Thursday, January 7, 2010

Evidence-Based Medicine in Clinically Complex Elders – Barking up the Wrong Tree?


For years, many academics (myself included) have decried the lack of rigorous clinical trial evidence to inform the care of clinically complex elders. Clinical trials of drug therapy routinely exclude patients of advanced age and those with multiple comorbidities, leading to a paucity of evidence in this vulnerable group.

However, perhaps we’ve been barking up the wrong tree. Evidence-based medicine is often ill-suited to the care of complex older adults, and more of the traditional sorts of randomized trials of drugs and devices will not give us the information we need. Consider the HYVET trial (http://content.nejm.org/cgi/content/abstract/358/18/1887). For years, observational evidence and limited trial data suggested that in the oldest old, high blood pressure was protective against mortality (the opposite of the relationship in younger people). HYVET, a randomized controlled trial which enrolled 3845 patients from 195 centers on 5 continents, found the opposite to be true. However, the study has been criticized for its lack of generalizability, since many patients were excluded from participation in the trial.

This leads to two conclusions. First, if we cannot get the data we need from this enormous undertaking, when will we ever? Second, imagine that this study did include a wider variety of patients. It would still be unclear how to apply its results. Would the results apply to a 90-year old man with multiple comorbidities and substantial functional decline, or to an 85-year old woman who is in excellent health? The great heterogeneity of the older population means that the population-averaged results that arise from a clinical trial do not necessarily apply to our individual, highly non-average patients. Subgroup analyses from studies can provide hints in this direction, but cannot truly account for the great heterogeneity of the older patient population. This is different than the traditional population of clinical trials, which with their younger, less comorbidly-ill patient samples apply more directly to the younger, less comorbidly-ill population of patients seen in clinical practice.

This doesn’t mean that we should abandon evidence-based medicine for the care of complex elders, nor does it mean that HYVET and other clinical trials offer little of value. However, the primary solution is not likely to arise from repeating traditional drug and device trials in clinically complex elders. Instead, trials that study integrative, cross-disease approaches to care in complex elders may hold greater promise, insofar as they may be more generalizable to a broad spectrum of this population and less dependent on the specific comorbidity or functional profile of individual patients. We will still need to use judgment to extrapolate study results to individual patients. However, by embracing complexity, these studies can offer a path forward that complements and may ultimately prove of greater value than the single-drug, single-disease focused interventions that are the mainstay of the traditional evidence-based literature.

4 comments:

Alex Smith said...

Thanks Mike for a terrific post. I have been thinking about a related issue recently regarding prognostic models. Louise Walter, noted today that most prognostic models are based on a single disease entity (eg cancer, heart failure). She and Sei Lee developed and published prognostic models for hospitalized and community dwelling older adults. No specific disease focus.

I'm intrgued when you say, "Instead, trials that study integrative, cross-disease approaches to care in complex elders may hold greater promise, insofar as they may be more generalizable to a broad spectrum of this population and less dependent on the specific comorbidity or functional profile of individual patients."

I see the prognostic models of older adults as similar, "embracing" the complexity of elders, as you say. Can you give some examples of trails that are "integrative, cross disease focused?"

Perhaps Seth Landefeld's Acute Care of the Elderly trial? That answers a question about a very sort of "geriatric" outcome - functional impairment - for which multi-method approaches are best suited. But are there examples of trials that address specific disease managment questions - like how do I treat the blood pressure of my frail 85 year old patient with multimorbidity, and mild cognitive and functional impairment? I guess I have trouble wrapping my brain around what such a trail would look like.

Mike Steinman said...

Thanks Alex for the thoughtful comment. I had a few ideas in mind when I talked about different kinds of trials for clinically complex elders. One such model is the one you pointed out – looking at different systems of care delivery, such as acute care for elders (ACE) units in hospitals, medical home models in outpatient clinics, etc to see if they improve outcomes. However, there may be interesting avenues to explore about trials that look at more vs. less aggressive care for the whole patient. One interesting model for this (and in part an inspiration for the post) was a study I stumbled on recently that evaluated usual care vs. very aggressive medication discontinuation for nursing home patients in Israel (http://www.ncbi.nlm.nih.gov/pubmed/17642388). In this trial, the authors found dramatically lower rates of mortality and hospitalization in patients assigned to aggressive discontinuation of medications. The results are so dramatic that they are a bit hard to accept at face value, and I don’t necessarily endorse the algorithm used by the authors. However, it’s an interesting idea and variants of this approach may help better define the risk/benefit ratio of more or less intensive approaches to prescribing. Finally, the diabetes literature has been a leader in looking at different treatment targets in young and healthy vs. old and clinically complex patients with diabetes. For example, a study published last month by Greenfield at al (http://www.ncbi.nlm.nih.gov/pubmed/20008761) found that patients with high levels of comorbidity derive less benefit from intensive glucose control than their healthier counterparts.

Chrissy Kistler said...

As an aside, my big issue with HYVET was not that it wasn't generalizable enough (though it could have been better); it was that my cardiology and gen med colleagues seem to have missed their target BP (150/90-if I recall correctly) and NOT 140/80 or as some seem to feel 120s/60s. They seem to forget the pilot HYVET study which showed increasing mortality at the lowest bps.

In addition, it makes perfect sense to me that as you increase in comorbid conditions you derive an incrementally smaller amount of benefit from ANY ONE intervention. When you already have CHF, CAD, CVA, DM, CRI, how much good does a statin really do? An ASA? And how much harm in terms of side effects and med interactions...

Great post, thanks Mike!

ken covinsky said...

Great post and discussion. It does seem that it is possible to take evidence based medicine too far. No study is ever perfectly applicable to the patient in front of you, so we just can't get out of needing to apply some art and judgement in deciding how to apply evidence to an individual patient.

One of the things I remember from Sackett's text on clinical epidemiology---rather than asking,"are the patients in this study like my patient" they suggest asking, "is my patient so different from the patients in the study that I don't think I can apply the results of the study to them?"

Still, often the problem is not that the studies don't match our patients--they sometimes don't even come close to hitting the target. Sometimes, even for predominantly geriatric conditions, the studies vastly underrepresent the elderly. For example, during my most recent attending month, our team reviewed some studies on COPD. It was remarkable, for this predominantly Geriatric condition, how few elderly subjects were in many of the clinical studies.

Whenever one applies for a NIH grant, one has to discuss the gender and minority inclusion and make a convincing case the study is enrolling a diverse patient population. I suspect this situation would improve if NIH added age to these criteria.