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
MITO-8 was a randomized, unmasked, phase 3 clinical trial in ovarian cancer with a crossover design. The registry records 215 participants and two treatment arms, with overall survival at 18 months registered as the primary endpoint.
| Feature | MITO-8 |
|---|---|
| Trial acronym | MITO-8 |
| ClinicalTrials.gov identifier | NCT00657878 |
| Phase | Phase 3 |
| Condition | Ovarian Cancer |
| Allocation | Randomized |
| Design model | Crossover |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 215 |
| Number of arms | 2 |
| Trial status | UNKNOWN |
| Start date | 2008-11 |
| Primary completion date | 2023-12 |
| Lead sponsor | National Cancer Institute, Naples |
| Sponsor type | OTHER |
2. Clinical Question
The central statistical question is whether the randomized treatment strategies evaluated in MITO-8 differ in overall survival at the registered 18-month time frame in patients with ovarian cancer.
Population
Patients enrolled in the phase 3 trial for the condition recorded as ovarian cancer.
Intervention
The randomized treatment strategies involved chemotherapy agents listed in the registry: stealth liposomal doxorubicin, carboplatin, paclitaxel, topotecan, and gemcitabine.
Comparator
The trial contains two randomized treatment arms. The registry extract identifies the interventions but does not provide an arm-by-arm regimen mapping.
Primary question
What is the comparative effect of the randomized treatment strategies on overall survival over the registered 18-month time frame?
3. Trial Design
A crossover design changes the interpretation of a randomized comparison because treatment received after the crossover can differ from the originally assigned treatment. For an overall-survival endpoint, this is particularly important: survival is measured over time, while treatment exposure can change during that same period.
Why randomization and crossover need to be considered together
Randomization establishes the initial comparison. Crossover subsequently introduces treatment exposure that is no longer determined solely by the original assignment. Consequently, an analysis based on randomized assignment answers a different question from an analysis based on treatment actually received.
The first preserves the treatment assignment created by randomization. The second can be affected by why, when, and how participants crossed over. A crossover therefore requires explicit attention to the estimand and analysis population.
4. Endpoints
| Endpoint | Registry definition / time frame | Role |
|---|---|---|
| Overall survival | Overall survival; 18 months | Primary endpoint |
The registry identifies overall survival as the primary endpoint and specifies an 18-month time frame. No additional endpoint definitions are included in the registry information available for this analysis.
What an overall-survival endpoint measures
Overall survival is a time-to-event endpoint in which the event is death. Unlike a simple binary endpoint assessed at a single visit, time-to-event analysis uses both the occurrence and timing of events and can accommodate participants whose complete event time is not observed during the study period.
The exact statistical origin and censoring conventions should follow the prespecified protocol or statistical analysis plan when those documents are available.
5. Statistical Methodology
Because overall survival is a time-to-event endpoint, the standard analytical framework would use methods designed for censored survival data. A crossover trial adds a second methodological issue: the analysis must distinguish the effect associated with randomized assignment from effects associated with treatment exposure after crossover.
Kaplan-Meier estimation
Kaplan-Meier estimation is commonly used to describe the survival distribution for each randomized group. It estimates the probability of remaining event-free through time while retaining information from participants who are censored before experiencing the event.
Here di represents the number of deaths at event time ti, while ni represents the number at risk immediately before that event time.
Log-rank comparison
A log-rank test is a standard method for comparing survival distributions between randomized groups. It evaluates whether the observed pattern of events over follow-up differs between groups under the test's assumptions.
Cox proportional-hazards model
A Cox model is commonly used to estimate a hazard ratio comparing two treatment groups. The hazard ratio is a relative measure of the instantaneous event rate under the model; it is not an absolute survival probability.
The interpretation depends on the model and the time-to-event data. It does not mean that the same percentage of individual participants experienced a reduction in their personal probability of death.
Crossover-aware analysis
For a crossover study, the most important analytical question is whether the primary comparison is defined by original randomized assignment or by subsequent treatment exposure. An intention-to-treat analysis maintains the randomized groups throughout follow-up. A treatment-received analysis instead incorporates exposure after randomization and may be affected by the reasons and timing of crossover.
When crossover is substantial, methods such as rank-preserving structural failure time approaches or inverse-probability weighting can sometimes be considered when the scientific objective specifically concerns treatment received without crossover. These methods require assumptions and detailed individual-level data. They should not be treated as interchangeable with the primary randomized comparison.
18-month survival
Because the registry specifies overall survival at 18 months, an analysis can report the estimated survival probability at that time point, together with an appropriate confidence interval. This is distinct from reporting a median survival time or a hazard ratio: each summarizes a different aspect of the survival experience.
Time-specific estimate
The 18-month survival probability answers the question: what proportion is estimated to remain alive at 18 months?
Hazard ratio
A hazard ratio compares modeled instantaneous event rates over follow-up rather than directly comparing survival probabilities at 18 months.
6. Planned Analysis
The registry identifies overall survival at 18 months as the primary endpoint, but no formal statistical analyses are posted in the ClinicalTrials.gov record available here. The appropriate analysis for this endpoint would ordinarily use time-to-event methods rather than a simple comparison of proportions.
Primary endpoint
Registered time frame: 18 months
A conventional analysis would estimate survival over time, compare the randomized groups using a survival-analysis framework, and report an effect measure with statistical uncertainty.
What would normally be reported
| Component | Purpose |
|---|---|
| Kaplan-Meier survival estimates | Describe the survival experience over follow-up and provide a time-specific estimate at 18 months. |
| Confidence interval | Quantify statistical uncertainty around the estimated survival probability or treatment-effect estimate. |
| Log-rank test | Provide a formal comparison of survival distributions between randomized groups. |
| Hazard ratio | Summarize the relative event rate between treatment groups under a Cox model. |
| Censoring rules | Define how participants without an observed death during follow-up contribute information. |
| Crossover analysis | Distinguish the randomized treatment effect from analyses incorporating post-randomization treatment changes. |
Why a binary 18-month comparison alone is incomplete
If participants have different lengths of follow-up or are censored before 18 months, simply classifying everyone as alive or dead at 18 months can discard information. Time-to-event methods retain the timing of observed deaths and appropriately incorporate censored observations under their assumptions.
7. Statistical Methods Explained
Why is overall survival analyzed as a time-to-event endpoint?
Because death can occur at different times, the timing of the event contains information. Survival analysis uses that timing rather than reducing the entire follow-up experience to a single binary outcome.
Why use Kaplan-Meier estimation?
Kaplan-Meier estimation provides a way to describe survival over time when some participants have not experienced the event by the end of their observed follow-up. Those participants are censored rather than treated as though they had experienced the event.
What does a hazard ratio measure?
A hazard ratio compares the modeled instantaneous event rates between groups. An HR of 1 would indicate equal modeled hazards; values below 1 indicate a lower estimated hazard in the numerator treatment group, while values above 1 indicate a higher estimated hazard. It is not the same as an 18-month survival difference.
Why does crossover matter for an intention-to-treat analysis?
Intention-to-treat analysis preserves the original randomized comparison even after participants change treatment. That preserves the causal structure created by randomization, but the observed difference in treatment exposure can make the randomized-group contrast less representative of a hypothetical comparison in which nobody crossed over.
Why can treatment-received analyses be difficult?
After crossover, treatment received is no longer necessarily independent of the participant's disease course. The decision or opportunity to cross over can depend on post-randomization information. Consequently, simply regrouping participants according to the treatment they eventually received can introduce post-randomization selection effects.
What does an 18-month survival estimate tell us?
It describes the estimated probability of remaining alive at a specified time point. It does not by itself describe the entire survival curve, the median survival time, or the relative hazard over follow-up.
8. Limitations
- No posted statistical analyses: the ClinicalTrials.gov record does not provide numerical statistical results for the registered primary endpoint in the information used here.
- Crossover: crossover changes treatment exposure after randomization and complicates interpretation of analyses based on treatment actually received.
- Endpoint detail: the registry identifies overall survival and an 18-month time frame but does not provide additional endpoint-definition detail in the available record.
- Arm-specific regimen information: the record lists five chemotherapy interventions but does not provide the arm-by-arm treatment combinations in the information available here.
- Unmasked design: the registry records no masking. Lack of masking can affect behavior, assessment, or other post-randomization processes depending on the endpoint and study procedures.
- Hazard-ratio assumptions: if a Cox model is used, its interpretation depends on model assumptions, including the proportional-hazards framework.
- Censoring: Kaplan-Meier and Cox analyses rely on assumptions concerning the relationship between censoring and the underlying event process.
- Missing statistical documents: the registry information available here does not provide the detailed statistical analysis plan needed to establish the exact prespecified analysis population, censoring rules, hypothesis-testing framework, or multiplicity procedures.
9. Why This Trial Matters Statistically
MITO-8 is a useful teaching case because it combines a randomized phase 3 design with a crossover structure and a time-to-event primary endpoint. That combination illustrates why trial design and statistical analysis cannot be interpreted independently.
| Concept | How it appears in MITO-8 |
|---|---|
| Randomization | The trial uses randomized allocation between two treatment arms. |
| Crossover design | The registry classifies the design model as crossover. |
| Time-to-event analysis | Overall survival is the registered primary endpoint. |
| Time-specific endpoint | The registered overall-survival time frame is 18 months. |
| Kaplan-Meier estimation | Provides a standard framework for describing survival with censored observations. |
| Hazard ratio | Provides a model-based relative comparison of event rates when a Cox model is appropriate. |
| Intention-to-treat principle | Preserves the randomized assignment when assessing the treatment strategies. |
| Post-randomization treatment | Crossover means that treatment exposure can differ from the original assignment. |
| Censoring | Participants whose complete survival time is not observed require appropriate handling in time-to-event analysis. |
| Estimand definition | The crossover design makes it important to distinguish the effect of assignment from the effect of treatment actually received. |
Randomization establishes the comparison; crossover changes exposure
This distinction is one of the most important statistical lessons from the design. Randomization creates a comparison that can support causal inference about assignment. Once participants cross over, however, the treatment they receive during follow-up is partly determined by post-randomization events and decisions.
Why overall survival is particularly informative in a crossover trial
Overall survival follows participants beyond the initial treatment decision. That makes it clinically meaningful as a time-to-event endpoint, but it also means that every post-randomization treatment change can potentially influence the observed survival experience.
10. Statistical Interpretation of the Primary Endpoint
The primary analysis would seek to characterize survival over time in the two randomized treatment groups and specifically estimate survival at the registered 18-month time point.
An estimated 18-month survival probability would represent the model-free or model-assisted estimate of the proportion of participants expected to remain alive at 18 months, subject to the censoring and survival-analysis assumptions used.
An 18-month survival estimate would not describe survival at every later time point, would not by itself establish a hazard ratio, and would not show how treatment exposure after crossover affected individual participants.
A confidence interval around an estimated survival probability or treatment effect quantifies statistical uncertainty associated with sampling and the analysis framework. It does not describe the range of outcomes that every individual participant can experience.
If a hypothesis test is reported, its p-value addresses compatibility of the observed data with the null hypothesis under the specified test. It does not measure the magnitude or clinical importance of the treatment effect.
11. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The trial is randomized and uses a crossover design, with overall survival at 18 months registered as the primary endpoint. The appropriate analytical framework is therefore centered on time-to-event methods and explicit handling of post-randomization treatment changes.
Clinical interpretation
The registry identifies ovarian cancer as the study condition and chemotherapy agents as the listed interventions. Without posted statistical analyses, the registry record does not provide a numerical estimate of the comparative effect on the primary endpoint.
12. Related Tutorials
Learn more about the methods used in this trial:
13. Related Calculators
14. Sources
- ClinicalTrials.gov: MITO-8, NCT00657878.
- PubMed: PubMed record 29462248.
Continue through the Clinical Biostats knowledge graph
Connect the crossover design in MITO-8 with deeper statistical explanations and practical tools for clinical-trial analysis.
15. Record Summary
MITO-8 is a randomized phase 3 trial in ovarian cancer with 215 enrolled participants, two treatment arms, and a crossover design. The registry lists overall survival at 18 months as the primary endpoint and identifies stealth liposomal doxorubicin, carboplatin, paclitaxel, topotecan, and gemcitabine among the study interventions. The statistical structure is therefore centered on a randomized time-to-event comparison complicated by changes in treatment exposure after crossover.
The key methodological lesson is that the treatment effect associated with randomized assignment is not automatically identical to the effect associated with treatment actually received. For an overall-survival endpoint, Kaplan-Meier estimation, appropriate comparison of survival distributions, and model-based effect measures such as the hazard ratio can describe the randomized groups, while crossover-aware analyses require additional assumptions and detailed individual-level data.