguish between the rate of aging, the mechanisms of aging, and the causes of aging. The mechanisms and causes of aging belong to the field of biology, whereas the rate of aging is the mathematical phenomenon of the age-related increase in the force of mortality. A mathematical phenomenon cannot explain the causes or mechanisms of aging, but it makes it possible to determine whether what we observe in biology can truly be considered aging.
If aging is defined as a set of biological processes, there is a risk of error because these processes alone, without a mathematical criterion, do not allow us to determine rigorously whether they actually lead to an age-related increase in the force of mortality.
In contrast, defining aging as the age-related increase in the force of mortality itself provides a precise and reproducible criterion. Once this criterion is established, it becomes possible to investigate which specific biological mechanisms generate this observable mathematical process.
The aging process is a biological phenomenon associated with a set of changes in the organism that lead to the progressive deterioration of function and an increased risk of mortality. The Gompertz law describes the rate at which mortality increases with age but does not explain the underlying biological processes. It is based on the observation that, with age, the risk of death from certain causes increases exponentially. To determine which biological processes are associated with aging, the following principle can be applied: a process is considered to be associated with aging if it leads to an increasing force of mortality that, in the foreseeable future, would become a cause of death even in the absence of other age-related changes. However, not all causes of mortality satisfy this criterion. For example, malignant neoplasms of the cervix do not exhibit an exponential age-related increase in the force of mortality and therefore cannot be classified as manifestations of aging. Thus, mathematical methods such as Gompertz analysis are necessary for classifying processes as aging-related or non-aging-related. They do not describe biological mechanisms but make it possible to evaluate the impact of biological processes on mortality. Without quantitative methods, we cannot objectively distinguish between different processes, making measurement essential for understanding aging.
The rate of aging, as a mathematical phenomenon, represents a fundamental concept that has been applied since 1975 and continues to be used today. The calculation of MRDT (Mortality Rate Doubling Time) for estimating the rate of aging on the basis of the Gompertz law was first proposed in 1975 by the American epidemiologist Richard G. Peto. To this day, many leading gerontologists worldwide use MRDT and the Gompertz function to assess the rate of aging [https://academic.oup.com/genetics/article/204/3/905/6066287] [https://academic.oup.com/genetics/article/208/4/1617/6084259] [https://pubmed.ncbi.nlm.nih.gov/34151374]. The rate of aging reflects the increase in the probability of death at each successive moment in time as age increases, accompanied by an increase in the force of mortality.
Regardless of the method of measurement, whether the Weibull distribution or the Gompertz law is used, the same phenomenon is observed: aging has a measurable rate, manifested by an increasing probability of death at successive moments in time as age advances.
The Gompertz law itself, which describes the exponential increase in the overall force of mortality with age, does not reveal the causal mechanisms of aging. Overall mortality represents the sum of mortality from many different causes and therefore provides only limited information for the analysis of biological processes. A more informative approach is to analyze the age-related dynamics of the force of mortality separately for different groups of diseases, or, even better, for each individual disease classified according to the International Classification of Diseases (ICD). Such an analysis makes it possible to determine which diseases truly exhibit the exponential age-related increase in risk that is characteristic of aging [https://pubmed.ncbi.nlm.nih.gov/30099284/].
An even more informative approach is to analyze not the force of mortality but the force of incidence of each disease. Death is the final outcome and can occur only once in each individual, whereas age-related diseases can accumulate simultaneously. Therefore, analyzing the age-related dynamics of disease incidence provides a more accurate quantitative description of aging than analyzing overall mortality. If the force of incidence of a disease increases according to the same pattern and at the same rate [https://pubmed.ncbi.nlm.nih.gov/30729179/] as the force of mortality from that disease, this provides additional evidence that the disease should be regarded as a manifestation of the aging process.
However, for the practical monitoring of aging, waiting until a disease develops is already too late. Every clinical diagnosis is preceded by a gradual decline in the function of an organ or physiological system. For example, the diagnosis of diastolic dysfunction is established only after diastolic cardiac function has fallen below the diagnostic threshold, and the diagnosis of chronic kidney disease is established only after the glomerular filtration rate has declined to the corresponding threshold. Therefore, continuous measurement of functional parameters makes it possible to monitor age-related changes long before clinical disease incidence occurs. These functional parameters are not arbitrary biomarkers. They have been validated in clinical medicine as predictors of future disease incidence and mortality and form the basis of modern diagnostic criteria (serving as the "gold standards"), making them among the most reliable tools for assessing the risk of age-related diseases.
Thus, the practical application of the Gompertz law does not lie in analyzing overall mortality as such, but in using age-related patterns of mortality and disease incidence to identify age-related diseases, after which it becomes possible to monitor an individual's aging trajectory through changes in the functions of organs and physiological systems. The accuracy of this approach could potentially be further improved by machine learning methods that estimate individual rates of age-related change and predict when functional parameters will cross diagnostic thresholds.
It is important to distinguish between age and aging. Age itself is not evidence of aging. Any age-related change can be regarded as a component of aging only if it has been clinically demonstrated to be associated with an increased risk of disease or death and has sufficient predictive value for these outcomes.
Researchers often overlook the very necessity of validating age-related changes against clinical risk. In many cases, they do not test whether the observed age-related changes are associated with clinically meaningful outcomes, such as increased mortality, disease incidence, or functional decline. This is particularly critical when developing artificial intelligence models trained exclusively on data from individuals classified as Health Group I. If such models are not longitudinal and do not use long-term follow-up of the same individuals, they largely lose the signal of aging. The main reason is that the proportion of practically healthy individuals declines rapidly with age: among people aged 21–36 years, it may be approximately 60%; among those aged 39–60 years, only 18–22%; and after the age of 60 years, only 2–3% [http://vestnik.mednet.ru/content/view/1434/30/lang,ru]. This creates a pronounced class imbalance. As a result, while minimizing the average training error, the algorithm systematically shifts its predictions toward younger ages, underestimating the biological age of older participants and overestimating it in younger ones. This reduces the model's ability to correctly identify age-related changes that are truly associated with aging rather than with the characteristics of the training dataset.
Thus, if a model is trained only on a dataset consisting of individuals from Health Group I, a pronounced age imbalance arises. As a result, the model "rejuvenates" older individuals and "ages" younger ones. Aging is not a disease, but diseases make it visible.
If a biological age model is trained on a dataset composed predominantly of individuals from Health Group I, it typically reproduces not the true relationship of "predicted age = chronological age" but a shallower trend. As a result, the model systematically overestimates age in younger individuals and systematically underestimates it in older individuals. In other words, the mean predicted age no longer matches the mean chronological age, and the slope of the regression line becomes less than one. With proper model calibration, the mean predicted age should coincide with the mean chronological age, while prediction errors should fluctuate randomly around this line without any systematic bias.
Such systematic bias creates a serious problem when evaluating potential geroprotectors. If biological age is measured with such a model before and after an intervention, chronological age increases over time faster than the age predicted by the model. Consequently, even in the complete absence of any effect of the intervention, it may appear that biological age is increasing more slowly than chronological age. This can create the false impression that aging has been slowed.
Therefore, models with this type of age-related miscalibration cannot reliably evaluate geroprotective effects. Without first verifying the absence of systematic prediction error across the entire age range, there is a substantial risk of obtaining false-positive results, in which compounds are incorrectly classified as geroprotectors because of statistical bias in the model rather than because they genuinely slow the aging process.
Unfortunately, most biological age models worldwide are developed using artificially health-selected cohorts. Many authors have highlighted this important problem. For example, in 2024, Hongqian Qi from Nankai University (China) proposed the DeepQA model [https://pubmed.ncbi.nlm.nih.gov/39757434/], which was trained on a mixed dataset including both healthy and diseased individuals. In the same study, Hongqian Qi also clearly demonstrated that models developed using health-selected cohorts produce compressed regression lines (slope <1), resulting in "rejuvenating" bias.
Many researchers attempt to predict age-related changes and then try to treat them. To achieve this, they develop first-generation biological age clocks. Instead of predicting risk, first-generation clocks are trained to predict chronological age. However, not everything that changes with age is necessarily harmful or life-limiting [https://www.biorxiv.org/content/10.1101/2024.06.06.597715v1]. Fingerprints change with age, but this does not kill us. Plants develop more growth rings as they age. Telomere length also changes with age in humans, although only weakly, with a correlation of approximately 9%. However, in humans, telomere length does not significantly predict all-cause mortality (HR 0.882, p = 0.092) or cardiovascular mortality (HR 0.897, p = 0.150) [https://pubmed.ncbi.nlm.nih.gov/29920523/].
First-generation biological age clocks were originally trained solely to predict chronological age. Chronological age itself was their target variable. Consequently, these models were optimized exclusively to minimize the error of age prediction rather than to predict the risk of death, the rate of aging, or rejuvenation. This fundamentally distinguishes them from second-generation clocks. Second-generation clocks are trained on clinical outcomes such as mortality, disease incidence, or other measures of risk. Therefore, deviations of their predictions from the average age-related trend contain information about individual differences in risk. These deviations constitute the informative signal for such models [https://www.biorxiv.org/content/10.1101/2024.06.06.597715v1]. The situation is different for first-generation clocks. Any deviation between the predicted age and the chronological age is not something the model was specifically trained to detect. For these models, such deviations represent only the residual error of age prediction. Even if, in a particular individual, this prediction error happens to coincide with a true change in mortality risk, such a coincidence does not mean that the model has learned to measure risk. It was never trained to do so.
The main reason is that age and the force of mortality are different variables. According to the Gompertz law, the force of mortality generally increases with age. However, even if an intervention reduces the force of mortality, R(t), chronological age, t, continues to increase inexorably. Therefore, the correlation between age and mortality at the population level does not imply that a model trained to predict age automatically becomes a measure of changes in mortality risk. In other words, first-generation clocks can predict only the variable on which they were trained—chronological age. They were never trained to recognize individual deviations in risk around the age-related trend. Consequently, their prediction errors cannot automatically be interpreted as accelerated or slowed aging, rejuvenation, or reduced mortality risk. Such conclusions require models that have been specifically trained and validated using the corresponding clinical outcomes.
First-generation clocks are a clear example of predicting age (chronological age), but not aging (risk).
Let us imagine, hypothetically, that it became possible to reduce certain age-related changes in humans that are unrelated to risk. Telomeres might become longer, yet mortality risk would remain unchanged [https://pubmed.ncbi.nlm.nih.gov/29920523/]. Fingerprints might come to resemble those of a younger person, but this likewise would have no effect on lifespan. By the same logic, one could simply change the date of birth in a passport, since that value also changes with age. But would doing so increase longevity?
A striking example of the distinction between age-related changes and aging is provided by the evolution of the diagnostic criteria for left ventricular diastolic dysfunction. The older diagnostic criteria primarily reflected age-related changes in the heart's diastolic function. However, many of these changes had not been validated as predictors of heart failure or mortality. As a result, the prevalence of diastolic dysfunction according to the old criteria increased dramatically with age and reached very high levels in the majority of older adults [https://jamanetwork.com/journals/jama/fullarticle/195749]. This indicated the accumulation of age-related changes rather than the identification of processes that determine clinical prognosis.
Therefore, even if an intervention improved such age-related parameters, this alone did not necessarily mean that it reduced the risk of heart failure or improved survival. In this case, treating age-related changes did not lead to an improvement in clinical prognosis.
For this reason, new diagnostic criteria for diastolic dysfunction were proposed in 2016, based primarily on markers of elevated left ventricular filling pressure. These criteria are much more strongly associated with the risk of heart failure and make it possible to identify not merely age-related changes but patients with a genuinely unfavorable prognosis [https://pubmed.ncbi.nlm.nih.gov/32741597/]. As a result, the diagnosis became considerably less common because the new criteria no longer classified a large number of older individuals as having the disease when they exhibited only age-related changes without evidence of increased risk.
This example clearly illustrates the fundamental difference between age and aging. Age-related changes may be very common, but by themselves they do not necessarily determine the probability of developing disease or dying. Aging, in contrast, should be defined by those changes that are accompanied by an increased risk of clinically significant outcomes. Therefore, any biomarkers and potential therapeutic interventions should be validated against the risks of heart failure, mortality, and other clinical outcomes rather than solely by their association with age. Otherwise, there is a risk of successfully treating age-related changes while having little or no impact on health, lifespan, or patient prognosis.
4. By Analyzing the Age-Related Dynamics of Diseases and Conditions Classified in the ICD, Practical Models for Managing Aging-Related Risks Can Be Developed
Analyzing the age-related dynamics of diseases according to the International Classification of Diseases (ICD) makes it possible to move from abstract discussions about aging to the practical management of lifespan. If it is known which diseases claim the greatest number of lives at different ages, it becomes possible to construct a model showing which causes of death impose the greatest limitations on maximum lifespan and which of them should be eliminated first.
For example, cardiovascular diseases are the leading cause of death worldwide [https://pubmed.ncbi.nlm.nih.gov/30099284/]. Our modeling indicates that their complete prevention could increase life expectancy to nearly 100 years. This means that the path toward radical life extension begins not with the search for yet another molecular marker, but with the elimination of the most important causes of age-related mortality.
For this reason, models based on mortality, disease incidence, and risk according to the International Classification of Diseases have direct practical significance. They make it possible to establish research priorities, estimate the potential benefit of each research direction, and identify which mechanisms truly limit human lifespan and which merely accompany aging while having little practical impact on prognosis.
This approach transforms the study of aging from the search for age-related changes into the search for the causes of premature death. These are precisely the causes that should become the primary targets of scientific research if the goal is not merely to explain aging, but to learn how to extend human life in a meaningful way.
5. Whether There Is a Single Biological Aging Process Remains an Open Scientific Question. At Present, There Is No Global Scientific Consensus on This Issue. Evolution May Shorten the Lifespan of Short-Lived Species Through a Small Number of Biological Mechanisms If This Increases Their Fitness and Reproductive Success. In Long-Lived Species, Lifespan May Be Determined Not by a Single Mechanism but by the Combined Effects of Many Biological Traits and Constraints.
Some researchers have proposed considering aging as a disease. However, within the framework of the Gompertz model, aging is not an isolated pathology but an integrated dynamic of increasing force of mortality that arises from the interaction of numerous age-associated diseases. Each of these diseases corresponds to a separate diagnostic entity in the International Classification of Diseases (ICD) and contributes its own age-dependent component to the overall risk of death. Since the overall force of mortality increases exponentially with age not because of a single cause but as a result of the cumulative effects of many pathological processes, aging cannot be reduced to a single diagnosable disease and does not fit within the existing nosological classification system.
Alan A. Cohen from the University of Sherbrooke challenges the concept of aging as a disease in his article [https://www.sciencedirect.com/science/article/pii/S0047637420301408]. He argues that the idea of a single entity called "aging" prevents researchers from asking the right scientific questions. Instead of saying "we treat" or "we slow aging," he suggests specifying exactly which disease or pathological process, as defined by the International Classification of Diseases (ICD), is being targeted. According to Cohen, aging is not a single biological process but merely a convenient term that groups together many different phenomena. He argues that the term "aging" may even hinder scientific progress because it creates the illusion of a single underlying mechanism that could be "fixed with a single pill."
At present, there is no convincing evidence for the existence of an independent disease called "aging" that directly causes death. The statement that a person "died of old age" is usually used when the immediate cause of death has not been established or when a comprehensive diagnostic evaluation was not performed. Autopsy studies, however, present a different picture. In nearly all cases, severe organ diseases or acute pathological processes that could have caused death are identified at autopsy. In other words, death is almost always associated with a specific pathological condition rather than with an abstract state of "old age." A representative example comes from a study of centenarians older than 100 years [https://pubmed.ncbi.nlm.nih.gov/16079208/], in which the authors challenge the concept of "healthy longevity." Approximately 60% of these individuals had been considered "healthy" during life by physicians or relatives. However, autopsy examination revealed potentially fatal diseases in every individual, and multiple chronic diseases were present in most of them. Cardiovascular diseases were the most common cause of death, whereas cancer was comparatively rare. The authors concluded that their autopsy findings do not support the concept of the "healthy centenarian."
Of course, the presence of a potentially fatal disease at autopsy does not by itself prove that it was the immediate cause of death. However, the opposite is equally true: the absence of a diagnosis during life does not mean that a person did not die from an age-related disease. It may simply indicate that the individual never sought medical attention for diagnosis or treatment. Therefore, the diagnosis of "death from old age" should primarily be regarded as evidence of uncertainty regarding the actual cause of death rather than as proof of the existence of an independent disease called "old age."
It is often argued that if the lifespan of short-lived animals can be dramatically extended by modifying a single gene, then humans probably possess a similar "master switch" controlling aging. However, evidence from evolutionary biology suggests that such a scenario is unlikely in humans.
A good example is the naked mole-rat. It lives approximately ten times longer than a mouse, but this remarkable longevity did not arise because of a single advantageous mutation. These animals have been evolving for approximately 73 million years. During this period, their genome underwent extensive changes: evolution eliminated approximately 320 genes, acquired about 750 new genes, and approximately 45 genes accumulated changes that have been associated with exceptional longevity. This indicates that the remarkable increase in lifespan observed in long-lived species results from the accumulation of numerous genetic changes rather than from the appearance of a single "longevity gene" [https://pubmed.ncbi.nlm.nih.gov/21993625/].
The situation is very different in short-lived model organisms. In the nematode Caenorhabditis elegans, modification of a single gene, age-1, increased lifespan by almost ninefold [https://academic.oup.com/genetics/article/204/3/905/6066287]. However, a subsequent experiment revealed an interesting pattern. When these long-lived mutants were placed together with ordinary short-lived worms, the short-lived worms displaced the long-lived mutants after several generations. Their faster rate of reproduction provided an evolutionary advantage. This demonstrates that, under certain conditions, evolution can indeed favor a short lifespan if doing so enhances the survival and reproductive success of the species.
6. There Are Many Theories and Proposed Mechanisms of Aging. Only Some of Them Are Supported by Robust Clinical Evidence and Become the Basis of Clinical Guidelines and Evidence-Based Medical Practice.
At present, there are many interventions that may or may not extend lifespan. This uncertainty arises because research regulations permit the publication of studies with statistically underpowered outcomes resulting from insufficient sample sizes. Statistical power is the probability of detecting a true effect if it actually exists. For example, twin studies failed to detect an association between physical activity and longevity when the sample consisted of 180 twin pairs [https://pubmed.ncbi.nlm.nih.gov/26666586/], but the association became detectable when the sample included 5,240 twin pairs [https://pubmed.ncbi.nlm.nih.gov/17493950/].
Statistically underpowered studies are not only unable to detect true differences in longevity with high reliability when such differences exist, but they also generate false discoveries that confuse science by giving rise to incorrect theories of aging. At p ≈ 0.05, the proportion of false discoveries may reach approximately 50–70% when statistical power is only 20–30% [John P. A. Ioannidis, University College London] [https://pubmed.ncbi.nlm.nih.gov/26064558/] [https://stats.stackexchange.com/questions/176384/do-underpowered-studies-have-increased-likelihood-of-false-positives]. In preclinical research, despite recommendations in international guidance documents, sample size calculations required to achieve adequate statistical power are performed only rarely [https://pubmed.ncbi.nlm.nih.gov/21956292/] [Malcolm Macleod, University of Edinburgh].
In exploratory fields characterized by multiple testing and small effect sizes, such as the molecular biology of aging, up to approximately 50–80% of statistically significant associations may fail to represent causal mechanisms and instead result from low statistical power and methodological bias [Andrew D. Higginson, University of Exeter] [https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.0020124] [https://pubmed.ncbi.nlm.nih.gov/27832072/]. Biology requires systematic re-evaluation. Today, up to 75% of all preclinical studies fail to be reproduced [https://nplus1.ru/material/2021/12/10/project-reproducibility-cancer-biology] [https://pubmed.ncbi.nlm.nih.gov/27225100] [https://techxplore.com/news/2022-04-robot-scientist-eve-one-third-scientific.html]. Because of the limited reliability of preclinical findings, many diagrams of aging mechanisms lose much of their scientific value. On the basis of such diagrams, it becomes impossible to determine whether what is being observed represents a genuine disease mechanism or simply an error, because the proportion of possible errors in signaling pathway connections (9.5%) is comparable to the proportion (<9.5%) of changes observed in disease [https://arxiv.org/abs/1411.7919] [Jing Ma, University of Michigan]. Such signaling and metabolic pathway diagrams can therefore generate false discoveries rather than identify the true causes of disease [https://arxiv.org/abs/2201.05593] [Sarah Mubeen, University of Bonn]. Classical diagrams of molecular aging pathways are useful for visualization, but they often become misleading when used to draw practical conclusions about disease. The pathways shown in these diagrams appear isolated, whereas in reality they are connected through numerous cross-interactions [https://arxiv.org/abs/0706.0194] [Mark B. Gerstein, Yale University].
Perhaps partly for this reason, only 8.7% of candidate drugs for cardiovascular disease identified through such preclinical research ultimately succeed when they enter clinical trials [https://pubmed.ncbi.nlm.nih.gov/20130567/]. Among novel cardiovascular compounds that entered Phase I clinical trials between 1993 and 2004, only approximately 3% (4 out of 134 compounds) had received regulatory approval by the time of analysis in June 2009.
The rapid development of artificial intelligence (AI) has introduced another challenge. AI systems are increasingly generating fabricated references in scientific manuscripts. This has already resulted in tens of thousands of publications containing false citations, and their number is growing at an accelerating rate, creating the risk of further worsening the reproducibility crisis in science [https://www.nature.com/articles/d41586-026-00969-z]. ChatGPT-5.0 also frequently cites retracted scientific articles [https://link.springer.com/article/10.1007/s11192-025-05484-y]. "Garbage in, garbage out." Today, the garbage may already exist within an apparently legitimate scientific publication complete with a DOI, journal, authors, and references.
Researchers from the The University of Texas decided to test whether AI models become less capable when they are fed the same kind of information overload that humans encounter—viral tweets, online hype, and meaningless social media posts. They do. They referred to this phenomenon as AI digital dementia [https://arxiv.org/pdf/2510.13928]. Potentially, a similar form of digital dementia may also arise when neural networks are trained on low-quality or noisy data.
When people perceive reality, they do not see it exactly as it is but only partially. The missing information is reconstructed by the brain on the basis of previous experience. For example, the human retina contains a blind spot, yet we do not notice it because the brain fills in the missing image [https://www.sciencedirect.com/science/article/pii/S0042698906003932]. Like neural networks, scientists who lack proper scientific information hygiene may encounter similar problems. They do not perceive reality itself but rather a version of reality reconstructed by the brain using prior knowledge—including a large proportion of false discoveries—combined with a relatively small amount of genuinely new information. When a scientist reads scientific papers, only part of the perceived information is truly new; the remainder is also reconstructed by the brain.
Every year, media headlines promise breakthroughs, cures, and revolutions. Time passes, and most of these promises collapse. But why would brilliant and highly intelligent people "systematically mislead the world"? This apparent paradox is precisely the point. There are no villains. The system simply rewards sensational findings rather than truth. A scientist's career depends heavily on the number of publications, making it more advantageous to conduct many small, statistically unreliable studies than a few large, statistically robust ones. Small but sensational studies almost always have low statistical power. According to the model developed by Andrew D. Higginson from the University of Exeter, the level of statistical power that is most advantageous for academic career advancement in such studies is only 10–40% [https://pubmed.ncbi.nlm.nih.gov/27832072/]. At this level of statistical power, a substantial proportion of the reported conclusions are likely to be false (Figure).
Many people argue that insufficient funding is allocated to research on aging and life extension. The evidence presented above suggests that the problem may not be a lack of funding but rather a system that hinders the achievement of genuine scientific progress. Approximately 18.3% of the U.S. gross domestic product is spent on healthcare. In addition, substantial financial resources are invested each year in biomedical research, including aging research. The more funding is directed toward career-advantageous studies characterized by low statistical power, multiple testing, and a high risk of bias, the more resources are spent on findings that are highly likely to be false. According to research by John P. A. Ioannidis from Stanford University, a large proportion of scientific studies lack sufficient reliability, and the probability that their conclusions are actually correct is low [https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.0020124].
If a substantial proportion of the fundamental biology of aging suffers from poor reproducibility, the question arises: where should we look for more reliable mechanisms of human aging? One answer already exists in clinical medicine. Modern guidelines on dyslipidemia [https://www.ahajournals.org/doi/10.1161/CIR.0000000000001423] are not merely protocols for lowering cholesterol. They constitute a map of the molecular targets of vascular aging because each target is linked not to an abstract age-related change but to the risk of myocardial infarction, stroke, and death.
The central idea is simple. Atherosclerosis develops over decades, and its molecular basis is the long-term exposure of the arterial wall to atherogenic lipoproteins. Therefore, medicine does not target "aging" in general but rather the specific molecules that create lifelong vascular exposure to damaging particles.
The first target is HMG-CoA reductase. This is the enzyme responsible for cholesterol synthesis in the liver. It is inhibited by statins. When cholesterol synthesis is reduced, the liver removes low-density lipoproteins from the bloodstream more efficiently, thereby reducing the exposure of blood vessels to atherogenic particles.
The second target is NPC1L1. This is the protein responsible for intestinal cholesterol absorption. It is inhibited by ezetimibe. The purpose of targeting this pathway is to reduce intestinal cholesterol absorption and further decrease the concentration of atherogenic lipoproteins.
The third target is PCSK9. This protein promotes the degradation of low-density lipoprotein receptors in the liver. The more active PCSK9 is, the fewer receptors remain on the surface of hepatocytes, and the less efficiently atherogenic particles are removed from the circulation. Modern medicine targets this pathway in two ways. The first involves monoclonal antibodies against PCSK9, such as evolocumab and alirocumab. The second uses small interfering RNA, such as inclisiran, which reduces PCSK9 synthesis at the messenger RNA level. This is no longer "general prevention" but a highly targeted molecular intervention.
The fourth target is ATP-citrate lyase. This enzyme acts upstream of HMG-CoA reductase in the cholesterol biosynthesis pathway. It is inhibited by bempedoic acid. This provides another strategy for reducing hepatic cholesterol production and lowering the burden of atherogenic lipoproteins.
The fifth target is apolipoprotein B. This is the principal structural protein of atherogenic lipoprotein particles. Any particle containing apolipoprotein B has the potential to contribute to the development of atherosclerosis. Therefore, apolipoprotein B is not merely a laboratory marker but an integrated molecular target representing the total burden of atherogenic lipoproteins.
The sixth target is lipoprotein(a). This genetically determined atherogenic lipoprotein is associated with the risk of atherosclerotic cardiovascular disease and calcific aortic valve stenosis. Its clinical importance lies in the fact that it may remain elevated even when low-density lipoprotein cholesterol levels are well controlled. Consequently, lipoprotein(a) represents an independent target of vascular aging, and RNA-based therapies designed to reduce apolipoprotein(a) production are currently under development.
The seventh target is ANGPTL3 (angiopoietin-like protein 3). It is inhibited by evinacumab. This target is particularly important in severe inherited lipid disorders because suppression of ANGPTL3 enhances hepatic clearance of very-low-density lipoprotein remnants and intermediate-density lipoproteins, thereby reducing atherogenic particles partly independently of the low-density lipoprotein receptor.
The eighth target is the microsomal triglyceride transfer protein. It is inhibited by lomitapide. This protein is required for the assembly of apolipoprotein B-containing lipoproteins, including very-low-density lipoproteins, low-density lipoproteins, and chylomicrons. Inhibition of this target reduces the production of atherogenic particles, although its clinical use is limited because of safety concerns.
The ninth target is apolipoprotein C-III. This protein inhibits the clearance of triglyceride-rich lipoproteins. The antisense oligonucleotide olezarsen binds to and degrades the messenger RNA encoding apolipoprotein C-III, thereby reducing its production and accelerating the clearance of triglycerides and very-low-density lipoproteins from the circulation.
The tenth target is vascular inflammation. In clinical guidelines, this is reflected by high-sensitivity C-reactive protein. Persistently elevated levels indicate an additional inflammatory component of cardiovascular risk. Thus, atherosclerosis is not merely a lipid disorder but also an inflammatory disease of the arterial wall.
Clinical medicine therefore already describes not a single molecular target of vascular aging but an entire system of interconnected targets. These include cholesterol synthesis, cholesterol absorption, degradation of low-density lipoprotein receptors, production of apolipoprotein B-containing lipoproteins, lipoprotein(a), triglyceride-rich lipoproteins, ANGPTL3, apolipoprotein C-III, and vascular inflammation. This represents a genuine clinical theory of vascular aging: a molecular target, a therapeutic intervention directed against that target, validation in humans, and evaluation of its effects on the risks of myocardial infarction, stroke, and death.
Another example is provided by the clinical guidelines on cardiovascular prevention [https://old.scardio.ru/content/Guidelines/Kardiovascular_profilaktika_2022.pdf]. These guidelines extend far beyond lipid metabolism and describe other evidence-based mechanisms of vascular aging, including endothelial dysfunction, chronic hyperactivation of the renin–angiotensin–aldosterone system, abnormalities of the sodium–glucose cotransporter 2 signaling pathway, thrombosis, sympathetic overactivation, chronic vascular inflammation, and residual cardiovascular risk. For most of these molecular and physiological targets, effective therapies already exist, with benefits confirmed by reductions in the risks of myocardial infarction, stroke, and cardiovascular mortality. If aging is viewed as the combined effect of age-associated diseases leading to an exponential increase in the force of mortality, then modern clinical guidelines effectively constitute a clinical theory of aging, presented not in a single document but as a system of evidence-based recommendations for individual diseases.
7. If Mathematical Survival Models Are Constructed for Individual Causes of Death Classified According to the ICD, and These Survival Curves Are Then Adjusted by the Magnitude of Mortality Reduction Already Demonstrated for the Corresponding Diseases and Risk Factors by Modern Medicine, the Combined Contribution of Individual Causes of Death to Overall Mortality Will Decrease. As a Result, the Model Will Predict an Increase in Population Life Expectancy. As New Effective Medical Interventions Become Available and Mortality from Diseases Classified in the ICD Continues to Decline, the Predicted Life Expectancy in Such a Model Will Increase Accordingly.
If, for each cause of death in the International Classification of Diseases (ICD), one uses not theoretical assumptions but evidence-based estimates of mortality reduction derived from clinical guidelines and incorporates them into a single population model, it becomes possible to estimate the projected median life expectancy under the comprehensive application of existing medical knowledge. Such a model does not answer the question of which treatment should be prescribed to a particular individual. Instead, it addresses a much more important question: what level of longevity could be achieved by a population if the full range of evidence-based preventive and therapeutic measures available through modern medicine and public health systems were implemented systematically?
In the Nestarenie Global project, we developed exactly such a model. For each disease category, estimates of mortality reduction were derived from clinical guidelines and studies with clinical endpoints. These effects were then integrated into a unified population mortality model. Under the assumptions adopted in the model, the simulations indicated that comprehensive implementation of the evidence-based interventions already available today could increase median life expectancy to approximately 100 years or beyond. This suggests that a substantial extension of human lifespan may be achieved not only through the discovery of entirely new technologies but also through the most complete implementation of existing evidence-based medical approaches across the entire population. It should be recognized that the results of the model depend on assumptions regarding interactions among risk factors, competing causes of death, and the completeness with which clinical effects are incorporated. Depending on these assumptions, projected life expectancy may be either overestimated or underestimated.
The principal value of such a model does not lie in producing a universal list of medications but in creating a quantitative protocol for longevity. It identifies which parameters should be brought to their optimal values—lipid profile, blood pressure, body weight, physical activity, prevention of infectious diseases, prevention of injuries, and other evidence-based risk factors—while the specific methods for achieving these goals are already described in detail in the corresponding clinical guidelines.
This is precisely the approach that we plan to implement by the end of 2026 in the Nestarenie.expert project. Beginning in the autumn, users will be able to model different longevity scenarios themselves by modifying individual risk factors through interactive tools and observing how their combined optimization affects projected life expectancy. Our goal is to transform fragmented clinical guidelines into a single quantitative model for managing longevity that is accessible to everyone.
8. The Age-Related Increase in the Force of Mortality and the Force of Disease Incidence Is Usually Preceded by a Gradual Decline in the Biological Functions of Organs and Physiological Systems. It Is on the Basis of These Functional Impairments That Diagnoses Are Subsequently Established According to the ICD. Therefore, Aging-Related Processes Can Be Monitored and Corrected Before Disease Develops by Targeting Measurable Functional Parameters in Real Time Rather Than Waiting for Clinical Outcomes or the Establishment of a Diagnosis.
The management of aging begins not with molecules but with function, because it is the impairment of organ and physiological system function that directly leads to disease, disability, and death.
The increase in the force of mortality and the force of incidence of age-related diseases is usually preceded by a gradual deterioration in the function of organs and physiological systems. It is this functional impairment that forms the basis for the diagnosis of many diseases classified in the International Classification of Diseases (ICD). Consequently, aging-related processes can be detected and corrected before disease develops, without waiting until diagnostic thresholds are crossed.
This approach is gradually becoming a new direction in modern gerontology. Whereas previous research focused primarily on identifying the molecular mechanisms of aging, an increasing number of investigators now propose shifting the emphasis toward measuring physiological function and the body's capacity to maintain performance and recover from stress. This does not mean that the molecular mechanisms of aging are no longer important. They remain a central subject of basic biological research. However, for the clinical management of lifespan, the crucial question is no longer which molecule has changed, but whether that change has resulted in a measurable deterioration of physiological function and an increased risk of disease or death.
A representative example is the editorial by Felipe Sierra and Viviana Perez, published in GeroScience in 2025 [https://link.springer.com/article/10.1007/s11357-025-01916-y]. The authors argue that the concept of the "Hallmarks of Aging," although it has played a major role in the development of gerontology, is gradually becoming insufficient for the practical management of aging. As scientific knowledge expands, the number of proposed molecular mechanisms continues to grow, increasing interactions are being identified among them, and attempts to define a single primary mechanism of aging are becoming progressively less realistic. As a consequence, the list of hallmarks may continue to expand almost indefinitely without bringing us closer to the ability to manage lifespan quantitatively.
Instead, the authors propose focusing on the common outcome of all these molecular changes—the gradual loss of organismal resilience. It is the declining ability of tissues and organs to maintain function and recover from stress that precedes the development of age-related diseases. Therefore, the primary object of observation should not be individual molecular processes but the physiological functions that determine human health.
This perspective is highly consistent with clinical medicine. Physicians do not diagnose disease on the basis of the activity of a signaling pathway or the number of senescent cells. A diagnosis is established when the function of an organ becomes impaired. Heart failure is diagnosed after deterioration of systolic or diastolic cardiac function, chronic kidney disease after a reduction in the glomerular filtration rate, chronic obstructive pulmonary disease after a decrease in forced expiratory volume, and osteoporosis after a reduction in bone mineral density. In every case, functional measurements constitute the basis for clinical decision-making.
Consequently, if function can be measured continuously, it becomes possible to manage aging before disease develops. For example, deterioration of diastolic cardiac function begins long before the onset of heart failure. Modern echocardiography allows these changes to be monitored in real time, while physical exercise, blood pressure control, weight reduction, and other interventions can improve diastolic function before a clinical diagnosis is established. Thus, the target of preventive intervention becomes the functional impairment that precedes disease.
This principle is universal. Most age-related diseases are preceded by a gradual decline in the function of the corresponding organ or physiological system. Therefore, the most rational strategy for managing aging is continuous monitoring of physiological function followed by corrective interventions before diagnostic thresholds are reached.
For this reason, the future of medicine will likely be associated not with an endless search for new molecular hallmarks of aging but with the development of quantitative models capable of continuously measuring organ function, predicting its future trajectory, estimating an individual's disease risk, and enabling timely intervention. This approach directly links the biology of aging with clinical medicine and provides the foundation for the practical management of healthy lifespan.
The principle of managing aging through the restoration of physiological function has already been implemented in modern clinical medicine. The goal of such interventions is not to target a hypothetical molecular mechanism of aging but to preserve or restore organ function before the diagnostic threshold for disease is reached.
One of the most illustrative examples is the American randomized controlled trial Reversing the Cardiac Effects of Sedentary Aging in Middle Age [https://pubmed.ncbi.nlm.nih.gov/29311053/], which investigated whether age-related deterioration of diastolic cardiac function could be reversed in healthy middle-aged adults. The authors demonstrated that a two-year program of regular physical exercise significantly increased left ventricular compliance and reduced age-related ventricular stiffness. In other words, the intervention was directed not at an individual molecular mechanism of aging but directly at the age-related decline in cardiac function, and this function was indeed improved, corresponding to an average reversal of cardiac functional age by approximately 23 years toward a younger physiological state.
This example illustrates a fundamentally different approach to rejuvenation. The object of intervention is not the hypothetical primary molecular causes of aging, which may differ among individuals and whose number is likely to be virtually unlimited, but rather the measurable functions of organs and physiological systems that determine the risks of disease and death. If function deteriorates, it can be detected, quantified, and in many cases improved before a clinical diagnosis is established.
For this reason, the future of medicine will likely depend not on the continual expansion of the list of molecular hallmarks of aging but on continuous monitoring of organ function, prediction of age-related functional trajectories using mathematical models and machine learning methods, and timely correction of identified functional impairments. This approach integrates the fundamental biology of aging with evidence-based clinical medicine and shifts gerontology from the search for hypothetical molecular causes toward the practical management of human health.
What is aging? At first glance, the answer seems to have been known for a long time. However, over the past several decades, the biology of aging has proposed dozens of different theories and hundreds of presumed mechanisms, without coming closer to a unified understanding of what aging actually is and how it can be managed. This article offers a different view of the problem. Instead of searching for an endless number of molecular causes, it considers what can be objectively measured, verified, and used in clinical practice: the age-related increase in mortality, the age-related incidence of diseases, and the gradual decline in organ function. This approach makes it possible to move from theoretical discussions about the mechanisms of aging to a practical strategy for managing human health and lifespan.
If Age-Related Changes Did Not Increase the Risk of Death and Serious Diseases, Aging Would Not Be a Major Medical Problem
Aging is not gray hair, wrinkles, or the age listed in a passport. If age-related changes did not increase the risk of death and serious diseases, aging would not be a major medical problem. Gray hair by itself does not prevent people from enjoying life. The real tragedy of aging lies elsewhere. With each passing year, the probability of myocardial infarction, stroke, cancer, dementia, loss of independence, and death increases at an ever faster rate. This is why modern medicine regards aging primarily as the greatest risk factor for most chronic diseases [https://www.nature.com/articles/s41467-021-23014-1].
The most alarming aspect is that this risk does not increase linearly but almost exponentially. In adults, the probability of death approximately doubles every 8–11 years (according to different estimates)—a pattern known as the Gompertz law, which has been repeatedly confirmed in large demographic datasets [https://pubmed.ncbi.nlm.nih.gov/30729179/] [https://biobank.ndph.ox.ac.uk/ukb/pub.cgi?id=1828]. Similarly, the accumulation of chronic diseases and functional impairments accelerates with age. Therefore, the main problem of aging is not changes in appearance but the increasingly rapid rise in the probability of severe disease, disability, and death [https://pubmed.ncbi.nlm.nih.gov/38622100/]. This is precisely what makes combating the biological causes of aging one of the most important challenges of modern medicine.
All these facts lead to the conclusion that any theory of aging must mathematically explain the actual age-related patterns of mortality. According to aging researchers Leonid Gavrilov and Natalia Gavrilova [https://pubmed.ncbi.nlm.nih.gov/38622100/], the ability to reproduce the laws of mortality is an important criterion for the adequacy of a model of aging.
2. Since the Main Medical Problem of Aging Is the Age-Related Increase in Mortality and Disease, These Phenomena Must Be Described Quantitatively and Mathematically
In the previous section, we established that any theory of aging must mathematically explain the actual age-related patterns of mortality. Mathematically, aging is a process or a set of processes that, with increasing age, lead to an increase in the probability of dying at each subsequent moment in time [https://pubmed.ncbi.nlm.nih.gov/25750242/] [https://pubmed.ncbi.nlm.nih.gov/9074828/], or, in other words, to an increase in the force of mortality. R(t) = A × e^(G×t), where R is the force of mortality, t is age, A is the baseline force of mortality, and G is the rate of increase in the force of mortality. In aging, G is always greater than zero.
Non-aging is a state in which the force of mortality does not increase with age. Mathematically, this means that in the equation R(t) = A × e^(G×t), the rate of increase in the force of mortality, G, is always equal to zero. In this case, the force of mortality R(t) remains constant and equal to A throughout life. It is important to understand that even under non-aging, the probability of dying still increases with age because the cumulative probability of death continues to rise. The more time passes, the greater the chance that a random fatal event will eventually occur, but the instantaneous probability of dying at any given moment remains unchanged.
In the Gompertz equation, the baseline force of mortality, A, reflects the overall level of age-related risks at the initial stage, whereas G describes the "rate of aging," that is, the rate of exponential increase in these risks with age. If we eliminate one of the causes of aging that has the same rate of risk increase (the same G) as the other causes, we primarily reduce A—the overall initial burden of age-related causes—but we do not substantially affect G, because the remaining causes continue to increase at nearly the same rate. Thus, by eliminating one of the causes of aging, we may extend lifespan without changing the rate of aging itself.
But where does mortality in young people come from if accidents and infectious diseases are excluded? Does the absence of an age-related increase in mortality imply an almost endless life? No. Even between the ages of 10 and 24 years, about 27% of all deaths worldwide are caused by non-communicable diseases, including cardiovascular diseases, cancer, diabetes, and other chronic disorders. Consequently, even in the absence of pronounced aging, humans retain a baseline risk of dying from disease. Aging does not create mortality from scratch; rather, it primarily multiplies an already existing risk [https://www.thelancet.com/journals/lancet/article/piis0140-6736%2821%2901546-4/fulltext].
Imagine two jars: one with a capacity of 2 liters and another with a capacity of 5 liters. We begin pouring water into both of them. Suppose that water flows quickly into one jar and slowly into the other. Intuitively, it may seem that the pouring speed is the most important factor because the faster-filling jar will be filled sooner. However, this is not always the case. If one jar is much larger, it may overflow later despite being filled more rapidly than a smaller jar into which water is poured slowly. What matters is not how quickly the water is poured but which jar overflows first. This is precisely the essence of the difference between the rate of aging and the baseline force of mortality when described in the language of the science of life and death.
In the Gompertz law, which describes how the risk of death changes with age, there are two main parameters. The first is the baseline force of mortality—the initial level of vulnerability of the organism. This is analogous to the size of the jar—the initial reserve of resilience. The second parameter is the rate of increase in the force of mortality, that is, the rate of aging. This is analogous to the speed at which water is poured into the jar. We often focus primarily on this speed, assuming that slower aging necessarily means a longer life. However, this is not always true.
A person with high baseline vulnerability (a small jar) may live a shorter life even if they age slowly. Conversely, a person with low initial vulnerability (a large jar) may survive to an advanced age despite a relatively rapid rate of aging. What matters is not only how fast the process proceeds but also where it begins. Life is not only about speed—it is also about capacity.
This is why lifespan cannot be judged solely by the rate of aging. As in the example of the jars, the final outcome—when the water overflows or when a person dies—depends on the combination of two factors: the baseline force of mortality and the rate at which it increases. In reality, what matters is not who ages faster but whose life "overflows" first.
Let us consider an example of how it is possible to have even a "zero" rate of aging while still dying much sooner than under aging. In naked mole-rats, the baseline force of mortality, A, is already as high in early life as it is in 72-year-old humans. This means that their life expectancy is comparable to that of 72-year-old humans—that is, half of them will die within the next 27 years. If humans stopped aging at the age of 72 years, they still would not live much longer because their force of mortality would already be high. In naked mole-rats, either the increase in mortality from age-related diseases is masked by very high mortality due to fighting, or, even if they truly do not age, they still cannot live very long because of their high baseline force of mortality according to the Gompertz model. Aging increases the probability of dying at each subsequent moment in time, whereas non-aging merely means that the force of mortality remains constant, not that it disappears. In humans aged 35 years, the probability of dying within one year is approximately 0.0015, whereas in naked mole-rats this value remains approximately 0.0365 per year even after 35 years of life, corresponding to the mortality level of a 72-year-old human. If the force of mortality remains constant, the expected remaining lifespan of the mole-rat would be 1 / 0.0365 = 27 years, meaning that half of the 35-year-old individuals would die within the next 27 years, reaching an age of about 62 years. However, this conclusion is valid only if the force of mortality truly does not increase with age, which has not been definitively established. Therefore, our goal should not be to combat the rate of aging itself, but rather any process that shortens life [https://link.springer.com/article/10.1007/s11357-024-01201-4].
3. Mathematical Models Make It Possible to Classify Biological Processes According to Their Relationship with the Age-Related Increase in Mortality and Disease. They Do Not Explain the Mechanisms of Aging but Allow Their Clinical Significance to Be Evaluated.
It is important to distinguish between the rate of aging, the mechanisms of aging, and the causes of aging. The mechanisms and causes of aging belong to the field of biology, whereas the rate of aging is the mathematical phenomenon of the age-related increase in the force of mortality. A mathematical phenomenon cannot explain the causes or mechanisms of aging, but it makes it possible to determine whether what we observe in biology can truly be considered aging.
If aging is defined as a set of biological processes, there is a risk of error because these processes alone, without a mathematical criterion, do not allow us to determine rigorously whether they actually lead to an age-related increase in the force of mortality.
In contrast, defining aging as the age-related increase in the force of mortality itself provides a precise and reproducible criterion. Once this criterion is established, it becomes possible to investigate which specific biological mechanisms generate this observable mathematical process.
The aging process is a biological phenomenon associated with a set of changes in the organism that lead to the progressive deterioration of function and an increased risk of mortality. The Gompertz law describes the rate at which mortality increases with age but does not explain the underlying biological processes. It is based on the observation that, with age, the risk of death from certain causes increases exponentially. To determine which biological processes are associated with aging, the following principle can be applied: a process is considered to be associated with aging if it leads to an increasing force of mortality that, in the foreseeable future, would become a cause of death even in the absence of other age-related changes. However, not all causes of mortality satisfy this criterion. For example, malignant neoplasms of the cervix do not exhibit an exponential age-related increase in the force of mortality and therefore cannot be classified as manifestations of aging. Thus, mathematical methods such as Gompertz analysis are necessary for classifying processes as aging-related or non-aging-related. They do not describe biological mechanisms but make it possible to evaluate the impact of biological processes on mortality. Without quantitative methods, we cannot objectively distinguish between different processes, making measurement essential for understanding aging.
The rate of aging, as a mathematical phenomenon, represents a fundamental concept that has been applied since 1975 and continues to be used today. The calculation of MRDT (Mortality Rate Doubling Time) for estimating the rate of aging on the basis of the Gompertz law was first proposed in 1975 by the American epidemiologist Richard G. Peto. To this day, many leading gerontologists worldwide use MRDT and the Gompertz function to assess the rate of aging [https://academic.oup.com/genetics/article/204/3/905/6066287] [https://academic.oup.com/genetics/article/208/4/1617/6084259] [https://pubmed.ncbi.nlm.nih.gov/34151374]. The rate of aging reflects the increase in the probability of death at each successive moment in time as age increases, accompanied by an increase in the force of mortality.
Regardless of the method of measurement, whether the Weibull distribution or the Gompertz law is used, the same phenomenon is observed: aging has a measurable rate, manifested by an increasing probability of death at successive moments in time as age advances.
The Gompertz law itself, which describes the exponential increase in the overall force of mortality with age, does not reveal the causal mechanisms of aging. Overall mortality represents the sum of mortality from many different causes and therefore provides only limited information for the analysis of biological processes. A more informative approach is to analyze the age-related dynamics of the force of mortality separately for different groups of diseases, or, even better, for each individual disease classified according to the International Classification of Diseases (ICD). Such an analysis makes it possible to determine which diseases truly exhibit the exponential age-related increase in risk that is characteristic of aging [https://pubmed.ncbi.nlm.nih.gov/30099284/].
An even more informative approach is to analyze not the force of mortality but the force of incidence of each disease. Death is the final outcome and can occur only once in each individual, whereas age-related diseases can accumulate simultaneously. Therefore, analyzing the age-related dynamics of disease incidence provides a more accurate quantitative description of aging than analyzing overall mortality. If the force of incidence of a disease increases according to the same pattern and at the same rate [https://pubmed.ncbi.nlm.nih.gov/30729179/] as the force of mortality from that disease, this provides additional evidence that the disease should be regarded as a manifestation of the aging process.
However, for the practical monitoring of aging, waiting until a disease develops is already too late. Every clinical diagnosis is preceded by a gradual decline in the function of an organ or physiological system. For example, the diagnosis of diastolic dysfunction is established only after diastolic cardiac function has fallen below the diagnostic threshold, and the diagnosis of chronic kidney disease is established only after the glomerular filtration rate has declined to the corresponding threshold. Therefore, continuous measurement of functional parameters makes it possible to monitor age-related changes long before clinical disease incidence occurs. These functional parameters are not arbitrary biomarkers. They have been validated in clinical medicine as predictors of future disease incidence and mortality and form the basis of modern diagnostic criteria (serving as the "gold standards"), making them among the most reliable tools for assessing the risk of age-related diseases.
Thus, the practical application of the Gompertz law does not lie in analyzing overall mortality as such, but in using age-related patterns of mortality and disease incidence to identify age-related diseases, after which it becomes possible to monitor an individual's aging trajectory through changes in the functions of organs and physiological systems. The accuracy of this approach could potentially be further improved by machine learning methods that estimate individual rates of age-related change and predict when functional parameters will cross diagnostic thresholds.
It is important to distinguish between age and aging. Age itself is not evidence of aging. Any age-related change can be regarded as a component of aging only if it has been clinically demonstrated to be associated with an increased risk of disease or death and has sufficient predictive value for these outcomes.
Researchers often overlook the very necessity of validating age-related changes against clinical risk. In many cases, they do not test whether the observed age-related changes are associated with clinically meaningful outcomes, such as increased mortality, disease incidence, or functional decline. This is particularly critical when developing artificial intelligence models trained exclusively on data from individuals classified as Health Group I. If such models are not longitudinal and do not use long-term follow-up of the same individuals, they largely lose the signal of aging. The main reason is that the proportion of practically healthy individuals declines rapidly with age: among people aged 21–36 years, it may be approximately 60%; among those aged 39–60 years, only 18–22%; and after the age of 60 years, only 2–3% [http://vestnik.mednet.ru/content/view/1434/30/lang,ru]. This creates a pronounced class imbalance. As a result, while minimizing the average training error, the algorithm systematically shifts its predictions toward younger ages, underestimating the biological age of older participants and overestimating it in younger ones. This reduces the model's ability to correctly identify age-related changes that are truly associated with aging rather than with the characteristics of the training dataset.
Thus, if a model is trained only on a dataset consisting of individuals from Health Group I, a pronounced age imbalance arises. As a result, the model "rejuvenates" older individuals and "ages" younger ones. Aging is not a disease, but diseases make it visible.
If a biological age model is trained on a dataset composed predominantly of individuals from Health Group I, it typically reproduces not the true relationship of "predicted age = chronological age" but a shallower trend. As a result, the model systematically overestimates age in younger individuals and systematically underestimates it in older individuals. In other words, the mean predicted age no longer matches the mean chronological age, and the slope of the regression line becomes less than one. With proper model calibration, the mean predicted age should coincide with the mean chronological age, while prediction errors should fluctuate randomly around this line without any systematic bias.
Such systematic bias creates a serious problem when evaluating potential geroprotectors. If biological age is measured with such a model before and after an intervention, chronological age increases over time faster than the age predicted by the model. Consequently, even in the complete absence of any effect of the intervention, it may appear that biological age is increasing more slowly than chronological age. This can create the false impression that aging has been slowed.
Therefore, models with this type of age-related miscalibration cannot reliably evaluate geroprotective effects. Without first verifying the absence of systematic prediction error across the entire age range, there is a substantial risk of obtaining false-positive results, in which compounds are incorrectly classified as geroprotectors because of statistical bias in the model rather than because they genuinely slow the aging process.
Unfortunately, most biological age models worldwide are developed using artificially health-selected cohorts. Many authors have highlighted this important problem. For example, in 2024, Hongqian Qi from Nankai University (China) proposed the DeepQA model [https://pubmed.ncbi.nlm.nih.gov/39757434/], which was trained on a mixed dataset including both healthy and diseased individuals. In the same study, Hongqian Qi also clearly demonstrated that models developed using health-selected cohorts produce compressed regression lines (slope <1), resulting in "rejuvenating" bias.
Many researchers attempt to predict age-related changes and then try to treat them. To achieve this, they develop first-generation biological age clocks. Instead of predicting risk, first-generation clocks are trained to predict chronological age. However, not everything that changes with age is necessarily harmful or life-limiting [https://www.biorxiv.org/content/10.1101/2024.06.06.597715v1]. Fingerprints change with age, but this does not kill us. Plants develop more growth rings as they age. Telomere length also changes with age in humans, although only weakly, with a correlation of approximately 9%. However, in humans, telomere length does not significantly predict all-cause mortality (HR 0.882, p = 0.092) or cardiovascular mortality (HR 0.897, p = 0.150) [https://pubmed.ncbi.nlm.nih.gov/29920523/].
First-generation biological age clocks were originally trained solely to predict chronological age. Chronological age itself was their target variable. Consequently, these models were optimized exclusively to minimize the error of age prediction rather than to predict the risk of death, the rate of aging, or rejuvenation. This fundamentally distinguishes them from second-generation clocks. Second-generation clocks are trained on clinical outcomes such as mortality, disease incidence, or other measures of risk. Therefore, deviations of their predictions from the average age-related trend contain information about individual differences in risk. These deviations constitute the informative signal for such models [https://www.biorxiv.org/content/10.1101/2024.06.06.597715v1]. The situation is different for first-generation clocks. Any deviation between the predicted age and the chronological age is not something the model was specifically trained to detect. For these models, such deviations represent only the residual error of age prediction. Even if, in a particular individual, this prediction error happens to coincide with a true change in mortality risk, such a coincidence does not mean that the model has learned to measure risk. It was never trained to do so.
The main reason is that age and the force of mortality are different variables. According to the Gompertz law, the force of mortality generally increases with age. However, even if an intervention reduces the force of mortality, R(t), chronological age, t, continues to increase inexorably. Therefore, the correlation between age and mortality at the population level does not imply that a model trained to predict age automatically becomes a measure of changes in mortality risk. In other words, first-generation clocks can predict only the variable on which they were trained—chronological age. They were never trained to recognize individual deviations in risk around the age-related trend. Consequently, their prediction errors cannot automatically be interpreted as accelerated or slowed aging, rejuvenation, or reduced mortality risk. Such conclusions require models that have been specifically trained and validated using the corresponding clinical outcomes.
First-generation clocks are a clear example of predicting age (chronological age), but not aging (risk).
Let us imagine, hypothetically, that it became possible to reduce certain age-related changes in humans that are unrelated to risk. Telomeres might become longer, yet mortality risk would remain unchanged [https://pubmed.ncbi.nlm.nih.gov/29920523/]. Fingerprints might come to resemble those of a younger person, but this likewise would have no effect on lifespan. By the same logic, one could simply change the date of birth in a passport, since that value also changes with age. But would doing so increase longevity?
A striking example of the distinction between age-related changes and aging is provided by the evolution of the diagnostic criteria for left ventricular diastolic dysfunction. The older diagnostic criteria primarily reflected age-related changes in the heart's diastolic function. However, many of these changes had not been validated as predictors of heart failure or mortality. As a result, the prevalence of diastolic dysfunction according to the old criteria increased dramatically with age and reached very high levels in the majority of older adults [https://jamanetwork.com/journals/jama/fullarticle/195749]. This indicated the accumulation of age-related changes rather than the identification of processes that determine clinical prognosis.
Therefore, even if an intervention improved such age-related parameters, this alone did not necessarily mean that it reduced the risk of heart failure or improved survival. In this case, treating age-related changes did not lead to an improvement in clinical prognosis.
For this reason, new diagnostic criteria for diastolic dysfunction were proposed in 2016, based primarily on markers of elevated left ventricular filling pressure. These criteria are much more strongly associated with the risk of heart failure and make it possible to identify not merely age-related changes but patients with a genuinely unfavorable prognosis [https://pubmed.ncbi.nlm.nih.gov/32741597/]. As a result, the diagnosis became considerably less common because the new criteria no longer classified a large number of older individuals as having the disease when they exhibited only age-related changes without evidence of increased risk.
This example clearly illustrates the fundamental difference between age and aging. Age-related changes may be very common, but by themselves they do not necessarily determine the probability of developing disease or dying. Aging, in contrast, should be defined by those changes that are accompanied by an increased risk of clinically significant outcomes. Therefore, any biomarkers and potential therapeutic interventions should be validated against the risks of heart failure, mortality, and other clinical outcomes rather than solely by their association with age. Otherwise, there is a risk of successfully treating age-related changes while having little or no impact on health, lifespan, or patient prognosis.
4. By Analyzing the Age-Related Dynamics of Diseases and Conditions Classified in the ICD, Practical Models for Managing Aging-Related Risks Can Be Developed
Analyzing the age-related dynamics of diseases according to the International Classification of Diseases (ICD) makes it possible to move from abstract discussions about aging to the practical management of lifespan. If it is known which diseases claim the greatest number of lives at different ages, it becomes possible to construct a model showing which causes of death impose the greatest limitations on maximum lifespan and which of them should be eliminated first.
For example, cardiovascular diseases are the leading cause of death worldwide [https://pubmed.ncbi.nlm.nih.gov/30099284/]. Our modeling indicates that their complete prevention could increase life expectancy to nearly 100 years. This means that the path toward radical life extension begins not with the search for yet another molecular marker, but with the elimination of the most important causes of age-related mortality.
For this reason, models based on mortality, disease incidence, and risk according to the International Classification of Diseases have direct practical significance. They make it possible to establish research priorities, estimate the potential benefit of each research direction, and identify which mechanisms truly limit human lifespan and which merely accompany aging while having little practical impact on prognosis.
This approach transforms the study of aging from the search for age-related changes into the search for the causes of premature death. These are precisely the causes that should become the primary targets of scientific research if the goal is not merely to explain aging, but to learn how to extend human life in a meaningful way.
5. Whether There Is a Single Biological Aging Process Remains an Open Scientific Question. At Present, There Is No Global Scientific Consensus on This Issue. Evolution May Shorten the Lifespan of Short-Lived Species Through a Small Number of Biological Mechanisms If This Increases Their Fitness and Reproductive Success. In Long-Lived Species, Lifespan May Be Determined Not by a Single Mechanism but by the Combined Effects of Many Biological Traits and Constraints.
Some researchers have proposed considering aging as a disease. However, within the framework of the Gompertz model, aging is not an isolated pathology but an integrated dynamic of increasing force of mortality that arises from the interaction of numerous age-associated diseases. Each of these diseases corresponds to a separate diagnostic entity in the International Classification of Diseases (ICD) and contributes its own age-dependent component to the overall risk of death. Since the overall force of mortality increases exponentially with age not because of a single cause but as a result of the cumulative effects of many pathological processes, aging cannot be reduced to a single diagnosable disease and does not fit within the existing nosological classification system.
Alan A. Cohen from the University of Sherbrooke challenges the concept of aging as a disease in his article [https://www.sciencedirect.com/science/article/pii/S0047637420301408]. He argues that the idea of a single entity called "aging" prevents researchers from asking the right scientific questions. Instead of saying "we treat" or "we slow aging," he suggests specifying exactly which disease or pathological process, as defined by the International Classification of Diseases (ICD), is being targeted. According to Cohen, aging is not a single biological process but merely a convenient term that groups together many different phenomena. He argues that the term "aging" may even hinder scientific progress because it creates the illusion of a single underlying mechanism that could be "fixed with a single pill."
At present, there is no convincing evidence for the existence of an independent disease called "aging" that directly causes death. The statement that a person "died of old age" is usually used when the immediate cause of death has not been established or when a comprehensive diagnostic evaluation was not performed. Autopsy studies, however, present a different picture. In nearly all cases, severe organ diseases or acute pathological processes that could have caused death are identified at autopsy. In other words, death is almost always associated with a specific pathological condition rather than with an abstract state of "old age." A representative example comes from a study of centenarians older than 100 years [https://pubmed.ncbi.nlm.nih.gov/16079208/], in which the authors challenge the concept of "healthy longevity." Approximately 60% of these individuals had been considered "healthy" during life by physicians or relatives. However, autopsy examination revealed potentially fatal diseases in every individual, and multiple chronic diseases were present in most of them. Cardiovascular diseases were the most common cause of death, whereas cancer was comparatively rare. The authors concluded that their autopsy findings do not support the concept of the "healthy centenarian."
Of course, the presence of a potentially fatal disease at autopsy does not by itself prove that it was the immediate cause of death. However, the opposite is equally true: the absence of a diagnosis during life does not mean that a person did not die from an age-related disease. It may simply indicate that the individual never sought medical attention for diagnosis or treatment. Therefore, the diagnosis of "death from old age" should primarily be regarded as evidence of uncertainty regarding the actual cause of death rather than as proof of the existence of an independent disease called "old age."
It is often argued that if the lifespan of short-lived animals can be dramatically extended by modifying a single gene, then humans probably possess a similar "master switch" controlling aging. However, evidence from evolutionary biology suggests that such a scenario is unlikely in humans.
A good example is the naked mole-rat. It lives approximately ten times longer than a mouse, but this remarkable longevity did not arise because of a single advantageous mutation. These animals have been evolving for approximately 73 million years. During this period, their genome underwent extensive changes: evolution eliminated approximately 320 genes, acquired about 750 new genes, and approximately 45 genes accumulated changes that have been associated with exceptional longevity. This indicates that the remarkable increase in lifespan observed in long-lived species results from the accumulation of numerous genetic changes rather than from the appearance of a single "longevity gene" [https://pubmed.ncbi.nlm.nih.gov/21993625/].
The situation is very different in short-lived model organisms. In the nematode Caenorhabditis elegans, modification of a single gene, age-1, increased lifespan by almost ninefold [https://academic.oup.com/genetics/article/204/3/905/6066287]. However, a subsequent experiment revealed an interesting pattern. When these long-lived mutants were placed together with ordinary short-lived worms, the short-lived worms displaced the long-lived mutants after several generations. Their faster rate of reproduction provided an evolutionary advantage. This demonstrates that, under certain conditions, evolution can indeed favor a short lifespan if doing so enhances the survival and reproductive success of the species.
6. There Are Many Theories and Proposed Mechanisms of Aging. Only Some of Them Are Supported by Robust Clinical Evidence and Become the Basis of Clinical Guidelines and Evidence-Based Medical Practice.
At present, there are many interventions that may or may not extend lifespan. This uncertainty arises because research regulations permit the publication of studies with statistically underpowered outcomes resulting from insufficient sample sizes. Statistical power is the probability of detecting a true effect if it actually exists. For example, twin studies failed to detect an association between physical activity and longevity when the sample consisted of 180 twin pairs [https://pubmed.ncbi.nlm.nih.gov/26666586/], but the association became detectable when the sample included 5,240 twin pairs [https://pubmed.ncbi.nlm.nih.gov/17493950/].
Statistically underpowered studies are not only unable to detect true differences in longevity with high reliability when such differences exist, but they also generate false discoveries that confuse science by giving rise to incorrect theories of aging. At p ≈ 0.05, the proportion of false discoveries may reach approximately 50–70% when statistical power is only 20–30% [John P. A. Ioannidis, University College London] [https://pubmed.ncbi.nlm.nih.gov/26064558/] [https://stats.stackexchange.com/questions/176384/do-underpowered-studies-have-increased-likelihood-of-false-positives]. In preclinical research, despite recommendations in international guidance documents, sample size calculations required to achieve adequate statistical power are performed only rarely [https://pubmed.ncbi.nlm.nih.gov/21956292/] [Malcolm Macleod, University of Edinburgh].
In exploratory fields characterized by multiple testing and small effect sizes, such as the molecular biology of aging, up to approximately 50–80% of statistically significant associations may fail to represent causal mechanisms and instead result from low statistical power and methodological bias [Andrew D. Higginson, University of Exeter] [https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.0020124] [https://pubmed.ncbi.nlm.nih.gov/27832072/]. Biology requires systematic re-evaluation. Today, up to 75% of all preclinical studies fail to be reproduced [https://nplus1.ru/material/2021/12/10/project-reproducibility-cancer-biology] [https://pubmed.ncbi.nlm.nih.gov/27225100] [https://techxplore.com/news/2022-04-robot-scientist-eve-one-third-scientific.html]. Because of the limited reliability of preclinical findings, many diagrams of aging mechanisms lose much of their scientific value. On the basis of such diagrams, it becomes impossible to determine whether what is being observed represents a genuine disease mechanism or simply an error, because the proportion of possible errors in signaling pathway connections (9.5%) is comparable to the proportion (<9.5%) of changes observed in disease [https://arxiv.org/abs/1411.7919] [Jing Ma, University of Michigan]. Such signaling and metabolic pathway diagrams can therefore generate false discoveries rather than identify the true causes of disease [https://arxiv.org/abs/2201.05593] [Sarah Mubeen, University of Bonn]. Classical diagrams of molecular aging pathways are useful for visualization, but they often become misleading when used to draw practical conclusions about disease. The pathways shown in these diagrams appear isolated, whereas in reality they are connected through numerous cross-interactions [https://arxiv.org/abs/0706.0194] [Mark B. Gerstein, Yale University].
Perhaps partly for this reason, only 8.7% of candidate drugs for cardiovascular disease identified through such preclinical research ultimately succeed when they enter clinical trials [https://pubmed.ncbi.nlm.nih.gov/20130567/]. Among novel cardiovascular compounds that entered Phase I clinical trials between 1993 and 2004, only approximately 3% (4 out of 134 compounds) had received regulatory approval by the time of analysis in June 2009.
The rapid development of artificial intelligence (AI) has introduced another challenge. AI systems are increasingly generating fabricated references in scientific manuscripts. This has already resulted in tens of thousands of publications containing false citations, and their number is growing at an accelerating rate, creating the risk of further worsening the reproducibility crisis in science [https://www.nature.com/articles/d41586-026-00969-z]. ChatGPT-5.0 also frequently cites retracted scientific articles [https://link.springer.com/article/10.1007/s11192-025-05484-y]. "Garbage in, garbage out." Today, the garbage may already exist within an apparently legitimate scientific publication complete with a DOI, journal, authors, and references.
Researchers from the The University of Texas decided to test whether AI models become less capable when they are fed the same kind of information overload that humans encounter—viral tweets, online hype, and meaningless social media posts. They do. They referred to this phenomenon as AI digital dementia [https://arxiv.org/pdf/2510.13928]. Potentially, a similar form of digital dementia may also arise when neural networks are trained on low-quality or noisy data.
When people perceive reality, they do not see it exactly as it is but only partially. The missing information is reconstructed by the brain on the basis of previous experience. For example, the human retina contains a blind spot, yet we do not notice it because the brain fills in the missing image [https://www.sciencedirect.com/science/article/pii/S0042698906003932]. Like neural networks, scientists who lack proper scientific information hygiene may encounter similar problems. They do not perceive reality itself but rather a version of reality reconstructed by the brain using prior knowledge—including a large proportion of false discoveries—combined with a relatively small amount of genuinely new information. When a scientist reads scientific papers, only part of the perceived information is truly new; the remainder is also reconstructed by the brain.
Every year, media headlines promise breakthroughs, cures, and revolutions. Time passes, and most of these promises collapse. But why would brilliant and highly intelligent people "systematically mislead the world"? This apparent paradox is precisely the point. There are no villains. The system simply rewards sensational findings rather than truth. A scientist's career depends heavily on the number of publications, making it more advantageous to conduct many small, statistically unreliable studies than a few large, statistically robust ones. Small but sensational studies almost always have low statistical power. According to the model developed by Andrew D. Higginson from the University of Exeter, the level of statistical power that is most advantageous for academic career advancement in such studies is only 10–40% [https://pubmed.ncbi.nlm.nih.gov/27832072/]. At this level of statistical power, a substantial proportion of the reported conclusions are likely to be false (Figure).
Many people argue that insufficient funding is allocated to research on aging and life extension. The evidence presented above suggests that the problem may not be a lack of funding but rather a system that hinders the achievement of genuine scientific progress. Approximately 18.3% of the U.S. gross domestic product is spent on healthcare. In addition, substantial financial resources are invested each year in biomedical research, including aging research. The more funding is directed toward career-advantageous studies characterized by low statistical power, multiple testing, and a high risk of bias, the more resources are spent on findings that are highly likely to be false. According to research by John P. A. Ioannidis from Stanford University, a large proportion of scientific studies lack sufficient reliability, and the probability that their conclusions are actually correct is low [https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.0020124].
If a substantial proportion of the fundamental biology of aging suffers from poor reproducibility, the question arises: where should we look for more reliable mechanisms of human aging? One answer already exists in clinical medicine. Modern guidelines on dyslipidemia [https://www.ahajournals.org/doi/10.1161/CIR.0000000000001423] are not merely protocols for lowering cholesterol. They constitute a map of the molecular targets of vascular aging because each target is linked not to an abstract age-related change but to the risk of myocardial infarction, stroke, and death.
The central idea is simple. Atherosclerosis develops over decades, and its molecular basis is the long-term exposure of the arterial wall to atherogenic lipoproteins. Therefore, medicine does not target "aging" in general but rather the specific molecules that create lifelong vascular exposure to damaging particles.
The first target is HMG-CoA reductase. This is the enzyme responsible for cholesterol synthesis in the liver. It is inhibited by statins. When cholesterol synthesis is reduced, the liver removes low-density lipoproteins from the bloodstream more efficiently, thereby reducing the exposure of blood vessels to atherogenic particles.
The second target is NPC1L1. This is the protein responsible for intestinal cholesterol absorption. It is inhibited by ezetimibe. The purpose of targeting this pathway is to reduce intestinal cholesterol absorption and further decrease the concentration of atherogenic lipoproteins.
The third target is PCSK9. This protein promotes the degradation of low-density lipoprotein receptors in the liver. The more active PCSK9 is, the fewer receptors remain on the surface of hepatocytes, and the less efficiently atherogenic particles are removed from the circulation. Modern medicine targets this pathway in two ways. The first involves monoclonal antibodies against PCSK9, such as evolocumab and alirocumab. The second uses small interfering RNA, such as inclisiran, which reduces PCSK9 synthesis at the messenger RNA level. This is no longer "general prevention" but a highly targeted molecular intervention.
The fourth target is ATP-citrate lyase. This enzyme acts upstream of HMG-CoA reductase in the cholesterol biosynthesis pathway. It is inhibited by bempedoic acid. This provides another strategy for reducing hepatic cholesterol production and lowering the burden of atherogenic lipoproteins.
The fifth target is apolipoprotein B. This is the principal structural protein of atherogenic lipoprotein particles. Any particle containing apolipoprotein B has the potential to contribute to the development of atherosclerosis. Therefore, apolipoprotein B is not merely a laboratory marker but an integrated molecular target representing the total burden of atherogenic lipoproteins.
The sixth target is lipoprotein(a). This genetically determined atherogenic lipoprotein is associated with the risk of atherosclerotic cardiovascular disease and calcific aortic valve stenosis. Its clinical importance lies in the fact that it may remain elevated even when low-density lipoprotein cholesterol levels are well controlled. Consequently, lipoprotein(a) represents an independent target of vascular aging, and RNA-based therapies designed to reduce apolipoprotein(a) production are currently under development.
The seventh target is ANGPTL3 (angiopoietin-like protein 3). It is inhibited by evinacumab. This target is particularly important in severe inherited lipid disorders because suppression of ANGPTL3 enhances hepatic clearance of very-low-density lipoprotein remnants and intermediate-density lipoproteins, thereby reducing atherogenic particles partly independently of the low-density lipoprotein receptor.
The eighth target is the microsomal triglyceride transfer protein. It is inhibited by lomitapide. This protein is required for the assembly of apolipoprotein B-containing lipoproteins, including very-low-density lipoproteins, low-density lipoproteins, and chylomicrons. Inhibition of this target reduces the production of atherogenic particles, although its clinical use is limited because of safety concerns.
The ninth target is apolipoprotein C-III. This protein inhibits the clearance of triglyceride-rich lipoproteins. The antisense oligonucleotide olezarsen binds to and degrades the messenger RNA encoding apolipoprotein C-III, thereby reducing its production and accelerating the clearance of triglycerides and very-low-density lipoproteins from the circulation.
The tenth target is vascular inflammation. In clinical guidelines, this is reflected by high-sensitivity C-reactive protein. Persistently elevated levels indicate an additional inflammatory component of cardiovascular risk. Thus, atherosclerosis is not merely a lipid disorder but also an inflammatory disease of the arterial wall.
Clinical medicine therefore already describes not a single molecular target of vascular aging but an entire system of interconnected targets. These include cholesterol synthesis, cholesterol absorption, degradation of low-density lipoprotein receptors, production of apolipoprotein B-containing lipoproteins, lipoprotein(a), triglyceride-rich lipoproteins, ANGPTL3, apolipoprotein C-III, and vascular inflammation. This represents a genuine clinical theory of vascular aging: a molecular target, a therapeutic intervention directed against that target, validation in humans, and evaluation of its effects on the risks of myocardial infarction, stroke, and death.
Another example is provided by the clinical guidelines on cardiovascular prevention [https://old.scardio.ru/content/Guidelines/Kardiovascular_profilaktika_2022.pdf]. These guidelines extend far beyond lipid metabolism and describe other evidence-based mechanisms of vascular aging, including endothelial dysfunction, chronic hyperactivation of the renin–angiotensin–aldosterone system, abnormalities of the sodium–glucose cotransporter 2 signaling pathway, thrombosis, sympathetic overactivation, chronic vascular inflammation, and residual cardiovascular risk. For most of these molecular and physiological targets, effective therapies already exist, with benefits confirmed by reductions in the risks of myocardial infarction, stroke, and cardiovascular mortality. If aging is viewed as the combined effect of age-associated diseases leading to an exponential increase in the force of mortality, then modern clinical guidelines effectively constitute a clinical theory of aging, presented not in a single document but as a system of evidence-based recommendations for individual diseases.
7. If Mathematical Survival Models Are Constructed for Individual Causes of Death Classified According to the ICD, and These Survival Curves Are Then Adjusted by the Magnitude of Mortality Reduction Already Demonstrated for the Corresponding Diseases and Risk Factors by Modern Medicine, the Combined Contribution of Individual Causes of Death to Overall Mortality Will Decrease. As a Result, the Model Will Predict an Increase in Population Life Expectancy. As New Effective Medical Interventions Become Available and Mortality from Diseases Classified in the ICD Continues to Decline, the Predicted Life Expectancy in Such a Model Will Increase Accordingly.
If, for each cause of death in the International Classification of Diseases (ICD), one uses not theoretical assumptions but evidence-based estimates of mortality reduction derived from clinical guidelines and incorporates them into a single population model, it becomes possible to estimate the projected median life expectancy under the comprehensive application of existing medical knowledge. Such a model does not answer the question of which treatment should be prescribed to a particular individual. Instead, it addresses a much more important question: what level of longevity could be achieved by a population if the full range of evidence-based preventive and therapeutic measures available through modern medicine and public health systems were implemented systematically?
In the Nestarenie Global project, we developed exactly such a model. For each disease category, estimates of mortality reduction were derived from clinical guidelines and studies with clinical endpoints. These effects were then integrated into a unified population mortality model. Under the assumptions adopted in the model, the simulations indicated that comprehensive implementation of the evidence-based interventions already available today could increase median life expectancy to approximately 100 years or beyond. This suggests that a substantial extension of human lifespan may be achieved not only through the discovery of entirely new technologies but also through the most complete implementation of existing evidence-based medical approaches across the entire population. It should be recognized that the results of the model depend on assumptions regarding interactions among risk factors, competing causes of death, and the completeness with which clinical effects are incorporated. Depending on these assumptions, projected life expectancy may be either overestimated or underestimated.
The principal value of such a model does not lie in producing a universal list of medications but in creating a quantitative protocol for longevity. It identifies which parameters should be brought to their optimal values—lipid profile, blood pressure, body weight, physical activity, prevention of infectious diseases, prevention of injuries, and other evidence-based risk factors—while the specific methods for achieving these goals are already described in detail in the corresponding clinical guidelines.
This is precisely the approach that we plan to implement by the end of 2026 in the Nestarenie.expert project. Beginning in the autumn, users will be able to model different longevity scenarios themselves by modifying individual risk factors through interactive tools and observing how their combined optimization affects projected life expectancy. Our goal is to transform fragmented clinical guidelines into a single quantitative model for managing longevity that is accessible to everyone.
8. The Age-Related Increase in the Force of Mortality and the Force of Disease Incidence Is Usually Preceded by a Gradual Decline in the Biological Functions of Organs and Physiological Systems. It Is on the Basis of These Functional Impairments That Diagnoses Are Subsequently Established According to the ICD. Therefore, Aging-Related Processes Can Be Monitored and Corrected Before Disease Develops by Targeting Measurable Functional Parameters in Real Time Rather Than Waiting for Clinical Outcomes or the Establishment of a Diagnosis.
The management of aging begins not with molecules but with function, because it is the impairment of organ and physiological system function that directly leads to disease, disability, and death.
The increase in the force of mortality and the force of incidence of age-related diseases is usually preceded by a gradual deterioration in the function of organs and physiological systems. It is this functional impairment that forms the basis for the diagnosis of many diseases classified in the International Classification of Diseases (ICD). Consequently, aging-related processes can be detected and corrected before disease develops, without waiting until diagnostic thresholds are crossed.
This approach is gradually becoming a new direction in modern gerontology. Whereas previous research focused primarily on identifying the molecular mechanisms of aging, an increasing number of investigators now propose shifting the emphasis toward measuring physiological function and the body's capacity to maintain performance and recover from stress. This does not mean that the molecular mechanisms of aging are no longer important. They remain a central subject of basic biological research. However, for the clinical management of lifespan, the crucial question is no longer which molecule has changed, but whether that change has resulted in a measurable deterioration of physiological function and an increased risk of disease or death.
A representative example is the editorial by Felipe Sierra and Viviana Perez, published in GeroScience in 2025 [https://link.springer.com/article/10.1007/s11357-025-01916-y]. The authors argue that the concept of the "Hallmarks of Aging," although it has played a major role in the development of gerontology, is gradually becoming insufficient for the practical management of aging. As scientific knowledge expands, the number of proposed molecular mechanisms continues to grow, increasing interactions are being identified among them, and attempts to define a single primary mechanism of aging are becoming progressively less realistic. As a consequence, the list of hallmarks may continue to expand almost indefinitely without bringing us closer to the ability to manage lifespan quantitatively.
Instead, the authors propose focusing on the common outcome of all these molecular changes—the gradual loss of organismal resilience. It is the declining ability of tissues and organs to maintain function and recover from stress that precedes the development of age-related diseases. Therefore, the primary object of observation should not be individual molecular processes but the physiological functions that determine human health.
This perspective is highly consistent with clinical medicine. Physicians do not diagnose disease on the basis of the activity of a signaling pathway or the number of senescent cells. A diagnosis is established when the function of an organ becomes impaired. Heart failure is diagnosed after deterioration of systolic or diastolic cardiac function, chronic kidney disease after a reduction in the glomerular filtration rate, chronic obstructive pulmonary disease after a decrease in forced expiratory volume, and osteoporosis after a reduction in bone mineral density. In every case, functional measurements constitute the basis for clinical decision-making.
Consequently, if function can be measured continuously, it becomes possible to manage aging before disease develops. For example, deterioration of diastolic cardiac function begins long before the onset of heart failure. Modern echocardiography allows these changes to be monitored in real time, while physical exercise, blood pressure control, weight reduction, and other interventions can improve diastolic function before a clinical diagnosis is established. Thus, the target of preventive intervention becomes the functional impairment that precedes disease.
This principle is universal. Most age-related diseases are preceded by a gradual decline in the function of the corresponding organ or physiological system. Therefore, the most rational strategy for managing aging is continuous monitoring of physiological function followed by corrective interventions before diagnostic thresholds are reached.
For this reason, the future of medicine will likely be associated not with an endless search for new molecular hallmarks of aging but with the development of quantitative models capable of continuously measuring organ function, predicting its future trajectory, estimating an individual's disease risk, and enabling timely intervention. This approach directly links the biology of aging with clinical medicine and provides the foundation for the practical management of healthy lifespan.
The principle of managing aging through the restoration of physiological function has already been implemented in modern clinical medicine. The goal of such interventions is not to target a hypothetical molecular mechanism of aging but to preserve or restore organ function before the diagnostic threshold for disease is reached.
One of the most illustrative examples is the American randomized controlled trial Reversing the Cardiac Effects of Sedentary Aging in Middle Age [https://pubmed.ncbi.nlm.nih.gov/29311053/], which investigated whether age-related deterioration of diastolic cardiac function could be reversed in healthy middle-aged adults. The authors demonstrated that a two-year program of regular physical exercise significantly increased left ventricular compliance and reduced age-related ventricular stiffness. In other words, the intervention was directed not at an individual molecular mechanism of aging but directly at the age-related decline in cardiac function, and this function was indeed improved, corresponding to an average reversal of cardiac functional age by approximately 23 years toward a younger physiological state.
This example illustrates a fundamentally different approach to rejuvenation. The object of intervention is not the hypothetical primary molecular causes of aging, which may differ among individuals and whose number is likely to be virtually unlimited, but rather the measurable functions of organs and physiological systems that determine the risks of disease and death. If function deteriorates, it can be detected, quantified, and in many cases improved before a clinical diagnosis is established.
For this reason, the future of medicine will likely depend not on the continual expansion of the list of molecular hallmarks of aging but on continuous monitoring of organ function, prediction of age-related functional trajectories using mathematical models and machine learning methods, and timely correction of identified functional impairments. This approach integrates the fundamental biology of aging with evidence-based clinical medicine and shifts gerontology from the search for hypothetical molecular causes toward the practical management of human health.