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
SKYSCRAPER-04 was a randomized, parallel-group phase 2 study evaluating tiragolumab plus atezolizumab versus atezolizumab monotherapy in participants with metastatic and/or recurrent PD-L1-positive cervical cancer.
| Feature | SKYSCRAPER-04 |
|---|---|
| Trial name | SKYSCRAPER-04 |
| NCT identifier | NCT04300647 |
| Phase | Phase 2 |
| Status | Completed |
| Condition | Cervical cancer |
| Population | Participants with metastatic and/or recurrent PD-L1-positive cervical cancer |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 172 |
| Arms | 2 |
| Interventions | Tiragolumab; atezolizumab |
| Lead sponsor | Hoffmann-La Roche |
| Sponsor type | Industry |
2. Clinical Question
The registered study evaluates tiragolumab plus atezolizumab and atezolizumab monotherapy in participants with metastatic and/or recurrent PD-L1-positive cervical cancer. The primary registered question is framed around Independent Review Committee (IRC)-assessed objective response rate during the pre-crossover period.
Population
Participants with metastatic and/or recurrent PD-L1-positive cervical cancer.
Intervention
Tiragolumab plus atezolizumab.
Comparator
Atezolizumab monotherapy.
Primary question
How does the registered pre-crossover IRC-assessed objective response endpoint compare between the randomized treatment groups?
3. Trial Design
Atezolizumab Monotherapy
- Atezolizumab was the study intervention in the monotherapy arm.
- Pre-crossover serious adverse events: 12/45.
Atezolizumab + Tiragolumab
- Tiragolumab was evaluated in combination with atezolizumab.
- Pre-crossover serious adverse events: 42/126.
4. Trial Timeline and Registry Status
Study start
The registry lists June 30, 2020 as the study start date.
Primary completion
The registry lists December 8, 2021 as the primary completion date.
Completed
The ClinicalTrials.gov record identifies the study as completed and indicates that results have been posted.
5. Primary Endpoint
| Endpoint | Registry definition | Time frame | Endpoint type |
|---|---|---|---|
| Pre-crossover Period: Independent Review Committee (IRC)-Assessed Objective Response Rate (ORR) | ORR was defined as the percentage of participants with a complete response (CR) or a partial response (PR) on two consecutive occasions ≥4 weeks apart, as determined by the IRC according to Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST v1.1). | From randomization up to approximately 17 months | Binary |
The registry definition identifies ORR as a binary response endpoint. At the participant level, the outcome can therefore be represented as whether the prespecified response criterion was achieved during the specified assessment period.
The registered endpoint is a proportion rather than a time-to-event endpoint. The ClinicalTrials.gov record does not provide the arm-specific numbers of responders or the corresponding formal statistical comparison.
6. Statistical Methodology
Randomization
The study was randomized with two parallel treatment arms. Randomization is important statistically because it is intended to make treatment assignment independent of participants' baseline characteristics, in expectation, allowing differences in outcomes between randomized groups to be interpreted as treatment-group differences under the trial design.
Binary endpoint analysis
The primary endpoint is explicitly classified in the registry data as binary. For a binary response endpoint such as ORR, the basic analysis compares the proportion of participants meeting the response definition in the two randomized groups.
These are general statistical measures that could describe a randomized binary endpoint. The ClinicalTrials.gov record does not specify which formal effect measure or hypothesis test was used for SKYSCRAPER-04.
Confidence intervals
A confidence interval around an ORR difference, ratio, or odds ratio quantifies statistical uncertainty associated with the estimated treatment contrast. The interval is distinct from the observed point estimate: two studies can have the same estimated response difference but substantially different precision if their sample sizes or event patterns differ.
Formal hypothesis testing
A formal comparison of two binary proportions can be conducted with methods such as a two-sample proportion test, chi-square test, or Fisher's exact test, depending on the prespecified statistical analysis plan and the observed cell counts. The ClinicalTrials.gov record does not identify the formal test used for the primary endpoint.
Independent Review Committee assessment
The primary endpoint is specifically described as IRC-assessed. Independent review is intended to provide a standardized assessment of tumor response using the registered RECIST v1.1 framework rather than relying solely on investigator assessment.
7. Results
Primary endpoint: IRC-assessed ORR
The registry reports the primary endpoint as a binary, pre-crossover IRC-assessed ORR measured from randomization up to approximately 17 months. The ClinicalTrials.gov record does not provide the number or percentage of participants with CR or PR in either randomized arm, and they do not provide a formal statistical comparison.
| Primary endpoint | What the ClinicalTrials.gov record reports |
|---|---|
| IRC-assessed ORR | Registered and results posted |
| Assessment period | From randomization up to approximately 17 months |
| Endpoint type | Binary |
| Statistical analysis | No the posted statistical analyses entry reported |
| Arm-specific ORR estimate | Not contained in the ClinicalTrials.gov record |
| Confidence interval | Not contained in the ClinicalTrials.gov record |
| p-value | Not contained in the ClinicalTrials.gov record |
The important statistical distinction is between an endpoint being reported and a formal treatment comparison being reported. The ClinicalTrials.gov record establishes that ORR was the primary binary endpoint and defines how response was determined, but it does not supply the arm-level response estimates or a statistical test.
Without those quantities, it would be inappropriate to infer the magnitude of the treatment effect from the existence of a posted endpoint alone. An ORR comparison would normally report the response proportion in each randomized group together with an appropriate measure of effect and its uncertainty.
A p-value, if available, would address compatibility with a specified null hypothesis under the chosen statistical model or test. It would not measure the size or clinical importance of the treatment effect. A confidence interval would provide the corresponding information about statistical precision.
Safety: serious adverse events by arm
| Period / treatment group | Affected / at risk | Observed proportion |
|---|---|---|
| Pre-crossover: Atezolizumab Monotherapy | 12/45 | 12/45 |
| Pre-crossover: Atezolizumab + Tiragolumab | 42/126 | 42/126 |
| Post Crossover: Atezolizumab + Tiragolumab | 1/17 | 1/17 |
The ClinicalTrials.gov record identifies these as serious adverse events by arm. They should not be treated as an efficacy analysis, and the post-crossover group should not be compared with the randomized pre-crossover groups as though it were a third randomized arm.
The affected/at-risk counts provide a descriptive safety summary. Because the ClinicalTrials.gov record does not identify a prespecified statistical comparison for serious adverse events, no comparative p-value or confidence interval should be inferred.
The denominator also matters. Safety proportions are conditional on the population represented by each period and arm. In particular, the post-crossover 1/17 observation comes from a post-crossover group rather than from the original randomized comparison.
8. Statistical Methods Explained
Why was randomization important in this trial?
Randomization creates the foundation for a causal comparison between the treatment groups. In a randomized parallel-group study, treatment assignment is determined by the trial's allocation process rather than by the participant's response or prognosis. This helps balance both measured and unmeasured prognostic factors in expectation.
Why is ORR analysed as a binary endpoint?
The registered primary endpoint classifies participants according to whether they achieve a complete response or partial response meeting the specified confirmation rule. That creates an outcome with two broad categories: response achieved or response not achieved. The resulting treatment comparison is therefore a comparison of proportions.
Why does the IRC matter?
The registry specifies an Independent Review Committee assessment using RECIST v1.1. Independent assessment can reduce the influence of knowledge of treatment assignment on radiologic response classification, although the ClinicalTrials.gov record identifies the overall study as unmasked. The statistical value comes from having a defined and independently assessed response criterion.
What does a binary treatment effect tell us?
A binary treatment effect summarizes how frequently an outcome occurs in one group relative to another. A risk difference describes the absolute difference in proportions, while a risk ratio and odds ratio describe relative comparisons on different scales. Each answers a slightly different question.
Why are confidence intervals important for ORR?
An observed response percentage is an estimate based on a finite number of participants. A confidence interval communicates how precisely the treatment effect has been estimated under the selected statistical method. Wider intervals indicate greater statistical uncertainty than narrower intervals, all else being equal.
Why should crossover be separated from the primary randomized comparison?
Crossover changes the treatment actually received after the initial randomized assignment. A randomized efficacy analysis generally preserves the original treatment assignment as the comparison framework, whereas outcomes observed after crossover can reflect exposure to treatments received after the initial assignment. This is why the ClinicalTrials.gov record distinguishes pre-crossover and post-crossover periods.
9. Crossover and Analysis Interpretation
The ClinicalTrials.gov record explicitly distinguish pre-crossover safety information from a post-crossover atezolizumab-plus-tiragolumab group. That distinction is statistically important because randomization applies to the initial treatment assignment, whereas treatment received after crossover is no longer equivalent to the original randomized assignment.
Pre-crossover
The reported serious-adverse-event counts are 12/45 for atezolizumab monotherapy and 42/126 for atezolizumab plus tiragolumab.
Post-crossover
The ClinicalTrials.gov record reports 1/17 affected by a serious adverse event in the post-crossover atezolizumab-plus-tiragolumab group.
These denominators should remain attached to their respective periods. Combining the three counts into a single comparison would discard the distinction between randomized treatment assignment and post-crossover exposure.
10. Endpoint Interpretation
The primary endpoint is not a time-to-event endpoint such as overall survival or progression-free survival. Instead, it is a response endpoint measured during a specified period after randomization.
| Statistical feature | Implication for SKYSCRAPER-04 |
|---|---|
| Endpoint scale | Binary |
| Outcome | IRC-assessed objective response |
| Response definition | CR or PR on two consecutive occasions ≥4 weeks apart |
| Assessment framework | RECIST v1.1 |
| Assessment period | From randomization up to approximately 17 months |
| Primary analysis population in the ClinicalTrials.gov record | Not specified beyond the registered endpoint description |
| Formal statistical test in the ClinicalTrials.gov record | Not in the posted results |
ORR is not the same as time-to-event efficacy
ORR asks whether a participant achieved a qualifying tumor response during the assessment period. It does not directly measure how long a participant remains alive, how long disease remains controlled, or how long a response persists unless those are separately defined endpoints. The ClinicalTrials.gov record identifies ORR as the primary endpoint but do not provide those additional efficacy estimates.
11. Statistical Interpretation of the Available Safety Counts
These are descriptive affected/at-risk counts from the ClinicalTrials.gov record. They are not accompanied by a statistical comparison in the ClinicalTrials.gov record.
The two pre-crossover denominators are different, so the counts alone should not be compared as raw numbers. A formal comparison would operate on proportions or another prespecified safety measure. Even then, interpretation would depend on the exact safety-analysis population, follow-up, event definitions, and statistical method.
This observation belongs to the post-crossover period and should be interpreted descriptively rather than as a randomized treatment-arm result.
12. What Is Known and What Is Not Reported in the Supplied Data
| Question | Supplied registry information |
|---|---|
| Was the study randomized? | Yes |
| Was it parallel group? | Yes |
| Was it masked? | No |
| What was the primary endpoint? | Pre-crossover IRC-assessed ORR |
| What was the endpoint type? | Binary |
| What was the endpoint time frame? | From randomization up to approximately 17 months |
| Were results posted? | Yes |
| Are formal statistical analyses reported? | No |
| Are primary-endpoint effect estimates reported? | No |
| Are primary-endpoint confidence intervals reported? | No |
| Are primary-endpoint p-values reported? | No |
| Are serious adverse-event counts reported? | Yes |
This distinction is central to responsible trial interpretation. A registry can contain a posted outcome without the ClinicalTrials.gov recordset containing every numerical component needed to reconstruct the formal inferential analysis.
13. Limitations
- No formal primary-endpoint analysis in the ClinicalTrials.gov record: no formal statistical analyses were posted, so no arm-specific ORR estimate, confidence interval, p-value, or effect measure can be reported from the ClinicalTrials.gov recordset.
- Results-posted does not mean all numerical results are available here: the ClinicalTrials.gov record states that results are posted and that 16 outcome measures are posted, but the ClinicalTrials.gov record does not contain their numerical efficacy results.
- Binary endpoint: ORR reduces tumor-response information to a response classification. It should not be interpreted as a direct measure of survival or duration of benefit.
- Crossover: the ClinicalTrials.gov record distinguishes pre-crossover and post-crossover observations. Post-crossover data should not be treated as another randomized arm.
- Safety denominators: serious-adverse-event counts are reported with their affected/at-risk denominators, but the statistical analysis population and follow-up framework for those counts are not further described in the ClinicalTrials.gov record.
- Open-label design: the trial is recorded as having no masking. The primary endpoint was nevertheless assessed by an Independent Review Committee according to the registered definition.
- Missing methodological details: the ClinicalTrials.gov record does not identify the formal hypothesis test, confidence-interval method, multiplicity strategy, missing-data/imputation method, interim-analysis framework, or stratification factors.
- Interpretation should remain registry-specific: additional claims about efficacy, subgroups, long-term follow-up, or detailed statistical procedures would require source information beyond the ClinicalTrials.gov record.
14. Why This Trial Matters Statistically
SKYSCRAPER-04 is a useful teaching case because the ClinicalTrials.gov record illustrates how a randomized clinical trial can combine a clearly defined binary efficacy endpoint with independent response assessment, an open-label design, and crossover-related safety reporting.
| Concept | How it appears in SKYSCRAPER-04 |
|---|---|
| Randomization | The trial uses randomized allocation across two treatment arms. |
| Parallel-group design | The study has two parallel treatment groups. |
| Binary endpoint | The registered primary endpoint is IRC-assessed ORR. |
| Independent review | ORR is assessed by an Independent Review Committee. |
| RECIST assessment | Response is defined according to RECIST v1.1. |
| Response confirmation | CR or PR must occur on two consecutive occasions ≥4 weeks apart. |
| Open-label design | The ClinicalTrials.gov record identifies the study as having no masking. |
| Crossover | The ClinicalTrials.gov record separates pre-crossover and post-crossover periods. |
| Safety denominators | Serious adverse events are reported as affected/at-risk counts by treatment period and group. |
| Registry-data interpretation | Posted outcomes do not necessarily provide all inferential statistics needed for a formal comparison. |
15. Related Tutorials
Learn more about the methods used in this trial:
16. Related Statistical Calculators
17. Sources
- ClinicalTrials.gov: NCT04300647 — SKYSCRAPER-04.
- PubMed: PMID 38858106.
Continue through Clinical Biostats
Explore statistical tutorials and calculators related to randomized clinical-trial methods and binary endpoints.
18. Record Summary
SKYSCRAPER-04 is a completed randomized phase 2, parallel-group study of tiragolumab plus atezolizumab versus atezolizumab monotherapy in participants with metastatic and/or recurrent PD-L1-positive cervical cancer. The registered primary endpoint is pre-crossover Independent Review Committee-assessed objective response rate, defined as the percentage of participants with a complete response or partial response on two consecutive occasions ≥4 weeks apart according to RECIST v1.1, measured from randomization up to approximately 17 months.
The ClinicalTrials.gov record establishes the study design, endpoint definition, enrollment, treatment groups, study dates, and serious-adverse-event counts. They also state that results are posted. However, no formal statistical analyses were posted, so the available data do not support reporting an arm-specific ORR estimate, confidence interval, p-value, or other formal primary-endpoint treatment effect. The appropriate statistical interpretation is therefore to distinguish the registered endpoint and its usual analytical framework from numerical inferential results that are not contained in the ClinicalTrials.gov recordset.