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
IALT was a randomized phase 3 clinical trial evaluating combination chemotherapy compared with no treatment following surgery in patients with lung cancer. The registry identifies the study as a treatment trial and reports an enrollment of 3300 participants.
| Feature | IALT |
|---|---|
| Acronym | IALT |
| NCT identifier | NCT00002823 |
| Title | Combination Chemotherapy Compared With No Treatment Following Surgery in Treating Patients With Non-small Cell Lung Cancer |
| Condition | Lung Cancer |
| Phase | Phase 3 |
| Status | Completed |
| Allocation | Randomized |
| Primary purpose | Treatment |
| Enrollment | 3300 |
| Lead sponsor | Gustave Roussy, Cancer Campus, Grand Paris |
| Sponsor type | Other |
2. Clinical Question
The registry title defines the central clinical comparison as combination chemotherapy versus no treatment following surgery in patients with non-small cell lung cancer. The study's primary purpose is recorded as treatment.
Population
Patients with non-small cell lung cancer following surgery, as described by the registered study title.
Intervention
The registry lists cisplatin, etoposide, vinblastine sulfate, vindesine, vinorelbine tartrate, and radiation therapy among the study interventions.
Comparator
The registered title identifies no treatment following surgery as the comparison condition.
Primary question
Does combination chemotherapy following surgery produce different clinical outcomes than no treatment following surgery?
3. Trial Design
Randomization is the central statistical feature of this design. When participants are randomly allocated to treatment conditions, the comparison between groups can be interpreted as a randomized treatment comparison rather than simply an observational association, provided the randomized assignment and subsequent analysis are handled appropriately.
4. Endpoints
The ClinicalTrials.gov record does not list any registered primary endpoints for IALT. Consequently, no registry-defined endpoint name, event definition, assessment window, or endpoint-specific time frame can be stated from the record.
| Endpoint information | Registry record |
|---|---|
| Registered primary endpoints | None listed |
| Endpoint definitions | Not listed in the registered primary-endpoint section |
| Endpoint time frames | Not listed in the registered primary-endpoint section |
5. Planned Analysis
No statistical analyses are posted to ClinicalTrials.gov for this record. The registry also does not list registered primary endpoints. The statistical interpretation therefore focuses on the design and on the analytical framework that would ordinarily be appropriate for a randomized post-surgical treatment comparison.
How a randomized treatment endpoint would normally be analysed
The appropriate statistical method depends on the exact endpoint definition. For a binary endpoint measured at a prespecified time point, the treatment groups could ordinarily be compared using a risk difference, risk ratio, odds ratio, or an appropriate regression model. For a continuous endpoint, the analysis could use a difference in means or a regression-based comparison. For a time-to-event endpoint such as time to death or time to recurrence, Kaplan-Meier estimation, a log-rank comparison, and a Cox proportional-hazards model are commonly used.
The choice among these methods should be driven by the prespecified endpoint and estimand rather than selected after observing the results. In particular, time-to-event methods account for different lengths of follow-up and right censoring, whereas a simple comparison of proportions does not.
If the registered primary endpoint were a time-to-event outcome, a hazard ratio could summarize the relative event rate between randomized groups. An HR below 1 would indicate a lower estimated instantaneous event rate in the numerator group, while an HR above 1 would indicate a higher estimated instantaneous event rate.
Why the endpoint definition comes first
A statistical method cannot be interpreted independently of the endpoint it is intended to estimate. For example, overall survival, disease-free survival, recurrence, response, and treatment toxicity have different event definitions and therefore require different estimands and potentially different analysis populations or censoring rules.
Why randomization remains important
Randomization establishes the treatment comparison before outcomes are observed. It helps balance both measured and unmeasured baseline factors in expectation and provides the foundation for causal interpretation. Statistical adjustment can improve precision or account for prespecified design factors, but it does not replace the role of randomization.
6. Statistical Methodology
Randomized comparison
IALT is registered as a randomized study. The fundamental statistical comparison is therefore between outcomes observed after assignment to the randomized treatment conditions. The treatment effect is conceptually the difference attributable to assignment, rather than merely the difference between patients who happened to receive different treatments.
Intention-to-treat principle
For a randomized clinical trial, an intention-to-treat analysis ordinarily evaluates participants according to their randomized assignment. This preserves the initial randomization and generally provides the clearest estimate of the effect of assigning a treatment strategy.
The exact analysis population used in IALT is not specified in the registry information presented here, so no particular population can be attributed to the trial.
Time-to-event analysis
If an outcome was measured as time from randomization or surgery to an event, the Kaplan-Meier estimator would ordinarily be useful because it accommodates right-censored observations. A participant who has not experienced the event by the last available follow-up can contribute information up to that time without being treated as though the event had occurred.
Here, di is the number of events at time ti and ni is the number at risk immediately before that time.
Hazard ratios
If the primary endpoint were a time-to-event outcome, a Cox proportional-hazards model would commonly be used to estimate a hazard ratio. The hazard ratio is a relative measure of the instantaneous event rate and should not be interpreted as a percentage of patients cured, a relative reduction in cumulative incidence, or a difference in median survival unless those quantities are separately estimated.
Confidence intervals
A confidence interval describes statistical uncertainty around an estimated treatment effect under the specified analysis framework. A narrow interval generally indicates greater statistical precision than a wide interval. Precision depends on the amount of information available, including the number of participants and, for time-to-event analyses, the number and timing of observed events.
P-values
A p-value evaluates the compatibility of the observed data with a specified null hypothesis under the statistical model and testing procedure. It does not measure the magnitude of a treatment effect, the probability that a treatment is effective, or the clinical importance of an observed difference.
7. Statistical Methods Explained
Why does randomization matter in IALT?
Randomization creates treatment groups by an allocation mechanism that does not depend on participants' future outcomes. This is the key design feature supporting a causal comparison between combination chemotherapy and no treatment. With adequate randomization, systematic differences in prognosis should be balanced in expectation rather than being selected by investigators or patients.
Why would an intention-to-treat analysis usually be preferred for efficacy?
Analysing participants according to their randomized assignment maintains the treatment groups created by randomization. Excluding participants after randomization because of treatment discontinuation, nonadherence, or other post-randomization events can change the composition of the groups and introduce bias.
When would Kaplan-Meier analysis be appropriate?
Kaplan-Meier analysis is appropriate when the outcome is measured as time until an event and some participants have not experienced the event by the end of their available observation. It uses information from those participants up to their censoring time rather than requiring every participant to have an observed event.
What would a hazard ratio mean if survival were the endpoint?
A hazard ratio would compare the estimated instantaneous event rates between randomized groups. An HR below 1 would indicate a lower estimated event rate in the treatment group relative to the comparator under the fitted model. It would not by itself specify the absolute survival probability at a particular time.
Why should the endpoint be defined before selecting the statistical test?
The endpoint determines what quantity is being estimated. A binary response, continuous measurement, and time-to-event outcome have different statistical structures. Selecting the endpoint and analysis framework after seeing which comparison produces the most favorable result can increase false-positive risk and make the resulting inference less reliable.
Why is a p-value not a measure of effect size?
The p-value depends on both the observed difference and the amount of statistical information. A very large randomized trial can produce a small p-value for a relatively modest effect, while a smaller study can produce a large p-value even when the point estimate is substantial. Effect estimates and confidence intervals are therefore needed to understand magnitude and precision.
8. Sample Size and Information
The registered enrollment is 3300 participants. In a randomized phase 3 study, a sample of this size can provide substantial information, but the statistical information available for any particular endpoint depends not only on the number enrolled but also on the endpoint definition, event frequency, follow-up, censoring, treatment allocation, and analysis population.
| Design quantity | Registered value | Statistical significance |
|---|---|---|
| Enrollment | 3300 | Defines the planned size of the enrolled study population |
| Allocation | Randomized | Supports causal treatment comparison |
| Phase | 3 | Identifies the registered development phase |
| Primary purpose | Treatment | Identifies the principal study purpose |
Enrollment alone does not determine statistical power. For a time-to-event endpoint, for example, the number of observed events can be more directly related to precision than the number enrolled. For a binary endpoint, the expected event proportion and allocation ratio affect the amount of information available. Because no formal power calculation or endpoint-specific assumptions are listed here, a trial-specific power calculation cannot be stated.
9. Intervention Structure
The registry lists six intervention entries: cisplatin, etoposide, vinblastine sulfate, vindesine, vinorelbine tartrate, and radiation therapy.
| Registered intervention | Type |
|---|---|
| Cisplatin | Drug |
| Etoposide | Drug |
| Vinblastine sulfate | Drug |
| Vindesine | Drug |
| Vinorelbine tartrate | Drug |
| Radiation therapy | Radiation |
The registered title describes the intervention as combination chemotherapy following surgery. The intervention list contains several agents and radiation therapy, but the registry information presented here does not specify the dosing schedule, sequencing, number of treatment cycles, or which intervention combinations were assigned to particular participants.
10. Trial Timeline
Study start
The registered study start date is February 1995.
Randomized phase 3 study period
The study was registered as a randomized phase 3 treatment trial with an enrollment of 3300 participants.
Primary completion
The registered primary completion date is January 2001.
Study status
The ClinicalTrials.gov record identifies the study as completed.
11. What Statistical Evidence Would Be Needed to Quantify Treatment Effect?
A randomized design establishes the framework for comparing treatment groups, but treatment-effect quantification requires endpoint-specific observations. For each endpoint, the analysis would ordinarily report an effect estimate together with a measure of uncertainty.
Binary outcome
A risk difference, risk ratio, odds ratio, or regression-based estimate could quantify the contrast between randomized groups.
Continuous outcome
A difference in means or an adjusted regression coefficient could describe the treatment contrast.
Time-to-event outcome
Kaplan-Meier estimates, survival probabilities, median event times, and a hazard ratio could describe the treatment contrast.
Safety outcome
Adverse-event rates, event counts, severity categories, and treatment-emergent event comparisons could be summarized by randomized group.
The appropriate estimand should be specified before analysis. For example, a time-to-event estimand may concern the difference in survival over a particular period, the ratio of hazards under a model, or another prespecified summary. These are not interchangeable quantities.
12. Statistical Interpretation of Randomized Treatment Effects
Randomization supports a comparison in which differences in outcomes between groups can be attributed to the assigned treatment strategy, subject to adherence, follow-up, analysis, and other features of the study. It is stronger for causal inference than simply comparing people who received different treatments outside a randomized design.
An effect estimate answers the question of how different the outcomes were between treatment groups under the chosen estimand. The magnitude and direction of that estimate should be considered separately from whether a hypothesis test crossed a prespecified significance threshold.
The confidence interval communicates how precisely the treatment effect was estimated. Two studies can have similar point estimates but substantially different confidence intervals if their information content differs.
A p-value addresses evidence against a null hypothesis under a specified statistical model. It does not determine whether an effect is clinically important. Clinical importance requires consideration of effect magnitude, uncertainty, endpoint relevance, treatment burden, and the context in which the result is applied.
13. Limitations and Interpretation Issues
- Registered endpoints: the ClinicalTrials.gov record does not list registered primary endpoints, so a trial-specific primary estimand cannot be identified from the record.
- No posted statistical analyses: no formal statistical analyses are posted to ClinicalTrials.gov for this record.
- No numerical efficacy results: the registry information presented here does not provide treatment-effect estimates, confidence intervals, p-values, event counts, or survival estimates.
- Intervention structure: several drugs and radiation therapy are listed, but the available registry information does not specify the detailed treatment-assignment structure.
- Randomization ratio: the record identifies the study as randomized but does not specify the allocation ratio in the information presented here.
- Blinding: no blinding status is specified in the available trial information.
- Stratification: no randomization or analysis stratification factors are specified in the available trial information.
- Analysis populations: no specific intention-to-treat, per-protocol, safety, or other analysis population definitions are provided in the available trial information.
- Follow-up: the study start and primary completion dates are recorded, but no endpoint-specific follow-up duration is provided.
- Multiplicity: no endpoint hierarchy, multiplicity procedure, or familywise error-control strategy is specified in the available trial information.
- Interim analysis: no interim-analysis plan or stopping boundary is specified in the available trial information.
- Missing data: no missing-data or imputation strategy is specified in the available trial information.
- Bayesian methods: no Bayesian analysis is specified in the available trial information.
14. Why This Trial Matters Statistically
IALT is a useful statistical teaching case because its registered design illustrates the central structure of a large randomized phase 3 treatment comparison: a defined clinical population, randomized allocation, a post-surgical treatment strategy, and a no-treatment comparator.
| Statistical concept | How it appears in the registered study |
|---|---|
| Randomization | The allocation is registered as randomized. |
| Phase 3 design | The study is registered as phase 3. |
| Treatment comparison | The registered title describes combination chemotherapy compared with no treatment following surgery. |
| Large enrollment | 3300 participants are registered as enrolled. |
| Endpoint definition | No registered primary endpoints are listed. |
| Statistical analysis | No statistical analyses are posted to ClinicalTrials.gov. |
| Clinical interpretation | Requires endpoint-specific effect estimates and uncertainty measures before quantitative treatment conclusions can be drawn. |
The study also demonstrates why trial design and analysis cannot be separated. Randomization determines the basic causal comparison, but the interpretation of that comparison depends on prespecified endpoints, analysis populations, missing-data handling, follow-up, and appropriate statistical models.
15. Safety and Adverse Events
The available ClinicalTrials.gov information does not provide serious adverse-event counts by randomized treatment group. Consequently, a numerical safety comparison cannot be constructed from the registry information presented here.
| Safety information | Registry record |
|---|---|
| Serious adverse events by arm | Not reported in the available trial information |
| Adverse-event rates by arm | Not reported in the available trial information |
| Grade-specific adverse events | Not reported in the available trial information |
| Treatment-related adverse events | Not reported in the available trial information |
In a randomized treatment trial, safety should ordinarily be analysed separately from efficacy. The relevant unit may be the participant, the adverse-event episode, the treatment exposure, or the time to a first event, depending on the safety endpoint. Counts alone may also be insufficient when exposure duration differs between groups.
16. Clinical Interpretation vs Statistical Interpretation
Statistical interpretation
The study design is randomized and phase 3, with 3300 enrolled participants. The available registry information does not provide endpoint-specific statistical estimates or formal statistical analyses.
Clinical interpretation
The registered clinical question concerns combination chemotherapy compared with no treatment following surgery in lung cancer. Quantitative clinical conclusions require endpoint-specific outcome data.
This distinction is important. A completed randomized trial establishes that a prespecified comparison was undertaken, but the design description alone does not establish the size, direction, precision, or statistical significance of the treatment effect.
17. Reading a Future Trial Result Correctly
If an endpoint-specific result is reported for IALT elsewhere, the first step is to identify exactly what was measured. A useful statistical reading should then proceed through several layers.
1. Identify the estimand
Determine the outcome, treatment contrast, analysis population, and time frame represented by the estimate.
2. Identify the effect measure
Determine whether the result is expressed as a risk difference, ratio, mean difference, hazard ratio, or another measure.
3. Examine uncertainty
Read the confidence interval together with the point estimate rather than relying on a p-value alone.
4. Assess design compatibility
Check whether the analysis preserves the randomized comparison and follows the prespecified endpoint and analysis framework.
For time-to-event outcomes, the reader should also consider censoring, follow-up, the number of events, and whether a proportional-hazards interpretation is appropriate. For repeated or multiple endpoints, multiplicity and endpoint hierarchy become important.
18. Four Core Statistical Questions Raised by IALT
Question 1: What does randomization establish?
It establishes the intended treatment comparison at baseline. It does not by itself guarantee equal numbers of participants in every prognostic subgroup or eliminate every source of bias after randomization.
Question 2: Why does sample size matter?
Larger samples generally provide more information, but statistical precision depends on the endpoint and the amount of information actually observed. For time-to-event outcomes, the number of events is particularly important.
Question 3: Why can a treatment effect be statistically uncertain even in a large study?
The answer depends on the endpoint, event frequency, variability, censoring, allocation, and analysis. Enrollment is only one component of statistical information.
Question 4: Why can a statistically significant result still be clinically modest?
Statistical significance concerns compatibility with a null hypothesis. Clinical importance concerns the magnitude and consequences of the treatment effect. The two concepts should not be treated as synonyms.
19. Sources
- ClinicalTrials.gov: IALT, NCT00002823.
Continue through the Clinical Biostats trial library
Clinical trial pages connect study design with the statistical concepts needed to understand randomized evidence, endpoints, treatment effects, uncertainty, and analysis choices.
20. Record Summary
IALT is registered as a completed, randomized phase 3 treatment trial with 3300 enrolled participants, beginning in February 1995 and reaching primary completion in January 2001. The registered title describes a comparison of combination chemotherapy with no treatment following surgery in patients with non-small cell lung cancer, and the registry lists cisplatin, etoposide, vinblastine sulfate, vindesine, vinorelbine tartrate, and radiation therapy among the interventions.
From a statistical perspective, the most important feature is the randomized design. Randomization provides the foundation for a causal treatment comparison, while endpoint definitions, analysis populations, follow-up, missing-data procedures, and statistical models determine how that comparison is quantified and interpreted.
The ClinicalTrials.gov record does not list registered primary endpoints or posted statistical analyses. It therefore does not provide the endpoint-specific effect estimates, confidence intervals, p-values, event counts, or safety-by-arm results needed for a quantitative statistical results section. The appropriate statistical framework depends on the endpoint ultimately analysed: binary, continuous, time-to-event, or safety outcomes each require their own estimand and analysis strategy.