A clinical trial in the United States costs, on average, between thirty and fifty thousand dollars per enrolled patient. This figure is the median across all therapeutic areas; in oncology it can run higher than a hundred thousand. The number is not a function of the drug or the disease. It is the cost of the apparatus surrounding the science — the protocols, the audits, the data capture, the regulatory filings, the institutional review boards, the contract research organizations, and the small army of nurses and coordinators whose full-time job is to keep all of this in motion.
The number has roughly tripled, in real terms, since 1990. The science has not gotten three times harder. The patients have not gotten three times harder to find. What has happened, instead, is that the regulatory and procedural overhead of doing a trial has expanded — incrementally, defensibly, and almost without anyone making a conscious decision — to a level at which the average new drug now costs about $2.6 billion to develop. Of that figure, the actual chemistry and biology probably accounts for less than a third. The rest is the cost of doing things the way we have decided they should be done.
Where the money goes
The single largest line item in a modern trial is data management. Every Phase III oncology study generates somewhere between five and ten thousand pages of data per patient — physiological measurements, lab results, imaging studies, adverse event reports, concomitant medication logs, quality-of-life surveys, informed consent documentation, and protocol deviation reports. All of this has to be captured electronically, entered into validated systems, double-checked against source documents, queried for inconsistencies, and ultimately submitted to regulators in a specific machine-readable format.
"We employed three full-time people whose only job was to check that the right boxes had been checked on the right forms. They were among the most highly paid people on the trial."
The next largest item is monitoring. Federal regulations require that an external party visit each clinical site, on a regular schedule, to verify that the protocol is being followed exactly. These visits cost about ten thousand dollars each. A medium-sized trial has dozens of sites and runs for years. The monitoring budget alone, in many trials, exceeds the cost of the drug being studied.
The third-largest item is the trial network itself. Most modern trials are run by contract research organizations — specialized firms that have built the infrastructure, expertise, and regulatory standing required to do this work. They are not cheap. A medium-sized CRO bills its sponsors at margins comparable to a top-tier law firm.
How we got here
None of this was, in any meaningful sense, designed. The regulatory architecture of the modern clinical trial is the cumulative result of approximately fifty years of incremental rule-making, almost all of it in response to specific failures. The 1962 Kefauver-Harris amendments — passed in response to thalidomide — established the modern requirement for proof of efficacy. The 1980s saw the Belmont Report and the systematization of institutional review boards. The 1990s brought ICH-GCP, a harmonized international standard for trial conduct. The 2000s brought electronic data capture mandates. Each addition was reasonable in isolation. The aggregate is a system in which it now costs more to generate the evidence for a drug than to discover the drug.
This is not a regulatory failure in the narrow sense. The FDA has not been captured. The standards are not arbitrary. The system catches real problems. It also, just as a matter of accounting, makes new drugs cost more, makes rare diseases unprofitable to study, and pushes a generation of would-be biotech founders into easier fields.
What could change
There are three reforms that, if seriously implemented, could reduce per-patient trial costs by a factor of two or more without compromising safety.
The first is real-world evidence. For many indications, particularly in chronic disease, the same questions a trial is asking could be answered by looking at electronic health records, registries, and insurance claims. The FDA has been cautiously expanding the use of real-world evidence since the 21st Century Cures Act, but the pace has been glacial. The bottleneck is not the data — which exists — but the regulatory willingness to treat it as primary.
The second is risk-based monitoring. The current model treats every trial site, every patient, and every data point as if it required the same level of scrutiny. In practice, the overwhelming majority of errors and protocol deviations cluster in a small subset of sites and a small subset of data fields. A monitoring approach that focused effort where errors actually occur — rather than uniformly across the trial — could cut costs substantially without any loss of safety.
The third is decentralized trials. Most of the cost of bringing patients to a trial site is the cost of bringing patients to a trial site. The pandemic forced a brief experiment in conducting visits remotely, dispensing drugs by mail, and capturing data from home devices. The results were generally encouraging. The infrastructure to do this routinely now exists. What is missing is the regulatory acceptance, in a wide enough range of indications, to make it the default rather than the exception.
The honest accounting
It is worth being clear about what is at stake. A drug that costs $2.6 billion to develop is a drug that the developer has to recover that cost on. The pricing of new medicines, the unwillingness of pharmaceutical companies to invest in rare diseases, the slowness with which new treatments reach the people who need them — all of these are, in part, downstream consequences of how much it costs to generate the evidence required to market a drug. The cost is not somebody else's problem.
The clinical trial system as it stands works. It produces safe drugs. It catches dangerous ones. But it costs what it costs, and what it costs is enough to deter most of the science we would otherwise be doing. The question, asked seriously, is whether we have calibrated correctly — whether the current level of caution is buying us safety in proportion to the medicines we are foregoing. Nobody seriously believes the answer is yes. The question is what to do about it.