-
Creating a Cohort of Cases – ICTR Workshop on Clinical
Registries
Josef Coresh, MD, PhD Professor of Epidemiology, Biostatistics
& Medicine
Johns Hopkins University Director, George W. Comstock Center for
Public
Health Research and Prevention Director, Cardiovascular
Epidemiology Training
Program
-
Outline • Cohort definition (see Gordis “Epidemiology” text for
overview)
– Membership criteria (“Case” Definition in a clinical cohort of
cases – but remember that case series is a weak design) -
Considering Referral Pathway - Considering Precohort Factors
• Data collection – Exposures, Treatments & outcomes (mostly
covered by other lectures)
• Examples of different cohorts to illustrate ideas: – ARIC
– CHOICE
– CLUE
• Discussion of planned cohorts by participants
-
Taxonomy of Designs
• Randomized Controlled Trial • Prospective Cohort Study
– Variations exist – non-concurrent (going back to old records
etc.)
• Case-Control Study • Cross-Sectional Study • Other Designs
– Quasi-Experimental
– Ecologic
– Case Report
-
The basic fighting unit was a cohort, composed of six centuries
(480 men plus 6 centurions). The legion itself was composed of ten
cohorts, and the first cohort had many extra men—the clerks,
engineers, and other specialists who did not usually fight—and the
senior centurion of the legion, the primipilus, or “number one
javelin.”
http://www.vroma.org/images/mcmanus_images/cohort.jpg�
-
pro·spec·tive Pronunciation: pr&-'spek-tiv also 'prä-",
prO-',prä-' Function: adjective Date: circa 1699 1 : relating to or
effective in the future 2 a : likely to come about : EXPECTED b :
likely to be or become
http://www.m-w.com/cgi-bin/dictionary?book=Dictionary&va=expected�
-
“Prospective” in Epidemiology
• Clearly defined cohort (group, sample) of persons at risk
followed through time
– For pre-defined outcomes
– And their relationship to “exposures” measured prior to the
outcome (reduces bias, e.g. recall; but confounding & effect of
subclinical disease remain)
• Data regarding exposures (risk factors, predictors) collected
prior to data on outcomes (endpoints)
• Research-grade data collection methods used for purpose of
testing hypothesis (?)
-
0
5
10
15
20
25
30
35
120 160 200 240Cholesterol, mg/dL
3-ye
ar C
VD M
orta
lity R
ate P
er 10
0
*Adjusted to the age of 60 years, female, Whites, HD and
non-smokers.
Overall
Distorted Associations – Reverse Causation? (Baseline
Subclinical Disease lower Cholesterol higher CVD)
Adjusted* 3-year cardiovascular mortality in Dialysis
Patients
Presence of Inflammation/Malnutrition
Absence of Inflammation/Malnutrition
Liu et al. JAMA 2004; 291(4):451-9.
-
Cohort - Membership
• Cohorts are defined at baseline and followed subsequently
(exception: open cohorts can continue to enrol during
follow-up)
• Reasons for selection: – Group of interest for follow-up (e.g.
specific disease -
brain cancer, MI, ESRD, “middle age”)
• Basis for Inferences: – Internal comparisons (within the
cohort) are strongest
(randomized; “exposure” measured prior to outcome)
– External comparisons are quite weak (e.g. case series)
• Selection: biases all external comparisons but only some
internal comparisons.
-
Why Do A Cohort Study?
• Get incidence data • Study a range of possible risk factors •
Establish temporal sequence (risk factor before outcome) • Get
representative data (of some population) • Prepare for randomized
controlled trial
– Effect size estimates
– Population of eligible participants (“registry”)
• Establish a research empire (not a good primary goal)
-
Types of Cohorts • Occupational (e.g. Asbestos workers) •
Convenience (e.g. Precursors, Nurses) • Geographic (e.g.
Framingham, ARIC) • Disease or Procedure
– Natural History (e.g. Syncope, Lupus)
– Outcomes Research (e.g. Dialysis, Cataracts)
-
Sources of Cohort Data
• Clinic Visits – Laboratory Assays
– Interview
– Physical Examination
– Imaging
– Physiologic tests
• Home visits • Mailed materials • Telephone Interview
• Medical Records • Administrative Data
– Medicare
– Medicaid
– Managed Care
– Veterans Admin
• Birth Records • Death Certificates • Specimen Bank
-
Challenges in Cohort Studies
• Possibly long duration • Possibly large sample size • Need to
recruit people “at risk” • Drop outs, Deaths, Other losses •
Concern about residual confounding • Multiple comparisons Type I
error
-
How to Exploit Cohort Design When Time is Short & Money is
Scarce
• Analyze existing data from another study • Piggy-back onto
on-going study • Choose hospital-based cohort • Choose short-term
outcome • Consider administrative data • Consider public-use data •
Consider non-concurrent design
-
Examples – Food for Thought
-
Results Drift – Even in a “good” lab Serum Creatinine Compared
to the Mean of All Labs:
College of American Pathologists (CAP) Data
Coresh J et al. Am J Kid Dis 2002;39:920-929
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
1/1/1992 1/1/1994 1/1/1996 1/1/1998 1/1/2000Date
Ser
um C
reat
inin
e D
iffer
ence
, mg/
dl
White Sands - Mean of All MethodsCleveland Clinic - Mean of All
MethodsAverage White Sands - Mean of All MethodsAverage Cleveland
Clinic - Mean of All Methods
-
Systematic Errors can be “corrected”
• NHANES 1988-1994 data can be “calibrated” to the cleveland
clinic foundation (CCF) 2006 standardized serum creatinine assay
using regression
01
23
420
06 C
CF
Cre
atin
ine
from
sto
red
sam
ple
(mg/
dL)
0 1 2 3 4Uncalibrated NHANES III (mg/dL)
Uncalibrated NHANES III vs 2006 CCF with identity line
-1.5
-1-.
50
.51
1.5
Diff
eren
ce (
CC
F 2
006
scr
- O
rigin
al N
H3
scr)
0 .5 1 1.5 2 2.5 3 3.5 4Mean ([CCF 2006 scr + Original NH3
scr]/2)
diff1 Fitted values
black lines are +/- 1.96*SDBland-Altman Plot for Creatinine
Selvin et al. Am J Kidney Dis. 2007; 50(6):918-26.
-
ARIC – Atherosclerosis Risk in Communities
• NHLBI cohort to study atherosclerosis – Community based sample
ages 45-64
– ~5 hour examination: interview, exam, phlebotomy, carotid
ultrasound (all standardized) • Baseline, 3, 6, 9 years … 25
years
– Annual telephone calls
– Chart abstraction of all hospitalizations
– Morbidity and Mortality Classification Committee review of CHD
outcomes
-
ARIC-NCS Calendar Year 1987-89 1990-92 1993-95 1996-99 2004-06
2011-13
Aim 1 Prevalence X
Stage 2 Eval 2637
Aim 4
ARIC-NCS Study Design Overview Exam 1 Exam 2 Exam 3 Exam 4
Brain MRI
Aim 3 8,220+phone
Genetics – Aim 5
R – Retinal photography
Aim 2
X 2,000**
Cognitive testing X X (n) 14,201 11,343
Brain MRI X 1,134 X 1,929
Stage 3 MRI
** Includes 357 dementia,852 MCI, 791 normal; 547 with 2
previous brain MRIs •Numbers updated to reflect 2011 start +
distant + no lower age limit
X X X X
X X X X
X X
X
R R
15,792 14,348 12,887 11,656 8220 examined more incl. phone
(n)
Median follow-up ,y 0 3 6 9 17 25
1,134
Vascular risk factors
Vascular markers
Age range,y 45-64 48-67 51-70 54-73 62-82 68-89
ARIC V5
Combined visit
X Echo-
cardiogram
X
-
ARIC – NCS: Aims 1) estimate the prevalence of dementia/MCI by
race and sex
in participants aged 70-89, 2) determine whether midlife
vascular factors (risk factors
and markers of macrovascular and microvascular disease) predict
dementia, MCI and cognitive change,
3) determine whether the associations between midlife vascular
factors and dementia/MCI differ by dementia/MCI subtype defined
clinically or by MRI signs,
4) identify cerebral markers associated with cognitive change,
including progression of MRI ischemic burden and atrophy across 3
MRI scans spanning 17 years, and
5) identify genomic regions containing susceptibility loci for
cognitive decline, using 106 SNPs spanning the genome.
-
Type of contact Content Sample for Stages 2 & 3
AFU Call
Clinic visit
Stage 1 (n=6886) (4/d * 5 d/wk)
Stage 2 – participant + proxy (2.3/d*3d/wk)
Stage 3 (2/d * 2d/wk)
Contract V5 + NCS Cognitive Function * MRI eligibility *
Schedule stage2 (+MRI for subset)?
Neuro** + retinal
MRI – same day as Stage 2 for dementia + normals (for borderline
cases MRI sampling depends on Stage 2)
(6.5 hours) (~3 hours) (~1 hour)
Home or LTC
Abbreviated exam Abbreviated – done with Stage 1
No MRIs
Overview of ARIC Visit 5 + NCS Data Collection
* Only applies to sampled individuals – sampling fractions based
on CF & ∆CF ** Skip the neuro exam on most (all but n=50)
normals
-
CHOICE Cohort Choices for Healthy Outcomes in Caring for
ESRD
• Study Design: national prospective cohort study (CHOICE;
PI:Powe & Klag & specimen bank Coresh)
• Study Population: – 1026 incident outpatient dialysis patients
– Enrolled between 10/95 and 06/98 (DCI + St. Raph) – Recruited
within a median of 45 days from 1st dialysis
(98% within 4 months) – From 81 dialysis clinics in 19 States –
Age 18 years or older, English or Spanish speaker – Provided
informed consent
• Main research topics: Dose & ModalityOutcomes
21
-
CHOICE Top Papers 119 cited 2,110 by 2010 by (Fink N* AND
(Coresh or Powe or Klag))
1. Association between cholesterol level and mortality in - Role
of inflammation dialysis patients and malnutrition . Author(s): Liu
YM, Coresh J, Eustace JA, et al. JAMA 2004 Times Cited: 209 2.
Traditional cardiovascular disease risk factors in dialysis
patients compared with the general population: The CHOICE study.
Author(s): Longenecker JC, Coresh J, Powe NR, et al. JASN 2002
Times Cited: 180 3. The timing of specialist evaluation in chronic
kidney disease and mortality Author(s): Kinchen KS, Sadler J, Fink
N, et al. Ann Int Med 2002 Times Cited: 176 4. Validation of
comorbid conditions on the end-stage renal disease medical evidence
report: The CHOICE study. Author(s): Longenecker JC, Coresh J, Klag
MJ, et al. JASN 2000 Times Cited: 141 5. Changes in serum calcium,
phosphate, and PTH and the risk of death in incident dialysis
patients: A longitudinal study. Author(s): Melamed ML, Eustace JA,
Plantinga L, et al. Kidney Int 2006 Times Cited: 96
-
CHOICE Top Papers 119 cited 2,110 by 2010 by (Fink N* AND
(Coresh or Powe or Klag))
6. MYH9 is associated with nondiabetic end-stage renal disease
in African Americans Author(s): Kao WHL, Klag MJ, Meoni LA, et al.
Nature Genetics 2008 Times Cited: 93 7. Timing of nephrologist
referral and arteriovenous access use: The CHOICE study Author(s):
Astor BC, Eustace JA, Powe NR, et al. Am J Kidney Dise 2001 Times
Cited: 92 8. Comparing the risk for death with peritoneal dialysis
and hemodialysis in a national cohort of patients with chronic
kidney disease Author(s): Jaar BG, Coresh J, Plantinga LC, et al.
Ann Int Med 2005 Times Cited: 86 9. Type of vascular access and
survival among incident hemodialysis patients: The choices for
healthy outcomes in caring for ESRD (CHOICE) study Author(s): Astor
BC, Eustace JA, Powe NR, et al. J Am Soc Nephrol 2005 Times Cited:
73 10. Comorbidity and other factors associated with modality
selection in incident dialysis patients: The CHOICE Study
Author(s): Miskulin DC, Meyer KB, Athienites NV, et al. J Am Soc
Nephrol 2002 Times Cited: 72
-
Research Opportunities in Washington County: From shoe-leather
epidemiology to genomics
Josef Coresh, MD, PhD Professor of Epidemiology, Biostatistics
& Medicine Johns Hopkins University Director, George W.
Comstock Center for Public Health Research and Prevention Ana
Navas-Acien, MD, PhD Assistant Professor, Environmental Health
Sciences & Epidemiology Sleep
Heart Health
Washington County, MD Johns Hopkins University
-
CLUE I & CLUE II Studies
CLUE I (1974) N=26,147 • Serum stored at -70o • Baseline
questionnaire
CLUE II (1989) N=32,894 • Plasma , RBC, DNA -70o • Toenail
sample • Baseline questionnaire • Food freq. questionnaire
-
The CLUE Specimen Banks: A paradigm for long-term,
population-based studies to evaluate cancer-related biomarkers
CLUE I (1974) N=26,147
Serum
Plasma WBC RBC
Follow-up for cancer outcomes through Washington County Cancer
Registry (medical record/treatment info available)
Active follow-up of CLUE II cohort: questionnaires
Key advantages: • large, prospective • population-based • long
term follow-up • specimens from multiple time points • specimens
obtained prior to diagnosis • multiple health outcomes
(8297 also gave to CLUE I)
Odyssey
CLUE II (1989) N=32,894
Baseline questionnaire – FFQ included in CLUE II
1996, 1998, 2000, 2003, 2007
-
Number of Deaths from CLUE I and CLUE II Volunteers as of
6/30/2009
Cause of Death ICD10* Clue I Clue II Clue I
& II Total
Heart Disease I20 – I51
1261
713
777
2751 Cancer C00 -C97 929 668 672 2269 Cerebrovascular I60 – I69
254 144 170 568
Chronic Lower Respiratory Disease
J40 –J47
222
125
121
468
Influenza, Pneumonia J10 –J18 149 61 72 282 Accident V01-
X59,
Y85, Y86 83 59 52 194
Nephritis, Nephritic syndrome, Nephrosis
N00 -N07, N17 -N19 N25 -N27
53
30
33
116
Total 5823 2379 2476 10678 All deaths 8299 4855
* ICD-8 and 9 used for previous years Underlying caues of death
data not available for 1999 CLUE I and 23 CLUE II participants (11
in CLUE I & II)
-
Thank you! (it takes a team)
CKD-Epi
ARIC Staff CHOICE Study
CVD-Epi Stein Hallan
Creating a Cohort of Cases – ICTR Workshop on �Clinical
RegistriesOutlineTaxonomy of DesignsSlide Number 4Slide Number
5“Prospective” in EpidemiologyDistorted Associations – Reverse
Causation?�(Baseline Subclinical Disease lower Cholesterol higher
CVD)�Adjusted* 3-year cardiovascular mortality in Dialysis
Patients�Cohort - MembershipWhy Do A Cohort Study?Types of
CohortsSources of Cohort DataChallenges in Cohort StudiesHow to
Exploit Cohort Design When Time is Short & Money is
ScarceExamples – Food for ThoughtResults Drift – Even in a “good”
lab�Serum Creatinine Compared to the Mean of All Labs: �College of
American Pathologists (CAP) DataSystematic Errors can be
“corrected”ARIC – Atherosclerosis Risk in CommunitiesSlide Number
18ARIC – NCS: AimsOverview of ARIC Visit 5 + NCS Data Collection
CHOICE Cohort�Choices for Healthy Outcomes in Caring for ESRDCHOICE
Top Papers�119 cited 2,110 by 2010 by (Fink N* AND (Coresh or Powe
or Klag))CHOICE Top Papers�119 cited 2,110 by 2010 by (Fink N* AND
(Coresh or Powe or Klag))Research Opportunities in Washington
County: From shoe-leather epidemiology to
genomics CLUE I & CLUE II
StudiesSlide Number 26Number of Deaths �from CLUE I and CLUE II
Volunteers�as of 6/30/2009Thank you! (it takes a team)