Walk into any cardiology ward on a Wednesday morning. The beds hold a 74-year-old with diabetes, stage 3 chronic kidney disease, and a recent heart failure admission. Next to her, an 81-year-old with atrial fibrillation, COPD, and a hemoglobin that’s been drifting downward for six months. Across the hall, a 67-year-old man recovers from his second stent placement—still smoking intermittently, still juggling six medications prescribed by three different specialists. This is the texture of modern medicine. A world of accumulation, polypharmacy, and physiological entropy. Now ask yourself: how many of these people resemble the subjects enrolled in the landmark trials that produced the guidelines pinned to their charts? The answer is a statistical near-zero. The clinical trial population is a carefully curated fiction, and the gap between that fiction and the messy reality of the bedside is a public health problem we’ve tolerated for far too long.
The Architecture of Exclusion
The problem doesn’t start with malice. It starts with methodology. The randomized controlled trial, for all its epistemological power, is built on a foundation of homogeneity. Its statistical logic demands it. To detect a treatment signal amid the noise of biological variation, researchers need a clean sample. Every variable introduces variance, and variance inflates the sample size required to achieve significance. The solution, generation after generation, has been to exclude complexity.
Look at a typical phase III trial protocol. The inclusion criteria read like a recipe for a patient who barely exists outside a university campus wellness program. Age limits, often set at 65 or 75, are standard. Comorbidities that might interfere with the drug’s metabolism or confuse the endpoint get carved out: renal impairment, hepatic dysfunction, a second active cancer, a recent cardiovascular event. Concomitant medications are restricted because of the dread of drug-drug interactions muddying the pharmacokinetic data. The result? A trial population that is younger, healthier, and pharmacologically simpler than the target population by a staggering margin.
Consider the evidence base for heart failure with reduced ejection fraction. The landmark trials—SOLVD, MERIT-HF, PARADIGM-HF—changed practice. Yet their participants had a mean age in the early sixties. The average age of a heart failure patient in the community is closer to 80. In PARADIGM-HF, patients with an eGFR below 30 mL/min were excluded. In real-world registries, over 40% of heart failure patients have chronic kidney disease stage 3 or worse. We’re extrapolating efficacy and safety data from a middle-aged, renally intact cohort to an elderly, renally compromised one, and calling it evidence-based medicine.
Oncology’s Particularly Acute Failure
Oncology amplifies this mismatch into an ethical crisis. Cancer is a disease of aging. The median age of a cancer diagnosis in the United States is 66. Yet a 2013 analysis in the Journal of Clinical Oncology found that patients over 65 represented only about 36% of enrollees in registration trials for new cancer drugs, despite making up nearly 60% of the incident cancer population. For patients over 75, the representation drops to a rounding error. These are the very patients who will fill the infusion chairs once the drug is approved.
The clinical consequences aren’t abstract. Older adults have reduced organ reserve. Their bone marrow is less resilient. Their sarcopenia alters drug distribution. A chemotherapy regimen tested in a 55-year-old with no other medical problems can be catastrophically toxic in a 78-year-old with mild renal impairment and a gait speed below 0.8 meters per second. We learn this not from the registration trials—which excluded that 78-year-old—but from post-marketing surveillance and the hard experience of community oncologists. The trial, in its pursuit of a clean answer, has abdicated its responsibility to inform the treatment of the majority.
The Polypharmacy Blind Spot
There’s a deeper layer of distortion. Even when trials include older patients, they select for a version of aging that’s unnaturally pristine. Trial participants, by virtue of the consent process and the demands of protocol adherence, are a self-selected elite. They have the cognitive capacity, social support, and transportation access to navigate a major academic center. They’re not the patient who misses appointments because the bus route changed, or who can’t read the consent form in English, or who’s taking a dozen medications that interact in ways no algorithm can fully predict. The trial population is systematically depleted of the very patients who are most vulnerable to both the disease and the treatment.
Polypharmacy is the elephant in the exam room. The average American over 65 takes four or more prescription drugs. A substantial fraction takes ten or more. These drug cocktails create a unique pharmacodynamic milieu that’s entirely unrepresented in a trial that demands a washout period or excludes common concurrent medications. When the new blockbuster drug is approved and layered onto this existing polypharmacy, we’re conducting an uncontrolled, unmonitored experiment on a massive scale. The clinical trial, for all its rigor, has provided almost no guidance for this scenario.
The Sex and Ancestry Distortion
The exclusion problem compounds along axes of sex and genetic ancestry. For decades, women of childbearing potential were excluded from early-phase trials out of a paternalistic fear of teratogenicity, creating a knowledge gap in dosing and efficacy that persists. Cardiovascular trials historically enrolled mostly men, leading to a systematic underestimation of sex-specific side effects and a one-size-fits-all approach to dosing that doesn’t account for differences in body composition and drug metabolism.
The representation of non-European ancestry remains dismal. A 2020 analysis in JAMA found that trials supporting FDA drug approvals from 2015 to 2019 enrolled a median of only 8% Black participants. This isn’t merely a matter of social justice; it’s a pharmacokinetic error. Genetic polymorphisms in drug-metabolizing enzymes like CYP2D6 and CYP2C19 vary markedly by ancestry. A drug dose calibrated in a predominantly white European cohort may be subtherapeutic or toxic in a population with a different distribution of metabolizer phenotypes. The trial, by treating humanity as a monolith, bakes imprecision into the prescription pad.
The Pragmatic Trial: A Partial Antidote
The research enterprise isn’t blind to this failure. The rise of pragmatic clinical trials, which aim to embed randomization into routine care with broad eligibility criteria, represents a necessary correction. Trials like the RECOVERY platform during the COVID-19 pandemic demonstrated that it’s possible to generate actionable evidence at enormous scale and speed within the chaos of real clinical practice. The inclusion criteria were essentially “the doctor thinks you have COVID-19,” and the results changed global practice within months.
Yet pragmatic trials remain the exception. They’re harder to fund, harder to execute, and produce messier data that challenges the tidy narratives prized by journals and regulators. They require a cultural shift in what we consider a valid answer. The scientific community must learn to tolerate a wider confidence interval in exchange for a result that actually applies to the patient in front of them. Precision is not the same as accuracy, and a precise estimate of the wrong population is simply a well-measured falsehood.
Regulatory and Industry Inertia
The forces maintaining the status quo are formidable. Industry sponsors, understandably, want the cleanest possible signal to secure regulatory approval. A broader population increases the risk of adverse events that could derail a filing. Regulators, despite public statements encouraging diversity, have historically accepted trial data from narrow populations without demanding strong post-marketing studies in the elderly or multimorbid. The FDA’s guidance on geriatric patients in clinical trials has existed for decades, but its enforcement remains tepid.
The solution isn’t to abandon the randomized trial, which remains the least bad tool we have for causal inference. The solution is to demand that the trial’s population mirror the disease’s population. This means mandating proportional enrollment of older adults, requiring explicit justification for excluding patients with common comorbidities, and funding the larger sample sizes needed to handle the resulting variance. It means building a research infrastructure that reaches into community hospitals and rural clinics, not just academic medical centers in wealthy zip codes. It means paying for the complexity instead of pretending it doesn’t exist.
The Clinical Translation Gap
At the bedside, the physician is left to perform a series of awkward mental adjustments. She reads the trial, notes the mean age of 58, the exclusion of renal disease, the single-agent background therapy, and then looks at her 79-year-old patient with an eGFR of 38 on five other drugs. She applies a vague, experience-based discount factor to the reported efficacy and an inflation factor to the toxicity. This isn’t science; it’s art informed by anxiety. The guidelines, derived from these trials, offer a class I recommendation with a footnote about limited data in the elderly. That footnote is doing an enormous amount of heavy lifting.
The clinical trial population does not represent real patients because the system has not been forced to make it do so. The costs of this failure are measured in preventable adverse drug reactions, futile treatments, and a creeping therapeutic nihilism among clinicians who have learned not to trust the literature. The fiction of the representative trial is a luxury the healthcare system can no longer afford. The patients are complex, and the evidence base must become complex enough to meet them.
Frequently Asked Questions
Why can’t researchers just include older, sicker patients in their trials?
They can, but it’s expensive and statistically inconvenient. Older, multimorbid patients have higher rates of competing events—death from other causes, hospitalizations, drug intolerances—that dilute the measurable treatment effect. This forces sponsors to enroll far larger sample sizes to achieve statistical significance. Current incentives reward smaller, faster trials that show a clean benefit, even if that benefit has uncertain relevance for the patients who will actually use the drug. Regulatory and funding bodies need to explicitly require and pay for the enrollment of representative populations.
How should I interpret a clinical trial for my older, multimorbid patient?
Start by looking at the baseline characteristics table, not the abstract. Compare the trial population’s age, renal function, and comorbidity burden to your patient. If your patient would have been excluded from the trial, the results are a rough hypothesis, not a prediction. Apply a healthy skepticism to both the magnitude of benefit and the tolerability. Seek out subgroup analyses by age and renal function, but recognize they’re often underpowered. When available, prioritize evidence from pragmatic trials or large, representative registries over tightly controlled explanatory trials.
Are regulators doing anything to fix this problem?
They’re making incremental moves. The FDA has issued multiple guidance documents urging greater diversity in clinical trials and has begun to require diversity plans for certain studies. However, these plans often lack enforceable teeth. Real change will require legislative mandates that tie drug approval to demonstrated representation of the target population, coupled with substantial public investment in a trial infrastructure that can reach beyond elite academic centers. Without such pressure, the economic logic of narrow trials will persist.