Clinical Benefit Rate

Clinical Benefit Rate Definition, Calculation and Limits as an Input to Oncology Economic Models

The clinical benefit rate (CBR) is a tumour-response endpoint reported in cancer trials; in the trials cited here it was a secondary endpoint. It counts patients whose best overall response is a complete or partial response, together with patients whose disease stays stable for at least a minimum period fixed in the protocol, such as 24 weeks or 6 months, and divides by the analysis population. It appears in trials of abemaciclib and ribociclib for hormone receptor positive metastatic breast cancer, and its name can make it look like a measure of patient benefit in economic evaluation. It is built from RECIST 1.1 categories, it differs from the objective response rate and the disease control rate, and an economic model can take only limited information from it.

The RECIST 1.1 categories that CBR is built from

In the solid tumour trials cited here, the categories come from RECIST 1.1, published by Eisenhauer and colleagues. Target lesions are classified into one of four categories. At each assessment, target lesions, non-target lesions and any new lesions combine into an overall response, and the patient's best overall response is the best of these recorded from the start of treatment, taking account of any requirement for confirmation.

• Complete response (CR) means the disappearance of all target lesions, with any pathological lymph node reduced to a short axis below 10 mm.
• Partial response (PR) means at least a 30% decrease in the sum of target lesion diameters, compared with the baseline sum.
• Progressive disease (PD) means at least a 20% increase in the sum, compared with the smallest sum on study, plus an absolute increase of at least 5 mm, or the appearance of new lesions.
• Stable disease (SD) means neither enough shrinkage for PR nor enough growth for PD.

RECIST 1.1 attaches timing rules to stable disease that matter for CBR. Stable disease counts as best response only if the measurements met the SD criteria at least once after study entry, at a minimum interval set in the protocol and in general not less than 6 to 8 weeks. Its duration runs from the start of treatment, or from randomisation in a randomised trial, until progression. Where the proportion of patients with stable disease for a minimum period is an endpoint, the guideline asks the protocol to specify that interval. Confirmation of CR or PR at a later assessment, generally 4 weeks later, is required in non-randomised trials where response is the primary endpoint, and not in randomised trials.

The formula and the choices hidden inside it

Once each patient has a single best overall response, the rate is a simple proportion. The protocol and analysis plan still have to fix the minimum duration of stable disease, the denominator and the assessor, and each choice moves the result. The minimum duration is part of the definition, so the rate is best reported with it, as below.

$$CBR(d) = \frac{n_{CR} + n_{PR} + n_{SD \geq d}}{N}$$

where $n_{CR}$ and $n_{PR}$ are the numbers of patients whose best overall response is complete or partial response, $n_{SD \geq d}$ is the number whose best response is stable disease lasting at least the minimum duration $d$, and $N$ is the number of patients in the analysis population.

Published trials fill in $d$ differently. In the single-arm MONARCH 1 study of abemaciclib, Dickler and colleagues defined CBR as CR plus PR plus SD lasting at least 6 months. The FDA's multi-discipline review of ribociclib (Kisqali, NDA 209092, supplement 1) records that the MONALEESA-7 and MONALEESA-3 trials defined it as confirmed CR or PR, or stable disease lasting 24 weeks or longer, under RECIST 1.1. The same review reports the rate for all patients and separately for patients with measurable disease at baseline, which gave 79.1% and 79.9% in the ribociclib arm of MONALEESA-7. Patients with only non-measurable disease cannot have stable disease in the RECIST sense; the guideline prefers the label non-CR/non-PD for them, so a protocol needs to state whether that category counts.

CBR beside the objective response and disease control rates

Three proportions are built from the same best-response table and differ only in how much stable disease they admit. MONARCH 1 defined the disease control rate (DCR) as CR plus PR plus SD with no minimum duration, and the clinical benefit rate as the same sum restricted to SD lasting at least 6 months. Definitions vary between trials, so the labels alone do not guarantee comparability.

The FDA's 2018 guidance on cancer trial endpoints explains why stable disease is kept out of the objective response rate. Stable disease can reflect the natural history of the disease, whereas tumour reduction is a direct therapeutic effect, and stable disease can be assessed more accurately through time to progression or progression-free survival. CBR therefore mixes a component the FDA treats as a direct drug effect with one it regards as better measured by a time-to-event endpoint.

Recomputing the MONARCH 1 rates

MONARCH 1 enrolled 132 women with refractory hormone receptor positive, HER2-negative metastatic breast cancer, and its Table 2 reports best overall response under RECIST 1.1 by investigator and by independent review. The counts below come from that table; the arithmetic is recomputed here.

1. Investigator assessment. There were no complete responses, 26 partial responses, 63 patients with stable disease, of whom 30 had stable disease lasting at least 6 months, 34 with progressive disease and 9 not evaluable, which sums to 132. The three rates are as follows.

$$ORR = \frac{0 + 26}{132} = 19.7\%, \quad CBR = \frac{0 + 26 + 30}{132} = 42.4\%, \quad DCR = \frac{0 + 26 + 63}{132} = 67.4\%$$

where the numerators are the response counts listed above and 132 is the number of enrolled patients, the denominator the paper uses. Durable stable disease more than doubles the response rate, and the 33 patients with shorter stable disease add a further 25 percentage points to reach the DCR.

2. Independent review. Independent reviewers found 23 partial responses, 20 patients with stable disease for at least 6 months and 7 with non-CR/non-PD for at least 6 months. The reported CBR of 50 patients, 37.9%, is reached only if the 7 non-CR/non-PD patients are included, since 23 + 20 + 7 = 50, and 50 of 132 patients is 37.9%. The same trial therefore gives 42.4% or 37.9% depending on who reads the scans and how non-measurable disease is handled.

3. Precision. Exact binomial (Clopper-Pearson) 95% confidence intervals computed for 56/132 and 50/132 give 33.9% to 51.3% and 29.6% to 46.7%, matching the published intervals.

4. Interpretation. With no control arm, 42.4% cannot be split into patients whose disease was held back by the drug and patients whose disease was growing slowly anyway. The study's median progression-free survival was 6.0 months, so a 6-month threshold for stable disease falls close to the point by which half of the cohort had progressed or died.

Using CBR in an economic model

The NICE Decision Support Unit's Technical Support Document 19 (TSD 19) found that 22 of the 30 most recent NICE cancer appraisals it reviewed used a partitioned survival model. The usual structure has progression-free, progressed and dead states whose occupancy comes from the progression-free survival and overall survival curves. A single proportion does not describe how state occupancy changes over time, so it cannot replace either curve.

A rate like CBR could enter a response-based structure instead. TSD 19 reports a NICE appraisal in multiple myeloma (TA311) in which a decision tree allocated patients to response categories before partitioned survival analysis, with response used as a surrogate endpoint because the survival data were immature, and lists the assumption that response is the sole determinant of overall survival among the concerns raised. Any such use falls under the surrogate outcome sections of the NICE technology appraisal manual (PMG36, sections 4.6.5 to 4.6.11), which, for cost-utility analyses, regard end points reflecting how a patient feels, functions or how long a patient lives as more informative than surrogate outcomes. They rank evidence for a surrogate relationship from biological plausibility (level 3) through consistent association (level 2) to randomised trials in which the effect on the surrogate matches a commensurate effect on the final outcome (level 1). Validation is specific to the population and technology type, and the uncertainty in the relationship should be quantified and carried into the probabilistic sensitivity analysis.

Illustrative example of one trial with three modelled benefits

The figures are illustrative. In this example, a model attaches a gain of 0.04 QALYs to each patient classed as benefiting, taking the class from the MONARCH 1 investigator counts scaled to a cohort of 1,000.

$$\text{QALYs} = 1{,}000 \times \frac{56}{132} \times 0.04 = 424.2 \times 0.04 = 16.97$$

where 1,000 is the cohort, 56/132 is the clinical benefit rate and 0.04 is the assumed QALY gain per benefiting patient. Using the objective response rate instead gives 1,000 × 26/132 = 197.0 patients and 7.88 QALYs, while the disease control rate gives 674.2 patients and 26.97 QALYs. The modelled benefit more than triples between the narrowest and widest label with no new evidence about patients. Unless the gain is linked by evidence to the stable disease included, the choice of endpoint label rather than the data decides the result. A comparative model would apply the gain to the difference in rates between arms; the effect of the label choice is the same.

Limitations and common misreadings

The name suggests more than the endpoint measures. The FDA guidance treats survival improvement as an appropriate measure of clinical benefit in oncology and names improved quality of life, physical functioning and tumour-related symptoms as more direct evidence of clinical benefit. Mittal and colleagues call CBR a potentially misleading surrogate measure, arguing that it rests on the false premise that anything short of progression is a therapeutic benefit. In its review of ribociclib, the FDA called CBR an exploratory measure only (MONALEESA-3) and, for both trials, did not concur with describing the CBR differences as improvements.

• Patient-level association is not surrogacy. In individual patient data from 11 randomised first-line trials in metastatic breast cancer, Burzykowski and colleagues found tumour response and disease control strongly associated with overall survival, yet the correlation between treatment effects on these endpoints and on survival was low and too imprecisely estimated to be informative, and no endpoint could be shown to be a good surrogate.
• Comparing survival by CBR status is biased. Anderson, Cain and Gelber showed that the usual comparison of responders with non-responders is biased in favour of responders and should not be used to infer treatment effectiveness, and grouping patients by CBR status builds in the same kind of bias, because a patient counted through stable disease has already stayed free of progression for at least $d$.
• CBR is not the share progression-free at $d$. A patient with a partial response who progresses soon afterwards still counts, so CBR cannot stand in for a point on the progression-free survival curve.
• Cross-trial comparisons need matching definitions. The minimum duration, confirmation rule, assessor, denominator and scan schedule all differ between trials, and RECIST 1.1 notes that the durations of response and stable disease depend on the frequency of follow-up.
• Single-arm rates lack a counterfactual. Without a control arm, durable stable disease cannot be attributed to treatment, which matters in slowly growing disease.

Sources

• Anderson JR, Cain KC, Gelber RD. Analysis of survival by tumor response. Journal of Clinical Oncology. 1983;1(11):710-719. https://doi.org/10.1200/JCO.1983.1.11.710
• Burzykowski T, Buyse M, Piccart-Gebhart MJ, Sledge G, Carmichael J, Lück HJ, et al. Evaluation of tumor response, disease control, progression-free survival, and time to progression as potential surrogate end points in metastatic breast cancer. Journal of Clinical Oncology. 2008;26(12):1987-1992. https://doi.org/10.1200/JCO.2007.10.8407
• Dickler MN, Tolaney SM, Rugo HS, Cortés J, Diéras V, Patt D, et al. MONARCH 1, a phase II study of abemaciclib, a CDK4 and CDK6 inhibitor, as a single agent, in patients with refractory HR+/HER2- metastatic breast cancer. Clinical Cancer Research. 2017;23(17):5218-5224. https://doi.org/10.1158/1078-0432.CCR-17-0754
• Eisenhauer EA, Therasse P, Bogaerts J, Schwartz LH, Sargent D, Ford R, et al. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). European Journal of Cancer. 2009;45(2):228-247. https://doi.org/10.1016/j.ejca.2008.10.026
• Mittal A, Kim MS, Dunn S, Wright K, Gyawali B. Frequently asked questions on surrogate endpoints in oncology-opportunities, pitfalls, and the way forward. eClinicalMedicine. 2024;76:102824. https://doi.org/10.1016/j.eclinm.2024.102824
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). London: NICE; 2022, last updated 13 October 2026. Sections 4.6.5 to 4.6.11.
• US Food and Drug Administration. Clinical Trial Endpoints for the Approval of Cancer Drugs and Biologics: Guidance for Industry. FDA; December 2018. https://www.fda.gov/media/71195/download
• US Food and Drug Administration, Center for Drug Evaluation and Research. Approval package and multi-discipline review: Kisqali (ribociclib), application 209092Orig1s001. FDA; 2018. https://www.accessdata.fda.gov/drugsatfda_docs/nda/2018/209092Orig1s001.pdf
• Woods B, Sideris E, Palmer S, Latimer N, Soares M. NICE DSU Technical Support Document 19: partitioned survival analysis for decision modelling in health care: a critical review. Sheffield: Decision Support Unit, ScHARR, University of Sheffield; 2017. https://sheffield.ac.uk/nice-dsu/tsds/partitioned-survival-analysis