Phase III Study of S-1 + Cisplatin vs Cisplatin in Cervical Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Phase III Study of S-1 + Cisplatin vs Cisplatin in Cervical Cancer (NCT00770874), a randomized phase 3 study. The primary outcome is Overall Survival.

2. Clinical Question

3. Trial Design

This was a Phase 3, randomized, open-label, multicenter, multinational, two-arm, parallel study of S-1 plus cisplatin compared with single-agent cisplatin. The registry lists an enrollment of 375. The trial status is completed. The registry record gives a start date of 2008-09 and a primary completion date of 2015-12.

Participants were assigned to one of two groups: S-1 plus cisplatin, designated Arm A, or cisplatin alone, designated Arm B. Because masking was none, the study was open-label rather than blinded. The registry does not report the randomization sequence method, allocation concealment, or any procedures intended to reduce the influence of open-label treatment knowledge on outcome assessment.

The primary outcome was overall survival in the full analysis set. The registry defines its measurement period as starting at randomization and ending at death from any cause, with assessment up to 296 events or the end of November 2015, whichever occurred earlier, at three-month intervals. This is a time-to-event outcome: participants can have different lengths of follow-up, and the event of interest is death.

The registry also names progression-free survival and safety as secondary outcomes. It reports one posted secondary outcome, but the detail needed to identify and interpret its result is not reported in the registry record. The trial’s original sample-size calculation, its planned statistical analysis in full, and any interim-analysis arrangements are not reported in the registry record. The planning illustrations based on the observed effect are separate from those original design details.

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

Analysis population for the primary outcome, as defined in the registry: Full analysis set

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: A total of 364 (188 in Arm A; 176 in Arm B) received study treatment. Two patients in Arm B were excluded in the efficacy analysis, because of insufficient informed consent, and missing efficacy data.

6. Primary Endpoint

The registry lists the primary outcome as Overall Survival, measured over From the date of randomization to death from any cause, assessed up to 296 events or the end of November 2015, whichever was earlier, each three months.

Overall survival measures the time from randomization until death from any cause. In this trial, the registry specifies that the outcome was assessed at three-month intervals and that the stated assessment period ended at 296 events or the end of November 2015, whichever came first. The primary analysis population is described as the full analysis set.

Time-to-event outcomes are useful when the timing of an event matters and follow-up duration may vary across participants. They can use information about both whether an event occurred and when it occurred, while accommodating participants whose event time is not observed during the study period. Such participants are generally treated as censored in time-to-event analyses, but the registry record does not describe the censoring rules or the handling of loss to follow-up for this trial.

The registry reports the median overall survival in each arm and a hazard ratio comparing the groups. A median is the time by which half of the analyzed group has experienced the event, as estimated from the survival data. A hazard ratio summarizes the relative event rate over follow-up under the assumptions of the analysis; it is not a direct comparison of the two medians. The posted record does not report the survival curves or enough methodological detail to assess the assumptions underlying the analysis.

7. Results Posted in the Registry

• S-1 + Cisplatin (Arm A): hazard ratio 0.84 (95% CI 0.67–1.05), reported p = 0.125 (as reported in the registry).

8. Statistical Interpretation

9. Statistical Methods Explained

What does a time-to-event analysis compare?

Overall survival is a time-to-event outcome because it records the time from randomization until death. Unlike an outcome measured once at a fixed visit, time-to-event data can include participants with different observation periods. Some participants may not have experienced the event by the end of their observed follow-up; their data do not establish an event time within that period. Standard survival methods can retain information from such participants up to the point their follow-up ends, subject to assumptions about censoring. The registry record does not describe the censoring definitions or give enough information to assess whether those assumptions were appropriate in this study.

The registry reports median overall survival by arm. In general, the median is the time at which the estimated survival function reaches one half. It gives a central point of the estimated time-to-event distribution, but does not describe the whole curve. Two arms can have different patterns before or after their medians, so the median values alone cannot summarize every aspect of the survival experience.

What does the hazard ratio express?

The posted comparison uses a hazard ratio and identifies the method as log rank. A hazard is the instantaneous event rate among participants who have not yet experienced the event, conditional on the analysis framework. A hazard ratio compares hazards between groups. It is not a ratio of median survival times, a difference in months, or the fraction of participants who died. The reported hazard ratio of 0.84 is oriented as S-1 plus cisplatin relative to cisplatin. The registry does not provide enough detail to determine whether the proportional-hazards assumption was assessed or how any departures from it would have been handled.

The log-rank method is a general approach for comparing time-to-event experience between groups. It uses event timing over follow-up rather than relying only on the medians. The registry labels the reported analysis as log rank and gives a hazard ratio, confidence interval, and p-value. It does not give further details about model specification, covariate adjustment, stratification, or the handling of ties. Those elements should not be inferred from the method label alone.

How should the uncertainty measures be read?

A confidence interval describes uncertainty around an estimated quantity under the statistical method used. The 95% confidence interval for the hazard ratio extends from 0.67 to 1.05. Since this interval includes 1, it includes the value corresponding to equal hazards in the comparison. The interval also includes values below and above 1, so the data are not precise enough to restrict the estimate to one direction. A confidence interval is not a range containing 95% of individual participants’ outcomes, nor is it a direct probability statement about the true hazard ratio.

The log-rank p-value is 0.125. In general, a p-value describes how unusual data at least as incompatible with a specified null model would be under that model and the test assumptions. Its interpretation depends on the analysis plan and the assumptions of the method. The registry does not report the planned significance threshold or whether any adjustment for multiple testing was intended; neither should be assumed.

What can be learned from the posted comparison?

The posted comparison indicates the estimated direction and uncertainty of the relative time-to-event result. It does not establish that any observed difference is caused by treatment, and a p-value alone cannot determine clinical importance. Interpretation also depends on trial conduct, the analysis population, missing follow-up, and the validity of the method’s assumptions. The registry record identifies the full analysis set but does not report its definition, participant flow, or detailed analysis procedures.

The study was open-label, which means treatment assignment was not masked. Overall survival is an event with a relatively objective definition, but open-label design can still be relevant to decisions around follow-up, co-interventions, or the timing and completeness of outcome ascertainment. The registry does not describe these procedures. This is a reason to distinguish the posted numerical comparison from a complete appraisal of study conduct.

10. Confidence Intervals and Precision

The median estimate in the S-1 plus cisplatin arm is 21.9 months, with a 95% confidence interval from 18.6 to 25.8 months. The median estimate in the cisplatin arm is 19.5 months, with a 95% confidence interval from 17.0 to 24.3 months. These intervals indicate uncertainty in the estimated median for each group. They do not give the range of survival times for individual participants.

For the between-group comparison, the sponsor reported a hazard ratio of 0.84 and a 95% confidence interval from 0.67 to 1.05. This interval includes 1, the value corresponding to equal hazards. It therefore does not confine the comparison to a single direction. The interval’s endpoints also show that the point estimate should not be treated as an exact or certain value. The log-rank p-value was 0.125.

Precision is affected by the amount and timing of event information, follow-up, and the variability in the data, among other factors. The registry’s outcome timeframe refers to assessment up to 296 events or the end of November 2015, whichever occurred earlier, but it does not report the observed number of deaths in the primary analysis or provide a detailed account of follow-up. Those omissions make it difficult to assess the sources of uncertainty beyond the reported intervals.

The confidence intervals for the arm-specific medians should not be used as a formal test of the difference between arms. The reported hazard-ratio interval and log-rank analysis are the posted inferential comparison. Precision concerns how narrowly an estimate is supported by the data; it is distinct from whether a result is clinically important.

11. 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 illustrations are expressed as total numbers of events for this time-to-event outcome, not participants per arm. They use the Schoenfeld approach and are planning illustrations based on the observed effect, rather than the trial’s own original sample-size calculation. The registry’s original design assumptions and sample-size justification are not reported in the registry record.

The listed assumptions for these illustrations are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to be detected is set in turn as 100%, 75%, or 50% of the observed effect. The event totals for each effect assumption are given for 80% and 90% power.

In plain language, an 80% power column gives the total event count targeted by the illustration to have an 80% chance of detecting the specified effect under its assumptions. The 90% power column gives the corresponding event count for a higher chance of detecting that effect, again under those assumptions. These are not participant counts per arm because the calculation is for total events in a time-to-event analysis.

The figures increase as the assumed effect becomes smaller: detecting a less pronounced difference generally requires more event information. These planning counts are not observed power and should not be read as a description of what the completed trial achieved. They are hypothetical planning quantities based on the stated effect fractions and assumptions, not a substitute for the trial’s unreported original design plan.

12. 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 = 0.840, so ln(HR) = -0.1744.
2. d = 4 × (1.96 + 0.842)² / [ln(0.840)]² = 1,032.8, rounded up to 1,033 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 <- 0.84                              # 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(770874)
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 = 1,033 (80% power), 1,383 (90% power)
 75% of observed effect: events = 1,837 (80% power), 2,458 (90% power)
 50% of observed effect: events = 4,132 (80% power), 5,531 (90% power)
Simulated power with 1,033 events: 0.806 (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.

13. Analyses Reported by the Sponsor

The registry reports the sponsor’s overall-survival comparison as a hazard ratio of 0.84, with a 95% confidence interval from 0.67 to 1.05 and a p-value of 0.125 using the log-rank method. The posted comparison is the appropriate source for the reported time-to-event inferential result. The median survival estimates are separately reported for each arm and do not allow the hazard ratio to be reconstructed from the medians alone.

No separate unadjusted hazard-ratio recalculation is used for this time-to-event comparison. The available arm-level medians and confidence intervals do not contain the event-time information needed to independently reproduce a log-rank hazard-ratio analysis. The registry’s reported analysis therefore remains the basis for describing the comparison.

The reported hazard ratio and the comparison entry agree. The interval includes 1, and the p-value is 0.125. These features should be described together rather than presenting the point estimate alone. The registry record does not report additional modeling details that would permit an independent assessment of adjustment, assumptions, or alternative analysis choices.

14. Multiplicity, Missing Data and Analysis Populations

The registry names progression-free survival and safety as secondary outcomes and indicates that one secondary outcome was posted. It does not report enough information to determine from the record which secondary outcome is represented by that count or to review a complete set of secondary outcome estimates. Details of the secondary outcome definitions, analysis populations, and statistical results are not reported in the registry record.

The record does not report whether the trial used a multiplicity strategy for the primary and secondary outcomes, including any hierarchy, adjustment, or rules for interpreting several tests. When multiple outcomes are examined, the chance of at least one apparently unusual result can differ from that for a single test. Any formal adjustment or prespecified testing order must come from the analysis plan; it cannot be inferred from the number of outcomes listed.

The primary outcome population is named as the full analysis set, but its definition and the treatment of participants with missing outcome data are not reported in the registry record. The registry also does not describe censoring rules, withdrawals, loss to follow-up, or sensitivity analyses for missing data. For overall survival, the event is death from any cause, while participants without an observed event by the end of follow-up may contribute censored information in a time-to-event analysis. The specific implementation in this trial is not detailed.

The record does not report an interim-analysis plan, stopping boundaries, or an adjustment to the primary analysis for interim looks. It would be inappropriate to assume such procedures were or were not used. The absence of these details limits assessment of the analysis plan and how fully the posted p-value reflects the trial’s intended inferential framework.

These reporting gaps do not alter the numerical results that are posted, but they restrict a fuller evaluation of their context. The reported primary analysis can be described as presented, while conclusions about missingness, multiplicity, interim monitoring, and analysis populations beyond the stated full analysis set remain uncertain.

15. Safety Results

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

Deaths

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 (6.4%), haemoglobin decreased (5.3%), thrombosis (5.3%), diarrhoea (4.8%), urinary tract infection (2.7%).

The registry lists serious adverse events for both treatment arms. In the S-1 plus cisplatin group, 72 of 188 participants are recorded as having a serious adverse event; in the cisplatin group, the corresponding count is 34 of 175. The adverse-event record also lists other adverse events for 188 of 188 in the first arm and 172 of 175 in the second. Death counts in the adverse-event groups are 7 and 3, respectively. These are descriptive registry counts and do not establish that an event was caused by treatment.

The listed serious-event terms include neutrophil count decreased at 6.4%, haemoglobin decreased at 5.3%, thrombosis at 5.3%, diarrhoea at 4.8%, urinary tract infection at 2.7%, and platelet count decreased at 2.7%. The registry does not specify an arm alongside these term percentages in the posted list, so they should not be attributed to either group separately.

Safety counts depend on the event definitions, observation period, reporting rules, and analysis population. Those details are not reported in the registry record. The denominators also differ between the arm-level serious and other-event summaries, so each count should be read with its stated denominator rather than treated as a direct comparison of rates. The figures describe what was posted; they do not by themselves establish comparative safety or causality.

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

The posted results describe overall survival in a randomized comparison of S-1 plus cisplatin with cisplatin alone. The median estimates were 21.9 months and 19.5 months, respectively. The sponsor-reported hazard ratio was 0.84, with a 95% confidence interval from 0.67 to 1.05, and a log-rank p-value of 0.125. The estimate points below 1, but its confidence interval includes 1 and also extends above 1. The result therefore has uncertainty in both the magnitude and direction of the relative comparison.

These statistics do not establish that one treatment is better or worse. A point estimate is not a definitive treatment effect, and the p-value is not a measure of clinical importance. The registry does not report the planned significance threshold, detailed analysis plan, observed event count, or full follow-up information needed to place the estimate in a more complete statistical context.

The findings also do not establish equivalence or prove that the treatments have identical outcomes. A confidence interval that includes 1 is compatible with no relative hazard difference, but it is also compatible with differences in either direction. Demonstrating equivalence or non-inferiority requires a prespecified framework and margins; these are not reported in the registry record.

Overall survival is an important time-to-event endpoint, but the registry’s posted summary does not provide every element needed to judge the overall evidence. The record gives median estimates and a hazard ratio, while further details about censoring, missing follow-up, analysis assumptions, secondary outcomes, and multiplicity are unavailable. The appropriate reading is therefore a cautious description of the reported estimates and their precision, without extending the result beyond what the record documents.

17. Why This Trial Matters Statistically

This trial illustrates why a time-to-event comparison should not be reduced to a difference between two median values. The registry reports arm-specific median overall survival as well as a hazard ratio with a confidence interval and log-rank p-value. Each statistic has a distinct interpretation: medians describe a point on each arm’s estimated survival distribution, while the hazard ratio summarizes a relative comparison over follow-up.

It also demonstrates the importance of reading a point estimate alongside its uncertainty. The reported hazard ratio is 0.84, but the 95% confidence interval ranges from 0.67 to 1.05 and includes 1. A point estimate alone would conceal the fact that the posted data do not constrain the comparison to a single direction. The p-value of 0.125 adds information about the test result but is not a measure of effect size, clinical importance, or the probability that a hypothesis is true.

The study’s randomized design supports a comparative analysis, while the open-label design and limited reporting of methods leave important questions for a complete appraisal. The registry identifies the primary analysis population and endpoint timeframe, but does not document several analysis details, including the original sample-size justification, multiplicity plan, or handling of missing data. Distinguishing posted evidence from unreported design information is essential when interpreting registry results.

Finally, the planning illustrations show how required event information depends on the effect a future study aims to detect and the chosen power target. Those illustrations are not the completed trial’s observed power and do not reveal the assumptions used when the trial was originally designed. This distinction prevents hypothetical planning quantities from being mistaken for evidence about the completed comparison.

18. 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 registry record is not a complete statistical report. It does not provide the original sample-size justification, planned significance threshold, detailed analysis plan, or interim-analysis arrangements. The planning illustrations use the observed effect under stated assumptions and should not be treated as a reconstruction of the trial’s original design.

Important details of the primary analysis are also absent. The record does not define the full analysis set, describe censoring or missing-data procedures, report the observed number of deaths in the analysis, or provide follow-up summaries. It identifies a log-rank method and a hazard ratio but does not report whether assumptions were assessed, whether covariates were included, or whether the analysis was stratified. These omissions limit independent evaluation and reproduction of the analysis.

The registry describes the study as open-label and gives no masking. It does not report the method of sequence generation or allocation concealment. Overall survival is based on death from any cause, but a fuller assessment would still benefit from information on outcome ascertainment and completeness of follow-up. The record does not provide these details.

The secondary outcomes named are progression-free survival and safety, while the record indicates one posted secondary outcome. The full secondary results and any multiplicity approach are not reported in the registry record. The serious-event term percentages are listed without an arm specified in the posted term list, and the available adverse-event summaries do not establish causation. These limits call for restraint when comparing safety or interpreting secondary outcomes.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Phase III Study of S-1 + Cisplatin vs Cisplatin in Cervical Cancer — NCT00770874.
• Registry record last updated 2019-06-21; status last verified 2019-03.
• Results first posted 2019-06-21.
• 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.

22. Record Summary

NCT00770874 is a completed Phase 3 randomized, open-label, parallel study comparing S-1 plus cisplatin with single-agent cisplatin in stage IVB, recurrent, or persistent carcinoma of the cervix. The registry lists an enrollment of 375 and describes the study as multicenter and multinational. The recorded start date is 2008-09, and the primary completion date is 2015-12.

The primary outcome was overall survival in the full analysis set, measured from randomization to death from any cause. Median overall survival was 21.9 months in the S-1 plus cisplatin arm and 19.5 months in the cisplatin arm. The sponsor-reported hazard ratio was 0.84, with a 95% confidence interval from 0.67 to 1.05 and a log-rank p-value of 0.125.

The registry also names progression-free survival and safety as secondary outcomes. The original sample-size justification and several details needed to assess the analysis plan, missing data, and multiplicity are not reported in the registry record.