Home : Evidence Based Medicine : Classification of Medical Research

Classification of Medical Research

Medical research is broken down into 2 main areas (5):

Syntheses

Syntheses are considered secondary research because they compile and summarize information from studies (see Individual Studies below) that have already been completed. As mentioned earlier, syntheses can save us a lot of time because they perform all the steps of EBM in one place. Ideally the synthesis frames a focused clinical question, finds all the best information, evaluates the information for validity, and synthesizes it so that we can apply it to our care.

The 2 most commonly used syntheses by healthcare providers are:

Overviews/Meta-analyses – This type of research compiles all the information from the individual studies (prioritizing randomized controlled trials – see below) to provide us with an overview review on a given question. When the overview incorporates statistics to provide a quantitative summary of the literature (for example, it says something like the studies combined show a 20% reduction in…), it is known as a meta-analysis. When the review frames the question well and finds, evaluates and synthesizes all the relevant information in a rigorous way, the overview is also referred to as a systematic review.

Practice Guidelines – This type of information compiles all the information from individual studies as well as meta-analyses to make recommendations on key clinical questions relating to a given disease. The way this is done varies. When experts arrive at their recommendations by agreement, the product is a “consensus-based” guideline. Guidelines that tie the strength of the recommendation to the strength of the evidence are considered to be more rigorous and are known as “evidence-based” guidelines.

Individual Studies

Studies are considered primary research. Individual studies directly experiment on or observe what happens in population groups. There are five main types of studies (ordered from the most rigorous to least):

The Randomized Controlled Trial (RCT) – This is the best type of study for evaluating issues of efficacy, tolerability and safety of an intervention or therapy. In this study design the researcher randomly assigns subjects to various groups that should end up being well matched in terms of their baseline characteristics. Each group is assigned a different therapy and the researchers then evaluate which group had the best outcomes. When a dummy pill – a placebo – is given to one of the study groups, the study is known as a randomized placebo-controlled trial. This type of study tells us if the therapy is better than doing nothing. It is also very important to mask both patients and researchers from what intervention or therapy they are getting to avoid ways they might bias the result. When this is done the study is known as a double-blind randomized controlled trial, which is a higher quality design. Other issues that are important are to ensure that at least 80% of the subjects in each arm of the study completed the trial and that patients were compliant with taking their medication.
While well-performed RCT’s provide useful information on efficacy (how well the drug works) and tolerability (nuisance and relatively frequent side effects like nausea), if the trial is relatively small (for example, a few hundred patients), it cannot provide reassurance on issues of safety such as infrequent life threatening events like liver or kidney failure, where events may happen in less than one in a thousand patients. For safety, you may need to turn to one of the other study designs listed below.

Cohort Studies – In this type of study design the researcher does not intervene but simply observes what happens to different populations over time after they are exposed to different substances. Because the researcher does not intervene, cohort studies are less expensive to do than RCT’s and are often large enough – sometimes on the order of thousands or tens of thousands of subjects – to provide information on efficacy, tolerability and safety.
However, this study design has a weakness. Since the observed groups are not randomized, the subjects in each group may not have equal baseline characteristics. For example, they may have differences in severity of disease. Therefore we cannot be sure that the outcomes observed in each group are entirely due to the therapy of interest, or is rather explained by other differences between the groups that are really the cause of the outcome. These “other explanations” are known as confounders. Researchers try to account and adjust for all the confounders they know about using statistics (such as logistic regression or multivariate analysis). Yet even when they have done this well, they cannot anticipate every possible confounder and cannot completely eliminate the possibility that something else might be explaining the outcome.

One of the main reasons why well-done RCT’s trump cohort and the other study designs below, is that they deal with both known and unknown confounders (because the confounders tend to randomize equally to each of the study arms, thereby canceling out their effects).

Case-Control Studies – Unlike RCT’s and cohort studies where groups are defined by what they are exposed to and followed forward in time to see what outcomes they have, in case-control studies, groups are selected according to their outcomes and researchers look back in time to see what each group was exposed to. For example one group with a relatively uncommon cancer might be compared to a group of similar patients who don’t have that cancer and explore the toxic or drug exposures each may have had. The design can also make associations between rare outcomes and particular exposures, like infrequent life threatening safety concerns in patients taking a particular drug.
However, this study design is highly susceptible to bias. For example patients who are diagnosed with cancer may ruminate more about what caused it and be more predisposed to remembering the exposures they have had in comparison to the group without cancer. In addition, this design has the same problem with confounding as cohort studies.

Cross-sectional studies – In this design the researcher collects all the outcomes and exposures in a single slice in time in populations of interest. Cross-sectional studies are also susceptible to bias and confounding can occur. Since this design represents a slice in time, when associations are made, it is often unclear which factors are causes (exposures) and which are effects (outcomes). Therefore cross-sectional studies are typically weaker than the RCT and cohort designs.

Case-Reports/Case Series – These are not study designs per se, but rather descriptions of what happened to individual patients or groups of patients. These studies do not include comparator or control placebo groups and therefore cannot make associations. Thus it is unclear whether the effects are from a treatment or due to the natural history of the disease. Rather than providing definitive answers these studies may stimulate the need for more rigorous studies to be done that include comparator groups to ascertain whether a therapy may indeed be beneficial or harmful.

Return to where you were by clicking the back button in your browser or one of the respective sections listed below:

Find: Syntheses

Find: Studies

Evaluating Research

Is the Research Believable?

Super Smart Health

Empowering Health and Wellbeing

Classification of Medical Research

Syntheses

Individual Studies