A Study to Investigate Tislelizumab (BGB-A317) Versus Placebo in Combination With Concurrent Chemoradiotherapy in Participants With Localized Esophageal Squamous Cell Carcinoma: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study to Investigate Tislelizumab (BGB-A317) Versus Placebo in Combination With Concurrent Chemoradiotherapy in Participants With Localized Esophageal Squamous Cell Carcinoma (NCT03957590), a randomized phase 3 study. The primary outcome is Progression-free Survival (PFS).

2. Clinical Question

3. Trial Design

The registry record describes a phase 3, randomized, double-blind, placebo-controlled, parallel-group study. Masking is listed as quadruple. The trial compared tislelizumab plus concurrent chemoradiotherapy with placebo plus concurrent chemoradiotherapy in participants with localized esophageal squamous cell carcinoma. Enrollment is recorded as 370.

The intervention descriptions specify intravenous tislelizumab or placebo once every 3 weeks alongside chemoradiotherapy for up to approximately 6 weeks, followed by continued study drug for up to approximately 24 months, described as about 35 cycles, or until specified treatment-ending events. The record lists paclitaxel, cisplatin, and radiotherapy among the interventions. It does not give a complete account of the chemoradiotherapy schedule in the arm descriptions.

The study start date is 2019-06-11. The status is completed, and the primary completion date is 2025-01-08. The primary analysis timeframe runs from randomization to the prespecified primary analysis data cut-off date of 08 January 2025; maximum time on study was 67 months.

The primary analysis set is the intention-to-treat population, defined as all randomized participants. The outcome summaries list 185 participants for each arm. The registry record does not report the full randomization procedure, the original sample-size justification, or the detailed statistical analysis plan. Those features should not be inferred from the enrollment or the observed arm sizes.

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: Eligible participants were randomized in a 1:1 ratio to receive either tislelizumab or placebo in combination with concurrent chemoradiotherapy (cCRT).

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: The Intention-to-Treat (ITT) analysis set includes all 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.

6. Primary Endpoint

The registry lists the primary outcome as Progression-free Survival (PFS), measured over From randomization to the prespecified primary analysis data cut-off date of 08 January 2025; maximum time on study was 67 months..

The primary endpoint is progression-free survival, a time-to-event measure. It counts time from randomization until the first documented disease progression or death from any cause, whichever occurs first. Disease progression was determined by a blinded independent review committee using RECIST v1.1. The registry description defines progression using changes in target lesions and also includes unequivocal progression of existing nontarget lesions or the appearance of new lesions.

Progression-free survival is used in trials to capture whether participants remain alive without documented disease progression over follow-up. It combines progression and death into one endpoint, so it is not the same as overall survival, which concerns time to death. The registry record reports progression-free survival estimates using the Kaplan-Meier method and gives a median for each group, along with a hazard-ratio comparison.

General statistical context: a time-to-event endpoint can use information from participants with different follow-up durations, including those whose event has not occurred by the analysis cut-off. Such observations are censored rather than treated as if an event occurred. Interpretation depends on the event definition, assessment schedule, follow-up, and assumptions underlying the analysis. The registry record does not provide detailed information about censoring rules beyond the endpoint definition and timeframe.

7. Results Posted in the Registry

• Tislelizumab + Chemoradiotherapy vs Placebo + Chemoradiotherapy: hazard ratio 0.92 (95% CI 0.68–1.25), reported p = 0.302 (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.

Overall Survival (OS)

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor; it was not recalculated.
• Result:
  • Tislelizumab + Chemoradiotherapy vs Placebo + Chemoradiotherapy: hazard ratio 1.10 (95% CI 0.83–1.46) (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 Tislelizumab + Chemoradiotherapy vs Placebo + Chemoradiotherapy and the control. The interval gives the range of effects that remain compatible with the data.

Overall Response Rate (ORR)

• 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:
  • Tislelizumab + Chemoradiotherapy: 27.6% (51/185) versus 38.9% (72/185) on control; risk difference -11.4 percentage points (95% CI -20.9 to -1.8), risk ratio 0.71 (0.53–0.95); unadjusted two-proportion z-test p = 0.020.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically lower for Tislelizumab + Chemoradiotherapy than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The secondary outcomes cover survival, tumor response, duration of response, quality of life, and treatment-emergent adverse events. The outcomes therefore address several different aspects of the trial, rather than providing multiple interchangeable measures of one effect. Some secondary estimates are posted, while numerical results for the quality-of-life scales and duration of response are not reported in the registry record.

The overall pattern does not give a uniform signal across the posted secondary outcomes. The survival comparison has uncertainty spanning either direction, while the response-rate comparison has a recalculated unadjusted result that favors the placebo group on the displayed response proportion. These secondary findings do not overturn the primary endpoint result. They add context, but the primary progression-free survival comparison remains the principal reported efficacy analysis.

General statistical context: when many outcomes or comparisons are examined, the chance of observing apparently notable results by chance can increase. Adjustment procedures or a prespecified testing hierarchy can address some multiplicity concerns, but the registry record does not report whether such procedures were used for the secondary outcomes. An unadjusted p-value should therefore be interpreted cautiously, particularly when several outcomes are considered together.

The registry record lists overall survival, response rate, duration of response, quality-of-life changes, and treatment-emergent adverse events among secondary outcomes. It does not report the complete analysis plan, the prespecified priority among secondary endpoints, or the numerical results for all listed outcomes. The secondary findings should be considered supportive and exploratory in context rather than as independent confirmations of the primary result.

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

• Overall Survival (OS)
• Change From Baseline in European Organization for Research and Treatment of Cancer (EORTC) Quality of Life Questionnaire-Oesophageal Cancer Module (OES18) Dysphagia, Reflux, Pain, and Eating Scales
• Change From Baseline in EORTC Quality of Life Questionnaire-Core 30 (QLQ-C30) Global Health Status (GHS)/Quality of Life (QOL), Physical Functioning, and Fatigue Scores
• Overall Response Rate (ORR)
• Duration of Response (DOR)
• Number of Participants With Treatment-emergent Adverse Events (TEAEs)

9. Statistical Interpretation

10. Statistical Methods Explained

Progression-free survival and overall survival are time-to-event outcomes. Their summaries differ from a simple proportion because each participant’s follow-up time matters, and some participants may not have had the event by the analysis cut-off. The registry record reports medians estimated with the Kaplan-Meier method and comparisons using hazard ratios. General statistical context: Kaplan-Meier estimation uses event times and accommodates censored observations, meaning follow-up ended before an event was observed. The method relies on assumptions about censoring and the way follow-up and events are recorded.

What does a median time-to-event estimate describe?

A median is the estimated time by which half of the relevant event-time distribution has experienced the event. It is one summary of the curve, not a complete description of the curve. Two groups can have similar medians while their curves differ at other times, or can have different medians while uncertainty around the estimates remains substantial. The registry record reports median progression-free survival and overall survival, but the upper confidence limits for those medians are not displayed in the posted results.

How should a hazard ratio be read?

A hazard ratio compares event rates between groups over follow-up. A value below the no-difference reference is conventionally read as a lower estimated event hazard for the first group; a value above it is read as a higher estimated event hazard. This is a relative, time-to-event summary, not a direct comparison of the number of events or a statement about the probability that an individual participant will experience an event. Its interpretation is clearest when the relative hazards are reasonably stable over time. The registry record does not report whether that assumption was assessed or whether additional time-varying analyses were conducted.

The primary reported hazard ratio is 0.92, with a 95% confidence interval from 0.68 to 1.25. The interval includes the no-difference reference and covers estimates in both directions. The log-rank p-value is 0.302. General statistical context: the log-rank test compares event-time distributions between groups while accounting for follow-up and censoring. A p-value measures how compatible the observed test statistic is with a specified null model; it is not the probability that the null hypothesis is true, nor does it measure the size or clinical importance of an effect.

What does a confidence interval add?

A confidence interval conveys uncertainty around an estimate under the analysis method and its assumptions. It identifies a range of estimates compatible with the observed data at the stated confidence level. It does not mean that the true effect has a particular probability of lying inside this one calculated interval. Wide intervals indicate less precision than narrow intervals, all else being equal. The primary and secondary hazard-ratio intervals should therefore be considered with their point estimates rather than read as definitive proof of either benefit or harm.

What is the role of the endpoint review process?

The registry definition states that progression was determined by a blinded independent review committee using RECIST v1.1. Blinded assessment is intended to limit the influence of knowledge of treatment assignment on judgments about disease progression. The record gives the assessment framework and event definition, but does not report further operational details, such as adjudication procedures for disagreements or the completeness of scheduled assessments.

The registry record identifies the intention-to-treat set as the primary analysis population. This population includes all randomized participants and preserves the randomized comparison in the analysis definition. The record does not report detailed handling of missing endpoint assessments, deviations from treatment, or other analysis populations. Those details matter because missing or unequally observed outcomes can affect time-to-event estimates and comparisons.

How should the response-rate comparison be understood?

Overall response rate is a binary outcome: participants are classified according to whether they met the response definition during the assessment period. The posted group estimates are proportions with confidence intervals. The recalculated risk difference compares those proportions by subtraction, while the unadjusted p-value tests a simple group comparison. General statistical context: a risk difference is expressed on an absolute scale, whereas a risk ratio or odds ratio is relative. Each scale answers a different question, and none alone establishes clinical importance. The registry record does not provide a sponsor-reported analysis for this response-rate comparison.

The secondary response-rate recalculation is explicitly unadjusted. It does not account for multiplicity, covariates, or any other analysis features not included in the posted summary-statistic calculation. The registry record does not state whether a different prespecified analysis was planned. For the primary time-to-event outcome, the sponsor-reported hazard ratio and test are the relevant reported analysis; the registry record indicates no discrepancy between that analysis and the displayed primary result.

11. Confidence Intervals and Precision

The primary hazard-ratio estimate is 0.92, with a 95% confidence interval from 0.68 to 1.25. The interval is broad enough to include hazard ratios on either side of the no-difference reference. This means the point estimate should not be interpreted as a precise estimate of the direction or size of a treatment effect. The log-rank p-value of 0.302 is consistent with the interval’s lack of a statistically clear separation from the no-difference reference.

The reported medians are 29.0 months and 28.9 months. The primary outcome is described as having a 95% confidence interval, but only the lower limits, 18.8 and 18.0 months, are present in the posted results; upper limits are not reported in the registry record. Those lower limits alone do not convey the full uncertainty around each median.

For overall survival, the reported hazard ratio is 1.10, with a 95% confidence interval from 0.83 to 1.46. That interval also covers both directions of a possible difference. A p-value is not reported for this comparison in the registry record. The posted medians are 39.2 months and 48.2 months, with lower confidence limits of 28.3 and 30.2 months; upper limits are not reported in the registry record.

General statistical context: precision reflects how narrowly an estimate is determined by the data and analysis. It is distinct from whether an estimate is statistically significant or clinically important. Confidence intervals also rely on the analysis assumptions and do not account automatically for biases, missing data, measurement limitations, or multiplicity. These results are best read as estimates with substantial uncertainty about the comparative time-to-event effect.

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 design assumptions. For the time-to-event outcomes, the displayed figures are total events, not participants per arm. For the overall response-rate comparison, the figures are participants per arm. The planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation; each illustration uses the stated fraction of the observed effect. The registry record does not report the trial’s own original sample-size assumptions or justification.

The 80% power column describes the number of events or participants the illustration estimates would be needed to have an 80% chance of detecting the chosen effect under those assumptions. The 90% power column uses a higher target chance of detecting that effect and therefore generally requires more information. These are planning quantities, not a claim about what the completed trial achieved.

Each row represents planning for a specified fraction of the observed effect. Planning for a smaller effect requires more information to distinguish it from random variation, so the participant or event requirement rises as the assumed effect becomes smaller. Time-to-event planning values should not be read as enrollment targets, because they are event counts; the response-rate values are per-arm participant counts.

This is not observed power. It is a prospective-style illustration using the observed effect as an input, not a measure of the probability that the completed trial detected an effect.

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 = 0.920, so ln(HR) = -0.0834.
2. d = 4 × (1.96 + 0.842)² / [ln(0.920)]² = 4,515.7, rounded up to 4,516 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 <- 0.92                              # 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(3957590)
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 = 4,516 (80% power), 6,046 (90% power)
 75% of observed effect: events = 8,028 (80% power), 10,748 (90% power)
 50% of observed effect: events = 18,063 (80% power), 24,182 (90% power)
Simulated power with 4,516 events: 0.804 (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 11,115 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 370 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 <- 28.9          # control-arm median, months (converted from the posted unit)
hr          <- 0.92          # 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: 10148         p (1-sided)   0.0001
 Events: 2039        ~HR at bound   0.8484
    Month: 22    P(Cross) if HR=1   0.0001
              P(Cross) if HR=0.92   0.0338
    IA 2: 67%                   Z   2.5114
     N: 11116         p (1-sided)   0.0060
 Events: 4078        ~HR at bound   0.9244
    Month: 35    P(Cross) if HR=1   0.0060
              P(Cross) if HR=0.92   0.5603
        Final                   Z   1.9930
     N: 11116         p (1-sided)   0.0231
 Events: 6117        ~HR at bound   0.9503
    Month: 60    P(Cross) if HR=1   0.0250
              P(Cross) if HR=0.92   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 record reports the primary comparison as a log-rank hazard ratio of 0.92, with a 95% confidence interval from 0.68 to 1.25 and p = 0.302. These are the sponsor’s reported primary analysis results. The record indicates no discrepancy between the posted primary analysis and the displayed result, so the sponsor-reported analysis is the appropriate basis for describing the primary finding.

The survival-outcome meaning specifies that only sponsor-reported hazard ratios are used. The displayed medians are descriptive summaries and should not be substituted for the sponsor’s comparative analysis. The confidence interval includes estimates in either direction, and the reported p-value does not indicate a statistically clear difference under the log-rank test.

The overall response-rate comparison is different: no sponsor-reported analysis is listed for that outcome. Its risk difference and unadjusted p-value are recalculations from the posted summary statistics. Such a recalculation is a simple, unadjusted comparison and may not match an analysis that accounts for prespecified methods, covariates, or other design features. The registry record does not report a sponsor analysis for that outcome, so the recalculation should be treated as supplementary rather than as the trial’s official conclusion.

16. Multiplicity, Missing Data and Analysis Populations

The registry record lists several secondary outcomes, including overall survival, quality-of-life measures, overall response rate, duration of response, and treatment-emergent adverse events. Examining multiple endpoints creates a multiplicity issue: even when there is no underlying treatment difference, one or more comparisons may appear notable by chance. General statistical context: prespecified testing sequences, family-wise error control, or other adjustment methods can help manage this issue.

Whether multiplicity adjustments or a testing hierarchy were used is not reported in the registry record. The unadjusted p-value for the recalculated response-rate comparison therefore should not be treated as if it had been adjusted for all secondary outcomes. The registry record also does not report whether the secondary analyses were confirmatory or exploratory.

The primary analysis population is reported as the intention-to-treat set, including all randomized participants. The outcome summaries list 185 participants in each treatment arm. The safety summaries list 185 participants at risk in the tislelizumab group and 184 in the placebo group for the displayed safety measures. The registry record does not explain the difference between the outcome and safety denominators.

Detailed missing-data procedures are not reported in the registry record. For time-to-event analyses, censoring is relevant when an event is not observed before follow-up ends, but the record does not provide the complete censoring rules or details of handling missing assessments. It also does not report the extent or pattern of missing data for the quality-of-life outcomes or the analysis populations used for each secondary endpoint. These omissions limit assessment of how robust the results are to incomplete follow-up or outcome data.

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 3% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: pneumonia (10.8%), immune-mediated lung disease (7.6%), platelet count decreased (3.2%), oesophageal fistula (2.7%), oesophageal obstruction (2.7%).

The safety summaries list serious adverse events affecting 84 participants in the tislelizumab plus chemoradiotherapy group and 63 participants in the placebo plus chemoradiotherapy group. The displayed numbers at risk are 185 and 184, respectively. Deaths are listed as 100 and 93, respectively. These are descriptive counts; they do not by themselves establish that treatment caused an event or explain the circumstances of a death.

The registry record identifies pneumonia at 10.8% and immune-mediated lung disease at 7.6% among the leading serious adverse-event terms. Other listed terms are platelet count decreased at 3.2%, oesophageal fistula at 2.7%, oesophageal obstruction at 2.7%, and radiation oesophagitis at 2.7%. These terms describe reported serious events and should be interpreted alongside the trial population, treatment exposure, and event definitions.

The record does not provide the full safety-event listing, event timing, severity detail, causal assessments, or statistical comparisons. The counts and percentages are not a substitute for a complete safety analysis, and differences between descriptive counts should not be interpreted as proof of a causal effect.

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

The primary result describes progression-free survival among participants in the intention-to-treat population. The median estimates are close, and the reported hazard-ratio confidence interval includes values compatible with either a lower or higher event hazard in the tislelizumab group. The log-rank p-value is 0.302. Taken together, the posted results do not establish a statistically clear difference in progression-free survival between the groups.

This interpretation does not prove that the treatments are equivalent. A result that does not show a statistically clear difference can reflect an effect that is small, an imprecisely estimated effect, or no meaningful difference; the confidence interval shows that uncertainty remains. The registry record does not report a noninferiority or equivalence analysis, so the results should not be described as demonstrating equivalence.

The secondary outcomes provide additional observations but do not change the status of the primary comparison. The overall survival estimate is uncertain, and the response-rate comparison is an unadjusted recalculation rather than a reported sponsor analysis. Numerical findings for some listed secondary outcomes are not reported in the registry record. Safety counts describe reported events but do not establish causation.

General statistical context: statistical evidence and clinical importance are related but distinct. A p-value is not a measure of the probability that treatment is beneficial, and a confidence interval does not directly establish whether an effect matters to patients. Clinical interpretation also depends on the endpoint, treatment burden, adverse events, follow-up, and patient priorities. The registry record does not provide enough information to make those broader judgments on its own.

These findings apply to the study population and treatment conditions recorded for this trial. They should not be generalized automatically to different disease settings, treatment regimens, or patient groups. No medical advice follows from the statistical results alone.

19. Why This Trial Matters Statistically

This trial illustrates how a randomized, placebo-controlled comparison can be analyzed with a time-to-event primary endpoint. Randomization provides the basis for comparing treatment groups, while the intention-to-treat definition preserves participants in the groups to which they were randomized for the primary analysis. The blinded independent review committee and RECIST v1.1 progression criteria make the event definition explicit in the registry record.

The results also demonstrate why a single point estimate is not enough. The primary hazard ratio is accompanied by a confidence interval that includes estimates in both directions, and the log-rank p-value does not indicate a statistically clear separation. The medians alone cannot convey the full shape of event-time distributions or the uncertainty in the comparison.

The trial further highlights the distinction between a sponsor-reported analysis and an unadjusted recalculation. The reported primary hazard ratio is the main comparative result. The response-rate recalculation can help describe the posted proportions, but it does not replace a prespecified sponsor analysis and is not adjusted for multiplicity.

Finally, the registry record illustrates the importance of complete reporting. Some outcomes are listed without numerical results, and detailed information about missing data, multiplicity, and analysis assumptions is not reported in the registry record. Such gaps affect how fully researchers can evaluate precision, robustness, and the relationship between primary and secondary findings.

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 posted record does not include the full statistical analysis plan, so the prespecified handling of multiplicity, missing data, censoring, and secondary analyses is not reported in the registry record. The original sample-size justification and design assumptions are also not reported in the registry record. Planning illustrations based on observed effects do not fill that gap.

The upper confidence limits for the posted median progression-free survival and overall survival estimates are absent. The record reports 95% confidence intervals for these endpoints but displays only lower limits for the medians. The sponsor-reported hazard-ratio intervals are available, but they do not provide every detail needed to assess model assumptions or the stability of hazards over follow-up.

Numerical results for some listed secondary outcomes, including duration of response and quality-of-life measures, are not reported in the registry record. The overall response-rate comparison has no sponsor-reported analysis listed, and its displayed comparative statistics are unadjusted recalculations. The extent and pattern of missing outcome data are not reported in the registry record.

The safety summaries are descriptive and do not provide full event-level detail, exposure-adjusted rates, or causal assessments. The registry record also does not explain why the safety denominators differ from the outcome-analysis group sizes. These limitations constrain detailed comparisons of safety and efficacy and call for restraint when interpreting the posted findings.

21. Related Tutorials

22. Related Statistical Calculators

23. Sources

• ClinicalTrials.gov: A Study to Investigate Tislelizumab (BGB-A317) Versus Placebo in Combination With Concurrent Chemoradiotherapy in Participants With Localized Esophageal Squamous Cell Carcinoma — NCT03957590.
• Registry record last updated 2026-04-13; status last verified 2026-03.
• Results first posted 2026-04-13.
• 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

This is a completed phase 3 randomized, double-blind, placebo-controlled study of tislelizumab versus placebo with concurrent chemoradiotherapy in localized esophageal squamous cell carcinoma. The recorded enrollment is 370, and the primary analysis population is the intention-to-treat set.

The primary endpoint was progression-free survival, assessed by a blinded independent review committee using RECIST v1.1. The reported hazard ratio is 0.92, with a 95% confidence interval from 0.68 to 1.25 and a log-rank p-value of 0.302. The posted results do not establish a statistically clear difference between the treatment groups.

Overall survival, response rate, quality-of-life outcomes, duration of response, and treatment-emergent adverse events are listed as secondary outcomes. Some secondary results are posted, while numerical results for certain outcomes and several analysis details are not reported in the registry record.