Observational Study Design

Question 1

What is the difference between observational study vs experimental study?

Answer 1

in experimental study the exposure is assigned by the researcher

Question 2

What are the 4 types of observational studies (analytical)?

Answer 2

• cohort
• case control
• cross-sectional

Question 3

What are the differences between the 4 observational studies (analytical)?

Answer 3

in cohort the exposure happens first and you follow your participants overtime to see the outcome

for case control, you pick the participants who already have the outcome and you try to identify the exposure

cross sectional is a snapshot of what is happening right now so the exposure and the outcome are measured at the same time.

Question 4

Descriptive study help identify….

Answer 4

patterns and generate hypothesis

Question 5

What are descriptive studies?

Answer 5

case-report
surveys
fact finding

Question 6

Descriptive studies can

Answer 6

give frequency and distribution of single variables

Question 7

Analytical studies aim to

Answer 7

associate exposure/intervention and outcome

Question 8

Analytical studies help

Answer 8

generate hypothesis
explore associations between diet and health
build up evidence to support a suggested effect

Question 9

Analytical studies cannot

Answer 9

show cause and effect

Question 10

Explain ecological studies

Answer 10

characterize population groups
does not link individual’s exposures to health outcomes
associations between group indicators of diet/nutritional status AND group indicators of health status

Question 11

Ecological studies need 2 population-based measures:

Answer 11

• one for exposure of interest

- one for health outcome

Question 12

What is the one important thing about ecological studies?

Answer 12

individuals may or may not be the same as those providing data for health outcome

Question 13

What is the difference between ecological studies and cross-sectional study?

Answer 13

in ecological studies individuals may or may not be the same as those providing data for health outcome

Question 14

Give an example of an ecological study

Answer 14

let’s say we want to look at fluoride in drinking water and disease outcome but we cannot measure individual fluoride intake so we use ecological study.

Question 15

What are the indices of dietary intake in ecological studies?

Answer 15

survey data collected for the study or pre-existing data

ex. FAO food balance sheet
ex. household budget survey
ex. individual survey data

Question 16

What the indices of health outcomes in ecological studies?

Answer 16

Government reports

WHO publications

Question 17

How do we analyze ecological studies?

Answer 17

Correlation coefficient “r”

multiple regression modelling

Question 18

What is multiple regression modelling?

Answer 18

estimate magnitude of associations while taking into account other factors

Question 19

What are the limitations of ecological studies?

Answer 19

“ecological fallacy”: when relationships observed for groups are assumes to hold for individuals
ex. higher disease rate in one country may reflect greater proportion of older people

Question 20

What is a cross-sectional study?

Answer 20

snapshot about the population (prevalence)

association between exposure and health

Question 21

What are the pros and cons of cross-sectional studies?

Answer 21

pros:
relatively cheap and simple

cons:
cannot determine causation, recall bias problems, confounders may be unequally distributed

Question 22

Define prevalence

Answer 22

the percentage of a population that is affected with a particular disease at a given time

Question 23

What are some methods of cross-sectional studies?

Answer 23

describing population characteristics (should be as nationally representative as possible, as accurate and as complete as possible)

NHANES
Canadian Health Measures Survey

Question 24

How to do the analysis of cross-sectional studies?

Answer 24

correlations

regression modelling

Question 25

What are the limitations of cross-sectional studies?

Answer 25

- exposure and outcome measured at the same time --> cause and effect? - difficult to draw conclusions about short-lived exposures (environmental chemicals) and chronic diseases - snapshot may not capture details of exposure from people with acute fetal diseases

Question 26

Why doesn't NHANES accept volunteers?

Answer 26

Bc it is free health examination and they might take advantage of that. You might get more sick people coming than healthy people and this can skew the results of the study. The results of the survey will overestimate the prevalence of disease and this would lead to self-selection bias

Question 27

Explain case-control studies

Answer 27

Matching cases (people with disease) with controls (people without disease) and measuring past exposure to dietary factors, estimate risk associated with the risk factor (retrospective)

Question 28

What are the pros and cons of case-control studies?

Answer 28

Pros: - quicker, cheaper - useful for rare conditions - can have greater statistical power Cons: - challenges: choice of appropriate control (confounders), measurement of previous dietary exposure (recall bias, reporting bias).

Question 29

What is reporting bias?

Answer 29

if people know what you are looking for then they might subconsciously no tell you the real answer for example underreporting fat intake in a study that looks at fat intake and breast cancer.

Question 30

What are the methods of case-control studies?

Answer 30

- chose population with adequate supply of cases & controls - need diversity of exposure to dietary risk factors being studies

Question 31

How to select cases for case-control studies?

Answer 31

- incident cases - newly diagnosed - prevalent cases - more common, easier to find - specificity of diagnosis!

Question 32

Where to find cases for case-control studies?

Answer 32

- hospitals - general practice records - general population

Question 33

How to select controls for case-control studies?

Answer 33

- should be from the same population or represent source population - balance confounders between 2 groups - should be independent of exposure status (both cases and controls from screening clinics-avoid self-selection bias) - can match more than 1 control with a case --> increase power - match cases & controls on variables such as age, sex, etc - beware of over-matching.

Question 34

What are the challenges in case-control studies?

Answer 34

- identify the past dietary behavior - disease with long pre-clinical phase: exposure may have occurred many years before diagnosis - accurate reporting of past diet - answers on past diet are influences by current eating patterns - some diseases are accompanied by changes in diet

Question 35

Why do we want diversity in exposure in case-control studies?

Answer 35

There could be an error associated with dietary measurements so just the rate of error in the dietary measurements might be able to hide the potential associations that you are looking for. So, if those in the study population are similar regarding the dietary exposure the differences between the dietary exposure might be smaller than your measurement error so you might not be able to catch anything. So, you need to have people with low-fat intake, medium, high fat intake... bc you just want to make sure that the measurement error of you assessing fat intake is smaller than the actual differences in the fat intake between people. You might miss and effect if the exposure is pretty constant between people.

Question 36

What are the limitations of selecting cases from prevalence?

Answer 36

need to be careful with how to interpret the associations bc the effect of the diet might be the effect on survival rather than the effect on the development of disease, if you are using prevalent cases the disease is already there so you are not necessarily looking at how the diet is affecting the disease progressing.

Question 37

Why do we want to get our controls from the same population as the cases?

Answer 37

to balance the confounders between two groups.

Question 38

What happens if you overmatch your cases with your controls in a case-control study?

Answer 38

you might be causing selection bias So, by doin that you are reducing the efficiency of detecting associations. Ex: fat intake and T2DM, if you match the cases and the controls on BMI you might be overmatching bc BMI is on the causal pathway between fat intake and diabetes development so if you make everybody equal in terms of BMI, you are now cancelling out the possibility of seeing an association between fat intake and T2DM.

Question 39

What are nested case-control studies?

Answer 39

- developed from cohort studies - subset of controls from cohort compared to incident cases - when exposure is difficult or expensive to obtain - use data previously collected from a large cohort study--> save time and $$ - reduce impact of recall bias

Question 40

What are the pros of nested case-control studies?

Answer 40

- use data previously collected from a large cohort study--> save time and $$ - reduce impact of recall bias

Question 41

How do we analyze case-control studies?

Answer 41

- odds ratio

Question 42

What is odds ratio?

Answer 42

odds that an event will occur given a particular exposure/odd of the event occurring in the absence of the exposure ex: if you had an odds ratio of 2, it means that the people who have had the exposure are twice as likely to have the disease than the people who did not have the exposure.

Question 43

What are the limitations of case-control studies?

Answer 43

- recall bias --> dietary measurement error | - choice of controls

Question 44

Explain cohort studies

Answer 44

Defining the population need large number of individuals --> generalize findings to population representative sample --> randomly select form subset of sampling frame stratified by age, gender, SES, etc.

Question 45

What is cluster sampling?

Answer 45

With cluster sampling, the researcher divides the population into separate groups, called clusters. Then, a simple random sample of clusters is selected from the population

Question 46

Cohort studies can measure...

Answer 46

incidence

Question 47

How do we analyze the cohort data?

Answer 47

Relative risk ratios if F/U time is similar for all individuals = cumulative incidence in the exposed group compared to the cumulative incidence in the unexposed group

Question 48

What is incidence?

Answer 48

probability of occurrence of a given medical condition in a population within a specified period of time.

Question 49

What are the limitations of cohort studies?

Answer 49

- very time consuming, expensive - need very large sample sizes and long F/U periods - methods of collecting and electronically capturing exposures can be time-consuming - new technologies - e.g. smartphone applications, online 24h recalls - not efficient/practical to study rare outcomes - attrition bias - if loss to F/U are related to exposure and outcome - selection bias

Question 50

What does STROBE stand for?

Answer 50

Strengthening the Reporting of Observational Studies in Epidemiology

Question 51

What is STROBE?

Answer 51

reporting guideline of 22 items that are essential for reporting observational studies. Helps reviewers evaluate how well your study was conducted.

Question 52

What are pilot studies? What are its goals?

Answer 52

- exploratory studies - small scale - assess feasibility of full-scale trial - test recruitment procedures - develop/test novel intervention, or research instruments - identify logistical challenges - support funding applications - information on variability and timescale of outcome responses

Question 53

Explain Randomized controlled trials

Answer 53

- Treatment and control groups - control group: placebo or on treatment - 2 basic RCT designs: parallel and cross-over

Question 54

Explain the parallel design in RCTs

Answer 54

- each participant receives one intervention | - comparison on a between-participant basis

Question 55

When should we use parallel design?

Answer 55

- longer term interventions - washout ineffective - intentionally returning to baseline is unethical

Question 56

Explain the cross-over design in RCTs

Answer 56

- each participant receives all interventions - comparison on a within-participant basis - increased power

Question 57

When should we use cross-over design?

Answer 57

- outcomes show large inter-participant variation - restricted participant availability - very short-term studies ex. postprandial glucose measurements ex. seasonal change

Question 58

What are some other RCT designs?

Answer 58

- factorial (you test many effects at once) | - cluster randomized design (you dont randomize individuals but you randomize clusters of individuals)

Question 59

What are quasi-experimental studies?

Answer 59

- similar to RCT but lack one or more key features | - no randomization, no control group

Question 60

When are quasi-experimental studies used?

Answer 60

used when - randomization may reduce effectiveness of intervention - randomization is unethical - randomization is impractical

Question 61

What are the types of controls can we use in quasi-experimental studies?

Answer 61

- concurrent | - historical

Question 62

When are quasi-experimental studies used most often which area of research?

Answer 62

often used in public health interventions

Question 63

What are the pros and cons of quasi-experimental studies?

Answer 63

+ easier, quicker, cheaper, less planning required | - ability to establish causality is compromised

Question 64

Give examples of quasi-experimental studies

Answer 64

- before and after study without control group - before and after study with 2 groups that are not equivalent ex. when you do an intervention study on prosthetics of pharmaceutical pill. Some people may prefer to take a poll, or some people might prefer to use prosthetic, so you want to give them a choice so that they are more compliant to your treatment.

Question 65

Internal validity

Answer 65

Internal validity is the extent to which a piece of evidence supports a claim about cause and effect, within the context of a particular study. It is one of the most important properties of scientific studies, and is an important concept in reasoning about evidence more generally.

Question 66

What is the null hypothesis?

Answer 66

no difference between intervention and control

Question 67

What are the considerations for the duration of the study?

Answer 67

- long enough to observe changes in outcome measure - based on previous studies and knowledge of physiology - short enough to be feasible (ethical reasons, costs/resources, avoid participant fatigue & non compliance) - post study F/U measures

Question 68

What are the considerations for the intervention?

Answer 68

- food, food component, nutrient, diet - study design must reflect intended use of intervention - report frequency of ingestion - amounts/dose based on previous data, underlying physiology, palatability, bioavailability, etc.

Question 69

What are the considerations for control group for an intervention?

Answer 69

- food substance that does not provide component being tested - composition should be documented - matches for sensory characteristics - easy with supplementation studies - foods/diets - blinding may not be possible - comparing diets --> scoring scheme for self-reported diet adherence - groups not receiving dietary advice will have lower scores

Question 70

What are the considerations for outcome measures in intervention studies?

Answer 70

- one primary outcome, one or more secondary outcome measures - power calculation based on primary outcome measure - specify - one time point? Several? - must have biological relevance - reduce measurement errors by standard protocols, train observers, average several instruments - analytical variability - duplicate or triplicate measurements

Question 71

What are the considerations for participants in intervention studies?

Answer 71

- eligibility criteria - must be accurately described - strict criteria

Question 72

What are the pros and cons of having a strict criteria for participants in an intervention study?

Answer 72

+ reduce inter-participant variation, more homogenous | - limits generalizability of findings

Question 73

Why do we randomize participants?

Answer 73

- study groups comparable with respect to known and unknown factors - potential confounders are equally distributed between groups - increases internal validity of study

Question 74

What is blinding?

Answer 74

concealment of intervention allocations

Question 75

What are the consequences of a small study?

Answer 75

fail to detect important differences

Question 76

What are the consequences of a large study?

Answer 76

waste resources, unethical

Question 77

What is a sequential design?

Answer 77

analyze results as they become available | + quickly identify inferior interventions, minimize # of people

Question 78

How to estimate the sample size? (Effect size)

Answer 78

specify magnitude of smallest meaningful difference in outcome variable

Question 79

Why should we consider the drop-out rate or non-compliance in the power calculations?

Answer 79

let's say around 20 drops out then you need to add that number to your sample size.

Question 80

Why to register trial in publicly accessible database before recruitment?

Answer 80

- reduce consequences of non-publication - discourage selective reporting of outcomes - can be a condition for publication

Question 81

What does CONSORT stand for?

Answer 81

Consolidated Standards of Reporting Trials