Evidence-based medicine faces fundamental challenges that threaten its core mission of integrating clinical expertise with rigorous research, according to Adam Cifu, MD, who delivered a lecture at a Toronto physicians' meeting. His analysis identified two competing explanations for what he calls evidence-based medicine's failures: inherent design limitations predicted by early skeptics and systematic corruption by special interests.
Foundational Flaws in Trial Design
Dr. Cifu highlighted points raised by Alvan Feinstein, MD, a Yale clinical epidemiologist who said that evidence-based medicine (EBM) was "doomed from the start." Dr. Feinstein identified three critical limitations: "A minority of our clinical decisions could ever be addressed by randomized control trials," most clinical work involves "soft, difficult to measure outcomes" rather than hard endpoints, and "evidence generated by studies is always for an average patient, not for the actual patient sitting in front of you."
A West of Scotland Study explored these limitations. Of 160,000 men initially screened, only 6,600 were ultimately enrolled fora a 96% attrition rate. Dr. Cifu calculated the implications: "I care for about 850 people, if all of my patients were considered for West of Scotland, only 17 would have actually been enrolled." With a number needed to treat of 20 for the primary endpoint, "this study that we consider gospel in a way, given how it was designed and what the run-in period was like, really applies to none of the patients I care for."
Post-Randomization Confounding
The Women's Health Initiative demonstrated what Dr. Feinstein termed "post-randomization empowerment"—differential treatment patterns that emerged after randomization. Initially, approximately 7% of women in each group used statins. By trial's end, a 5% differential had developed, with the placebo group receiving substantially more statin therapy. Dr. Cifu suggested an alternative interpretation of the trial's increased cardiovascular events in the hormone therapy group. "The increase in events in the estrogen group was not because of the dangerous, toxic effect of estrogen, but it was because estrogen was basically protecting people or prohibiting people from getting lipid-lowering therapy, statins, that would have protected them from coronary heart disease," he said.
Low Positive Predictive Value
Dr. John Ioannidis's framework for assessing trial reliability revealed concerning mathematics. In therapeutic areas where only 10% of interventions prove superior to standard care—as documented in sepsis research—the positive predictive value of a statistically significant result drops to 64%, assuming no other biases exist. "That's if there are no other problems, no other biases, no publication bias, nothing," said Dr. Cifu.
Additional factors further degrade reliability. "Most of the trials we read are designed, somewhat brilliantly, to have positive effects, publication bias, analytic bias. There's just the absence of the trials that reproduce positive findings. So often positive studies that are false positives are never actually demonstrated to be false positives because they're never retested."
Medical Reversal Phenomenon
Research examining high-impact journals documented the scale of premature practice adoption. A study analyzing the New England Journal of Medicine, JAMA, and Lancet over 10 to 15 years identified 396 reversals—practices subsequently proven ineffective or inferior. "About 14% of the trials that looked at an accepted medical practice showed that practice to be ineffective," Dr. Cifu reported. The study appeared in eLife in 2019, authored by Herrera-Perez et al.
Industry-Influenced Trial Design
Dr. Cifu presented three examples of methodologically sound trials with concerning design choices that favored novel therapies:
PARADIGM-HF (2014): This trial compared sacubitril-valsartan (Entresto) to enalapril in heart failure with reduced ejection fraction. The run-in period required tolerating enalapril 10 mg twice daily for two weeks, followed by single-blind sacubitril-valsartan for 4 to 6 weeks before randomization. The control arm received enalapril 10 mg twice daily rather than the maximum dose of 20 mg twice daily. "About 15% of the patients were on enalapril. And by the time they got into the study, they ended up on a reduced dose of enalapril," Dr. Cifu noted. He characterized the design: "A run-in period of two to three times longer for the study drug sets up a trial that the standard therapy is going to look worse."
EMPEROR-Preserved (2021): This trial of empagliflozin in heart failure with preserved ejection fraction showed reduced heart failure hospitalizations (hazard ratio 0.79, p<0.05) but no significant difference in cardiovascular death. Dr. Cifu observed, "These results look like diuretic treatment for heart failure. Diuretics keep people out of the hospital, but they don't change mortality at all." Later analysis revealed empagliflozin did not reduce total hospitalizations despite decreasing heart failure-specific admissions.
STEP-TEENS (2022): This semaglutide trial in adolescents with obesity used percentage change in BMI at 68 weeks as the primary endpoint. Dr. Cifu questioned the study's clinical relevance: "You really need to ask yourself, do we want to treat teens with a drug that helps them to lose weight only while they're on the drug? This is a treatment for obesity, not a cure for the treatment." He identified the appropriate research question. "How safe are these drugs when you start them at 13 and your teen continues them until they're 23, 33, or 43."
Proposed Solutions
Dr. Cifu outlined requirements for EBM restoration: "This demands unbiased research. It demands trials that are actually designed by actual physicians, practicing physicians, to answer important clinical questions that they identify." He advocated for reducing publication volume to improve quality. "That probably means fewer but better, meaning experimental studies, instead of more weak observational studies. That's trouble because that means fewer journals. It means shorter CVs."
Regarding conflicts of interest, he proposed gradual guideline committee reform. "The AHA/ACC guidelines over the next, I don't know, 5 years, are going to transition from 50% people who've taken pharmaceutical money to 40%, to 30%, to 20%, to 10%, to 0%, then people are going to start saying no to money because that's how they want their career to progress."
For medical education, he suggested restructuring curricula. "Medical education should prioritize pragmatism over reductionism, clinical science over basic science... maybe we should begin medical school curricula with critical appraisal and translating that data to patients."
Dr. Cifu concluded by paraphrasing Winston Churchill. "Indeed, it's been said that evidence-based medicine is the worst form of practice, except for all other forms that have been tried."