A Clinical Trial of Vorinostat (MK0683, SAHA) in Combination With FDA Approved Cancer Drugs in Patients With Advanced Non-Small Cell Lung Cancer (NSCLC)(0683-056): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Clinical Trial of Vorinostat (MK0683, SAHA) in Combination With FDA Approved Cancer Drugs in Patients With Advanced Non-Small Cell Lung Cancer (NSCLC)(0683-056) (NCT00473889), a randomized phase 2, phase 3 study. The primary outcome is Overall Survival.

2. Clinical Question

The registry identifies the study as randomized and double-blind, with a parallel-group design. It describes a Phase II/III study incorporating an initial Phase II component. The planned clinical comparison concerned survival after adding vorinostat, rather than replacing paclitaxel or carboplatin. The registry's secondary outcomes were progression-free survival and the number of participants with a disease response to treatment.

Enrollment was 253. The listed arm counts for the primary outcome were 126.0 for vorinostat plus paclitaxel plus carboplatin and 127.0 for placebo plus paclitaxel plus carboplatin. The record does not give a separate description of eligibility criteria beyond the cancer stage and condition named in the trial record, nor does it report further details about the treatment schedule or dose.

3. Trial Design

This was a randomized, double-blind, parallel-group Phase II/III trial sponsored by Merck Sharp & Dohme LLC. Its official title describes a comparison of paclitaxel plus carboplatin combined with either vorinostat or placebo in patients with Stage IIIB non-small-cell lung cancer with pleural effusion or Stage IV non-small-cell lung cancer. The registry describes the trial as including an initial Phase II component within a Phase III study.

The intervention group was assigned vorinostat, intravenous paclitaxel, and intravenous carboplatin. The comparator group was assigned placebo, intravenous paclitaxel, and intravenous carboplatin. Both regimens therefore included the same named chemotherapy drugs; the randomized contrast was vorinostat versus placebo alongside those drugs. The meaning of double-blind is that masking was recorded in the registry, but the record does not specify which participant or study personnel roles were masked, or how masking was maintained.

The registry lists enrollment of 253 and reports 126.0 participants in the vorinostat group and 127.0 in the placebo group for the primary outcome. The overall survival population was intention-to-treat, defined as all randomized participants analyzed according to their randomized group. This preserves the randomized comparison in the stated analysis population, regardless of treatment received after assignment.

The recorded study start was 2007-05 and primary completion was 2008-12. The study status is terminated. The reason for termination is not reported in the registry record. The record does not report a sample-size justification, a detailed randomization procedure, stratification factors, treatment duration, dose schedule, or a more detailed account of the Phase II-to-Phase III transition. Those details cannot be inferred from the brief descriptions of design and intervention.

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.

Before assignment: Randomized participants were stratified by Stage (IIIB versus IV), geographic region and eligibility for treatment with bevacizumab prior to treatment assignment.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Intention-to-treat (ITT) population is defined as all randomized participants. Participants are counted in the group to which they are randomized.

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.

6. Primary Endpoint

The registry lists the primary outcome as Overall Survival, measured over Start of treatment to death.

The primary endpoint was overall survival, measured in months. The registry defines it as the time from randomization until death due to any cause. For a participant without a documented death at the time of final analysis, the specified approach was to censor the observation at the date of the last follow-up. The listed analysis population was intention-to-treat, including all randomized participants and retaining each participant in the group to which they were assigned.

Overall survival is a time-to-event endpoint: the outcome includes both whether death was recorded and how long each participant was followed before death or censoring. It is used to describe the time course to death from any cause, rather than a particular cause of death or a tumor response alone. That makes the endpoint directly interpretable as survival time, while also requiring careful handling of participants whose event has not been documented by the analysis cutoff.

The registry reports median overall survival of 11.0 months in the vorinostat group and 14.0 months in the placebo group. It lists full ranges of 0.2 to 17.3 months and 0.03 to 18.7 months, respectively. These are arm-specific summaries; the ranges are not confidence intervals for the medians. The record reports a stratified log-rank test but does not provide a hazard ratio or confidence interval for the between-group effect. Further detail about the timing and completeness of follow-up is not reported in the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is a time-to-event endpoint. Each participant contributes time from randomization until death or until the last follow-up date if death had not been documented by final analysis. The registry specifies the latter observation as censored. Censoring means that the event time is not known to have occurred during the observed follow-up; it does not mean that the participant is known to have survived indefinitely. The registry gives the rule but does not report how many participants were censored, their follow-up times, or reasons for incomplete observation.

What does the median describe?

In general, the median survival time is the time by which half of the estimated survival distribution has experienced the event, with the remaining half not yet having experienced it. It is a summary of each arm's survival experience, not a direct estimate of the difference between arms. The registry reports 11.0 months for vorinostat plus paclitaxel plus carboplatin and 14.0 months for placebo plus paclitaxel plus carboplatin. It also reports full ranges, from 0.2 to 17.3 months and from 0.03 to 18.7 months, respectively. A full range reflects the reported endpoints of the observed values; it is not a confidence interval and does not communicate the precision of the median.

In general, estimating a median from censored time-to-event data requires methods that account for the fact that some event times are only partly observed. The registry does not provide a survival curve or details on the estimation method used for the medians. The stated endpoint definition and censoring rule alone do not establish all the analysis choices behind the reported summaries.

What does the log-rank test assess?

The registry reports a stratified log-rank test for the comparison, with p = 0.992. In general, a log-rank test compares the pattern of events over follow-up between groups, taking account of when participants remain at risk and when events occur. A stratified version performs the comparison within strata and combines information across them. The registry does not identify the stratification factors, so their clinical or analytic role cannot be described. The test is a comparison of survival distributions, not a test that compares only the two medians.

A p-value is a summary of the compatibility of the observed test statistic with a null model, conditional on the test assumptions. It is not a probability that a treatment works, nor does it measure the size of a difference. The value p = 0.992 does not provide evidence of a difference according to the reported test, but it cannot show that the groups are equivalent. In general, a confidence interval around an effect estimate helps communicate the range of values compatible with the data. No comparative hazard ratio or confidence interval is reported in the registry record.

Why is a hazard ratio not interchangeable with a median?

In general, a hazard ratio compares the event rates between groups over time under a time-to-event model. It addresses a relative contrast across follow-up, while the median is a point on each group's survival-time distribution. Neither is a simple substitute for the other. A hazard ratio and its confidence interval would add an estimate of comparative magnitude and precision, though interpretation depends on the model and its assumptions. The record contains no hazard ratio, so no hazard-ratio recalculation or independent estimate of that effect can be made from the listed medians and ranges.

The registry reports only the sponsor's stratified log-rank result for the primary comparison. The available summary values do not include the event times or risk sets needed to reconstruct that test. The record also indicates that values for a two-arm comparison were missing or not estimable for the median-time endpoint. Consequently, no additional unadjusted two-arm estimate is available from the posted summaries. Any calculation based only on subtracting the medians would not reproduce a time-to-event test and would not account for censoring or the timing of events.

How should the analysis population be understood?

The stated intention-to-treat population includes all randomized participants and counts each in the assigned group. In general, this approach retains the randomized comparison and avoids redefining groups based on subsequent treatment received. It does not, by itself, resolve missing outcome data or establish that follow-up was complete. The record does not report exclusions after randomization, protocol deviations, treatment discontinuations, or details of missingness beyond the censoring definition.

In general, statistical methods for time-to-event data rely on assumptions about censoring, including that censoring does not systematically conceal event experience in a way that invalidates the comparison. The registry record does not provide information to assess those assumptions. It also does not report a proportional-hazards assessment, model diagnostics, or other detailed statistical plan. No conclusions about those analytic features should be drawn from the method name alone.

What can planning illustrations tell readers?

In general, sample-size calculations are prospective planning exercises based on assumptions about the anticipated effect, significance level, allocation, and analysis method. A planning illustration based on an observed effect is not the trial's original design justification and is not observed power. The record describes a normal-approximation planning approach with two-sided significance level 0.05 and equal allocation, and states that the effect to detect is a fraction of the observed effect. It does not report the original sample-size justification or additional trial-design assumptions. Any planning illustration should therefore be read as conditional on its stated assumptions, not as evidence that the trial was designed around those values or that it had a particular probability of finding an effect.

10. Analyses Reported by the Sponsor

The registry reports the sponsor's primary comparison as a stratified log-rank test, with p = 0.992. It does not give a hazard ratio, confidence interval, or the stratification factors. Those omissions limit interpretation of the comparative effect's magnitude and precision. The reported arm medians and full ranges are descriptive summaries and cannot independently reproduce the sponsor's time-to-event test.

The record indicates that a two-arm recalculation was not available for this median-time endpoint because values were missing or not estimable. Therefore, there is no independent recalculated comparative statistic to place alongside the sponsor's test. The registry's discrepancy indicator is false; there is no recorded conflict between a recalculation and the sponsor's result. This does not add information beyond the sponsor's reported p-value, and it does not turn that p-value into an effect estimate.

In general, a recalculation from summary statistics may be useful as a check when the required values are available, but it is not interchangeable with the sponsor's prespecified analysis. Here, the registry does not provide the event-level information needed to recreate the stratified log-rank test, nor does it provide a hazard ratio to independently assess. The sponsor-reported test is therefore the only reported inferential comparison for overall survival in this record. Its p-value should be read alongside the descriptive medians, while recognizing that the magnitude and precision of a relative effect are not reported.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists overall survival as the primary outcome and identifies progression-free survival and the number of participants with a disease response to treatment as secondary outcomes. It reports three outcome measures and three analyses. However, a detailed statistical analysis plan, including how the primary and secondary outcomes were ordered or jointly interpreted, is not reported in the registry record.

Multiplicity arises when several outcomes, comparisons, time points, or interim looks are tested. In general, testing multiple hypotheses can increase the chance of obtaining at least one apparently notable result by chance. Methods such as a prespecified testing sequence or adjustment can address that issue, but the record does not report whether any such strategy was planned or used. The presence of multiple listed outcomes alone does not reveal the analysis hierarchy or whether adjustment was necessary.

The primary analysis population is described as intention-to-treat: all randomized participants are counted in their assigned group. For overall survival, participants without a documented death by final analysis were to be censored at last follow-up. The record does not report how many participants were censored, the amount of follow-up by group, the number of deaths, missing-data patterns, or sensitivity analyses. It also does not state whether other analysis populations were used.

The registry does not report interim analyses, stopping boundaries, or monitoring procedures. The trial status is terminated, but the reason for termination is not reported in the registry record. Termination status alone does not establish whether the decision was related to efficacy, safety, recruitment, funding, or another cause.

In general, the interpretation of a time-to-event analysis depends on how censoring and follow-up are handled. The registry supplies a censoring definition but does not provide enough detail to evaluate whether censoring differed between groups or whether the underlying assumptions were examined. These missing details constrain assessment of multiplicity, missing data, and the robustness of the primary comparison; they should not be filled in by assuming a particular protocol or statistical plan.

12. 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 3% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: pneumonia (8.9%), non-small cell lung cancer (8.9%), febrile neutropenia (8.1%), neutropenia (7.3%), asthenia (4.8%).

The registry reports serious adverse events for groups with 124 participants at risk in each arm. Serious adverse events affected 63 participants in the vorinostat group and 45 participants in the placebo group. Other adverse events affected 114 participants and 117 participants, respectively. The registry does not provide a death count in these safety summaries. The reason the safety group counts differ from the primary-outcome arm counts of 126.0 and 127.0 is not reported in the registry record.

The most frequent listed serious-event terms were pneumonia (8.9%) and non-small-cell lung cancer (8.9%), followed by febrile neutropenia (8.1%), neutropenia (7.3%), asthenia (4.8%), and pyrexia (4.0%). These figures describe the listed terms, not an overall comparison of safety between randomized groups. The record does not identify which arm each term-specific percentage refers to, provide arm-specific term counts, or explain whether the terms can overlap within participants.

In general, an adverse-event count is a descriptive safety measure and does not by itself establish that the study treatment caused an event. Interpretation also depends on definitions, ascertainment, exposure, follow-up, and the context of the underlying condition. Details of those aspects are not reported in the registry record. The listed affected-participant counts and terms should therefore be read as registry safety summaries, not as causal estimates or a complete safety profile.

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

The posted overall-survival summaries show median values of 11.0 months for vorinostat plus paclitaxel plus carboplatin and 14.0 months for placebo plus paclitaxel plus carboplatin. The sponsor's stratified log-rank test reports p = 0.992. Taken together, these are a pair of descriptive arm summaries and a test result that does not indicate evidence of a difference in survival distributions. They do not establish that the regimens have identical effects.

The record does not report a hazard ratio or confidence interval. Without a comparative effect estimate and its precision, the size and range of plausible relative effects cannot be assessed from the record. The full ranges reported for each arm are not confidence intervals and do not resolve this limitation. A non-significant result is not a demonstration of equivalence or non-inferiority; those conclusions require an appropriate design, prespecified margin, and analysis, none of which is reported.

The result also does not speak directly to the listed secondary outcomes of progression-free survival or disease response. Their values are not included in the information described for this primary analysis. Likewise, the safety summaries provide affected-participant counts and serious-event terms, but do not establish causation or a complete between-group assessment of harms.

In general, randomized assignment helps make groups comparable at baseline in expectation, but the strength of conclusions still depends on the analysis, follow-up, missing data, and reporting. The record identifies an intention-to-treat population and a censoring rule, yet does not provide numbers censored, follow-up details, or event counts. These limitations make it difficult to judge the precision and robustness of the reported survival comparison.

The trial's terminated status is recorded, but the reason is not reported in the registry record. It should not be interpreted as evidence for or against a treatment effect. The narrow conclusion supported by the posted information is that the registry reports the stated medians and a stratified log-rank p-value, while leaving the magnitude and precision of a comparative survival effect unreported.

14. Why This Trial Matters Statistically

This trial is statistically informative as an example of why an arm-specific median and a between-group test answer different questions. The medians describe each group's survival experience at a particular point in its distribution. The stratified log-rank test compares event patterns over follow-up. A reader should not treat the medians as if they were themselves the full comparative analysis, especially when censoring is part of the endpoint definition.

The record also illustrates the distinction between evidence of a difference and evidence of equivalence. The reported p-value is 0.992, but that value does not establish that the randomized regimens are equivalent. To assess the magnitude and precision of an effect, an estimate such as a hazard ratio together with a confidence interval would be useful; neither is reported. The absence of an estimate limits what can be learned from the result even though the registry does report a test.

Another statistical issue is transparency about analysis choices. The test is identified as stratified, but the stratification factors are not reported. The record specifies an intention-to-treat population and a censoring rule, but does not describe the extent of censoring or give detailed follow-up information. These details matter to evaluating how well a time-to-event analysis represents the randomized comparison.

The trial is described as Phase II/III, randomized, double-blind, and parallel-group, and its status is terminated. The record does not report the original sample-size justification, interim analyses, or reason for termination. Recognizing those omissions helps separate the information that the registry does contain from assumptions about the conduct or design that cannot be verified from the record.

Finally, safety figures require their own careful reading. Aggregate counts of affected participants and a list of serious-event terms do not establish causation or provide a complete comparative safety assessment. The statistical value of this record lies in identifying what is reported, distinguishing descriptive summaries from inferential results, and making clear where missing estimates and design details constrain interpretation.

15. 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 main limitation for interpreting overall survival is the absence of a reported hazard ratio and confidence interval. The registry gives arm-specific median survival and full ranges, plus a stratified log-rank p-value, but those items do not quantify the comparative effect and its precision in the same way as an effect estimate with an uncertainty interval. The full ranges should not be mistaken for confidence intervals.

The stratified log-rank method is named, but its stratification factors are not reported. The record does not provide event counts, numbers censored, survival curves, follow-up summaries, or enough event-level information to reconstruct the test. It states that participants without documented death at final analysis were censored at last follow-up, but does not describe the amount or pattern of censoring. The method assumptions therefore cannot be assessed from the available record.

The analysis population is defined as intention-to-treat, but details of participant flow after randomization, exclusions, treatment discontinuation, adherence, or protocol deviations are not reported. Safety counts are based on 124 participants at risk in each arm, while the primary-outcome groups contain 126.0 and 127.0 participants; the reason for the difference is not reported in the registry record. The event-term percentages are not assigned to specific arms in the listed information.

The trial is marked terminated, but the reason is not reported. The record does not report the original sample-size justification, interim analyses, stopping rules, detailed multiplicity strategy, or full statistical analysis plan. It also does not provide the detailed results for the listed secondary outcomes in the information used for this analysis. These gaps prevent a complete assessment of design assumptions, robustness, secondary findings, and the context of termination.

These limitations do not invalidate the reported summaries, but they restrict how far the conclusions can go. The medians and p-value can be described as recorded; the relative effect's magnitude, precision, and sensitivity to analytic choices remain unreported.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: A Clinical Trial of Vorinostat (MK0683, SAHA) in Combination With FDA Approved Cancer Drugs in Patients With Advanced Non-Small Cell Lung Cancer (NSCLC)(0683-056) — NCT00473889.
• Registry record last updated 2015-07-03; status last verified 2015-06.
• Results first posted 2009-10-12.
• 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.

19. Record Summary

NCT00473889 is a randomized, double-blind, parallel-group Phase II/III trial of vorinostat with paclitaxel and carboplatin versus placebo with paclitaxel and carboplatin in Stage IIIB or Stage IV non-small-cell lung cancer. The registry lists enrollment of 253 and an intention-to-treat primary population.

The primary outcome was overall survival from randomization to death from any cause, with censoring at last follow-up for participants without documented death at final analysis. Median survival was 11.0 months in the vorinostat group and 14.0 months in the placebo group. The sponsor-reported stratified log-rank test gave p = 0.992. A hazard ratio and confidence interval are not reported in the registry record.

The record lists progression-free survival and disease response as secondary outcomes. It reports serious adverse events affecting 63 participants in the vorinostat group and 45 in the placebo group, with 124 participants at risk in each group. Study status is terminated; the reason is not reported in the registry record.