Surrogate endpoints are indirect measures of clinical benefit, commonly used to determine disease status and monitor the therapeutic effects of treatments. In contrast to clinical outcomes, which measure how patients feel, function, or survive, surrogate endpoints measure changes expected to be proxies for these outcomes, such as the level of a biomarker in the blood or the size of a tumor. These surrogate endpoints are commonly used in clinical trials, because they can be measured more readily and reflect responses to treatment more quickly than clinical outcomes, fostering shorter and less expensive studies. However, even as surrogate endpoints have been championed, their reliability has been questioned. Examples have accumulated where surrogate endpoints have failed to predict the clinical outcomes they purported to represent. Just two recent examples are the poor correlation between progression-free survival and quality of life measures or overall survival among patients with cancer,^{Reference Prasad, Kim, Burotto and Vandross1} and the poor correlation between HbA1c to microvascular and macrovascular outcomes among patients with diabetes mellitus,^{Reference Yudkin, Lipska, Montori, Rodriguez-Gutierrez and McCoy2} are just two recent examples.

In this issue of JLME, several groups of authors describe the discussions that took place at a 1-day symposium to address the challenges and opportunities for biomarker validation, including as surrogate endpoints. Highlighted at the symposium were new pathways for biomarker investigation, including testing treatments with the assistance of algorithms, taking better advantage of data sharing platforms, and standardizing querying systems for published literature and systematic reviews. However, much work remains to be done, including in standardizing the procedures by which validation occurs, which will require collaboration between the National Institutes of Health (NIH), the Food and Drug Administration (FDA), and other stakeholders.

The FDA, in particular, plays a key role in the use and validation of biomarkers as surrogate endpoints. Over the past several decades, the FDA's sophistication on this topic has grown steadily, beginning with a burgeoning recognition of surrogate endpoints as conceptually distinct from clinical endpoints in the 1980s; to the explicit endorsement of approvals relying on surrogate endpoints through the Accelerated Approval pathway in 1992; to the establishment of programs that aim to validate biomarkers as surrogate endpoints. Now, surrogate endpoints are a fixture in new drug approvals, with almost half of drugs approved between 2005-2012 relying solely on pivotal trials using surrogate endpoints.^{Reference Downing, Aminawung, Shah, Krumholz and Ross3} As the 21st Century Cures Act explicitly calls on the FDA to rely more on biomarkers and other surrogate measures in assessing the efficacy of both drugs and devices, the next era is now upon us, surely ushering in even greater use.

Given the FDA's widespread and still growing use of surrogate endpoints, ensuring that evidence generated using surrogate endpoints reliably informs patient care is more important than ever. However, the execution of many existing initiatives encouraging surrogate endpoint use shows there to be room for improvement. For instance, the statutory language for Accelerated Approval calls for the confirmation of surrogate endpoint benefit through postmarketing studies with clinical endpoints.⁴ However, one study conducted by FDA scientists found that the majority of such studies designated as “confirmatory” by the FDA did not actually demonstrate benefit using clinical endpoints, but instead did so using surrogate endpoints, effectively rendering no validation of surrogate endpoint use.^{Reference Beaver, Howie and Pelosof5} The FDA also has room to expand its use of other levers by which it can encourage stronger validation of biomarkers. For instance, the FDA's ability to negotiate mandatory postmarketing requirements (PMRs), or even to elicit voluntary postmarketing commitments (PMCs) at the time of approval, could be used to ensure that studies are conducted to confirm the benefit of drugs approved on the basis of surrogate endpoints, even in cases where a special regulatory program, such as Accelerated Approval, is not invoked. Given that the majority of drugs supported by studies using surrogate endpoints fall into this category,^{Reference Pease, Krumholz, Downing, Aminawung, Shah and Ross6} there is considerable room to use and enforce this mechanism as a means to generate the data necessary for validation.^{Reference Wallach, Ross, Naci, Wallach, Egilman and Dhruva7}

Fostering the conduct of confirmatory studies such as these will require collaboration with industry and research institutions, which are key partners in the effort to understand and validate biomarkers for use as surrogate endpoints. One way in which collaboration has already been fruitful has been in the development of data sharing platforms, such as Clinical Study Data Request (CSDR)^{Reference Strom, Buyse, Hughes and Knoppers8} and the Yale Open Data Access (YODA) Project,^{Reference Ross, Waldstreicher and Bamford9} by which industry sponsors make individual patient-level data (IPD) from their clinical trials available to other investigators, sometimes with research institutions serving as stewards for data access. In particular, the availability of IPD through these platforms enables secondary analyses by researchers and furthers the ability to use existing data to answer novel clinical questions, including meta-level questions about biomarker validity.

The promise of using biomarkers as surrogate endpoints in advancing our understanding and management of disease is evident. But to benefit patients, biomarkers used for this purpose need to be valid and reliable, with expected benefits confirmed in clinical trials. Continued collaborations between regulators, industry, and researchers will be necessary to avoid the pitfalls of untested surrogates and enable them to fully realize their potential.