Radiation Therapy, Paclitaxel, and Carboplatin With or Without Trastuzumab in Treating Patients With Esophageal Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Radiation Therapy, Paclitaxel, and Carboplatin With or Without Trastuzumab in Treating Patients With Esophageal Cancer (NCT01196390), a randomized phase 3 study. The primary outcome is Disease-free Survival (DFS).

2. Clinical Question

The trial was randomized, open-label, and used a parallel-group design. Its primary endpoint captured several kinds of disease events, including persistence or recurrence in the local or regional area, distant metastases, a new second primary cancer, or death from any cause. This broad definition makes the endpoint a measure of time without a recorded disease event, rather than a measure limited to tumor recurrence.

The clinical question is specifically about the addition of trastuzumab to the chemoradiation regimen. The comparison does not evaluate trastuzumab in isolation or separate its contribution from the other treatments. The registry describes a planned surgery-related response outcome and other secondary measures as well, but the primary statistical comparison is disease-free survival between the randomized groups.

3. Trial Design

This was a randomized phase III trial with parallel groups and no masking. The registry records enrollment of 203. The primary analysis population is described as eligible randomized participants. The listed groups were chemoradiation and trastuzumab, and chemoradiation without trastuzumab. The National Cancer Institute sponsored the trial.

Both groups received radiation therapy with paclitaxel and carboplatin. The described radiation schedule was once daily, 5 days a week for 5.5 weeks. Paclitaxel and carboplatin were administered on days 1, 8, 15, 22, 29, and 36. The trastuzumab group also received trastuzumab on days 1, 8, 15, 22, 29, 36, and 57. The registry lists surgery as an intervention, and the pathologic complete response outcome was assessed at surgery.

The primary outcome was disease-free survival, analyzed using Kaplan-Meier estimates and a log-rank comparison. The registry says the analysis was to occur after 162 events were reported. It does not report the actual event count used in the analysis. The listed maximum follow-up at the time of the primary analysis was 8.0 years. The trial status is completed; the start date and primary completion date are recorded as 2011-02-14 and 2019-12-23.

Participants and clinical teams were not masked, according to the registry. Whether that could have influenced care, outcome assessment, or follow-up cannot be determined from the registry record. Random assignment supports comparison of the groups, but the record does not report details of the randomization procedure or allocation concealment.

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: After registration, sites were required to submit participant tumor tissue for central human epidermal growth factor receptor 2 (HER2) testing. Only HER2-positive participants continued to randomization. In total, 571 participants were registered and 203 were randomized.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Eligible randomized participants

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: Eligible participants

6. Primary Endpoint

The registry lists the primary outcome as Disease-free Survival (DFS), measured over From randomization to last follow-up. Maximum follow-up at time of analysis was 8.0 years..

Disease-free survival (DFS) is the time from randomization to the first defined disease event or the last known follow-up. The registry counted local or regional persistence or recurrence, distant metastases, a new second primary cancer, and death from any cause as events. Participants who had surgery with an R2 resection, or who did not have surgery and had a positive endoscopic biopsy or no biopsy, were considered to have local persistence of disease.

Participants without a recorded event by their last known follow-up were censored at that time. Censoring means that the analysis uses their observed time without treating the last follow-up as if an event had occurred. The registry states that DFS rates were estimated with the Kaplan-Meier method.

General: a time-to-event endpoint uses both whether an event was observed and how long participants were followed before an event or censoring. It can therefore use follow-up information from participants whose observation ends before an event is recorded. DFS is useful when the research question concerns how long participants remain free of a broad set of clinically relevant events. Its interpretation depends on the event definition and on the handling of censoring. The registry record gives the event definition, but does not report the actual event count analyzed or detailed censoring diagnostics.

7. Results Posted in the Registry

• Chemoradiation and Trastuzumab (Arm 1) vs Chemoradiation (Arm 2): hazard ratio 0.97 (95% CI 0.69–1.36), reported p = 0.850 (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.

Percentage of Participants With Pathologic Complete Response at Surgery

• Endpoint type: Binary: the percentage of participants with the event (counts rebuilt from the posted percentages and group sizes).
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Chemoradiation: 29.5% (23/78) versus 26.8% (22/82) on control; risk difference 2.7 percentage points (95% CI -11.3 to 16.6), risk ratio 1.10 (0.67–1.80); unadjusted two-proportion z-test p = 0.709.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Chemoradiation and the control. The interval gives the range of effects that remain compatible with the data.

Overall Survival

• 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:
  • Chemoradiation and Trastuzumab vs Chemoradiation: hazard ratio 1.01 (95% CI 0.69–1.47), reported p = 0.950 (as reported in the registry).
• Interpretation:
  • The 95% confidence interval includes no difference (a ratio of 1), so these data do not show a statistically significant difference between Chemoradiation and Trastuzumab vs Chemoradiation and the control. The interval gives the range of effects that remain compatible with the data.

The registry lists secondary outcomes spanning tumor response at surgery, overall survival, adverse events, quality of life, quality-adjusted survival, molecular correlates, and cardiac events. The posted numerical results described in this record include pathologic complete response and overall survival. The other listed secondary measures are not reported in the registry record in the registry record for this analysis.

Across the two posted secondary comparisons, there is no consistent pattern that clearly reinforces a separation between the randomized groups on the primary endpoint. The pathologic response comparison does not show a clear statistical difference, and the overall-survival comparison is also statistically inconclusive. This broad pattern is compatible with the primary DFS result, but it is not independent confirmation of that result: secondary outcomes address different questions and may have different analysis populations, timing, and sources of uncertainty.

Secondary findings generally carry less confirmatory weight when a registry record does not report how multiplicity was handled. Testing several outcomes increases the opportunity for a chance finding, and a nominal p-value does not account for that unless an appropriate procedure was specified and used. The registry record does not report a multiplicity adjustment or a hierarchy for the secondary outcomes. Therefore, any individual secondary result should be interpreted cautiously and in context rather than treated as a separate definitive conclusion.

The listed secondary domains may provide complementary information about response, survival, safety, and participant experience, but the results for those domains cannot be inferred from the primary endpoint. Their absence from the reported results means their direction, precision, and relationship to DFS are not reported in the registry record.

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

• Percentage of Participants With Pathologic Complete Response at Surgery
• Overall Survival
• Frequency of Highest Grade Adverse Event Per Participant
• Percentage of Patients With Improvement in the Functional Assessment of Cancer Therapy - Esophagus (FACT-E) Esophageal Cancer Subscale (ECS) Subscale After Treatment
• Quality-adjusted Survival
• Molecular Correlates of Efficacy
• Number of Participants With Any Cardiac Adverse Events Regardless of Attribution

9. Statistical Interpretation

10. Statistical Methods Explained

The primary endpoint is time-to-event data. Each participant contributes a duration from randomization until a recorded disease event or last known follow-up. Participants with no event at their last follow-up are censored. General: censoring allows the analysis to retain partial follow-up rather than treating every participant without an observed event as if they had an event at the end of observation. It relies on assumptions about how censoring relates to the event process; the registry record does not report diagnostic checks of those assumptions.

How is the median estimated?

The registry says DFS rates were estimated using the Kaplan-Meier method. General: the Kaplan-Meier estimator updates the estimated proportion remaining event-free as events occur, while accounting for participants who are censored. The median is the time at which the estimated event-free curve reaches its midpoint. It is not the arithmetic mean, and it does not imply that every participant had follow-up until that time. The median estimates reported are 19.6 months and 14.2 months for the two groups, each with its own 95% confidence interval.

What does the hazard ratio compare?

The primary comparison uses a hazard ratio as reported by the sponsor, estimated using a log-rank method. General: a hazard ratio describes the relative rate of events over time between groups, rather than a simple ratio of the median times or a comparison of the percentage with an event at one fixed time. A hazard ratio of 1 represents equal hazards. The reported estimate is 0.97, and its 95% confidence interval is 0.69 to 1.36. Since the interval includes 1, it does not exclude equal hazards at the confidence level represented by that interval.

The log-rank test compares the groups' event-time experience across follow-up. Its p-value is 0.85. General: a p-value describes how incompatible the observed data are with a specified null hypothesis under the test's assumptions; it is not the probability that the null hypothesis is true. A large p-value does not establish no effect, and a small p-value would not by itself measure clinical importance. The registry reports the test and result but does not give further detail about model checks or additional covariate adjustment.

How should the secondary analyses be understood?

Pathologic complete response is a binary outcome reported as a percentage of participants assessed at surgery, with 95% confidence intervals. The sponsor reports a chi-squared test with p=0.71. General: a chi-squared test assesses whether observed categorical outcome counts are compatible with equal proportions under its null hypothesis. The registry also reports a risk-difference re-calculation, but the sponsor's reported result is the primary source for the sponsor analysis. The two groups' reported percentages and confidence intervals are not a substitute for the test or for a prespecified adjusted analysis.

Overall survival is another time-to-event endpoint. The registry reports medians, 95% confidence intervals, a hazard ratio, and a log-rank p-value. General: interpreting overall survival follows similar principles to DFS, but the endpoint is death rather than the broader DFS event definition. The time origin is randomization, and the registry describes follow-up through last follow-up.

The registry record does not report the trial's original sample-size justification, detailed statistical analysis plan, handling of missing data for the secondary outcomes, or multiplicity procedure. It records that the planned DFS analysis was to occur after 162 reported events, but does not state the event count at the analysis. These omissions limit assessment of whether all reported analyses followed a prespecified plan.

What does over-dispersion mean in general?

General: over-dispersion occurs when outcome variability is greater than a statistical model expects. It can make standard errors too small if the model does not account for the extra variability, leading to confidence intervals that are too narrow and tests that are too optimistic. The posted primary analysis is a time-to-event comparison, and no over-dispersion assessment is reported. This general concept does not establish that over-dispersion occurred in this trial.

11. Confidence Intervals and Precision

The 95% confidence intervals communicate uncertainty around the estimates. For DFS, the median estimate in the trastuzumab group is 19.6 months, with an interval from 13.5 to 26.2 months; the chemoradiation group's median is 14.2 months, with an interval from 10.5 to 23.0 months. These intervals describe uncertainty in each group's median estimate, not a direct interval for the difference between medians.

The more direct comparative estimate is the reported hazard ratio of 0.97, with a 95% confidence interval from 0.69 to 1.36. The interval includes 1, the value corresponding to equal hazards. It also spans values on either side of 1, so the data do not pin down the direction of the relative hazard with high precision. The p-value of 0.85 is consistent with the absence of a clear difference in the reported test, but it cannot show that the effects are exactly equal.

For overall survival, the reported hazard ratio is 1.01, with a 95% confidence interval from 0.69 to 1.47. This interval also includes 1 and permits a range of effects in either direction. The medians have their own intervals, which overlap, but overlap between separate group intervals is not itself a formal test of the comparative effect. The reported hazard ratio and log-rank result are the appropriate summaries of the sponsor's comparison.

General: a confidence interval is a range of values compatible with the data and the statistical method, under its assumptions. It is not a guarantee that the true value lies in the interval, nor does it indicate the probability that one treatment is better. The width of an interval reflects precision: wider intervals leave more uncertainty. Here, the intervals permit both possible directions of relative effect, so the posted results should not be interpreted as demonstrating equivalence. The registry record does not report a prespecified equivalence margin or a noninferiority analysis.

12. Planning a Similar Trial: Sample Size

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

The planning figures are illustrations based on the observed effect, not the trial's original sample-size calculation. For time-to-event outcomes, the figures are total events required; for the binary pathologic complete response outcome, they are participants per arm. The displayed primary DFS planning values are total-event targets, while the pathologic complete response values are per-arm participant targets. They are not estimates of the trial's actual enrollment or event count.

The assumptions for these illustrations are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. Each row considers detecting a stated fraction of the observed effect. The 80% power column describes the event or participant target that, under those assumptions and a specified effect, would give an 80% chance of rejecting the null hypothesis if that effect were true. The 90% power column describes the corresponding target for a 90% chance.

Smaller effects require more information to distinguish them from no difference. Accordingly, the planning rows for smaller fractions of the observed effect show larger required event or participant counts than the rows for the full observed effect. These figures are planning illustrations, not observed power. They do not say how likely the trial was to detect an effect after the observed data were collected, and they should not be read as the original protocol's design targets. The registry record does not report the original sample-size justification.

For the time-to-event planning illustration, read each target as the total number of events rather than a number of participants in each group. For the pathologic complete response illustration, read each target as participants in each arm. The distinction matters because participants can be followed for different periods and may not all experience an event, whereas the binary outcome's planning units are participants allocated to each group.

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 = 0.970, so ln(HR) = -0.0305.
2. d = 4 × (1.96 + 0.842)² / [ln(0.970)]² = 33,840.0, rounded up to 33,841 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 <- 0.97                              # 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(1196390)
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 = 33,841 (80% power), 45,303 (90% power)
 75% of observed effect: events = 60,161 (80% power), 80,538 (90% power)
 50% of observed effect: events = 135,361 (80% power), 181,209 (90% power)
Simulated power with 33,841 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 59,327 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 203 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 <- 14.2          # control-arm median, months (converted from the posted unit)
hr          <- 0.97          # 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: 46868         p (1-sided)   0.0001
 Events: 15279        ~HR at bound   0.9417
     Month: 19    P(Cross) if HR=1   0.0001
               P(Cross) if HR=0.97   0.0338
     IA 2: 67%                   Z   2.5114
      N: 59328         p (1-sided)   0.0060
 Events: 30558        ~HR at bound   0.9717
     Month: 30    P(Cross) if HR=1   0.0060
               P(Cross) if HR=0.97   0.5603
         Final                   Z   1.9930
      N: 59328         p (1-sided)   0.0231
 Events: 45836        ~HR at bound   0.9816
     Month: 60    P(Cross) if HR=1   0.0250
               P(Cross) if HR=0.97   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 sponsor's primary analysis reports a hazard ratio of 0.97, a 95% confidence interval of 0.69 to 1.36, and a log-rank p-value of 0.85. The recalculated summary agrees with the sponsor's reported primary estimate and test; the registry record does not flag a discrepancy. The sponsor's analysis therefore remains the basis for describing the primary comparison.

For overall survival, the sponsor-reported hazard ratio is 1.01, with a 95% confidence interval of 0.69 to 1.47 and a log-rank p-value of 0.95. For pathologic complete response, the sponsor reports a chi-squared p-value of 0.71. The unadjusted recalculation is consistent with that reported p-value after rounding. These results are not evidence that the trial's sponsor analysis should be replaced by a recalculation.

General: recalculations from posted summary statistics can help readers check the scale or consistency of a result, but they may not reproduce the sponsor's participant-level analysis or its handling of time-to-event data. In this record, the available comparisons do not show a discrepancy requiring a different reading. No additional adjusted analysis is reported in the registry record.

16. Multiplicity, Missing Data and Analysis Populations

The registry record does not report whether the analyses used a multiplicity adjustment, a hierarchical testing strategy, or another prespecified approach to control the chance of false-positive findings across outcomes. This matters because the trial lists multiple secondary outcomes, and testing many questions can produce apparently notable results by chance even when no corresponding effect exists. Without a reported plan, secondary p-values should not be treated as if each were the sole confirmatory test.

The record describes the primary analysis population as eligible randomized participants. It does not report further details about analysis populations, such as whether the primary analysis followed an intention-to-treat principle or how participants who became ineligible after randomization were handled. The record also does not report the number of participants included in each primary analysis beyond the group counts attached to the outcome results.

For DFS, the endpoint definition says that participants without a recorded event were censored at last known follow-up. That describes the endpoint's handling of incomplete event observation, but the registry record does not report a broader missing-data strategy, the amount or pattern of missing follow-up, or sensitivity analyses for informative censoring. Missing outcome data for the secondary endpoints and their analysis procedures are not reported in the registry record.

The planned timing of the primary analysis is described as after 162 events were reported, but the actual event count at analysis is not reported. Interim analyses, stopping rules, and any changes to the statistical plan are not reported in the registry record. These gaps do not invalidate the reported estimates, but they limit the ability to assess prespecification, completeness, and the robustness of the analyses from registry information alone.

17. 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: dehydration (10.4%), sepsis (6.3%), nausea (6.2%), hypotension (5.3%), white blood cell decreased (5.2%).

The registry lists serious adverse event figures for participants at risk and participants affected in each group. In the chemoradiation-and-trastuzumab group, 95 participants were listed as at risk and 51 as affected; in the chemoradiation group, 96 were listed as at risk and 34 as affected. These are group-level counts, not a comparison adjusted for follow-up duration or participant characteristics. No comparative safety test is reported in the registry record.

The listed serious adverse event terms include dehydration at 10.4%, sepsis at 6.3%, nausea at 6.2%, hypotension at 5.3%, white blood cell decreased at 5.2%, and neutrophil count decreased at 4.2%. The registry also records deaths as 53 and 55 in the respective groups. The relationship between those death counts and the adverse-event categories, and the specific time window for these counts, are not reported in the registry record.

General: serious adverse event counts describe reported events or affected participants as defined by the registry, but they do not by themselves establish that a treatment caused an event. Differences in exposure time, ascertainment, and event attribution can affect comparisons. The registry record does not provide enough detail to make a causal safety comparison.

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

The posted primary result is an estimated DFS hazard ratio close to 1, with a confidence interval that includes values below and above 1 and a log-rank p-value of 0.85. This means the analysis does not provide clear statistical evidence of a difference between the groups on the primary endpoint. It does not establish that the regimens are equivalent, nor does it prove that trastuzumab has no effect.

The estimate is uncertain: the confidence interval remains compatible with relative hazards in either direction. The median DFS estimates differ numerically, but medians alone do not determine the comparative treatment effect. The hazard ratio and its interval, together with the endpoint definition and follow-up, are more informative for the reported time-to-event comparison.

The secondary results do not materially resolve that uncertainty. The overall-survival comparison is inconclusive, and the pathologic complete response test does not show a clear difference. Other listed secondary measures do not have results reported in the registry record. The lack of a reported multiplicity procedure further limits the confirmatory interpretation of secondary analyses.

General: a statistically inconclusive result is not the same as proof of no effect. A trial can fail to distinguish a treatment effect from no difference when the estimate is imprecise, even if the true effect is not exactly zero. Conversely, a numerical difference in a group estimate does not by itself establish a treatment effect. This analysis describes the registry's posted results and their precision; it does not provide medical advice or determine which regimen should be used.

19. Why This Trial Matters Statistically

This trial illustrates why a treatment comparison should be read through its endpoint definition, analysis method, and uncertainty rather than through a single median or p-value. DFS includes persistence or recurrence, distant metastases, a new second primary cancer, and death, so it captures a broader event experience than recurrence alone. The clinical meaning of the result depends on that definition and on how follow-up and censoring were handled.

It also shows the distinction between an estimate and a conclusion. The reported hazard ratio is near the reference value for equal hazards, but its confidence interval extends to both sides of that value. The p-value does not establish equivalence, and the median estimates do not replace the comparative time-to-event analysis. Statistical interpretation must account for the uncertainty in both the relative effect and the group-specific medians.

The trial's secondary outcomes span survival, response, safety, and participant-centered measures. That breadth can be valuable, but it also makes prespecification and multiplicity important. When those methods are not reported, secondary results should be interpreted with restraint and should not be elevated to confirmatory evidence on their own.

Finally, registry records can leave important questions unanswered even when they report headline estimates. Details of the original sample-size justification, analysis populations, missing-data strategy, multiplicity, and interim monitoring are not reported. Recognizing these limits is part of sound biostatistical interpretation, not a reason to substitute speculation for the posted evidence.

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 record does not report the original sample-size justification, so the planning illustrations based on observed effects should not be mistaken for the trial's design assumptions. Although the primary analysis was planned after 162 events, the actual event count at analysis is not reported. Detailed information on interim analyses, stopping rules, and changes to the analysis plan is also not reported.

The primary analysis population is described as eligible randomized participants, but the record does not provide enough detail to determine whether all randomized participants were analyzed under an intention-to-treat approach or how exclusions were handled. The primary outcome counts are reported by group, but additional participant flow and missingness details are not available in the registry record. The record describes censoring for DFS but does not report sensitivity analyses or checks of censoring assumptions.

The study was open-label, with no masking. The registry record does not describe how potential knowledge of assignment was addressed in outcome assessment or follow-up. The primary endpoint includes several event types, and the record does not report event-specific results that would show which components contributed most to the overall DFS result.

The secondary outcomes include several domains, but a multiplicity plan is not reported. Results for several listed secondary measures are not reported in the registry record, limiting a broader assessment of survival, quality of life, quality-adjusted survival, molecular correlates, cardiac events, and other safety outcomes. The available summary also does not report detailed adverse-event attribution, exposure-adjusted comparisons, or the time window attached to the recorded deaths.

These limitations constrain interpretation of precision, completeness, and reproducibility from registry information alone. They do not permit assumptions about unreported methods or outcomes.

21. Related Tutorials

22. Related Statistical Calculators

23. Sources

• ClinicalTrials.gov: Radiation Therapy, Paclitaxel, and Carboplatin With or Without Trastuzumab in Treating Patients With Esophageal Cancer — NCT01196390.
• Registry record last updated 2026-06-02; status last verified 2026-01.
• Results first posted 2021-05-04.
• 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

NCT01196390 is a completed randomized phase III, open-label, parallel-group trial sponsored by the National Cancer Institute. It enrolled 203 and compared chemoradiation with trastuzumab against chemoradiation without trastuzumab in eligible participants with HER2-overexpressing esophageal adenocarcinoma.

The primary outcome was DFS, defined as time from randomization to a disease event or last known follow-up. The reported hazard ratio was 0.97, with a 95% confidence interval of 0.69 to 1.36 and a log-rank p-value of 0.85. The posted secondary results include overall survival and pathologic complete response; other listed secondary results are not reported in the registry record.

The record does not report the original sample-size justification, multiplicity procedure, detailed missing-data strategy, or actual event count at primary analysis. The estimates are therefore best read with attention to their uncertainty and to the limits of the available registry information.