Personalized Medicine: Motivation, Challenges and Progress

Personalized Medicine

Abstract

Personalized medicine is based on the idea that individuals have unique characteristics at molecular, physiological, environmental, and behavioral levels, requiring tailored interventions.
Emerging technologies like DNA sequencing, proteomics, imaging, and wireless health monitoring have confirmed significant inter-individual variation in disease processes.
The review covers motivation, historical context, enabling technologies, successes, setbacks, vetting, deployment, and future directions of personalized medicine, including fertility and sterility treatments.
It also addresses current limitations and argues that personalized medicine is becoming inevitable due to biological realities and increasingly efficient assays and deployment strategies.
Keywords: Precision medicine; biomarkers; patient monitoring; genomics

Introduction

High-throughput biomedical assays (DNA sequencing, proteomics, imaging, wireless monitoring) reveal inter-individual variation in disease effects and mechanisms.
This raises questions about tailoring disease treatment, monitoring, and prevention to individual biochemical, physiological, environmental, and behavioral profiles.
Personalized medicine is used interchangeably with individualized and precision medicine, though subtle distinctions exist.
Challenges include regulatory approval and acceptance by stakeholders (physicians, executives, insurers, patients) due to the need to prove that personalized medicine outperforms traditional strategies.
Tailored therapies like autologous CAR-T cell transplants and mutation-specific medicines such as ivacaftor can be very expensive.

History and Motivation of Personalized Medicine

The review considers the history and motivation of personalized medicine, context on what personalized medicines strategies have emerged, limitations slowing their advance, and what is on the horizon.
Strategies for proving that personalized medicine protocols and strategies can outperform traditional medicine protocols and strategies are considered
Importance of distinguishing examples and challenges in personalized disease prevention, health monitoring, and treatment.

Archibald Garrod and The Precursors Of Personalized Medicine

Archibald Garrod studied ‘inborn errors of metabolism’ like alkaptonuria, albinism, cystinuria, and pentosuria.
Garrod's work on alkaptonuria showed that affected family members had outlying values in biochemical assays (e.g., urine).
He concluded alkaptonuria was due to a specific ‘altered course of metabolism.’
Garrod suggested these conditions are extreme examples of chemical behavior variations present in minor degrees everywhere.
He believed metabolic differences could explain varying disease susceptibilities and manifestations.
Garrod's work occurred amidst debates about genetics; he referred to ‘factors’ influencing disease, consistent with the modern notion of genes.
Mendel's findings connected specific phenotypes to breeding protocols in peas, anticipating genetics.
William Provine's book discusses the debate on how genes could explain phenotypic variation.
‘Mendelians’ (William Bateson, Hugo de Vries) focused on discrete factors responsible for inheritance patterns.
‘Biometricians’ (Karl Pearson) focused on continuous phenotypes like height, questioning how to reconcile them with discrete factors.
Ronald Fisher reconciled the debate by suggesting many genes contribute in a small way to a phenotype.
The collective effect of these factors creates continuous variation in phenotypes.
This has been validated by high-throughput genetic technologies, showing individuals vary widely with subsets of genetic variants.
Some genetic variants arise as de novo mutations, unique to an individual.
Phenotypic variation explains differences in disease susceptibilities and intervention responses.
Personalized medicine considers genetic studies, environmental exposures, developmental phenomena, epigenetic changes, and behaviors.
Hogben and Sim emphasized the need for clinical practice to consider patient-specific characteristics to determine appropriate interventions.
Greater patient information, utility vetting, and strategies for incorporating study results into future care are needed.
Practical issues in implementing personalized medicine include patient follow-up, satisfaction assessment, and comparison with other interventions.
These issues are often overlooked amid the development of profiling technologies and evidence for inter-individual variation.

Early Examples of Personalized Medicine

Personalized medicine can be applied to disease treatment, early detection, and prevention.
Early examples focused on genetically-mediated pharmacokinetic aspects of drugs due to understanding of drug-metabolizing enzymes.
Weber's book provides an excellent introduction to pharmacogenetics, especially genetic variants in drug-metabolizing enzymes.
Warfarin: A blood-thinning medication, targets the VKORC1 gene and is metabolized by the CYP2C9 gene. Genetic variation in both genes leads to variation in the pharmacodynamic and pharmacokinetic properties of Warfarin across individuals.
The FDA recommends dosing based on an individual's genotype in the VKORC1 and CYP2C9 genes.
VKORC1
CYP2C9
Primaquine (PQ): Used for malaria management. Some soldiers treated with PQ became jaundiced and anemic, exhibiting acute haemolytic anaemia (AHA).
Individuals exhibiting AHA carried variants in the G6PD gene.
Current practice involves genotyping patients for G6PD variants before PQ administration.
G6PD
I matinib: Used to treat chronic myelogenous leukemia (CML). Inhibits tyrosine kinase, increased by the bcr-abl fusion (Philadelphia chromosome).
Imatinib is given only to CML patients with this fusion event.
bcr-abl

Contemporary Examples Of Personalized Medicine

Driven by success of drugs like warfarin, PQ, and imatinib that work based on a patient's genetic profile.
Interest has expanded into personalized disease surveillance and prevention.

Mutation-Specific Therapies

Identifying genetic profiles and crafting therapies targeting those profiles.
Ivacaftor: Treats cystic fibrosis (CF) in individuals with specific pathogenic mutations in the CFTR gene.
CFTR gene has a gate-like structure controlling salt movement in and out of cells.
Ivacaftor opens the gate for longer periods in the presence of mutations causing gate closure.
Only useful for CF patients with this specific gating problem.
Connections between genetic variants and drug efficacy/side effects are growing.
The US FDA provides a list of approved drug-based interventions requiring a test: https://www.fda.gov/Drugs/ ScienceResearch/ucm572698.htm.
The Personalized Medicine Coalition (PMC) considers the practical implications of approved interventions.

Immunotherapies

Aim to prime or trigger an individual's immune system to attack cancer.
Exploit unique sets of genetic alterations (neo-antigens) in tumor cells to raise an immune response.
Harvest T cells from a patient, modify them to recognize and target neo-antigens, then reintroduce them to attack tumor cells.
Cytotoxic T cell therapies can be very patient-specific.
Neo-antigen profiles can be unique, and autologous constructs (patient's own T cells) are modified.
Attempts to make allogeneic constructs (one individual's T cells modified and introduced into another's body) are pursued.

Personalizing Early Detection Strategies

Monitoring individuals susceptible to a disease should use ‘personal thresholds’ instead of ‘population thresholds’.
Population thresholds are derived from epidemiologic data (e.g., cholesterol > 200 for heart disease, systolic blood pressure > 140 for hypertension).
Personal thresholds are established from legacy values collected on an individual over time.
Significant deviations from historical values indicate health status change, irrespective of population thresholds.
Drescher et al. explored personal thresholds for longitudinal CA125 levels in women, some developing ovarian cancer.
Personal thresholds captured ovarian cancer at the same time or earlier than population thresholds, almost a year earlier on average.
As costs and convenience improve, personal thresholds will likely become the rule in health monitoring.

Personalizing Disease Prevention

Using genetic information to develop personalized disease prevention strategies.
Genetic information can decrease disease risk and complications from standard treatment and screening strategies.
Colorectal Cancer: Remains the third leading cause of cancer deaths despite being highly preventable.
Liao et al. (2012) reported improved survival and decreased cancer-specific deaths with postoperative aspirin in patients with somatic mutation in the PIK3CA gene.
Nan et al. (2015) reported varying effects of aspirin use on colorectal cancer risk depending on genotype.
Aspirin use can have serious side effects; limiting its use to those predicted to benefit based on genotype would be ideal.
Jeon et al. (2018) reported the use of expanded risk prediction models for determining when to begin colorectal cancer screening.
Current guidelines use age and family history; Jeon et al. showed that environmental exposure and genetic profile could change screening recommendations by 12 years for men and 14 years for women.
AUC value for a model including environmental and genetic factors was 0.63 for men and 0.62 for women, compared to 0.53 (men) and 0.54 (women) with only family history.
Improvement over models without genetic or environmental information justifies their use, although there is still room for imporevement.

Testing Personalized Medicines

Recognition of personalized medicine as a paradigm is relatively recent; not enough time has elapsed to show it works broadly.
Questions arise about how to vet or test the utility of personalized medicine.
Three emerging strategies: N-of-1 clinical trials, intervention-matching trials, and adaptive clinical trials.
These strategies borrow from traditional randomized clinical trials (RCTs) but deviate significantly from population-based RCTs.

N-of-1 Clinical Trials

If there is no reason to believe that any one of a set of different interventions matches an individual's profile better than others, then there is ‘equipoise’ among those interventions.
Trials focusing on an individual's response to different interventions to determine an optimal intervention are referred to as ‘N-of-1’ or single subject trials.
N-of-1 trials use cross-over designs (e.g., ‘ABABAB’) to compare interventions by alternating them and collecting data on the individual's response.
Randomization, blinding, washout periods, multiple endpoints, and other design elements can be used in N-of-1 trials.
N-of-1 trials must accommodate serial correlation between observations and carry-over effects from one intervention to another.
Cross-over trials are impractical in acute or life-threatening conditions; sequential N-of-1 designs with real-time monitoring are proposed.

Intervention Matching Trials

If evidence suggests patient profile features can identify interventions that might work, the question arises as to how to test that the hypothesis that providing interventions to those individuals based on these ‘matches’ leads to better outcomes than providing those individuals interventions based on some other scheme or strategy.
Testing individual matches may require many small clinical trials.
‘Basket’ and ‘umbrella’ trials test matching strategies against alternative intervention methods in oncology.
Basket trials enroll individuals regardless of cancer tissue, while umbrella trials focus on a single tissue.
Each patient's tumor is profiled, usually via DNA sequencing, to identify actionable ‘driver’ perturbations.
Drugs are matched to tumor perturbations (e.g., EGFR inhibitor for mutated and overexpressed EGFR).
The trial tests whether using different intervention baskets based on the matching scheme results in better outcomes than other intervention methods
If the matching scheme fails, it does not necessarily mean the interventions or personalized medicine concept are flawed.
Some basket trials only have a single basket and no comparison group but rely on determining which patient profiles appear to be associated with better outcomes for the intervention being tested.
Intervention matching schemes are likely to become the rule, especially with computational environments like IBM's ‘Watson’ system.
Watson includes a database extracted from medical literature, linking patient information to outcomes, enhanced by statistical methods.
Watson can identify perturbations in a tumor and suggest interventions.
Use of IBM's Watson in clinical settings has led to discussions about how best to test and deploy such as a system as a way of supporting physicians' decisions about an intervention choice for individual patients.

Adaptive Clinical Trials

Have as one of their focal points a desire to minimize the amount of time a patient is on what is likely to be an inferior therapy.
Adaptive trials minimize the time a patient is on an inferior therapy.
In personalized medicine, if there is equipoise or if an untested or conventional intervention is being used, the evaluation of the effects of each intervention on an individual to determine the best one for that individual (as in a very elaborate N-of-1 study) might be impractical or cause more harm than good because some, if not all, of the interventions might not actually benefit that individual.
Biomarkers reflecting response or adverse effects are collected and monitored to determine if an intervention is not working.
If signs indicate failure, the individual could cross-over to a new intervention.
Adaptive designs can be difficult to implement and analyze but are often seen as more ethical.
Adaptive components can be added to N-of-1, aggregated N-of-1, and intervention-matching trials.
The work of Murphy and colleagues minimizes the time a patient is on an inferior treatment.

Emerging and Next-Generation Personalized Medicine Strategies

Four recent research and clinical activities: patient-derived cell avatars, intense individualized diagnostic and monitoring protocols, personalized digital therapeutics, and personalized medicine approaches in fertility treatment.

Patient-Derived Cellular Avatars

Cells harvested from individuals can be used to generate additional cell types relevant to a patient's condition without directly biopsying the affected tissue via pluripotency induction (iPSC).
Allows the creation of a ‘disease in a dish’ cellular model of a patient's condition.
These in vitro cellular ‘avatars’ can be studied to identify key molecular pathologies that might give an indication as to how best to treat an individual patient of interest.
iPSC technologies can be extended with CRISPR to create isogenic cells with and without the mutation in question. Comparison of these cells allows direct insight into the effects of the mutation while controlling for all relevant genetic background effects associated with the patient's genome.
Partial organs or ‘organoids’ can be created from cells, providing greater insight into molecular pathologies and cell:cell interactions.
Integrating patient avatars with other information and protocols can achieve truly personalized medical care.
Schork and Nazor describe the motivation and integration of different aspects of patient diagnosis, intervention choice, and monitoring, using, among other things, patient avatars.
Cell-based patient avatars can accommodate personalized drug screening: thousands of drugs tested against a patient's cells or organoids (possibly modified with CRISPR) to correct molecular defects.
Approved drugs could be tested for efficacy under a repurposing protocol.
Patient-derived cells have shown success in cancer settings for drug screening.
The biggest concern is whether in vitro models capture relevant in vivo pathobiology and drug response information.
A more direct strategy for in vivo experimental cancer intervention choice could involve implanting a device into a patient's tumor in vivo and then delivering different drugs through that device to see which ones have an effect.

Intensive Personalized Health Monitoring

Inexpensive genotyping and sequencing allow individuals and providers to assess genetic risk.
Health monitoring devices, online blood assays, and inexpensive imaging allow continuous monitoring.
Combining genetic risk assessment with intense health monitoring makes sense.
Individuals with unique diseases have benefited from genetic diagnoses, uncovering pathogenic mechanisms or targets for pharmacotherapies.
Individuals have monitored their health intensely to identify health status changes attributable to genetic susceptibilities.
Monitoring individuals for health status changes is not trivial, however, if the measures being collected have not been evaluated in a population.
The community is recognizing the utility of establishing ‘personal thresholds’ for measures as opposed to ‘population thresholds.
As noted, population thresholds are established from epidemiologic and population survey data and include often-used thresholds for determining disease status such as a cholesterol level greater than 200 for heart disease or a systolic blood pressure greater than 140 mmHg for hypertension.Personal thresholds are established from longitudinal or legacy values of a measure collected on an individual and may be unique to the individual in question and their use in some settings suggests that they work better than population thresholds.

Digital Therapeutics and Personalized App Content

Smart phones can collect health data and provide personalized advice, feedback, and coaching.
This has led to ‘digital therapeutics:’ smart phone apps designed to treat medical or psychological conditions.
Content varies based on what is learned about the individual and their response, personalizing the app.
Many digital therapeutics have undergone evaluation for their ability to engage users and benefit them.
The FDA has guidelines for registering digital therapeutics as health technologies and has begun approving them.
The first approved digital therapeutic (for substance abuse) was approved by the FDA in 2017.

Personalized Interventions Involving Fertility and Sterility

Personalized medicine strategies can be applied to fertility treatments.
Leveraging ‘real world’ data from reproductive medicine clinics (EMR systems) can shed light on variations in fertility rates and intervention responses.
Digital medicine proposals include smart phone apps providing personalized coaching content to enhance pregnancy.
Genetic variants influencing fertility can be used to support diagnoses or personalized intervention plans.
Adaptive trial designs can assess the utility of personalized approaches to raising awareness about time to conception and fertility.
Emerging strategies to enhance fertility go beyond traditional ovarian stimulation.
Cryopreserving oocytes and ovarian tissue for later implantation is highly personalized.
This procedure works if preserved tissues are viable and not damaged; cells could be corrected for genetic defects using gene editing.
A futuristic personalized fertility intervention involves cell reprogramming to generate sperm and egg cells from other cells (e.g., skin cells) for fertilization – ‘in vitro gametogenesis.’

Conclusions

Personalized medicine is a necessity due to clinically meaningful inter-individual variation.
Modern technologies like DNA sequencing, proteomics, and wireless monitoring have enabled the identification of this variation.
Future challenges involve improving the efficiency of individual characterization and vetting personalized medicines.
Interventions that work ubiquitously still matter, but they might be very hard to identify going forward.
Large data collections needed to identify factors distinguishing groups could raise privacy concerns.
Strategies from banking, marketing, and social media industries could be used in health care settings.
Developing more efficient ways of developing personalized medicines (for example, with respect to cell replacement therapies or mutation-specific drugs that work for a small fraction of patients) is crucial to meet the demands of all patients.
Paying for personalized medicine may be complicated due to initial higher costs.
Better strategies to educate and train health care professionals about personalized medicine are needed.