Study Design

February 06, 2023

Information

In this second video, we discuss study designs that are commonly used in clinical research, including experimental and observational studies and systematic reviews.

ID9450

To CiteDCA Citation Guide

00:02<v Maria>My name is Maria Ciarlegio</v>
00:04and I'm a faculty member
00:05in the Department of Biostatistics
00:08at the Yale School of Public Health.
00:11In this video series,
00:12I will introduce the clinical research process
00:15to prepare you to collaborate with a statistician.
00:20In the second video,
00:21we'll discuss some study designs that are commonly used
00:25in clinical research.
00:27The study design used to answer your research questions
00:31is determined by the goal of the research
00:34and the feasibility of different designs,
00:36including the availability of resources.
00:39For example, a randomized clinical trial
00:42can require a good amount of administrative support
00:46in addition to the required clinical involvement.
00:50The first major distinction between study types
00:53is whether the study investigator decides
00:56if a participant receives the exposure
00:59or the intervention.
01:01Studies where the investigator assigns the exposure
01:05are experimental studies,
01:07and studies where the investigator
01:09does not assign the exposure are observational studies.
01:14In observational studies,
01:15the investigator observes or documents
01:18the presence of exposures and outcomes
01:21as they naturally occur in a population or a sample.
01:26There are both descriptive
01:28and analytical observational studies.
01:31A descriptive observational study
01:33is used to describe characteristics
01:36of a sample or population.
01:38There's no comparator group
01:40because the goal is simply to describe,
01:43not formally compare.
01:45For example, you can use a descriptive study
01:48to estimate the prevalence of a disease
01:50or report the results of a patient survey of symptoms.
01:55They can also be used to inform the design
01:57of a future comparative or analytical study.
02:00For example, you may not have a good estimate
02:03of baseline prevalence of disease
02:05for a future study testing a new treatment on that disease.
02:10A descriptive study can be used to estimate this value
02:14to help you design that future study.
02:16The first type of descriptive study is the case report,
02:20which usually describes a patient presenting with an unusual
02:25or complicated disease.
02:27When more than one patient is described,
02:30it becomes a case series.
02:32Case reports and case series are useful for defining cases,
02:36generating hypotheses about the causes of disease,
02:40or use in clinical education.
02:42Finally, descriptive cross-sectional studies
02:46collect information on the presence or the level of one
02:50or more characteristic at one point in time.
02:54Those characteristics can include risk factors
02:57or different exposures and outcomes,
03:00such as the presence of disease.
03:03If the goal is to report the distribution
03:05of one or more of the characteristics,
03:07then the cross-sectional study is descriptive.
03:11However, if the goal is to assess the relationship
03:13between say presence of an exposure
03:16and presence of disease,
03:18then the cross-sectional study is analytical.
03:21And we'll talk about analytical studies next.
03:24But the takeaway on cross-sectional studies
03:26is that they provide a snapshot
03:28of the frequency of disease and patient characteristics
03:32at one point in time.
03:35Analytical observational studies include a comparator
03:39or control group.
03:40The goal is to formally establish
03:43or quantify an association between exposures and outcomes.
03:47Again, the exposures are naturally determined
03:49in observational studies,
03:51unlike experimental studies
03:53where the investigator assigns exposure usually
03:56in the form of different treatments
03:58the temporal direction of the study determines
04:01the type of analytical observational study.
04:04If the study assesses exposures
04:06and outcomes at the same point in time,
04:08the study is cross-sectional.
04:10Analytical cross-sectional studies
04:12as opposed to descriptive cross-sectional studies, again,
04:16are used to assess the relationship
04:18between exposure and disease.
04:21An analytical cross-sectional study was conducted in Japan
04:26to assess the relationship
04:27between non-alcoholic fatty liver disease
04:30and periodontal disease.
04:32Magnetic resonance elastography
04:34was used to measure liver stiffness
04:36in these patients with NAFLD.
04:38They found an association between P. gingivalis positivity
04:42and number of periodontal pockets and liver stiffness.
04:46The idea here is that the investigators
04:49are looking at these patients at a snapshot in time,
04:52not over time, in this cross-sectional study.
04:56A cohort study follows patients forward in time
05:00for development or occurrence of the outcome.
05:04Investigators identify a group of patients
05:06without the outcome or disease of interest.
05:09And of these patients,
05:10some are exposed to an exposure of interest
05:13and some are unexposed.
05:16We follow the exposed and unexposed groups forward in time
05:20for development of the outcome.
05:22If we observe a higher incidence of the outcome
05:26in the exposed,
05:27then the exposure is associated
05:29with an increased risk of the outcome.
05:32A strength of the cohort study
05:34is that we know the exposure preceded the outcome.
05:37However, it's necessary to wait for the development
05:40of the outcome in prospective cohort studies.
05:43So they can be slow to conduct,
05:46especially in the case of rare outcomes.
05:50We often perform retrospective cohort studies.
05:54Here, the start of the study can occur
05:56after some patients have already developed
05:59the outcome of interest.
06:00Exposure status in the past
06:03is established using existing data, such as medical records.
06:08The goal is to establish a cohort of individuals
06:11without the outcome at a fixed point in the past
06:14and determine their exposure status at that point in time.
06:18Then determine if the patient
06:20subsequently develops the outcome.
06:23A limitation of retrospective cohort studies
06:26is that exposure status is not assessed
06:29by the study investigators
06:31because we often rely on medical records.
06:34So the desired exposure data may not be available.
06:38A case control study is an alternative design
06:42that begins by identifying a group of individuals
06:45with the outcome of interest.
06:47A similar control group without the outcome
06:50is also identified.
06:52Through chart reviews or interviews,
06:55the investigator then determines past exposure status.
06:59Unlike a retrospective cohort study
07:01that begins by identifying study participants
07:05based on the exposure and then assesses outcomes,
07:08the case control study begins by assessing the outcome
07:12and then determines past exposures.
07:15Case control studies are useful when studying rare outcomes
07:19or diseases with long latency.
07:21However, limitations include recall bias
07:24in recalling exposure,
07:26problems with selecting comparable controls,
07:29as well as some analytical issues.
07:33The other major arm of study designs
07:36are the experimental studies.
07:38The gold standard of evidence is the clinical trial,
07:41in particular, the randomized clinical trial.
07:44Here, the investigator randomly assigns participants
07:48to different exposures, for example,
07:50sorafenib or placebo in patients with advanced liver cancer.
07:55Usually each participant has an equal chance
07:58of being assigned to the two groups,
08:00although the allocation ratio
08:01does not have to be one-to-one.
08:04Here we have an example of a randomized control trial.
08:07This study randomly assigned patients
08:09with liver cancer to sorafenib or placebo.
08:14A traditional parallel group randomized clinical trial
08:17resembles a prospective cohort study,
08:20except for the important difference of randomization.
08:24Randomization is an important element
08:26of clinical trials
08:28because it protects against selection bias
08:30and should balance both known
08:32and unknown confounding factors between the exposure groups.
08:38Additional commonly used randomized trial designs
08:41include crossover designs, factorial designs,
08:45and cluster randomized clinical trials.
08:49In non-randomized control trials,
08:51participants are assigned to different interventions
08:54without following a random procedure.
08:57For example, assignment may be made according
09:00to investigator preference.
09:02This design is susceptible to bias
09:04due to the potential differences
09:06in patient characteristics between the two groups.
09:10Another type of study is the systematic review.
09:13The goal here is to review
09:15and synthesize all available evidence
09:18on a specific research question.
09:20Systematic reviews are carried out according to
09:23a pre-specified protocol that defines the question,
09:27describes the scope of the review,
09:29and the criteria and methodology that will be used.
09:33Begin by specifying the review question,
09:36the population you're studying,
09:38the interventions or exposures of interest,
09:41outcomes of interest,
09:42and study designs that should be included in the review.
09:46Then you'll list the eligibility criteria
09:49for the studies to be included.
09:51For example, you may only want to include evidence
09:54from randomized control trials
09:56or large scale observational studies.
09:59You'll then search PubMed and other databases
10:03for articles or studies that meet these criteria.
10:07An important step is assessing the quality of the studies
10:11and critically assessing how well the studies were done.
10:14Any potential issues with the design may introduce bias,
10:18so it's important to critique the strengths and weaknesses
10:21of the evidence you've accumulated.
10:23At this stage,
10:24you can also comment on gaps in the evidence,
10:27such as patient populations not represented in the research.
10:31Then you extract the data or the study results
10:34that are summarized in the manuscripts
10:36and the supplemental materials.
10:37And a final step involves synthesizing the results,
10:40or combining and analyzing the results
10:43from multiple studies.
10:45This can be a qualitative synthesis
10:48in which you summarize
10:49how the research you found fit together,
10:52describe the strengths and the weaknesses
10:54of the body of evidence,
10:56identify gaps in areas of future research,
10:59or it can be a quantitative synthesis.
11:03Quantitative synthesis is usually in the form
11:06of a meta-analysis.
11:09A meta-analysis is an analytical way of formally combining
11:14or pooling results from different sources.
11:17In this study, the goal was to estimate the effect
11:20of antiviral therapy on liver stiffness
11:23in patients with hepatitis B at six months,
11:25one year, two years, three years,
11:28and five years after beginning treatment.
11:31This flow chart shows the process of study collection.
11:34In the end, 24 studies were included in the analysis.
11:39Typically, a forest plot,
11:41which is the graph shown here on the right,
11:43displays the point estimates
11:45and 95% confidence intervals from each study,
11:48along with the pooled estimate.
11:51At six months from the start of therapy,
11:53data pooled from eight studies
11:55with 968 patients showed a significant decline
11:59in liver stiffness by 2.21 kilopascals
12:03as compared to pre-treatment LSM.
12:07In this video, we discussed some commonly used designs
12:11in clinical research, including observational studies,
12:14experimental studies, and systematic reviews.
12:17The next video, which is the third video in this series,
12:21will discuss the data collection process and variable types.
12:25Understanding variable types will prepare us
12:28for the fourth video in this series
12:30on sample size determination.