Chemotherapy With or Without Bevacizumab in Treating Patients With Stage IB, Stage II, or Stage IIIA Non-small Cell Lung Cancer That Was Removed By Surgery: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Chemotherapy With or Without Bevacizumab in Treating Patients With Stage IB, Stage II, or Stage IIIA Non-small Cell Lung Cancer That Was Removed By Surgery (NCT00324805), a randomized phase 3 study. The primary outcome is Overall Survival.

2. Clinical Question

The clinical question concerns whether adding bevacizumab to postoperative chemotherapy changes survival after surgery. The registry record identifies overall survival as the primary endpoint and says the study failed to meet it. The reported comparison is close to a hazard ratio of 1, and its confidence interval spans values on either side of 1; the data therefore do not establish a clear difference between the randomized groups. This finding should be read as an estimate with uncertainty, not as proof that the groups are identical.

3. Trial Design

This was a phase III, randomized, parallel-group trial with no masking. It enrolled 1501 patients. The record lists 749 patients in the chemotherapy arm and 752 in the chemotherapy-plus-bevacizumab arm. The study began on 2007-07-19, had primary completion on 2015-10-20, and is recorded as completed.

The comparison was between adjuvant chemotherapy alone and the same general chemotherapy approach with bevacizumab added. The record lists cisplatin, docetaxel, gemcitabine hydrochloride, pemetrexed disodium, and vinorelbine tartrate among the interventions. The available arm description indicates that patients received one of several chemotherapy regimens; it does not supply a complete description of every regimen. The bevacizumab arm is described as receiving chemotherapy as in the comparison arm, with bevacizumab added on day 1 and repeated every 21 days for up to 1 year. The chemotherapy description gives up to 4 courses, with treatment subject to progression or unacceptable toxicity.

The population was patients with completely resected non-small cell lung cancer in the listed stage groups, including stage IB disease of at least 4 cm. Overall survival was measured from randomization to death from any cause, with follow-up described as up to 10 years. Patients thought to be alive at final analysis were censored at their last contact date. The record does not report the original sample-size justification, the allocation ratio as a protocol assumption, or additional details of the randomization procedure. It also does not report masking beyond stating that the trial was unmasked.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All enrolled patients

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

Population: All randomized patients

6. Primary Endpoint

The registry lists the primary outcome as Overall Survival, measured over From registration to death, up to 10 years.

Overall survival measures the time from randomization until death from any cause. It is a time-to-event endpoint: the analysis accounts for both when deaths occur and how long patients are followed without a recorded death. Patients who were thought to be alive at the final analysis were censored at their last contact date, according to the registry definition.

Overall survival is used to assess the duration of survival without limiting the event to a particular cause. In general, a time-to-event measure can use information from participants who have different lengths of follow-up, rather than treating everyone as if their observation ended at the same time. Censoring records that an event was not observed during the available follow-up; it does not establish what happened after the last contact.

The primary endpoint is reported as a median in months, with a 95% confidence interval, and the between-arm comparison is reported as a Cox regression hazard ratio. The record gives a median of 85.8 months for the chemotherapy-plus-bevacizumab arm, with a lower confidence limit of 74.9 months; an upper limit is not posted. A median and a hazard ratio describe different aspects of the time-to-event data: the median is a point on the survival-time distribution, while the hazard ratio compares event rates over follow-up under the Cox model. The chemotherapy-arm median is not reported in the registry record.

7. Results Posted in the Registry

• Arm I (Chemotherapy) vs Arm II (Chemotherapy, Bevacizumab): hazard ratio 0.99 (95% CI 0.82–1.19), reported p = 0.900 (as reported in the registry).

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Disease-free Survival

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor (Regression, Cox); it was not recalculated.
• Result:
  • Arm I (Chemotherapy) vs Arm II (Chemotherapy, Bevacizumab): hazard ratio 0.99 (95% CI 0.86–1.15), reported p = 0.950 (as reported in the registry).
• Interpretation:
  • The 95% confidence interval includes no difference (a ratio of 1), so these data do not show a statistically significant difference between Arm I (Chemotherapy) vs Arm II (Chemotherapy, Bevacizumab) and the control. The interval gives the range of effects that remain compatible with the data.

Disease-free survival is the secondary endpoint with results posted. Its reported hazard ratio is 0.99, with a 95% confidence interval from 0.86 to 1.15 and p = 0.95. The interval includes 1, and the estimate is close to the reference value. In broad terms, the primary and posted secondary comparisons both show estimates near 1 with intervals that include values on either side of 1; neither posted comparison establishes a clear separation between the randomized groups.

The secondary endpoint adds context but does not replace the primary overall-survival result. A secondary outcome can address a related but distinct aspect of follow-up, and its estimate can be informative about the pattern of results. It should not be treated as a second opportunity to declare the primary question answered, especially when the primary endpoint was not met.

In general, testing multiple outcomes increases the chance of a chance-positive result if each is interpreted as a standalone confirmatory test. Adjustments or a prespecified hierarchy can control how results are interpreted, but the registry record does not report whether any multiplicity procedure or hierarchy was used. For that reason, unadjusted secondary results carry less evidential weight than the prespecified primary analysis. The record lists other outcome measures, but results for additional secondary endpoints are not reported in the registry record.

The registry describes the disease-free survival timeframe as from registration to death, up to 10 years, but does not supply a further definition of the events included in that endpoint in the posted description. Details of its analysis population and censoring rules are not reported in the registry record. These gaps constrain any more detailed comparison of the secondary endpoint with overall survival.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Disease-free Survival

9. Statistical Interpretation

10. Statistical Methods Explained

The primary endpoint is a time-to-event outcome, and the reported between-group analysis uses Cox regression. The sponsor-reported statistic is a hazard ratio, with a 95% confidence interval and p-value. No recalculated time-to-event estimate is substituted for that analysis; the reported result is the relevant comparison in the registry record.

What does a hazard ratio describe?

In general, a hazard ratio compares the event rates between groups over the observed follow-up, conditional on the model. A value of 1 represents no relative difference in the modeled hazards. Values below or above 1 indicate the direction of the estimated comparison, depending on the order in which the groups are defined. The posted comparison is labeled chemotherapy versus chemotherapy plus bevacizumab, and its hazard ratio is 0.99. This is an estimate from the full time-to-event analysis, not a ratio of the two reported median survival times.

The Cox model is commonly used for time-to-event data because it can incorporate event timing and censoring. In general, censoring means that an event was not observed by the end of the person’s available follow-up; it is not the same as confirming that the event never occurred. The registry says that patients thought to be alive at final analysis were censored at their last contact date. Details such as the model covariates, checks of model assumptions, and handling of other forms of missing follow-up are not reported in the registry record.

How should the confidence interval and p-value be read?

In general, a confidence interval expresses the statistical uncertainty around an estimate under the analysis assumptions. The primary 95% confidence interval runs from 0.82 to 1.19. Because it includes 1, it does not rule out the reference value of no relative difference in hazards. Its range also allows estimates on either side of that reference, so the point estimate alone should not be treated as precise evidence of a directional difference. A p-value of 0.90 is the reported test result for the comparison; it is not the probability that the null hypothesis is true, nor does it quantify the size or clinical importance of an effect.

The registry identifies the primary test as a superiority analysis. In general, a superiority test asks whether the data support a difference in the specified direction or any difference, according to the test definition. Not meeting a superiority endpoint does not show equivalence. The registry record does not report an equivalence or noninferiority margin or an analysis designed to establish either conclusion.

What do medians add?

A median time-to-event value is the time at which the estimated event-free proportion reaches the midpoint of the distribution. It offers an interpretable summary of follow-up experience, but it does not capture the entire pattern of events. The record posts a median overall survival of 85.8 months for the chemotherapy-plus-bevacizumab arm and gives a lower confidence limit of 74.9 months; the upper confidence limit is not reported. A median for the chemotherapy-alone arm is not reported in the registry record. The Cox hazard ratio and medians are different summaries, and they need not convey identical information.

What is not specified about the analysis?

The record identifies all enrolled patients as the primary analysis population. It does not report whether the analysis was explicitly intention-to-treat, the covariates included in the Cox model, or how protocol deviations were treated. It also does not report the study’s original sample-size assumptions, interim analyses, stopping rules, or a detailed statistical analysis plan. These omissions prevent a full audit of the prespecified analysis choices from the registry record alone.

In general, interpretation of a Cox model relies on assumptions about how the relative hazards behave over time and about censoring. The registry does not report diagnostic checks or a detailed account of these assumptions. This is a limit on what can be assessed from the posted summary, not evidence that a particular assumption was violated.

11. Confidence Intervals and Precision

The primary hazard ratio is 0.99, with a 95% confidence interval from 0.82 to 1.19. The estimate is close to 1, but the interval is broad enough to include values below and above 1. It therefore leaves uncertainty about both the direction and size of the relative difference. The p-value of 0.90 is consistent with the registry’s statement that the primary endpoint was not met.

In general, a confidence interval describes the range of values reasonably compatible with the data and the statistical model, at the stated confidence level. It is not a guarantee that the true effect lies within its limits, and it does not assign equal probability to all values in the interval. A confidence interval crossing 1 for a hazard ratio indicates that the data do not exclude the no-relative-difference reference at that confidence level. It does not prove that the true hazard ratio is exactly 1.

The median overall survival estimate is 85.8 months in the chemotherapy-plus-bevacizumab arm, with a reported lower limit of 74.9 months. The upper confidence limit and chemotherapy-arm median are not reported in the registry record, so the median-based comparison cannot be fully assessed from these entries. The hazard-ratio interval gives the posted comparative precision for the primary analysis.

Precision is also distinct from practical importance. A precise estimate can be small, and an imprecise estimate can leave a range of potentially relevant effects unresolved. The registry record does not provide a threshold for clinical importance or a prespecified margin for equivalence. The primary interval should consequently be interpreted as uncertainty around the estimated comparison, rather than as a verdict about all possible differences between the regimens.

12. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The planning figures are expressed as total event counts because these are time-to-event analyses, not as participants per arm. The calculations use the observed effect as a reference and show event requirements for planning around 100% of that effect, 75% of that effect, and 50% of that effect. For each assumed effect, the columns for 80% and 90% power indicate the total events needed to have the stated probability of detecting that effect under the planning model, if the assumptions hold.

These are planning illustrations, not a report of the trial’s original design assumptions. The stated calculation assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation; the effect to detect is the stated fraction of the observed effect. The registry does not report the trial’s own original sample-size assumptions. The posted figures are total events, not total participants or events per arm.

Smaller effects require more participants or, for this time-to-event illustration, more total events to detect them at a given power. The 90% power column requires more events than the 80% column for the same effect assumption. For example, the entry for planning around 100% of the observed effect gives 310,819 total events for 80% power and 416,098 for 90% power. These figures are extraordinarily large in relation to the observed trial enrollment and should be understood only as outputs of the stated illustration, not as achievable targets for this trial.

This is not observed power. It does not describe the probability that the completed trial detected an effect, and it should not be used to reinterpret the observed p-value. The calculations are hypothetical planning quantities based on the observed estimate and listed assumptions, not evidence about the trial’s prespecified design or the number of events actually observed.

13. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

d is the total number of events needed (Schoenfeld's formula) with 1:1 allocation and proportional hazards.

Worked calculation for this trial

1. Hazard ratio HR = 0.990, so ln(HR) = -0.0101.
2. d = 4 × (1.96 + 0.842)² / [ln(0.990)]² = 310,818.3, rounded up to 310,819 events in total for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
hr <- 0.99                              # observed hazard ratio
events_needed <- function(power, frac = 1) ceiling(round(4 * (z_a + qnorm(power))^2 / log(hr^frac)^2, 6))   # Schoenfeld, 1:1 allocation
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: events = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(events_needed(0.80, f), big.mark = ","), format(events_needed(0.90, f), big.mark = ",")))

# Simulation: given d total events under proportional hazards, events in the treatment arm are Binomial(d, hr / (1 + hr)).
# The score (log-rank type) statistic is (D1 - d / 2) / sqrt(d / 4).
set.seed(324805)
d    <- events_needed(0.80)
reps <- 20000
D1 <- rbinom(reps, d, hr / (1 + hr))
z  <- (D1 - d / 2) / sqrt(d / 4)
cat(sprintf("Simulated power with %s events: %.3f (%d replicates)\n", format(d, big.mark = ","), mean(abs(z) > z_a), reps))

Output from running the code

100% of observed effect: events = 310,819 (80% power), 416,098 (90% power)
 75% of observed effect: events = 552,566 (80% power), 739,729 (90% power)
 50% of observed effect: events = 1,243,274 (80% power), 1,664,390 (90% power)
Simulated power with 310,819 events: 0.797 (20000 replicates)

The simulation uses the fact that, with d events in total and proportional hazards, the number of events in the treatment arm is binomial. It is an approximation that conditions on the total number of events, and it should land close to the target power.

14. Analyses Reported by the Sponsor

The registry reports the sponsor’s Cox regression analysis for overall survival as a hazard ratio of 0.99, with a 95% confidence interval from 0.82 to 1.19 and p = 0.90. It also states that the study failed to meet its primary endpoint. The posted interpretation should therefore be anchored in that reported analysis rather than a separate unadjusted reconstruction.

The registry record does not flag a discrepancy between the sponsor’s analysis and a recalculation. For the time-to-event result, the displayed comparison uses the sponsor-reported hazard ratio. This is important because a simple reconstruction from summary values would not reproduce the full event-time analysis, its censoring information, or any model details. Such a reconstruction should not be treated as a replacement for the sponsor’s Cox model.

In general, sponsor-reported analyses may incorporate participant-level event timing and censoring that are not available in a summary record. The record names Cox regression but does not report the full model specification or its diagnostics. The absence of a flagged discrepancy does not resolve those reporting limitations; it means only that no conflict is indicated in the registry entries used for this result.

15. Multiplicity, Missing Data and Analysis Populations

The registry identifies overall survival as the primary endpoint and lists disease-free survival as a secondary outcome with results posted. The record also lists five outcome measures and two analyses, but it does not describe a prespecified hierarchy for testing endpoints or report an adjustment for multiple comparisons. Whether any multiplicity strategy was used is not reported in the registry record.

This matters because repeated testing across outcomes can increase the chance of obtaining at least one apparently notable result even when no underlying difference exists. In general, a prespecified testing order or adjustment can clarify how evidence across several outcomes should be interpreted. Without a reported strategy, secondary findings should be treated as supportive or exploratory rather than as independent confirmation of the primary result. The primary analysis remains the central test identified in the record.

The registry says that all enrolled patients formed the primary population. It does not state whether this was formally an intention-to-treat analysis, describe exclusions after randomization, or provide a detailed account of missing outcome data. For overall survival, the record says that participants thought to be alive at final analysis were censored at last contact. That explains one aspect of follow-up handling, but does not provide the number or pattern of censored observations, reasons for loss to follow-up, or sensitivity analyses for missingness. Those details are not reported in the registry record.

Interim analyses, stopping rules, and the full statistical analysis plan are also not reported in the registry record. Their absence from the posted information does not show that they were absent from the conduct of the trial; it limits what can be verified from the registry summary. The record does not provide enough information to assess whether analysis decisions were prespecified in detail or how deviations, if any, were handled.

16. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: neutrophil count decreased (37.1%), hypertension (27.2%), fatigue (12.5%), nausea (10.1%), hyponatremia (9.0%).

The registry reports serious adverse events affecting 424 of 738 participants at risk in the chemotherapy arm and 563 of 735 at risk in the chemotherapy-plus-bevacizumab arm. Other, non-serious adverse events are reported for 474 of 738 at risk in the chemotherapy arm and 509 of 735 at risk in the combination arm. These are posted counts with their listed at-risk denominators; they should be read as event reporting in the registry, not as an adjusted comparison of risk.

The listed serious adverse event terms include neutrophil count decreased at 37.1%, hypertension at 27.2%, fatigue at 12.5%, nausea at 10.1%, hyponatremia at 9.0%, and anemia at 7.0%. The registry summary does not make clear in this entry whether these term percentages refer to a particular arm or to a pooled listing, so they should not be assigned to one treatment group. Death counts among adverse-event outcomes are not reported in the registry record.

In general, adverse-event counts depend on how events were defined, collected, and attributed, and the raw counts alone do not establish causation. Further details on severity grading, exposure-adjusted rates, event timing, and statistical comparisons are not reported in the registry record. These safety entries describe what was posted and do not by themselves provide a complete benefit-risk assessment.

17. What the Result Does, and Does Not, Mean

The primary estimate is consistent with little relative separation in the observed overall-survival hazards: the reported hazard ratio is 0.99. Its 95% confidence interval, from 0.82 to 1.19, includes the no-relative-difference reference and allows estimates on either side. The reported p-value is 0.90, and the registry states that the study failed to meet its primary endpoint. Taken together, these entries do not establish a statistically significant superiority difference in the primary comparison.

That conclusion is narrower than saying the treatments are identical or that no possible benefit or harm exists. The confidence interval does not establish equivalence, and no equivalence or noninferiority margin is reported. The result also does not show that the true effect is precisely the point estimate. It is an estimate from this study, with uncertainty that should remain visible in interpretation.

The posted primary outcome concerns overall survival among the trial’s enrolled population with resected non-small cell lung cancer in the listed stages, comparing the randomized adjuvant regimens. It does not answer questions about other patient groups, other treatment schedules, other outcomes, or follow-up beyond what the registry describes. The posted median for one arm cannot complete a median-based comparison because the comparator median and the upper confidence limit are not reported in the registry record.

These results are not individualized predictions and should not be used as medical advice. In general, trial estimates describe group-level outcomes under the study’s design and analysis. The registry summary does not provide enough information to determine how the result applies to an individual’s circumstances. The most defensible statistical reading is therefore limited to the posted comparison: no clear difference was established on the primary superiority endpoint, while the uncertainty interval does not prove that all meaningful differences are absent.

18. Why This Trial Matters Statistically

This trial illustrates why a point estimate should be read together with its confidence interval and the trial’s stated endpoint conclusion. The primary hazard ratio of 0.99 is close to 1, but the interval from 0.82 to 1.19 shows that the estimate is not a precise demonstration of equality. The p-value and the registry’s statement that the primary endpoint was not met provide additional context, but neither removes the uncertainty represented by the interval.

It also demonstrates the distinction between superiority and equivalence. A superiority analysis that does not meet its endpoint cannot, by itself, establish that two regimens have the same effect. That requires a different inferential framework and a prespecified margin, neither of which is reported in the registry record. This distinction is important when communicating a result that is not statistically significant.

The trial also highlights differences between time-to-event summaries. A median survival value is an easily understood point in the follow-up distribution, while a Cox hazard ratio uses event timing across follow-up and is the posted comparative analysis. Where the registry supplies only one arm’s median and an incomplete confidence interval, the hazard ratio remains the more complete reported comparison.

Finally, the secondary and safety entries demonstrate why statistical interpretation depends on reporting context. Unadjusted secondary findings require caution when multiplicity procedures are not documented, and adverse-event counts need their definitions, denominators, and attribution context. The record makes several central results available but leaves important analytical details unreported, so conclusions should be limited to the information documented.

19. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The overall-survival result is reported with a hazard ratio and confidence interval, but the registry does not provide the Cox model specification, covariates, or diagnostic checks. It also does not report the number of deaths, the distribution of follow-up, or detailed censoring information beyond censoring participants thought to be alive at their last contact. These omissions limit independent assessment of the time-to-event analysis.

The median overall survival is available only for the chemotherapy-plus-bevacizumab arm, with a lower confidence limit but no posted upper limit. The chemotherapy-alone median is not reported in the registry record. This prevents a complete descriptive comparison of the arm-specific medians. The primary endpoint’s definition and timeframe are stated, but the posted results do not supply further detail on missing data or sensitivity analyses.

The record does not report the trial’s original sample-size assumptions, interim analyses, stopping rules, or complete statistical analysis plan. Nor does it report whether multiplicity was handled through an adjustment or hierarchical testing strategy. Those omissions constrain assessment of prespecification and the interpretation of secondary outcomes.

The chemotherapy arm description names several possible regimens but does not provide complete details for each regimen in the posted entry. The serious adverse event term percentages are not clearly assigned to an arm in the summary, and adverse-event deaths are not reported. The registry also does not provide enough detail in the posted disease-free survival definition to assess all event components or analysis choices. These are reporting limitations; they do not, by themselves, indicate that the trial was conducted incorrectly.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Chemotherapy With or Without Bevacizumab in Treating Patients With Stage IB, Stage II, or Stage IIIA Non-small Cell Lung Cancer That Was Removed By Surgery — NCT00324805.
• Registry record last updated 2026-06-23; status last verified 2026-05.
• Results first posted 2018-02-14.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

23. Record Summary

NCT00324805 is a completed, randomized, unmasked, parallel-group phase III trial of adjuvant chemotherapy with or without bevacizumab for completely resected stage IB, stage II, or stage IIIA non-small cell lung cancer. Enrollment was 1501. The primary endpoint was overall survival, measured from randomization to death from any cause, with follow-up described as up to 10 years.

The sponsor-reported Cox analysis gave a hazard ratio of 0.99, a 95% confidence interval from 0.82 to 1.19, and p = 0.90. The registry states that the study failed to meet its primary endpoint. Disease-free survival is the secondary endpoint with results posted. Important details, including the original sample-size assumptions, multiplicity approach, and full analysis plan, are not reported in the registry record.