Name:
Harm (Observational Studies): Mitchell Levine, MD, MSc, discusses harm (observational studies).
Description:
Harm (Observational Studies): Mitchell Levine, MD, MSc, discusses harm (observational studies).
Thumbnail URL:
https://cadmoremediastorage.blob.core.windows.net/387d2387-68e1-45b0-878a-791c3480a210/thumbnails/387d2387-68e1-45b0-878a-791c3480a210.jpg?sv=2019-02-02&sr=c&sig=uv%2FoDiJvtNgZLTeovXtgmQJD%2BF1N6ca9cj4ItTGlUKg%3D&st=2023-09-21T08%3A05%3A00Z&se=2023-09-21T12%3A10%3A00Z&sp=r
Duration:
T00H07M53S
Embed URL:
https://stream.cadmore.media/player/387d2387-68e1-45b0-878a-791c3480a210
Content URL:
https://asa1cadmoremedia.blob.core.windows.net/asset-934fd8ef-cd50-4413-8d3b-e2389ce501d3/6830486.mp3
Upload Date:
2022-02-23T00:00:00.0000000
Transcript:
Language: EN.
Segment:0 .
>> I'm Joan Stephenson, Editor of JAMA's Medical News and Perspectives section. Today, I have the pleasure of speaking with Dr. Mitchell Levine about evaluating observational studies that examine potentially harmful exposures to either medical interventions or environmental agents, a topic that's examined in Chapter 12 of Users' Guides to the Medical Literature. Dr. Levine, why don't you introduce yourself to our listeners? >> Hi. I'm Mitchell Levine. I'm a professor in clinical epidemiology and biostatistics and in the Department of Medicine at McMaster University.
>> Dr. Levine, could you describe the approach clinicians should take when considering how to use information from an article in the medical literature about harm to guide their practice? >> When a clinician has an article that evaluates an association between an exposure and a harmful outcome, critical in that evaluation is to look at the type of methodology that was used to make that association. Ideally, they would have a randomized clinical trial to make that association. But alternatively, if that's not available, then there could be observational studies, either cohort or case control design.
When they're thinking about that association, though, they should try and find the best article in the literature for doing it, and they should be looking at randomized control trials as being the best source. So, even though they're looking at an association of an article in their hand, what they should be asking themselves, is this the best possible methodology available to evaluate this association? And if not, maybe they should go back and try and check to make sure that they do have the best evidence before they go to the next step.
>> What are the strengths and weaknesses in different study designs, including randomized control trials, cohort studies, and case control studies? >> The randomized control trial gives the strongest evidence in support of an association, because through randomization, variables that might influence the outcome are hopefully balanced between the two groups, the exposed and the unexposed group. This is ideal because then if there is a difference in the outcome between the two groups, one will assume that the difference is related to the exposure and to not some unknown important variable that has influenced outcomes.
That's the major strength of a randomized control trial. The disadvantage of randomized control trials is that they can't be conducted to evaluate harmful associations when in advance of the study one assumes that there's going to be a harmful association without much benefit. And you can't do a study with an RCT to evaluate that question. We do find harmful outcomes sometimes reported in randomized control trials. But it wasn't really intended as a trial to identify the harmful outcomes.
So, they're not as feasible, and they're not as likely to be used to directly answer a question about a harmful outcome. It would obviously be unethical if one really suspected that the harmful outcome would occur, and then to conduct a randomized control trial. On the other hand, we have cohort and case control designs, which are observational studies that are very useful for evaluating harmful associations. These two studies, though, have limitations, because they are more prone to being influenced by bias or confounding in their design and can actually lead to wrong or misleading conclusions.
Their feasibility is what is very attractive for evaluating that association. They can be done without concern for ethical issues of exposing people to harmful outcomes. And they could also be done in a manner that's much more feasible in terms of both time and numbers of patients that need to be assessed. The downside, as I say, is very strong, relating to problems with confounding and bias, which ultimately could lead to an erroneous conclusion. >> What factors should clinicians consider when evaluating the results of a study?
>> In the randomized control trial, the same principles that apply to evaluating a therapeutic intervention would apply to assessing the harmful outcomes. And I believe this has been discussed elsewhere. So, I'm going to spend more time talking about the considerations that physicians need when there are studies using observational designs, either the cohort or the case control design. The most important consideration when one is using a cohort design for evaluating association is determining whether the two groups, the exposed and the unexposed groups, have really equal opportunity for the outcome.
And that doesn't mean at the end that they will have the same amounts of outcome, but that they had to have the equal opportunity for that outcome; either having the same predisposition towards that outcome, aside from the exposure, and having the same abilities for the outcome to be detected in the two groups. If the outcome was not an equal opportunity for both the exposed and the unexposed group, then one will end up with misleading results, or certainly has the potential to have misleading results. On the other hand, for the case control study, the most important question the reader has to ask of the article is in the cases and the controls did they have equal opportunity to being exposed?
So, it's going in the opposite direction. But an equally important question, do both the cases and controls have equal opportunity for exposure and detection of that exposure. Should it be that one group was more likely to have been exposed than the other then, in fact, the results that one would obtain by the analysis of a case control design could be misleading. >> What question should clinicians ask themselves when applying study results to their patients? >> Probably the fundamental question to ask is whether the patient that they have in front of them could have been in that study that they are using for evaluating that association.
If the patient has the same characteristics that would have allowed them to be in the study, then that's really the first step to saying it is likely that these results are going to apply to my patient. The second question would be is was the environment and the exposure and the way they measured the outcomes similar to what would happen in the real world that my patient lives in? And if the answer is yes to that, then I think they passed step two for saying that the results of the study are really quite applicable to the patient that I'm managing.
>> Is there anything else you would like to tell our listeners about using observational studies to determine the risk of harm? >> I think the critical thing is to realize that cohort and case control designs are essential for evaluating harmful associations because of their feasibility. The problem is they do have limitations, and that one could be misled by some of the results, particularly if those earlier questions are not adequately addressed, the equal opportunity for the outcome and the equal opportunity for the exposure.
In those cases, a healthy skepticism about the results really needs to be held, and one might want to say before you come to a definitive conclusion and agree that an association exists between an exposure and an outcome, but to look at more evidence or to wait for time to see if there's any new studies that come along to support the same kind of conclusion. >> Thank you, Dr. Levine, for your discussion of using observational studies to determine the risk of harm. For additional information about this topic, JAMAevidence subscribers can consult the online chapter on Harm and Observational Studies.
And that's Chapter 12 in Users' Guide to the Medical Literature. This has been Joan Stephenson of JAMA talking with Dr. Mitchell Levine for JAMAevidence.
