Drug Approvals in Oncology: Striking the Right Balance Between Saving Lives and Patient Safety

In 1962, the Kefauver-Harris Amendment to the United States Federal Food, Drug, and Cosmetic Act mandated for the first time that pharmaceutical companies seeking drug approval must provide proof of clinical benefit (efficacy) from well-controlled studies. Before this amendment, drug companies mainly needed to document the safety of the drug they intended to market, per the requirements of the 1938 Federal Food, Drug, and Cosmetic Act. The Kefauver-Harris Amendment also removed time constraints on the US Food and Drug Administration (FDA) adjudication and approval of new drug applications, which was previously restricted to 180 days. This was a landmark piece of legislation, and the FDA as we know it would not exist without this amendment.

Clinical benefit for regulatory purposes is defined as an improvement in overall survival or a validated patient-reported outcome, and the FDA typically requires evidence of such benefit from 2 randomized controlled trials in order to approve a new drug. However, because of the serious and often fatal nature of cancer, the agency over the years has progressively relaxed its requirements for proof of efficacy in order to allow potentially lifesaving drugs to be available to the public in a timely manner. Often, a single randomized trial revealing clinical benefit is enough to secure regular approval of an oncologic drug. However, even this requirement can mean a delay of years before a drug is approved, which means that many patients may die waiting for an effective treatment to become commercially available. Thus, for serious illnesses like cancer, the FDA rightly accepts surrogate end points such as response rate or progression-free survival that are considered likely to accurately predict the true outcome of interest (overall survival). The use of surrogate end points allows drugs to come into the market faster, and through postmarketing surveillance the occasional drug that is inappropriately introduced can later be identified and withdrawn from the market. The current standards are meant to provide the optimal balance between speedy delivery of promising drugs when dealing with deadly diseases vs keeping the public safe from useless or harmful drugs.

In the current issue of Mayo Clinic Proceedings, Kim and Prasad report that in many drug approvals, the surrogate end points used were not well validated or verified to consistently predict the outcome of interest. They recommend changes in the system so that unverified surrogate end points are not used for new drug approvals.

The study by Kim and Prasad is the product of a time-consuming, thoughtful, and careful study that examined 55 oncologic drug approvals from 2009 through 2014. They found that a published analysis of the strength of the association between the surrogate end point used and overall survival had been formally studied in only 30 approvals (55%). When they graded the type of studies done, the strength of the surrogate-survival association was assessed through a high-quality (level 1) analysis in only 19 approvals (35%). Further, in most such analyses, the association was weak.

On the surface, the findings of Kim and Prasad can be interpreted as evidence that the FDA is more lenient, and is willing to approve oncologic drugs more readily, than ever before. The authors' findings can also be used to advocate for changes in the way drugs are approved in the United States and the need to place a greater standard for proof of efficacy before drug approval and clinical application. Kim and Prasad argue that when surrogate end points are used for drug approval, high-quality published analysis documenting a consistent and strong association between the surrogate and overall survival is necessary. They recommend that the FDA reconsider the current practice of approving drugs in the absence of such evidence. In some instances, such as regular approval of marginally effective drugs in solid tumors, these recommendations do make sense. However, having been on both sides of the equation—both as an investigator on regulatory trials and as a clinical oncologist who treats patients with cancer—I have a different view.

The data provided by Kim and Prasad are accurate. However, there are many aspects in the interpretation of those data that one must be wary of, so that we do not impede progress in drug development and accessibility. First, the existence (or nonexistence) of a published analysis of the strength of the association between the surrogate end point and overall survival is not the same as whether there is (or is not) information to support an association between the two. In many instances, the association is assumed on the basis of historical trials for which a formal analysis of the surrogate end point's validity and utility was never published. In other areas, such as uncommon cancers, there are simply not enough studies or investigators to document the association. It is also possible that a formal analysis of the strength of the association was done and submitted to the FDA by investigators or the pharmaceutical company but was never published. The primary FDA reviewer who examined the application may have also performed such an analysis.

When bortezomib was first studied for the treatment of multiple myeloma, there was no study that analyzed whether response to therapy (a surrogate) correlated with overall survival in the disease. However, investigators were unanimous that the dramatic responses seen with bortezomib in just 2 or 3 patients in the first phase 1 study was sufficient proof that the drug would be lifesaving. Historically, myeloma was a disease with limited treatment options, and new drugs simply did not work. As a result, no analysis had been done to evaluate the strength of a surrogate-survival association. Waiting for a phase 3 trial to prove survival benefit would have meant years of delay in bringing the drug to market. The FDA did not approve bortezomib on the basis of an unvalidated surrogate end point: it did so on the basis of input from numerous experts who were convinced of the strength of the surrogate-survival association, the lack of a formal analysis notwithstanding.

Second, what level of proof should we demand to decide whether a surrogate reasonably predicts overall survival? Yes, it is great to have many high-quality published studies examining the strength of the surrogate-survival association, but a quick look at Tables 1 and 2 in the article by Kim and Prasad indicates that such analyses are often lacking in uncommon cancers. In these cancers, there is a paucity of data to conduct such analysis, and if such data existed, there is a paucity of investigators interested in contributing to such studies. For these life-threating cancers, why should there be a need to have anything more than a biologically plausible association (considered level 3 by Kim and Prasad) for the surrogate-survival correlation?

Finally, we come to the crux of the issue. Drug approvals for life-threatening diseases need to balance speed and safety. Speed can save lives by bringing new active drugs to treat diseases like cancer to the public in a timely manner. Safety can save lives by preventing deaths due to drug-induced toxicity. But which approach would save more lives, (1) allowing drugs to be approved rapidly for serious fatal diseases like cancer using promising surrogate end points, with or without published high-quality validation analysis, or (2) demanding proof of the association between surrogate and survival, and if such proof cannot be produced, requiring survival benefit in a randomized trial? Ironically, short of a randomized trial, there is no definitive way to answer this question.

In my view, the FDA has struck the right balance between the need to save lives and safety. By programs such as “breakthrough status,” “priority review,” “fast-track designation,” and “accelerated approval,” they have minimized delays in the entry of effective lifesaving drugs into the market by many, many years. In my field of work (multiple myeloma), bortezomib, carfilzomib, pomalidomide, and daratumumab were all approved by the FDA on the basis of single-arm uncontrolled studies that used response rate as a surrogate end point. In each of these instances, we assume that in many countries where there are more restrictive rules for drug approval, numerous patients must have died before they had access to the possible benefits of these drugs. The concerns laid out by Kim and Prasad can be addressed by stricter postmarketing enforcement policy to remove useless and/or dangerous drugs from the market and by aggressive policing to make sure any required confirmatory phase 3 trials are performed in a timely manner. The current FDA standards for oncology drug approval are rigorous but fair and should not be further relaxed, especially with new precision medicine drugs for which some have advocated a more liberal approval policy.

In a way, the system the FDA currently has in place accepts a minimal risk that a useless drug may be approved. This is a small price to pay for early access to cancer drugs, and thankfully, it is rare that a harmful, totally useless cancer drug is approved. After all, the drug that started it all and led to the Kefauver-Harris Amendment, thalidomide, was eventually approved in the United States to treat newly diagnosed multiple myeloma on the basis of a surrogate end point (overall response rate) that is still not validated as a reliable surrogate for overall survival in previously untreated patients.^,  Even with a notorious drug like thalidomide, the high probability of saving lives trumped concerns about safety or inadequate proof of efficacy, and time has confirmed that this was the right decision.