48

Precision medicine is positioned to revolutionize the healthcare landscape, and the utilization of advanced

clinical development technologies is key to supporting precision medicine trials. These technologies can

harness a variety of disparate data sources at richer and greater volumes to build more accurate predictive

models for different patient subsets. We discuss briefly the nature of precision medicine, the FDA’s initiatives

encouraging the development of targeted therapies and the importance of the revolution in genomics for

precision medicine. We describe the five Vs of big data and the pivotal role of big data in precision medicine.

We present four clinical trial technologies that provide essential capabilities for precision medicine:

(1) Continuous collection, aggregation, and integration of myriad data sources; (2) Adaptive and iterative

study design and execution as new information becomes available; (3) Maintenance and management of data

throughout the study life cycle; and (4) Advanced analytics for research discovery. Precision medicine shows

great promise for developing exploratory hypotheses from its broad variety of data and iterative approach,

evaluating these hypotheses, and producing safer, more effective treatments for targeted patient subgroups

with advantageous benefit-risk profiles. The complexities of precision medicine study designs require

substantial planning, investment, and commitment from all stakeholders.

Clinical Trial Technologies for Precision Medicine: The Current State of the Art by Ruthanna Davi, Ph.D.,Therese Dolan, Jeffrey Wiser, and Steven Schwager, Ph.D.

1. Introduction

Precision medicine is an approach to medicinal

practice that tailors decisions to a patient’s

individual characteristics: genetic, anatomical,

physiological, environmental, and lifestyle.

Precision medicine is defined by the National

Academies of Science as “an emerging

approach for disease treatment and prevention

that takes into account individual variability

in genes, environment, and lifestyle for each

person” (National Academies of Science, 2017).

Individually tailored healthcare is a rapidly

developing, exciting alternative to the

traditional practices that are often described

as one-size-fits-all, based on the best course

of prevention or treatment for the general

population. In this approach, if an initial

treatment fails, the patient moves on to the

next standard treatment, again based on the

population. The challenge of this method

of practicing medicine have long been

recognized.

Clinicians and patients today have access to a

rapidly expanding range of data. Drug regula-

tory authorities and the practice of medicine

itself are adapting to this new information

era, which holds promise both to further our

understanding of disease etiology and to

deliver better patient care and optimal

outcomes (Heckman-Stoddard, 2014;

Woodcock, 2017).

Torkamani et al. (2017) recently expanded on

the concept of precision medicine by defining

high-definition medicine as “the data-driven

practice of medicine through the utilization of

highly detailed, longitudinal, and multi-para-

metric measures of the determinants of health

to modify disease risk factors, detect disease

processes early, drive precise and dynamically

adjusted interventions, and determine

preventative and therapeutic intervention

efficacy from real-world outcomes.” Establishing a

personal baseline through frequent assessment

of the determinants of health can enable early

detection and intervention, sometimes before

diseases manifest clinically.

The objective of this paper is to describe the

current state of affairs for sponsors who are

considering a precision medicine approach and

seeking to create a development path that will

maximize the chance of success.

1.1. Precision Medicine and the FDA

The U.S. Food and Drug Administration (FDA)

has embraced the possibilities offered by

precision medicine. This is reflected in a recent

New England Journal of Medicine article by

Janet Woodcock, director of FDA’s Center for

Drug Evaluation and Research (CDER), and Lisa

LaVange, director of the Office of Biostatistics

at CDER (Woodcock, 2017) that discusses the

need to capitalize on similarities among trials

(e.g., shared control group and recruiting

efforts) to gain efficiencies in drug development

using umbrella, basket, or platform trials and

cites I-SPY2 trial and Lung-Map as examples of

such innovation.

The FDA recently announced the Precision

Medicine Initiative (PMI), with the overarching

goals of facilitating the identification of genetic

drivers of disease and developing better

diagnostic tools and treatments (FDA, 2016).

The FDA has issued two PMI draft guidances

and established the precisionFDA portal

for oversight, analysis, and validation of

methodology related to next-generation

sequencing (NGS), also known as

high-throughput sequencing. NGS tests are

used to identify the genomic differences in

individuals (FDA, 2016) that are the basis of

the genetic component of precision medicine.

The FDA has fostered numerous initiatives that

encourage drug sponsors to develop targeted

therapies and has provided guidance on the

regulatory pathway for such medicines. FDA’s

Enrichment Guidance (FDA, 2012) describes

approaches to studying therapeutics when

the efficacy or safety of that therapeutic is

expected to vary across subgroups. The FDA

guidance on adaptive designs provides advice

for incorporating analyses of accumulating

data at prospectively planned time points

(FDA, 2010). Adaptations such as modifications

in the treatment randomization ratios,

dropping or adding an arm, or revising a

patient’s treatment based on their inadequate

response to an initial treatment are all

examples of how adaptive designs can fuel

precision medicine and are discussed in the

guidance. The Co-development Guidance

provides advice for obtaining contemporane-

ous marketing authorization for a therapeutic

product and its corresponding companion

diagnostic (FDA, 2016). And, expected soon is

the Complementary Diagnostics Guidance.

This guidance is expected to lay out a

framework that allows for a diagnostic device

that aids in the benefit-risk assessment of

the therapeutic product (but it is not essential

as with the companion diagnostic).

The FDA has approved an increasing number of

medicines that are indicated for defined subsets

of patients, including the first ever-approval

of a drug [Keytruda (pembrolizumab)] to treat

patients whose cancers have a specific genetic

feature, or biomarker, rather than being based

on the location of origin of the cancer.

Importantly, these precision medicine

approvals are not limited to oncology

indications. For instance, the FDA has approved

two products for treatment of severe asthma

in patients with an eosinophilic phenotype

[Cinqair (reslizumab), Nucala (mepolizumab)].

In March 2017, FDA concurrently approved a

drug [Zejula (niraparib)] and a complementary

diagnostic test [BRACAnalysis CDx] to identify

patients with germline BRCA mutations, the

subset of patients expected to benefit most

from Zejula. A complementary diagnostic test

is defined as one that aids in the benefit-risk

decision(s) related to a particular therapeutic

product but is not essential to the safe and

effective use of the product, as is the case with

a companion diagnostic. Draft guidance on the

topic of therapeutic products developed with

a complementary diagnostic test is expected

from the FDA soon (Mansfield, 2015). Overall, the

FDA continues to solidify a workable regulato-

ry platform that encourages innovation while

maintaining efficacy and safety as priorities.

49

PR

EC

ISIO

N T

OO

LS

51

PR

EC

ISIO

N T

OO

LS

1.2. Genomic Data and the Current State of

Precision Medicine

The ongoing revolution in genomics and

molecular biology continues to improve our

fundamental understanding of human biology

and disease; at the same time, it presents new

opportunities for precision medicine. For

example, not long ago, breast cancer was

largely defined as a single disease. Today,

based on molecular subtyping, breast cancer

is recognized to consist of several distinct

molecular subtypes (Bartlett, 2017). The

biomarkers that distinguish these molecular

subtypes or influence the risk of cancer

development have been translated into

targeted clinical practice (Marchina, 2010;

Godet, 2017; Slamon, 2001). Similarly, the

clinical phenotype of depression has long

been known to encompass several subtypes

(e.g., major depressive disorder, bipolar

disorder, and atypical depression) and to

show high variability regarding treatment

response. This has often made the diagnosis

As precision medicine continues to propagate

into clinical practice, it is increasingly impor-

tant for clinicians to be trained to understand

precision medicine and to use it effectively

for the benefit of their patients. The National

Human Genome Research Institute (a research

institute of the National Institutes of Health

(NIH)) recently awarded several training grants

that will instruct medical doctors in applying

computational tools to disease-gene discovery,

patient genome interpretation, and big data

management in research and clinical settings

(National Human Genome Research Institute,

2016).

Drug developers may view precision medicine

as a daunting and uncertain task, and utilizing

large datasets (i.e., big data) effectively for

precision medicine requires careful planning,

investment, and discussion among stakehold-

ers. Establishing the appropriate infrastructure

is a crucial step in dealing with huge volumes

of information and the complexities of

adaptive precision medicine study designs.

and treatment of depression challenging for

clinicians. However, exciting new biomarker

data may soon aid clinical practice by providing-

molecular insights to guide accurate diagnosis

and prediction of treatment response (Buzuk,

2016; McIntyre, 2016; Strawbridge, 2017; Yue,

2016).

Increasingly, precision medicine is transitioning

from a specialized treatment approach to an

aspect of standard clinical practice. As of

September 2017, approximately 179 studies

(~50% actively recruiting) were listed in

clinicaltrials.gov as search returns for the term

“precision medicine,” with approximately

21% of these trials being funded by industry.

This is a gross underestimate of the actual

number of trials that can be deemed precision

medicine trials, because most of these

trials are not explicitly labeled as such in

clinicaltrials.gov. Based on combinations

of keywords related to biomarkers, next

generation sequencing and synonymous

terms, we can roughly suggest that this

number is an order of magnitude greater.

52

To predict outcomes accurately in a patient

subgroup, precision medicine demands large

volumes of data from often disparate data

streams, some of which are generated

continuously. One challenge for drug

developers is how to seamlessly ingest,

aggregate, integrate, and analyze these

datasets in an easy-to-use workflow that

can help uncover meaningful relationships

resulting in improved patient outcomes.

Precision medicine is an exploratory and

iterative journey, so the data platform of

choice must be flexible enough to adapt to

emerging data and to allow investigators to

redefine – in real time – patient clusters into

more precise subsets and to redesign studies

to enable refined hypothesis testing. A useful

starting point is to adopt a flexible, integrated,

and intelligent clinical development platform

as the engine that will drive the precision

medicine program. These are crucial features

in an optimized precision medicine approach.

The remainder of this paper discusses the

relevance of big data to precision medicine

trials and some key clinical development

technologies that can facilitate the

development of targeted therapies.

“ Phenotypic data, drug interaction data, family history data. Making sense of that data is a huge challenge, The key is to teach physicians to effectively use the new tools to tackle big data.”(National Human Genome Research Institute, 2016)

The ability to redefine patient clusters and adapt study designs in real time are key attributes of a platform to support a precision medicine approach.

“As drugs become more precise and trials become more complex, it will be increasingly important to exploit optimized clinical trial technologies.The platform should seamlessly ingest, aggregate, integrate, and facilitate analysis of disparate data types so that the ideal target patient population can be identified quickly and reliably.”

53

PR

EC

ISIO

N T

OO

LS

2. Big Data and the Path to Precision Medicine

A precision medicine approach to drug

development relies on large datasets to cluster

patients into increasingly more precise subsets.

This requires that investigators measure

patients on numerous attributes of many

different types, ranging from genotype to

phenotype and beyond (Beckmann, 2016;

Dinov, 2016; Dumitrescu, 2017). Simply

ingesting and managing data of such scope

and complexity are formidable tasks, and

conceiving, executing, and interpreting

appropriate analyses are even more formidable.

However, we must meet these challenges in

order to achieve the tremendous gains that

precision medicine promises over one-size-

fits-all population-based medicine.

The precision medicine approach generates a

unique and comprehensive patient profile by

integrating a wide range of medical data.

Taken in the aggregate, these data encompass

enormous volume, velocity, variety, veracity,

and value. We can describe these attributes –

the five Vs of big data – briefly as:

Volume — Enormous amounts of data are

generated by many data streams of different

types, including over 3 billion nucleotides

in the human genome; hundreds or thousands

of laboratory test results and vital signs;

wearable sensor readings of dozens of kinds;

environmental data on dozens or hundreds

of relevant biological, chemical, and

meteorological features; medical imaging

data containing many thousands or millions

of pixels per image; and more.

The utility of genetic information to drive

precision medicine has been clearly

demonstrated (e.g., trastuzumab (Herceptin)

for women with HER-2 positive breast cancer).

Other important data streams also offer vast

possibilities for additional lines of investiga-

tion. Electronic health records (EHRs) are an

example of big data that can help inform a

precision medicine approach. A recent article

by Dumitrescu et al. (2017) concluded that

current and new EHR technology will provide

international standards for applications that

facilitate precision medicine using the

complex healthcare information in health,

social, economic, behavioral, and environmental

data. This article reported the results of a

genome-wide association study for resistant

hypertension linked to EHRs and genotypes.

The authors demonstrated the utility of DNA

biobanks linked to EHRs for classifying

resistant hypertension cases and controls

and for performing large-scale genome-wide

association studies. Evans (2016) utilized an

algorithm based on a combination of billing

codes, laboratory values, text queries, and

prescription records. A study by Szlosek (2016)

reported the use of machine learning and

natural processing algorithms to automate the

evaluation of clinical decision support in EHR

systems. The authors showed that a machine

learning algorithm can be used to identify

abnormal head CT scans in free-text health

care data with a high degree of accuracy.

These examples demonstrate that using big

data to support a precision medicine approach

has moved beyond a conceptual framework.

These and other well-documented studies

showcase how big data technology can

enable precision medicine. However, it is our

experience that some drug developers are not

prepared to handle the complexities inherent

in the five Vs of big data, which can quickly

overwhelm the clinical development platform

and the supporting business processes not

designed for them.

Velocity

Many data streams produce new data at

rates of dozens or hundreds of points per

second (e.g., sensor data, environmental

data, imaging data).

Variety

The information collected for a patient

includes structured and unstructured data

from vital signs, laboratory tests,

genomic sequencing, patient behaviors,

social media, wearable sensor readings,

point-of-care notes, photos, medical

images, and more.

Veracity

It is critical for precision medicine trials

to have high-quality data. Missing and

inaccurate observations can rarely be

completely eliminated, but every effort

must be made to minimize their occur-

rence. Poor data quality will produce

errors in data interpretation and the

analyses required for precision medicine.

Value

The impressive scope of big data is

not an end in itself, but a means to the

end of valuable results: better therapeutic

and preventative treatments. Precision

medicine must manage the volume,

velocity, and variety of big data and

maximize the veracity of the data;

however, to be successful, precision

medicine must extract value from the

data with analytical algorithms whose

results yield improved treatments for

patients.

In precision medicine as well as traditional

medicine, data quality is a wide-ranging,

serious issue that poses a grave threat to the

validity of clinical research. Poor planning

often results in poorly standardized data,

missing or incomplete clinical datasets,

incompatible conceptualizations of data

types and elements, and infrastructure that is

inadequate for processing robust healthcare

data (Anderson, 2014; Field & Sansone, 2016;

Mead, 2006; Richesson, 2007).

The following section highlights clinical trial

technologies that make precision medicine

possible, not only by enabling the efficient

collection, aggregation, and integration of

data, but by empowering the exploratory and

iterative processes that drive the precision

medicine approach via powerful analytics.

3. Clinical Trial Technologies to Facilitate

Precision Medicine

Advanced clinical trial technologies allow

investigators to seamlessly collect and integrate

massive amounts of data from sources across

the clinical care continuum. An adaptable

system can leverage both historical datasets

and in-study datasets to update and optimize

models that allow for re-running the cohort

selection, risk stratification, and subsequent

intervention assignment. Further, advanced

predictive analytics enable the development

of new treatment models that can adapt to

new information as it is uncovered, with

subsequent forecasts and trials becoming

increasingly accurate as the volume of data

expands.

As discussed earlier, an important first step to

achieving a precision medicine approach is

to utilize a clinical technology platform that

ingests a wide variety of phenotypic data,

integrates the data into standardized data

models, and unifies these data models with

analytic technologies. This is a key determinant

of an optimized precision medicine approach.

Huge volumes of data, poorly standardized data

models, or integration challenges between data

models and analytical technologies can quickly

overwhelm ill-equipped data centers and

business processes, and without an adequate

infrastructure in place, the entire platform can

collapse, leading to lost time, money, and even

valuable data.

The platform should provide these capabilities,

which we list and then discuss in more detail:

l Continuous collection, aggregation, and

integration of myriad data sources.

l Adaptive and iterative study design and

execution as new information is made

available.

l Maintenance and management of data

throughout the study life cycle.

l Advanced analytics for research discovery

during the study and post-study.

3.1. Continuous Collection, Aggregation, and

Integration of Myriad Data Sources

To benefit fully from insights available with a

big data approach, massive amounts of data

must be collected, aggregated, and integrated,

either in real time (e.g., sensor data) or from

archives (e.g., unstructured data such as

point-of-care notes, histological images,

and photos). By capturing a wider variety of

non-traditional data derived as close to the

source as possible, big data offers amazing

opportunities for precision medicine. Below

are brief descriptions of the types of data that

are integral to informing a precision medicine

approach.

Laboratory data and vital signs continue to

serve as crucial data points. These data provide

temporally important insights; for example,

blood work can evaluate the status of a

biomarker to aid in diagnosis or prognosis of

a disease.

With the advent of NGS, genetic analyses can

have increasing impact on precision medicine;

NGS can simultaneously target hundreds of

genes that are known to be important for a

disease (Roychowdhury, 2016). The genetic

testing firm Concert Genetics estimates that

there are more than 65,000 genetic testing

products available on the market, with 10 new

products introduced per day. As additional

genetic data are verified, this will allow for

the definition of more precise stratification

of patient subsets and better assessments of

interactions between genotypes and clinical

phenotypes derived from biobank data (Azimi,

2016). New high-quality genome assemblies

are increasingly offering ethnicity-specific

reference sequences that can be used to

improve precision medicine studies in certain

populations, such as Asians (e.g., Seo, 2016;

Qiao, 2016; Shi, 2016).

Imaging data are a medically rich source that

can also be a resource-intense burden. Much

of the available technology is unable to analyze

these data effectively without careful deep

learning or neural network techniques (Dinov,

2016; Geraci, 2017). Continual advances in this

realm are generating new ways to use technol-

ogy to evaluate and even interpret images.

In histology, for instance, much of the routine

work of diagnosis follows algorithmic decision

trees. Recent studies have reported success in

using deep learning to increase the accuracy

and efficiency of histopathological diagnoses

(Djuric, 2017; Litjens, 2016). In another example,

deep convolutional neural networks were

shown to have potential for general and highly

variable tasks across many fine-grained object

categories: investigators demonstrated artificial

intelligence (AI) to be capable of classifying

skin cancer with a level of competence

comparable to dermatologists (Esteva, 2017).

Digital biomarkers represent a valuable stream

of citizen-generated data from sensors

embedded in a variety of devices (e.g.,

wearables, implantables, and smartphones).

These digital biomarkers can play a vital role

55

PR

EC

ISIO

N T

OO

LS

56

in helping to identify more precise subsets

of patients, based on their ability to provide

longitudinal data in on a real-world setting (as

opposed to a single snapshot in time in

a doctor’s office). More advanced intuitive

designs in microelectromechanical systems

are providing researchers with abundant data

to aid the precision medicine efforts for a

variety of phenotypes of interest (Redmond,

2014). In addition to health-related data, a

variety of behavioral and social data can also

be used to explore social networks and social

determinants of health for patient outcomes.

Prior to initiating a precision medicine trial,

sponsors must develop their study designs

based on the types of data that will be

collected. Sponsors must also be prepared to

address data quality deficiencies that result

from aggregating disparate data sources with

varying degrees of structure, and they must

be able to integrate data quality assessment

functionality to ensure the ongoing validity

and reproducibility of findings. Developments

in AI are expected to make data collection and

integration even more efficient in the near

future. For example, AI is actively being used

in the form of chatbots for patient triage (e.g.,

Microsoft’s Health bot), and this technology is

also being explored to support the burdensome

process of creating clinical notes and pheno-

typing disease (Deliberato, 2017; Geraci, 2017).

As data streams become increasingly structured

and streamed in real time, the precision

medicine approach will become even more

efficient and accelerated.

3.2. Adaptive and Iterative Study Design

and Execution as New Information is Made

Available

Precision medicine studies must offer

investigators the capability to adapt study

designs throughout the life cycle of the study,

based on newly added information or

newly discovered relationships in the data.

The precision medicine approach is initially

exploratory, and it requires iteration to

redefine patient clusters into more precise

subsets and to redesign studies to enable

refined hypothesis testing.

Investigators must find the right balance

of the number of initial hypotheses that

will be tested (i.e., exploratory analyses),

since too many hypotheses can be viewed

as unfavorable and lacking a defined research

objective. Likewise, too many post hoc

analyses constitute data dredging. While,

understanding causal relationships in data

to surface actionable insights that are truly

governing improvements can have a significant

impact. There is a need to use appropriate

methods, avoiding model misspecifications,

and careful selection of data to avoid issues

of endogeneity, which can manifest from rate

dependence or state dependence.

Methodological innovations in trial design

include the ability to evaluate multiple

treatments in more than one patient type or

disease within the same overall trial structure

(Woodcock, 2017). The key to this approach

is to generate a master protocol, a single

overarching protocol designed to answer

multiple questions (e.g., studies involving

multiple drugs and/or diseases) and to enable

the use of “innovative statistical approaches to

study designs and analysis, enabling a broader

set of objectives to be met more effectively

than would be possible in independent trials”

(Woodcock, 2017). The master protocol can

accommodate different clinical trial designs,

including umbrella, basket, and platform trials

(Woodcock, 2017).

Key tasks in designing a precision

medicine study:

l Generate the protocol — establish solid

scientific foundation for predictive

precision medicine trial via pre-protocol

literature research to derive several initial

hypotheses to be tested (may result in

generating a new master protocol or

capitalizing on existing trial infrastructures).

l Plan adequately for the complex trial

designs and real-time decision making that

is the foundation of precision medicine

trials, including a basis for decisions about

treatments to pursue and those to

discontinue.

l Consider a wide array of data types and

depth to be measured, based on the range

of relevant factors and outcomes.

l Construct the infrastructure required to

handle data flow, data quality, data sharing,

trial logistics, and real-time ability to analyze

trial data and redesign trials according to

new information.

l Evaluate innovative methods for analyses

and recognize the care required to interpret

results.

After the study objectives and measurable

outcomes have been finalized in the

protocol, downstream analyses may

leverage some of these clinical trial

designs:

l Longitudinal cohort studies: Precision

medicine cohorts can be created to

evaluate the relationships between genotype

and phenotype over time, such as the NIH

sponsored program All of Us (Collins, 2015;

National Institutes of Health, 2017).

l Basket trials: Enable a larger recruitment

pool, since this design tests whether a drug is

effective in patients characterized by a single

genetic alteration and whether patients may

have tumors of different histologic types and

tissue of origin (Simon, 2017; Redig, 2015;

American Society of Clinical Oncology, 2017).

For instance, the National Cancer Institute

Molecular Analysis for Therapy Choice

(NCI-Match) is a Phase II clinical trial in

which patients who share a common genetic

mutation for a given cancer are sorted into

“baskets,” or treatment arms, regardless of

cancer type.

l Umbrella trials: Assign patients to a

particular treatment arm based on their

cancer type and genetic marker. This design

can have many treatment arms. The Lung

Master Protocol (Lung-MAP) study aims to

rapidly identify drug therapies for particular

cancer types, making it a useful design for

cancers with wide genetic heterogeneity

(Woodcock, 2017).

l N-of-1 studies: In these trials, which are

often used in evaluation of musculoskeletal

or pulmonary conditions, a patient serves

as both control and experimental treatment.

Although seemingly counterintuitive to the

goal of precision medicine, which typically

harnesses large amounts of data, this design is

especially useful for low-frequency, rare

conditions (Schork, 2015; Duan, 2013; Lillie,

2011).

l Adaptive trials: Investigators can adapt study

goals based on emerging data, thus provoking

a revision of hypotheses to be tested

(Heckman-Stoddard, 2014). Treatment arms

can be altered or biomarker strategies adjusted

depending on drug efficacy data (Woodcock,

2017). This paradigm is potentially conducive

to more rapid drug approvals for precision

medicine (Printz, 2013).

Figure: Schematic highlighting the flow of umbrella and basket trials (Redrawn from Woodcock, 2017).

57

PR

EC

ISIO

N T

OO

LS

59

PR

EC

ISIO

N T

OO

LS

The precision medicine approach requires

rigorous pre-trial discussion between sponsors,

clinical trial management firms, and the

governing regulatory agency to ensure

alignment with the protocol; therapeutic

approaches that will be tested to improve

patient outcomes; and timing of regulatory

submission filing (Woodcock, 2017;

Schwaederle, 2015).

Precision medicine trials necessitate a steadfast

approach with considerable advance planning,

because the protocols and trial designs can be

quite complex.

3.3. Maintenance and Management of

Data Throughout the Study Life Cycle

The complexities underlying precision

medicine trials require substantial planning

for data maintenance and management over

the entire life cycle of the study. The highly

adaptive nature of precision medicine trials

demands that high-quality data be maintained

throughout the trial.

“The use of a single system for clinical data management will enable shorter start-up times as the protocol is expanded to incorporate new investigations. The use of a single central randomization system facilitates the addition of new therapies with minimal disruption. Real-time access to the genomic, proteomic, pathological, and imaging data streams is requisite for the adaptive features of I-S P Y 2 [Neoadjuvant and Personalized Adaptive Novel Agents to Treat Breast Cancer].”(Woodcock, 2017)

The infrastructure for logistical efficiencies

that establish optimal data quality is a

necessary component for the success of a

precision medicine trial. Engaging in early

discussions with experts in clinical trial

technologies can help establish the trajectory

for success.

3.4. Advanced Analytics for Research

Discovery During the Study and Post-study

Precision medicine trials rely heavily on

clinicians and researchers to consider emerging

data in new ways and to develop innovative

strategies to fully leverage the massive body

of data generated.

Promoting high-quality data relies on stringent,

consistently applied definitions of data quality

as well as quality assurance and control

procedures to ensure the validity of the

incoming data and reproducibility of results.

Data standards should be maintained to assess

quality on a variety of dimensions (Kahn, 2016;

Weiskopf, 2013). The breadth of information

generated in precision medicine often warrants

the development of procedures and statistical

models and algorithms to ensure rapid and

uninterrupted data flow (Woodcock, 2017;

Garraway, 2013).

60

“The collection and processing of biomarker samples is extremely important to the integrity and validity of the genetic data used in a precision medicine trial. Optimized clinical trial

technologies use a combination of risk indicators and advanced pattern detection, to identify sites that may be lacking in the

collection and processing of samples according to defined protocol procedures. The faster any quality issue with

biomarker samples can be identified, the faster mitigation strategies can be deployed.”

Both during the study and following study

close-out, advanced analytics can help uncover

newly observed patient behaviors, biomarkers

of response, new subgroups of responsive

patients, new diagnostic criteria or risk factors,

novel treatment paradigms, etc. This process

can result in revised or even completely novel

studies with increasingly focused hypotheses

to be tested.

Post hoc or exploratory subgroup examination

has traditionally been criticized for several

well-documented reasons, including selection

bias. However, developments in advanced

analytics — including machine learning, data

mining, and statistical methods — emphasize

critical statistical principles and the importance

of performing multiplicity adjustments to

adjust for selection bias inherent in subgroup

search. Some of these techniques also treat

subgroup investigation as a special case of

model selection (Lipkovich, 2017; Benjamini,

2005). Such advances are favorable to the type

of post-study analyses required to generate

novel insights to drive equally novel precision

medicine approaches.

Figure: The utility of biomarkers in post-study analyses (Redrawn from Boyapati, 2016).

62

“Machine learning algorithms have invigorated post-study analyses by minimizing biases while allowing

for novel insights that can be subsequently tested in a prospective trial.”

In planning a precision medicine trial, proper

consideration must be given to the clinical

development software platform that will be

used for post-study analytics. The software

should be user-friendly and integrated with

the existing trial infrastructure, to provide

for seamless workflows and to avoid the hefty

costs and extra work required to integrate

disparate solutions and adjusting to the various

environments.

4. Summary

Precision medicine is positioned to revolutionize

the healthcare landscape, and the utilization

of advanced clinical development technologies

is key to supporting precision medicine trials.

These technologies can harness a variety of

disparate data sources at richer and greater

volumes to build more accurate predictive

models for different patient subsets.

As the number of data streams increases across

the continuum of clinical care, the promise

is that precision medicines will become

increasingly targeted. However, the complexities

associated with evolving study designs are

substantial, and the precision medicine

approach demands solid planning, investment,

and commitment from all stakeholders.

An integrative platform should be designed

to house the clinical data; streamline

standardized workflow in data collection,

analysis, and iterative hypothesis testing; and

provide a satisfactory clinical user experience.

Our need to evaluate exploratory hypothesis

testing requires the thoughtful construction

of innovative clinical trials that are adaptive

to data generated in real time. A platform that

allows an iterative research process increases

the likelihood of success for sponsors

embarking on a precision medicine approach.

The promise of safer, more effective treatments

through targeting of specific patient subgroups

with the most advantageous benefit-risk profile

is what propels researchers on the precision

medicine journey.

References

American Society of Clinical Oncology, 2017. Clinical Trial Design and Methodology. [Online] Available at: https://www.asco.org/research-progress/clinical-trials/clini-cal-trial-resources/clinical-trial-design-and-methodology [Accessed August 24 2017].Anderson, V. H., Weintraub, W. S., Radfod, M. J., Kremers, M. S., Roe, M. T., Shaw, R. E., … Tcheng, J. E. (2014). Standardized Cardiovascular Data for Clinical Research, Registries, and Patient Care: A Report from the Data Stand-ards Workgroup of the National Cardiovascular Research Infrastructure Project. A collaboration of the Duke Clinical Research Institute and th, 61(18), 1835–1846. https://doi.org/10.1016/j.jacc.2012.12.047.StandardizedAzimi, M., Schmaus, K., Greger, V., Neitzel, D., Rochelle, R., & Dinh, T. (2016). Carrier screening by next-generation se-quencing: health benefits and cost effectiveness. Molecular Genetics & Genomic Medicine, 4(3), 292–302. https://doi.org/10.1002/mgg3.204Bartlett, J. M. S., & Parelukar, W. (2017). Breast cancers are rare diseases—and must be treated as such. NPJ Breast Cancer, 3, 11. https://doi.org/10.1038/s41523-017-0013-yBeckmann, J. S., & Lew, D. (2016). Reconciling evi-dence-based medicine and precision medicine in the era of big data: challenges and opportunities. Genome Medicine, 8, 134. https://doi.org/10.1186/s13073-016-0388-7Benjamini, Y., & Leshno, M. (2005). Statistical methods for data mining. In Data Mining and Knowledge Discovery Handbook (pp. 565-587). Springer US.Boyapati, R. K., Kalla, R., Satsangi, J., & Ho, G. T. (2016). Biomarkers in search of precision medicine in IBD. The American journal of gastroenterology, 111(12), 1682-1690.Buzuk, G., Lojko, D., Ow, M., & Ru, M. (2016). Depression with atypical features in various types of affective disor-ders, 50(4), 827–838.Collins, F. S., & Varmus, H. (2015). A New Initiative on Pre-cision Medicine. New England Journal of Medicine, 372(9), 793–795. https://doi.org/10.1056/NEJMp1500523Deliberato, O. R., Celi, A. L., & Stone, J. D. Clinical Note Creation, Binning, and Artificial Intelligence. JMIR Med Inform, 5(3), e24. https://doi.org/10.2196/medinform.7627Dinov, I. D. (2016). Methodological challenges and analytic opportunities for modeling and interpreting Big Healthcare Data. GigaScience, 5(1), 12. https://doi.org/10.1186/s13742-016-0117-6Djuric, U., Zadeh, G., Aldape, K., & Diamandis, P. (2017). Precision histology: how deep learning is poised to revi-talize histomorphology for personalized cancer care. npj Precision Oncology (Vol. 1). https://doi.org/10.1038/s41698-017-0022-1Duan, N., Kravitz, R. L., & Schmid, C. H. (2013). Single-pa-tient (n-of-1) trials: a pragmatic clinical decision meth-odology for patient-centered comparative effectiveness research. Journal of clinical epidemiology, 66(8), S21-S28. Dumitrescu, L., Ritchie, M. D., Denny, J. C., El Rouby, N. M., McDonough, C. W., Bradford, Y., … Crawford, D. C. (2017). Genome-wide study of resistant hypertensionidentified from electronic health records. PLoS ONE, 12(2), e0171745. https://doi.org/10.1371/journal.pone.0171745Esteva, A., Kuprel, B., Novoa, R. A., Ko, J., Swetter, S. M., Blau, H. M., & Thrun, S. (2017). Dermatologist-level classification of skin cancer with deep neural networks. Nature,542(7639), 115–118. https://doi.org/10.1038/nature21056Evans, R. S. (2016). Electronic Health Records: Then, Now, and in the Future. Yearbook of Medical Informatics, Suppl 1, S48-61. https://doi.org/10.15265/IYS-2016-s006

FDA. (2010). Guidance for Industry Adaptive Design Clin-ical Trials Guidance for Industry Adaptive Design Clinical Trials. Available at: https://www.fda.gov/downloads/drugs/guidances/ucm201790.pdf FDA. (2016). Precision Medicine Initiative. [Online] Availa-ble at: https://www.fda.gov/scienceresearch/specialtopics/precisionmedicine/default.htm [Accessed August 22 2017].Field, D., & Sansone, S. (2006). A Special Issue on Data Standards. OMICS: A Journal of Integrative Biology, 10(2), 84–93.Ford, E., Carroll, J. A., Smith, H. E., Scott, D., & Cassell, J. A. (2016). Extracting information from the text of electron-ic medical records to improve case detection: a systematic review. Journal of the American Medical Informatics Asso-ciation : JAMIA, 23(5), 1007–1015. https://doi.org/10.1093/jamia/ocv180Garraway, L. A., Verweij, J., & Ballman, K. V. (2013). Preci-sion Oncology: An Overview. Journal of Clinical Oncology, 31(15), 1803–1805. https://doi.org/10.1200/JCO.2013.49.4799Geraci, J., Wilansky, P., de Luca, V., Roy, A., Kennedy, J. L., & Strauss, J. (2017). Applying deep neural networks to unstructured text notes in electronic medical records for phenotyping youth depression. Evidence-Based Mental Health. https://doi.org/10.1136/eb-2017-102688Godet, I., & Gilkes, D. M. (2017). BRCA1 and BRCA2 muta-tions and treatment strategies for breast cancer. Integrative cancer science and therapeutics, 4(1). Heckman-Stoddard, B. M., & Smith, J. J. (2014). Precision medicine clinical trials: defining new treatment strategies. Seminars in Oncology Nursing, 30(2), 109–116. https://doi.org/10.1016/j.soncn.2014.03.004Kahn, M. G., Callahan, T. J., Barnard, J., Bauck, A. E., Brown, J., Davidson, B. N., ... & Liaw, S. T. (2016). A harmonized data quality assessment terminology and framework for the secondary use of electronic health record data. eGEMs, 4(1).Lillie, E. O., Patay, B., Diamant, J., Issell, B., Topol, E. J., & Schork, N. J. (2011). The n-of-1 clinical trial: the ultimate strategy for individualizing medicine?. Personalized medi-cine, 8(2), 161-173.Lipkovich, I., & Dmitrienko, A. (2017). Tutorial in biosta-tistics: data‐driven subgroup identification and analysis in clinical trials. Statistics in medicine, 36(1), 136-196. Litjens, G., Sánchez, C. I., Timofeeva, N., Hermsen, M., Nagtegaal, I., Kovacs, I., … van der Laak, J. (2016). Deep learning as a tool for increased accuracy and efficiency of histopathological diagnosis, 6, 26286. Retrieved from http://dx.doi.org/10.1038/srep26286Marchina, E., Fontana, M. G., Speziani, M., Salvi, A., Ricca, G., Di Lorenzo, D., … Barlati, S. (2010). BRCA1 and BRCA2 genetic test in high risk patients and families: counselling and management. Oncology Reports, 24(6), 1661–1667.Mansfield, E. Precision Medicine at the FDA [PDF docu-ment]. Retrieved from FDA Web site available at: https://www.fda.gov/downloads/advisorycommittees/commit-teesmeetingmaterials/scienceboardtothefoodanddrugad-ministration/ucm473506.pdf McIntyre, R. S. (2016). Implementing treatment strategies for different types of depression. The Journal of Clinical Psychiatry, 77 Suppl 1, 9–15. https://doi.org/10.4088/JCP.14077su1c.02Mead, C. N. (2006). Data interchange standards in health-care IT--computable semantic interoperability: now possi-ble but still difficult, do we really need a better mousetrap? Journal of Healthcare Information Management : JHIM, 20(1), 71–78.

National Academies of Science. (2017). EnabliNational Insti-tutes of Health, 2017. All of Us. [Online] Available at: https://allofus.nih.gov/ [Accessed August 23, 2017].Phillips, K. A., Deverka, P. A., Sox, H. C., Khoury, M. J., Sandy, L. G., Ginsburg, G. S., … Douglas, M. P. (2017). Making genomic medicine evidence-based and patient- centered : a structured review and landscape analysis of comparative effec-tiveness research, (April). https://doi.org/10.1038/gim.2017.21Printz, C. (2013). I-SPY 2 may change how clinical trials are conducted. Cancer, 119(11), 1925–1927. https://doi.org/10.1002/cncr.28172Qiao, W., Yang, Y., Sebra, R., Mendiratta, G., Gaedigk, A., Desnick, R. J., & Scott, S. A. (2016). Long‐read single molecule real‐time full gene sequencing of cytochrome P450‐2D6. Human mutation, 37(3), 315-323.Redig, A. J., & Jänne, P. A. (2015). Basket Trials and the Evolu-tion of Clinical Trial Design in an Era of Genomic Medicine. Journal of Clinical Oncology, 33(9), 975–977. https://doi.org/10.1200/JCO.2014.59.8433ng Precision Medicine: The Role of Genetics in Clinical Drug Development: Proceedings of a Workshop. Washington (DC): The National Academies Press. https://doi.org/10.17226/24829National Human Genome Research Institute, 2016. New train-ing grants prime doctors to tackle genomic medicine. [Online] Available at: https://www.genome.gov/27565602/2016-news-feature-new-training-grants-prime-doctors-to-tackle-genom-ic-medicine/ [Accessed August 21, 2017].Redmond, S. J., Lovell, N. H., Yang, G. Z., Horsch, A., Lukow-icz, P., Murrugarra, L., & Marschollek, M. (2014). What Does Big Data Mean for Wearable Sensor Systems?: Contribution of the IMIA Wearable Sensors in Healthcare WG. Yearbook of Medical Informatics, 9(1), 135–142. https://doi.org/10.15265/IY-2014-0019Richesson, R. L., & Krischer, J. (2007). Data Standards in Clinical Research: Gaps, Overlaps, Challenges and Future Directions. Journal of the American Medical Informatics Asso-ciation, 14(6), 687–696. https://doi.org/10.1197/jamia.M2470Roychowdhury, S., & Chinnaiyan, A. M. (2016). Translating cancer genomes and transcriptomes for precision oncology. CA: a cancer journal for clinicians, 66(1), 75-88. Schork, N. J. (2015). Personalized medicine: time for one-person trials. Nature, 520(7549), 609-611.Schwaederle, M., Zhao, M., Lee, J. J., Eggermont, A. M., Schilsky, R. L., Mendelsohn, J., … Kurzrock, R. (2015). Impact of Precision Medicine in Diverse Cancers: A Meta-Analysis of Phase II Clinical Trials. Journal of Clinical Oncology, 33(32), 3817–3825. https://doi.org/10.1200/JCO.2015.61.5997Seo JS, Rhie A, Kim J, Lee S, Sohn MH, Kim CU, …, Kim C. (2016). De novo assembly and phasing of a Korean human genome. Nature. Oct 13;538(7624):243-247.Shi, L., Guo, Y., Dong, C., Huddleston, J., Yang, H., Han, X., ... & Lintner, K. E. (2016). Long-read sequencing and de novo assembly of a Chinese genome. Nature communications, 7. Simon, R. (2017). Clinical Trials in Precision Medicine. In Verma, M. & Barh, D. Editors, Progress and Challenges in Precision Medicine (209-2015). New York: Elsevier.Slamon, D. J., Leyland-Jones, B., Shak, S., Fuchs, H., Paton, V., Bajamonde, A., … Norton, L. (2001). Use of Chemotherapy plus a Monoclonal Antibody against HER2 for Metastatic Breast Cancer That Overexpresses HER2. New England Journal of Medicine, 344(11), 783–792. https://doi.org/10.1056/NEJM200103153441101Strawbridge, R., Young, A. H., & Cleare, A. J. (2017). Biomark-ers for depression: recent insights, current challenges and future prospects. Neuropsychiatric Disease and Treatment, 13, 1245–1262. https://doi.org/10.2147/NDT.S114542Szlosek, D. A., & Ferrett, J. (2016). Using Machine Learning and Natural Language Processing Algorithms to Automate the Evaluation of Clinical Decision Support in Electronic Medical Record Systems. eGEMs, 4(3), 1222. https://doi.org/10.13063/2327-9214.1222Torkamani, A., Andersen, K. G., Steinhubl, S. R., & Topol, E. J. (2017). High-Definition Medicine. Cell, 170(5), 828-843.Woodcock, J., & LaVange, L. M. (2017). Master Protocols to Study Multiple Therapies, Multiple Diseases, or Both. New England Journal of Medicine, 377(1), 62-70. http:// 10.1056/NEJMra1510062Yue, Y., Jiang, H., Yin, Y., Zhang, Y., Liang, J., Li, S., … Yuan, Y. (2016). The Role of Neuropeptide Y mRNA Expression Level in Distinguishing Different Types of Depression. Frontiers in Aging Neuroscience, 8, 323. https://doi.org/10.3389/fnagi.2016.00323

Steven Schwager, PhD, is a Research Principal in the

Data Science group at Medidata Solutions and a Professor

Emeritus of Statistics and Biological Statistics at Cornell

University. He joined the faculty at Cornell after earning

his PhD in statistics at Yale University. His research

includes statistical methodology and theory and

collaborative projects in a wide range of biological

and physical sciences and other fields. He works at

Medidata on developing statistical and analytical

methods for improving the design, execution, and

analysis of clinical trials, including Clinical Trial

Genomics, operational analytics, and synthetic control

arms.

Jeffery Wiser, recently joined Medidata as Genomics

Portfolio Director to grow the new Clinical Trial Genomics

(CTG) product. Jeff has spent his 20 year career leading

the development of life sciences software with a focus

on the integration of clinical and ‘omics’ data. For the

last 10 years, he has served as the Program Manager or

Co-Investigator on multiple National Institute of Health

funded projects, most notably the ImmPort project, a

public/private partnership with UCSF and Stanford

University. ImmPort serves as a repository for the

collection, QC, curation and sharing of clinical and

mechanistic assay results generated by researchers

funded by the National Institute of Allergy and Infectious

Diseases (NIAID), and was recently designated a Nature

Scientific Data recommended repository for immunology

and cytometry. Previously, Jeff served for 9 years in

management roles at Gene Logic Inc. building clinical

data capture, biobanking, LIMS, and gene expression

analysis software for what was at the time the largest

microarray gene expression database in the world. He

holds a MSE from the Unversity of Pennsylvania in

Bioengineering and a BS from Cornell University in

Mechanical Engineering.

Therese Dolan is a Global Marketing Director

overseeing Precision Medicine and Study Conduct

for Medidata Solutions. She joined Medidata after

spending most of her 20 year career in Life Sciences

working in global marketing for the Commercial and

the Licensing Divisions for Merck & Co. Therese has

had the opportunity to work with some of the leading

scientific experts in the field for multiple disease states

discussing the future state of science in terms of study

and treatment, which often included some strand of

what has now become the topic of Precision Medicine.

The passion behind these discussions have carried

forward as Therese works towards supporting the

industry in achieving its vision of Precision Medicine

through use of technology for clinical trials.

Ruthie Davi, PhD. has worked in the area of

pharmaceutical product development, and specifically

the FDA, over 20 years. In her current position as a

Director in Data Science at Medidata Solutions , she

is part of a team focused on providing centralized

systematic monitoring and other novel clinical trial

tools. Ruthie has held numerous positions within the

FDA, including Statistical Reviewer, Team Leader, and

Deputy Director, where she implemented innovative

statistical methods and made recommendations for

clinical trial designs tailored to the regulatory setting.

She has conducted the statistical review, represented

the Agency at Advisory Committee meetings, and

provided recommendations regarding FDA marketing

approval for numerous New Drug and Biologic Licensing

Applications. With a particular interest in pediatric

clinical trials, she was an active member of FDA’s

Pediatric Review Committee. Ruthie holds a Ph.D. in

biostatistics from George Washington University.

63

PR

EC

ISIO

N T

OO

LS