STUDY DESIGN - LEARNING OUTCOMES

Question 1

Describe the main features of case reports and series.

Answer 1

Most basic type of descriptive study and documents an individual’s medical experience. A clinician may see an interesting case and describe what he/she has seen. A case series is an extension of a case report and is based on a small group of individuals. These studies are useful for hypothesis generating, but because there is no comparison group, their use is limited and statistical relationships between exposure and outcome cannot be assessed.

Question 2

Describe the main features of ecological studies.

Answer 2

Ecological studies look at the association between exposure and disease on a population or area level rather than on an individual level. With ecological studies any associations seen are on a population level, and we cannot assume that this transfers to an individual level. To assume that associations seen on a population level apply to the individual level is called the ecological fallacy.

Question 3

Describe the main features of cross-sectional studies.

Answer 3

A cross-sectional study collects observations on people at one point in time, thereby providing a ‘snap shot’ of the health of the population. Cross-sectional studies only provide information about disease prevalence and not disease incidence. Study subjects should be selected so they are representative of the target population.

Question 4

Describe the main features of case-control studies.

Answer 4

The case-control study design is useful for looking at risk factors for rare diseases. However, it cannot look at prevalence or incidence of disease. It only looks at whether associations with exposure exist. In a case-control study cases are selected on the basis or presence or absence of disease. A group of individuals with the disease are selected (cases), along with a group of individuals without the disease (controls). Disease frequency cannot be measured because subjects are chosen or sampled according to their disease status.

Question 5

Describe the main features of cohort studies.

Answer 5

Cohort studies measure the incidence of disease. In the traditional prospective cohort study, the study population is ‘free of disease’ at the beginning and the exposure variables are measured and then the population is measured through time to determine their disease outcome, and to compare the risk of disease in those who are exposed and those who are unexposed.

Question 6

Describe the main features of randomised controlled trials.

Answer 6

Controlled trials are human experiment where an investigator implements the intervention (or exposure) or interest and determines which individuals are exposed and unexposed. Randomisation is often used within trials to ensure that the exposed and unexposed groups are similar with respect to all other factors (including other risk factors for the outcome). An RTC is thus equivalent to a laboratory experiment where the investigator changes one variable at a time, keeping all other variables constant.

Question 7

When is it appropriate to use case reports and series?

Answer 7

Case reports and series are useful for generating a hypothesis.

Question 8

When is it appropriate to use ecological studies?

Answer 8

Ecological studies are a good first step in investigating possible exposure-disease relationships especially when there are restraints on time and money.

They are good for hypothesis generating.

They are also good when investigating an exposure which has little variation between individuals within a population/area, but large variations between populations/areas (e.g. some dietary factors).

Question 9

When is it appropriate to use cross-sectional studies?

Answer 9

Cross-sectional studies are good for measuring prevalence of disease. Only measure prevalent (i.e. current) cases of disease.

Good for measuring some continuous measure such as blood pressure or level of a protein in serum.

Gives a snap shot of the health of a particular population at a particular point in time.

Question 10

When is it appropriate to use case-control studies?

Answer 10

Case-control studies are useful for looking at risk factors for rare disease.

A case control study is appropriate when you have a single disease of interest because that is how we are doing our recruitment - it is more difficult to look at multiple outcomes than it would be for a cross-sectional study for example.

When the disease may be rare and you would have to use a very large cross-sectional or cohort study to get anybody with your outcome of interest.

It is useful when we are only looking at associations rather than trying to assess prevalence.

Also useful when looking at exposure(s) that are relatively common.

Question 11

When is it appropriate to use cohort studies?

Answer 11

Cohort studies are appropriate when you want to measure the incidence of a disease. This would traditionally involve starting with a cohort that is free of disease and following them through.

Question 12

When is it appropriate to use randomised controlled trials?

Answer 12

Randomised controlled trials are appropriate in cases where you want to test the effect of a certain exposure in comparison to a control exposure or placebo.

Question 13

List the advantages and disadvantages of case reports and series.

Answer 13

Advantages:

• Allows reporting on unusual medical cases
• Results used to generate hypotheses
• Generates evidence of possible new disease
• Rapid feedback of current events in medical community

Disadvantages:

• Cannot be used to assess statistical associations
• Could just be reporting a medical oddity

Question 14

List the advantages and disadvantages of ecological studies.

Answer 14

Advantages:

• Inexpensive and quick to conduct
• Exposure and disease information is often more readily available by area
• Differences in exposures often greater between areas than between individuals in one area

Disadvantages:

• Results cannot be extrapolated to the individual level
• Systematic differences between areas in recording disease frequency can occur
• Quality of diagnosis
• Differences in classification
• Completeness of reporting
• Sampling of the population can distort the results of an individual
• Data often not available on confounders

Question 15

List the advantages and disadvantages of cross-sectional studies.

Answer 15

Advantages:

• Can be used to examine how much disease there is in a population, and to look at cross-sectional associations between exposures and disease
• Can look at more than one disease and more than one exposure
• Can be relatively inexpensive and quick to conduct and drop-out is not a problem as no follow up

Disadvantages:

• Disease and exposure are measured at the same time therefore no temporal association can be made
• Not suitable for studying rare exposures or rare outcomes
• High possibility of recall / reporting bias

Question 16

List the advantages and disadvantages of case-control studies.

Answer 16

Advantages:

• There is no need to follow people up over time which can be costly, time-consuming and subject to drop-outs
• It reduces the number of disease free people that need to be studied – this is particularly important when the disease of interest is rare
• It is possible to study multiple exposures

Disadvantages:

• No estimate of disease frequency can be obtained
• Bias can be a problem, particularly selection bias
• It is possible to look at one outcome only

Question 17

List the advantages and disadvantages of cohort studies.

Answer 17

Advantages:

• Measures incidence of disease
• Multiple disease outcomes can be examined relating to one exposure
• Often multiple exposures for one disease can be examined
• Exposure is measured before the development of the disease so we can assess the temporal relationship between exposure and disease
• Good assessment of level / duration of exposure and therefore allows us to look at dose relationship between exposure and disease
• Extremely useful for measuring the effect of rare exposures
• Selection / recall bias could potentially be very low

Disadvantages:

• Prospective studies are time consuming and extremely expensive
• Retrospective studies rely on the availability of good and accurate records
• Results can be effected due to loss of follow up (response rate)
• Can be inefficient for the evaluation of rare diseases
• Can be inefficient for exposure-disease relationships with a long latency period (i.e. it takes a long time for the exposure to influence the disease or the disease to develop after exposure)
• Ascertainment of outcomes could be influenced by exposure

Question 18

List the advantages and disadvantages of randomised controlled trials.

Answer 18

Advantages:

• Rigorous evaluation of a single exposure or treatment
• Prospective design (can assess causation)
• Potentially eradicates bias and confounding

Disadvantages:

• Expensive and time consuming. This means that many trials are never done, or completed trials are too small or have follow-up periods that are too short. It also means that most trials are funded by large research bodies or drug companies (who may dictate the research agenda)
• There is a risk of hidden bias (e.g. did blinding really work?)
• Limited generalisiability – often exclude high risk groups such as the elderly, children or pregnant women

Question 19

What are the design and analysis issues associated with case reports and series?

Answer 19

It is just a descriptive study and documents an individual’s medical experience. There is no comparison group so their use is limited and statistical outcomes between exposure and outcome cannot be assessed. Not really any analysis associated with case reports and series. Use limited - good place to start.

Question 20

What are the design and analysis issues associated with ecological studies?

Answer 20

Confounding variables are a problem for ecological studies. To do something about confounding at the analysis stage you have to have collected the data and if you are using previously published data this may be an issue as that info may not be in there. The fact they we are dealing with a population level rather than an individual level can also cause problems.

We display ecological data as a scatter plot. Each dot on the scatter plot will represent a population or area. For example one scatter plot for an ecological data may have cigarettes per capita plotted on x, CHD deaths plotted on the Y, and then the dot representing the pop or area.

We need to think about confounding factors for apparent associations between two variables found in ecological studies.

We need to conduct analysis and interpretation of ecological studies at a population level and not an individual level. We have no data at the individual level. Without the information at the individual level often the correct interpretation of any relationship found can’t be established. If we want to extrapolate findings to the individual we would need to conduct further studies.

Question 21

How do we display ecological data?

Answer 21

We display ecological data as a scatter plot.

Question 22

Why do we use case-control studies for rare outcomes?

Answer 22

Because if we used any of the other study designs we would have to use large populations to pick up enough cases to analyse to look for associations.

Question 23

What is the usual analysis method for categorical variables?

Answer 23

chi-squared test and generation of odds-ratios

Question 24

Can you use the risk ratio in case control studies?

Answer 24

No

Question 25

What is the purpose of a multivariate analysis?

Answer 25

A multivariate analysis takes into account more than one variable.

Question 26

What could we deduce if our unadjusted and adjusted odds-ratios showed a big change?

Answer 26

We could deduce that the factors that were being taken into account for calculation of the adjusted odds ratio are having a confounding effect.

Question 27

How can you alter the study design of a case-control study to help control for confounding?

Answer 27

We can perform matching in case-control studies where one or more controls is paired to each case based on variables thought to be confounders. This helps control for confounding, but need to allow for matched design in the analysis. It does this by increasing precision of odds ratios. Shouldn't match on exposure of interest. (always need variation in our exposure to detect any association).

Question 28

What types of studies are cheaper - cohort studies or case-control studies?

Answer 28

Case-control studies are cheaper.

Question 29

Why might one say that cohort studies are more valuable than case-control studies?

Answer 29

Case-control studies can only look at one outcome whereas cohort studies can look at multiple exposures and outcomes. Case-control studies use retrospective measures of exposure whereas cohort studies use prospective measures. For case-control studies exposure is measured at one time whereas for cohort studies we can measure changes in exposure. Case-control studies are more susceptible to selection bias than cohort studies. For cohort studies we can measure incidence of disease which is not possible using case-control studies.

Question 30

Why might case-control studies be said to be easier to complete than cohort studies?

Answer 30

Case-control studies use fewer people, are usually faster to complete and cheaper than cohort studies.

Question 31

What is the equation for incidence risk?

Answer 31

Incidence risk = Number of new cases of disease during given time period / number of people initially at risk However, we need a closed cohort to be able to carry out incidence risk calculations.

Question 32

What is the equation for incidence rate?

Answer 32

Incidence rate = Number of new cases of disease / Total person-time at risk

Question 33

What are the 3 different interpretations of an incidence rate result of '7 per 100 person-years'?

Answer 33

- 7 new cases of disease in 100 people followed for 1 year - 7 new cases of disease in 10 people followed for 10 years - 7 new cases of disease in 1 person followed for 100 years

Question 34

How is prevalence related to incidence?

Answer 34

Prevalence = Incidence x Average duration of study P = I x D

Question 35

What measure of effect can we use to examine differences in incidence between two groups?

Answer 35

We can use the rate ratio. Rate ratio = Rate of disease amongst exposed / Rate of disease amongst unexposed

Question 36

What do we use to adjust for the effects of confounders in cohort studies?

Answer 36

Poisson regression - Poisson regression can give you adjusted rate ratios to take into account confounding factors. We need to always compare the unadjusted with the adjusted statistics to see if there is any difference - no difference means that confounding is probably not having any effect.

Question 37

How can you change the design of a cohort study to reduce the time it takes to conduct a study?

Answer 37

We could do a retrospective cohort study. Here we need to use a cohort where there is existing data on exposure and disease. This overcomes the problems of loss to follow up, long latency periods and is cheaper and quicker to perform than prospective cohorts. However, we need good exposure data with low misclassification for it to be successful.

Question 38

Where might bias arise in cohort studies?

Answer 38

- loss to follow up - selection bias (not so much - less than case-control studies) - healthy worker effect

Question 39

What are the advantages of cohort studies and why might we want to use them?

Answer 39

- multiple disease outcomes can be examined - assesses temporal relationship between exposure and disease - good assessment of level / duration of exposure to look at dose / response - selection / recall bias could potentially be very low

Question 40

What are the disadvantages of cohort studies?

Answer 40

- time consuming - extremely expensive - retrospective studies rely on availability of good and accurate records - results can be affected due to loss of follow up - inefficient if effect of exposure has long latency period (would need a long follow up which would increase the cost)

Question 41

How is chance minimised in a clinical trial?

Answer 41

Chance is minimised by estimating study power, i.e. studying enough people

Question 42

How is confounding minimised in clinical trials?

Answer 42

Confounding is removed by randomisation

Question 43

How is bias removed in clinical trials

Answer 43

Bias is removed by blinding

Question 44

Why might you want to use other study designs when clinical trials are accepted to be the gold standard?

Answer 44

Randomised controlled trials can only test beneficial intervention. Clinical trials are expensive and can take a long time. Clinical trials may not be generalisable as we use restricted samples in clinical trials.

Question 45

What are the objectives of a randomised control trial?

Answer 45

1. To test for a causal hypothesis of a disease | 2. To measure the effect of a particular intervention

Question 46

We can use randomised control trials to both test for a causal hypothesis and to measure the effect of a particular intervention. What types of randomised control trials could we use to measure the effect of a particular intervention?

Answer 46

Within this group of studies where you are measuring an effect there are two main types. 1. Explanatory trials - the objective of these trials is to measure the efficacy of an intervention of a new drug before it is licensed to try and find the optimal conditions for that drug. 2. Pragmatic trials - the objective is to measure the effectiveness of an intervention when actually implemented in practice.

Question 47

How can we increase generalisability in randomised control trials?

Answer 47

We can increase generalisability in randomised control trials by minimising exclusion criteria.

Question 48

How can we guard against Type I error in RCTs?

Answer 48

Type I error is when we reject the null hypothesis when it is actually true (i.e. we say results are statistically significant when they actually are not). We can protect against this by making our p-values smaller.

Question 49

How can we protect against Type II error in RCTs?

Answer 49

Type II error is when we we don't reject the null hypothesis when it is false (i.e. we declare there is no statistical significance when there actually is). To protect against this we need to make sure our sample size is large enough by carrying out sample size calculations.

Question 50

How are subjects allocated to the intervention in randomised controlled trials?

Answer 50

Using the process of randomisation. Randomisation breaks confounding by randomly allocating the intervention.

Question 51

What are the possible methods of randomisation?

Answer 51

1. Tossing a coin 2. Random numbers 3. Permuted numbers 4. Minimisation We can use systematic randomisation also.

Question 52

Even with randomisation we can still get differences between groups - how can we take this into account?

Answer 52

We need to take baseline measurements after subjects are recruited.

Question 53

Describe allocation concealment.

Answer 53

Treatment should be allocated after the subject is recruited to the study and after you have taken the baseline measurements. If the subjects are allocated to treatment groups in advance and these are known to the subject, investigator, or both, this may affect whether the individual is introduced into the study or not and you don't want that to introduce bias. What we should have is a blind randomisation sequence so that the investigator and the subjects don't know which group they are going in to. This may be done using sequentially sealed opaque enveloped and it may be controlled by a central allocation office. If we don't blind the investigator and study subjects we may get bias.

Question 54

Who can be blinded when considering allocation into an RCT?

Answer 54

1. Study subject 2. People recruiting subjects and delivering the intervention 3. People conducting the evaluation (outcome assessor) 4. Those doing the analyses

Question 55

Summarise randomisation for RCTs

Answer 55

Randomisation is important because it eliminates the role of confounding factors (both known and unknown). There are many different methods of randomisation. We need to make sure the method of randomisation is truly random and won't introduce bias into the study.

Question 56

Describe the most basic randomised control trial design type.

Answer 56

This is the parallel group design. Subjects are randomised wither to receive placebo or active drug.

Question 57

Describe the RCT crossover design.

Answer 57

The crossover design is an extension of the parallel group design. 1. We first recruit or subjects and they are allocated to either the drug treatment or the placebo group. 2. After a period of time the subjects swap treatment groups so essentially all of your subjects get both treatments. This is good for stable diseases like asthma or high blood pressure. An important point to think about is the period of time between the two treatments. You want a 'washout' period of time where the patients return back to baseline before they start the new treatment. The main benefit of this crossover design is that you only need half as many people because each person is participating in both groups just at a different time.

Question 58

Describe an RCT within patient design.

Answer 58

This is similar to a cross over trial but subjects get two treatments at the same time. This is only suitable for very specific conditions. For example if we were looking at a drug to treat skin problems on the scalp the randomisation would come in to determine which side of the scalp was being treated by which treatment for each person. The benefit of within patient design is that you need a smaller sample size and you also rule out some of the between individual differences because one person is receiving both treatments. The main advantage over the crossover trial design is that there isn't any washout period because you are doing both treatments at the same time.

Question 59

Describe the equivalence trial design of RCT.

Answer 59

Equivalence trials aim to show whether a new treatment is as effective as a standard treatment. These trials are normally based on the standard parallel group designs to assess the difference between the two groups. Before you carry out an equivalence trial there needs to be some investigation in order to assess whether the new and old treatment are better than placebo (they might well be as good as each other but that doesn't really mean much if they are no better than the placebo). It is important to note that a non-significant p-value doesn't necessarily mean that they are equivalent - it may mean that sample size wasn't big enough resulting in lack of power.

Question 60

Describe the cluster trial design of RCT.

Answer 60

Cluster trials are similar to parallel group studies but here we look at groups rather than individuals which are randomised to an intervention.

Question 61

List the different types of RCT designs.

Answer 61

1. Parallel 2. Crossover 3. Within patient 4. Cluster

Question 62

What do you need to consider when you are analysing clinical trials?

Answer 62

We want to be very clear when reading these reports as to the author's primary hypothesis and whether this matches the primary outcome measure. Have they actually just focussed on a secondary outcome that wasn't really mentioned before they started the trial? What was the evidence preceding this trial? does it match the primary hypothesis and the primary outcome? There should be less emphasis on secondary outcomes than primary outcomes. We also want to look at which people stuck with the trials and what they did about loss to follow up. Why did people drop out of the trials? was it related to the treatment they were receiving or to their final outcomes? Intention to treat principle to get around loss to follow-up in RCTs.

Question 63

What is the intention to treat principle?

Answer 63

For pragmatic trials we want to analyse for the treatment group for which people were randomised to - although we get the initial randomisation at the beginning people may have stopped treatment or swapped group during trial (we need to think about people's reasons for this). With the intention to treat principle even though people may have swapped groups by the end of the trial, those people are analysed as if they have stayed in the initial treatment group. The reason we do this is because people who do this may have important differences which are related to the outcome from those that were compliant and took the treatment for which they were first allocated. Excluding those people that have moved or analysing people according to their final group essentially takes out the benefits given to you by that initial randomisation. You would be introducing bias into your results by basing it on which group they actually ended up in. If there are a large proportion of people swapping from one treatment to another that should be reported as part of the outcomes from that trial. Pragmatic trials should always be conducted using the intention to treat principle. There are however some instances where you might want to analyse the data based on the treatment actually received and that would be in explanatory or per-protocol analysis trials. This is where if some people do swap you would analyse them based on the group they ended up in rather than the group that they were initially allocated to. Explanatory trials are used to test the safety of a new drug before it is licensed and optimise the dose.

Question 64

Would you want to use the intention to treat principle in explanatory trials?

Answer 64

There are however some instances where you might want to analyse the data based on the treatment actually received and that would be in explanatory or per-protocol analysis trials. This is where if some people do swap you would analyse them based on the group they ended up in rather than the group that they were initially allocated to. Explanatory trials are used to test the safety of a new drug before it is licensed and optimise the dose.

Question 65

What kinds of RCTs would you not want to analyse using intention to treat?

Answer 65

Explanatory trials

Question 66

What are the advantages of RCTs?

Answer 66

- Rigorous evaluation of a single exposure or treatment - experimental rather than observational - It is a prospective design so it can assess causation as we have temporal measure of exposure and final outcome - potentially eradicated bias and confounding using randomisation and blinding

Question 67

What are the disadvantages of RCTs?

Answer 67

- RCTs are expensive and time consuming - can lead to sample sizes being too small or too short - can also mean they may be funded by a large body who has dictated the research agenda - Risk of hidden bias - was it actually properly blinded? who was blinded? did the blinding truly work? - Limited generalisability - often exclude high risk groups, elderly, pregnant women, children