In deciding whether to undergo a certain treatment for cancer or another serious illness, a patient needs to know if it is indeed the safest and most effective option available.

A complicating factor in choosing the best option is the difference between the goals of the scientists who test and approve new drugs and the goals of a person seeking treatment. The goal of a randomized trial is to determine whether the benefits of the treatment outweigh its harms among patients who are similar to those who are enrolled in the trial. This is important information, to be sure, but the patient needs to know something more specific: Is the treatment safer and more effective for people like me—those with similar health characteristics?

A trial answers both questions only when a patient happens to be similar to the participants in the clinical trial. But that is often not the case. The volunteers in clinical trials tend to be whiter, younger, healthier and more likely male than real-world patients. For example, individuals aged 65 years or older account for about two thirds of new cancer diagnoses in the U.S., but less than one third of cancer clinical trial participants. One study found that patients age 75 years or older account for only10 percent of trial participants but made up 30 percent of patients with cancer. Women make up fewer than 50 percent of participants in studies of cancer, HIV, heart disease and other conditions. A review of therapeutic cancer trials from 2003 to 2016 found that non-Latinx whites were far more likely to be enrolled than Black or Latinx patients. Across all clinical trials for cancer, Black and Latinx people are less well represented than they were 20 years ago.

There are several reasons for this lack of inclusion. Some scientific protocols explicitly exclude patients who are older, sicker, or who have functional impairments. For example, more than half of randomized trials for ischemic heart disease explicitly excluded elderly patients from enrollment, mainly because they tend to have serious chronic illnesses or are physically frail, which constitutes a de facto age restriction. A similar dynamic often results in the exclusion of Black patients, who face inequities in access to health care and insurance, chronic illnesses and racism that affects how they interact with the health system. In particular, inappropriate treatment of Black patients and research participants in the past has contributed to a lingering mistrust of the medical profession, which makes some individuals hesitant to sign up for a trial.

The poor representation of real-world patients in trials dramatically limits the ability of doctors to distinguish which patients are likely to benefit from a given treatment. The example of bevacizumab, an antibody that inhibits the growth of blood vessels in tumors, is a case in point. In a randomized trial of 878 patients with advanced lung cancer, patients who received bevacizumab lived 12 months on average—about two months longer than patients in the comparison group. About one in four trial participants were age 70 years or older. By comparison, half of all patients diagnosed with lung cancer in the U.S. are older than 70 years. This age difference is troubling because an analysis of the 224 patients in the trial older than 70 years found that bevacizumab was not associated with increased survival—and the risk of toxicity was twice as high. Of course, 224 patients over the age of 70 is a small sample size, which makes it hard to say if these results are meaningful. This kind of uncertainty is commonplace for many drugs and many patient groups. Yet new treatments still diffuse into clinical practice: Despite this unfavorable benefit-to-risk ratio, up to 20 percent of older patients with lung cancer were receiving this treatment shortly after FDA approval, studies show.

One obvious way around this dilemma is to build upon our existing clinical trial infrastructure— that is, to keep enrolling patients using the same tools we’ve always used, largely focused in large academic medical centers—and simply strive to make the studies larger. This approach would be expensive and still might not offer a broad enough population coverage, and of course it would not solve the enrollment problem.

Technology presents us with a more realistic option: the use of electronic medical records to gather data about drug safety and efficacy in the real world, outside trials, after regulatory approval. By increasing our ability to harness and analyze data, digital technology has led to a growing appreciation among medical scientists for the role that real-world data can play in informing clinicians about the appropriateness of treatments for specific groups of patients. That is, after studies conducted primarily in younger patients lead to FDA approval for a new drug, scientists subsequently explore what happens when older persons receive the drug in routine clinical practice.

A drawback to this approach is that patients who weren’t represented in the controlled and well-monitored environment of a randomized trial will wind up being real-world test subjects—without even knowing it; they will take the drug before its risks and benefits are accurately determined in their age group. In this respect, this solution is not ideal, but it is better than the current system, where data is not systematically collected once a drug is approved.

An approach starting to gain momentum in the medical community is to embed clinical research within clinical care, inviting patients in routine clinical practice to participate in studies. After all, while nearly 70 percent of children with cancer are enrolled in research studies, fewer than 5 percent of adults with cancer are enrolled. By broadening study enrollment to many more types of patients, we can accelerate the rate of developing new evidence. And by ensuring that the patients in the studies reflect the populations that may be eligible for the new treatment, the new evidence will be much more relevant to patients, clinicians, and the broader public. This critical issue applies to nearly all diseases. In COVID-19 studies, for instance, there is already evidence that elderly and minority patients are being excluded.

The importance of real-world data should be evident to fans of sports. Every spring, baseball teams start their training sessions brimming with optimism that this will be the year they make it to the World Series. A team with a roster full of stars might “look good on paper.” But we don’t really know which team is best until they actually compete against other teams—on home and away fields—over the course of a long season. By the same token, gathering data about drugs after they’ve reached clinical practice and continuing to test them iteratively, comparing treatments against alternatives, in diverse patient populations, would help doctors better prescribe winners.

Cary P. Gross, MD, is a professor of medicine and epidemiology at the Yale School of Medicine.

More in this Report