How to assess treatment benefit when efficacy and safety outcomes compete in clinical trials

Clinical trials rarely produce a simple picture of treatment benefit. A new treatment may improve an important efficacy outcome while also causing more adverse events. When efficacy and safety outcomes in clinical trials point in different directions, how should the overall treatment benefit be assessed?

This is a common challenge in clinical development. Looking at each outcome separately provides important information, but it can make it difficult to understand the overall treatment effect when several outcomes matter at the same time.

One approach is to use a statistical method that evaluates multiple outcomes together while accounting for their relative importance.

How can safety and efficacy outcomes be assessed together?

One way to address this challenge is to define a hierarchy of clinically meaningful outcomes and evaluate treatments according to that hierarchy.

This is the principle behind Generalized Pairwise Comparisons (GPC) and its associated measure, Net Treatment Benefit (NTB).

Rather than analyzing each outcome independently, GPC compares patients between treatment groups across a predefined sequence of outcomes.

For example, efficacy might be considered first. If two patients are considered similar on that outcome, the comparison can move to the next priority, such as safety or tolerability.

The result is an integrated comparison across multiple outcomes.

NTB expresses the balance between favorable and unfavorable patient comparisons. It can therefore provide a quantitative measure of treatment effect when efficacy and safety outcomes need to be considered together.

A clinical trial example: S-1 versus UFT

A recent re-analysis of the JFMC 35-C1 trial provides an example of this approach in oncology.

The trial compared one year of adjuvant S-1 with uracil/tegafur (UFT) in patients with stage II/III rectal cancer. The original trial showed a significant improvement in relapse-free survival with S-1, but the two treatments also had different safety profiles.

The researchers therefore used NTB to assess the treatments jointly across both efficacy and safety outcomes.

They prioritized:

Relapse-free survival
Grade ≥3 symptoms
Grade ≥3 laboratory abnormalities

When these outcomes were analyzed together in this hierarchy, the multivariate NTB was 8.8% in favor of S-1 (95% CI, 2.7% to 14.9%; P = .014).

Importantly, the analysis did not simply add efficacy and safety results together. The hierarchy determined how the outcomes were considered when comparing patients between treatment groups.

The study also showed why the choice of outcomes and their prioritization mattered. In sensitivity analyses, the NTB was generally positive in patients younger than 70 years but was not statistically significant in those aged 70 years or older.

What does this mean for clinical development?

Integrating multiple outcomes can provide an additional perspective when conventional efficacy and safety analyses leave an important treatment trade-off unresolved.

Potential applications include:

Comparing treatments when efficacy and safety outcomes point in different directions
Incorporating multiple clinically meaningful outcomes into a treatment comparison
Making the relative priority of outcomes explicit
Incorporating patient or clinician preferences into outcome prioritisation
Providing an additional quantitative analysis to support benefit-risk discussions

NTB does not replace conventional efficacy or safety analyses. Instead, it can complement them by providing an integrated measure of treatment effect across prioritized outcomes.

From separate outcomes to a more complete treatment comparison

The S-1 versus UFT analysis illustrates an important issue in clinical development: the treatment with the strongest result on a single efficacy endpoint may not necessarily have the same overall profile once other clinically relevant outcomes are considered.

Assessing multiple outcomes together can provide a different perspective on treatment benefit, particularly when efficacy, safety, tolerability or quality of life involve meaningful trade-offs.

Net Treatment Benefit provides a statistical framework for making these comparisons explicit.

Frequently Asked Questions

How can efficacy and safety outcomes be assessed together in a clinical trial?

Efficacy and safety outcomes can be assessed together by defining a hierarchy of clinically meaningful outcomes and analysing them jointly. Methods such as Generalized Pairwise Comparisons (GPC) and Net Treatment Benefit (NTB) allow patients to be compared across multiple outcomes according to their predefined priority.

How can treatment benefit be assessed when efficacy and safety outcomes compete?

When efficacy and safety outcomes point in different directions, each outcome can be important without providing a clear overall treatment comparison. A hierarchical analysis such as Net Treatment Benefit can evaluate these competing outcomes together, making the relative importance of efficacy, safety and other outcomes explicit.

What statistical methods can combine multiple clinical outcomes?

Several approaches can be used to analyse multiple clinical outcomes, depending on the study objectives and types of outcomes involved. Generalized Pairwise Comparisons can compare treatments across a hierarchy of outcomes, while Net Treatment Benefit provides a summary measure of the favourable and unfavourable comparisons.

Can patient preferences be used to prioritize clinical trial outcomes?

Yes. Patient preferences can help inform which outcomes are considered most important and how they should be prioritised in an analysis. This can provide evidence for defining an outcome hierarchy before applying methods such as Net Treatment Benefit.