AI is Driving a Return to Holistic Medicine

Ancient medicine as practiced by the Chinese and Greeks thousands of years ago emphasized the importance of restoring balance and harmony in the body through a delicate blend of healing and careful observation. Greek physicians, influenced heavily by Hippocrates, incorporated diet, exercise, environment and lifestyle into health plans and emphasized the connection between physical and mental well-being. In subsequent centuries, the focus of healthcare shifted to acute and chronic diseases, which remains the norm today.  

In the last few decades, however, the emerging field of “scientific wellness” has begun to bring us back to the ancient notions of balance and harmony. Studies suggest that health is a combination of genetics (equivalent to what the ancients called fate), behaviors and environmental exposures for each individual. As a result, many of the central tenets of ancient holistic approaches to medicine are once again entering mainstream thinking of scientists and physicians. We are starting to appreciate that the human body is a unified system that we can assess through observation and come to understand through holistic systems-biology approaches. 

Driving this newfound appreciation for holistic medicine is the power of individual data-driven medicine. By sequencing each patient’s genome, a doctor can obtain advance warning of disease risks and information on the relative efficacy of different treatments. By routinely gathering tens of thousands of signals on a patient’s physiology, behaviors and environmental exposures, the doctor can come to a deep understanding of a patient’s phenome—the collection of all the dynamic characteristics of an individual’s biology.  

With a detailed picture of a patient’s genome and phenome, a physician can optimize wellness and healthy aging for each person by ameliorating the deficiencies that data-driven health reveals. In many cases, the doctor can anticipate disease before it advances far and nip it in the bud. When a patient is on the cusp of a transition that may lead to disease down the road, the doctor can intervene with preventive measures—including lifestyle changes, dietary modifications and preventive treatments—to reduce the likelihood of the onset of disease. Personalized recommendations based on the unique features of each patient’s genome, behaviors and environmental exposures give a picture not only of current disease but also the health trajectory of a patient that may lead to wellness optimization, healthy aging and ultimately lead to prevention of disease. Such an approach has the added benefit of enlisting the active participation of patients in their own healthcare decisions.  

Most doctors don’t yet practice this way, of course, but as our technology and knowledge advance, more and more will. This kind of high-tech, holistic approach is already beginning to transform healthcare from a reactive model focused primarily on treating diseases to a proactive model that emphasizes prevention and personalized care. It is our best bet to help people stay healthy throughout their full lifespan. 

A proof of concept 

The data-driven, scientific-wellness approach to healthy aging has been tested in real-world longitudinal studies in thousands of individuals. A few years ago, the scientific wellness company, Arivale, founded by Leroy Hood, Nathan Price and Clayton Lewis, gathered participants and administered extensive testing that measured a wide range of biological and behavioral factors. We were able to show striking improvements in wellness and healthy aging for most of the participants. Over four years of the program, more than 150 people transitioned from wellness to many different chronic diseases. The rich dataset gathered from this cohort allowed us to investigate the complex interplay of genetics, environment and lifestyle in determining health outcomes. 

The statistical analyses of the data yielded correlations that could be translated into hundreds of actionable possibilities—behavioral or clinical intervention strategies for reaching specific improved wellness and aging outcomes. For instance, ten people we had been routinely testing developed some form of cancer. We went back over the data we’d collected prior to their diagnoses and found multiple elevated physiological signals that had occurred months or years before their cancers were clinically apparent. As a result, doctors can now test for these signals, or biomarkers, in their patients to detect early signs of cancer before the onset of disease, giving them an opportunity to develop preventative interventions. Given that the treatment of chronic diseases consumes well over 80 percent of the $4.5 trillion spent each year on healthcare in the U.S., this approach could lead to substantial savings. 

One of Arivale’s major accomplishments was to develop a model that could predict an individual’s biological age—the age your body says you are, as distinct from your calendar age. The lower a person’s biological age is relative to chronological age, the better they are aging. It was found that biological age was correlated with wellness and disease phenotypes, suggesting that it may be a more accurate indicator of health status than chronological age. This approach allows for the calculation of not just overall biological age but also ages specific to each organ. In other words, your heart might be “younger” than your birthday, but your liver may be several years “older.” With this data come various possibilities for taking action that can lower one’s biological age globally or in individual organs. For instance, during Arivale’s wellness program, women lost an average of 1.5 years of biological age per year.  

Another key finding was the derivation of an algorithm for accurately measuring an individual’s body fat. The classical body-mass index (BMI) calculation, which uses height and weight measurements, does not accurately assess about 30 percent of the population, mainly because it doesn’t distinguish weight from muscle or fat. Our alternative, biological BMI (bBMI), is more responsive to metabolic changes and can accurately assess all individuals. This is a powerful metric to use in the context of the new anti-fat drugs. 

This research further led to discoveries of relationships between the gut microbiome and the response to common pharmaceuticals such as statins. The microbiome’s ability to enhance or diminish the effect of statins on low-density lipoprotein (LDL) cholesterol may be more pronounced than any genetic predisposition. Studies demonstrated associations between the composition of gut microbiota and factors such as body mass index, sleep duration, healthy aging and risks of developing certain diseases. 

Other studies contributed to our understanding of the genetic basis of health and disease. Large population studies (including hundreds of thousands of people) have identified collections of genes (10s to 100s) that contribute to many chronic diseases and from these polygenic scores one can derive the genetic risks of individuals for these diseases with their whole genome sequences. We found these genetic risks can influence how people respond to lifestyle interventions. For example, those with genes that confer a high risk for high LDL cholesterol (a proxy for heart disease) and high levels of LDL cholesterol in their blood can only bring it down by using drugs such as statins; in contrast, for those with low-risk LDL genes, diet and exercise can suffice. This raises the possibility that treatments should vary from one person to the next depending on their genetic profiles. We are now beginning to obtain data that will prove valuable for identifying the early stages of chronic diseases (such as diabetes, cancer, Alzheimer’s disease and heart disease), which may open the door to identifying biomarkers that could be used for early detection and intervention. 

Researchers at the Institute for Systems Biology are now exploring ways of assessing a person’s frailty, which we define as the diminishment of certain vital functions. Based on measurements of various molecules in the blood, a measurement of frailty would be independent of chronological age. We are also developing an AI algorithm that can quantify fragility, which we think would yield a deeper understanding of the underlying biological processes associated with aging.  

The right kind of data 

Artificial intelligence technologies rely on data to train predictive models. But human biomedical datasets tend not to represent the full diversity of the human population. Even though people who are white or of European descent make up about a quarter of the world population, they account for three-quarters of the world’s sequenced genomes. With such an imbalance, we cannot hope to develop therapeutics that are effective for everyone. 

Developing diverse datasets is also central to building accurate models of human diseases. Studies have shown that people with the APOE gene, which provides instructions for making a protein that helps transport cholesterol and fats in the bloodstream and is a risk factor in Alzheimer’s disease, generates very different genetic risks across different ethnicities. Individuals of Japanese descent who have the two bad copies of the APOE gene are almost three times more likely to develop Alzheimer’s than white people with the same genes. By contrast, among Hispanic populations with the same genetic profile, there appears to be no risk. South Asian populations show a higher prevalence of diabetes.  

Sex is another fundamental biological variable that can significantly influence gene expression and phenotypic outcomes. Ignoring sex-based differences in data can lead to biased models that fail to generalize to diverse populations. In 2016, the National Institutes of Health in the U.S. mandated that researchers include sex as a biological variable and justify single-sex investigations. Before that, however, biomedical studies were conducted almost exclusively on male animals and humans. As a result, almost all aspects of women’s health—from reproductive health and pregnancy to menopause—are data deserts for scientists. 

Socioeconomic status is another complex factor that can influence health and disease. Its impact on access to healthcare, nutrition and environmental factors are well documented. Healthcare data scientists are just beginning to accumulate the genome and longitudinal phenome data that will provide a myriad of new insights in this area. 

Overcoming bias in AI models presents a related challenge. Because models trained on biased data can perpetuate existing inequalities, we can help to mitigate bias by ensuring that the data used to train AI models is representative of the broader population. We must also establish standards of data collection so datasets gathered by different groups of scientists representing diverse racial populations can be compared and analyzed with one another.  

Data-driven medicine is likely to help drive down costs of healthcare and increase efficiency, but we do not yet have the financial data to make a strong economic argument. According to a 2021 report from Deloitte, a shift to wellness-focused healthcare in the U.S. could save trillions of dollars in chronic-disease care, potentially shifting that spending to wellness, healthy aging and disease prevention. The way to generate data to back up this claim is to target major chronic diseases that are associated with unhealthy aging, such as diabetes and Alzheimer’s.  

Phenome Health is advocating a large-scale research and clinical project, Human Phenome Initiative (HPI), to improve the quality and magnitude of health-related data. The HPI would generate a comprehensive database of genomes and phenomes that can be linked to clinical, mental, sociological and environmental information. The goal is to collect and analyze phenotypic data on a diverse population of a million or more individuals over a span of 10 years. The hope is that this biomedical resource will provide compelling data for strikingly improving individual wellness, healthy aging and initiate wide-spread early detection and prevention of chronic diseases. HPI may also provide the data necessary for researchers to identify the causal factors that contribute to various diseases and health conditions through the deployment of new AI techniques.  

HPI hopes to generate the right kind of data to power the next wave of health and biotechnology breakthroughs. The data and resulting insights will lead, we believe, to the validation of scientific wellness to improve the healthy aging of each individual. It will demonstrate the tremendous healthcare cost savings that come from preventing chronic diseases. It will catalyze the cost reduction of phenome-profiling technologies, making the technologies more accessible. And it will move wellness and prevention into the home by empowering individuals with new digital health technologies. Just as the Human Genome Project initiated a 100-million-fold decrease in the cost of DNA sequencing, we believe HPI will also lead to exponential decreases in the cost of phenomic technologies.  

AI-driven healthcare 

Wearables and AI avatars have a tremendous potential to democratize healthcare. By collecting and analyzing vast amounts of data, wearables and AI avatars can make it possible to detect health problems early, develop treatment plans tailored to each individual and provide remote (at-home) monitoring. They can also potentially lower healthcare costs, increase access to care and advance medical research. Addressing challenges such as data privacy, interoperability and ethical considerations is crucial to fully realizing the potential of these technologies.  

Ensuring the security of personal health data requires robust measures and explicit informed consent. Seamless data sharing and communication across different devices and platforms, which is crucial to achieve the full potential of these technologies, requires standardized protocols and data formats. It’s also essential to address biases in AI algorithms, prevent discrimination and protect patient autonomy.  

In the future, each of us might have an AI avatar that acts as a health companion, collecting and analyzing data from our wearable devices, electronic health records and other sources of data. Using advanced algorithms, such a health companion would make recommendations about sleep, managing stress, diet, exercise and repairing deficiencies arising from one’s genome or phenome. It would track changes in an individual’s health over time, identifying potential risks and areas for improvement as they occur.  

The AI companion would also serve as a bridge between people and their doctors. For example, if it detected a decline in heart-rate variability, it might recommend stress-reduction techniques and schedule a check-up with your cardiologist. By providing personalized guidance and support, it could help individuals take a more proactive approach to their health and well-being.  

The healthy-aging approaches of gerontology (study of aging) and scientific wellness will soon converge. As a result, the human healthspan will be extended into the 90s and beyond. People will live productive and creative lives for 20 or more extra years than they do now.  

We are moving toward a transformation from the current disease-oriented healthcare system to one that actively optimizes for wellness and healthy aging and prevents disease. Through continued advancements in technology, this data-driven health approach will be democratized and made accessible to all populations. In five years, millions of patients across the world will probably have access to the kind of clinical care we have been describing in this article. Scientists will also be able to execute nationwide data-driven health studies. 

Finally, the widespread deployment of AI will catalyze change in almost every aspect of healthcare, including using AI to empower any physician with the domain expertise necessary to take a holistic approach to practicing medicine—just as our predecessors did thousands of years ago.  

As more and more people live longer, healthier lives, we may eventually need to grapple with a different problem: What to do with that extra decade or two of healthy life?

Find out more about Phenome Health’s efforts to transform the future of health care here.

Explore the emerging science of healthspan in other stories in this special report.