Surrogate Endpoints in Regulatory Decision‐Making

ABSTRACT

To support approval, FDA requires substantial evidence of effectiveness that demonstrates a drug improves meaningful clinical outcomes as measured by how a patient feels, functions, or survives. Effectiveness is measured directly (e.g., by patient‐reported outcome or other clinical outcome assessment) or indirectly (i.e., by use of a surrogate endpoint (SE) that predicts how a patient will feel, function, or survive). Use of biomarkers as SEs can hasten drug development for patients with unmet medical needs.

1. Regulatory Framework

Drugs and biologics can receive FDA approval based on traditional approval (TA) or the accelerated approval (AA) pathway, which is available for products for serious or life‐threatening conditions with unmet needs. Clinical trial endpoints that directly measure the way a patient feels, functions, or survives, for example, overall survival for a cancer drug, and validated surrogate endpoints (SEs) that predict clinical outcomes can serve as the basis for traditional approval. SEs that are considered reasonably likely to predict meaningful clinical outcomes but do not reach the level of a validated SE can serve as the basis for AA.

When the SE is known to predict clinical benefit based on strong evidence supporting the expected relationship between treatment effects on the SE and a meaningful clinical outcome, the surrogate is considered a validated SE. For serious or life‐threatening conditions with unmet medical need, a drug or biologic can be approved based on AA using an SE that is reasonably likely to predict a particular clinical benefit. A reasonably likely surrogate endpoint (RLSE) is supported by epidemiologic, human genetic and genomic, pathophysiological, therapeutic, human pharmacological, or clinical trial evidence to predict a meaningful clinical benefit, but there may be uncertainty as to the relation of the SE to clinical benefit.

Due to inherent uncertainty, AA using an RLSE may be appropriate when no validated SE is available and when clinical outcomes cannot realistically be measured in the time frame of a pre‐marketing clinical trial, such as when the disease course is long and relevant clinical outcome events are infrequent. To qualify for AA, the severity of the condition, its rarity, and the lack of alternative treatments must also be considered, and the treatment under study should represent a meaningful benefit over existing treatments.

For products approved under AA, sponsors conduct a post‐marketing confirmatory trial to verify the clinical benefit. Given the difficulties enrolling a confirmatory trial when a treatment is already approved under AA, the FDA may require that the confirmatory study be underway at the time of AA. Recent statutory provisions provide streamlined procedures for expedited withdrawal of approval if the confirmatory trial is not conducted with due diligence or if the trial fails to verify the clinical benefit, among other reasons [1].

2. Evidentiary Considerations

Determining whether an SE is “reasonably likely” to predict clinical benefit is a matter of scientific judgment that involves an assessment of the evidence supporting the biologic plausibility of the relationship between the disease, the endpoint, and the specific clinical outcome the SE is intended to predict (Figure 1) [1]. Each situation is unique, so there is no single approach to demonstrating that an SE is an RLSE. For rare diseases, the FDA applies flexibility in the evaluation of the type and quantity of data a pharmaceutical company sponsor is required to provide for individual drug development programs [2]. The FDA Biomarker Qualification Program provides a framework for the review of new biomarkers, including SE, within an intended context of use for use in regulatory decision‐making [3].

FIGURE 1.

Potential sources of data to support a surrogate endpoint. Source: www.fda.gov.

In general, for a biomarker [4] to be used as an SE to support drug approval, the assay used to measure biomarker levels should undergo thorough and robust analytic validation to determine its reliability, accuracy, sensitivity, and specificity [5]. Accuracy is particularly important in situations where the aim of therapy is to reduce the biomarker level below a particular threshold level. Optimally, the biomarker is measured in a relevant tissue or compartment; however, when this is not practical—for example, when the biomarker is measured in blood, but the pathology relates to peripheral nerve or the brain—evidence should be provided that drug effects on the biomarker measurements reliably reflect drug effects on levels in the compartment of interest.

Clinical validation concerns the development and assessment of evidence relating the biomarker to the clinical outcome of interest. Epidemiologic evidence correlating the biomarker to clinical outcomes provides important evidence linking the two. However, correlation does not prove causation, so epidemiologic data alone are not sufficient to show that a biomarker can be used as an SE. A thorough understanding of disease pathophysiology, disease progression, the causal pathway of disease, the biological relevance of the biomarker, the mechanism of action of the drug, and its distribution are foundational for establishing biologic plausibility of the proposed SE. Of note, conditions can have multiple causal pathways, creating the potential for uncertainty about the use of a biomarker as a SE. When there is a single pathway causing disease, an SE may apply in a manner agnostic to the specific therapeutic intervention—for example, elimination of a microbial pathogen in certain infectious diseases. In contrast, other SEs are specific to a particular drug or class of drugs. Regardless, a clear understanding of the mechanism of action of the drug, its distribution, and the ability of the biomarker to reflect effects on the disease process in relevant tissues or compartments at therapeutic doses provides important supportive evidence for biologic plausibility.

Translational nonclinical models—and in some cases in vitro NAMs (New Alternative Methods)—can provide important supportive information to the extent that they recapitulate important aspects of the human disease. They can provide evidence that the biomarker is on the causal pathway and that biomarker levels relate to outcomes relevant to the human disease. When nonclinical studies demonstrate that treatment with the therapeutic agent in doses relevant to the human both reduces levels of the SE and impacts relevant outcomes, they provide particularly persuasive evidence.

The strongest evidence supporting clinical validation of a biomarker as an SE comes from clinical trial evidence demonstrating that therapies that impact the biomarker beyond a specific threshold also impact clinical outcomes. While clinical evidence linking the biomarker to the intended clinical benefit is particularly persuasive, this type of data is not available for many rare diseases. When the pathophysiology of the disease and the role of the SE are well understood, interventional trial data may not be required to support a prior agreement with the Agency on the acceptability of an RLSE for use in a pivotal clinical trial for accelerated approval. Studies supporting clinical validation should provide an estimate of the minimal change in the biomarker, or threshold, that would reliably predict a meaningful clinical benefit to determine whether a therapy is likely to provide meaningful benefit, to facilitate benefit–risk assessment, and to establish the appropriate size and duration for a confirmatory trial. Of note, the size of the confirmatory trial may also be determined by the need for long‐term safety data.

3. Case Examples

The following examples illustrate some key principles underlying the acceptance of SEs for drug approval (Table 1).

TABLE 1.

Surrogate endpoint evaluation.

Approval pathway	Indication	Measure	Corresponding endpoints	Predicted benefit
RLSE for AA [tofersen]	SOD1‐ALS	NfL	Reduction in CSF NfL level	Improved disease progression and survival
RLSE for AA	MASH with fibrosis	Liver histopathology	Proportion of patients with improvement in fibrosis and no worsening of inflammation	Improved liver‐related long‐term outcome events
Validated SE for TA	HIV disease	HIV RNA	Percent of patients with reduction to undetectable	Lower risk of disease progression or death
Validated SE for TA [palovarotene]	FOP	HO	Reduction in volume of new HO	Improved CAJIS score

Abbreviations: AA, accelerated approval; CAJIS, Cumulative Analogue Joint Involvement Scale (CAJIS) score; CSF, cerebrospinal fluid; FOP, Fibrodysplasia Ossificans Progressiva; HIV, human immunodeficiency virus; HO, heterotrophic ossification; MASH, metabolic dysfunction‐associated steatohepatitis; NfL, neurofilament light; RSLE, reasonably likely surrogate endpoint; SE, surrogate endpoint; SOD1‐ALS, Superoxide Dismutase 1‐Amyotrophic Lateral Sclerosis; TA, traditional approval.

Clinical development programs relying upon SEs must meet standards for substantial evidence of effectiveness [6]. In the case of very rare conditions, this may mean reliance on a single adequate and well‐controlled trial plus confirmatory evidence. An effect on an SE in a single trial is generally not sufficient by itself. For the tofersen drug development program for Superoxide Dismutase 1 (SOD1)‐Amyotrophic Lateral Sclerosis (ALS) [7] targeted at reducing SOD1 levels, the pivotal trial did not demonstrate statistical significance on the clinical endpoint, the ALS functional rating scale. Substantial evidence of effectiveness for AA was provided by the level of reduction in plasma neurofilament light (NfL), a biomarker of neuronal damage that is prognostic for disease progression and survival in ALS, as an RLSE supported by the mechanistic evidence of reduction in SOD1 protein, the prognostic value of NfL in ALS, and a correlation with clinical outcomes. A decrease in SOD1 in cerebrospinal fluid (CSF) demonstrating that the drug impacted its target provided confirmatory evidence [6].

Evidence supporting an SE can come from multiple sources besides randomized controlled trials. Liver histopathology is accepted as an RLSE for metabolic dysfunction‐associated steatohepatitis (MASH) with fibrosis. MASH patients are at risk of progression to cirrhosis and end‐stage liver disease. However, clinical events are infrequent and are not practical for use in the time frame of a typical pre‐marketing clinical trial. In MASH, the complex pathophysiology of the disease and causal pathway have made finding surrogate markers other than liver histology challenging. However, epidemiologic data tie inflammation and advanced fibrosis to adverse clinical outcomes, although clinical trial data relating changes in these biomarkers to clinical outcomes are not yet available. A 2018 draft guidance recognized a demonstration of resolution of steatohepatitis with no worsening of liver fibrosis, or improvement of liver fibrosis with no worsening of steatohepatitis, or both as potential RLSEs. Currently, non‐invasive tests are also being explored as potential RLSE [8, 9].

HIV RNA levels for HIV disease demonstrate the use of convergent evidence to support an SE. Its use was accepted as a validated SE for HIV disease for traditional approval based on the strong understanding of the pathophysiology of the disease, the place of the biomarker on the causal pathway, epidemiologic data, and treatment trial data relating the biomarker to disease progression or death. To confirm the validity of HIV RNA, the FDA conducted a patient‐level analysis of more than 5000 subjects in the late 1990s, which indicated a direct association between sustained reduction in HIV RNA and a lower risk of disease progression or death. In addition, a meta‐analysis was published with similar findings [10].

Well‐designed natural history studies can support validation of SEs. Fibrodysplasia Ossificans Progressiva (FOP) is a rare, severely disabling disease caused by a gain‐of‐function mutation in the activin A type 1 receptor, causing heterotrophic ossification (HO) in connective tissue, joints, and muscle. Palovarotene received TA for FOP based on the reduction in the volume of new HO, an imaging‐based SE. HO was accepted as a validated SE for traditional approval based on a strong understanding of the disease pathophysiology and natural history data showing a correlation between HO volume and the Cumulative Analogue Joint Involvement Scale (CAJIS) score, which measures gross mobility limitations. No clinical trial data were available to relate changes in the SE to subsequent clinical outcomes.

4. Conclusion

Appropriate use of biomarkers as RLSEs can hasten drug development and ensure novel treatments get to patients with unmet needs more quickly. Establishing a rigorous approach to the assessment of biomarkers for use as RLSEs will expedite drug development for serious diseases with unmet medical needs.

Funding

The authors have nothing to report.

Disclosure

Disclaimer: The opinions expressed in this manuscript are those of the authors and should not be interpreted as the position of the US Food and Drug Administration.

Conflicts of Interest

The authors declare no conflicts of interest.

Jeng L. J. B. and Siegel J., “Surrogate Endpoints in Regulatory Decision‐Making,” Clinical and Translational Science 18, no. 12 (2025): e70445, 10.1111/cts.70445.