This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the CANDOR ClinicalTrials.gov record.
1. Trial at a Glance
CANDOR was a randomized phase 3 parallel-group trial in patients with relapsed and/or refractory multiple myeloma. The study compared carfilzomib and dexamethasone alone with the same regimen combined with daratumumab, using progression-free survival as the registered primary endpoint.
| Feature | CANDOR |
|---|---|
| Trial name | CANDOR |
| Phase | Phase 3 |
| Condition | Relapsed Multiple Myeloma; Refractory Multiple Myeloma |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 466 |
| Arms | 2 |
| Primary endpoint | Progression-free Survival (PFS) as Assessed by the Independent Review Committee (PA DCO Only) |
| Primary endpoint type | Time-to-event |
| Hypothesis type | Superiority |
| Lead sponsor | Amgen |
| ClinicalTrials.gov | NCT03158688 |
2. Clinical Question
The central statistical question was whether adding daratumumab to carfilzomib and dexamethasone was associated with a different time-to-progression-or-death experience than carfilzomib and dexamethasone alone in patients with relapsed and/or refractory multiple myeloma.
Population
Patients with relapsed multiple myeloma or refractory multiple myeloma enrolled in the CANDOR phase 3 trial.
Intervention
Carfilzomib, dexamethasone and daratumumab, identified in the registry analysis as the KdD group.
Comparator
Carfilzomib and dexamethasone, identified as the Kd group.
Primary question
Does the KdD regimen improve progression-free survival relative to Kd in the intent-to-treat population?
3. Trial Design
Carfilzomib and dexamethasone
- Carfilzomib
- Dexamethasone
- Comparator group for the primary progression-free survival analysis
Carfilzomib, dexamethasone and daratumumab
- Carfilzomib
- Dexamethasone
- Daratumumab
- Experimental group for the primary progression-free survival analysis
The registry describes the allocation as randomized, the design model as parallel, and masking as none. The primary purpose was treatment. These design features establish a straightforward two-group randomized comparison, while the absence of masking is relevant when considering potential sources of bias in outcomes that involve clinical assessment.
4. Randomization, Stratification, and Analysis Population
The primary analysis was performed in the Intent to Treat Population. The registry analysis text identifies three stratification factors assessed at randomization: International Staging System stage at screening, prior proteasome inhibitor exposure, and number of prior lines of therapy.
| Feature | Registry-supported detail |
|---|---|
| Analysis population | Intent to Treat Population |
| ISS stage at screening | Stage 1 or 2 vs Stage 3 |
| Prior proteasome inhibitor exposure | Yes vs no |
| Number of prior lines of therapy | 1 vs ≥ 2 |
| Stratified methods | Log-rank test and Cox model for PFS; stratified Mantel-Haenszel methods for binary endpoints |
Stratification is important because the analysis does not simply compare all participants as though these baseline factors were irrelevant. Instead, the reported time-to-event comparison and several categorical analyses account for the randomization strata.
5. Endpoints
| Endpoint | Registry definition / time frame | Type |
|---|---|---|
| Progression-free Survival (PFS) as Assessed by the Independent Review Committee (PA DCO Only) | From randomization until the PA DCO date of 14 July 2019; the longest treatment duration as of the DCO was 102.3 weeks. PFS was defined as the time from randomization to the earlier of disease progression or death due to any cause. Participants were evaluated for disease response and progression according to the International Myeloma Working Group-Uniform Response Criteria (IMWG-URC) as assessed by an Independent Review Committee (IRC). | Time-to-event |
| Overall Response (OR) as Assessed by the Independent Review Committee (PA DCO Only) | From randomization until the PA DCO date of 14 July 2019; the longest treatment duration as of the DCO was 102.3 weeks. | Binary |
| Minimal Residual Disease Negative Complete Response Rate (MRD[-]CR) at 12 Months as Assessed by the Independent Review Committee | 12 Months (8- to 13-month window) | Binary |
| Overall Survival | Up to 58 months after the first participant was enrolled. | Time-to-event |
| Percentage of Participants Who Achieved Minimal Residual Disease Negative (MRD[-]) Status as Assessed by Next Generation Sequencing at 12 Months | 12 Months (8- to 13-month window) | Binary |
| Quality of Life Core Module (QLQ-C30) Global Health Status/Quality of Life Scores For Baseline Up to the First Follow-Up Visit After the Last Dose | Baseline (Day 1 pre-dose) up to 236.3 weeks (longest treatment duration as of the FA DCO) | Continuous |
The ClinicalTrials.gov record identifies the registered primary endpoint as progression-free survival. The other posted analyses are secondary endpoints. The page therefore separates the primary PFS result from the secondary efficacy and quality-of-life analyses rather than treating all reported endpoints as equally confirmatory.
6. Primary Result: Progression-Free Survival
The primary endpoint was analyzed in the intent-to-treat population using a stratified log-rank test. The treatment effect was summarized with a stratified Cox model hazard ratio comparing KdD with Kd.
Stratified hazard ratio for progression or death
95% CI: 0.464–0.854 · P = 0.0014
Analysis population: Intent to Treat Population
| Primary endpoint | Kd | KdD | Effect estimate | p-value |
|---|---|---|---|---|
| Progression-free survival | Carfilzomib + dexamethasone | Carfilzomib + dexamethasone + daratumumab | Stratified Cox HR 0.630 (95% CI 0.464–0.854) | 0.0014 |
The hazard ratio of 0.630 means that the estimated instantaneous hazard of progression or death for KdD relative to Kd was 0.630 under the reported stratified Cox model. Equivalently, 1 − 0.630 = 0.370, so the estimated hazard was approximately 37% lower for KdD than Kd under that model.
The hazard ratio does not mean that 37% of participants avoided progression or death, nor does it mean that every participant experienced exactly a 37% reduction in risk. It is a relative model-based measure of the event hazard over follow-up.
The 95% confidence interval of 0.464 to 0.854 describes statistical uncertainty around the estimated hazard ratio. It does not describe the range of individual patient outcomes. Because the interval remains below 1, the estimated direction of the treatment comparison is consistent with a lower hazard for KdD under the fitted model.
The p-value of 0.0014 addresses the strength of evidence against the null hypothesis specified for the statistical comparison. It is not a measure of how large or clinically important the treatment effect is. Effect size is better described by the hazard ratio and its confidence interval, together with absolute time-to-event information when available.
The registry reports a stratified Cox model and stratified log-rank analysis. As with any Cox-model hazard ratio, interpretation depends on the suitability of the proportional-hazards framework for summarizing the treatment contrast over time. The ClinicalTrials.gov record does not provide enough information to independently assess that assumption.
Why the stratification matters
The PFS comparison incorporated the randomization stratification factors: ISS stage at screening, prior proteasome inhibitor exposure, and number of prior lines of therapy. This makes the reported estimate a stratified treatment comparison rather than an unadjusted ratio derived from pooling all participants without regard to those factors.
The corresponding relative reduction in estimated hazard is 37%, obtained directly from 1 − 0.630. This is a statement about the modeled hazard, not a statement that 37% of participants experienced a particular outcome.
7. Secondary Result: Overall Response
Overall response was a binary secondary endpoint assessed by the Independent Review Committee. The intent-to-treat population was compared using a stratified Cochran-Mantel-Haenszel analysis, with the odds ratio estimated by the stratified Mantel-Haenszel method.
Odds ratio for overall response
95% CI: 1.184–3.129 · P = 0.0040
Odds ratio reported as KdD/Kd
An odds ratio of 1.925 means that the estimated odds of overall response were 1.925 times as high in KdD as in Kd under the reported stratified Mantel-Haenszel analysis.
An odds ratio is not the same as a risk ratio or a difference in response percentages. Without the underlying response counts or percentages, the odds ratio should not be converted into an absolute response-rate difference.
The 95% confidence interval of 1.184 to 3.129 describes uncertainty around the estimated odds ratio. The interval is entirely above 1, consistent with higher estimated odds of response for KdD under the reported analysis.
The p-value of 0.0040 provides evidence against the corresponding null hypothesis in the reported statistical framework. It does not measure the magnitude of the response benefit; the odds ratio and its confidence interval describe the estimated effect more directly.
8. Secondary Result: MRD-Negative Complete Response at 12 Months
The registry reports a secondary endpoint for minimal residual disease negative complete response at 12 months, using an 8- to 13-month assessment window. The analysis was performed in the intent-to-treat population with a stratified Cochran-Mantel-Haenszel approach.
Odds ratio for MRD-negative complete response
95% CI: 2.364–25.858
Odds ratio reported as KdD/Kd
An odds ratio of 7.819 indicates that the estimated odds of achieving MRD-negative complete response at the specified 12-month assessment were 7.819 times as high for KdD as for Kd in the reported stratified analysis.
This does not mean that 7.819 times as many participants achieved MRD-negative complete response, because an odds ratio operates on odds rather than probabilities or percentages.
The confidence interval, 2.364 to 25.858, is substantially wider than the point estimate. That width indicates considerable statistical uncertainty about the magnitude of the odds ratio, even though the interval remains above 1.
No response percentages are reported in the ClinicalTrials.gov record for this endpoint, so an absolute response-rate interpretation cannot be reconstructed without adding information outside the ClinicalTrials.gov recordset.
9. Secondary Result: Overall Survival
Overall survival was analyzed as a secondary time-to-event endpoint in the intent-to-treat population. The registry reports a stratified log-rank analysis and a Cox proportional-hazards effect measure.
Stratified hazard ratio for overall survival
95% CI: 0.595–1.033 · P = 0.0417
Time frame: up to 58 months after the first participant was enrolled
The reported hazard ratio of 0.784 indicates an estimated instantaneous hazard of death for KdD equal to 0.784 times that for Kd under the reported stratified Cox analysis. The direct arithmetic implication is an estimated hazard reduction of 21.6%, calculated as 1 − 0.784.
The hazard ratio is not an absolute survival difference and does not mean that 21.6% of participants survived because of treatment. It summarizes a relative event-hazard comparison.
The 95% confidence interval of 0.595 to 1.033 is relatively broad and includes 1. This means that the interval includes hazard-ratio values representing no difference as well as values below 1. The point estimate alone therefore does not describe the full uncertainty of the treatment comparison.
The reported p-value of 0.0417 is a test statistic under the specified analysis framework; it is not a probability that the treatment effect is real, nor is it a measure of effect size.
The registry notes that the p-value came from a 1-sided log-rank test controlling for the randomization stratification factors, while the hazard ratio and 95% confidence interval came from stratified Cox proportional-hazards models. The distinction between the one-sided test and the two-sided confidence interval should be preserved when interpreting the reported statistics.
10. Secondary Result: MRD-Negative Status at 12 Months
A second MRD endpoint evaluated the percentage of participants achieving minimal residual disease negative status by next generation sequencing at 12 months, using an 8- to 13-month window.
Odds ratio for MRD-negative status
95% CI: 2.007–9.656
Odds ratio reported as KdD/Kd
The reported odds ratio of 4.403 indicates that the estimated odds of achieving MRD-negative status at 12 months were 4.403 times as high in KdD as in Kd under the stratified Mantel-Haenszel analysis.
As with the other odds-ratio analyses, this is not a four-fold increase in probability. Odds and probabilities are mathematically related but are not interchangeable measures.
The 95% confidence interval of 2.007 to 9.656 indicates uncertainty around the magnitude of the estimated association. Because the interval is entirely above 1, the reported estimate is directionally consistent with higher odds of MRD-negative status in KdD.
11. Secondary Result: Quality of Life
The registry reports a continuous quality-of-life endpoint based on the QLQ-C30 Global Health Status/Quality of Life score from baseline through the first follow-up visit after the last dose. The analysis used a linear mixed-effects model in the intent-to-treat population.
Mean difference in final values
95% CI: -2.52 to 2.35 · P = 0.9480
Baseline (Day 1 pre-dose) up to 236.3 weeks
The estimated final-value mean difference was -0.08, comparing KdD with Kd under the reported mixed-effects analysis. The estimate is very close to zero relative to the scale of the reported uncertainty.
The 95% confidence interval of -2.52 to 2.35 includes zero and spans modest differences in either direction. The interval therefore indicates that the data are compatible with a range of small treatment-group differences in the modeled outcome.
The p-value of 0.9480 is not an effect-size measure. A p-value close to 1 does not prove that the groups are identical; it indicates that the observed data provide little evidence against the null comparison under the specified model.
The mixed-effects approach is appropriate to the longitudinal nature of the endpoint because repeated measurements from the same participant are statistically related. The ClinicalTrials.gov record identifies the model and several covariates but does not provide enough detail to reconstruct the complete covariance structure or all model parameters.
12. Statistical Methodology
Log-rank test for time-to-event endpoints
The registry reports the log-rank test for the primary PFS analysis and for overall survival. A log-rank test compares the observed and expected event experience between randomized groups across follow-up while accommodating right-censored observations.
For CANDOR, the reported PFS analysis was stratified according to the factors assessed at randomization. The overall-survival analysis was likewise described as controlling for the randomization stratification factors.
Stratified Cox proportional-hazards model
The primary PFS treatment effect was reported as a stratified Cox model hazard ratio. The overall-survival analysis also used a Cox proportional-hazards effect measure. A Cox model estimates a relative hazard while allowing the underlying baseline hazard to remain unspecified.
For the primary PFS analysis, HR = 0.630. Values below 1 correspond to a lower estimated instantaneous event hazard for KdD under the fitted model.
Cochran-Mantel-Haenszel analysis
Binary endpoints including overall response and the MRD-related outcomes were analyzed using stratified Mantel-Haenszel methods. The overall-response analysis specifically reports a Cochran-Mantel-Haenszel Chi-Square test for the p-value and a stratified Mantel-Haenszel odds ratio for the effect estimate.
Mixed-effects model for longitudinal quality of life
The quality-of-life endpoint was analyzed using a linear mixed-effects model. This class of model is designed for repeated observations because measurements from the same participant are not independent. The registry-reported analysis notes identify treatment, baseline QLQ-C30 GHS/QoL score, randomization stratification factors, and interactions among model terms as components of the analysis.
Intent-to-treat analysis
The posted efficacy analyses used the intent-to-treat population. The defining principle is that participants remain associated with their randomized treatment group for efficacy analysis. This helps maintain the comparability established by randomization.
13. Statistical Methods Explained
Why was a log-rank test used for PFS?
Progression-free survival is a time-to-event endpoint. Participants can experience progression or death at different times, while others may remain event-free when follow-up ends. The log-rank test is designed for comparing such event-time distributions while accounting for censoring.
What does a hazard ratio of 0.630 mean?
It means that the estimated instantaneous hazard of progression or death under KdD was 0.630 times the corresponding estimated hazard under Kd in the reported stratified Cox model. The complementary calculation, 1 − 0.630, gives a 37% relative reduction in the estimated hazard. It does not mean a 37% reduction in every participant's individual probability of progression or death.
Why use a stratified analysis?
The trial incorporated ISS stage, prior proteasome inhibitor exposure, and number of prior lines of therapy into randomization stratification. The reported stratified analyses preserve those design factors when estimating and testing the treatment comparison.
Why use an odds ratio for overall response?
Overall response is binary: a participant either meets the response definition or does not. An odds ratio compares the odds of response between treatment groups. It is not the same as a risk ratio, and it should not be interpreted as a percentage-point difference in response.
Why is the confidence interval important?
A point estimate alone does not show how precisely the treatment effect has been estimated. The confidence interval provides a range of values compatible with the statistical estimation framework. For example, the primary PFS hazard ratio is 0.630 with a 95% CI of 0.464–0.854, while the MRD-negative complete response odds ratio is 7.819 with a much wider 95% CI of 2.364–25.858.
Why was a mixed-effects model used for quality of life?
The quality-of-life endpoint contains measurements over time. Repeated observations from the same participant tend to be correlated, so treating every observation as independent would generally be inappropriate. A mixed-effects model provides a framework for modeling longitudinal observations while accounting for participant-level repeated measurement structure.
14. Understanding the Primary PFS Result
What the estimate says
The reported stratified Cox HR of 0.630 indicates a lower estimated hazard of progression or death for KdD relative to Kd.
What the estimate does not say
It does not state the proportion of participants who benefited, the absolute difference in survival time, or the probability that an individual participant will benefit.
What the CI says
The 95% CI of 0.464–0.854 quantifies uncertainty around the estimated hazard ratio under the reported model.
What the p-value says
The p-value of 0.0014 addresses evidence against the null hypothesis in the specified test framework; it does not quantify effect magnitude.
This distinction between effect size, precision, and hypothesis testing is central to interpreting randomized trial results. The hazard ratio answers a relative treatment-effect question, the confidence interval describes uncertainty around the estimate, and the p-value addresses the statistical test.
15. Secondary Endpoints as a Statistical Pattern
| Endpoint | Analysis | Effect measure | Estimate | 95% CI | p-value |
|---|---|---|---|---|---|
| Overall Response | Cochran-Mantel-Haenszel | Odds ratio | 1.925 | 1.184–3.129 | 0.0040 |
| MRD[-]CR at 12 Months | Cochran-Mantel-Haenszel | Odds ratio | 7.819 | 2.364–25.858 | Not reported in the ClinicalTrials.gov record |
| Overall Survival | Log-rank / stratified Cox | Hazard ratio | 0.784 | 0.595–1.033 | 0.0417 |
| MRD[-] status at 12 Months | Cochran-Mantel-Haenszel | Odds ratio | 4.403 | 2.007–9.656 | Not reported in the ClinicalTrials.gov record |
| QLQ-C30 Global Health Status/Quality of Life | Linear mixed-effects model | Mean difference | -0.08 | -2.52 to 2.35 | 0.9480 |
The pattern illustrates why a clinical trial cannot be reduced to a single p-value. CANDOR reports time-to-event, binary, and longitudinal continuous endpoints, and each requires a different statistical framework. The interpretation of an HR, odds ratio, and mean difference must therefore remain tied to the scale of the endpoint being analyzed.
16. Serious Adverse Events
The ClinicalTrials.gov record reports serious adverse events by arm as affected participants over participants at risk. These figures should be treated as counts and denominators rather than converted into a new percentage unless the conversion is explicitly needed and allowed by the data rules.
| Arm | Affected / at risk |
|---|---|
| Kd - Carfilzomib and Dexamethasone | 80 / 153 |
| KdD - Carfilzomib, Dexamethasone and Daratumumab | 211 / 308 |
The denominators in this safety summary are not the same as the overall enrollment of 466. That distinction is important: a safety endpoint may use an exposure-defined or otherwise restricted population rather than every participant enrolled in the trial.
17. Trial Timeline
Trial start
The CANDOR study began enrollment on 13 June 2017.
Primary analysis cutoff
The registered primary PFS endpoint used the PA DCO date of 14 July 2019. The longest treatment duration as of that data cutoff was 102.3 weeks.
Trial status
The ClinicalTrials.gov record identifies the study status as COMPLETED.
18. What the Different Effect Measures Mean
| Effect measure | CANDOR endpoint | Interpretation |
|---|---|---|
| Hazard ratio | PFS; Overall Survival | Relative comparison of estimated event hazards over time under a Cox-model framework. |
| Odds ratio | Overall Response; MRD endpoints | Relative comparison of the odds of a binary outcome between treatment groups. |
| Mean difference | QLQ-C30 Global Health Status/Quality of Life | Difference between modeled mean outcome values on the reported score scale. |
These measures should not be compared numerically as though they were interchangeable. An odds ratio of 1.925 and a hazard ratio of 0.630 describe fundamentally different statistical quantities because their endpoints and models are different.
19. Why the Confidence Intervals Matter
Confidence intervals provide essential context for every reported treatment estimate. The point estimate is only one value supported by the observed data and model. The interval communicates the uncertainty surrounding that estimate.
PFS
HR 0.630 with 95% CI 0.464–0.854. The interval is entirely below 1, while the point estimate represents a 37% lower estimated hazard under the reported model.
Overall response
OR 1.925 with 95% CI 1.184–3.129. The interval is entirely above 1, consistent with higher estimated odds of response for KdD.
Overall survival
HR 0.784 with 95% CI 0.595–1.033. The interval includes 1 and therefore communicates greater uncertainty about the magnitude of the relative hazard difference.
Quality of life
Mean difference -0.08 with 95% CI -2.52 to 2.35. The interval spans zero and covers differences in either direction.
The width of an interval is also informative. The MRD-negative complete response odds ratio has a point estimate of 7.819 but a 95% CI extending from 2.364 to 25.858. The large range indicates that the magnitude of the estimated odds ratio is substantially less precise than the point estimate alone might suggest.
20. Limitations
- Registry-level information: this analysis is restricted to the ClinicalTrials.gov record. It does not add publication-level results or statistical details that are absent from the ClinicalTrials.gov record.
- Incomplete absolute endpoint reporting: several binary analyses provide odds ratios and confidence intervals but not the corresponding treatment-group response percentages or counts in the ClinicalTrials.gov record.
- Hazard-ratio interpretation: a single Cox hazard ratio is a model-based summary. Its interpretation depends on the adequacy of the proportional-hazards framework for the underlying event processes.
- One-sided versus two-sided inference: the overall-survival analysis reports a one-sided log-rank p-value while its confidence interval is two-sided. These should be interpreted according to their stated conventions.
- Unreported p-values: the ClinicalTrials.gov record does not provide p-values for every secondary endpoint with a posted estimate and confidence interval.
- Safety denominators: serious adverse events are reported using affected/at-risk counts of 80/153 and 211/308 rather than the full enrollment of 466.
- Unmasked design: the registry identifies masking as none. Outcomes involving assessment or participant-reported information can therefore have different potential sources of bias than outcomes that are less susceptible to knowledge of treatment assignment.
- No independent reconstruction: the ClinicalTrials.gov record does not include participant-level event and censoring data, so a Kaplan-Meier curve or reconstructed median survival cannot be generated without adding external information.
21. Why This Trial Matters Statistically
CANDOR is a useful statistical teaching case because it combines randomized treatment comparison with several distinct endpoint types. The primary endpoint is time-to-event and uses stratified survival methods; response and MRD endpoints are binary and use stratified categorical methods; and quality of life is longitudinal and uses a mixed-effects model.
| Concept | How it appears in CANDOR |
|---|---|
| Randomization | Randomized phase 3 parallel-group comparison. |
| Intention-to-treat analysis | Primary and secondary efficacy analyses use the intent-to-treat population. |
| Time-to-event endpoint | Progression-free survival is the registered primary endpoint. |
| Log-rank test | Used for the primary PFS comparison and overall survival. |
| Hazard ratio | Used for PFS and overall survival treatment-effect estimation. |
| Stratified analysis | Stratification accounts for ISS stage, prior proteasome inhibitor exposure, and number of prior lines of therapy. |
| Cochran-Mantel-Haenszel test | Used for the overall-response analysis and reflected in the MRD analyses. |
| Odds ratio | Used for binary response and MRD endpoints. |
| Mixed-effects model | Used for longitudinal QLQ-C30 Global Health Status/Quality of Life scores. |
| Confidence intervals | Reported around hazard ratios, odds ratios, and the quality-of-life mean difference. |
The trial therefore illustrates an important principle of clinical biostatistics: the endpoint determines the appropriate statistical model. A time-to-event endpoint should not be interpreted with the same effect measure used for a binary response endpoint, and a repeated quality-of-life measure requires attention to within-participant correlation.
22. Related Tutorials
Learn more about the methods used in this trial:
23. Related Calculators
24. Sources
- ClinicalTrials.gov: CANDOR — NCT03158688. The official trial registry record and source for the trial data used on this page.
Continue through the Clinical Biostats statistical pathway
Use the trial's endpoint types to explore the underlying statistical methods and related calculation tools.
25. Record Summary
CANDOR provides a compact example of how several statistical frameworks operate within a single randomized clinical trial. The registered primary endpoint was progression-free survival, analyzed in the intent-to-treat population with a stratified log-rank test and a stratified Cox model. The reported hazard ratio was 0.630 with a 95% confidence interval of 0.464–0.854 and a p-value of 0.0014.
The secondary analyses demonstrate how the statistical method changes with the endpoint. Overall response and MRD outcomes were evaluated with stratified categorical methods and odds ratios, while overall survival used time-to-event methodology and quality of life used a linear mixed-effects model. These analyses should therefore be interpreted on their own statistical scales rather than collapsed into a single measure of treatment effect.
The ClinicalTrials.gov record also illustrate why denominators matter. Serious adverse events were reported as 80/153 in the Kd group and 211/308 in the KdD group. Those figures use arm-specific at-risk denominators rather than the full enrollment of 466 and should be interpreted accordingly.