Name:
Prognosis: Adrienne Randolph, MD, MSc, discusses prognosis.
Description:
Prognosis: Adrienne Randolph, MD, MSc, discusses prognosis.
Thumbnail URL:
https://cadmoremediastorage.blob.core.windows.net/b734433a-4277-4c6f-a740-f111204e54c1/thumbnails/b734433a-4277-4c6f-a740-f111204e54c1.jpg?sv=2019-02-02&sr=c&sig=2%2F1GjvZvxVAH%2BvMU61XzTt3nETn84x64m9gjxB8v47I%3D&st=2024-11-05T06%3A27%3A50Z&se=2024-11-05T10%3A32%3A50Z&sp=r
Duration:
T00H12M40S
Embed URL:
https://stream.cadmore.media/player/b734433a-4277-4c6f-a740-f111204e54c1
Content URL:
https://cadmoreoriginalmedia.blob.core.windows.net/b734433a-4277-4c6f-a740-f111204e54c1/6830494.mp3?sv=2019-02-02&sr=c&sig=LCFgpzOLzdyrE9zpYGlfdNFNwA73XBfPgVfwDaFD9bE%3D&st=2024-11-05T06%3A27%3A50Z&se=2024-11-05T08%3A32%3A50Z&sp=r
Upload Date:
2022-02-28T00:00:00.0000000
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. Adrienne Randolph about prognosis, a topic that is examined in Chapter 18 of User's Guides to the Medical Literature. Dr. Randolph, why don't you introduce yourself to our listeners? >> I'm Dr. Adrienne Randolph. I am a pediatric critical care physician at Boston Children's Hospital, and I also am an Associate Professor of Anesthesia and Pediatrics at Harvard Medical School.
>> Dr. Randolph, why is knowing a patient's prognosis helpful to clinicians? >> Well, clinicians help patients not only by diagnosing what is wrong with them, but by administering treatment that does more good than harm, and also by indicating what the future is likely to hold. The ideas of prognosis examine the possible outcomes of a disease and the probability with which they can be expected to occur. These studies help clinicians achieve those second and third goals. Knowledge of a patient's prognosis can help clinicians to balance the risk and benefit of treatment.
For example, if a patient's going to get well anyways, clinicians should not recommend expensive or potentially toxic treatments. In fact, if the patient has really no risk of having a negative outcome, maybe they don't need any treatments at all if they're going to get well all by themselves. And if a patient, however, is destined to have a poor outcome no matter what you do, no matter what a physician does, aggressive therapy may only prolong that patient's suffering. So studies of prognosis help clinicians present the future course of a patient's illness to that patient so that they can offer either reassurance and hope, or prepare the patient for even potential death or long-term disability, >> Are clinical instincts sufficient to make an accurate assessment of prognosis?
>> Well, clinical instinct gained through years of experience can sometimes lead to an accurate estimate of prognosis, but built-in clinical biases often lead to excessive optimism or pessimism, and patients really deserve the best advice based on unbiased evidence. Many clinicians are not aware of factors that can result in biasing their estimates. For example, the last patient that they treated and that patient's outcome, or the last few patients they treated and those patients' outcomes, may bias them either positively or negatively, depending on what those outcomes were.
And that is not really an accurate estimate when, really, they should be focused on the last 50 or 100 patients and their outcomes. So it's very important to do studies where these estimates that could bias or systematically overestimate or underestimate the likelihood of positive or adverse outcomes, that those are controlled. >> Which study designs are useful for investigating issues of prognosis?
>> Well, observational studies, which are cohort or case-control studies. Cohort studies are actually where a population of patients is put together and then closely followed up over time, or case-control studies where patients with the conditions are identified, and then control patients who don't have the conditions that meet certain criteria are identified, and those patients are followed up and compared. Those are common study designs for prognosis studies, but even randomized trials can often be used to assess prognosis in patient populations.
The control group would tell the usual outcome without the therapy, and then the treatment group would tell what is the outcome with the therapy? And if the therapy is not efficacious, then the study can actually be used as a cohort study and the two groups can be combined. So all those types of study designs can be used to investigate issues of prognosis. >> When assessing a study's validity, how can clinicians determine if the sample of patients was representative?
>> Well, first, it's important to determine whether the patients pass through some sort of filter before entering the study. For example, what is the sequence of referrals leading the patients from a primary to a tertiary care center where the study may take place? So patients in that tertiary care center might differ from the population in the community hospitals, and population-based studies of risk often differ from the risk assessed in patients evaluated in clinic.
For example, if you are trying to assess what's the rate or risk of febrile seizures recurring, if you look at population-based studies across the entire population of patients, you may get a lower risk estimate than if you were just assessing patients who are evaluated in a neurology clinic, so it's really important to look at where those patients came from who are in the study. >> Dr. Randolph, what is an adjusted analysis, and how is it helpful in evaluating factors that may influence patient outcome?
>> Well, an adjusted analysis is sometimes complicated for clinicians to understand, but I'll give a couple of examples. It's important that clinicians look at all prognostic factors together and look at them in relationship to each other. For example, if sickness but not age truly determines the patient's outcome, and sicker patients tend to be older, investigators who don't simultaneously consider age and sickness together might mistakenly conclude that age is the only important prognostic factor.
So here's another example that was from the Framingham Heart Study where they examined risk factors for stroke. They reported that the rate of stroke in patients with atrial fibrillation and rheumatic heart disease was 41 per 1,000 person years, similar to the rate for patients with atrial fibrillation but without rheumatic heart disease. Patients with rheumatic heart disease were, however, much younger than those who did not have rheumatic heart disease. So in order to properly understand the influence of rheumatic heart disease, you had to separately look at the relative risk of stroke in young people with and without rheumatic heart disease, and the risk of stroke in elderly people with and without rheumatic heart disease.
These separate investigations looking at all of these risk factors, both together and separately, is called an adjusted analysis. Once these investigators adjusted for age, they found the rate of stroke was six-fold greater in the patients with rheumatic heart disease and atrial fibrillation than in those patients with atrial fibrillation who didn't have rheumatic heart disease. So without adjusting simultaneously for the age and rheumatic heart disease, one would have been misled on what were the risk factors.
>> What role does follow-up play in prognostic studies? >> Well, follow-up is very, very important because investigators who lose track of a large number of patients compromise the validity of the study. The reason is because those who aren't followed up could be systematically at higher or lower risk of the outcome than those not followed. In fact, when assessing an outcome such as mortality, if patients die, they may be harder to find in follow-ups.
So there may be more patients who died earlier who are now lost to follow-up. So that would bias the outcome of the study if one assumes that all of those patients that weren't followed up survived. So trying to follow up all patients completely is one of the most important factors of the validity of a prognostic study. >> Dr. Randolph, what other considerations should clinicians keep in mind when assessing the risk of bias within a prognostic study?
>> Well, besides making sure that follow-up is sufficiently complete, one of the other major factors that are really important is looking at those outcome criteria themselves and making sure that they're objective and unbiased. So outcome events that are objective and easily measured, such as death or height, some of these are outcome criteria that a person's subjective judgment or diagnostic acumen would not influence and that there wouldn't be variability amongst practitioners, or low variability amongst practitioners.
But some of these outcome criteria that require a lot of judgment, or may vary depending on how people are diagnosing things, could lead to variability amongst studies and inaccuracy as well. So it's really important that investigators clearly specify and define those target outcome measures. And whenever possible, they should try to not use subjective measures but use objective measures.
And if they do use subjective measures, they need to assess what is the agreement amongst clinicians. If three people evaluated that same patient, two or three people, would they all come to the same conclusion on that outcome? >> Once clinicians have decided that a study of prognosis is at low risk of bias, are there any challenges they may face in interpreting the results, and do you have any tips for handling these challenges? >> Well, two things that they need to understand when interpreting the results are how likely the outcomes are over time and how precise are those estimates of the outcome over time.
One common way that outcome studies are presented are looking at survival curves. So that they look at, for example, how often a patient might have a second heart attack after the first heart attack and at what time period would that outcome occur. So looking at those survival curves, one of the things that happens is that the earlier follow-up periods will include results from more patients than the later follow-up periods, and the more results there are, the more precise the confidence limits are around the estimate of that outcome over time.
So those are things that the person needs to understand in interpreting. It's important to understand how to look at survival and how to look at how things change over time, especially as those patients who've had an event now are no longer followed up. So once they have the event, they're no longer in the curve, and the curve has fewer and fewer patients as it goes over time over being followed many years. >> Is there anything else you would like to tell our listeners about prognosis?
>> Well, one important thing to understand about prognostic studies is that applying their results to patients in your patient care is not as straightforward as one would think, because one factor that could strongly influence patient outcome is therapy. Therapeutic strategies vary markedly among institutions and change over time as new treatments are available or old treatments regain popularity. To the extent that those treatments are beneficial or detrimental, the overall patient outcome might improve or become worse, so it's important to look at the context of the evidence that you're using to determine if it's still relevant in the time period that you're practicing and with the therapies you're using, and also if the study patients and their management in that study that you're trying to apply to your patient population is similar to those patients in your practice that you're applying the prognostic estimates to.
>> Thank you, Dr. Randolph, for your discussion of prognosis. For additional information about this topic, JAMAevidence subscribers can consult the online chapter on prognosis. That's Chapter 18 in User's Guide to the Medical Literature. This has been Joan Stephenson of JAMA talking with Dr. Adrienne Randolph for JAMAevidence.