| |
About Scientific Consulting
|
Balancing knowledge needs, scientific rigor,
time lines, and budgetary realities - four challenges faced as much in
industry as in private/academic settings.
In Biopharma Research
Unfortunately, biopharma phase 4
and 5 studies are often criticized on methodology and credibility.
True, there have been "quick-and-dirty" studies: poorly designed,
underpowered, ambiguous data models, inappropriate statistical analyses,
questionable results reporting, and opportunistic communications. Just as true, the majority of phase
4 and 5 studies have significant scientific merit - perhaps not always at
par with registration studies, but certainly at par with most
investigator-initiated and academic studies.
The issue is not whether Phase 4 and 5
studies are just as good as Phase 3 studies. Instead, the issue is
which approach is most appropriate for the strategic and clinical
purpose(s). Pre-registration and
post-marketing studies are distinctly different activities
along the scientific lifecycle of a compound. Think, however, what you would
not know if you only had the constrained data sets of controlled
trials - but not the "real world" depth and breadth of data on practice
patterns and outcomes. Think, for a moment, how much you would be
able to say about patients if all you had was knowledge from relatively
homogeneous trial samples - but no access to large databases of
patients within and across full ranges of disease states. Think,
finally, about the new indications you may not have discovered - and the
"back-to-the-future" opportunity you would have missed: no real world data
to launch new registration trials for supplemental registrations.
Biopharma studies fall in different classes
- each with different levels of tolerance for design and data quality, yet
all sharing a common quality threshold. The task is to balance
purpose with scientific rigor - yet always to design studies that will
stand up to the scientific standards of tits class. You will not
hear us advocate (and certainly not see us do) poor research.
Instead, we work with our clients on finding the right scientific strategy
and putting the markers in place for quality and tolerance.
This is the scientific consulting expertise
you can expect from
MATRIX45:
solid science, pragmatically
adapted to strategic purpose, time lines, and budgetary realities.
We may be able to recommend one particular scientific approach, or present
you with different scenarios and their respective advantages and
disadvantages.
In Private/Academic Research
The principles of scientific rigor cross all
settings of healthcare research. In their own way,
academic/private researchers confront similar challenges of balancing
knowledge needs, scientific merit, timelines, and budgets. On the
other hand, private/academic research is often different from biopharma
research in intent: basic research, whether at the laboratory bench or in
the clinical field, to find determinants, identify processes, examine
models, or test new methods.
Here too
MATRIX45
brings valuable expertise. Some of us
started in academic research, some of us are still involved in it.
We know firsthand the conceptual, methodological, and statistical
challenges. We too have had to balance the relevance (and at times
urgency) of major scientific questions with realistic designs - and with
funding constraints (see also our section on
Development & Funding).
At the Interface
There is an added dimension in healthcare
research that interfaces basic and applied research. Conducting
clinical biopharma research requires efficient and productive
collaboration between the two sectors. Biopharma needs top academic
scientists who are at the forefront of both basic and applied research.
Academic/private researchers need the collaborative (and funding)
mechanisms that enable them to answer new clinical questions.
Again,
MATRIX45
staff bring experience in both
arenas. The research careers of some of us have encompassed the
basic-to-applied continuum and the interface between sectors. We'd
like to believe it sets us apart.
Top of Page
Top of Section |
Across Paradigms And
Methodologies
|
In absolute terms the randomized controlled
trial (RCT) is the gold standard for assessing the efficacy of a treatment
in individuals - whether a drug treatment, a nonpharmacological
intervention, or a healthcare delivery model.
Is the RCT the only valid model, i.e. the
only model that will help us understand the antecedents, determinants, and
outcomes of interventions as they exist "out there"? In other words,
do RCTs yield the necessary and sufficient knowledge
to safely and effectively improve healthcare to patients, families, and
communities? What if RCTs are just not possible; either because too
many confounds cannot be neutralized, ethical issues prevail, multilevel influences must be
examined, randomization is impossible, or costs are plainly prohibitive?
Just to add another dimension ... The
quality of any study is determined to a good extent by the quality of
measurement of its endpoints or outcomes. In the physiological
arena, this may not be as much of a problem: while (surrogate) markers may
not be perfect, variability in measurement is more often due to laboratory
(and equipment) differences than "true" errors in measurement.
Things get more difficult when measuring
behavioral, emotional, or cognitive attributes. Just the mere
plethora of measures for particular traits, attributes, or conditions
(let's call them human characteristics) underscores the challenges of
measurement. Should we measure these characteristics as
comprehensively and completely as possibly, while also being able to zoom
in on particular dimensions (the case for large batteries)? Should
we instead focus on simple but accurate tools that screen well and give us
a solid overall assessment (the case for quick, easy-to-use, yet reliable
tools)? Perhaps a cascading model is appropriate, where we start out
with a top-level screening assessment and, based on embedded triggers,
move on to more in-depth scales and batteries? Or is there a step
(or cascading level) in between, where screening tools provide more than
top-level evaluation and enable clinical researchers to identify a
patient's dimensions of behavior, emotion, and cognition that may require
further assessment - without imposing the up-front burden of large
psychometric batteries.
How do we at
MATRIX45
think about these issues?
Without questioning the gold standard status
of the RCT, we are convinced that a comprehensive and well-rounded
lifecycle program of research may require
the combination and integration of various paradigms and methodologies.
Indeed,
multiple questions inevitably raise the probability of multiple
methodologies - and choices will need to be made.
Further, by necessity RCTs constrain the
variability at the "left side" of the equation. The left side is
clear cut, unambiguous, and under control: "each patient gets something
versus nothing or something else." RCTs also constrain the "right
side" of the equation: the efficacy endpoints to be achieved within a
limited time period, and potential safety issues during that same time period.
Consequently, many questions go unanswered
even as the efficacy and safety of an intervention are being established. Are the controlled conditions of the RCT replicable in
day-to-day clinical practice? Which patients benefit a lot,
which less, and which not at all? Where and how may variability in
clinical practice occur? What outcomes are associated with this
variability?
Even more questions ... Can we detect new
applications for this intervention: from new indications for a given drug
to new patient populations for a given healthcare delivery model? If
clinical guidelines are available, how do actual practice patterns meet
the treatment standards and targeted outcomes advocated by these
guidelines?
Comprehensive and well-rounded research
programs on the lifecycle of a clinical intervention (a drug, an
integrated healthcare solution, a healthsystems delivery model, ...) will
require the use of multiple paradigms and methodologies. At
MATRIX45,
we are prepared to assist clients in the various challenges involved in this
process: designing lifecycle research programs, with special emphasis on
post-RCT studies; bringing together the various stakeholders in such
programs - scientific, strategic, and clinical; and implementing and
disseminating the knowledge generated by such lifecycle programs. We
are experienced in trials, epidemiological studies, measurement
validation, and patient registries. We have used both controlled and
noncontrolled methods to examine causality. We have looked back
on data, and we have used them to project into the future. At times,
we even have abandoned "hard" quantitative methods to find answers that
only "soft" qualitative methods might yield.
A note on the measurement of behavioral,
emotional, and psychosocial attributes... The field of
psychodiagnostics is too broad, too diverse, and too deep for anyone to
master (according to our academic psychodiagnostic experts). This
has forced us to make some choices at
MATRIX45.
One, we would like to move beyond the use of relatively simple rating
scales in clinical research, even if some have attained "gold
standard" status (a particular, convenient, not-too-demanding clinician
rating scale for depression comes to mind). Two, we want to fill the
gap between top-level screens and "deep-down" batteries, in part by
assessing the extent to which the "screens" may yield some
intermediate-level insights in a patient's psychological capabilities.
Three, regardless of scope and depth, good research depends on good
measurement of attributes that arte difficult to measure - not only in the
psychiatric and neurological domain, but in health behavior in general.
Summarized, at
MATRIX45
you will find the "toolboxes" for successful comprehensive research
programs - in objectives, design, measurement, and analysis. Importantly, you will also find the frameworks and
rationales for knowing when, how, and for what purpose to use the tools in
our toolboxes.
Top of Page
Top of Section
|
Types of Studies And
Investigations |
In our careers, we have design or conducted the following
types of studies and investigations:
Controlled trials
Pharmaco-epidemiologic studies
Observational studies
Practice pattern analysis
Outcomes analysis
Practice <=> outcomes analysis
Conversion trials
Natural-entry trials
Registries
New indicators analysis
Pattern recognition and signal analysis
Instrument development and psychometric
analysis
Causal modeling
Multilevel modeling
Evidence-based practice and outcomes improvement
Surveillance studies
Top of Page
Top of Section
|
Our Network of Experts
|
Our academic affiliations in North America
and Europe, coupled with a network of clinical opinion leaders and
methodological specialists, have enabled us to build a "virtual" expert
staffing model that covers the full range of scientific methodologies and
statistical models. On the other hand, we also have the integrity to
tell prospective clients when we do not have (and cannot find) the
required expertise.
Top of Page
Top of Section
|
Case Studies |
Drawing upon our
principals' career-long scientific consulting work, the following are some
(public domain) examples:
Data Integrity in Phase 4 & 5 Studies
Blockbuster Under Competitive Threat
Core Data Model for
Phases 2 through 5
Assessing Multilevel
Determinants of Treatment Outcomes
Differentiating Dimensions of Depression and Dementia in Older Adults
Top of Page
|
Data Integrity in Phase 4 & 5
Studies |
"Should our [nonregistration]
study have the same rigor of source document verification, monitoring and
auditing as a registration trial?" - a question often asked in
biopharma.
The answer is neither yes nor no. Rigor is important, but needs to be
balanced against such factors as purpose, intended use of data, allowable
margin of error, to name a few. Especially in large international pharmaco-epidemiologic and outcomes studies, the rigor of trials may not
be necessary nor affordable. Thus, we work with clients on
identifying the required boundaries of methodology and data integrity, while
acknowledging the inherent drawbacks of various adjustments. We also
work with clients on developing cost-efficient methods of assuring data
integrity, from random (remote) audits to advanced statistical pattern
detection. Defensibility of the findings, internally or externally,
is the guiding principle.
Return to list of examples
|
Blockbuster Under Competitive Threat
|
A blockbuster drug,
an
early biotechnology success story and in a few years’ time the standard of
care, comes under competitive threat as similar agents gain approval
and reimbursement. To assert its scientific leadership, the
manufacturer decides to sponsor the independent development of best
practice guidelines and to launch a longitudinal survey. It wants to
establish a data base, while also grappling with the strategic and
scientific questions that drive design, implementation, analysis, and
dissemination.
Fast forward through a
hectic 2-month pre-launch period and a 6-month data collection effort, and
here is a 6-month longitudinal database on over 14,000 patients – and the survey expanding from one world region to a global initiative. The
data are analyzed, presented to key opinion leaders, and released for
dissemination – along with the best practice guidelines – less than a
year after the survey was designed. The findings reveal a
significant gap between best and “real world” practice, and documents the
outcomes of various practice patterns. Innovative medical education
and clinician support strategies are developed in response, while
extending the core survey concept into other approved or emerging
indications. All along, sales in the initial study region accelerate
rapidly: clinicians are treating more patients, but
are not necessarily treating them better. Concurrent with
surveys being launched across indications in major markets, global drug
revenues for this drug increase by >275% over 4 years – from blockbuster
to megabuster.
Return to list of examples
|
A Core Data Model For Phases 2-5 |
There is often a gap (if not
chasm) between the data models of pre-registration studies and those of post-marketing studies.
Completing a promising phase 2
study on a potential next-to-market product with a better side-effect
profile, the company sought to develop a continuous data model for phases
3 through 5 - able to accommodate both trial and non-trial studies.
Working with client staff and key opinion leaders, we defined the core
data elements of primary endpoints, secondary endpoints, and outcome
variables to be collected throughout the compound's lifecycle. We
also specified flexible extended data elements that can be added to study
protocols and data projects on an "as-needed" basis. The result is
data continuity: in definition, measurement, and interpretation. In
turn, it enables integration of datasets within and across phases into
meta-datasets. Importantly, it assures consistency and continuity in
interpreting results.
Return to list of examples
|
Assessing Multilevel
Determinants of Treatment Outcomes
|
Imagine a drug trial in
which the efficacy of a new drug is established: a limited but
statistically significant treatment response is observed in the treatment
versus the placebo group. However, there is some uneasiness about
the findings: the "between-groups" results are significant, but there are
placebo and treatment patients responding at similar levels of slight
improvement ("the overlapping distributions" problem)
What might have caused the
overlapping response distribution curves? Of all possible patient assignments into subsamples, randomization might have generated subsamples with inherent
(and often undetectable) biases. Even before randomization, perhaps
inclusion and exclusion criteria were not as tightly specified. But
then, would we want subsamples that are so contrived in their homogeneity
that they fail to represent real-world patients (and cause
excessively restrictive drug labeling)? Endpoints may not have been
operationalized effectively or measured accurately. Data exploration
may not have shed light on some distributional characteristics [etc. etc.
etc.].
Let's examine the
variability/overlap issue from another but complementary angle.
Numerous centers participated in the trial, recruited from many countries
and several continents. Might geography be a proxy for different
cultures, approaches to healthcare, organization of healthcare delivery,
and case mix? Most patients were recruited from academic medical centers or affiliated hospitals,
not community hospitals. Perhaps several clinicians were involved in
screening patients; was experience a factor? Did all centers and
clinicians comply 100% with the study protocol, or might there have been
(known or undetected) deviations? At the end of the "patient supply
chain", who was responsible for recruiting patients, obtaining their
informed consent, orienting them to the study, guiding them through the
protocol, collecting their data over the
duration of the study period, getting patients to answer sensitive questions
(including about their compliance!), and transmitting patient data to the
trial coordination center.
Clinical outcomes, in trials
and in other studies, may be influenced by many factors. Patient
randomization should equalize confounds at the individual patient level.
However, beyond the patient, there are many other factors, at different
levels of organization, that may influence outcomes.
At
MATRIX45,
we believe that in the future both controlled and noncontrolled studies
will increasingly use multilevel analysis models to better understand the
contextual determinants of clinical outcomes.
Return to list of examples
|
Differentiating Dimensions of
Depression and Dementia in Older Adults
|
Depression and dementia are
common clinical problems in outpatient and inpatient populations. In
a systematic program of research, we sought for ways to use common
screening instruments to provide more than an overall assessment of mood
or cognition. We examined whether such instruments had embedded
dimensions of clinical status that could give a more differentiated
screening and guide next-step evaluation. As screening
instruments are often used in (definitive) clinical research, we were also
interested in strengthening these instruments conceptually so that they
could provide more than a yes/no screening answer and instead bring more
differentiated assessment dimensions to the fore.
As the references
here show, widely used scales for depression and dementia enable both
clinicians and researchers to go beyond the "yes/no" screening question.
Embedded dimensions enable that intermediate step in the cascading process
from screening to in-depth assessment by focusing in on key dimension.
Return to list of examples
|
Scientific Consulting Services |
Study / project design using
various paradigms and methodologies
Protocol development
Data model development, from
single study to lifecycle models
Corporate database
development
Measurement, instrumentation,
scaling and (psycho)diagnostics
Statistical analysis
planning
Advisory Board / Expert
Panel / Key Opinion Leader Development and Support
Top of Page
|
|