Principles and Practice of Clinical Trial Medicine

536 Pages·2008·24.06 MB·English

Dedication
This endeavor would not have been possible had we not met as classmates (and dormitory mates) at Harvard Medical
School. This book is dedicated to our families, whose support and encouragement made the endeavor possible. We also
would like to thank everyone at Elsevier for their hard work and invaluable help and making the process an exciting and
enjoyable one.
RYC and BYL
PPrreelliimmss--PP337733669955..iinndddd vv 55//2266//22000088 1122::0055::2299 PPMM
About the Authors
Richard Chin, MD
San Francisco, California, USA
Dr. Chin is a Harvard trained, Board Certifi ed internist with drug development across a wide range of specialties, includ-
ing oncology, immunology, ophthalmology, neurology, endocrinology, and cardiovascular medicine, among others. He
has been responsible for over 40 Investigational New Drug (IND) Applications and 8 New Drug Applications (NDAs)/
Biologic License Applications (BLAs). Some of the drugs Dr. Chin has overseen include ranibizumab, natalizumab, ten-
ecteplase, bapineuzumab, ziconotide, and omalizumab. He has both US and international experience, and has overseen
a wide variety of functions, including Clinical Development, Regulatory, Biostatistics, Quality Assurance/Compliance,
Chemistry Manufacturing, and Control, Safety and Medical Affairs.
Dr. Chin is currently President and CEO of a NASDAQ-listed a biotechnology company. Previously, he was Senior
Vice President and Head of Global Development for Elan Corporation, and before that served as the Head of Clinical
Research for the Biotherapeutics Unit at Genentech. Dr. Chin holds a Medical Degree from Harvard Medical School. He
received the equivalent of a J.D. with honors from Oxford University, England under a Rhodes Scholarship. He graduated
with a Bachelor of Arts in Biology, magna cum laude, from Harvard University. He previously served on the adjunct clini-
cal faculty at Stanford Medical School.
Bruce Y. Lee, MD, MBA
University of Pittsburgh
Pittsburgh, Pennsylvania, USA
Dr. Lee is currently an Assistant Professor of Medicine and Biomedical Informatics, and Epidemiology at the University
of Pittsburgh. He is also Core Faculty in the Section for Decision Sciences and Clinical Systems Modeling, the Center for
Research in Health Care, and the RAND-University of Pittsburgh Health Institute. He is a Co-Investigator for the National
Institutes of Health (NIH) Modeling of Infectious Diseases Agent Study (MIDAS) research and informatics network, and
a Co-Investigator for a Bill & Melinda Gates Foundations Grant on the evaluation of candidate vaccine technologies using
computational models.
His previous positions include serving as Senior Manager at Quintiles Transnational, working in biotechnology
equity research at Montgomery Securities, and co-founding Integrigen, a biotechnology company. His consulting experi-
ence includes a large variety of clients, ranging from large and small pharmaceutical, biotechnology, and medical device
companies.
Dr. Lee has authored three books, as well as numerous research publications, review articles, and book chapters.
Dr. Lee received his B.A. from Harvard University, M.D. from Harvard Medical School, and M.B.A. from the Stanford
Graduate School of Business. He is board-certifi ed in Internal Medicine, having completed his residency training at the
University of California, San Diego.
BBIIOO--PP337733669955..iinndddd iixx 55//2233//22000088 66::5599::3311 AAMM
Introduction
While we may not recognize it, we all use the skills physicians? Pharmaceutical industry professionals? Statis-
necessary to conduct and interpret “ clinical trials ” every ticians? Academics? Clinical research specialists? Regulatory
single day. Sampling and comparing one restaurant, article professionals? Ethicists? Medical students? Nursing stu-
of clothing, television show, fi tness club, vacation loca- dents? Medicine residents? Graduate students? Post-doctoral
tion, date, job candidate, or client to another is effectively fellows? Epidemiologists? Engineers? Pharmacologists?
conducting a clinical trial. Evidence and results from these Pharmacists? Biologists? Pharmaceutical or medical device
mini-trials guide your choices and decisions throughout executives? The more we thought about it, the more we
the day. Should you buy that swimsuit? Is it better than the realized that the audience could be quite broad. Both of our
one in the other store? Are the Philadelphia Eagles or the career journeys have taken us through a variety of func-
Pittsburgh Steelers a better football team? Which sushi tions and domains in industry, academics, and business. We
restaurant has the best salmon? What job will provide the have seen the investment, research, technical, management,
best experience? Where should you live and what house teaching, writing, consulting, and clinical practice realms
should you buy? Is it better to hire Job Candidate #1 or Job of the health care industry. In the end, while each area may
Candidate #2? Faulty “ trial ” design, data, or interpretation have different jargon, cultures, personalities, and perspec-
leads to inaccurate assessments and perhaps poor decisions. tives, the guiding principles are the same. A good clinical
For example, using the wrong criteria will result in hiring trial at an academic institution is a good one in industry and
the wrong person for a job. We all have suffered from that vice-versa.
impulse buy, that forehead-slapping wrong decision, or that Therefore, we wrote this book with a broad audience in
bad choice of friend, employee, or signifi cant other. mind, trying to minimize the jargon and explain any impor-
Formal and informal clinical trials are a large part of our tant terminology in the process. The goal was to write a
lives. If you use, produce, study, purchase, invest in, or con- book that could be easily understood regardless of your
duct research in drugs, medical devices, or any type of health background, especially since people from so many differ-
care intervention, understanding the science and operations of ent backgrounds are involved in clinical trials. In fact, in
formal clinical trials can only help. Today, even understand- many professions, understanding the jargon and terminol-
ing many major news items requires at least some knowl- ogy is half the battle.
edge of clinical trials. Whenever a drug or medical device is Moreover, regardless of your interest and function in
recalled, a medical intervention is debunked, or a new ther- the clinical research world, knowing the general concepts
apy hits the market, clinical trial design, conduct, or analy- of all aspects of clinical trials can be very advantageous.
sis is at the heart of the evidence or the controversy. Health In many ways, the clinical research world has become far
care is such a major business that even seemingly unrelated too specialized. Many individuals stay ensconced within
industries and professions can be dramatically affected by a their areas of knowledge and expertise. But the best clini-
successful or unsuccessful clinical trial. Flaws in a clinical cal researchers or trialists have broad knowledge bases that
trial that force a major drug or device to be pulled from the span statistics, regulatory affairs, ethics, clinical medicine,
market can alter many lives and rock the economy. science, basic probability, data management, and trial and
Therefore, during our planning stages for P rinciples and personnel management. The ones that stand out, are most
Practices of Clinical Trial Medicine , confi ning the book ’ s marketable, and do the best work cannot afford to say, “ I
audience was diffi cult. Should this book be geared toward just do not need to know that because it is not in my area. ”
xi
PPRREE--PP337733669955..iinndddd xxii 55//2233//22000088 66::5599::4433 AAMM
xii Introduction
Designing, conducting, and analyzing a clinical trial is Chapter 9 (Patient Selection and Sampling) reviews consid-
like designing, building, and using a house. Recognizing erations when choosing patients for your trial, and Chapter 10
a house ’ s design and construction helps you realize its (Dosing and Intervention) analyzes how medical interventions
potential use. For example, a thin-walled house may cause should be administered to patients. All of these are as impor-
problems during the winter. Very cramped rooms may not tant factors and parameters to clinical trials as ceiling height,
facilitate hosting a party. At the same time, anticipating the room size, lighting, house temperature, and other features are
house ’ s use aids its design and construction. Your design to house construction. Chapter 11 Epidemiology, Decision
of a beach house likely will differ signifi cantly from your analysis and simulation offers additional tools that may help
design of a farm house or a city dwelling. in the planning and analysis of trials, and is analogous to the
The “ building a house” analogy helps illustrate the gen- model building and “ roughing-in ” phase of house building,
eral organization of our book. The Principles and Practices where you visualize how a house might look.
of Clinical Trial Medicine contains fi ve sections. Section I Section IV covers practical logistical issues involved in
introduces the fi eld of clinical research with Chapter 1 delin- conducting a trial. This is analogous to concerns that arise
eating some general theory and Chapter 2 covering impor- when actually building the house. For example, from where
tant legal, ethical, and regulatory issues. The materials do you procure building materials? Which forms should you
in this section are analogous to all of the rules and regula- complete when ordering such materials? Which nails should
tions that govern the construction of a house: ranging from you use? Where should you place the beams? How do you
general engineering and architectural principles to zoning select and supervise the contractor? All of these types
laws and building codes. Just as you can ’ t build any kind of of issues are discussed in Chapter 12 Study Execution.
house anywhere you choose (e.g., Igloos do not belong in Recruiting Patients and Choosing Trial Locations (Chapter
Philadelphia or San Francisco), you must understand general 13) are such an important part of conducting trials that a
clinical research theory and comply with legal, ethical, and separate chapter is devoted to the topic.
regulatory principles when designing and conducting a trial. Finally, Section V discusses how to analyze the results
Section II focuses on the general design of clinical trials. If of clinical trial. In our building analogy, this is similar
you imagine a clinical trial to be a house, statistics (Chapter 3) to using and inhabiting the house. Data is the output of a
are the tools used to build the house. The fi nal design of the clinical trial, just as a house is the end product of house
house depends heavily on the tools that you have at your dis- construction. Chapter 14, Assessing Data Quality and Trans-
posal. Sure you can rely on others to choose and wield the forming Data, is akin to inspecting the house and making
tools… but would you truly know and trust the house? To be the fi nal adjustments and reworking anything that needs to
truly competent at clinical research, you have to know your be reworked. If the stairs are not to code, they need to be
tools, even if you have specialists to employ them. Measures redone, and if the painters overpainted the moldings, they
and Variables (Chapter 4) are the construction materials for need to be repainted. Data similarly need to the cleaned and
the house. Construction materials help determine the house’ s transformed, to ameliorate missing or unreliable data points.
appearance and utility. Building a house resistant to harsh ele- Chapter 15, Analysis of Data, is akin to decorating the
ments may be diffi cult without good quality bricks or cinder house and moving the furniture into the appropriate rooms.
blocks. Similarly, studying heart disease may be challenging You manipulate the data that has been gathered and prepared.
without accurate echocardiograms, electrocardiograms, and This allows you to then interpret the data, which is the subject
blood pressure measurements. Study Groups (Chapter 5) and of Chapter 16 Data interpretation and conclusions. This is
Periods, Sequences and Design (Chapter 6) are the rooms akin to moving into the house and living in it. This is the acid
and corridors of the house. Changing these will dramatically test. No matter how well-built or well-decorated the house
change the house’ s functionality and purpose. Having no is, if you don’ t enjoy living in it, all has been for naught.
kitchen makes cooking and hosting dinner parties diffi cult. Similarly, the ultimate end product of a clinical trial is a con-
An indoor garage allows you to shield your car from the ele- clusion that is actionable for the treatment of future patients.
ments. Similarly, comparing two medical interventions nor- So whether you are new to the world of clinical trials or
mally requires employing at least two different study groups. have been conducting clinical research for many years, we
Seeing the long-term effects of a drug necessitates patients hope that this book serves you well. The importance and
being on the drug for a long period of time. use of clinical trials will only continue to grow in the future.
Section III takes a closer look at an array of important ele- Concomitantly, trial design and conduct will face increasing
ments in clinical trial design. Endpoints (Chapter 7) are special scrutiny. In many cases, lives of innumerable patients and
measures and variables that serve as the outcomes of the trial. signifi cant amount of time and resources will be riding on
So, continuing our building analogy, endpoints are the key them. Will you be ready?
construction materials that determine the worth, strength, and
use of the house. Chapter 8 (Economics and Patient Reported Richard Chin, MD
Outcomes) discusses some special types of endpoints, Bruce Y. Lee, MD, MBA
PPRREE--PP337733669955..iinndddd xxiiii 55//2233//22000088 66::5599::4433 AAMM
Section I
Overview
SSEECC11--PP337733669955..iinndddd 11 55//2222//22000088 55::4477::1144 PPMM
Chapter 1
Overview of Clinical Research Medicine
1.1 CLINICAL RESEARCH MEDICINE CTM is primarily a methodological science, in that it
is primarily concerned with h ow to best answer such ques-
1.1.1 Defi nition of Clinical Research tions, not w hat the specifi c answer is. In other words,
Medicine CTM is concerned not with the answer to questions such
as, “ Do ACE inhibitors slow the progression of renal dis-
Let us begin by defi ning the science that is the focus of this ease? ” Regardless of the answer, if the results are defi ni-
book. We call this science clinical trial medicine (CTM) tive, then CTM has served its purpose. Nor is it concerned
and defi ne it as the science of designing, conducting, and with which clinical questions to study or how to apply the
interpreting clinical trials. Its goal is to understand and results to specifi c patients. Rather, it is concerned with
improve methods for determining whether an intervention, determining what is the best way to design trials to answer
such as a drug, a device, or a procedure, improves clini- such questions.
cal outcome in patients. For example, it might address a To put it another way, the goal of CTM is not to be able
question such as, “ How can one determine whether or not to declare, “ ACE inhibitors slow the progression of renal
angiotensin converting enzyme (ACE) inhibitors slow the disease.” Its goal is to be able to say, “ A double-blind,
progression of renal disease? ” Or it might answer a ques- placebo-controlled study using measured creatinine as an
tion such as, “ How can one determine whether or not endpoint at 6 months will answer the question, but a single
patients with angina benefi t from coronary artery bypass arm study using calculated creatinine at 4 weeks will not. ”
surgery? ” CTM is a broad fi eld that addresses issues such As an analogy, CTM is to clinical medicine what an archi-
as types of patients to enroll in a trial, appropriate size of tect is to the house builder, or what a coach is to an athlete.
a trial, and ways of maximizing the amount and quality of
information elicited from a trial.
Put another way, a clinical trial is concerned with fi nd-
ing therapies not for an individual person but rather for a
Principles of medicine
group of patients with a disease. This is different from clini- Clinical research
(Body of medical
cal practice where the goal is to treat individual patients. As knowledge)
an illustration, a question for a practicing clinician treating
a patient may be, “ Will administration of ibuprofen to Mr. X
who has pain in his knee improve his symptoms? ” In a
Practice
clinical trial, the question may be, “ Will administering ibu- of
profen to patients with arthritis decrease their symptoms?” medicine
It is not suffi cient that ibuprofen improves knee pain in Mr.
X; the goal of a clinical trial is whether as a group, most (or
suffi cient proportion) of patients with knee pain of a certain
type benefi t. Clinical trials can eventually lead to improved Various specialties: cardiology,
therapies for a large group of patients if the treatment is ophthalmology, etc.

demonstrated to be effective. FIGURE 1.1 Technical research medicine in context.
3
CChh000011--PP337733669955..iinndddd 33 55//2244//22000088 22::0077::4455 PPMM
4 SECTION I | Overview
In this way, CTM is more similar to other methodological Anecdotal
fi elds such as statistics, education, or epistemology than to
most other branches of medicine.
Aggregate Medical Case
CTM
data knowledge series
CTM as Methodological Science
The body of data generated by CTM is usually added to spe-
cifi c disciplines such as cardiology, oncology, and gastroen-
terology. Only the methodological advances are added to Epidemiological
the body of CTM knowledge. For example, the answer to the data
question, “ Does administering ACE inhibitors to patients
FIGURE 1.2 Sources of medical knowledge.
with diabetes slow the progression of renal disease? ” would
enter the body of knowledge for nephrology. The answer to
the question, “ What is the best way to answer questions of
Evidence-Based Medicine
this type where disease progression is slow, and surrogate
markers are only partially validated? ” would enter the body of
knowledge for CTM. It is often thought that “ evidence-based medicine ” is a mod-
ern development. This term is often used with the implication
physicians in the past practiced medicine without relying on
evidence. This is inaccurate. Our forbearers in medicine prac-
ticed a form of evidence-based medicine, but the sources of
1.1.2 Epistemology of Medicine
evidence were different. They did not have the luxury of bas-
ing their decisions on aggregate data from large trials – they
As was previously mentioned, CTM is only one of several
had little access to such data. They relied on anecdotal data
possible ways of generating medical knowledge. Indeed,
and small case series.
knowledge acquired through clinical trials, especially pro- The new paradigm of modern evidence-based medicine is
spective, randomized, controlled clinical trials, is the excep- different mostly in that it asserts a hierarchy of evidence, plac-
tion rather than the rule in medicine. Historically – and even ing randomized controlled clinical trials at the top and others
today – much of the body of medical knowledge was based below that. This hierarchy is appropriate in most instances,
on other types of evidence, such as personal experience, since in most cases, data generated from randomized, pro-
historical knowledge, case reports, and observational stud- spective clinical trials is more robust than anecdotal data.
ies (Figure 1.2). Commonly, intuition and pathophysiologi- It is however not always appropriate, as will be discussed.
cal rationale have also played important roles in shaping
medical thinking. Habits and practice patterns based on
informal knowledge have been handed down from one gen- Clinical trials are expensive, diffi cult to conduct, and
eration of physicians to the next, usually without formal suffer from some signifi cant validity fl aws. Conducting clin-
verifi cation or validation. ical trials for every therapy is neither practical nor prudent,
In many instances, traditional methods worked ade- and knowledge generated from other methods is not infre-
quately, and even now, clinical trials are not always nec- quently both necessary and helpful. However, for many dis-
essary. For much of medicine, particularly those branches eases, a clinical trial is the most reliable tool for establishing
not concerned with intervention, traditional sources of a causal relationship between intervention and outcome. It
knowledge are acceptable. These include branches such as may sometime be the only way of establishing effectiveness
diagnosis, prognosis, education, and monitoring. Even for of a new therapy and developing a new treatment. This is
interventions, CTM is not always the most rigorous nor because of randomization, prospective treatment assignment,
the most practical way of generating data. In cases where and large aggregate data sets that characterize well-designed
it is impossible to blind treatments, in cases where it clinical trials (Figure 1.3).
would be unethical to randomize patients, and in cases Unlike anecdotal data or small case series, where patient
of extremely rare diseases, formal, rigorous clinical trials histories can be individually studied and understood, clini-
may not be the best option. As an example, for advanced cal trials have too many patients to allow analysis on an
colon cancer, where the survival is less than 5% at 5 years, individual level. Rather, they require that aggregate data be
effi cacy of a drug that achieves 100% survival at 10 years analyzed. Large sample size is a major strength of CTM but
can be established even without a controlled clinical analysis of aggregate data is neither easy nor intuitive, and
trial. A small case series may be suffi cient to establish fraught with cognitive illusions and intellectual fallacies. In
effi cacy. order to avoid inaccurate or spurious conclusions, clinical
CChh000011--PP337733669955..iinndddd 44 55//2244//22000088 22::0077::4455 PPMM
Chapter 1 | Overview of Clinical Research Medicine 5
CTM Anecdotal data Case series Epidemiology
Prospective? Yes Yes Sometimes Sometimes
Active intervention? Yes Yes No No
Assigned, unbiased intervention? Yes No N/A N/A
Aggregate data? Yes No Sometimes Yes
Blinded? Yes No No No
Representative of clinical practice? No Yes Yes Yes
Unbiased, random sample? No No Sometimes Sometimes
FIGURE 1.3 Characteristics of different sources of medical knowledge.
trials must be designed, executed, and analyzed in a rigorous controlled randomized clinical trials were performed. The
way. CTM provides the tools to accomplish this. trials conclusively demonstrated that digoxin is ineffective
in inducing cardioversion.
1.1.3 Case Studies, Personal
1.1.4 Epidemiology and Observational
Observations, and Case Series
Data
As noted above, most medical knowledge has been gener-
One formal – as opposed to anecdotal – source of knowl-
ated through accumulation of anecdotal data by individual
edge is epidemiology. This discipline relies on surveys, close
physicians over time, and passed down through generations.
recording of aggregate data, retrospective studies, registries,
Informal experience, such as personal observations, historical
and prospective nonrandomized studies. Unlike case histo-
knowledge, case studies, and case series are effective ways
ries, epidemiology relies on a quantitative data set collected
of accumulating, clarifying, and disseminating such knowl-
in a consistent enough manner to allow mathematical and
edge. In many instances, this type of knowledge is invaluable,
statistical analysis.
and has been critical in advancement of medical care. Some
examples include case series of Wegener ’ s Granulomatosis
and retrospective analysis of Reye ’ s syndrome. In addition,
Example: Matches and Lung Cancer
the intellectual tools required to observe, describe, and assim-
ilate such knowledge usually come readily to an average
Correlation does not establish causation. A classic demonstra-
physician because they draw upon the arts of physical exami-
tion of this is the relationship between matches and lung can-
nation and diagnosis. These are skills used in daily practice.
cer. There is a strong correlation between carrying of matches
The drawback to this approach is that the number
and risk of lung cancer, but matches do not cause cancer.
of cases any individual physician or even an institution
Rather, many people who carry matches do so because they
encounters is usually small. It is also diffi cult to enlarge the smoke, and smoking causes lung cancer. A randomized trial
data set by pooling the experience of multiple physicians. that assigned one group to carry matches and another not to
Data collected by different persons, under different condi- carry matches would fi nd that there was no difference in lung
tions, documented in different ways, can be diffi cult to col- cancer rates between the two groups.
late and interpret. In addition, knowledge obtained this way
is commonly fraught with confounding factors. In short,
the probability of inaccurate conclusions based on these Epidemiology is a sister discipline of CTM. The two dis-
types of sources tends to be higher than conclusions drawn ciplines are similar to each other, particularly with respect
from randomized, controlled clinical trials. to the inferential method of drawing conclusions. However,
Many treatments that were supported by conventional epidemiology, unlike clinical trials, does not involve an
wisdom, and many practices based on knowledge distilled active intervention and therefore does not normally lead
from thousands of physician-years of expert experience have to causal inferences. It can establish correlations between
subsequently been demonstrated to be erroneous. A clas- patient characteristics or therapies and outcome, but corre-
sic example is the digoxin and cardioversion. Digoxin had lations do not establish causation in and of themselves.
previously enjoyed general acceptance as being effi cacious Formal observational knowledge can come from cross-
for cardioversion of atrial fi brillation until well-designed sectional surveys, case-controlled studies, and cohort studies.
CChh000011--PP337733669955..iinndddd 55 55//2244//22000088 22::0077::4455 PPMM
6 SECTION I | Overview
A cross-sectional study is a survey that collects risk factors to work when in fact it has no effect. Just by chance, some
and outcome data in a group at one point of time and exam- patients ’ symptoms will spontaneously regress.
ines the data for correlations. A case-controlled study com- Clinical trials, with (usually) large aggregate data sets,
pares a group of patients with a disease to a group without, randomization, and blinding, can often overcome these
and explores risk factors in the past. A cohort study exam- issues of variability and noise.
ines a group with a risk factor and one without and follows The second argument against relying on informal observa-
them prospectively for the disease. tions or nonrandomized studies is the diffi culty in distinguish-
ing between a result due to a bias vs. a result due to a real
effect. For example, when a patient and/or the treating phy-
sician know that a therapy is being administered, there may
Note: Epidemiology and Safety Data
be a placebo effect. For example, a physician conducting a
psoriasis trial might under-report the area of body affected by
As an aside, one aspect of CTM that overlaps with epidemiol-
psoriasis when he or she is measuring the response.
ogy is the monitoring and analysis of safety data. This branch
Another potential source of bias is imbalance between
of CTM relies heavily on epidemiology, and will be discussed
the treatment groups. The group receiving the interven-
in a later section of this book.
tion, for example, may be younger and healthier than the
one not receiving it. The outcome in that group may be bet-
ter than the control group, not because of the intervention
but because they were healthier to begin with. In addition,
1.2 RATIONALE FOR CLINICAL TRIALS there can be regression to the mean in waxing and waning
diseases, in that any therapy administered during fl ares will
1.2.1 Scientifi c Rationale seem to improve symptoms in some patients, just due to the
natural history of the disease.
As noted above, there are multiple paths to medical knowl- Clinical trials can ameliorate or eliminate these biases
edge. Reliable medical knowledge, particularly about effi - and issues. Blinding can reduce the placebo effect and ran-
caciousness of interventions, is diffi cult though to glean domization can reduce imbalances in patient characteristics
for most diseases and most interventions. There are several between the groups.
reasons for this, and a modern randomized, controlled, pro- The third argument against informal observations as the
spective clinical trial can address most of these factors. sole source of medical knowledge is the hazards of mul-
The fi rst reason is the diffi culty of determining whether tiple post hoc analyses. Given any set of data, and given
an outcome represents a true signal or just background suffi cient numbers of analysis of subgroups and endpoints,
noise. For example, administering a new compound to one it is possible to link almost any therapeutic intervention to
patient (or even to 10) and observing that the patient recov- an outcome. In other words, if the data is analyzed enough
ers from pneumonia doesn’ t establish that the drug cured times in enough different ways, one can often fi nd a con-
the infection. This is because most diseases have variable vincing association between therapy and outcome. For
outcome; anecdotal evidence are subject to tremendous example, it is often possible to fi nd correlations between
biases and confounding factors; and most drugs only work even patently trivial characteristics such as zodiac signs
partially. The patient could have spontaneously recovered and response rate. On average, looking at the data in 20 dif-
from pneumonia, as many patients do. ferent ways can be expected to yield one spurious associa-
There is great variability in the onset, course, and out- tion with a p value of 0.05 or less.
come of many common diseases. For example, only a few Prospective clinical trials prespecify one primary end-
of the patients exposed to M. leprae contract leprosy; only point. This minimizes the risk of spurious results. By
a fraction of the patients who harbor H . pylori develop convention, a randomized, prospective clinical trial that
ulcers; and only some of the patients who experience a demonstrates a difference between a treated group and an
myocardial infarction develop lethal arrhythmias. There is untreated group for the pre-specifi ed primary endpoint with
also great variability in response to many therapies. Statins a p value of less than 0.05 is accepted as having established
prevent cardiovascular events in only a fraction of patients a causal relationship. This convention does not eliminate
who receive them; infl iximab induces a response in only the possibility of spurious results, but does make it much
a fraction of rheumatoid arthritis patients; and aspirin less likely, and establishes a common language and com-
relieves headache in only some patients, some of the time. mon ground for decisions on whether an intervention was
As another example, psoriasis is a disease with a wax- effective.
ing and waning course. Most drugs for the disease work in The fourth and the most important rationale for limiting
some patients but only sporadically, and in other patients reliance on informal knowledge, especially knowledge based
never. Small uncontrolled series or trials in psoriasis often on retrospective data, is the need to establish causation.
will yield misleading results because some drugs will seem Although in some special cases, it is possible to establish
CChh000011--PP337733669955..iinndddd 66 55//2244//22000088 22::0077::4466 PPMM