Why an Uncontrolled Study Cannot Carry a Claim

Answer first: the rule is a single sentence in 21 CFR 314.126

Paragraph (e) of 21 CFR 314.126 states it without hedging. Uncontrolled studies or partially controlled studies are not acceptable as the sole basis for the approval of claims of effectiveness.

The same paragraph then adds what such studies can do. Carefully conducted and documented, it says, they may provide corroborative support of well-controlled studies regarding efficacy, and may yield valuable data regarding safety of the test drug. They will be considered on their merits in the light of the principles listed in the section, with the exception of the requirement for comparison of the treated subjects with controls.

And it closes by naming what falls below even that. Isolated case reports, random experience, and reports lacking the details which permit scientific evaluation will not be considered.

So the regulation draws two lines, not one. Below the first line, an uncontrolled study is real evidence about safety and can support a controlled result. Below the second, a collection of anecdotes is not evidence in this system at all.

The rule is also stated in terms of what a claim rests on rather than in terms of what a study is called. The same section says that reports of adequate and well-controlled investigations provide the primary basis for determining whether there is substantial evidence to support claims of effectiveness for new drugs. Paragraph (e) is the mirror image of that sentence: the primary basis has to come from somewhere, and it cannot come from a study with nothing to compare against.

Uncontrolled studies are not an aberration — phase 3 contains them

It would be easy to read the rule above as describing something that happens off to one side of drug development. It does not. The definition of phase 3 at 21 CFR 312.21 describes phase 3 studies as expanded controlled and uncontrolled trials, performed after preliminary evidence suggesting effectiveness has been obtained.

That single phrase resolves a confusion that shows up constantly in marketing copy. A trial can be large, late-stage, run under a federal application, and reported in detail, and still not have a comparison group. The stage label does not carry the design.

It also explains why the two rules coexist. Development programs generate a mixture: some studies establish an effect against a comparison, others accumulate exposure and safety information in people who all received the drug. Both are useful. Only the first kind can be the sole basis for a claim that the drug works.

Corroborative support is a real category, and it has a shape

The phrase in the regulation is corroborative support of well-controlled studies. That word — corroborative — assumes something already exists to corroborate. An uncontrolled study is doing a supporting job in a structure whose load is carried elsewhere.

The safety half of the sentence is where uncontrolled data does its heaviest work. Under 21 CFR 312.33 a sponsor's annual report must include a narrative or tabular summary showing the most frequent and most serious adverse experiences by body system, a summary of all safety reports submitted during the year, a list of subjects who died during participation with the cause of death for each, and a list of subjects who dropped out in association with any adverse experience, whether or not thought to be drug related. None of those requirements depends on a control group existing.

FDA's draft guidance for weight-reduction drugs shows the same asymmetry in how a program is sized. It says the number of subjects necessary to demonstrate efficacy in each individual trial is generally smaller than the number needed to adequately assess safety, and that a sponsor can either increase the size of the two adequate and well-controlled trials necessary to support approval or base the safety analysis on integrated data from multiple adequate and well-controlled trials.

The same guidance warns about pooling. Integrated analyses of safety data should generally be stratified by trial, it says, because stratification is important to prevent confounding — it names Simpson's paradox — that can occur when pooling trials with different randomization ratios and populations with different risks of adverse events. It also lists differences to consider across studies, including different treatment durations, different trial populations, and different methods of adverse event ascertainment such as questionnaire versus general inquiry.

The regulation also has an escape hatch, and it is narrow and public

Paragraph (c) of 314.126 lets the Director of the Center for Drug Evaluation and Research waive, in whole or in part, any of the design criteria for a specific clinical investigation — on the Director's own initiative or on the petition of an interested person, either before the investigation or in the evaluation of a completed study.

The conditions on that petition tell you how the agency thinks about it. A petition must set out clearly and concisely the specific criteria from which waiver is sought, why those criteria are not reasonably applicable to the particular investigation, what alternative procedures are to be or have been employed, and what results have been obtained. It must also state why the investigation will yield, or has yielded, substantial evidence of effectiveness notwithstanding nonconformance with the criteria for which waiver is requested.

In other words, the waiver route does not lower the evidentiary standard. It asks the sponsor to explain how the standard is met by other means, in writing, in advance or on review.

There is a version of this rule with teeth during a trial

The consequences are not confined to the approval decision. Under 21 CFR 312.42, FDA may place a proposed or ongoing investigation that is not designed to be adequate and well-controlled on clinical hold on several grounds that are specific to that kind of study.

Among them: that there is reasonable evidence the study is impeding enrollment in, or otherwise interfering with the conduct or completion of, a study that is designed to be an adequate and well-controlled investigation of the same or another investigational drug; that insufficient quantities of the drug exist to conduct both; that the drug has been studied in one or more adequate and well-controlled investigations that strongly suggest lack of effectiveness; that another drug under investigation or approved for the same indication and available to the same patient population has demonstrated a better potential benefit-risk balance; or that the sponsor of the controlled study is not actively pursuing marketing approval with due diligence.

Read as a group, those grounds treat controlled trials as a scarce public resource. An uncontrolled study that competes with one for participants or supply can be stopped for that reason alone.

For a reader comparing commercial pages, the takeaway is a question rather than a verdict. When a page cites a study, the first thing to establish is what the treated group was compared with. If the answer is nothing, the regulation's own position is available and quotable: such a study may support safety conclusions and may corroborate a controlled result, and it cannot be the sole basis for a claim that the drug works.