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
PISCES was a randomized phase 3 trial evaluating patient preference between pazopanib and sunitinib in carcinoma of the kidney. The primary preference analysis compared two treatment sequences and used a sequence-adjusted Prescotts test in a modified-intent-to-treat population.
| Feature | PISCES |
|---|---|
| Phase | Phase 3 |
| Condition | Carcinoma, Renal Cell |
| Brief title | Patient Preference Study of Pazopanib Versus Sunitinib in Advanced or Metastatic Kidney Cancer |
| Design | Randomized, quadruple-masked, treatment-purpose study |
| Allocation | Randomized |
| Enrollment | 169 |
| Interventions | Pazopanib; sunitinib |
| Primary endpoints | 2 binary endpoints |
| Results posted | Yes |
| Statistical analyses posted | 1 |
| Lead sponsor | Novartis Pharmaceuticals |
| ClinicalTrials.gov | NCT01064310 |
2. Clinical Question
The central question was not simply whether one drug produced a better clinical outcome. The registered primary endpoint asked which of the two study drugs participants preferred after experiencing both double-blind treatment periods.
Population
Participants in a phase 3 study of pazopanib versus sunitinib for advanced or metastatic kidney cancer, with the condition registered as carcinoma, renal cell.
Intervention
Pazopanib was one of the two study drugs evaluated in the patient-preference comparison.
Comparator
Sunitinib was the other study drug in the randomized treatment-sequence comparison.
Primary question
What was the difference in the percentage of participants who preferred pazopanib versus sunitinib after both treatment periods?
3. Trial Design
The trial was randomized and quadruple-masked. Its structure is important because the primary outcome was a within-participant preference after exposure to both drugs, rather than a conventional parallel-group efficacy endpoint in which each participant remains on only one randomized treatment.
Sunitinib followed by pazopanib
- Sunitinib 50 mg in the first treatment period
- Pazopanib 800 mg in the second treatment period
- Participants subsequently completed the Patient Preference Questionnaire at the end of treatment of both study drugs
Pazopanib followed by sunitinib
- Pazopanib 800 mg in the first treatment period
- Sunitinib 50 mg in the second treatment period
- Participants subsequently completed the Patient Preference Questionnaire at the end of treatment of both study drugs
4. Endpoints
| Registered endpoint | Time frame | Type | Formal analysis posted |
|---|---|---|---|
| Number of Participants With Preference for Pazopanib Versus Sunitinib as Assessed by the Patient Preference Questionnaire (PPQ) | End of treatment of both study drugs (maximum of 22 weeks) | Binary | Yes |
| Number of Participants Answering "Yes," "no," or Not Applicable (N/A) to the Question of Whether the Indicated Factors Influenced Their Preference for Sunitinib or Pazopanib Treatment as Assessed by the Patient Preference Questionnaire | End of treatment of both study drugs (maximum of 22 weeks) | Binary | No formal statistical analysis posted |
5. Statistical Methodology
Primary analysis population
The primary analysis used a Modified-Intent-to-Treat (mITT) Population. The registry definition describes this as participants who received at least one dose of study treatment from each treatment period and who did not have documented progressive disease (PD) at the end of Treatment Period 1 and completed the patient preference questionnaire.
This population definition is different from a conventional all-randomized intention-to-treat population. Requiring exposure to at least one dose in each treatment period means that participants who did not complete exposure to both treatments could be excluded from the primary preference analysis.
Prescotts test
The registered statistical method for the primary preference endpoint was Prescotts test. The analysis compared the two randomized treatment sequences and estimated the difference in the percentage of participants who preferred pazopanib versus sunitinib.
The registry analysis notes that the estimated value was adjusted for sequence. This adjustment is particularly relevant to a crossover-style preference design because the first treatment and second treatment are not statistically interchangeable positions.
The reported estimate is therefore a percentage-point comparison, not a hazard ratio, odds ratio, risk ratio, or mean difference.
Confidence interval
The primary estimate was accompanied by a two-sided 90% confidence interval. Confidence intervals are useful because they show the range of values compatible with the observed data and the statistical model under the stated confidence level.
Hypothesis type
The registered hypothesis type was superiority. The formal primary analysis therefore evaluated a treatment-preference difference rather than a non-inferiority margin or equivalence margin.
Covariate adjustment
The registry identifies covariate adjustment as an additional concept in the analysis text and explicitly states that the preference estimate was adjusted for sequence. That adjustment is a key part of interpreting the reported percentage difference.
Intention-to-treat considerations
The analysis is described as using a modified-intent-to-treat population rather than the full randomized population. This preserves some of the randomized structure while restricting the analysis to participants who received at least one dose during each treatment period, according to the registry definition.
6. Primary Result: Patient Preference
Sequence-adjusted preference difference
Two-sided 90% CI: 37.0–61.5 · P < 0.001
Effect measure: percentage of participants
The primary analysis compared Sunitinib 50 mg followed by Pazopanib 800 mg with Pazopanib 800 mg followed by Sunitinib 50 mg. The reported estimate represents the difference in the percentage of participants who preferred pazopanib versus sunitinib, with adjustment for treatment sequence.
| Primary endpoint component | Reported result |
|---|---|
| Analysis population | Modified-Intent-to-Treat (mITT) Population |
| Groups compared | Sunitinib 50 mg followed by Pazopanib 800 mg vs Pazopanib 800 mg followed by Sunitinib 50 mg |
| Method | Prescotts test |
| Effect measure | Percentage of participants |
| Estimate | 49.26 |
| Confidence interval | Two-sided 90% CI 37.0–61.5 |
| P-value | <0.001 |
| Hypothesis | Superiority |
| Adjustment | Adjusted for sequence |
The 49.26 estimate is a difference in percentages, expressed on a percentage-point scale. In the context of the registry analysis, it represents the estimated difference in the percentage of participants who preferred pazopanib versus sunitinib after accounting for treatment sequence.
It does not mean that 49.26% of all enrolled participants preferred pazopanib. It is also not a relative risk, odds ratio, probability of clinical benefit, survival effect, or percentage reduction in an adverse event rate.
The two-sided 90% CI of 37.0–61.5 describes statistical uncertainty around the sequence-adjusted percentage difference. It does not mean that individual patient preferences vary only within this interval.
The P < 0.001 result addresses evidence against the null hypothesis under the specified statistical procedure. It does not measure the size or clinical importance of the preference difference. The estimate and confidence interval are needed to understand magnitude and precision.
Finally, the mITT population and the sequence adjustment matter. Because the primary analysis required participants to receive at least one dose in each treatment period, the estimate should not automatically be interpreted as the preference effect that would have been obtained from every randomized participant.
7. What the Primary Estimate Means Statistically
A useful way to read the primary result is to separate three quantities that are often confused: the effect estimate, the confidence interval, and the P-value.
Effect estimate
The reported 49.26 is the estimated sequence-adjusted difference in the percentage of participants preferring pazopanib versus sunitinib.
Confidence interval
The two-sided 90% CI of 37.0–61.5 communicates uncertainty around the estimated percentage difference.
P-value
P < 0.001 indicates strong statistical evidence against the null hypothesis under the reported test. It is not an effect-size measure.
Sequence adjustment
The analysis accounts for whether participants received sunitinib or pazopanib first, an important feature of a two-period preference design.
8. The Second Primary Endpoint
The second registered primary endpoint concerned the number of participants answering "Yes," "no," or Not Applicable (N/A) to the question of whether indicated factors influenced their preference for sunitinib or pazopanib, as assessed by the Patient Preference Questionnaire.
| Feature | Registry information reported |
|---|---|
| Endpoint | Number of Participants Answering "Yes," "no," or Not Applicable (N/A) to the Question of Whether the Indicated Factors Influenced Their Preference for Sunitinib or Pazopanib Treatment as Assessed by the Patient Preference Questionnaire |
| Time frame | End of treatment of both study drugs (maximum of 22 weeks) |
| Endpoint type | Binary |
| Results posted | Yes |
| Formal statistical analysis in the ClinicalTrials.gov record | No |
For a binary endpoint of this kind, a formal comparison would ordinarily specify the analysis population, comparison of response proportions, an appropriate confidence interval, and a hypothesis test if a confirmatory comparison were intended. The ClinicalTrials.gov record does not provide such a formal statistical analysis for this endpoint, so no numerical treatment effect or P-value is assigned here.
9. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment group. These figures should be read as affected participants divided by the stated at-risk population.
| Group | Participants with serious adverse events | At risk | Reported affected / at risk |
|---|---|---|---|
| Sunitinib | 35 | 148 | 35/148 |
| Pazopanib | 30 | 153 | 30/153 |
| Open Label Pazopanib | 15 | 84 | 15/84 |
These safety data should not be combined into a single treatment comparison. In particular, Open Label Pazopanib is presented as a separate reported group in the ClinicalTrials.gov record, with 15 affected participants among 84 at risk.
10. Crossover, Treatment Sequence, and Open-Label Treatment
PISCES is statistically distinctive because participants were evaluated after exposure to both study drugs. That makes treatment sequence a central design variable rather than a minor stratification detail.
The primary analysis directly reflects this feature: the registry reports that the preference difference was adjusted for sequence. Without accounting for sequence, a preference response could potentially reflect not only which drug was preferred but also which drug was experienced first.
The registry also states that the study remained open to allow subjects currently on treatment continued access to treatment with open label pazopanib. This creates an additional distinction between the controlled treatment periods used for the preference assessment and later treatment access.
Controlled preference assessment
The primary endpoint was assessed at the end of treatment of both study drugs, with a maximum time frame of 22 weeks.
Sequence adjustment
The primary statistical estimate was adjusted for whether the participant received sunitinib or pazopanib first.
Open-label access
The study remained open so subjects currently on treatment could continue access to open label pazopanib.
Interpretive boundary
A preference endpoint after two treatment periods should not be interpreted as a conventional parallel-group survival or response endpoint.
11. Statistical Methods Explained
Why was sequence adjustment used?
In a two-period treatment-preference study, the order of exposure can matter. A participant receiving sunitinib first and pazopanib second may respond to the questionnaire differently from a participant receiving pazopanib first and sunitinib second, even if the underlying treatment preferences are otherwise comparable. The registry therefore reports a preference estimate adjusted for sequence.
What does 49.26 mean?
The value 49.26 is an estimated difference in the percentage of participants who preferred pazopanib versus sunitinib. It is expressed as a percentage-point effect measure. It should not be read as a hazard ratio, relative risk, or probability that an individual participant will prefer pazopanib.
Why is the confidence interval 90% rather than 95%?
The registry analysis specifies a two-sided 90% confidence interval. The confidence level is part of the prespecified statistical reporting framework. A 90% interval is narrower than a 95% interval calculated from the same underlying information, but the appropriate interpretation is always tied to the confidence level that was actually reported.
What does P < 0.001 tell us?
The P-value describes the compatibility of the observed result with the null hypothesis under the reported statistical test. It does not quantify the size of the preference difference and does not tell us how likely the null hypothesis itself is to be true.
Why use a modified-intent-to-treat population?
The primary analysis population required participants to have received at least one dose of study treatment during each treatment period, according to the registry definition. This makes the population particularly relevant to a preference question that requires exposure to both treatments, but it also means the primary estimate is not simply an all-randomized analysis.
What is the role of masking?
The trial was registered as quadruple-masked. Blinding can reduce the influence of expectations on participant-reported outcomes, which is especially relevant when the endpoint is a subjective preference assessment. The ClinicalTrials.gov record does not provide additional detail about the individual masking roles, so no more specific masking assignment is inferred here.
Is this a conventional intention-to-treat analysis?
No. The registry-reported statistical analysis identifies a Modified-Intent-to-Treat population. The distinction matters because conventional ITT generally retains all randomized participants in their assigned groups, whereas the registry-reported mITT definition requires exposure to at least one dose in each treatment period and contains an additional exclusion criterion.
12. Randomization and Internal Validity
Randomization is important even in a patient-preference study. It determines the sequence assigned to participants rather than allowing participants or investigators to choose which drug is received first.
The statistical analysis then adjusts the preference comparison for sequence, preserving the design feature that the registry identifies as relevant to the primary estimate.
Quadruple masking also has a particular relevance here because preference is a participant-reported endpoint. If participants knew which treatment they were receiving, expectations about the drugs could potentially influence their reported preference. The registry's quadruple-masked designation indicates that masking was part of the registered design.
13. Intention-to-Treat and the mITT Population
It is useful to distinguish randomization from analysis population. Randomization occurs at the beginning of the trial, while the analysis population is defined according to subsequent eligibility for the particular analysis.
| Concept | Role in PISCES |
|---|---|
| Randomization | Participants were randomized to one of two treatment sequences. |
| mITT analysis | The primary preference analysis used a Modified-Intent-to-Treat population. |
| Exposure requirement | The registry-reported definition requires at least one dose from each treatment period. |
| Primary estimand | Sequence-adjusted difference in the percentage preferring pazopanib versus sunitinib. |
The practical implication is that the primary result describes the population defined for the mITT analysis. It should not automatically be generalized to every randomized participant without considering who was excluded by the analysis definition.
14. Confidence Intervals and Precision
The primary estimate was 49.26, with a two-sided 90% CI of 37.0–61.5. The interval communicates uncertainty around the estimated sequence-adjusted percentage difference.
A confidence interval is not a list of individual patient outcomes. Nor does it mean that there is a 90% probability that the fixed underlying effect lies between 37.0 and 61.5. Its formal interpretation is tied to the repeated-sampling framework used to construct the interval.
The width of the interval also matters. A relatively narrow interval provides more precision than a very wide interval, all else equal. For PISCES, the reported interval gives substantially more information about the estimated magnitude than the P-value alone.
15. P-Values and Effect Size
The PISCES primary analysis provides a useful example of why statistical significance and effect magnitude should be separated.
Effect size
49.26 is the reported sequence-adjusted percentage difference.
Uncertainty
37.0–61.5 is the two-sided 90% confidence interval around the estimate.
Evidence against the null
P < 0.001 is the reported P-value from Prescotts test.
What the P-value is not
It is not the probability that pazopanib is preferred, and it is not a measure of clinical importance.
A small P-value can arise with different effect sizes depending on sample size and variability. Conversely, an effect estimate can be substantial while uncertainty remains large. Reporting the estimate, confidence interval, and P-value together avoids collapsing these distinct statistical concepts into a single number.
16. Missing Data and Analysis-Set Considerations
The ClinicalTrials.gov record does not provide a detailed missing-data or imputation strategy for the Patient Preference Questionnaire. It does, however, define the primary analysis using a modified-intent-to-treat population requiring exposure to both treatment periods.
That distinction is important. If a participant did not receive treatment during both periods, the participant may not satisfy the registry-reported mITT definition. This can change the population contributing to the preference analysis and therefore the target population to which the primary estimate most directly applies.
17. Multiplicity and the Two Primary Endpoints
PISCES registered 2 primary endpoints, while the ClinicalTrials.gov record reports only 1 statistical analysis. The first primary endpoint has a formal superiority analysis with an estimate, confidence interval, and P-value. The second endpoint is reported as an outcome measure but does not have a formal statistical analysis in the ClinicalTrials.gov record.
| Endpoint | Role | Formal analysis reported | Reported statistical result |
|---|---|---|---|
| Preference for pazopanib versus sunitinib | Primary | Yes | Estimate 49.26; two-sided 90% CI 37.0–61.5; P < 0.001 |
| Factors influencing preference | Primary | No | No formal comparison reported |
This is an important registry-analysis distinction.
18. Trial Timeline
Study start
The registered study start date was 2010-05-17.
Randomized treatment study
The phase 3 study evaluated pazopanib and sunitinib using two randomized treatment sequences and a Patient Preference Questionnaire.
Primary completion
The registered primary completion date was 2011-10-19.
Registry status
The trial is listed as COMPLETED, with results posted and one statistical analysis posted in the ClinicalTrials.gov record.
19. Why Patient Preference Is a Distinct Statistical Endpoint
Patient preference is fundamentally different from endpoints such as overall survival, progression-free survival, tumor response, or a laboratory measurement. The endpoint asks participants to integrate their experiences with both treatments and then select a preference response.
That makes the questionnaire a patient-reported binary outcome within a treatment-sequence framework. The statistical challenge is therefore not simply estimating the percentage of patients who prefer each treatment; the analysis must also account for the fact that participants experienced treatments in different orders.
Subjective endpoint
Preference is reported by the participant rather than directly measured as a physiological or time-to-event outcome.
Within-participant exposure
The questionnaire follows treatment with both study drugs rather than a single permanent treatment assignment.
Sequence effect
The registry explicitly accounts for treatment order in the primary analysis.
Binary analysis
The primary endpoint is classified as binary in the ClinicalTrials.gov record.
20. Important Limitations and Interpretation Issues
- mITT population: the primary analysis did not use an unrestricted all-randomized population; the registry-reported definition requires at least one dose from each treatment period and includes an additional exclusion criterion.
- Sequence dependence: because the primary outcome follows exposure to both drugs, treatment order is an important design feature and was explicitly included in the analysis.
- Preference is not survival: a patient preference difference does not directly estimate an effect on overall survival, progression, tumor response, or another clinical time-to-event endpoint.
- Second primary endpoint: although the registry reports the second primary outcome measure, the ClinicalTrials.gov record contains no formal statistical analysis for it.
- Safety denominators: the registry-reported serious-adverse-event groups have different denominators, including a separate open-label pazopanib group.
- No registry-reported imputation method: the available data does not specify a formal missing-data or imputation procedure for the preference questionnaire.
- Open-label access: the study remained open to allow subjects currently on treatment continued access to open label pazopanib, which is an important feature when interpreting the study's later treatment experience.
- Registry scope: the statistical interpretation on this page is limited to the facts and analyses contained in the ClinicalTrials.gov record.
21. Why This Trial Matters Statistically
PISCES is a useful teaching case because it shows how a randomized clinical trial can have a primary endpoint that is neither a conventional continuous outcome nor a time-to-event outcome. Instead, it combines randomization, treatment sequence, patient-reported preference, binary analysis, covariate adjustment, confidence intervals, and a modified-intent-to-treat population.
| Concept | How it appears in PISCES |
|---|---|
| Randomization | Participants were randomized to one of two treatment sequences. |
| Blinding | The registered masking designation was quadruple. |
| Binary endpoint | The primary preference endpoint is classified as binary. |
| Patient-reported outcome | Preference was assessed with the Patient Preference Questionnaire. |
| Sequence adjustment | The reported preference difference was adjusted for sequence. |
| Prescotts test | The registered formal method for the primary preference analysis. |
| Confidence interval | The primary estimate has a two-sided 90% CI of 37.0–61.5. |
| P-value | The primary analysis reports P < 0.001. |
| mITT analysis | The primary analysis uses a Modified-Intent-to-Treat population. |
| Multiple primary endpoints | Two primary endpoints were registered, while one formal statistical analysis is reported. |
| Safety denominators | Serious adverse events are reported separately for sunitinib, pazopanib, and open label pazopanib. |
22. Statistical Concepts in This Trial
This trial can serve as a practical introduction to several core statistical concepts.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Calculators
25. Sources
- ClinicalTrials.gov: PISCES — NCT01064310.
- PubMed: PubMed record 37146227.
- PubMed: PubMed record 35011579.
- PubMed: PubMed record 26457466.
- PubMed: PubMed record 24687826.
Continue through the Clinical Biostats statistical pathway
Use the trial's design and analysis concepts to explore broader tutorials and statistical calculators.
26. Record Summary
PISCES provides a useful example of how clinical-trial statistics change when the principal outcome is patient preference after exposure to two randomized treatment sequences. The registered primary analysis used a Modified-Intent-to-Treat population, Prescotts test, sequence adjustment, a percentage-based effect measure, a two-sided 90% confidence interval, and a superiority hypothesis.
The reported primary estimate was 49.26, with a two-sided 90% CI of 37.0–61.5 and P < 0.001. The estimate should be interpreted as a sequence-adjusted difference in the percentage of participants preferring pazopanib versus sunitinib, not as a relative risk, survival effect, or probability of benefit for an individual patient.
The trial also illustrates why the analysis population, treatment sequence, masking, endpoint definition, and availability of formal statistical analyses matter when interpreting registry results. The second registered primary endpoint has results posted as an outcome measure but no formal statistical analysis in the ClinicalTrials.gov record, so it should not be assigned an unreported treatment effect or P-value.