← Clinical Trials
Breast Neoplasms Phase 3 Time-to-Event NCT00490139

ALTTO: Complete Statistical Analysis of Lapatinib and Trastuzumab in Breast Neoplasms

An independent statistical analysis of the randomized phase 3 ALTTO study evaluating lapatinib and trastuzumab treatment strategies in participants with breast neoplasms, with emphasis on disease-free survival, overall survival, recurrence endpoints, and the trial's time-to-event methodology.

ALTTO  ·  BIG 2-06/N063D  ·  Completed  ·  Enrollment 8381
Scope of this record

This page separates reported trial results from statistical interpretation. Numerical results and trial-design facts on this page are restricted to the ClinicalTrials.gov record and the listed linked publications.

Registry context: 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

ALTTO was a randomized, parallel, open-label phase 3 study with 4 arms and 8381 participants enrolled. The registered primary endpoint was Disease-Free Survival (DFS) at the Primary Analysis, a time-to-event endpoint analyzed through comparisons involving lapatinib plus trastuzumab and trastuzumab.

8381
Enrollment
Phase 3
4
Arms
Randomized parallel design
0.84
Primary DFS HR
95% CI 0.71–1.00
0.048
Primary DFS P-value
Lapatinib + trastuzumab vs trastuzumab
FeatureALTTO
Trial nameALTTO (Adjuvant Lapatinib And/Or Trastuzumab Treatment Optimisation) Study; BIG 2-06/N063D
PhasePhase 3
ConditionNeoplasms, Breast
AllocationRandomized
Design modelParallel
MaskingNone
Primary purposeTreatment
Enrollment8381
InterventionsLapatinib; Trastuzumab
Primary endpointDisease-Free Survival (DFS) at the Primary Analysis
Primary endpoint typeTime-to-event
Primary statistical methodsLog-rank test; Cox proportional-hazards model
Effect measureHazard ratio
Hypothesis typeSuperiority
Trial statusCompleted
Start2007-05-16
Primary completion2013-12-06
Lead sponsorNovartis Pharmaceuticals
Sponsor typeIndustry

2. Clinical Question

The registered ALTTO study examined treatment strategies involving lapatinib and trastuzumab in participants with breast neoplasms. The primary statistical question was framed around Disease-Free Survival at the Primary Analysis, with superiority comparisons reported for lapatinib plus trastuzumab versus trastuzumab and trastuzumab followed by lapatinib versus trastuzumab.

Population

Participants enrolled in the phase 3 ALTTO study for the condition recorded as Neoplasms, Breast.

Interventions

Lapatinib and trastuzumab were the registered interventions. The results include comparisons involving lapatinib plus trastuzumab and trastuzumab followed by lapatinib.

Comparator

Trastuzumab serves as the comparator in the reported primary and secondary statistical comparisons reported in the ClinicalTrials.gov record.

Primary question

How do the specified lapatinib-containing treatment strategies compare with trastuzumab for the registered Disease-Free Survival endpoint?

3. Trial Design

ALTTO was randomized, parallel, unmasked, and intended for treatment. The study enrolled 8381 participants across 4 arms. The registry identifies lapatinib and trastuzumab as the study interventions.

01
Randomize 8381 enrolled
02
4 arms Lapatinib and/or trastuzumab strategies
03
Follow-up Time-to-event outcomes
04
Primary analysis Disease-Free Survival
05
Long-term 10-year follow-up outcomes
REPORTED COMPARISON · L + T

Lapatinib Plus Trastuzumab

  • Reported as the treatment group in the primary DFS comparison with trastuzumab.
  • Also evaluated for OS, TTR, TTDR, and CNS recurrence at the 10-year follow-up.
REPORTED COMPARISON · T

Trastuzumab

  • Comparator group for the principal reported treatment comparisons.
  • Used as the reference group for the reported hazard ratios.
REPORTED COMPARISON · T → L

Trastuzumab Followed by Lapatinib

  • Compared with trastuzumab for the primary DFS analysis.
  • Also evaluated for OS, TTR, TTDR, and CNS recurrence at the 10-year follow-up.
FOURTH REGISTERED ARM

Lapatinib

  • The lapatinib-alone arm was discontinued prior to the primary analysis due to futility.
  • DFS data for this arm were collected up to the first interim analysis, with data cut-off 11th July 2011.
  • For the reported primary OS analysis, zero participants were analyzed in the lapatinib arm.
Important design distinction: The existence of 4 randomized arms does not mean that every possible pairwise comparison is represented in the statistical analyses posted on ClinicalTrials.gov. The formal results provided here concern the specific comparisons reported in the trial data.

4. Randomization, Stratification, and Analysis Population

The reported primary analyses used the Intent-to-Treat (ITT) Population. The primary treatment hazard-ratio estimates were based on Cox proportional-hazards models adjusted for the trial's stratification factors.

FeatureReported information
Analysis populationIntent-to-Treat (ITT) Population
Stratification factor 1Chemotherapy timing
Stratification factor 2Hormone receptor status
Stratification factor 3Axillary lymph node status
Cox model roleEstimate treatment hazard ratio while adjusting for the stratification factors
Log-rank roleReported method for the primary DFS and primary OS comparisons

The ITT framework is particularly important here because one randomized arm was discontinued before the primary analysis. The registry explicitly notes that the lapatinib-alone arm was discontinued prior to primary analysis because of futility, so the information available for that arm differs from the information available for the other reported comparisons.

5. Registered Primary Endpoint

EndpointRegistry definition / time frameType
Disease-Free Survival (DFS) at the Primary Analysis From randomization until the date of the first occurrence of disease recurrence, a contralateral invasive breast cancer, a second primary cancer, or death from any cause, assessed up to approximately 10 years Time-to-event

The registry definition states that Disease-Free Survival was defined as the interval between randomization and the date of first occurrence of disease recurrence (local, regional or distant), a contralateral invasive breast cancer, a second primary cancer or death without recurrence. The registry-reported definition also specifies rules concerning deaths in participants whose clinical follow-up was ongoing and who had no recurrence or contralateral breast cancer.

Because DFS is a time-to-event endpoint, participants who have not experienced the defined event by the end of their available observation can contribute censored follow-up rather than being treated as if the event occurred at the end of observation.

6. Results: Primary Disease-Free Survival

The registry contains two formal primary analyses for the registered DFS endpoint. Both use the ITT population and a log-rank comparison, with the treatment hazard ratio estimated from a Cox proportional-hazards model adjusted for chemotherapy timing, hormone receptor status, and axillary lymph node status.

Lapatinib Plus Trastuzumab vs Trastuzumab

Disease-Free Survival hazard ratio

0.84

95% CI: 0.71–1.00   ·   P = 0.048

Two-sided confidence interval · Superiority hypothesis · ITT population

FeatureReported result
ComparisonLapatinib Plus Trastuzumab vs Trastuzumab
Analysis methodLog-rank test
Effect measureHazard Ratio (HR)
Estimate0.84
95% CI0.71–1.00
P-value0.048
Analysis populationIntent-to-Treat (ITT)
Model adjustmentChemotherapy timing, hormone receptor status, and axillary lymph node status
Clinical Biostats interpretation

An HR of 0.84 means that, under the fitted Cox model, the estimated instantaneous rate of a DFS event in the lapatinib-plus-trastuzumab group was 0.84 times that in the trastuzumab group. Expressed as a simple relative interpretation, 1 − 0.84 = 0.16, so the estimated hazard was 16% lower under the model.

The HR does not mean that 16% of participants avoided an event, that every participant had a 16% lower individual risk, or that the absolute probability of remaining disease-free was reduced or increased by exactly 16 percentage points.

The 95% CI of 0.71–1.00 describes statistical uncertainty around the estimated hazard ratio under the analysis framework. Its upper boundary reaches 1.00, so the estimate is close to the conventional null value on the hazard-ratio scale.

The P = 0.048 value addresses evidence against the null hypothesis within the specified statistical test. It is not a measure of the size or clinical importance of the treatment effect. The magnitude of the estimated effect is described by the HR and its confidence interval.

The interpretation also depends on the Cox proportional-hazards framework. A single HR summarizes the relative hazard under that model; it is not a direct description of how the treatment effect behaves at every individual time point.

Trastuzumab Followed by Lapatinib vs Trastuzumab

Disease-Free Survival hazard ratio

0.96

95% CI: 0.81–1.13   ·   P = 0.610

Two-sided confidence interval · Superiority hypothesis · ITT population

FeatureReported result
ComparisonTrastuzumab Followed by Lapatinib vs Trastuzumab
Analysis methodLog-rank test
Effect measureHazard Ratio (HR)
Estimate0.96
95% CI0.81–1.13
P-value0.610
Analysis populationIntent-to-Treat (ITT)
Model adjustmentChemotherapy timing, hormone receptor status, and axillary lymph node status
Clinical Biostats interpretation

An HR of 0.96 corresponds to an estimated hazard equal to 96% of the hazard in the trastuzumab group under the fitted model. The estimate is therefore close to the null value of 1.00.

The 95% CI of 0.81–1.13 spans 1.00. This means the interval includes both values below and above the null hazard ratio, so the reported estimate has substantial uncertainty about the direction and magnitude of the relative effect.

The P = 0.610 value is not an effect-size measure. A large P-value does not prove that the two treatment strategies are identical; it indicates that the observed data do not provide strong evidence against the null hypothesis under this particular test.

As with the first comparison, the HR is model-based and should be interpreted in the context of censoring, the ITT analysis population, the prespecified stratification factors, and the proportional-hazards framework.

7. Primary Analysis: What the Two DFS Comparisons Show Statistically

The two primary DFS comparisons illustrate why a clinical trial with several randomized treatment strategies cannot be summarized adequately by one number. The lapatinib-plus-trastuzumab comparison produced an HR of 0.84, while the trastuzumab-followed-by-lapatinib comparison produced an HR of 0.96. Their confidence intervals were 0.71–1.00 and 0.81–1.13, respectively.

Primary DFS comparisonHR95% CIP-value
Lapatinib Plus Trastuzumab vs Trastuzumab0.840.71–1.000.048
Trastuzumab Followed by Lapatinib vs Trastuzumab0.960.81–1.130.610

These are separate randomized comparisons within a multi-arm study. The first HR should not be used to infer the effect of the trastuzumab-followed-by-lapatinib strategy, and the second HR should not be used to infer the effect of the combination strategy. Each estimate answers its own prespecified comparison.

Multiplicity matters: the ClinicalTrials.gov record identifies multiple statistical analyses and multiple treatment comparisons, but do not provide a multiplicity-adjustment procedure or an alpha allocation for the primary comparisons. Accordingly, this page does not impose an additional multiplicity interpretation that is not contained in the ClinicalTrials.gov record.

8. Secondary Disease-Free Survival at the 10-Year Follow-Up

The registry also reports DFS at the 10-year follow-up using Cox proportional-hazards models. Both reported comparisons use the ITT population and adjust for chemotherapy timing, hormone receptor status, and axillary lymph node status.

ComparisonHR95% CIAnalysis
Lapatinib Plus Trastuzumab vs Trastuzumab 0.887 0.77–1.02 Cox proportional-hazards model
Trastuzumab Followed by Lapatinib vs Trastuzumab 0.914 0.80–1.05 Cox proportional-hazards model

Both 10-year DFS hazard-ratio estimates are below 1.00, but both corresponding 95% confidence intervals include 1.00. The long-term estimates should therefore be read as estimates with uncertainty rather than as exact measures of a persistent treatment difference.

How to read the 10-year DFS estimates

The HR of 0.887 for lapatinib plus trastuzumab versus trastuzumab corresponds to an estimated hazard approximately 11.3% lower under the fitted model, calculated directly from 1 − 0.887. The HR of 0.914 for trastuzumab followed by lapatinib corresponds to an estimated hazard approximately 8.6% lower.

Those derived percentages are descriptive transformations of the reported HRs; they are not absolute reductions in the probability of recurrence or death. The confidence intervals remain essential because they quantify uncertainty around the estimates.

9. Overall Survival Results

Overall Survival (OS) was a secondary time-to-event endpoint in the ClinicalTrials.gov record. At the Primary Analysis, the time frame was from randomization until death due to any cause, with a median follow-up of 4.5 years.

Primary-analysis OS comparisonHR95% CIP-valueMethod
Lapatinib Plus Trastuzumab vs Trastuzumab 0.80 0.62–1.03 0.078 Log-rank test; Cox model for HR
Trastuzumab Followed by Lapatinib vs Trastuzumab 0.91 0.71–1.16 0.433 Log-rank test; Cox model for HR
Clinical Biostats interpretation

For lapatinib plus trastuzumab versus trastuzumab, the reported OS HR of 0.80 corresponds to an estimated 20% lower instantaneous hazard of death under the Cox model. The 95% CI of 0.62–1.03 includes 1.00, so the estimate is compatible with a range of relative effects that includes the null value.

For trastuzumab followed by lapatinib versus trastuzumab, the OS HR was 0.91, with a 95% CI of 0.71–1.16. This interval also includes 1.00.

The P-values, 0.078 and 0.433, should not be interpreted as probabilities that a treatment is effective or ineffective. They quantify the evidence against the specified null hypothesis under the corresponding test; they do not measure treatment magnitude, patient-level benefit, or clinical importance.

10. Overall Survival at the 10-Year Follow-Up

The registry supplies additional OS analyses assessed up to approximately 10 years. These analyses used Cox proportional-hazards models in the ITT population and adjusted for chemotherapy timing, hormone receptor status, and axillary lymph node status.

ComparisonHR95% CIFollow-up
Lapatinib Plus Trastuzumab vs Trastuzumab 0.853 0.70–1.03 Approximately 10 years
Trastuzumab Followed by Lapatinib vs Trastuzumab 0.863 0.71–1.04 Approximately 10 years

The two long-term OS estimates are close to one another on the hazard-ratio scale, with both point estimates below 1.00 and both 95% confidence intervals including 1.00. These results illustrate the difference between an estimated treatment effect and the precision with which that effect is estimated.

Relative effect

An HR below 1 indicates a lower estimated instantaneous event rate in the first-listed treatment group under the Cox model.

Uncertainty

The 95% confidence interval communicates how much statistical uncertainty surrounds the reported point estimate.

Time horizon

The 10-year analyses describe a later follow-up period and should not be silently combined with the primary-analysis estimates.

Population

The registry-reported long-term analyses use the Intent-to-Treat population.

11. Time to Recurrence

Time to Recurrence (TTR) was assessed from randomization until the date of first occurrence of a disease recurrence, up to approximately 10 years. The registry-reported analysis text states that TTR was analyzed using competing risks methodology, with death without event treated as a competing risk.

ComparisonHR95% CIMethod reported in the ClinicalTrials.gov record
Lapatinib Plus Trastuzumab vs Trastuzumab 0.811 0.68–0.96 Not reported; competing risks methodology specified
Trastuzumab Followed by Lapatinib vs Trastuzumab 0.930 0.79–1.10 Not reported; competing risks methodology specified
Why competing risks change the interpretation

For TTR, death without recurrence prevents a later recurrence from being observed. Treating such a death as an ordinary censoring event can answer a different statistical question because the competing event is informative about what can subsequently occur.

The registry analysis therefore identifies competing-risks methodology specifically for TTR. The reported HR of 0.811 for lapatinib plus trastuzumab corresponds to an estimated 18.9% lower hazard on the reported relative scale, while the HR of 0.930 corresponds to an estimated 7.0% lower hazard. These transformations describe the reported HRs; they do not represent absolute reductions in recurrence probability.

12. Time to Distant Recurrence

Time to Distant Recurrence (TTDR) was defined from randomization until the first occurrence of distant recurrence, assessed up to approximately 10 years. The registry-reported analysis text specifies competing risks methodology with death without event as a competing risk.

ComparisonHR95% CI
Lapatinib Plus Trastuzumab vs Trastuzumab 0.852 0.71–1.02
Trastuzumab Followed by Lapatinib vs Trastuzumab 0.968 0.81–1.16

The lapatinib-plus-trastuzumab estimate of 0.852 is below 1.00, whereas the trastuzumab-followed-by-lapatinib estimate of 0.968 is close to 1.00. The corresponding confidence intervals are important because the first includes 1.00 and the second also includes 1.00.

Statistical caution: TTDR is not interchangeable with DFS or TTR. The event definition is different, and the ClinicalTrials.gov record specifically identify death without the event as a competing risk for TTDR.

13. Time to Central Nervous System Recurrence

Time to Central Nervous System (CNS) Recurrence was assessed from randomization until the first CNS recurrence, up to approximately 10 years. The registry-reported analysis text specifies competing risks methodology with death without event as a competing risk.

ComparisonHR95% CI
Lapatinib Plus Trastuzumab vs Trastuzumab 0.986 0.74–1.31
Trastuzumab Followed by Lapatinib vs Trastuzumab 0.980 0.74–1.31

Both CNS recurrence hazard-ratio estimates are very close to 1.00, and their confidence intervals are comparatively broad. The statistical message is therefore primarily one of substantial uncertainty around these estimates rather than a precise estimate of a large relative effect.

Reading a wide confidence interval

A confidence interval such as 0.74–1.31 covers values below and above 1.00. The point estimate alone, therefore, is insufficient to characterize the evidence. The interval shows that the statistical uncertainty is large relative to the distance between the point estimate and the null value.

14. Secondary Endpoint Results at a Glance

EndpointLapatinib Plus Trastuzumab vs TrastuzumabTrastuzumab Followed by Lapatinib vs Trastuzumab
DFS, Primary AnalysisHR 0.84; 95% CI 0.71–1.00; P = 0.048HR 0.96; 95% CI 0.81–1.13; P = 0.610
DFS, 10-Year Follow-UpHR 0.887; 95% CI 0.77–1.02HR 0.914; 95% CI 0.80–1.05
OS, Primary AnalysisHR 0.80; 95% CI 0.62–1.03; P = 0.078HR 0.91; 95% CI 0.71–1.16; P = 0.433
OS, 10-Year Follow-UpHR 0.853; 95% CI 0.70–1.03HR 0.863; 95% CI 0.71–1.04
TTR, 10-Year Follow-UpHR 0.811; 95% CI 0.68–0.96HR 0.930; 95% CI 0.79–1.10
TTDR, 10-Year Follow-UpHR 0.852; 95% CI 0.71–1.02HR 0.968; 95% CI 0.81–1.16
CNS Recurrence, 10-Year Follow-UpHR 0.986; 95% CI 0.74–1.31HR 0.980; 95% CI 0.74–1.31

This table is useful for pattern recognition, but it should not be read as a single composite treatment effect. The endpoints have different definitions, different follow-up descriptions, and, for recurrence outcomes, a specified competing-risks framework.

15. Statistical Methodology

Kaplan-Meier estimation

The ClinicalTrials.gov record identifies the major endpoints as time-to-event outcomes and identify survival-analysis methods including the log-rank test and Cox proportional-hazards model. Kaplan-Meier estimation is the standard descriptive framework for displaying a time-to-event distribution, although the registry-reported statistical-analysis fields do not explicitly identify Kaplan-Meier as a posted method for each individual ALTTO analysis.

Conceptual survival function
S(t) = ∏ti ≤ t (1 − di/ni)

Here, di is the number of events at time ti and ni is the number at risk immediately before that time. Censoring allows participants to contribute information up to their last observed event-free time.

Log-rank test

The registry reports the log-rank test for the primary DFS and primary OS comparisons. Conceptually, the log-rank test compares the observed and expected numbers of events between treatment groups across the observed follow-up while accounting for the timing of events.

Cox proportional-hazards model

The reported treatment hazard ratios were based on Cox proportional-hazards models. For the primary analyses, the model adjusted for chemotherapy timing, hormone receptor status, and axillary lymph node status.

Hazard-ratio interpretation
HR < 1  →  lower estimated instantaneous event rate in the first-listed treatment group

The HR is a relative, model-based measure. It is not an absolute risk difference, an absolute probability of avoiding an event, or a percentage of patients who benefit.

Stratified analysis

The registry-reported primary-analysis notes explicitly state that stratification was by chemotherapy timing, hormone receptor status, and axillary lymph node status. The Cox model adjusted for these factors when estimating the treatment hazard ratio.

Intention-to-treat analysis

The efficacy analyses reported in the registry use the Intent-to-Treat population. An ITT comparison retains participants according to randomized assignment rather than redefining the treatment groups according to what treatment was actually received. This preserves the treatment comparison created by randomization.

Competing risks

The TTR, TTDR, and CNS recurrence analysis notes specifically state that competing-risks methodology was used, with death without the event treated as a competing risk. This is an important distinction from ordinary time-to-event analysis because death can prevent the recurrence event from occurring later.

16. Interim Analysis and Futility

The ClinicalTrials.gov record explicitly connect the lapatinib-alone arm to an interim analysis. DFS data for that arm include data collected up to the first interim analysis, with a data cut-off of 11th July 2011, because the lapatinib-alone arm was discontinued prior to the primary analysis.

Why interim analysis matters

An interim analysis evaluates accumulating trial information before the nominal end of follow-up. It can lead to continuation, modification, or discontinuation of a treatment arm depending on prespecified monitoring rules.

Futility decision

The ClinicalTrials.gov record states that the lapatinib-alone arm was discontinued prior to the primary analysis due to futility.

The existence of an interim analysis has direct consequences for interpreting the lapatinib-alone data. The registry does not provide a complete primary-analysis result for that arm in the statistical analyses posted on ClinicalTrials.gov. Therefore, the lapatinib-alone arm should not be assigned an invented primary-analysis hazard ratio.

Alpha spending: The trial data identify interim analysis / alpha spending as a statistical concept associated with the reported analyses, but do not supply a specific alpha-spending function, boundary, or numerical alpha allocation. This page therefore does not attribute a particular interim boundary to ALTTO.

17. Multiplicity and Multiple Comparisons

ALTTO contains multiple randomized arms and multiple reported endpoints and comparisons. The ClinicalTrials.gov record identifies 14 statistical analyses, including 2 primary-endpoint analyses with estimates and confidence intervals.

FeatureRegistry-supported information
Statistical analyses posted14
Primary endpoint analyses2
Primary analyses with estimate + CI2
Hypothesis typeSuperiority
Confidence intervals95%, two-sided
Multiplicity procedureNot reported in the ClinicalTrials.gov record

Multiplicity is important because every additional formal statistical comparison creates another opportunity for a result to appear unusual under a collection of null hypotheses. However, the ClinicalTrials.gov record does not specify an adjustment procedure or familywise error strategy, so no such procedure should be inferred.

The same principle applies to the numerous secondary endpoints. A nominal confidence interval or P-value for an individual secondary analysis does not automatically establish that the analysis was powered or error-controlled as an independent confirmatory test.

18. Missing Data, Censoring, and the Analysis Population

The ClinicalTrials.gov record does not provide a missing-data or imputation method. Accordingly, no specific imputation procedure is attributed to ALTTO here.

For the time-to-event endpoints, censoring is an inherent part of survival analysis: participants can remain event-free when their observed follow-up ends. The important statistical point is that a censored participant is not treated as having experienced the endpoint at the censoring time.

Arm-specific information matters: the lapatinib-alone arm had a different information history because it was discontinued for futility before the primary analysis. The registry-reported DFS data for that arm extend only to the first interim analysis, with data cut-off 11th July 2011.

19. Safety Results

The ClinicalTrials.gov record reports serious adverse events by randomized arm as affected participants divided by participants at risk.

ArmSerious adverse eventsAffected / at risk
Lapatinib Plus TrastuzumabSerious adverse events379 / 2061
Trastuzumab Followed by LapatinibSerious adverse events299 / 2076
LapatinibSerious adverse events394 / 2056
TrastuzumabSerious adverse events251 / 2076
Serious adverse events: affected participants / at risk
Lapatinib + Trastuzumab
379 / 2061
Trastuzumab → Lapatinib
299 / 2076
Lapatinib
394 / 2056
Trastuzumab
251 / 2076

The affected/at-risk values are reported counts rather than statistical effect estimates. Because the ClinicalTrials.gov record does not provide a formal safety comparison, this page does not convert these counts into an inferential treatment comparison or attach a P-value that was not reported.

20. Statistical Methods Explained

Why was a Cox proportional-hazards model used?

The registered primary endpoint is time-to-event, and the reported effect measure is a hazard ratio. The Cox model provides a framework for estimating a relative hazard while incorporating the timing of events and censoring and, in ALTTO's reported primary analyses, adjusting for the specified stratification factors.

What does an HR of 0.84 mean?

An HR of 0.84 means that the estimated instantaneous event rate for the first-listed treatment group was 84% of that for the comparator under the fitted model. A simple transformation, 1 − 0.84, gives a 16% relative reduction in the estimated hazard. It does not mean a 16-percentage-point increase in disease-free survival or that 16% of patients benefited.

Why use stratification factors in the Cox model?

The primary analysis notes identify chemotherapy timing, hormone receptor status, and axillary lymph node status as stratification factors. Adjusting the Cox model for these factors incorporates the trial's randomization structure into estimation of the treatment hazard ratio.

Why does the P-value not measure effect size?

A P-value describes the compatibility of the observed data with a specified null hypothesis under the statistical model and test. It depends on the estimated effect, its variability, and the amount of information. The HR describes the estimated relative effect, while the confidence interval describes uncertainty around that effect.

Why are TTR, TTDR, and CNS recurrence treated differently?

The registry-reported analysis notes specify competing-risks methodology for these recurrence endpoints because death without the endpoint is a competing event. A competing event prevents the endpoint of interest from occurring later, so treating it as ordinary censoring can answer a different statistical question.

Why should the lapatinib-alone arm not be given an invented primary result?

The ClinicalTrials.gov record states that the lapatinib-alone arm was discontinued prior to primary analysis due to futility. Its DFS information was collected only through the first interim analysis, with data cut-off 11th July 2011, and the registry-reported primary OS analysis states that zero participants were analyzed in that arm. A primary-analysis HR for that arm is therefore not present in the ClinicalTrials.gov record.

21. Understanding the Hazard Ratio Across ALTTO Endpoints

Endpoint / comparisonHRSimple interpretation of point estimate
Primary DFS: Lapatinib + Trastuzumab vs Trastuzumab0.8416% lower estimated hazard
Primary DFS: Trastuzumab → Lapatinib vs Trastuzumab0.964% lower estimated hazard
10-Year DFS: Lapatinib + Trastuzumab vs Trastuzumab0.88711.3% lower estimated hazard
10-Year DFS: Trastuzumab → Lapatinib vs Trastuzumab0.9148.6% lower estimated hazard
Primary OS: Lapatinib + Trastuzumab vs Trastuzumab0.8020% lower estimated hazard
Primary OS: Trastuzumab → Lapatinib vs Trastuzumab0.919% lower estimated hazard
10-Year OS: Lapatinib + Trastuzumab vs Trastuzumab0.85314.7% lower estimated hazard
10-Year OS: Trastuzumab → Lapatinib vs Trastuzumab0.86313.7% lower estimated hazard

These percentage transformations are simple arithmetic interpretations of the reported HRs. They should not be confused with absolute treatment effects. For example, an HR of 0.80 does not imply that 20% of patients were prevented from dying, nor does it specify the difference in survival probability at any particular time point.

Confidence intervals are indispensable

Consider the primary OS estimate of 0.80 with a 95% CI of 0.62–1.03. The point estimate suggests a lower hazard, but the interval includes 1.00. The confidence interval therefore supplies information that the point estimate alone cannot: the data are compatible with a range of plausible relative effects extending through the conventional null value.

22. Longitudinal Statistical Story

2007-05-16

Trial begins

The ALTTO study start date in the ClinicalTrials.gov record is 2007-05-16.

11th July 2011

First interim data cut-off for lapatinib-alone data

The registry-reported DFS analysis states that data for the lapatinib-alone arm include data collected up to the first interim analysis, with this data cut-off.

Primary analysis

Lapatinib-alone arm discontinued for futility

The lapatinib-alone arm was discontinued prior to primary analysis due to futility. The principal DFS comparison reported an HR of 0.84 for lapatinib plus trastuzumab versus trastuzumab.

2013-12-06

Primary completion

The registry profile lists 2013-12-06 as the primary completion date.

10-Year Follow-Up

Long-term time-to-event analyses

The ClinicalTrials.gov record includes DFS, OS, TTR, TTDR, and CNS recurrence analyses assessed at approximately 10 years.

23. Limitations and Interpretation Issues

24. Why This Trial Matters Statistically

ALTTO is a useful statistical teaching case because it combines a randomized multi-arm design with an interim futility decision, stratified time-to-event analysis, multiple treatment comparisons, long-term follow-up, and competing-risks methodology for recurrence endpoints.

Statistical conceptHow it appears in ALTTO
RandomizationRandomized parallel phase 3 study with 4 arms
ITT analysisPrimary and secondary efficacy analyses use the Intent-to-Treat population
Time-to-event endpointsDFS, OS, TTR, TTDR, and CNS recurrence
Hazard ratioPrimary and secondary treatment effects are expressed as HRs
Confidence intervalReported effects use two-sided 95% confidence intervals
Log-rank testReported for the primary DFS and primary OS comparisons
Cox modelUsed to estimate HRs with adjustment for stratification factors
Stratified analysisAdjustment for chemotherapy timing, hormone receptor status, and axillary lymph node status
Interim analysisLapatinib-alone data include information through the first interim analysis
FutilityLapatinib-alone arm discontinued before primary analysis due to futility
Competing risksDeath without event treated as a competing risk for TTR, TTDR, and CNS recurrence
MultiplicityMultiple statistical analyses and treatment comparisons require careful interpretation

25. Statistical Interpretation vs Clinical Interpretation

Statistical interpretation

The primary DFS comparison of lapatinib plus trastuzumab versus trastuzumab produced an HR of 0.84 with a 95% CI of 0.71–1.00 and P = 0.048. The alternative sequence comparison produced an HR of 0.96 with a 95% CI of 0.81–1.13 and P = 0.610.

What the statistics do not establish alone

These estimates do not by themselves quantify absolute benefit, individual patient benefit, or the probability that a treatment will produce a particular outcome for an individual participant.

Long-term interpretation

The 10-year analyses provide additional hazard-ratio estimates for DFS, OS, TTR, TTDR, and CNS recurrence, but each endpoint has its own definition and uncertainty.

Safety interpretation

Serious adverse events were reported as affected participants divided by participants at risk for each arm. These descriptive safety counts are separate from the efficacy hazard ratios.

26. A Practical Reading Strategy for ALTTO

A useful way to read the ALTTO results is to move from the design to the endpoint and then to the estimator rather than beginning with the P-value.

  1. Identify the randomized comparison. Confirm which treatment group is being compared with trastuzumab.
  2. Identify the endpoint. DFS, OS, TTR, TTDR, and CNS recurrence are different time-to-event outcomes.
  3. Check the analysis population. The registry-reported efficacy analyses use the ITT population.
  4. Read the hazard ratio. HR below 1 indicates a lower estimated instantaneous event rate in the first-listed group.
  5. Read the confidence interval. The interval shows the uncertainty surrounding the HR and whether it includes 1.00.
  6. Then read the P-value. It describes evidence against the specified null hypothesis under the test; it does not measure effect size.
  7. Check the design context. Interim analysis, early arm discontinuation, stratification, multiple comparisons, and competing risks all affect how the result should be interpreted.
A compact statistical summary
Treatment comparison → time-to-event endpoint → log-rank / Cox model → HR + 95% CI → P-value → design context

This sequence prevents the P-value from becoming the entire interpretation of a clinical-trial result.

27. Related Tutorials

Learn more about the methods used in this trial:

28. Related Calculators

29. Sources

Continue through Clinical Biostats

Connect the statistical concepts in ALTTO with deeper tutorials, calculators, and additional clinical-trial analyses.

30. Record Summary

ALTTO provides a detailed example of how a randomized multi-arm clinical trial can generate several layers of time-to-event evidence. The primary registered endpoint was Disease-Free Survival at the Primary Analysis, with formal comparisons of lapatinib plus trastuzumab versus trastuzumab and trastuzumab followed by lapatinib versus trastuzumab. The corresponding HRs were 0.84 and 0.96, with 95% confidence intervals of 0.71–1.00 and 0.81–1.13.

The statistical story extends beyond the primary DFS analysis. The ClinicalTrials.gov record includes primary-analysis OS, 10-year DFS and OS, TTR, TTDR, and CNS recurrence results, together with competing-risks methodology for recurrence endpoints. They also document ITT analysis, adjustment for chemotherapy timing, hormone receptor status, and axillary lymph node status, and an interim analysis that preceded discontinuation of the lapatinib-alone arm for futility.

Clinical Biostats methodology: A rigorous trial-results page should identify the randomized comparison, preserve the registered endpoint definition, distinguish the analysis population from the treatment actually received, report the effect estimate together with its confidence interval, and then interpret the P-value in the context of the trial design. ALTTO is particularly useful for demonstrating why multi-arm comparisons, interim decisions, stratification, and competing risks must remain visible in the statistical interpretation.