This page separates reported trial results from statistical interpretation. All numerical trial results presented here are taken from the ClinicalTrials.gov record data for NCT01544179. The registry provides the official trial record.
1. Trial at a Glance
IMPRESS was a completed, randomized, phase 3, parallel-group trial in non-small cell lung cancer. The registry describes the study as quadruple-masked and compares gefitinib treatment beyond progression in addition to chemotherapy with chemotherapy alone.
| Feature | IMPRESS |
|---|---|
| Trial name | IMPRESS |
| NCT identifier | NCT01544179 |
| Phase | Phase 3 |
| Condition | Non-Small Cell Lung Cancer |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 265 |
| Interventions | Gefitinib; placebo; pemetrexed; cisplatin |
| Lead sponsor | AstraZeneca |
| Status | Completed |
2. Clinical Question
The trial addresses whether continuing gefitinib beyond progression while adding it to chemotherapy produces a different progression-free survival outcome than chemotherapy alone.
Population
Participants enrolled in the phase 3 IMPRESS study with non-small cell lung cancer.
Intervention
Gefitinib treatment beyond progression in addition to chemotherapy; the registry intervention list includes gefitinib, pemetrexed, and cisplatin.
Comparator
Placebo in addition to chemotherapy, with placebo, pemetrexed, and cisplatin included in the registry intervention list.
Primary question
Does gefitinib plus chemotherapy improve the registered progression-free survival endpoint compared with placebo plus chemotherapy?
3. Trial Design
Gefitinib-containing strategy
- Gefitinib
- Gefitinib treatment beyond progression
- Chemotherapy included pemetrexed and cisplatin in the registry intervention list
Placebo-containing strategy
- Placebo
- Chemotherapy
- Pemetrexed and cisplatin were included in the registry intervention list
4. Trial Timing and Registry Record
Trial start
The registry lists March 15, 2012 as the study start date.
Primary completion
The registry lists May 5, 2014 as the primary completion date.
Registry status
The study status is recorded as completed, with results posted and 12 outcome measures reported.
5. Primary Endpoints
| Endpoint | Registry definition / time frame | Statistical analysis |
|---|---|---|
| Progression-Free Survival (Site Read, Investigator Assessment) | Time from randomisation until the date of objective disease progression as defined by RECIST version 1.1 or death by any cause in the absence of progression. Radiologic evaluations were carried out every 6 weeks from randomization until documented progression, withdrawal of consent, loss to follow up, death or the primary data cut off (DCO) for the analysis. | Cox proportional-hazards model; full analysis set (all treated patients) |
| Median Progression-Free Survival (Site Read, Investigator Assessment) | Same registered PFS definition and assessment framework. The registry separately identifies median PFS as a primary endpoint. | The ClinicalTrials.gov record contains one formal primary PFS analysis; it does not provide a separate formal statistical-analysis record for the median PFS endpoint. |
6. Primary Results: Progression-Free Survival
The registry reports one formal primary statistical analysis for progression-free survival. The analysis compared gefitinib with placebo using a Cox proportional-hazards model in the full analysis set, defined in the ClinicalTrials.gov record as all treated patients.
Hazard ratio for progression or death
95% CI: 0.65–1.13 · P = 0.273
Two-sided superiority analysis; full analysis set.
| Primary endpoint | Comparison | Analysis population | Method | Effect estimate |
|---|---|---|---|---|
| Progression-Free Survival (Site Read, Investigator Assessment) | Gefitinib vs Placebo | Full analysis set (all treated patients) | Cox proportional-hazards model | HR 0.86 (95% CI 0.65–1.13), P = 0.273 |
The estimated hazard ratio of 0.86 means that the fitted Cox model estimated a lower instantaneous rate of progression or death in the gefitinib group relative to the placebo group, with the model-based estimate corresponding to approximately a 14% lower estimated hazard.
That statement does not mean that 14% of patients avoided progression, that individual patients experienced a 14% reduction in their personal risk, or that the absolute probability of progression or death was reduced by 14 percentage points. A hazard ratio is a relative, model-based time-to-event measure.
The 95% confidence interval of 0.65–1.13 describes uncertainty around the estimated hazard ratio under the statistical model and sampling framework. Because the interval extends across 1, the data are compatible with a range of relative effects that includes no difference in hazard as well as effects in either direction.
The P = 0.273 value is a measure of statistical evidence against the null hypothesis specified for the superiority comparison. It is not a measure of the size, importance, or probability of the treatment effect. The magnitude of the estimate is described by the hazard ratio, while its statistical uncertainty is described by the confidence interval.
Because the analysis uses a Cox proportional-hazards model, interpretation of a single hazard ratio also depends on the proportional-hazards framework being a reasonable description of the treatment-specific hazards over the relevant follow-up. The ClinicalTrials.gov record does not report a formal assessment of that assumption.
7. Secondary Efficacy Results
The registry contains formal statistical analyses for several secondary endpoints. Time-to-event outcomes were analyzed with Cox proportional-hazards models, while binary outcomes were analyzed with logistic regression. The effect measures were hazard ratios and odds ratios, respectively.
Overall Survival
Median Overall Survival at Time of PFS Analysis
95% CI: 1.05–2.52 · P = 0.029
Gefitinib vs placebo; full analysis set; Cox proportional-hazards model.
The reported hazard ratio of 1.62 means that the fitted model estimated a higher instantaneous rate of death in the gefitinib group than in the placebo group during the analyzed follow-up. Numerically, the estimated hazard was 62% higher under the fitted model; this is not the same as saying that 62% more patients died.
The 95% CI of 1.05–2.52 indicates uncertainty around that estimate and lies above 1. The reported P = 0.029 indicates statistical evidence against the null hypothesis used for this superiority analysis, but the P-value itself does not quantify the magnitude of the effect.
Because this is a time-to-event endpoint, censoring and the proportional-hazards model are important to interpretation. The ClinicalTrials.gov record does not provide enough information to assess those features beyond the reported Cox analysis.
Objective Response Rate
| Endpoint | Method | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|---|
| Objective Response Rate (ORR) (Site Read Data) | Logistic regression | Odds ratio | 0.92 | 0.55–1.55 | 0.760 |
| Disease Control Rate (DCR) | Logistic regression | Odds ratio | 1.39 | 0.74–2.62 | 0.308 |
For ORR, the odds ratio of 0.92 compares the odds of the binary response outcome between the randomized groups under the logistic model. An odds ratio of 1 would represent equal odds; an odds ratio of 0.92 is close to 1. The 95% CI of 0.55–1.55 spans 1, and the reported P-value is 0.760.
For DCR, the odds ratio of 1.39 indicates higher estimated odds of the binary disease-control outcome in the gefitinib group under the fitted model. The 95% CI of 0.74–2.62 spans 1, and the P-value is 0.308. An odds ratio should not be interpreted as a risk ratio or percentage-point difference in response probability.
Quality-of-Life Endpoints
| Endpoint | Endpoint type | Analysis population | Method | Effect estimate | 95% CI | P-value |
|---|---|---|---|---|---|---|
| Improvement in Trial Outcome Index | Binary | Evaluable-for-QoL | Logistic regression | OR 0.92 | 0.53–1.59 | 0.770 |
| Time to Worsening in Trial Outcome Index | Time-to-event | Evaluable-for-QoL | Cox proportional-hazards model | HR 0.91 | 0.68–1.21 | 0.507 |
| Improvement in FACT-L Total Score | Binary | Evaluable-for-QoL | Logistic regression | OR 0.91 | 0.54–1.53 | 0.725 |
| Time to Worsening in FACT-L Total Score | Time-to-event | Evaluable-for-QoL | Cox proportional-hazards model | HR 0.92 | 0.69–1.23 | 0.575 |
| Improvement in Lung Cancer Subscale | Binary | Evaluable-for-QoL | Logistic regression | OR 1.01 | 0.61–1.68 | 0.959 |
| Time to Worsening in Lung Cancer Subscale | Time-to-event | Evaluable-for-QoL | Cox proportional-hazards model | HR 0.89 | 0.66–1.20 | 0.437 |
The quality-of-life analyses illustrate why the analysis population matters. These endpoints were evaluated in the evaluable-for-QoL population rather than the full analysis set used for the primary PFS analysis. The binary endpoints ask whether improvement occurred, whereas the time-to-worsening endpoints treat deterioration as a time-to-event outcome.
8. Statistical Methodology
Cox proportional-hazards model
The registry reports Cox proportional-hazards models for the primary PFS analysis and for the secondary overall-survival and quality-of-life time-to-worsening analyses. The model relates the instantaneous event rate to treatment and potentially other covariates according to the specified analysis model.
For a binary treatment indicator, exp(β) represents the hazard ratio comparing the treatment groups under the fitted model.
The key advantage of a time-to-event model is that it can incorporate different follow-up times and right-censored observations. A participant who has not experienced progression or death by the end of observable follow-up can still contribute information up to the censoring time.
Hazard ratio
The hazard ratio is the principal effect measure for the time-to-event analyses in the registry. Values below 1 correspond to a lower estimated instantaneous event rate in the first group of the comparison; values above 1 correspond to a higher estimated instantaneous event rate.
These estimates point in different directions for progression-free survival and overall survival. They should therefore be interpreted separately rather than collapsed into a single summary of the trial.
Logistic regression
Logistic regression was used for binary endpoints including objective response rate, disease control rate, and improvement in quality-of-life measures. The corresponding effect measure was the odds ratio.
An OR of 1 represents equal odds. The OR should not be read as a percentage change in the probability of the outcome unless the underlying probabilities and model context justify that interpretation.
Intention-to-treat analysis
The registry-reported analysis text identifies intention-to-treat analysis as an analysis concept for the primary PFS and secondary overall-survival analyses. The primary PFS statistical-analysis record describes its analysis population as the full analysis set (all treated patients). Because the ClinicalTrials.gov record uses both terms, they should not be silently treated as identical definitions beyond what the record explicitly states.
Kaplan-Meier estimation
Kaplan-Meier estimation is the standard descriptive framework for time-to-event outcomes such as PFS and OS because it accounts for censoring and estimates the event-free survival function over time. The registry-reported statistical-analysis fields specifically identify Cox proportional-hazards models for the formal comparisons; they do not provide a separate Kaplan-Meier analysis field or numerical median estimates.
Here di is the number of events at time ti, and ni is the number at risk immediately before that time.
9. Statistical Methods Explained
Why was a Cox model used for progression-free survival?
PFS is a time-to-event endpoint: participants can experience progression or death at different times, while some observations may be censored. A Cox model allows the treatment comparison to use the timing of events rather than reducing the endpoint to a simple yes/no outcome at one fixed date. Its primary effect measure is the hazard ratio.
What does a PFS hazard ratio of 0.86 mean?
Under the fitted model, 0.86 represents the estimated hazard of progression or death in the gefitinib group relative to the placebo group. Equivalently, 1 − 0.86 = 0.14, so the estimate corresponds to approximately a 14% lower estimated hazard. It does not represent a 14% absolute reduction in the probability of progression or death.
Why does the confidence interval matter?
The point estimate is only one summary of the treatment comparison. The 95% CI of 0.65–1.13 for PFS shows the uncertainty surrounding the estimated hazard ratio. Because the interval includes 1, the data do not exclude a null treatment difference under the stated model and confidence-interval framework.
Why is the P-value not an effect size?
The P-value addresses statistical evidence against a null hypothesis. It is affected by both the observed data and the amount of information available. The magnitude of an effect is better described by the effect estimate itself, such as HR 0.86 or OR 0.92, together with its confidence interval.
Why are odds ratios used for response endpoints?
Objective response and disease control are binary outcomes. Logistic regression models the probability of a binary outcome and naturally expresses the treatment contrast as an odds ratio. An OR of 0.92 for ORR is not equivalent to saying that response probability was 8% lower.
Why are the quality-of-life analyses based on a different population?
The quality-of-life endpoints were analyzed in the evaluable-for-QoL population, while the primary PFS analysis was based on the full analysis set described as all treated patients. This distinction matters because restricting an analysis to evaluable participants can change the population contributing information to the estimate.
Why should PFS and OS be interpreted separately?
PFS measures time to progression or death, whereas OS measures time to death. They are related but distinct endpoints. The IMPRESS registry results illustrate this directly: the reported PFS HR is 0.86, while the reported OS HR is 1.62. A single overall statement about "survival benefit" would obscure the difference between these endpoints.
10. Secondary Endpoint Interpretation
Response
The ORR odds ratio was 0.92 with a 95% CI of 0.55–1.55 and P = 0.760. The confidence interval spans both sides of an odds ratio of 1.
Disease control
The DCR odds ratio was 1.39 with a 95% CI of 0.74–2.62 and P = 0.308. The estimate is above 1, but its confidence interval includes 1.
Trial Outcome Index
Improvement had OR 0.92, while time to worsening had HR 0.91. Both confidence intervals include their respective null values.
FACT-L
Improvement had OR 0.91, while time to worsening had HR 0.92. These estimates describe different quality-of-life questions and should not be treated as interchangeable.
The Lung Cancer Subscale results similarly separate improvement from time to worsening. Improvement had an OR of 1.01 (95% CI 0.61–1.68; P = 0.959), while time to worsening had an HR of 0.89 (95% CI 0.66–1.20; P = 0.437).
11. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment arm as affected participants divided by participants at risk. These figures are reported separately from the efficacy analyses.
| Safety measure | Gefitinib | Placebo |
|---|---|---|
| Serious adverse events | 38/132 | 28/132 |
The serious-adverse-event figures show the number of affected participants and the number at risk in each reported arm. They should not be converted into a relative risk, odds ratio, or hypothesis test without additional analysis assumptions and the underlying event definitions.
Safety and efficacy answer different statistical questions. The primary PFS comparison estimates a treatment effect on a time-to-event endpoint, whereas the reported serious-adverse-event counts describe the number of affected participants among those at risk. A treatment comparison therefore cannot be summarized by combining the PFS hazard ratio with the serious-adverse-event counts into a single numerical "benefit-risk" measure.
12. Analysis Populations
| Analysis | Population reported | Why it matters |
|---|---|---|
| Primary PFS | Full analysis set (all treated patients) | The reported treatment effect is tied to this analysis population. |
| Overall survival | Full analysis set | The OS hazard ratio is estimated in the full analysis set. |
| ORR / DCR | Full analysis set | Binary efficacy comparisons use the full analysis set. |
| Quality-of-life endpoints | Evaluable-for-QoL | Only participants meeting the registry's QoL-evaluable definition contribute to these analyses. |
The distinction between analysis populations is statistically important. Randomization establishes comparability at allocation, but the set of participants contributing to an individual analysis can differ. The ClinicalTrials.gov record explicitly identifies this difference between the full analysis set and the evaluable-for-QoL population.
13. Endpoint Timing and Assessment
| Endpoint group | Registry time frame | Type |
|---|---|---|
| Progression-Free Survival | Radiologic evaluations were carried out every 6 weeks from randomization until documented progression, withdrawal of consent, loss to follow up, death or the primary data cut off (DCO) for the analysis. | Time-to-event |
| Overall Survival | Baseline and then every 6 weeks after randomization until objective disease progression. OS is then assessed 8 weekly following PFS progression up to PFS analysis data cut off. | Time-to-event |
| Quality-of-life endpoints | At visits 2-8, then every 6 weeks until progression, at progression or treatment discontinuation, and every 8 weeks after progression until PFS analysis data cut off. | Binary or time-to-event |
14. Crossover, Factorial Design, Interim Analysis, and Bayesian Methods
The ClinicalTrials.gov record identifies a two-arm randomized parallel design. They do not report a factorial design, crossover analysis, Bayesian statistical method, or interim-analysis procedure.
| Design topic | What the ClinicalTrials.gov record supports |
|---|---|
| Non-inferiority margin | Not reported in the ClinicalTrials.gov record; the posted hypothesis type is superiority. |
| Crossover | No crossover procedure is reported in the ClinicalTrials.gov record. |
| Factorial design | No; the registered design model is parallel with 2 arms. |
| Multiplicity procedure | Not reported in the ClinicalTrials.gov record. |
| Interim analysis | No interim-analysis procedure is reported in the ClinicalTrials.gov record. |
| Missing-data / imputation method | No specific imputation method is reported in the ClinicalTrials.gov record. |
| Stratification | No stratification factors are reported in the ClinicalTrials.gov record. |
| Bayesian methods | No Bayesian method is reported in the ClinicalTrials.gov record. |
15. Confidence Intervals Across the Trial
The IMPRESS results illustrate why confidence intervals are essential alongside point estimates. They describe the statistical precision of an estimated effect and allow the reader to see which effect sizes remain compatible with the data under the stated framework.
| Endpoint | Effect | 95% CI | Null value |
|---|---|---|---|
| Progression-Free Survival | HR 0.86 | 0.65–1.13 | 1 |
| Overall Survival | HR 1.62 | 1.05–2.52 | 1 |
| Objective Response Rate | OR 0.92 | 0.55–1.55 | 1 |
| Disease Control Rate | OR 1.39 | 0.74–2.62 | 1 |
| Improvement in Trial Outcome Index | OR 0.92 | 0.53–1.59 | 1 |
| Time to Worsening in Trial Outcome Index | HR 0.91 | 0.68–1.21 | 1 |
| Improvement in FACT-L Total Score | OR 0.91 | 0.54–1.53 | 1 |
| Time to Worsening in FACT-L Total Score | HR 0.92 | 0.69–1.23 | 1 |
| Improvement in Lung Cancer Subscale | OR 1.01 | 0.61–1.68 | 1 |
| Time to Worsening in Lung Cancer Subscale | HR 0.89 | 0.66–1.20 | 1 |
For hazard ratios and odds ratios, the null value is 1. A confidence interval crossing 1 means that a no-difference effect remains among the values compatible with the reported confidence-interval procedure. It does not mean that every value inside the interval is equally plausible, nor does it establish that there is no treatment effect.
16. P-Values and Statistical Evidence
Primary PFS
HR 0.86; 95% CI 0.65–1.13; P = 0.273.
Overall survival
HR 1.62; 95% CI 1.05–2.52; P = 0.029.
ORR
OR 0.92; 95% CI 0.55–1.55; P = 0.760.
DCR
OR 1.39; 95% CI 0.74–2.62; P = 0.308.
The P-values should be interpreted in conjunction with the effect estimates, confidence intervals, endpoint definitions, and analysis populations. In particular, the statistical evidence for one endpoint does not automatically determine the interpretation of another endpoint.
A P-value answers a question about compatibility with a null hypothesis under the specified statistical framework. It does not answer how large an effect is, whether an effect is clinically important, or what probability a hypothesis is true. Those questions require different information.
17. Limitations
- Registry-level information: This analysis is limited to the ClinicalTrials.gov record. Detailed protocol and statistical-analysis-plan information not present in that data has not been inferred.
- Primary endpoint reporting: The registry separately lists PFS and median PFS as primary endpoints, but the ClinicalTrials.gov record contains one formal primary PFS analysis rather than a separate formal analysis for median PFS.
- No median PFS value: Although median PFS is registered as an endpoint, the ClinicalTrials.gov record does not provide a median PFS estimate or confidence interval.
- Limited survival detail: The ClinicalTrials.gov record provides a hazard ratio, confidence interval, and P-value for the OS analysis but do not provide a median OS estimate or Kaplan-Meier survival probabilities.
- Proportional-hazards assumption: Cox-model hazard ratios rely on the model's assumptions. The ClinicalTrials.gov record does not report a formal diagnostic for proportional hazards.
- Different analysis populations: Primary PFS and OS analyses use the full analysis set, while quality-of-life analyses use the evaluable-for-QoL population. Estimates therefore do not all describe exactly the same participant set.
- Multiplicity: The trial has multiple primary and secondary endpoints, but the ClinicalTrials.gov record does not report a multiplicity-adjustment procedure. Individual P-values should therefore be interpreted in the context of the complete testing strategy rather than treated as isolated proof of an overall treatment effect.
- Missing-data handling: No specific missing-data or imputation strategy is reported in the ClinicalTrials.gov record.
- Safety detail: The ClinicalTrials.gov record is limited to serious adverse events by arm and does not provide a broader adverse-event profile.
- No subgroup results: The ClinicalTrials.gov record does not contain subgroup effect estimates, so no subgroup conclusions are drawn.
18. Why This Trial Matters Statistically
IMPRESS is a useful statistical teaching case because the registry contains multiple endpoint types analyzed with two major regression frameworks: Cox proportional-hazards models for time-to-event outcomes and logistic regression for binary outcomes. It also illustrates why the numerical interpretation of a trial depends on the endpoint, effect measure, confidence interval, and analysis population.
| Concept | How it appears in IMPRESS |
|---|---|
| Randomization | The study used randomized allocation. |
| Parallel design | The registered design model has 2 parallel arms. |
| Quadruple masking | The registry reports quadruple masking. |
| Time-to-event analysis | PFS, OS, and time-to-worsening endpoints were analyzed as time-to-event outcomes. |
| Cox model | Used for PFS, OS, and time-to-worsening analyses. |
| Hazard ratio | Used as the effect measure for time-to-event comparisons. |
| Logistic regression | Used for ORR, DCR, and improvement endpoints. |
| Odds ratio | Used as the effect measure for binary endpoints. |
| Confidence intervals | 95% two-sided confidence intervals were reported for the posted estimates. |
| P-values | Reported for each posted statistical analysis. |
| Analysis populations | Full analysis set and evaluable-for-QoL populations are distinguished. |
| Multiple endpoints | Two registered primary endpoints and multiple secondary endpoints are present. |
19. A Deeper Statistical Reading of the Results
PFS and OS should not be collapsed into one conclusion
The primary PFS analysis produced an HR of 0.86, whereas the secondary OS analysis produced an HR of 1.62. These are estimates for different endpoints and should be interpreted independently. PFS incorporates objective progression or death; OS is based on death. A treatment comparison can therefore produce different estimates across these endpoints.
Relative measures do not replace absolute measures
The registry results are primarily expressed as hazard ratios and odds ratios. These are useful relative measures, but they do not directly provide the absolute probability of progression, death, response, or disease control. The ClinicalTrials.gov record does not include enough information to calculate valid absolute treatment differences without making additional assumptions.
The direction of an effect depends on the endpoint
For the time-to-event analyses, an HR below 1 means a lower estimated event hazard in the gefitinib group, while an HR above 1 means a higher estimated event hazard. For the binary analyses, an OR below 1 means lower estimated odds of the specified outcome, while an OR above 1 means higher estimated odds. The meaning of the direction therefore depends on both the effect measure and the event being modeled.
Statistical significance and precision are different concepts
The OS estimate has a 95% CI of 1.05–2.52 and P = 0.029, while the PFS estimate has a 95% CI of 0.65–1.13 and P = 0.273. These statistics describe different relationships between the estimated effect, its uncertainty, and the null value. They should not be interpreted simply as labels attached to the entire trial.
Quality-of-life endpoints require careful population interpretation
Quality-of-life improvement and worsening analyses were conducted in the evaluable-for-QoL population. This means that their estimates describe a selected analysis population rather than necessarily the same population used for the primary PFS analysis. The distinction is especially important when comparing effect estimates across endpoint families.
20. What the Hazard Ratio Does — and Does Not — Mean
The PFS HR of 0.86 means that the estimated instantaneous rate of progression or death in the gefitinib group was 0.86 times that in the placebo group under the fitted Cox model.
It does not mean that 14% of patients benefited, that progression was delayed by a fixed percentage for every patient, or that the absolute probability of progression or death was reduced by 14 percentage points.
The 95% CI of 0.65–1.13 quantifies uncertainty around the estimated PFS hazard ratio. It is not a range of outcomes that individual patients can expect and should not be interpreted as a distribution of individual treatment effects.
The OS HR of 1.62 means that the fitted model estimated a higher instantaneous rate of death in the gefitinib group than in the placebo group. This is a model-based relative comparison and is not equivalent to a 62% absolute increase in mortality.
21. Statistical Interpretation vs Clinical Interpretation
Statistical interpretation
The primary PFS Cox analysis estimated HR 0.86 with a 95% CI of 0.65–1.13 and P = 0.273. The secondary OS analysis estimated HR 1.62 with a 95% CI of 1.05–2.52 and P = 0.029.
Clinical interpretation
Clinical meaning cannot be reduced to a single P-value or hazard ratio. It requires consideration of the endpoint, magnitude and precision of the estimate, absolute outcomes when available, safety, and the population represented by the trial.
The ClinicalTrials.gov record does not provide enough information to make a broader clinical assessment beyond these statistical observations. In particular, median survival values, detailed baseline characteristics, subgroup estimates, and a comprehensive safety profile are not included in the ClinicalTrials.gov record.
22. Results Not Reported in the Supplied Registry Data
| Information | Status in the ClinicalTrials.gov record |
|---|---|
| Median PFS estimate | Not provided in the ClinicalTrials.gov record. |
| Median OS estimate | Not provided. |
| Kaplan-Meier survival probabilities | Not provided. |
| Baseline characteristics | Not provided. |
| Subgroup estimates | Not provided. |
| Detailed adverse-event profile | Not provided beyond serious adverse events by arm. |
| Interim-analysis procedure | Not reported in the ClinicalTrials.gov record. |
| Multiplicity adjustment | Not reported in the ClinicalTrials.gov record. |
| Missing-data / imputation strategy | Not reported in the ClinicalTrials.gov record. |
| Bayesian methods | Not reported in the ClinicalTrials.gov record. |
These omissions are important because a statistical analysis should distinguish between what was measured and formally reported and what would require additional trial documents or source data.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Statistical Calculators
25. Sources
- ClinicalTrials.gov: NCT01544179 — IMPRESS.
- Linked publication: PubMed record for PMID 26159065: PubMed PMID 26159065.
Continue with Clinical Biostats
Explore statistical tutorials and calculators covering survival analysis, regression, effect measures, confidence intervals, and clinical-trial methodology.
26. Record Summary
IMPRESS provides a useful example of how a randomized phase 3 clinical trial can generate different statistical questions across multiple endpoint types. The primary progression-free survival endpoint was analyzed with a Cox proportional-hazards model and produced an HR of 0.86 (95% CI 0.65–1.13; P = 0.273). A secondary overall-survival analysis produced an HR of 1.62 (95% CI 1.05–2.52; P = 0.029). Binary outcomes such as objective response rate and disease control rate were analyzed with logistic regression, while quality-of-life endpoints included both binary improvement measures and time-to-worsening analyses.
The most important statistical lesson is that these results cannot be reduced to one number. Each estimate belongs to a particular endpoint, analysis population, model, and effect measure. Hazard ratios describe relative event hazards; odds ratios describe relative odds; confidence intervals describe statistical precision; and P-values describe evidence against a specified null hypothesis. The differences among these quantities are central to interpreting the IMPRESS results accurately.