Net Treatment Benefit: A Patient-Centered Approach to Composite Endpoints

Moving beyond traditional composite endpoints

Patient-centricity has become an increasingly important principle in clinical research and treatment evaluation. There is a growing movement to involve patients more actively in clinical trial design and decision-making, with patient preferences, quality of life, and individual values increasingly recognized as important considerations when assessing treatment success.

This has also increased interest in statistical approaches that can evaluate multiple clinical outcomes together. Traditional composite endpoints combine several outcomes into a single endpoint, but they may not always reflect the relative clinical importance of those outcomes. A composite endpoint can, for example, give equal importance to events that patients and clinicians may consider very different in terms of their impact.

Net Treatment Benefit (NTB) offers an alternative approach. NTB is estimated using generalized pairwise comparisons (GPC), allowing multiple prioritized outcomes to be evaluated within a single analysis. Outcomes such as efficacy, safety, and quality of life can be considered according to a predefined hierarchy, with the possibility of incorporating thresholds for clinical relevance.

By allowing the relative importance of outcomes to be defined before the analysis, NTB can provide a more comprehensive assessment of treatment effects while also creating an opportunity to incorporate patient input into the definition of what constitutes a meaningful treatment benefit.

The methodology behind Net Treatment Benefit

NTB is a measure of treatment effects estimated using generalized pairwise comparisons (GPC), a statistical method that stems from the well-known Mann-Whitney-Wilcoxon non-parametric test.¹ GPC provides a global assessment of treatment effects by hierarchically integrating different types of outcomes, such as efficacy, safety, and quality of life.

This approach is related to the broader concept of hierarchical composite endpoints, in which multiple outcomes are considered together according to a predefined order of priority. Unlike a conventional composite endpoint, where the occurrence of any component may determine the outcome, GPC evaluates outcomes sequentially according to their clinical priority.

The logic of GPC is relatively straightforward. Each patient in the experimental treatment group is compared with each patient in the control group. The comparison begins with the highest-priority outcome. For example, in an oncology trial, overall survival might be ranked first, followed by disease progression, treatment-related adverse events, and quality of life.

The outcomes of each pairwise comparison are classified as:

• Favorable: the patient receiving the experimental treatment has a better outcome.
• Unfavorable: the patient receiving the control treatment has a better outcome.
• Neutral: neither patient has a meaningfully better outcome.

Thresholds for clinical relevance can also be predefined. These thresholds specify the minimum difference between two patients that is considered clinically meaningful.

If a pair is classified as neutral on the highest-priority outcome, the comparison proceeds to the next outcome in the hierarchy. For example, if two patients have sufficiently similar survival outcomes, their treatment-related adverse events may then be compared. If the comparison remains neutral, the analysis continues to the next prioritized outcome.

This process continues through the hierarchy until the pair is classified as favorable, unfavorable, or remains neutral across all outcomes.

The resulting Net Treatment Benefit is calculated as the difference between the probability of a favorable outcome and the probability of an unfavorable outcome. A positive NTB indicates that the experimental treatment has a net benefit over the control treatment, while a negative value indicates a net harm.

Composite endpoints, hierarchical outcomes and Net Treatment Benefit

Composite endpoints are widely used in clinical trials because they allow several clinically relevant events to be evaluated within a single statistical framework. However, conventional composite endpoints generally treat the components as part of the same endpoint, even when the components differ substantially in clinical importance, frequency, or impact on patients.

Hierarchical approaches provide a way to address some of these challenges by explicitly prioritizing the outcomes. Generalized pairwise comparisons are one such approach and have been used to construct measures including the win ratio (WR) and NTB.

The win ratio has attracted considerable interest as a relative measure of treatment effect. However, relative measures can make it more difficult to understand how the individual outcomes contribute to the overall treatment effect. The treatment effect may be driven predominantly by one outcome, while the contribution of other outcomes remains less apparent.

GPC has been used in cardiovascular trials, including the ATTR-ACT trial of tafamidis in patients with transthyretin amyloid cardiomyopathy.² The trial used a hierarchical composite outcome incorporating all-cause mortality followed by hospitalization for cardiovascular causes.

NTB provides an alternative way of expressing the result of a GPC analysis. As an absolute measure, it represents the difference between the probability of favorable and unfavorable pairwise comparisons. This can make the overall treatment effect easier to interpret and can provide a clearer view of how individual outcomes contribute to the result.

Rather than asking only whether a treatment performs better on a single endpoint, or whether a predefined composite endpoint has been met, a hierarchical analysis can provide a broader assessment of how treatment benefits and harms compare across outcomes that matter to patients.

Using multiple outcomes to assess benefit and risk

The potential value of NTB extends beyond conventional efficacy endpoints. Clinical decisions often involve trade-offs between benefits and harms. A treatment may improve an important efficacy outcome while increasing toxicity, for example, or provide similar efficacy while offering better tolerability or quality of life.

Conventional trial designs can make these trade-offs difficult to evaluate within a single treatment effect measure. Safety and quality of life outcomes are often analyzed separately from the primary efficacy endpoint, leaving decision-makers to consider several results simultaneously.

NTB can bring these outcomes together within a predefined hierarchy. For example, an analysis could prioritize overall survival, followed by disease progression, serious adverse events, and quality of life. The resulting treatment effect reflects the predefined priorities while preserving information about the individual outcomes included in the analysis.

This approach may also be relevant when evaluating treatments in non-inferiority settings. As discussed in a recent publication in The Lancet Oncology, non-inferiority trials can place outcomes such as toxicity and other patient-relevant considerations in secondary positions, while narrow statistical margins can make it difficult for alternative treatments to demonstrate that they offer a meaningful overall advantage.³

A hierarchical GPC analysis using NTB can instead frame the question around the balance between treatment benefits and harms. Rather than focusing exclusively on whether an alternative treatment is sufficiently close to the standard treatment on a predefined endpoint, the analysis can consider whether the overall balance of prioritized outcomes favors one treatment.

Incorporating patient preferences into Net Treatment Benefit

One of the potential advantages of NTB is that the hierarchy of outcomes can be informed by patient preferences. This creates an opportunity to incorporate the patient perspective into the definition and prioritization of outcomes used in the analysis.

In traditional clinical trial frameworks, outcomes such as overall survival or disease-free survival often receive the greatest emphasis in treatment evaluation, while quality of life, symptom burden, and adverse effects may be assessed separately. Yet the relative importance of these outcomes can vary considerably between patients and disease settings.

For some patients, a small improvement in survival may be more important than a reduction in treatment-related toxicity. For others, avoiding severe adverse effects or preserving day-to-day functioning may be a higher priority. These trade-offs are particularly relevant in chronic diseases and situations where treatments offer different combinations of benefits and risks.

Patient preference research can therefore help inform the selection and prioritization of outcomes included in an NTB analysis. Patient preferences could be elicited through approaches such as surveys, discrete choice experiments, or conjoint analysis.

This does not mean that NTB automatically incorporates patient preferences. Rather, patient preferences can be used as an input into the design of the outcome hierarchy and the thresholds used to define clinically meaningful differences.

The FDA has encouraged sponsors to engage with patients early in the development process, including during clinical trial design.⁴ Understanding which outcomes patients value most can therefore provide an important foundation for developing treatment effect measures that better reflect patient priorities.

The need for better patient preference methods

Despite growing interest in patient-focused drug development, important methodological challenges remain. There is currently no universally accepted approach for translating patient preferences into the prioritization of outcomes within treatment effect measures.

Existing methods, including surveys, discrete choice experiments, and conjoint analysis, provide potential approaches for eliciting patient preferences. However, their use and interpretation can vary across development programs.

The FDA has previously recognized quantitative patient preference assessment as an evolving area of research.⁵ Continued methodological development is therefore needed to determine how patient preferences can be elicited, quantified, and incorporated consistently into clinical trial design and analysis.

For pharmaceutical and biotechnology companies, this represents an opportunity to move beyond simply collecting patient input and toward using that input to influence how treatment effects are defined and evaluated.

Toward more patient-centered composite endpoints

Composite endpoints and hierarchical outcome measures provide ways to bring multiple clinical outcomes into a single analytical framework. However, the way these outcomes are combined matters.

Net Treatment Benefit provides a framework for evaluating prioritized outcomes through generalized pairwise comparisons and expressing the result as an interpretable absolute measure of treatment effect. By incorporating efficacy, safety, quality of life, and other patient-relevant outcomes into a predefined hierarchy, NTB can help characterize the overall balance of treatment benefits and harms.

Importantly, the framework also creates an opportunity to incorporate patient preferences into the design of the analysis. As methods for eliciting and quantifying those preferences continue to develop, they could play a greater role in determining which outcomes are prioritized and how clinically meaningful differences are defined.

The broader goal is not simply to create another type of composite endpoint. It is to develop treatment effect measures that better reflect the multidimensional nature of clinical decisions and, ultimately, the outcomes that matter most to patients.

Authors: Tom Mann is Clinical Solutions Engagement Lead; Sarah Kosta, PhD, is Clinical Trial Solutions Analyst; and Samuel Salvaggio, PhD, is Senior Trial Design Lead; all with One2Treat.

References

1. Buyse, M. Generalized Pairwise Comparisons of Prioritized Outcomes in the Two-Sample Problem. Stat Med. 2010; 29(30):3245-3257.
2. Maurer, M.S.; Schwartz, J.H.; Gundapaneni, B., et al. Tafamidis Treatment for Patients with Transthyretin Amyloid Cardiomyopathy. N Engl J Med. 2018; 379(11):1007-1016.
3. Tannock, I.F.; Buyse, M.; De Backer, M.; et al. The Tyranny of Non-Inferiority Trials. The Lancet Oncology. 2024; 25(10):e520-e525.
4. FDA. FDA Patient-Focused Drug Development Guidance Series for Enhancing the Incorporation of the Patient’s Voice in Medical Product Development and Regulatory Decision Making. February 14, 2024.
5. FDA. Patient Preference Information: Voluntary Submission, Review in Premarket Approval Applications, Humanitarian Device Exemption Applications, and De Novo Requests, and Inclusion in Decision Summaries and Device Labeling. August 2016.