EPIDEMIOLOGY – TUTORIAL 5

STUDY DESIGN
Types of Epidemiologic Research
There are two broad types of epidemiologic studies, experimental and observational. An
experimental study uses randomization to allocate subjects to different categories of the
exposure. An observational study does not use randomization. (For additional information
on randomization, please refer to the end of this chapter.) In experimental studies, the
investigator, through randomization, determines the exposure status for each subject, then
follows them and documents subsequent disease outcome. In an observational study, the
subjects themselves, or perhaps their genetics, determine their exposure, for example,
whether to smoke or not. The investigator is relegated to the role of simply observing
exposure status and subsequent disease outcome. Experimental studies in epidemiology
usually take the form of clinical trials and community intervention trials. The objective of
most clinical trials is to test the possible effect, that is, the efficacy, of a therapeutic or
preventive treatment such as a new drug, physical therapy or dietary regimen for either
treating or preventing the occurrence of a disease. The objective of most community
intervention trials is to assess the effectiveness of a prevention program. For example, one
might study the effectiveness of fluoridation, of sex education, or of needle exchange.

Most epidemiologic studies are observational. Observational studies are broadly identified
as two types: descriptive and analytic. Descriptive studies are performed to describe the
natural history of a disease, to determine the allocation of health care resources, and to
suggest hypotheses about disease causation. Analytic studies are performed to test
hypotheses about the determinants of a disease or other health condition, with the ideal
goal of assessing causation.

1. Figures 1-4 give brief summaries or extracts from the methods sections of different
studies published in various journal.

Figure 1:
Subjects were followed from enrolment (1st November 2005 to 31st December 2007) to 31st
December 2011. Incident and prevalent cancer cases were identified by cross-linkage with
population-based cancer registries in the 11 geographical areas, and were coded according to
the International Classification of Diseases for Oncology, 3rd edition (ICD-O-3). These
registries meet high-quality criteria: the completeness and quality of data are regularly
assessed by the Comité National des Registres [French Institute of Health and Medical
Research (INSERM), National Cancer Institute (Inca) and the French Institute for Public Health
Surveillance (InVS)], and data are regularly published by IARC in Cancer Incidence in the series
Five Continents (CI5). Matching with cancer registries was based on married and maiden
names, first names, gender, date and place of birth, place of residence, vital status and date
of death (if applicable). Only malignant tumors (exclusion of in-situ) were used in this analysis,
with the exception of non-melanoma skin cancers due to their non-exhaustive registration.
Vital status and place of residence were checked annually using the MSA data, the French
National Postal Service (La Poste) and the French National Death Index (Répertoire National
pour l’Identification des Personnes Physiques). People moving outside the AGRICAN area
(0.8%) were no longer followed for cancer diagnosis. Person-years accumulation was
calculated from the date of reception of the enrolment questionnaire and ended at cancer
diagnosis, date of death (11%), date of loss to follow-up (less than 2% of subjects) or 31st
December 2011, whichever occurred first. For a given type of cancer, diagnosis between 1st
January 2005 (date of implementation of the most recent registry in the areas concerned) and
enrolment was considered as prevalent and excluded from analysis. Earlier information was
not considered because of a major heterogeneity between registries (implementation of
registries extended over 30 years) in order to avoid selection bias between areas.
(Lemarchand et al., 2017)
Please identify this following:
1. Outcome of interest
2. How do researcher know the cancer status of study participant?
3. Type of study design

Figure 2:
Eligible subjects were men, younger than 76 years at diagnosis or selection, residents of
Greater Montreal, registered on Quebec’s permanent electoral list (continually updated) and
Canadian citizen. Cases were all patients newly diagnosed with primary Prostate Cancer,
actively ascertained through pathology departments across the main French hospitals (7 out
of 9 hospitals) in the Montreal area between 2005 and 2009. Based on registry information,
this covered at least 80% of all Prostate Cancer cases diagnosed in the area during the study
period. The degree of aggressiveness of Prostate Cancer, defined by the Gleason score, was
extracted from prostate biopsy pathology reports. Control subjects were selected
concurrently from the population-based provincial electoral list of French-speaking men, and
frequency-matched to cases by 5-year age groups.
Eligible subjects were sent an introductory package and were reached by telephone by
interviewers to set up an appointment. In-person interviews collected socio-demographic and
lifestyle characteristics, family history of cancer, medical and Prostate Cancer screening
histories, and self-reported weight and height. Subjects reported their overall physical activity
level at work (Very/Moderately/Not very active), at home (Very/Moderately/Not very active)
and their engagement in any leisure physical activity during adulthood, along with their
frequency of use of 44 fruit and vegetables.
For subjects who reported ever consuming alcohol once a month for one year or more,
lifetime alcohol consumption was recorded for beer, wine and spirits. Drinks were reported
in commonly used servings, i.e., 375 ml for beer, 125 ml for wine and 45 ml for spirits. For
each beverage type, each time the pattern of intake changed, participants were asked to
report their drinking habits, including the time period (age started and age ended) and the
frequency of drinking (number of drinks per month, week or day). This allowed taking into
account changes in intake levels over the lifespan.
Lifetime cumulative exposure variables were created for each type of beverages. These were
defined as the product of the average number of drinks consumed per day and the duration
of drinking in years for beer, wine, and spirits, and expressed as drink-years.
A composite exposure variable was also constructed to express the cumulative exposure to
total alcohol by taking into account the ethanol content by volume of each type of beverages.
Using the quantities of 14.8 g of ethanol per drink of beer, 11.8 g per drink of wine, and 14.2 g
per drink of spirits, we calculated the cumulative intake in total grams of ethanol. The latter
was divided by 14, which corresponds to the average amount of ethanol per drink weighted
by the proportion of each type of drink in our study population, to estimate the total amount
of alcohol used, in drink-years, standardized for ethanol content. (Demoury, Karakiewicz and
Parent, 2016)
Please identify this following:
1. Exposure of interest and how they measure it

2. Outcome of interest and how they measure it

3. Type of study design

Figure 3:
Participants were Caucasian men and women aged 20–49, recruited through random digit
dialing (RDD) from the population of Metro Vancouver, Canada. We aimed for a similar
number of cases and controls, with no matching. Cases were persons with hip pain defined as
pain in the groin or upper thigh in the past 12 months that lasted longer than 6 weeks or
occurred on three or more occasions. Controls were persons without hip pain selected
randomly from the source population for the cases. Pregnant women and persons with
bilateral hip replacement were excluded. Subjects who agreed to participate in the study filled
out a self-administered questionnaire and obtained X-rays of both hips.
In the self-administered questionnaire, we asked about any experience of pain, stiffness or
discomfort in the left and right groin or the front of the upper thigh as well as duration and
frequency of these symptoms. The questions were based on guidelines from the
literature and extensively validated through cognitive interviewing prior to the study. Hip
injury was defined by the following question: Have you ever had a hip injury that required you
to use a walking aid (e.g., cane or crutch) for at least 1 week?
All subjects completed a previously validated self-administered online lifetime PA
questionnaire that covered occupational, domestic and recreational activities. Detailed
questions asked about the amount of time per day spent sitting, standing, standing while
holding objects (>50 lb/23 kg, >20 lb/9 kg, or a child), walking, walking while carrying objects
(>50 lb/23 kg, >20 lb/9 kg, or a child), moving, pushing objects (>75 lb/34 kg, >50 lb/23 kg, or
a stroller), using heavy tools, squatting and kneeling. We also asked about the frequency (per
hour of activity) of squatting/bending knees without/with lifting objects (<20 lb/9 kg,
>20 lb/9 kg) and climbing/descending stairs or ladders. The recreational activity component
consisted of a list of 68 sports and recreational activities. For each type of activity that was
performed on at least 100 occasions during lifetime, detailed questions were asked about the
age activity started and ended, average number of months per year, frequency per week (or
month/year), minutes/hours on each occasion, and typical time per hour of activity spent
sitting, standing, walking, running/jogging, squatting or knee bending with/without lifting or
force, and jumping. We calculated energy expenditure in terms of metabolic equivalents
(METs) using MET values for eight generic types of activity (sitting, standing, walking, running,
squatting, jumping, kneeling, and climbing stairs) and five specific sport activities (swimming,
skating, skiing, biking, and rowing) obtained from the Compendium of Physical Activities. The
average number of MET-hours per year was calculated by multiplying hours of activity for
each activity type by the corresponding MET values, summing up over lifetime (from age 10)
and dividing by the number of years.
Standardized X-rays of the pelvis, anteroposterior (AP) view and Dunn view of both hips were
obtained, as described in detail in a previous publication. For the AP pelvis view, the subject
was in a weight-bearing position, with legs internally rotated 15°. For the bilateral Dunn view,
the subject was supine and the hip was positioned in 45° flexion and 20° abduction while
maintaining neutral rotation. Cam morphology was assessed in Dunn view and pincer
morphology was assessed in AP view. Cam was defined as alpha angle >55°. Pincer was
defined as lateral centre edge (LCE) angle >40° or positive crossover sign (figure of 8) 31. CPM
was considered to be present if CPM was present in at least one hip. These radiographic
features were measured by a trained reader using standardized methods. In a reliability study
in 50 subjects using the same rater, intra-rater prevalence-adjusted bias-adjusted kappa for
CPM diagnosis was 0.76. (Kopec et al., 2017)

Please identify this following:

1. Exposure of interest and how they measure it
2. Outcome of interest and how they measure it
3. Type of study design

Figure 4:
Between April 2009 and October 2012 (the study period), a multicenter study was conducted
in Japan. Newly diagnosed cases of pertussis were recruited at 4 collaborating hospitals in 4
different areas of Japan: (from north to south) Chiba, Saitama, Mie, and Fukuoka. Eligible
cases were newly diagnosed pertussis patients who satisfied the clinical criteria for pertussis
and whose age at diagnosis was less than 30 years. The clinical criteria for pertussis were:
cough lasting for more than 7 days with one or more symptoms (paroxysmal cough, whoop,
or posttussive (post cough) vomiting) and one of isolation of Bordetella pertussis, positive
results by the loop-mediated isothermal amplification (LAMP) method, serodiagnosis (for
paired serum samples at the acute phase and at the recovery phase, at least twofold increase
of IgG antibody for pertussis toxin (PT-IgG) or fourfold increase of agglutinin titer, while for a
single serum sample at the acute phase, PT-IgG of 10 EU/mL or more among unvaccinated
subjects or 100 EU/mL or more among vaccinated subjects) or epidemiologically linked to a
confirmed pertussis case. During the study period, a pertussis outbreak occurred in Saga
University [14], where one of the investigators worked. Thus, if cases diagnosed in Saga
University satisfied the clinical criteria in the present study, they also contributed to the
present study.
Regarding the recruitment of control subjects, each case was asked to provide up to five
friend controls, of the same age (or school grade) and sex as the case. Exclusion criteria were:
presence of lasting cough for more than 1 week during the 1 month prior to case diagnosis.
During the study period, however, it turned out that some cases (particularly preschool
children) did not have any friends and could not provide any friend controls. Thus, not only
friend controls but also hospital controls were recruited for cases who were enrolled since
April 2012. Collaborating hospitals were encouraged to select up to five hospital controls
among patients without pertussis, matching for age and sex.
The following information was obtained by means of a self-administered questionnaire
completed by each child's parent or guardian: sex, date of birth; history of pertussis; history
of DTaP vaccination, number of vaccinations, vaccination dates, vaccine manufacturer and
vaccine lot number if vaccinated; underlying illnesses (heart disease, renal disease, liver
disease, diabetes mellitus, anemia, asthma, other respiratory diseases, tonsillitis, atopic
dermatitis, allergic rhinitis, allergic conjunctivitis, immunodeficiency, epilepsy); history of
steroid treatment for more than one month; total room space in the house (m 2); number of
family members; contact with a confirmed pertussis case during the recent one month; and
contact with a person with a lasting cough during the last month. In Japan, the vaccination
history is usually recorded in an immunization record book maintained by individuals. Thus,
the information collected about vaccination status was confirmed by the immunization
record. When missing answers or illogical data were detected by research technicians,
research technicians conducted a telephone interview to complete the data.
In addition, for pertussis cases, the following clinical findings were reported by the
pediatricians-in-charge using a standardized questionnaire: date at symptom onset; date at
diagnosis; disease symptoms (paroxysmal cough, whoop, posttussive vomiting, fever,
dyspnea, and seizures); and laboratory examinations (culture isolation of B. pertussis and
results by the LAMP method, and PT-IgG and agglutinin titers in the acute and recovery
phases). (Ohfuji et al., 2015)
Please identify this following:
1. Exposure of interest and how they measure it
2. Outcome of interest and how they measure it
3. Type of study design
2. Complete Table 1 to show the relative strengths and limitations of the main study designs,
scoring each one on a scale from 1 = poor (e.g. not good to investigate a rare disease or
very expensive) to 5 = excellent (e.g. very good to investigate rare exposure or very quick
to do) and then discuss with your tutor.

Table 1. Comparing the strengths and weaknesses of different study designs.

Consideration Cross- Case- Cohort RCT

sectional control
Investigation of rare disease or
outcome
Investigation of a rare exposure
Testing multiple effects of an
exposure
Study of multiple exposures
Establishing temporality*
Give a direct measure of incidence
Explore exposures which change
over time
Time required
Low cost
Ethical problems
*i.e. that the exposure came before the outcome