Hospital leaders often tell clinicians to strictly embrace evidence-based medicine (EBM) because patients will do better and the hospital will avoid pay-for-performance losses. In chart reviews, EBM looks easy: diagnosis A means the clinician uses medicines B and C.
I love and fully embrace evidence. But this sort of nudging tramples on the founding principle of EBM. One sentence from the late David Sackett, MD, a pioneer in EBM, reveals the challenge: Evidence-based medicine is the conscientious explicit and judicious use of current best evidence in making decisions about the care of individual patients.
A novel new study highlights the important modifiers Sackett chose for that sentence. Notice Sackett's choice to modify the noun, patients. It's not just patients; it is individual patients. EBM means applying evidence in light of an individual’s many health issues (age, frailty, etc.), personal values, and beliefs.
As patients live longer and with more comorbid conditions, the job of an evidenced-based practitioner has become harder. Anyone with internet access can look up the guidelines or the results of a randomized controlled trial, but the challenge comes when deciding whether or not the patient in front of you is similar to those enrolled in the clinical trials underpinning the evidence.
How, for example, do you determine the best blood pressure goal for a patient with multiple risk factors, hypertension, and diabetes.
Strict application of the evidence takes us to the ACCORD-BP trial, which showed that a systolic blood pressure target of 120 mm Hg, compared with a target below 140 mm Hg, did not reduce the rate of the composite outcome of fatal and nonfatal major cardiovascular events. Plus, patients who received intensive treatment had more adverse effects. Evidence, therefore, points to the higher target.
But a thinking, judicious clinician has doubts. She is seeing a robust patient who has signs of hypertensive disease, and she knows this man; he likes aggressive approaches to problems.
She also wonders about the SPRINT trial, which showed that treatment to a lower BP goal resulted in serious reductions in cardiac events. She knows SPRINT excluded patients with diabetes, but asks: Is this patient closer to those enrolled in SPRINT?
Here's where a nifty new approach to evidence translation comes in. Luke Laffin, MD, from the Cleveland Clinic, and Stephanie Besser, MSA, MSPA, and Francis Alenghat, MD, PhD, from the University of Chicago, know that there are often big differences between a trial's inclusion/exclusion criteria and the actual characteristics of enrolled patients. The patient in the clinic might technically meet a trial's inclusion criteria but be poorly represented by the actual baseline characteristics of the patients enrolled in the study. Or, the patient might have a single exclusion criterion (e.g., diabetes) but resemble the trial population in many other ways.
Laffin and his colleagues took advantage of the fact that the NHLBI-funded SPRINT and ACCORD-BP trials have open databases.
The team calculated a Trial Score by mapping how close each person in the SPRINT trial was (± standard deviation) to the mean for each of six continuous variables: age, systolic BP, fasting glucose, non-HDL, creatinine, and BMI. They then performed multivariate regression with these variables, using major cardiac adverse events as the dependent variable.
This created a bell-curve distribution of Trial Scores that defined three areas: a data-rich zone of patients with scores below the 90th percentile (SD, <1.6), a data-limited zone of those with scores between the 90th and 97.5th percentiles (SD, 1.6 - 2.13), and a data-free zone of those with scores above the 97.5th percentile.
The big discovery came when they compared Trial Scores with outcomes for patients in ACCORD-BP. Although only 28% of ACCORD-BP patients had scores that fell into the data-rich zone of SPRINT, those who did had similar time-to-event rates for intensive and standard BP control. Conversely, event rates of ACCORD-BP patients with Trial Scores outside of SPRINT's data-rich zone were significantly different.
In other words, ACCORD patients who looked like average SPRINT patients benefited from intense BP control, whereas those far from the average SPRINT patient did not.
The technique has major implications for the thinking clinician.
Take the example of an 80-year-old woman with a creatinine level of 2 mg/dL who meets the inclusion criteria for the SPRINT trial, Alenghat suggested by email.
Of the more than 9300 SPRINT participants, only 35 were older than 80 years and had a creatinine level of 2 mg/dL or higher, and only 10 of those were women. This patient, therefore, falls into the data-free zone of SPRINT.
The Trial Score offers careful and judicious clinicians a way to look beyond the inclusion and exclusion criteria of a trial.
There are limitations to this analysis. The Trial Score does not account for categorical variables, such as smoking status or sex. And, importantly, clinicians will still have to decide if a trial's exclusion criterion is related to significant safety concerns. Another issue is that Laffin's team used individual patient-level data to generate the Trial Score, so the technique cannot be used unless trialists share their data.
Alenghat said he plans further analyses on other trials. I look forward to that. Imagine if we had open data on the MitraClip trials.[5,6] We could test the theory that COAPT-like patients in Mitra-FR benefited from the device.
In addition, a Trial Score might help ICD decisions. In SCD-HeFT, for instance, the entry EF was 35% or less, but the average EF of enrolled patients was 25%. This is important because subgroup analyses suggest that patients with EFs closer to entry criteria get less benefit from ICDs.[7,8] Could a Trial Score help stratify the likelihood of ICD benefit, or perhaps allow for the comparison of ICD trials?
I would not get hung up on the mathematics involved in this technique. I foresee an app that will calculate a Trial Score for an individual patient at the bedside.
As algorithms gain influence and threaten the humanity of medical practice, this group's discovery could not be more welcome or timely. This study and technique reinforce Sackett's calls for clinicians to stop and think at the bedside.
Cite this: John M Mandrola. Treating the Individual, Not the 'Average' Patient - Medscape - Dec 06, 2018.