1. Acknowledgment: EPIET/EUPHEM
Choice of reference group for comparison Tunis, 29th October 2014
Acknowledgment: EPIET/EUPHEM
Dr Adela Paez
MediPIET Scientific Coordinator – ECDC, Sweden

2. Learning objectives Source population
How to choose your comparison group:
What is a good control?

3. Epidemiologists .... compare
occurrence of outcome depending on presence or absence of exposure
disease incidence in different time-periods in a population
First, we need to understand what is the source population!

4. Cases Controls Source population Exposed Sample Unexposed
The source population is the
source of the cases
The control group is the sample of
the source population representative
with regard to exposure
Controls

5. The control group “The comparison group serves to provide
an estimate of the exposure distribution in the source population
from which the cases originate.”
Rothman KJ, 1986

6. Control characteristics
If controls represent source population, they need to:
be representative of exposures in source population
be identified as cases if they had disease under study
have same exclusion and restriction criteria as cases

7. Example of an outbreak 32 cases of Salmonella Enteritidis
Granada, Spain (population 230,000)
onset through September 2014
age range 15 – 48 years
19 male, 13 female
Not belonging to a defined group such as attendants of a common event, etc.
No recent travel abroad
What type of study would you conduct?
What is the source population?

8. Source population Residents of Granada aged above 14 years
during September 2014
who have not recently travelled abroad

9. Source population Residents of Granada aged above 14 years
during September 2014
who have not recently travelled abroad
Controls should then be
representative of this population

10. Case definition
Resident of Granada aged above 14 years with isolate of Salmonella Enteritidis in faecal sample during September 2014
Exclusion: Travel abroad in week before illness

11. How to select controls? Aim for random sample of source population
Not always feasible

12. Selecting controls (examples)
Population / community
General practitioners
Neighbourhood
Friends or acquaintances
Hospital

13. Population / community controls
Is there a list or register of source population?
Such a list should
be complete
contain all cases
be readily accessible
identify specified characteristics, e.g. age
Take random sample

14. or... random digit dialling
using residential directories or mobile numbers
quick and easy
but may be biased in selection
telephone ownership
availability
geographical area
participation

15. Population / community controls
Advantages
Best choice if the source population = community
Good representation of the source population
Cost and time efficient
Disadvantages
Possible only in countries where population register exist
Depends on the completeness of register

16. General practitioners
Advantages
well defined catchment population
easily accessible
cost and time efficient
Disadvantages
underrepresentation of mobile population
underrepresentation of private care users
possible only in countries where GP registration required by law

17. Neighbourhood controls
Advantages
no need for population register
similar socio-economic status and
environmental exposure
Disadvantages
might be too similar to cases
low co-operation
may be time consuming, expensive

18. Friends / family controls
Advantages
Good matching for social and
genetic factors
Can be quick and easy
People motivated to participate
Disadvantages
Effectively link the cases and controls
May not be able to detect a difference
in exposure of importance

19. Hospital controls Advantages Disadvantages
useful if all cases identified from hospital register
easy to identify
cost and time efficient
Disadvantages
different catchments for different diseases
may not be representative of source population
overmatching on exposures for other diseases

20. Controls may not be easy to find
Representative of exposure
Exposure should not be a criteria for selecting

21. Example of an outbreak 32 cases of Salmonella Enteritidis
Granada, Spain (population 230,000)
onset through September 2014
age range 15 – 48 years
19 male, 13 female
no recent travel abroad

22. Source population Residents of North Yorkshire aged above 14 years
during April 2012
who have not recently travelled abroad
Controls should then be
representative of this population

23. Which reference group ? How would you select controls ? (2 minutes!)
You are in charge of the case control study!
How would you select controls ?
No population register or list is available
Please discuss with a person next to you
(2 minutes!)

24. Some common questions Immune populations 100% ill
Non-cases as controls

25. 1. Immune subjects Difficult to identify their immune status
May have been cases in the past
Not part of the source population
May introduce bias and under-estimation of effect
OR towards 1

26. 2. Which control group if 100% ill ?
What if close to 100% of population ill?
Try to have severity
Cases: severe cases
Control group: less severe cases

27. Outbreak of S. Enteritidis in a Kebab restaurant, North East London February 2005
Food specific attack rates of severe case (AR), relatives risks (RR), 95% confidence intervals (95% CI) amongst cases.
Food eaten
Food not eaten
Severe
Total
AR% RR
95% CI
Salad 10 12
83.3 32 71
45.1
1.9
Chips 11 14
78.5 31 69
44.9
1.8
Sauce 59
52.5 24
45.8
1.2 57
54.4 26
42.3
1.3
Source: Giraudon I et al, EPIET, London HPA

28. Low attack rate: non-cases likely to represent exposure in source pop
3. Non-cases as controls
Cases
Low attack rate: non-cases likely to represent exposure in source pop
Non- cases
Source population
start
end
High attack rate: non-cases unlikely to represent exposure in source population
Cases
Source population
Non- cases
start
end

29. Outbreak of food borne disease in a nursing home
Outbreak of food borne disease in a nursing home. 100 residents, 40 cases
Cohort
Source
Population
Non
cases
Cases
ARe = 60%
ARu = 10%
RR = 6 E 36 24 60 E 4 36 40

30. Outbreak of food borne disease in a nursing home
Outbreak of food borne disease in a nursing home. 100 residents, 40 cases
Cohort
Potential control groups
Sample
Non
cases
Source
Population
Non
cases
Cases E 36 24 60 12 E 4 36 40 18
RR = 6
OR = 13.5

31. Outbreak of food borne disease in a
nursing home. 100 residents, 40 cases
Cohort
Potential control groups
Sample
Source
Population
Source
Population
Sample
Non cases
Non
cases
Cases E 36 24 60 12 30 4 36 40 18 20 E
RR = 6
OR = 13.5
OR = 6

32. To sum up…

33. Key points in choosing controls
Define source population
Think of the cohort you could have done
Aim for representative sample
Review pros and cons of available options

34. Characteristics of good controls
Come from the same population as the cases
Have the same opportunity of exposure as cases
Could be recruited as cases if diseased
Have exposure window identical to cases

35. Dealing with imperfect control groups
Examine the limitations of your control group with respect to each criteria
Assess in which way the limitation will affect the odds ratio
Interpret your results in light of this review

36. Conclusions No control group is perfect
Consider the advantages and disadvantages of the available options
Aim to minimise error within constraints of
resources and
urgency of study

37. Be prepared to defend your choice…

38. Research question (1/2)
12% Hepatitis C Virus (HCV) prevalence in Egypt
How would you identify current risk factors for HCV infection? What type of study would you conduct?
Hint: we need to identify new HCV infections
Difficulty: the majority (80%) of acute HCV infections are asymptomatic
The highest HCV prevalence in the world occurs in Egypt at an estimated 12% [1], i.e. 10 to 20 fold higher than in Northern Europe [2] or in the United States [3]. The bulk of chronic infection is age-related [4] and occurs among persons of rural origin. Cohort studies have estimated a 9% prevalence and 0.8/ 1000 person-years incidence in Upper Egypt, and a 24% prevalence and 6.8/1000 incidence in the Nile Delta [5,6].
The widespread schistosomiasis treatment campaigns with intravenous tartar emetic, carried out in the countryside in the 60’s- early 80’s, ignited this epidemic through reuse of insufficiently sterilised needles and syringes [7]. Since then, cross-sectional studies have shown unsafe injection practices, history of blood transfusion, invasive medical procedures, and instrument-assisted birth deliveries as associated with HCV infection [8–10]. Intra-familial transmission may also have played an important role, as evidenced in two recent cohort studies

39. Research question (1/2)
Provided you conduct a case-control study with acute symptomatic hepatitis C patients identified at the Fever Hospitals in Cairo,
How would you select controls?
The highest HCV prevalence in the world occurs in Egypt at an estimated 12% [1], i.e. 10 to 20 fold higher than in Northern Europe [2] or in the United States [3]. The bulk of chronic infection is age-related [4] and occurs among persons of rural origin. Cohort studies have estimated a 9% prevalence and 0.8/ 1000 person-years incidence in Upper Egypt, and a 24% prevalence and 6.8/1000 incidence in the Nile Delta [5,6].
The widespread schistosomiasis treatment campaigns with intravenous tartar emetic, carried out in the countryside in the 60’s- early 80’s, ignited this epidemic through reuse of insufficiently sterilised needles and syringes [7]. Since then, cross-sectional studies have shown unsafe injection practices, history of blood transfusion, invasive medical procedures, and instrument-assisted birth deliveries as associated with HCV infection [8–10]. Intra-familial transmission may also have played an important role, as evidenced in two recent cohort studies

40. MediPIET Scientific Coordinator
Thank you!
Dr Adela Paez
MediPIET Scientific Coordinator

41. References
Rothmann KJ, Greenland S. Modern epidemiology. Lippincott-Raven 1998.
Hennekens CH, Epidemiology in Medicine. Lippincott-Williams and Wilkins 1987.

42. Investigation 2 Neonatal ward of hospital X
30 cases of Klebsiella pneumoniae (2 deaths)
The hospital epidemiologist wants to identify potential risk factors
How would you select controls?

43. Investigation 3
250 cases of tick-borne encephalitis (TBE) reported anually in two regions (population: 8 milion) of country X
10% of infected develop disease
Infection and vaccination lead to long-lasting immunity
Risk factors not well established
Contact with ticks (occupational and recreational outdoor activities)
How would you select controls to identify risk factors?

44. Investigation 4
In January 2014, the Public Health Agency of Sweden observed a nationwide increase in notified listeriosis cases between October 2013 and March 2014.
Isolates from 32 cases had identical Pulsed Field Gel Electrophoresis (PFGE) patterns, suggesting a common source. The median age of cases was 77 years and 23 (72%) were female.
It’s now June How would you select the controls? What questions would you ask?

45. What if: Only non-cases as controls?
If attack rate is low,
non-cases likely to represent exposures in source population
can use them as controls
If attack rate is high,
non-cases unlikely to represent exposure in source population
OR may be over-estimated
More accurate not to exclude cases from control group
Concept of case-cohort studies!

46. Case-cohort design
Aim of control group: properly reflect the exposure in the source population
Source pop originally includes people who will later can develop the disease
Controls selected from all individuals at risk at the start of the study
sampled regardless whether or not they will fall ill
No need to document disease status among controls
OR estimates relative risk