A Randomized, Double-Blinded, Placebo-Controlled Study of Sorafenib in Patients With Advanced Hepatocellular Carcinoma: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Randomized, Double-Blinded, Placebo-Controlled Study of Sorafenib in Patients With Advanced Hepatocellular Carcinoma (NCT00492752), a randomized phase 3 study. The primary outcome is Overall Survival.

2. Clinical Question

The central statistical question is about the distribution of time to death, not simply whether a participant was alive at one selected visit. A time-to-event analysis can use information from participants whose follow-up ends before death is observed, provided their observation is handled as censoring. The registry defines censoring for participants alive at analysis using their last contact, and its population description also refers to censoring at the last follow-up or data cutoff for participants alive or lost to follow-up.

Overall survival is a direct endpoint for comparing time until death from any cause. It does not, by itself, explain why survival times differ, establish the pathway through which any difference arose, or describe every dimension of a participant’s experience. The other outcomes listed in the registry address some distinct questions, including symptomatic progression, tumor progression, disease control, response, and patient-reported measures.

3. Trial Design

The registry describes a randomized, triple-masked, parallel-group phase 3 study comparing sorafenib with placebo in patients with advanced hepatocellular carcinoma. The stated enrollment is 226. The record is marked completed and identifies Bayer as sponsor. The study start is recorded as 2005-10 and primary completion as 2007-03.

Participants were assigned to oral sorafenib or matching placebo. Sorafenib was described as 400 mg twice daily, with two permitted dose-reduction levels for treatment-emergent adverse events related to study treatment: 400 mg once daily and 400 mg every 2 days. Placebo tablets matching in appearance were given twice daily. The registry calls the design triple-masked; it does not specify in the record which three roles were masked, so those roles should not be inferred.

The posted overall-survival arm summaries list 150 participants in the sorafenib group and 76 in the placebo group. These arm figures are used as recorded in the outcome results and need not be treated as a description of every analysis population or every safety exposure period. The registry also reports a placebo open-label-only safety group of participants who switched to sorafenib.

The study’s stated purposes included assessing whether participants receiving sorafenib would live longer, evaluating patient-reported outcomes, examining tumor growth or shrinkage and metastases, and determining pharmacokinetics in patients with liver cancer. Secondary outcomes include symptomatic progression, time to progression, disease control, changes in FACT-Hepatobiliary measures, tumor response, duration of response, and time to response.

The registry record does not report the trial’s original sample-size justification, the detailed randomization procedure, or which roles were included in the masking description. It also does not report a detailed statistical analysis plan in the information summarized for these results. Those omissions limit reconstruction of the design decisions from the registry alone.

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: 271 subjects were enrolled in a 28-day screening period; 226 subjects were randomized either to Sorafenib or placebo (2:1 ratio) (intent-to-treat \[ITT\] population: for efficacy analysis); 224 subjects received at least one dose of study drug (safety population: for safety analysis). Majority of screen failures did not meet the inclusion criteria.

Double Blind Treatment

Reasons for not completing this period:

Follow-up and/or Open Label Sorafenib

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: In this study the overall survival was measured for the ITT population from the date of randomization until the date of death due to any cause. For patients alive or lost to follow-up at the time of analysis, time to death was to be censored at their last date of follow-up, or at the data cut-off of 09 Aug 2007 (23 months after randomization).

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 From randomization of the first subject until the data cut-off date approximately 23 months after start of randomization.

The primary endpoint is overall survival, reported as a median time in days with a 95% confidence interval. The registry defines overall survival as time from randomization to death due to any cause. Participants still alive at analysis were censored at their last date of contact. The population description specifies the intention-to-treat population and describes censoring at last follow-up, or at the data cutoff, for participants alive or lost to follow-up.

A median survival time marks the point at which the estimated survival distribution reaches one-half; it is a summary of the distribution, not a prediction of an individual participant’s survival time. The registry reports a median of 198 days for sorafenib, with a 95% confidence interval from 169 to 230 days, and 127 days for placebo, with a 95% confidence interval from 114 to 166 days.

General explanation: Overall survival is commonly used in time-to-event research because the event is death from any cause, rather than a judgment about a particular symptom or a radiographic change. Censoring allows follow-up information to contribute when the event has not been observed by the participant’s last contact or the analysis cutoff. Such methods rely on assumptions about how censoring relates to the event process; the registry’s posted description does not provide enough detail to assess those assumptions.

The median alone does not describe the full survival curve or the timing of all deaths. The registry’s hazard-ratio analyses provide a separate comparison of event rates over follow-up, while their confidence intervals indicate uncertainty in those estimates.

7. Results Posted in the Registry

• Sorafenib (Nexavar, BAY43-9006) vs Placebo: hazard ratio 0.68 (95% CI 0.50–0.93), reported p = 0.014 (as reported in the registry).
• Sorafenib (Nexavar, BAY43-9006) vs Placebo: hazard ratio 0.62 (95% CI 0.45–0.86), reported p = 0.003 (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.

Time to Symptomatic Progression (TTSP)

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor (Log Rank); it was not recalculated.
• Result:
  • Sorafenib (Nexavar, BAY43-9006) vs Placebo: hazard ratio 0.90 (95% CI 0.67–1.22), reported p = 0.498 (as reported in the registry).
  • Sorafenib (Nexavar, BAY43-9006) vs Placebo: hazard ratio 0.88 (95% CI 0.64–1.21), reported p = 0.438 (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 Sorafenib (Nexavar, BAY43-9006) vs Placebo and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval includes no difference (a ratio of 1), so these data do not show a statistically significant difference between Sorafenib (Nexavar, BAY43-9006) vs Placebo and the control. The interval gives the range of effects that remain compatible with the data.

Time to Progression (TTP)

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor (Log Rank); it was not recalculated.
• Result:
  • Sorafenib (Nexavar, BAY43-9006) vs Placebo: hazard ratio 0.57 (95% CI 0.42–0.79), reported p < 0.001 (as reported in the registry).
  • Sorafenib (Nexavar, BAY43-9006) vs Placebo: hazard ratio 0.54 (95% CI 0.38–0.77), reported p < 0.001 (as reported in the registry).
• Interpretation:
  • The 95% confidence interval excludes no difference (a ratio of 1), so the hazard of the event is statistically lower for Sorafenib (Nexavar, BAY43-9006) vs Placebo than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.
  • The 95% confidence interval excludes no difference (a ratio of 1), so the hazard of the event is statistically lower for Sorafenib (Nexavar, BAY43-9006) vs Placebo than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The secondary time-to-event outcomes show a mixed pattern rather than a uniform shift across every measured question. The reported time-to-symptomatic-progression comparisons have confidence intervals that include 1 and p-values of 0.498 and 0.438. The time-to-progression comparisons have intervals below 1 and p-values of 0.001 and 0.001. Their endpoints represent different events, so their results should not be treated as interchangeable measures of the primary outcome.

The posted medians also differ by endpoint: time to symptomatic progression is reported as 105 days for sorafenib and 103 days for placebo, while time to progression is reported as 84 days and 41.5 days, respectively. These summaries give context for the secondary comparisons, but do not replace the associated time-to-event analyses or establish what happened for every participant.

Overall survival is the primary endpoint and the main basis for the trial’s central statistical question. Secondary outcomes can help describe other aspects of disease course and reported symptoms, but a result on one secondary endpoint does not prove a change in overall survival, and the primary result does not automatically establish a result on every secondary endpoint. The record lists further outcomes, including disease control, patient-reported measures, tumor response, and response timing; outcome titles alone are not results.

Unadjusted secondary results carry less weight when several outcomes are examined because repeated testing increases the chance of finding apparently strong evidence by chance. General explanation: A multiplicity strategy can control that risk by defining which comparisons are confirmatory and how statistical thresholds are allocated. The registry information summarized for these results does not report whether such a strategy was used. Therefore the secondary findings should be viewed as supportive or exploratory evidence unless the full study analysis plan establishes their testing status.

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

• Time to Symptomatic Progression (TTSP)
• Time to Progression (TTP)
• Disease Control
• Change in Functional Assessment of Cancer Therapy (FACT) Hepatobiliary Symptom Index-8 (FHSI-8) Score From Baseline to Cycle 1 and Cycle 3
• Change in Functional Assessment of Cancer Therapy-Hepatobiliary (FACT-Hep) Score From Baseline to Cycle 3 and End of Treatment
• Number of Participants With Different Tumor Response
• Duration of Response
• Time to Response

9. Statistical Interpretation

10. Statistical Methods Explained

The registry identifies the primary endpoint as a median time-to-event outcome and reports hazard ratios from Log Rank analyses. These are related but distinct summaries. The median describes a location in an estimated event-time distribution. A hazard ratio compares the event hazard between groups over follow-up, subject to the assumptions and implementation of the survival analysis. Neither statistic is a simple count comparison.

What does censoring mean for overall survival?

General explanation: In a time-to-event study, a participant may reach the end of available follow-up without the event being observed. That participant’s information is censored at the last known follow-up or the analysis cutoff: the analysis knows that the event had not been observed up to that point, but does not know the later event time. The registry defines overall survival as time from randomization to death from any cause and states that participants alive at analysis were censored at their last contact. Its population description also mentions censoring at last follow-up or the data cutoff for participants alive or lost to follow-up.

Censoring is not the same as treating a participant as if the event never occurred. Survival methods retain the observed follow-up time and account for the fact that the later event time is unknown. These methods generally require that censoring not create a systematically misleading comparison of future event risk between groups. The registry record does not report diagnostic checks or further details that would allow that assumption to be evaluated.

How should the median and hazard ratio be distinguished?

General explanation: The median survival time is the time at which the estimated survival probability reaches one-half. It offers an interpretable summary in the same time units as follow-up, but it does not summarize every feature of the survival curve. A hazard is an instantaneous event rate among participants who have not yet had the event, and a hazard ratio compares that rate between groups. The hazard ratio is not the ratio of the medians, nor does it directly state the probability that an individual will experience the event.

The registry gives median overall survival of 198 days for sorafenib and 127 days for placebo. It reports two hazard-ratio estimates from Log Rank analyses, each with a 95% confidence interval and p-value. The record does not identify the reason for the two analyses in the information. They should therefore be read as two reported analyses rather than assigned an unreported rationale.

What do the confidence intervals and p-values convey?

General explanation: A confidence interval expresses uncertainty in an estimated quantity under the statistical procedure used. For a hazard ratio, the value 1 represents equal hazards between groups. An interval wholly below 1 is consistent with a lower estimated hazard in the intervention group under that model; an interval spanning 1 includes values compatible with no hazard difference as well as values in either direction. The width of an interval reflects precision, but its interpretation also depends on the analysis assumptions and data.

A p-value measures how unusual data at least as incompatible with a specified null hypothesis would be under that hypothesis and the test’s assumptions. It is not the probability that the null hypothesis is true, nor does it quantify the size or clinical importance of an effect. The record reports p-values for its Log Rank analyses, but the registry summary does not provide a broader account of the prespecified testing strategy or multiplicity control.

Why does the analysis use time-to-event methods?

General explanation: Time-to-event methods are useful when participants have different lengths of follow-up or when some event times are not observed before data collection ends. They use both the timing of events and the periods during which participants remain under observation without an event. A simple comparison of event counts would not represent those varying observation times in the same way. The sponsor used Log Rank analyses for the reported hazard ratios, as recorded in the registry.

The registry describes the overall-survival analysis as using the intention-to-treat population. That label generally refers to analyzing participants according to their randomized group, regardless of later treatment changes, but the registry’s own detail about handling every possible circumstance is limited. The safety record separately identifies a placebo open-label-only group of participants who switched to sorafenib. The available summary does not establish how each such participant contributed to every efficacy analysis.

What can be learned from the analysis record?

The registry reports median estimates, confidence intervals, hazard ratios, and Log Rank p-values for the primary outcome, as well as corresponding time-to-event results for two secondary outcomes. The primary outcome’s reported hazard-ratio intervals lie below 1; the time-to-symptomatic-progression intervals include 1, while the time-to-progression intervals lie below 1. These are descriptions of the posted estimates, not proof that all analysis assumptions hold or that every outcome was tested according to a prespecified hierarchy.

The record does not report an interim-analysis plan, a detailed missing-data analysis beyond the stated censoring rules, or the trial’s original sample-size assumptions. It also does not report adjustment for multiple outcome comparisons in the information. Absence of those details in the registry record should not be taken as proof that a procedure was or was not used; it means the procedure cannot be assessed from this record.

11. Confidence Intervals and Precision

For overall survival, the registry reports a median of 198 days for sorafenib, with a 95% confidence interval from 169 to 230 days. For placebo, the median is 127 days, with an interval from 114 to 166 days. These intervals convey uncertainty in the estimated medians. They are not ranges expected to contain most participants’ survival times.

The two reported overall-survival hazard-ratio estimates have 95% confidence intervals that lie below 1. The registry’s reported estimates are 0.678 with an interval from 0.496 to 0.927, and 0.621 with an interval from 0.450 to 0.857. Their intervals indicate uncertainty around the relative hazard estimates; neither should be interpreted as a precise prediction for an individual. The registry does not explain why it reports two estimates or how readers should distinguish their analysis populations.

General explanation: A confidence interval is informative about precision as well as direction. A narrower interval indicates less statistical uncertainty on the estimate’s scale than a wider interval, while an interval extending across the null value signals that the data are compatible with a wider range of effects. Interval width depends on event information, follow-up, and the analysis method, among other factors. The interval does not account for every possible design or reporting limitation.

The primary median intervals overlap, while the hazard-ratio intervals are the registry’s direct relative comparison. Overlap or non-overlap between separate group-specific intervals is not, by itself, a substitute for the reported between-group test. The trial’s comparison should be read from its stated Log Rank results and their confidence intervals rather than inferred from the median intervals alone.

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 sample-size calculations are planning illustrations based on the observed effect, not the original trial design. For the time-to-event outcomes, each entry is a required total number of events, not participants per arm. The assumptions recorded for these illustrations are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. They consider detecting the stated fraction of the observed effect. The trial’s own original design assumptions are not reported in the registry record.

The 80% power column represents a planning target under which a study with the stated assumptions would have an 80% chance of detecting an effect of the specified size, if that effect and the model assumptions were correct. The 90% power column uses a higher target chance of detection. These are design calculations, not guarantees about a future study’s outcome. Since the time-to-event figures count events, translating them into a participant enrollment target would require additional information about follow-up and the expected proportion of participants who experience the event; that translation is not given by these event totals.

The rows compare planning for the full observed effect with planning for smaller fractions of that effect. Smaller effects need more information to distinguish them from random variation, so the illustrations require more events when the target is a smaller fraction. The values depend on the observed estimate and the listed simplifying assumptions. They are not observed power: observed power is not a useful substitute for interpreting the estimate and its confidence interval after a study has been completed.

Read the event totals as conditional illustrations, not as a recommendation that a future trial should use any particular target. The calculations do not establish the appropriate enrollment, follow-up duration, allocation, or testing plan for another study. The registry does not report that these planning calculations were used when the trial was designed.

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 = 0.678, so ln(HR) = -0.3882.
2. d = 4 × (1.96 + 0.842)² / [ln(0.678)]² = 208.4, rounded up to 209 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 <- 0.6783                              # 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(492752)
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 = 209 (80% power), 279 (90% power)
 75% of observed effect: events = 371 (80% power), 496 (90% power)
 50% of observed effect: events = 834 (80% power), 1,116 (90% power)
Simulated power with 209 events: 0.802 (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. Group-Sequential Design Analysis

Many time-to-event trials, especially in oncology, stop or report early when interim data are overwhelming, so a single fixed-sample number understates the real design problem. The calculation below is an illustrative group-sequential design built from this trial's posted control median and hazard ratio. The registry record does not state the planned number of looks or the spending function, so the design is a planning illustration, not a reconstruction of the protocol. Assumed: one-sided α = 0.025, 90% power, three equally spaced looks on information, Lan-DeMets O'Brien-Fleming spending, 24 months of accrual, 60 months in total and a 1% monthly dropout hazard.

Under these assumptions the design needs about 304 participants in total, with the boundaries below.

Fixed design versus group-sequential alternatives

Group-sequential designs cost a small increase in maximum sample size and events compared with the fixed design, in exchange for the chance to stop early for overwhelming efficacy. O'Brien-Fleming spending keeps early boundaries very conservative and the final boundary close to the fixed-design value, whereas Pocock spending spends more alpha early.

Sensitivity to a smaller true effect

Planning for exactly the observed effect is optimistic, because observed effects in published trials tend to be somewhat larger than the true effect. The table repeats the main design (three looks, O'Brien-Fleming spending, 90% power) for smaller true hazard ratios.

For context, this trial enrolled 226 participants.

Cross-check, upper boundaries: gsDesign 3.71, 2.511, 1.993 | rpact 3.71, 2.511, 1.993. Agreement between the two packages confirms the boundary calculation.

R code and output

suppressWarnings(suppressMessages({ library(gsDesign); library(rpact) }))
med_control <- 4.17534246575342          # control-arm median, months (converted from the posted unit)
hr          <- 0.6783          # hazard ratio used for planning
lambdaC     <- log(2) / med_control
eta         <- 0.01               # dropout hazard per month (assumption)
acc_months  <- 24                 # accrual duration (assumption)
study_months<- 60                 # total study duration (assumption)

design <- function(k, sfu, beta, h = hr) gsSurv(k = k, test.type = 1, alpha = 0.025, beta = beta, sfu = sfu,
  lambdaC = lambdaC, hr = h, hr0 = 1, eta = eta, gamma = 1, R = acc_months, T = study_months, minfup = study_months - acc_months)

# Main design: 3 looks, Lan-DeMets O'Brien-Fleming spending, one-sided alpha 0.025, 90% power
x <- design(3, sfLDOF, 0.10)
cat(sprintf("N_TOTAL|%s\n", format(ceiling(sum(x$gamma * x$R)), big.mark = ",")))
for (i in 1:3) cat(sprintf("GS|%d|%.1f|%.0f|%.3f|%.5f|%.5f\n", i, x$T[i], x$n.I[i], x$upper$bound[i], pnorm(-x$upper$bound[i]), cumsum(x$upper$spend)[i]))
cat("\n"); print(gsBoundSummary(x))

# Comparison of designs
cmp <- list(list("Fixed design, one analysis", 1, sfLDOF), list("Two looks, O'Brien-Fleming", 2, sfLDOF),
            list("Three looks, O'Brien-Fleming", 3, sfLDOF), list("Three looks, Pocock", 3, sfLDPocock))
cat("\n")
for (b in c(0.20, 0.10)) for (d in cmp) {
  y <- design(d[[2]], d[[3]], b)
  cat(sprintf("DES|%s|%d|%.0f|%s|%.1f\n", d[[1]], round(100 * (1 - b)), tail(y$n.I, 1), format(ceiling(sum(y$gamma * y$R)), big.mark = ","), tail(y$T, 1)))
}

# Sensitivity: what if the true effect is smaller than the observed one? (3 looks, O'Brien-Fleming, 90% power)
cat("\n")
for (fr in c(1, 0.75, 0.5)) {
  y <- design(3, sfLDOF, 0.10, exp(log(hr) * fr))
  cat(sprintf("SENS|%.2f|%.3f|%.0f|%s\n", fr, exp(log(hr) * fr), tail(y$n.I, 1), format(ceiling(sum(y$gamma * y$R)), big.mark = ",")))
}

# Independent cross-check of the boundaries with rpact
r <- getDesignGroupSequential(kMax = 3, alpha = 0.025, beta = 0.10, sided = 1, typeOfDesign = "asOF", informationRates = x$timing)
cat(sprintf("\nCross-check, upper boundaries: gsDesign %s | rpact %s\n", paste(round(x$upper$bound, 3), collapse = ", "), paste(round(r$criticalValues, 3), collapse = ", ")))

Method: LachinFoulkes 
    Analysis               Value Efficacy
   IA 1: 33%                   Z   3.7103
      N: 176         p (1-sided)   0.0001
  Events: 94        ~HR at bound   0.4650
   Month: 14    P(Cross) if HR=1   0.0001
             P(Cross) if HR=0.68   0.0338
   IA 2: 67%                   Z   2.5114
      N: 286         p (1-sided)   0.0060
 Events: 188        ~HR at bound   0.6931
   Month: 23    P(Cross) if HR=1   0.0060
             P(Cross) if HR=0.68   0.5603
       Final                   Z   1.9930
      N: 304         p (1-sided)   0.0231
 Events: 282        ~HR at bound   0.7886
   Month: 60    P(Cross) if HR=1   0.0250
             P(Cross) if HR=0.68   0.9000



Cross-check, upper boundaries: gsDesign 3.71, 2.511, 1.993 | rpact 3.71, 2.511, 1.993

15. Analyses Reported by the Sponsor

The registry reports the sponsor’s Log Rank hazard-ratio analyses for overall survival and the two posted secondary time-to-event outcomes. Those estimates and their confidence intervals are the appropriate basis for describing the sponsor-reported comparisons. The record contains no discrepancy flag for the primary comparison, so there is no stated conflict requiring the sponsor’s analysis to be displaced by a separate recalculation.

The two primary analyses report hazard ratios of 0.678 and 0.621, each with a 95% confidence interval below 1. Their reported p-values are 0.014 and 0.003. The registry does not explain in the summarized record why two primary estimates appear, or specify a distinct analysis population for each; the estimates should not be merged or assigned different meanings without that information.

Any recalculation from posted summary statistics is unadjusted and cannot recreate a full participant-level survival analysis. It cannot account for the precise event and censoring times or verify the model and analysis choices. Where a discrepancy is recorded, the sponsor’s own analysis takes precedence; no such discrepancy is indicated for this primary comparison. The registry record does not provide enough detail to independently audit every aspect of the reported analysis.

16. Multiplicity, Missing Data and Analysis Populations

The registry presents one primary outcome, overall survival, and a set of secondary outcomes that includes time to symptomatic progression, time to progression, disease control, patient-reported scores, tumor response, duration of response, and time to response. It reports results for the primary endpoint and two time-to-event secondary endpoints in the record. The multiplicity approach for the outcome family is not reported in the registry record.

General explanation: When researchers test multiple outcomes or comparisons, the chance of at least one apparently unusual result under a collection of null hypotheses can rise. Prespecified hierarchies, adjusted significance thresholds, or other procedures may address that issue. Without a reported plan, it is not possible to determine whether the secondary p-values were adjusted, whether outcomes were tested in a sequence, or which findings were intended as confirmatory. This uncertainty is one reason to give the primary endpoint greater interpretive weight and to treat the secondary results more cautiously.

The overall-survival description states that participants alive at analysis were censored at their last contact. Its population description also says participants alive or lost to follow-up were censored at their last follow-up or at data cutoff. Further details about missing data, loss to follow-up, sensitivity analyses, or other handling of incomplete outcome information are not reported in the registry record. The record also does not report interim analyses or their timing.

The primary analysis is identified as using the intention-to-treat population. That is useful context, but the summarized record does not provide a complete account of analysis populations for both reported hazard-ratio estimates or every secondary endpoint. It does not report the trial’s original sample-size justification or a detailed statistical analysis plan. These missing details prevent a full assessment of prespecification and robustness from the registry record alone; they do not establish that a particular procedure was omitted.

17. 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: death not associated with ctcae term, disease progression nos (16.7%), death not associated with ctcae term, multi - organ failure (16.7%), neurology - other (16.7%), pneumothorax (16.7%), liver dysfunction (10.7%).

The safety summary distinguishes serious adverse events from other adverse events and lists participant counts at risk and affected. In the sorafenib group labeled “All (Double-Blind and Open Label Phase),” 149 participants were at risk for the serious-event summary and 76 were affected; 144 were affected in the other-adverse-event summary. The placebo group with the same phase label lists 75 at risk, 34 affected by serious adverse events, and 65 affected by other adverse events. A separate placebo open-label-only group, described as participants who switched to sorafenib, lists 6 at risk, 3 affected by serious events, and 4 affected by other events.

These safety groups have different labels and exposure histories, so their counts should not be read as a clean randomized comparison. The registry lists serious-event terms with percentages, including death not associated with a CTCAE term, disease progression NOS; death not associated with a CTCAE term, multi-organ failure; neurology—other; pneumothorax; liver dysfunction; and upper gastrointestinal hemorrhage. The record does not state the denominator or counting basis for these term percentages in the information. Death counts are not reported in the registry record.

General explanation: Adverse-event counts describe events recorded during the specified safety observation, not necessarily events caused by study treatment. These summaries do not establish comparative safety on their own and should be interpreted alongside the study’s definitions, exposure periods, and fuller safety reporting.

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

The registry’s overall-survival results show higher reported median survival in the sorafenib group than in the placebo group, and both reported primary hazard-ratio confidence intervals lie below 1. This is statistical evidence of a group-level difference in the posted time-to-event analyses. It does not mean that every participant in the sorafenib group lived longer than every participant in the placebo group, or that the estimated group comparison predicts an individual outcome.

A hazard ratio describes relative event hazards under the reported analysis, not an absolute increase in survival time or a direct probability of benefit. The medians are useful summaries, but they do not show the full survival curves or explain how the groups may have differed at particular times. The registry does not report enough detail to evaluate all assumptions, follow-up patterns, or analysis choices underlying the results.

The secondary results are not uniform. Time-to-progression comparisons show hazard-ratio intervals below 1, while time-to-symptomatic-progression intervals include 1. These outcomes measure different things and do not each provide the same evidence. Several other secondary outcomes are listed, but their results are not included in the summarized record. Multiplicity procedures are also not reported, making unadjusted secondary findings less conclusive.

These statistical findings do not establish the mechanism behind a group difference, guarantee a particular treatment response, or answer every clinical question about benefits and harms. The safety summaries provide counts for groups with different phase and exposure labels, but do not on their own establish treatment causation or permit a fully adjusted randomized safety comparison. The record’s omissions concerning the original design assumptions, missing-data analyses, and detailed analysis plan further constrain interpretation.

The appropriate reading is therefore specific: the registry reports a primary survival comparison favoring a lower estimated event hazard in the sorafenib group under the stated analyses, with uncertainty represented by the reported intervals. That result is a study-level statistical finding, not medical advice, a certainty about an individual, or a complete assessment of the treatment’s effects.

19. Why This Trial Matters Statistically

This trial illustrates why a clinical time-to-event result should be read through several linked summaries. The medians give survival times in days, while the hazard ratios compare event hazards over follow-up and the confidence intervals convey uncertainty. These statistics answer related but nonidentical questions. Treating a hazard ratio as a ratio of medians, or as an individual probability, would change its meaning.

The endpoint definition and censoring rules matter as much as the headline estimate. Overall survival is measured from randomization to death from any cause, with participants still alive censored at last contact under the registry’s description. Time-to-event methods make use of incomplete follow-up, but depend on assumptions about censoring and follow-up. The registry reports the basic definition without enough detail to assess every assumption.

The record also demonstrates the importance of distinguishing a primary outcome from a family of secondary outcomes. The posted secondary time-to-event results do not all have the same pattern, and the multiplicity approach is not reported. This makes it important not to treat each nominal p-value as an equally definitive conclusion. Clear prespecification and reporting of the analysis plan would help readers understand the intended role of each result.

Finally, the registry contains two reported primary hazard-ratio analyses without explaining their distinction, and its safety summaries include groups with different labels and exposure histories. Transparent descriptions of analysis populations, event handling, and safety denominators are necessary for independent interpretation. The posted findings remain informative, but their meaning is strongest when kept within the limits of the reported design and analysis details.

20. 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 reports two hazard-ratio analyses for overall survival but does not explain why both are presented or identify a distinct analysis population for each in the summarized information. This makes it difficult to determine whether they represent different analysis sets, analysis choices, or another distinction. The results should be reported as separate sponsor analyses rather than reconciled through an unsupported explanation.

The record’s primary-completion date and stated timeframe are not a substitute for detailed follow-up information. The endpoint timeframe refers to a data cutoff approximately 23 months after start of randomization, but the summarized registry record does not provide participant-level follow-up distributions, numbers censored by group, or a detailed account of loss to follow-up. These details would help readers assess the information supporting the survival estimates.

The registry lists an intention-to-treat population for overall survival and describes censoring for participants alive or lost to follow-up, but further missing-data and sensitivity-analysis procedures are not reported. The original sample-size justification, interim-analysis plan, and multiplicity approach are also not reported in the registry record. Consequently, the relationship between the posted calculations and the trial’s original design cannot be established from these entries.

The adverse-event summaries combine double-blind and open-label phases for the principal safety groups and separately identify a small placebo open-label-only group of participants who switched to sorafenib. Those labels and periods complicate direct comparison. The denominator and counting basis for the listed serious-event percentages are not stated in the summarized information, and death counts are not reported in the registry record.

Finally, this interpretation is limited to the outcomes and analysis details recorded. The presence of an outcome title does not establish that its results are available in this record, and the registry summaries cannot substitute for a full report containing protocol, analysis-plan, participant-flow, and detailed safety information.

21. Related Tutorials

22. Related Statistical Calculators

23. Sources

• ClinicalTrials.gov: A Randomized, Double-Blinded, Placebo-Controlled Study of Sorafenib in Patients With Advanced Hepatocellular Carcinoma — NCT00492752.
• Registry record last updated 2014-04-16; status last verified 2014-03.
• Results first posted 2011-01-13.
• 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.

24. Record Summary

The registry describes a completed, randomized, triple-masked, parallel-group phase 3 study of sorafenib versus matching placebo in patients with advanced hepatocellular carcinoma. Enrollment is listed as 226, with Bayer as sponsor. Sorafenib was administered orally twice daily, with permitted dose reductions for treatment-emergent adverse events related to study treatment.

The primary endpoint was overall survival from randomization to death from any cause, analyzed in the intention-to-treat population with censoring described for participants alive or lost to follow-up. The registry reports median overall survival of 198 days for sorafenib and 127 days for placebo, alongside two Log Rank hazard-ratio analyses whose 95% confidence intervals lie below 1.

Secondary time-to-event outcomes show different patterns: the reported time-to-symptomatic-progression intervals include 1, whereas the time-to-progression intervals lie below 1. The registry record does not report the multiplicity approach, original sample-size justification, or detailed missing-data and interim-analysis plans. Safety summaries include groups spanning different treatment phases and should be interpreted with their labels and exposure histories in mind.