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
COMET-2 was a randomized, parallel-group, quadruple-masked phase 3 treatment trial enrolling men with previously treated symptomatic castration-resistant prostate cancer. The registry reports 119 randomized participants across two treatment arms and a primary endpoint based on confirmed pain response at Week 6 and Week 12.
| Feature | COMET-2 |
|---|---|
| Phase | Phase 3 |
| Population | Men with previously treated symptomatic castration-resistant prostate cancer |
| Conditions | Prostate Cancer; Castration Resistant Prostate Cancer; Pain; Prostatic Neoplasms |
| Design | Randomized, parallel-group, quadruple-masked treatment trial |
| Allocation | Randomized |
| Enrollment | 119 |
| Number of arms | 2 |
| Primary endpoint | Pain Response at Week 6 Confirmed at Week 12, Week 12 Reported |
| Primary analysis population | ITT population of 119 participants: 61 cabozantinib and 58 mitoxantrone plus prednisone |
| Status | Terminated |
| Lead sponsor | Exelixis |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT01522443 |
2. Clinical Question
The central statistical question was whether cabozantinib produced a different proportion of confirmed pain responders than mitoxantrone plus prednisone in men with previously treated symptomatic castration-resistant prostate cancer.
Population
Men with previously treated symptomatic castration-resistant prostate cancer.
Intervention
Cabozantinib.
Comparator
Mitoxantrone plus prednisone.
Primary question
Does cabozantinib improve the prespecified confirmed pain-response rate relative to mitoxantrone plus prednisone?
3. Trial Design
Cabozantinib
- Cabozantinib was the intervention assigned to this randomized group.
- The group was included in the ITT primary pain-response analysis.
- The group was also included in the ITT overall-survival and bone-scan-response analyses.
Mitoxantrone plus prednisone
- Mitoxantrone plus prednisone was the comparator regimen.
- The group was included in the ITT primary pain-response analysis.
- The group was also included in the ITT overall-survival and bone-scan-response analyses.
4. Endpoints
| Endpoint | Registry definition / time frame | Statistical analysis |
|---|---|---|
| Primary: Pain Response at Week 6 Confirmed at Week 12 | Pain response was measured at Week 6 and Week 12 by self-reports of subjects. The prespecified primary analysis was defined as ≥ 30% from baseline in the average daily worst pain intensity score during a 7-day reporting period, with neither a concomitant increase in average daily use of any opioid narcotic type nor addition of any new opioid narcotic type relative to baseline. | Cochran-Mantel-Haenszel; superiority hypothesis |
| Pain Progression | At a given time point, defined as ≥ 30% increase compared with baseline in the average daily worst pain intensity score during a 7-day reporting period or either an increase in average daily use of any type of opioid narcotic or addition of a new opioid narcotic type compared with baseline. | Part of the registry definition of the pain endpoint. |
| Secondary: Overall Survival | Measured at the time of randomization until 78 deaths. | Log-rank; hazard ratio; superiority hypothesis |
| Secondary: Bone Scan Response | Measured at the end of Week 12 as determined by the IRF. | Cochran-Mantel-Haenszel; superiority hypothesis |
The endpoint structure is statistically heterogeneous. Pain response and bone scan response are binary responder outcomes, whereas overall survival is a time-to-event endpoint. Consequently, the appropriate treatment-effect summaries differ across endpoints: CMH methods compare response proportions while the log-rank framework compares time-to-event experience over follow-up.
5. Statistical Methodology
Intention-to-treat analysis
The primary pain-response analysis was based on the ITT population of all 119 randomized participants: 61 assigned to cabozantinib and 58 assigned to mitoxantrone plus prednisone. The same randomized ITT population was used for the reported overall-survival and bone-scan-response analyses.
The ITT principle preserves the treatment comparison created by randomization. It avoids redefining the primary comparison based on treatment received, adherence, or later clinical events.
Cochran-Mantel-Haenszel analysis
The registry reports the Cochran-Mantel-Haenszel method for the primary pain-response comparison and for bone scan response. For the pain-response analysis, the CMH analysis was stratified by Eastern Cooperative Oncology Group performance status and prior cabazitaxel use. The bone scan response analysis was likewise stratified by baseline Eastern Cooperative Oncology Group performance status and prior cabazitaxel use.
Instead of treating every participant as belonging to one undifferentiated population, a stratified CMH analysis accounts for the specified stratification factors while estimating the treatment comparison across those strata.
Log-rank analysis
Overall survival was analyzed using the log-rank test. The registry states that the log-rank test was stratified by baseline Eastern Cooperative Oncology Group performance status and prior cabazitaxel use.
The log-rank test is designed for comparing time-to-event distributions between randomized groups while incorporating the timing of observed deaths and the presence of right-censored observations. It therefore uses more information than simply comparing the proportion alive at one fixed time point.
Hazard ratio
The registry reports a hazard ratio as the effect measure for overall survival. A hazard ratio below 1 represents a lower estimated instantaneous rate of death in the cabozantinib group relative to the mitoxantrone-plus-prednisone group under the fitted time-to-event comparison.
A hazard ratio is a relative time-to-event measure. It is not a probability, not a percentage of patients responding, and not a direct statement that every patient experiences the same proportional reduction in risk.
6. Primary Result: Pain Response
The primary endpoint was Pain Response at Week 6 Confirmed at Week 12, with pain response measured at Week 6 and Week 12 through subject self-reports. The analysis population was the randomized ITT population of 119 participants.
Cochran-Mantel-Haenszel test
Cabozantinib vs Mitoxantrone/Prednisone
Superiority hypothesis; CMH stratified by Eastern Cooperative Oncology Group performance status and prior cabazitaxel use.
| Primary endpoint | Analysis population | Method | Hypothesis | P-value |
|---|---|---|---|---|
| Pain Response at Week 6 Confirmed at Week 12 | ITT: 119 participants; 61 cabozantinib, 58 mitoxantrone plus prednisone | Cochran-Mantel-Haenszel, stratified by Eastern Cooperative Oncology Group performance status and prior cabazitaxel use | Superiority | 0.773 |
The reported P-value of 0.773 is the statistical evidence from the prespecified CMH comparison of pain-response proportions. Under the superiority framework, this result does not provide statistical evidence of a difference in the primary pain-response endpoint at the conventional levels commonly used for confirmatory testing.
The P-value is not an effect-size measure. It does not say that the treatment effect is "77.3%" or that the two treatments are equivalent. A P-value quantifies how compatible the observed data are with the null hypothesis used for the statistical test; it does not measure the magnitude or clinical importance of an effect.
The registry does not report a confidence interval or the treatment-specific percentage of pain responders for this posted analysis. Consequently, the statistical interpretation is centered on the reported CMH P-value rather than on a reported absolute or relative response difference.
The analysis was stratified by Eastern Cooperative Oncology Group performance status and prior cabazitaxel use. Stratification is important because these factors were incorporated into the CMH comparison rather than ignored in the primary analysis.
7. Secondary Result: Overall Survival
Overall survival was measured from randomization until 78 deaths. The analysis used the ITT population of 119 randomized participants: 61 in the cabozantinib group and 58 in the mitoxantrone-plus-prednisone group. The registry identifies 06 October 2014 as the data cutoff date for this primary analysis.
Hazard ratio for death
95% CI: 0.44–1.10 · P = 0.121
Log-rank analysis stratified by baseline Eastern Cooperative Oncology Group performance status and prior cabazitaxel use.
| Endpoint | Analysis population | Method | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|---|---|
| Overall Survival | ITT: 119 randomized participants; 61 cabozantinib, 58 mitoxantrone plus prednisone | Stratified Log Rank | Hazard Ratio | 0.70 | 0.44–1.10 | 0.121 |
The reported HR of 0.70 means that the estimated instantaneous rate of death in the cabozantinib group was 0.70 times that in the mitoxantrone-plus-prednisone group under the reported time-to-event analysis. Expressed as a simple relative interpretation, this corresponds to an estimated 30% lower hazard of death.
The HR does not mean that 30% of patients benefited, that survival time increased by 30%, or that each individual patient's probability of death was reduced by exactly 30%. Hazard is a time-dependent event-rate concept, not a direct probability.
The 95% confidence interval of 0.44–1.10 indicates substantial uncertainty around the estimated hazard ratio. Because the interval includes 1, the reported interval is compatible with no relative difference in hazard as well as with hazard ratios below 1.
The P-value of 0.121 does not measure the size of the HR. It addresses the statistical evidence under the superiority hypothesis. The magnitude of the estimated effect and its precision must therefore be read from the HR and confidence interval rather than from the P-value alone.
Overall survival was analyzed using a stratified log-rank approach, with stratification by baseline Eastern Cooperative Oncology Group performance status and prior cabazitaxel use. The registry reports the endpoint in terms of deaths rather than a fixed calendar-time survival proportion.
8. Secondary Result: Bone Scan Response
Bone Scan Response was measured at the end of Week 12 as determined by the IRF. The analysis was conducted on the randomized ITT population of 119 participants: 61 cabozantinib and 58 mitoxantrone plus prednisone.
Cochran-Mantel-Haenszel test
Cabozantinib vs Mitoxantrone/Prednisone
Superiority hypothesis; CMH stratified by baseline Eastern Cooperative Oncology Group performance status and prior cabazitaxel use.
| Endpoint | Analysis population | Method | Hypothesis | P-value |
|---|---|---|---|---|
| Bone Scan Response at Week 12 | ITT: 119 participants; 61 cabozantinib, 58 mitoxantrone plus prednisone | Cochran-Mantel-Haenszel, stratified by baseline Eastern Cooperative Oncology Group performance status and prior cabazitaxel use | Superiority | <0.001 |
The reported P-value of <0.001 indicates strong statistical evidence against the null hypothesis used for the superiority comparison of bone scan response at Week 12.
This P-value does not quantify how large the difference in bone scan response was. The registry entry reports the statistical test and P-value but does not provide a treatment-specific responder percentage or confidence interval in the posted statistical analysis.
The absence of a reported effect estimate means the magnitude and precision of the bone scan response difference cannot be characterized from this posted analysis using an absolute risk difference, relative risk, odds ratio, or confidence interval.
The analysis was stratified by baseline Eastern Cooperative Oncology Group performance status and prior cabazitaxel use, which aligns the statistical comparison with the reported stratification structure.
9. Comparing the Three Reported Analyses
| Endpoint | Role | Type | Method | Effect measure reported | P-value |
|---|---|---|---|---|---|
| Pain Response at Week 6 Confirmed at Week 12 | Primary | Binary responder endpoint | Cochran-Mantel-Haenszel | Percentage of responders; treatment-specific percentages not reported in the posted analysis | 0.773 |
| Overall Survival | Secondary | Time-to-event endpoint | Log Rank | HR 0.70; 95% CI 0.44–1.10 | 0.121 |
| Bone Scan Response at Week 12 | Secondary | Binary responder endpoint | Cochran-Mantel-Haenszel | Percentage of responders; treatment-specific percentages not reported in the posted analysis | <0.001 |
These results illustrate why a clinical trial cannot be summarized adequately by a single P-value. The primary pain-response analysis and the secondary bone-scan-response analysis used CMH methods because the outcomes are responder proportions, whereas overall survival required a time-to-event method. Moreover, the reported survival analysis provides an HR and confidence interval, giving substantially more information about effect magnitude and precision than a P-value alone.
10. Statistical Methods Explained
Why was a Cochran-Mantel-Haenszel test used for pain response?
Pain response is expressed as a percentage of responders, making it a categorical outcome rather than a continuous measurement or time-to-event outcome. The CMH method allows the treatment comparison to be stratified by Eastern Cooperative Oncology Group performance status and prior cabazitaxel use. This preserves the specified stratified structure while testing the treatment comparison across strata.
What does the primary P-value of 0.773 mean?
A P-value of 0.773 indicates that the observed pain-response data do not provide strong statistical evidence against the null hypothesis used in the superiority test. It does not establish that the treatments are identical, nor does it quantify the size of any possible difference.
Why was overall survival analyzed differently?
Overall survival is defined through the timing of death, with follow-up potentially ending before death for some participants. This creates a time-to-event analysis problem. The log-rank method is designed for comparing survival distributions while incorporating the timing of events and censoring.
What does an HR of 0.70 mean?
An HR of 0.70 represents a relative estimated hazard of 0.70 for death in the cabozantinib group compared with the mitoxantrone-plus-prednisone group. A simple interpretation is a 30% lower estimated hazard. It is not equivalent to a 30% lower absolute probability of death or a 30% longer survival time.
Why is the 95% confidence interval important?
The 95% CI of 0.44–1.10 communicates the precision of the reported HR estimate. It shows that the data are compatible with a range of possible relative hazard values, including values below and above 1. A point estimate alone would conceal this uncertainty.
Why should the survival P-value and HR be read together?
The HR describes the estimated relative treatment effect, while the P-value addresses statistical evidence under the specified hypothesis. Neither replaces the other. The HR of 0.70 and its 95% CI of 0.44–1.10 describe the magnitude and uncertainty of the survival comparison, while P = 0.121 describes the result of the reported hypothesis test.
Why does the primary endpoint receive particular attention?
The primary endpoint was Pain Response at Week 6 Confirmed at Week 12. Its prespecified status means that it represents the principal efficacy question identified for the trial. Secondary findings, such as overall survival and bone scan response, provide additional information but answer different questions and should not be treated as interchangeable with the primary endpoint.
11. Understanding the Pain-Response Definition
The registry's pain-response definition is more specific than a simple improvement in a pain score. A participant had to achieve a decrease of at least 30% from baseline in the average daily worst pain intensity score during a 7-day reporting period, while also avoiding both a concomitant increase in average daily opioid use of any opioid narcotic type and the addition of a new opioid narcotic type relative to baseline.
Magnitude criterion
Pain intensity had to decrease by at least 30% from baseline during the specified 7-day reporting period.
Opioid-use criterion
The response definition also required no concomitant increase in average daily use of any opioid narcotic type.
New-opioid criterion
Addition of any new opioid narcotic type relative to baseline was also excluded from the response definition.
Confirmation
The primary endpoint was pain response at Week 6 confirmed at Week 12.
This composite definition is statistically important because it prevents the endpoint from being interpreted as a simple change in a numerical pain score. The responder classification incorporates both reported pain intensity and opioid use. Consequently, the CMH comparison concerns the proportion meeting the complete prespecified response definition.
12. Pain Progression and Its Relationship to Response
The registry also defines pain progression at a given time point. Progression was defined as a ≥ 30% increase from baseline in the average daily worst pain intensity score during a 7-day reporting period, or either an increase in average daily use of any type of opioid narcotic or addition of a new opioid narcotic type compared with baseline.
| Concept | Registry definition |
|---|---|
| Pain response | ≥ 30% from baseline in average daily worst pain intensity during a 7-day reporting period, with neither increased average daily opioid use nor addition of a new opioid narcotic type relative to baseline. |
| Pain progression | ≥ 30% increase from baseline in average daily worst pain intensity during a 7-day reporting period or either increased average daily opioid use or addition of a new opioid narcotic type compared with baseline. |
Response and progression should not be treated as mirror-image endpoints. The response definition requires a specified improvement together with restrictions on opioid escalation, whereas progression can be triggered by worsening pain or opioid escalation. Their statistical interpretation therefore depends on their individual registry definitions.
13. Analysis Populations and Stratification
The registry consistently identifies the randomized ITT population as the basis for the three posted efficacy analyses. This is an important feature of the statistical design because the same randomized population anchors the treatment comparison across the primary and reported secondary endpoints.
| Endpoint | Population | Stratification |
|---|---|---|
| Pain Response | ITT, 119 participants: 61 cabozantinib and 58 mitoxantrone plus prednisone | Eastern Cooperative Oncology Group performance status and prior cabazitaxel use |
| Overall Survival | ITT, 119 randomized participants: 61 cabozantinib and 58 mitoxantrone plus prednisone | Baseline Eastern Cooperative Oncology Group performance status and prior cabazitaxel use |
| Bone Scan Response | Randomized ITT, 119 participants: 61 cabozantinib and 58 mitoxantrone plus prednisone | Baseline Eastern Cooperative Oncology Group performance status and prior cabazitaxel use |
The use of the same randomized ITT population is particularly useful when interpreting results across endpoints. It reduces the possibility that apparent differences between outcomes arise simply because substantially different efficacy populations were analyzed.
14. Overall Survival: Reading the Hazard Ratio Correctly
The reported HR of 0.70 is a relative time-to-event estimate comparing the instantaneous rate of death between the randomized treatment groups. A value below 1 favors the numerator group in the reported comparison, here cabozantinib.
The HR is not a direct estimate of the probability that an individual patient will die. It also does not mean that every patient experiences a 30% reduction in risk, because the hazard ratio summarizes the relative event rate across the analyzed time-to-event experience.
The 95% CI of 0.44–1.10 expresses uncertainty around the estimated HR. Its width matters: the observed HR of 0.70 should not be interpreted without recognizing that the interval extends from 0.44 to 1.10.
The P-value of 0.121 is evidence from the reported hypothesis test. It does not indicate that the effect is 12.1% in size or that there is a 12.1% probability that the treatment has no effect.
15. Confidence Intervals and Precision
Among the posted analyses, the overall-survival endpoint is the one for which the registry reports a confidence interval around an effect estimate. The combination of the point estimate and interval is more informative than the HR alone.
| Measure | Reported value | Statistical role |
|---|---|---|
| Hazard ratio | 0.70 | Estimated relative treatment effect for overall survival |
| 95% confidence interval | 0.44–1.10 | Quantifies uncertainty around the HR estimate |
| P-value | 0.121 | Hypothesis-test result under the superiority framework |
A confidence interval containing 1 should not be converted into a statement that the treatments are proven identical. It means that the interval includes a value corresponding to equal hazard as well as values representing lower or higher relative hazard. The evidence therefore needs to be described using the full interval rather than the point estimate alone.
16. Safety Results
The registry reports serious adverse events by randomized treatment arm. These data provide an important separate dimension of the trial because safety events are not summarized by the efficacy hypothesis tests.
| Safety measure | Cabozantinib | Mitoxantrone/Prednisone |
|---|---|---|
| Serious adverse events | 16/60 | 15/57 |
The denominators in the serious-adverse-event summary are 60 and 57, respectively, while the primary efficacy ITT population consists of 61 and 58 randomized participants. This distinction illustrates why a safety denominator should not automatically be substituted for an efficacy denominator.
17. Trial Timeline
Trial start
The COMET-2 trial began in March 2012 according to the registry.
Primary completion
The registry lists October 2014 as the primary completion date.
Overall-survival data cutoff
The posted overall-survival analysis used the ITT population at the time of primary analysis with a data cutoff date of 06 October 2014.
Enrollment stopped before planned size
The registry states that enrollment was stopped before the planned study population size of 246 was reached; 119 participants were enrolled.
18. Sample Size and Early Termination
The registry states that the planned study population size was 246, but enrollment stopped after 119 participants had been enrolled. The final enrolled population was therefore substantially smaller than the planned study population.
Planned population
246 participants were planned.
Enrolled population
119 participants were enrolled.
ITT population
The primary pain-response analysis included all 119 randomized participants.
Statistical consequence
A smaller-than-planned sample can reduce the precision of estimates and the ability to detect effects, particularly for secondary endpoints.
This limitation is particularly relevant when interpreting the overall-survival confidence interval of 0.44–1.10. The point estimate is below 1, but the interval is broad enough to include 1. The smaller enrolled population should therefore remain part of the interpretation of the uncertainty around the reported results.
19. Multiplicity and Endpoint Interpretation
The registry identifies one primary endpoint and two posted secondary endpoint analyses. The primary endpoint was pain response at Week 6 confirmed at Week 12. Overall survival and bone scan response were secondary endpoints.
| Endpoint | Role | Statistical question |
|---|---|---|
| Pain Response at Week 6 Confirmed at Week 12 | Primary | Does the proportion meeting the prespecified pain-response definition differ between treatment groups? |
| Overall Survival | Secondary | Do the randomized groups differ in time to death? |
| Bone Scan Response at Week 12 | Secondary | Does the proportion meeting the bone-scan-response definition differ between treatment groups? |
The endpoint hierarchy matters because a secondary P-value should not automatically be interpreted as though it were the result of the primary confirmatory question. The reported P < 0.001 for bone scan response describes strong statistical evidence for that secondary comparison, but it does not change the statistical role of the primary pain-response endpoint.
20. What the Three Results Say Together
The three posted efficacy analyses point in different statistical directions. The primary pain-response comparison has P = 0.773. The overall-survival analysis reports HR 0.70 with a 95% CI of 0.44–1.10 and P = 0.121. The bone scan response comparison reports P < 0.001.
| Finding | Interpretation |
|---|---|
| Pain response: P = 0.773 | The reported CMH superiority test does not provide strong statistical evidence of a difference in the primary pain-response endpoint. |
| Overall survival: HR 0.70 | The estimated hazard of death was lower in the cabozantinib group, with a reported HR below 1. |
| Overall survival: 95% CI 0.44–1.10 | The survival estimate is uncertain and the interval includes 1. |
| Overall survival: P = 0.121 | The reported superiority test does not provide strong statistical evidence against the null hypothesis at conventional significance levels. |
| Bone scan response: P < 0.001 | The reported CMH superiority test provides strong statistical evidence against its null hypothesis for the Week 12 bone scan response comparison. |
The most important statistical lesson is that an individual endpoint must be interpreted according to its own definition, analysis population, effect measure, and hypothesis test. A strong result on one secondary endpoint does not retroactively change the result of the primary endpoint, just as a nonsignificant result on one endpoint does not erase information from another prespecified analysis.
21. Limitations
- Early termination: the registry states that enrollment stopped before the planned study population size of 246 was reached, with only 119 participants enrolled.
- Precision: the overall-survival HR of 0.70 has a 95% CI of 0.44–1.10, indicating uncertainty around the point estimate.
- Primary endpoint reporting: the posted pain-response analysis provides the CMH P-value but does not report treatment-specific responder percentages or a confidence interval.
- Bone scan response reporting: the posted analysis reports P < 0.001 but does not provide treatment-specific responder percentages or a confidence interval.
- Different denominators: the efficacy analyses use the randomized ITT population of 119, whereas the reported serious-adverse-event counts use denominators of 60 and 57.
- Endpoint differences: pain response and bone scan response are binary responder endpoints, while overall survival is a time-to-event endpoint; their P-values and effect measures are therefore not directly interchangeable.
- Hazard-ratio interpretation: an HR is a relative time-to-event measure and should not be interpreted as an absolute risk difference or an individual-level treatment effect.
22. Why This Trial Matters Statistically
COMET-2 is a useful teaching example because the registry contains three distinct efficacy analyses within a relatively small randomized phase 3 trial. The primary endpoint is a clinically structured responder outcome analyzed using a stratified CMH method, while overall survival requires a time-to-event framework and bone scan response returns to a stratified responder analysis.
| Concept | How it appears in COMET-2 |
|---|---|
| Randomization | 119 participants randomized to two parallel treatment groups. |
| Intention-to-treat analysis | The primary pain-response analysis used all 119 randomized participants. |
| Stratification | CMH analyses were stratified by Eastern Cooperative Oncology Group performance status and prior cabazitaxel use. |
| Categorical endpoint analysis | Pain response and bone scan response were analyzed with CMH methods. |
| Time-to-event analysis | Overall survival was analyzed with a stratified log-rank test. |
| Hazard ratio | Overall survival was summarized with HR 0.70. |
| Confidence interval | The overall-survival HR had a 95% CI of 0.44–1.10. |
| Endpoint hierarchy | Pain response was primary; overall survival and bone scan response were secondary. |
| Safety denominators | Serious adverse events were reported as 16/60 and 15/57. |
| Early termination | Enrollment stopped before the planned population of 246 was reached. |
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Calculators
25. Sources
- ClinicalTrials.gov: COMET-2, NCT01522443.
- PubMed: PubMed record.
- PubMed: PubMed record.
- PubMed: PubMed record.
- PubMed: PubMed record.
Continue through the Clinical Biostats statistical library
Explore statistical tutorials and calculators for categorical endpoints, confidence intervals, hazard ratios, and time-to-event analysis.
26. Statistical Summary
COMET-2 illustrates the importance of matching the statistical method to the endpoint. The primary pain-response endpoint used a stratified Cochran-Mantel-Haenszel analysis in the 119-participant ITT population and reported P = 0.773. Overall survival was analyzed with a stratified log-rank method and reported an HR of 0.70 with a 95% CI of 0.44–1.10 and P = 0.121. Week 12 bone scan response was analyzed using a stratified CMH method and reported P < 0.001. Serious adverse events were reported as 16/60 for cabozantinib and 15/57 for mitoxantrone plus prednisone.