Evidence-based medicine is a term that attendings like to pepper into their teaching rounds, and medical students like to conspicuously incorporate into their assessment and plan. There is so much clinical trial data out there that it is impossible to stay up to date on the most recent guidelines or the newest landmark trial that all of our time goes into scrutinizing the results. However, we rarely delve into the same stringent analysis of the metrics used to quantify the results as we do about the result itself. Such scrutiny is necessary to answer the most basic fundamental question (that we put so much stock into): How much evidence is in evidence-based medicine?
It is a well-established statement that angiotensin receptor-neprilysin inhibitors (ARNI) provide benefits in patients with heart failure (HF) with reduced ejection fraction. This data comes from the PARADIGM-HF trial and is the main positive headline conclusion of that study. However, taking an objective look at the evidence presented in that trial, one will see that although ARNI did show benefit against medium-dose enalapril, it failed to show any differences in outcomes when compared to stronger ACE inhibitors (ACEi) such as ramipril or even valsartan alone. We also have fairly new data from the 2021 LIFE trial that concluded ARNI is not superior to valsartan alone, yet collectively we still hold ARNI to an anointed status even when the evidence might tell a different story.
A sub-group analysis of LIFE by Vader et al. investigated how patients were able to tolerate ARNI and found that the main drivers of intolerance were largely dictated by low BP, symptoms of low BP (dizziness, fatigue), and renal disease. Furthermore, they were able to predict ARNI intolerance by stratifying patients who had low BP, valvular disease, insulin use, electrolyte disturbances, or who were not using ACEi/ARB. The rate of predicted intolerance increased to over 50 percent if they had over four of these predictive variables.
I postulate that this dichotomy is due to the fact that HF trials generally recruit patients who have adequate socioeconomic status and who are healthy enough to walk into their clinics. This selective recruitment is very different from the real-world high-risk patients with multiple comorbidities who were included in LIFE. It demonstrated, along with the sub-group analysis by Vader et al., that the highly regarded ARNI did not benefit these patients, many of whom were unable to tolerate it in the first place. Anecdotally, ARNI is usually started on an inpatient basis for HF exacerbation in fragile patients; yet the evidence of its benefit comes mainly from outpatient, mostly Caucasian male clinic patients with high socioeconomic status. LIFE serves as a good reminder that vulnerable patients with multiple comorbidities and advanced HF often do not inherit the same benefits from ARNI as their stable ambulatory counterparts.
The concept of higher-risk patients benefiting less from staple therapies is not only applicable in the HF landscape. There seems to be a bias in placing more emphasis on trials that recruit the optimal patient profile and to generalize medical therapy guidelines that were extrapolated from such trials than those that represent realistic patients with their comorbidities – but we seem to ignore those. For example, take CORONA using statins in HF patients, the 2005 trial AURORA, and the 2009 4D trials that looked at statin use in ESRD. All three of these trials investigated high-risk patients with multiple comorbidities, and all of them showed no benefit from statins. The evidence has revealed that the driving assumption that high-risk patients benefit from common therapies or interventions is often false.
This can be seen in a variety of other trials that we routinely cite to justify our actions as “evidence-based medicine”:
- Watchman not meeting noninferiority vs. warfarin in primary endpoint for stroke and systemic embolism in PREVAIL.
- Offering Watchman to patients intolerant of anticoagulation when these exact patients were excluded in trials showing benefit.
- Not having convincing evidence that AF ablation reduces clinical outcomes in CABANA, and there are currently no placebo-controlled trials looking at quality of life.
Yet, somehow these interventions are so heavily ensconced when there isn’t enough evidence to support their definitive use.
There is a meta-analysis published by the Journal of Clinical Epidemiology in 2022 that I feel does not get the attention it deserves. The analysis looked at the data behind over 1,500 Cochrane interventions and objectively analyzed whether or not there was high-quality evidence showing clear benefits. The authors specifically used interventions that had Cochrane reviews due to their widespread implementation. They analyzed the data based on GRADE criteria where the primary endpoint had to have low bias, be statistically significant, and provide effective therapy. With such a high standard, you would think that a good number of these interventions would meet such stringent criteria to justify their widespread use. Shockingly, of those 1,500 interventions, only 5 percent were firmly rooted in “evidence-based medicine.” Furthermore, statistically significant harm was found in 8 percent of interventions.
There is tremendous bias, particularly when we talk about evidence-based medicine. Perhaps it’s the financial aspect of health care or flat-out confirmation bias, but I feel that we tend to look in the other direction when there are trials that show limited benefit in interventions that are lucrative to perform by the clinician. So the next time you want to label something as “evidence-based” medicine to justify its utilization, stop and think just how much evidence went into its designation – the results might just surprise you.
Benjamin Borokhovsky is a medical student.