Clinical Research and Epidemiology
Robert S. Sandler, M.D., M.P.H.
University of North Carolina
Chapel Hill, North Carolina
What is epidemiology?
Epidemiology
● Study of the distribution and determinants of disease
in populations
Clinical epidemiology
● Science of making predictions about individual
patients using the tools of epidemiology
History
John Snow
Map of cholera cases 1854
History
Epidemiology today
N Engl J Med 2012;367:1704-13
Epidemiology today
Williamsburg
Crown Heights
Borough Park
N Engl J Med 2012;367:1704-13
IBD is a challenge for epidemiologists
Epidemiologic methods work great when
● Acute onset illness – cholera, mumps
● Point source – Broad Street pump, yeshiva
● Rare illness cluster – angiosarcoma of the liver and
vinyl chloride
IBD
● Gradual onset in children and adults
● Persists for years
● No known etiology
Spectrum of epidemiologic research
Disease burden (how many/how much)
Etiology risk factors for disease)
Diagnosis (evaluation of medical tests)
Natural history/prognosis
● What happens to people with disease
● Risk factors for disease outcomes
Treatment effects
● Intended
● Unintended
Methods of clinical research
Observational Studies
● Investigators just observe (measure) exposures and
outcomes
Interventional Studies
● Investigators assign exposure and measure
outcomes (clinical trials of drugs or surgery)
Observational Studies
“You can observe a lot just by watching.”
- Yogi Berra
Descriptive
Analytical
● Cross-sectional
● Cohort
● Case-control
Descriptive Studies
Burden of Disease
● Incidence – how many new cases/time
● Prevalence – how many cases currently
● Morbidity
● Mortality
● Costs
● Days of work/school lost (disability days)
Descriptive studies
Incidence
Prevalence
Prevalence ~
incidence x duration
Analytical studies
Associations between exposures and outcomes
● Is smoking (or diet) associated with risk of IBD?
● Are specific genes associated with risk of IBD?
● Are results of a diagnostic test associated with diagnosis
of IBD?
● Is treatment X associated with improvement in Crohn’s
disease activity
● Is treatment Y associated with a particular side effect?
● Is diet (or other lifestyle factors) associated with IBD
flares?
Cross-sectional studies
Exposure and outcome measured at same time
Outcome?
Source Population
Sample
Exposure?
Cross-sectional
Assemble study population
Measure IBD status (yes/no)
Measure smoking (yes/no)
Find that people with Crohn’s more likely to smoke
● i.e. smoking (exposure) is associated with CD
(outcome)
Advantages: Quick and inexpensive
Disadvantages:
● No temporality
● Cannot differentiate cause and effect
Cross-sectional
CCFA Partners: Internet based research study
>12,000 individuals with IBD completed an online
survey about steroid use and many patient outcomes
Cohort study
Baseline
Exposed
Disease
Unexposed
Disease
Measure exposure
Time
Study begins here
Cohort study
Assemble study population free of disease
Measure smoking (yes/no)
Follow study population for long time
Identify those who develop IBD
Compare incidence (# new cases) of IBD in smokers vs
nonsmokers
Advantages: Can identify temporal relationship, assess
causality, determine incidence
Disadvantages:
● Long follow-up needed ($$$$$$$)
● If outcome is rare (IBD), need very large population
NSAIDs and Aspirin
Nurses’ Health Study I is a prospective cohort of 121,700
U.S. female registered nurses, aged 30 to 55 years
Aspirin/NSAID use ascertained by self-report
Over 1,295,317 person-years of follow-up
123 cases of CD and 117 cases of UC
CD
UC
NSAID > 15
days/month
1.6 (1.0-2.6)
1.9 (1.2-3.0)
Aspirin > 15
days/month
1.0 (0.6-1.5)
1.0 (0.6-1.7)
Ananthakrishnan, Annals Internal Medicine, 2012
Case-control study
Exposed?
Disease
Exposed?
No
Disease
Time
Study begins here
Case control study
Match cases (people with outcome (IBD)) to controls
Measure prior exposures
Compare exposures in cases and controls
● Are IBD cases more likely to report low fiber intake in
past than controls
Advantages:
● Good for rare diseases
● More efficient (time and sample size, ↓$$$$)
● Can identify temporal relationships
Disadvantages:
● Recall of exposure
● Sample selection (cases and controls should arise
from same population)
Case control study of potential risk factors for IBD
Cases with CD (n = 364) and UC (n = 217), ages 18-50 yr
Controls were drawn from Manitoba Health
Subjects were administered a multi-item questionnaire
Bernstein, Am. J. Gastro, 2006
What do statistics tell us?
Magnitude of effect
Precision of results
Statistical significance - role of chance
Statistical significance not clinical significance
Potential pitfalls: validity and bias
Internal validity: degree to which the study findings
reflect the truth
● Bias: Systematic difference between observed
findings and the truth
● Chance
● Statistics assess role of chance, but do not
address bias
External validity: degree to which the study findings are
applicable to other populations
Types of bias
Selection/referral bias
● Distortion in the results due to a difference b/t
subjects who participated and those who did not
Measurement bias
● Distortion in results due to error in measurement
of the exposure, outcome or both
Confounding
Direct
Exposure
Effect
(smoking)
Indirect
Effect
Confounder
(alcohol)
Outcome
(ulcer)
Clinical trial
Intervention ‘A’
Outcome
Standardized
Follow-up
&
Co-interventions
Study
population
Intervention ‘B’
Outcome
What does randomization accomplish?
If treatments are assigned at random, then:
● Treatments are unrelated to patient characteristics
● Achieves balance among different treatment groups
(same mix of smokers, disease severity, across
groups)
● No confounding
● Even if confounder not known or measured!
Clinical trial
● 526 patients randomly assigned to receive
intravenous ustekinumab or placebo
● Primary outcome (end point) was a clinical response
at 6 weeks
Clinical trial
Clinical trial
Why Placebo?
If patients know they are getting a study medication, they
may feel better
If doctors know their patients are getting a medication, they
may perceive them as doing better
“Double-blind, placebo-controlled”
● All subjects get something
● Neither subjects nor investigators know what
Randomization and concealed allocation required to prove
efficacy
Required by FDA
Ethical considerations
Honest disagreements exist about the preferred treatment
of the condition (equipoise)
● Neither intervention/treatment should be considered
inferior, or potentially harmful
The RCT must produce results which are convincing
enough to resolve the dispute
● Makes the patients’ participation is worthwhile
Comparative Effectiveness Research
If all we have is placebo controlled trials, then we don’t
know
● What medicines are more effective than others
● What medicines are safer than others
We need head-to-head comparisons, but
● Pharma often not willing to take the risk
● FDA not require it
Observational studies?
Clinical Effectiveness Research
Patients who enroll in clinical trials are:
● Young to middle aged adults
● Free of co-morbidity (other health problems)
● Cared for at large research hospitals
What about. . .
● Kids
● Elderly
● Those with other health conditions
Observational studies?
Safety
Clinical trials too small to assess medication safety
● 51% of drugs have label changes due to major
safety issues discovered after marketing
● 20% of drugs get new “black box” warnings after
marketing
● 4% of drugs are ultimately withdrawn for safety
reasons
Too Small
Probability of detection
Number of exposed
5000
99%
4000
95%
3000
90%
Lymphoma
2000
REACH
CERTIFI
1000
1/10
1/100
Incidence
n=212
n=526
1/1000
37
Safety research
Objective: To determine whether initiation of TNFantagonists compared with immunomodulators is
associated with an increased risk of serious infections
requiring hospitalization
Methods:
Cohort study using de-identified data from insurance
companies
For IBD, TNF- antagonist initiators compared to 6MP/AZA
initiators, matched by propensity score
Grijalva, JAMA, 2011
Hospitalization for Infection
No difference in hospitalizations
for infection
Mean age 58 (versus 35 in
SONIC)
What makes a good study
Is there a primary research question?
Is the research question important?
Is the study design appropriate to the research question?
Are the investigators qualified?
Is the study population well-defined?
Are exposures and outcomes well-defined?
Feasible?
● Sample size needed relative to time and budget?
● Barriers for recruitment
 Burden (for patients and providers)
 Pool of eligible subjects
 Competing studies
Ethical?
Review
Conclusion
Broad spectrum of clinical and epidemiological research:
burden, etiology, diagnosis, prognosis, treatment effects
Many study designs
● Advantages and disadvantages to each
Focused and highly relevant clinical question essential
Many pitfalls to be considered
● Bias
● Ethics
● Practical considerations