Basics of Scientific Research

Class #4

Observational studies:
cohort study, case-control study,
cross-sectional study
Ekaterine Karkashadze, MD, MS
Study designs
Experimental • Studies preventions and treatments for diseases
• Investigator actively manipulates which groups
▪ Randomized control trial
receive the agent under study
• Derives highest level evidence

Observational ▪ Studies causes, preventions and treatments for

diseases
▪ Cohort
▪ Investigator passively observes as nature takes its
▪ Case-control course
▪ Cross-sectional ▪ Prone to errors more than experimental
Hierarchy of evidence

Randomized
control trial
Cohort study

Case-Control Study

Cross-sectional study
Measures of disease frequency
Prevalence
Relates to Existing cases of disease/condition;
Proportion of people that has a disease at a given point in time (e.g. how many
people in the country has hepatitis C as for today?)

Incidence
Relates to NEW cases of disease/condition;
Proportion or rate showing occurrence of new cases during a period of time.
• Cumulative incidence (expressed as proportion) – proportion of people who
newly developed a disease over a period of time;
• Incidence rate (expressed N cases per 100 person years (person time)) – how
many new cases of a disease developed per 100 person years.
Cohort studies
Cohort studies
• Although experimental studies are scientifically more rigorous, they are often
infeasible because of challenges related to patient recruitment, high costs, and
ethical considerations

• Studying impact of smoking on fetus development in pregnant women

• Experimental study (randomized controlled trial): Assign newly pregnant women to
either smoking or nonsmoking ⌧ unethical
• Observational study: compare women who choose to smoke during pregnancy with
those who choose not to smoke 🗹 ethical
 Real-life Example of Importance of Cohort
Studies
• In 2013 three randomized controlled trials (experimental studies) documented high
efficacy of a new drug to treat HIV/AIDS – Dolutegravir (DTG)
• Treatment with Dolutegravir (DTG) was widely implemented worldwide

DTG 50 mg QD + 2 NRTIs*
ART-naive pts (n = 411)
SPRING-2
(active controlled, VL ≥ 1000 c/mL
RAL 400 mg BID + 2 NRTIs*
double blind) (N = 822) (n = 411)

ART-naive pts DTG 50 mg QD + ABC/3TC QD

SINGLE VL ≥ 1000 c/mL (n = 414)
(active controlled, HLA-B*5701 neg
double blind) CrCL > 50 mL/min EFV/TDF/FTC QD
(N = 833) (n = 419)

DTG 50 mg QD + 2 NRTIs*
ART-naive pts (n = 242)
FLAMINGO
(open label) VL ≥ 1000 c/mL DRV/RTV 800/100 mg QD + 2 NRTIs*
(N = 484) (n = 242)
 Real-life Example of Importance of Cohort
Studies

2.5 • In 2018 observational study from

Botswaba called “Tsepamo”
Neural Tube Deffects

2
reported higher rates of Neural
(%, 95% CI)

1.5
Tube Defects in infants born to
1 0.94
HIV+ mothers who at the time
0.5
conception were taking
0.12 0.05 0.00 0.09
0
DTG Any non DTG EFV DTG HIV- Dolutegravir based treatment
ART
Conception Pregnancy

Conception
Real-life Example of Importance of Cohort
Studies
• Based on results of “Tsepamo” study World Health Organization
changed the HIV/AIDS treatment guidelines

• Dolutegravir is not recommended in:

• Women during the first trimester of pregnancy
• Women of childbearing potential not using effective contraception
 Cohort study design
Cohort study on association between tobacco smoking and heart disease

Follow-up Assessing outcomes

Outcome
Exposed (smoking) No outcome
Exposure of
Eligible subjects
interest: smoking
Not exposed (not Outcome
smoking) No outcome

Self decision to smoke or not

Types of cohort studies
Type of population Follow-up
studied
Closed Does not gain member
No losses occur
All persons start follow-up at same time and are
followed for the same amount of time
(e.g. 100 persons followed for 2 years)
Open Members come and go
Losses may occur
All individuals do not start follow-up at the same time
and each person is followed for different amount of
time (e.g. of 100 persons 30 are followed for 2 years,
30 are followed for 1.5 years, 20 are followed for 6
months, 20 are followed for 1 year)
“Diamond princess” as Closed Cohort

• “Diamond Princess” - A cruise ship with COVID-19 outbreak.

• A passenger from Hong Kong embarked in Yokohama, sailed one segment of the
itinerary, and disembarked in Hong Kong on 25 January. Six days after leaving the
ship, he tested positive for COVID-19 on 1 February. The ship was quarantined in
Yokohama on Feb 4.
As of today’s date:
• Number of passengers – 3711.
• Number of those who tested positive – 712.
• Number who died – 13.
• Observation started on Feb 4 and continued to final outcome.
 Timing of Cohort Studies
Prospective Cohort Study
* Exposure *

* Disease *
Study starts

Retrospective Cohort Study

* Exposure *

* Disease *
Study starts

Ambidirectional Cohort Study

* Exposure *

* Disease *

Study starts
Directionality of cohort studies
• Regardless of timing the directionality is ALWAYS
forward in a cohort study (logic of reasoning is cause
leading to effect)

Exposure Outcome
Real-life example of retrospective cohort
• Mortality and Causes of Death Among HIV-Infected Individuals in the Country of Georgia:
1989–2012
• Study conducted in 2013-2014

Exposure of
interest:
3554 HIV patients Stage 3 diseases
at the time of
HIV diagnosis
 Real-life example of retrospective cohort
• Mortality and Causes of Death Among HIV-Infected Individuals in the Country of Georgia:
1989–2012
• Study conducted in 2013-2014
Follow-up until 31 December 2012

Exposure of
interest: Stage 3 (n=1715)
3554 HIV patients Stage 3 diseases
at the time of
HIV diagnosis No stage 3 (n=1639)
 Real-life example of retrospective cohort
• Mortality and Causes of Death Among HIV-Infected Individuals in the Country of Georgia:
1989–2012
• Study conducted in 2013-2014
Follow-up until 31 December 2012 Outcome

Exposure of 632 died

interest: Stage 3 (n=1715) 1084 alive
3554 HIV patients Stage 3 diseases
at the time of 147 died
HIV diagnosis No stage 3 (n=1639) 1492 alive
 Real-life example of retrospective cohort
• Mortality and Causes of Death Among HIV-Infected Individuals in the Country of Georgia:
1989–2012
• Study conducted in 2013-2014
Follow-up until 31 December 2012 Outcome

Exposure of 632 died

interest: Stage 3 (n=1715) 1084 alive
3554 HIV patients Stage 3 diseases
at the time of 147 died
HIV diagnosis No stage 3 (n=1639) 1492 alive

Forward directionality
Exposure Outcome
 Real-life example of retrospective cohort
• Mortality and Causes of Death Among HIV-Infected Individuals in the Country of Georgia:
1989–2012
• Study conducted in 2013-2014
Follow-up until 31 December 2012 Outcome

Exposure of 632 died

interest: Stage 3 (n=1715) 1084 alive
3554 HIV patients Stage 3 diseases
at the time of 147 died
HIV diagnosis No stage 3 (n=1639) 1492 alive

Mortality among exposed: 632 / 1715 * 100% = 36.9%

Mortality among not exposed: 147 / 1639 * 100% = 9.0%
 Real-life example of retrospective cohort
• Mortality and Causes of Death Among HIV-Infected Individuals in the Country of Georgia:
1989–2012
• Study conducted in 2013-2014
Follow-up until 31 December 2012 Outcome

Exposure of 632 died

interest: Stage 3 (n=1715) 1084 alive
3554 HIV patients Stage 3 diseases
at the time of 147 died
HIV diagnosis No stage 3 (n=1639) 1492 alive

Mortality among exposed: 632 / 1715 * 100% = 36.9% Risk of mortality if HIV patient has Stage 3 disease
Mortality among not exposed: 147 / 1639 * 100% = 9.0% at the time of HIV diagnosis: 36.9% / 9.0% = 4.1
Case-control studies
Case-control study
• Case-control studies are always retrospective: both exposure and
outcome already have occurred
• Directionality is always backward
Exposure Outcome (case)
• Cohort vs. Case Control
• Cohort: Does tobacco smoking cause lung cancer?
• Case-control: Was lung cancer caused by tobacco smoking?
• Bottom-line: Both cohort and case-control studies are looking at the
same thing just from the different angle
Case-control study design

Smoking Lung cancer (cases) 1. Identify cases

Smoking No lung cancer (controls) 2. Identify controls

• The most difficult part in case-control studies is the selection of

controls
Case-control study: Selection of controls
• Controls are a sample of the population that produced the cases
• More specifically, controls are chosen to represent the level of
risk/protective factors in the source population
• Types of control
• Population controls (e.g. same neighborhood)
• Hospital or clinic
• Case : control ratio – 1:1, 1:2
Example of wrong selection of controls
• Hypothetical study: The role of Pap-smear in preventing cervical
cancer

• Cases: 250 women identified through hospital chart review

• Controls: 250 women from the same neighborhoods as cases,

interviewed at their homes during weekday working hours
Example of wrong selection of controls
• Hypothetical study: The role of Pap-smear in preventing cervical
cancer

• Cases: 250 women identified through hospital chart review

• Controls: 250 women from the same neighborhoods as cases,

interviewed at their homes during weekday working hours

Missed women who were at work

• Wrong selection of controls leads to wrong conclusion 🡪 Pap-smear
does not prevent cervical cancer
• In reality Pap-smear significantly reduces the risk of cervical cancer
Cross-sectional study
Cross-sectional study
• Sometimes also referred as Prevalence Study, because it is usually
conducted to estimate prevalence (burden) of certain disease in a
population at a single point of time
• Temporal ambiguity: Exposure
Outcome
• Injection drug use leads to hepatitis C, if somebody has hepatitis C and
reports injection drugs in cross-sectional, we cannot conclude that injection
drugs use led to hepatitis C. Person may have started injection drug use after
contracting hepatitis C.
Why to conduct cross-sectional study?
• Snapshot of current situation 🡪 informs authorities about the state of
health of the population
• Planning purpose: to guide resource allocation, public health program
• Monitor diseases occurrence over time in specific population groups
Multiple cross-sectional studies over time to
monitor change in prevalence
HIV Prevalence Among MSM in Georgia, 2005-2018
25%

20% 20.7%
20.3%
15%
13.0%
10%

7.0%
5%
4.3% 3.7%

0%
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2017 2018

Action: introduced pre-exposure prophylaxis (PrEP) program in 2017

 5.4% of adults in Georgia
have chronic hepatitis C

150 000 persons

• Georgia launched national elimination program with specific targets to

• Diagnose 135 000 persons
• Treat 128 000 persons
• Achieve cure in 120 000 persons
• Respective action plan developed with budgetary assumptions
Sampling
• Sampling: selecting persons who will participate in the study

Total
Study Population
sample
(7 000)
(3.7 million)

Needs to be representative of the entire population

Sampling approaches
Probability Sampling
• Everybody in the population has
equal chance of being selected for the
study
• Simple random sampling
• Stratified random sampling
• Stratified multistage cluster sampling
• Ensures representativeness of entire
population!
Non Probability Sampling
• Does not give all the individuals in the
population equal chance of being
selected for the study
• Convenience sampling
• Consecutive sampling
• Snowball sampling
• Is not representative of entire
population
Questions?
Measures of Diseases Frequency
• Cohort study
• Cumulative incidence
• Incidence rate

• Cross-section study
• Prevalence

• Case-control study
• Disease frequency is not calculated in case-control study
2 X 2 Table

Outcome No outcome TOTAL

Exposed a C a+c
Not exposed b D b+d
TOTAL a+b c+d a+b+c+d
Cumulative Incidence in Cohort Studies
Outcome No outcome TOTAL
Exposed a c a+c
Not exposed b d b+d
TOTAL a+b c+d a+b+c+d

• Cumulative incidence is calculated in cohort studies when all persons

enrolled in the cohort have the same period of follow-up (closed cohort)

• If cohort member are followed for different periods of time then use
incidence rate (open cohort)
 Cumulative Incidence in Cohort Studies
Outcome No outcome TOTAL
Exposed a c a+c
Not exposed b d b+d
TOTAL a+b c+d a+b+c+d

Number of new cases (outcome)

Cumulative incidence (CI) = X 100%
Total population
CI is expressed as %, range 0% to 100%

CI(exposed) = a X 100% CI(unexposed) = b X 100% CI(total) = a+b X 100%

a+c b+d a+b+c+d
 Example
• 570 young persons engaging in high risk behavior and 240 young persons
not engaging in high risk behavior were followed and clinically monitored
from 2014 to 2016 to evaluate association between high risk behavior and
new cases of sexually transmitted infection (STI).
• New cases of STI were documented in 46 persons engaged in high risk
behavior in 6 persons not engaged in high risk behavior.

Outcome No outcome TOTAL

Exposed a c a+c
Not exposed b d b+d
TOTAL a+b c+d a+b+c+d
 Example
STI No STI TOTAL
High risk behavior 46 (a) 524 (c) 570 (a + c)
No high risk 6 (b) 234 (d) 240 (b+d)
52 758 810
TOTAL (a+b) (c+d) (a+b+c+d)

CI(exposed) = a X 100% =
a+c

CI(unexposed) = b X 100% =
b+d

CI(total) = a+b X 100% =

a+b+c+d
Incidence rate (IR) in cohort study
No. Person-
Years
A (Diagnosed with disease) Accrued
8

B (Died) 12

C (Lost) 6

D 20

E 14
Total population (A- 60
E)
1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014
Observation period

IR look at occurrence of diseases within the total time spent at risk

5 Patients (A-E) were at risk for 60 person-years and 1 person developed disease
Incidence rate (IR) in cohort study

• Incidence rate (IR) is (1) the number of people in a population who

developed disease divided by (2) the sum of "time at risk" each member of
the population contributed while in the study
• 1 case/ 60 person-years (PY) = 0.02 cases per 1 person-year of follow-up
• Usually this is multiplied by 100 (or 1 000, 100 000)
• 1 case / 60 PY *100 = 2 cases per 100 PY of follow-up
• IR of 2 / 100 PY means that if we follow 100 persons for 1 year period 2
persons will develop disease
 Prevalence
• Proportion of individuals in a population with disease or condition at a
specific point of time
• Expressed in %

No. of observed cases at time t

Prevalence
Total no. of individuals at time t

EXAMPLE: Researchers decided to evaluate prevalence of chronic Hepatitis C in a region

X with 8 458 adult residents. Anti-HCV test was used to determine Hepatitis C status. The
test was positive in 169 persons. Calculate the prevalence.

Prevalence of Hepatitis C among adult population of region X:

169 / 8 458 * 100% = 1.99 % (About 2%) 43

 Exercise
• 300 persons aged 40-50 years old were followed for 5 years to determine
association between regular alcohol consumption and hip fracture. 100
persons reported daily consumption of alcohol and 200 persons reported
that they do not consume alcohol on regular basis. At the end of follow-up
10 persons who consumed alcohol daily had hip fracture. Among persons
who did not consume alcohol on daily basis also 10 persons had hip
fracture.

• Fill out 2 X 2 table?

• What is Cumulative Incidence (CI) of hip fracture in exposed population?
• What is Cumulative Incidence (CI) of hip fracture in unexposed population?
• What is Cumulative Incidence (CI) of hip fracture in total population?
 Cumulative Incidence in Cohort Studies
Outcome No outcome TOTAL
Exposed a c a+c
Not exposed b d b+d
TOTAL a+b c+d a+b+c+d

Number of new cases (outcome)

Cumulative incidence (CI) = X 100%
Total population
CI is expressed as %, range 0% to 100%

CI(exposed) = a X 100% CI(unexposed) = b X 100% CI(total) = a+b X 100%

a+c b+d a+b+c+d
Cumulative Incidence in Cohort Studies
Fracture No fracture TOTAL
Daily alcohol 10 90 100
No daily alcohol 10 190 200
TOTAL 20 280 300

CI(exposed) = a X 100% = 10 X 100% = 10%

a+c 100
10
CI(unexposed) = b X 100% = X 100% = 5%
b+d 200
a+b 20
CI(total) = X 100% = X 100% = 6.6%
a+b+c+d 300
Homework
• Reading from the course book (Ann. K. Allen. Research skills for
Medical Students.SAGE Publication Inc. 2012)
• Pages 55-71