← Clinical Trials
Advanced / Metastatic Breast Cancer Phase 3 HER2-Low NCT04494425

DESTINY-Breast06: Complete Statistical Analysis of Trastuzumab Deruxtecan in Advanced or Metastatic Breast Cancer

An independent statistical analysis of the randomized phase 3 DESTINY-Breast06 trial comparing trastuzumab deruxtecan (T-DXd) with investigator's choice chemotherapy in a hormone receptor-positive, HER2-low population with advanced or metastatic breast cancer.

Trial start: 2020-07-24  ·  Primary completion: 2024-03-18  ·  Enrollment: 866
Scope of this record

This page separates reported trial results from statistical interpretation. All numerical trial results presented here are taken from the ClinicalTrials.gov record data. The registry provides one posted formal statistical analysis for the primary endpoint.

Registry note: This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.

1. Trial at a Glance

DESTINY-Breast06 is a randomized, parallel-group phase 3 trial evaluating trastuzumab deruxtecan versus investigator's choice chemotherapy in hormone receptor-positive, HER2-low advanced or metastatic breast cancer. The registry reports one primary time-to-event endpoint and one formal statistical analysis for that endpoint.

866
Enrollment
2 treatment arms
3
Phase
Phase 3
0.62
PFS HR
95% CI 0.52–0.75
<0.0001
P-value
Superiority analysis
FeatureDESTINY-Breast06
TrialDESTINY-Breast06
ClinicalTrials.gov identifierNCT04494425
PhasePhase 3
StatusActive, not recruiting
PopulationAdvanced or metastatic breast cancer; primary analysis in the hormone receptor-positive, HER2-low population
AllocationRandomized
Design modelParallel
MaskingNone
Primary purposeTreatment
Enrollment866
Arms2
Lead sponsorAstraZeneca
Sponsor typeIndustry
Primary endpointProgression-Free Survival (PFS) Assessed by Blinded Independent Central Review (BICR) in the Hormone Receptor-Positive (HR+), Human Epidermal Growth Factor Receptor 2 (HER2)-Low Population

2. Clinical Question

The primary statistical question is whether the time from randomization to progression or death differs between participants assigned to trastuzumab deruxtecan and those assigned to chemotherapy in the hormone receptor-positive, HER2-low population.

Population

Participants with advanced or metastatic breast cancer, with the posted primary analysis focused on the hormone receptor-positive, HER2-low population.

Intervention

Trastuzumab deruxtecan (T-DXd).

Comparator

Investigator's choice chemotherapy, represented in the registry by capecitabine, paclitaxel, and nab-paclitaxel.

Primary question

Does T-DXd improve progression-free survival relative to investigator's choice chemotherapy?

3. Trial Design

01
Randomize866 enrolled
02
Parallel arms2 treatment groups
03
TreatT-DXd or chemotherapy
04
AssessResponse evaluations q6w
05
AnalyzePFS by BICR
INTERVENTION

Trastuzumab deruxtecan

  • Trastuzumab deruxtecan (T-DXd)
COMPARATOR

Investigator's choice chemotherapy

  • Capecitabine
  • Paclitaxel
  • Nab-paclitaxel
Allocation
Randomized.
Structure
Parallel-group design with 2 arms.
Masking
None.
Primary purpose
Treatment.

Statistically, randomization is important because it establishes the treatment assignment mechanism before outcome information is observed. In a randomized comparison, the primary efficacy analysis can preserve the treatment groups created by that randomization rather than attempting to reconstruct treatment groups retrospectively from observed outcomes.

4. Endpoints

EndpointRegistry definition / assessmentAnalysis
Primary PFS endpoint Progression-Free Survival (PFS) Assessed by Blinded Independent Central Review (BICR) in the Hormone Receptor-Positive (HR+), Human Epidermal Growth Factor Receptor 2 (HER2)-Low Population. PFS per RECIST 1.1 assessed by BICR was defined as the time from the date of randomization until the date of PD, as defined or death (by any cause in the absence of progression), regardless of whether the participant withdrew from randomized therapy or received another anticancer therapy prior to progression. PD was defined as at least a 20% increase in the sum of diameters of target lesions (TLs), taking as reference the smallest previous sum of diameters (nadir), this included the baseline sum if that was the smallest on study. In addition to the relative increase of 20%, the sum must demonstrate an absolute increase of at least 5 millimeter (mm) from nadir. Median PFS was calculated using Kaplan-Meier method and its confidence interval (CI) using Brookmeyer-Crowley method. Stratified log-rank test; stratified Cox proportional-hazards model for HR and CI

The registry's registry-reported time frame is: “Response evaluations performed at screening, every 6 weeks (q6w) ± 1 week from randomization for 48 weeks, and then every 9 weeks (q9w) ± 1 week, starting at Week 48 until PD, up to PCD of 18 March 2024 (maximum of approximately 43.85 months).” The primary endpoint is therefore a time-to-event outcome rather than a simple proportion measured at one fixed visit.

Why the endpoint structure matters: PFS incorporates both the timing of progression and deaths before progression. Participants who have not experienced the defined event by the point at which their outcome information ends can contribute censored follow-up, which is why ordinary comparisons of proportions are not an adequate description of the primary analysis.

5. Analysis Population and Stratification

The posted primary analysis was conducted in the HER2-low population, described in the registry as the subset of participants in the ITT population with HER2 immunohistochemistry (IHC) 2+/in situ hybridization (ISH)- and IHC 1+ as determined per the interactive registry description.

Analysis componentRegistry-supported description
Underlying efficacy frameworkIntention-to-treat population
Primary analysis populationHER2-low subset of the ITT population
HER2 categories specifiedIHC 2+/ISH- and IHC 1+
Groups comparedT-DXd vs Chemotherapy
Stratified analysisYes
Covariates used in Cox modelPrior CDK4/6 inhibitor use and HER2 IHC expression

The registry specifically states that the Cox model adjusted for prior cyclin-dependent kinase (CDK)4/6 inhibitor use (yes versus no) and HER2 IHC expression (IHC 1+ versus IHC 2+/ISH-). This is important because the reported hazard ratio is not simply an unadjusted ratio of event rates; it comes from a stratified Cox proportional-hazards model incorporating these factors.

6. Primary Result: Progression-Free Survival

The registry contains one posted formal statistical analysis for the primary endpoint. The comparison was performed using a stratified log-rank test, with the hazard ratio and confidence interval calculated from a stratified Cox proportional-hazards model.

Progression-Free Survival

HR 0.62

95% CI: 0.52–0.75   ·   P < 0.0001

Comparison: T-DXd vs Chemotherapy

Primary endpointT-DXdChemotherapyEffect estimate
PFS by BICR in HR+, HER2-low population — — HR 0.62 (95% CI 0.52–0.75); P < 0.0001

The registry does not provide a median PFS or time-specific PFS percentages in the ClinicalTrials.gov record. Those quantities are therefore not added here. The formal result that is available is the hazard ratio, its two-sided 95% confidence interval, and the reported P-value.

Clinical Biostats interpretation

An HR of 0.62 means that, under the fitted stratified Cox model, the estimated instantaneous hazard of progression or death was about 38% lower with T-DXd than with chemotherapy over the analyzed follow-up. The calculation is the direct interpretation of 1 − 0.62 = 0.38.

The HR does not mean that 38% of participants avoided progression, that every participant had a 38% reduction in their individual probability of progression, or that median PFS was reduced or increased by a particular percentage. A hazard ratio is a relative time-to-event measure, not an absolute risk difference.

The 95% CI of 0.52–0.75 describes uncertainty around the estimated hazard ratio under the model and sampling framework. It does not describe the range of individual treatment effects among participants. The interval also provides information that a single point estimate cannot: it shows the statistical precision of the estimated relative effect.

The P-value < 0.0001 addresses the strength of evidence against the relevant null hypothesis under the prespecified statistical framework. It is not a measure of the size of the treatment effect. A very small P-value does not mean that the effect is proportionally large, just as a larger P-value would not by itself quantify clinical importance.

The HR and CI were produced using a stratified Cox proportional-hazards model. Consequently, interpretation of a single HR should be made with awareness of the proportional-hazards framework. If the relative hazards vary materially over time, one summary HR can compress a more complicated time-varying treatment effect into a single number.

The primary comparison used a stratified log-rank test, while the HR and CI used the corresponding stratified Cox model. The analysis therefore combines a hypothesis test of the time-to-event distributions with a model-based estimate of relative treatment effect.

7. How to Read the Primary Result

Hazard ratio
HR = 0.62  →  estimated hazard in T-DXd relative to chemotherapy

Because the HR is below 1, the estimated event hazard is lower in the T-DXd group under the fitted model. The complementary interpretation, 1 − 0.62 = 0.38, corresponds to an approximately 38% lower estimated hazard.

Relative measure

The hazard ratio describes a relative difference in the event hazard. It does not directly give an absolute difference in months of PFS or the absolute probability of remaining progression-free.

Precision

The 95% CI of 0.52–0.75 gives the statistical uncertainty reported around the estimated HR of 0.62.

Hypothesis test

The reported P-value is evidence from the specified statistical test; it is not an effect-size metric and should not be interpreted as the probability that the treatment works.

Time-to-event context

The analysis accounts for event timing and censoring rather than reducing the endpoint to whether an event occurred by one arbitrary calendar date.

8. Statistical Methodology

Stratified log-rank test

The registry reports a stratified log-rank test as the formal comparison method. The log-rank test is designed for comparing time-to-event distributions between treatment groups while using information across the observed follow-up period.

In contrast with a simple comparison of event proportions, the log-rank framework uses the ordering and timing of events. Participants who are censored can still contribute information to the comparison before their censoring time.

Conceptual comparison
Observed events versus expected events under the null hypothesis

The log-rank statistic accumulates treatment-group differences across event times. Stratification allows those comparisons to be made within defined strata and then combined according to the stratified test framework.

Stratified Cox proportional-hazards model

The hazard ratio and its confidence interval were calculated using a stratified Cox proportional-hazards model. The registry states that the model adjusted for prior CDK4/6 inhibitor use and HER2 IHC expression, with ties handled using the Efron approach.

Conceptual Cox model
h(t | X) = h0(t) exp(βX)

The treatment coefficient is transformed to obtain a hazard ratio. In a stratified Cox analysis, the baseline hazard can differ across strata while the treatment effect is estimated within the common modeling framework.

Covariate adjustment

Adjustment is particularly relevant when the analysis specifies prognostic or design-related factors that may influence the timing of progression or death. Here, the registry identifies prior CDK4/6 inhibitor use and HER2 IHC expression as adjustment factors.

The purpose of covariate adjustment is not to turn a randomized comparison into an observational analysis. Rather, prespecified adjustment can account for relevant variation while estimating the treatment effect. The key distinction is that treatment assignment remains anchored to the randomized comparison, while the Cox model incorporates the specified covariates when estimating the hazard ratio.

Intention-to-treat framework

The posted analysis population is described as a HER2-low subset of the ITT population. The ITT principle maintains treatment assignment as the basis for efficacy comparison, which is important because post-randomization treatment changes, discontinuation, and other events should not simply redefine the treatment groups after randomization.

Efron handling of tied event times

The registry specifies that ties were handled using the Efron approach. Ties occur when multiple participants have events recorded at the same observed time. Cox-model implementations need a rule for handling those tied event times; the Efron method is one such approach.

9. Statistical Methods Explained

Why was a log-rank test used?

PFS is a time-to-event endpoint, so the analysis needs to use both whether an event occurred and when it occurred. The log-rank test is designed for this setting and compares the treatment groups across the observed event-time experience rather than using only a single end-of-study proportion.

What does an HR of 0.62 mean?

An HR of 0.62 means that the estimated hazard of progression or death under the fitted model was 0.62 times the corresponding hazard in the chemotherapy group. Equivalently, 1 − 0.62 = 0.38, so the estimated hazard was approximately 38% lower with T-DXd. This is not the same as saying that the probability of progression or death was 38% lower for every individual participant.

Why is the confidence interval important?

The point estimate of 0.62 is only one estimate from the observed data. The 95% CI of 0.52–0.75 communicates the statistical uncertainty around that estimate. A confidence interval is therefore more informative than a P-value alone when assessing the precision of a treatment-effect estimate.

Why doesn't the P-value measure treatment benefit?

The reported P-value of <0.0001 is evidence against the null hypothesis under the specified test. It does not tell us that the treatment effect is “99.99% certain,” nor does it quantify the magnitude of benefit. Effect size is communicated by the hazard ratio, while uncertainty is communicated by the confidence interval.

Why was the Cox model stratified?

The registry reports a stratified Cox model and identifies prior CDK4/6 inhibitor use and HER2 IHC expression as adjustment factors. Stratification and covariate adjustment can account for important differences in baseline risk or design variables while estimating the treatment effect.

What does the BICR designation add?

The primary endpoint was assessed by blinded independent central review. For a radiologically determined endpoint such as PFS, independent central review provides a prespecified assessment framework intended to reduce the influence of knowledge of treatment assignment on progression determinations.

10. Censoring and the Time-to-Event Framework

PFS differs fundamentally from a binary endpoint such as “progressed by the end of the study.” Participants enter the risk set at randomization and are followed until progression, death, or the applicable end of observed event information. Participants without a qualifying event at the end of their observed follow-up contribute censored information.

Event timing

An event occurring early and an event occurring later are not treated as equivalent observations in a time-to-event analysis.

Censoring

A censored participant contributes information up to the censoring point rather than being treated as though the event never occurred.

Death

The registry definition includes death from any cause in the absence of progression as a PFS event.

Post-randomization therapy

The registry-reported endpoint definition specifies that PFS is assessed regardless of whether a participant withdrew from randomized therapy or received another anticancer therapy prior to progression.

This structure is one reason the hazard ratio cannot be translated directly into a statement such as “62% of patients remained progression-free.” The HR summarizes relative event hazard; it is not a survival probability.

11. Safety Results

The ClinicalTrials.gov record reports serious adverse events by treatment arm as affected participants over participants at risk:

Safety measureT-DXdChemotherapy
Serious adverse events, affected / at risk88 / 43467 / 417
How to read these safety data: the registry field reports counts as “affected/at risk.” These figures are presented exactly as reported in the registry rather than converting them into newly calculated percentages. Serious adverse events are also distinct from the primary PFS endpoint and should not be combined with the efficacy result into a single numerical measure.

The denominators in the registry-reported serious-adverse-event field are 434 for T-DXd and 417 for chemotherapy. The registry extract in the ClinicalTrials.gov record does not provide a broader adverse-event table, grade-specific event frequencies, or individual adverse-event categories, so those details are not added.

12. Trial Timeline

2020-07-24

Trial start

The registry lists 2020-07-24 as the study start date.

2024-03-18

Primary completion

The registry lists 2024-03-18 as the primary completion date.

Current registry status

Active, not recruiting

the ClinicalTrials.gov record lists the study status as ACTIVE_NOT_RECRUITING.

13. What the Primary PFS Result Does — and Does Not — Mean

What it means

The estimated hazard ratio of 0.62 indicates a lower estimated hazard of progression or death in the T-DXd group relative to chemotherapy under the fitted stratified Cox model.

What it does not mean

It does not mean that 62% of participants were progression-free, that 38% of participants were prevented from progressing, or that every participant experienced the same proportional reduction in risk.

What the confidence interval adds

The 0.52–0.75 interval indicates the statistical uncertainty around the HR estimate of 0.62. It is not an interval containing the PFS times of individual patients.

What the P-value adds

The reported P < 0.0001 quantifies evidence against the null hypothesis under the specified statistical test. It should not be used as a substitute for the hazard ratio or confidence interval when describing the magnitude and precision of the treatment effect.

14. Stratification and Covariate Adjustment: Why They Matter

The registry reports two related but conceptually distinct elements: a stratified log-rank test and a stratified Cox proportional-hazards model that adjusts for prior CDK4/6 inhibitor use and HER2 IHC expression.

ElementRole in the analysis
Stratified log-rank testFormal comparison of time-to-event experience between T-DXd and chemotherapy.
Stratified Cox modelEstimation of the hazard ratio and its confidence interval.
Prior CDK4/6 inhibitor useAdjustment factor: yes versus no.
HER2 IHC expressionAdjustment factor: IHC 1+ versus IHC 2+/ISH-.
Efron approachMethod used to handle tied event times in the Cox model.

For statistical interpretation, it is useful not to collapse all of these components into the phrase “the trial used a Cox model.” The hypothesis test and the effect-estimation model have different purposes. The log-rank test provides the formal comparison, while the Cox model supplies the reported relative effect estimate and its confidence interval.

15. Why a Single Hazard Ratio Is Not the Whole PFS Story

A hazard ratio is an efficient summary of a time-to-event comparison, but it does not contain every clinically or statistically relevant feature of the underlying survival distributions.

For example, a hazard ratio does not directly provide a median PFS, a PFS probability at a specified time point, or the absolute difference between treatment groups at a particular time. None of those quantities is contained in the ClinicalTrials.gov record, so they are not reconstructed here.

Hazard ratio

Summarizes the relative event hazard under the Cox model.

Confidence interval

Describes statistical uncertainty around the estimated hazard ratio.

Kaplan-Meier curve

Would describe estimated event-free probability over time if the underlying event and censoring data were available.

Absolute PFS measures

Would describe the event-free probability or time scale directly, but are not reported in the registry-reported primary-analysis data.

16. Multiplicity, Interim Analysis, and Other Design Features

The ClinicalTrials.gov record supports a primary superiority analysis, but they do not provide details about an alpha-spending plan, interim-analysis schedule, multiplicity hierarchy, or formal power calculation. Those elements are therefore not inferred from the P-value or reconstructed from the observed hazard ratio.

Design topicWhat the ClinicalTrials.gov record supports
Hypothesis typeSuperiority.
Primary endpointOne registered time-to-event endpoint: PFS by BICR in the HR+, HER2-low population.
Formal statistical analysis postedOne.
Interim analysisNo interim-analysis details are provided in the ClinicalTrials.gov record.
Multiplicity adjustmentNo multiplicity procedure is provided in the ClinicalTrials.gov record.
Non-inferiority marginNot applicable to the reported superiority hypothesis; no non-inferiority margin is provided.
CrossoverNo crossover information is provided in the ClinicalTrials.gov record.
Factorial designNo factorial design is reported; the design model is parallel.
Bayesian methodsNo Bayesian method is reported.
Missing-data / imputation methodNo specific imputation method is provided in the ClinicalTrials.gov record.

This distinction is important. Absence of a reported design feature in the ClinicalTrials.gov record is not evidence that a particular procedure was or was not used in an underlying protocol or statistical analysis plan. For this page, unsupported methods are simply not attributed to the trial.

17. Limitations

18. Why This Trial Matters Statistically

DESTINY-Breast06 provides a compact teaching example of how a modern randomized oncology trial connects clinical design to survival-analysis methodology. The trial uses randomization, a time-to-event primary endpoint, independent central review, a stratified log-rank test, a stratified Cox model, covariate adjustment, confidence intervals, and a prespecified superiority hypothesis.

ConceptHow it appears in DESTINY-Breast06
RandomizationThe allocation is randomized.
Parallel designThe registry specifies a parallel design with 2 arms.
Intention-to-treatThe primary analysis population is described as a HER2-low subset of the ITT population.
Time-to-event endpointPFS is the registered primary endpoint.
Blinded independent central reviewThe primary PFS endpoint is assessed by BICR.
Log-rank testA stratified log-rank test is the reported formal comparison.
Hazard ratioThe primary PFS effect measure is HR 0.62.
Confidence intervalThe 95% CI is 0.52–0.75.
Covariate adjustmentThe Cox model adjusts for prior CDK4/6 inhibitor use and HER2 IHC expression.
Stratified Cox modelUsed to calculate the HR and CI.
Efron methodUsed for tied event times in the Cox model.
Superiority testingThe registry identifies the hypothesis type as superiority.

The particularly useful statistical lesson is the separation between the endpoint, the hypothesis test, and the effect estimate. PFS defines what is being measured. The stratified log-rank test addresses the treatment-group comparison. The stratified Cox model provides the HR and CI used to describe the relative effect.

19. Primary Result in Statistical Context

The reported statistical signal

0.62

HR for PFS: 0.62  ·  95% CI 0.52–0.75  ·  P < 0.0001

Three pieces of information should be read together. First, the hazard ratio provides the estimated relative effect. Second, the confidence interval provides information about precision. Third, the P-value describes evidence against the null hypothesis under the specified statistical test.

None of these quantities, individually or collectively, supplies the complete clinical description of PFS. In particular, the ClinicalTrials.gov record does not contain a median PFS or absolute PFS rates, so those measures cannot be substituted into the interpretation.

Statistical caution: the phrase “38% lower risk” is commonly used as shorthand for an HR of 0.62, but “risk” can be ambiguous in ordinary language. More precisely, the registry-supported interpretation is that the fitted model estimates approximately a 38% lower hazard of progression or death for T-DXd relative to chemotherapy.

20. Clinical Interpretation vs Statistical Interpretation

Statistical interpretation

The primary PFS comparison produced an HR of 0.62 with a two-sided 95% CI of 0.52–0.75 and a reported P-value of <0.0001 using the specified stratified analysis framework.

Clinical interpretation

The statistical result indicates a difference in the time-to-progression-or-death experience between the randomized treatment groups. The magnitude of that difference should be interpreted alongside absolute PFS measures when those measures are available.

Keeping these interpretations separate is important. Statistical evidence addresses the compatibility of the observed data with the null hypothesis and quantifies the estimated relative effect. Clinical interpretation additionally requires context about absolute benefit, treatment burden, adverse events, patient characteristics, and the duration and pattern of benefit. The ClinicalTrials.gov record does not provide all of those quantities.

21. Serious Adverse Events and Statistical Interpretation

The serious-adverse-event data illustrate another important distinction in clinical-trial statistics: efficacy and safety are different outcome domains with different denominators and different interpretation frameworks.

ArmAffectedAt riskReported format
T-DXd8843488/434
Chemotherapy6741767/417

The safety figures should not be interpreted as a formal superiority or non-inferiority comparison because the ClinicalTrials.gov record does not provide a statistical analysis of serious adverse events. They are descriptive arm-level counts as reported in the ClinicalTrials.gov record.

Similarly, the serious-adverse-event counts should not be combined mathematically with the PFS HR to create a composite benefit-risk score. Doing so would impose a weighting and statistical structure that is not contained in the registry data.

22. Data Interpretation Checklist

QuestionAnswer for this trial
What was randomized?Participants were randomized in a 2-arm parallel treatment trial.
What was the primary endpoint?PFS assessed by BICR in the HR+, HER2-low population.
What kind of endpoint is PFS?Time-to-event.
What was the formal test?Stratified log-rank test.
What was the effect measure?Hazard ratio.
What was the estimate?0.62.
What was the 95% CI?0.52–0.75.
What was the P-value?<0.0001.
What was the hypothesis type?Superiority.
What model produced the HR?Stratified Cox proportional-hazards model.
Which covariates were specified?Prior CDK4/6 inhibitor use and HER2 IHC expression.
How were ties handled?Efron approach.

23. Related Tutorials

Learn more about the methods used in this trial:

24. Related Calculators

25. Sources

Continue through Clinical Biostats

Explore the statistical concepts behind randomized trials, survival analysis, confidence intervals, and clinical-trial effect measures.

26. Record Summary

DESTINY-Breast06 provides a clear example of a randomized phase 3 time-to-event analysis. The ClinicalTrials.gov record describes an 866-participant, 2-arm, parallel randomized trial comparing T-DXd with investigator's choice chemotherapy. Its registered primary endpoint is PFS assessed by BICR in the HR+, HER2-low population, with response evaluations performed at screening, every 6 weeks ± 1 week from randomization for 48 weeks, and then every 9 weeks (q9w) ± 1 week, starting at Week 48 until PD, up to PCD of 18 March 2024 (maximum of approximately 43.85 months).

The formal analysis uses a stratified log-rank test, while the hazard ratio and two-sided 95% confidence interval are estimated from a stratified Cox proportional-hazards model. The model adjusts for prior CDK4/6 inhibitor use and HER2 IHC expression, with ties handled by the Efron approach. The reported HR of 0.62, 95% CI 0.52–0.75, and P-value <0.0001 describe the primary statistical comparison.

The most important statistical lesson is that these numbers answer different questions. The HR describes relative event hazard, the confidence interval describes uncertainty around that estimate, and the P-value describes evidence against the null hypothesis under the specified testing framework. None of them, by itself, is a median PFS, an absolute PFS probability, or a measure of benefit for an individual patient.

Clinical Biostats methodology: A trial-results page should reconstruct the statistical story of the trial without extending the evidence beyond what is reported. For DESTINY-Breast06, that means emphasizing the randomized time-to-event framework, BICR assessment, stratified log-rank test, Cox hazard ratio, confidence interval, covariate adjustment, and the distinction between statistical evidence and absolute clinical measures.