Gemcitabine + Carboplatin + Nivolumab Versus Gemcitabine + Oxaliplatin + Nivolumab in Cisplatin-ineligible Patients With Metastatic Urothelial Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Gemcitabine + Carboplatin + Nivolumab Versus Gemcitabine + Oxaliplatin + Nivolumab in Cisplatin-ineligible Patients With Metastatic Urothelial Cancer (NCT03451331), a randomized phase 2 study. The primary outcome is Objective Response Rate (ORR).

2. Clinical Question

The clinical question is about the response proportions observed with two regimens in a randomized phase 2 comparison. The registry record reports an objective response rate of 69.6% for Arm A, with a 95% confidence interval of 47.0% to 87.0%, and 33.3% for Arm B, with a 95% confidence interval of 15.0% to 57.0%. These estimates describe the posted response data; they do not by themselves establish why the arms differed or what would happen for an individual patient.

The primary outcome counts complete and partial responses, rather than stable disease, progression-free survival, overall survival or adverse events. Those other outcomes are separate questions. The record lists adverse events, duration of response, progression-free survival and overall survival as secondary outcomes, but their detailed results are not reported in the registry record described for this comparison. The posted primary result therefore addresses tumor response as defined by the trial, not the full balance of benefits and harms or longer-term outcomes.

3. Trial Design

This is a randomized phase 2 trial with a parallel-group design and no masking. Participants were assigned to Arm A, gemcitabine plus carboplatin plus nivolumab, or Arm B, gemcitabine plus oxaliplatin plus nivolumab. The registry record says randomization was stratified by lymph node only and/or unresectable primary metastatic status. Further details of the randomization procedure are not reported in the registry record.

The registry lists enrollment of 49 and describes the trial as completed. The recorded start date is 2018-05-10, and primary completion is recorded as 2023-06-07. The primary outcome timeframe is up to a maximum of 50 months. Those dates and the stated follow-up window describe the registry entry; they do not establish how long each participant was followed.

The primary endpoint is objective response rate. The record reports 23 participants in Arm A and 21 in the posted response comparison for Arm B. Its population note says that, out of 23 subjects in Arm B, two subjects were not evaluable for best response. The registry does not explain in the posted record how this evaluability handling was specified or how it relates to every analysis population.

The trial compares regimens that share gemcitabine and nivolumab but differ in the platinum drug. As a randomized phase 2 study, it provides a comparative signal in the enrolled population. The registry record does not report the trial's original sample-size justification, detailed protocol, or a prespecified statistical analysis plan. Those design details should not be inferred from the results or from the planning illustration based on the observed effect.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Out of 23 subjects in Arm B, two subjects were not evaluable for best response.

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 Objective Response Rate (ORR), measured over Up to a maximum of 50 months.

Objective response rate (ORR) is the percentage of participants who achieve a confirmed complete response or partial response, according to RECIST 1.1. In the registry description, complete response means disappearance of all target lesions. Partial response requires at least a 30% decrease in the sum of the longest diameters of target lesions. The description also defines progressive disease by an increase of at least 20% in tumor burden relative to nadir or the appearance of one or more new lesions; stable disease does not meet the criteria for complete response, partial response or progressive disease.

ORR is used as a binary outcome: each evaluable participant is classified as having achieved the specified response or not. The reported percentage summarizes the proportion meeting the response definition; it does not convey how long the response lasted. Duration of response, progression-free survival and overall survival are listed separately as secondary outcomes.

As a general statistical point, a binary response measure offers a direct way to compare the proportion meeting a defined threshold in each arm. It is not a complete account of treatment experience: it does not by itself describe the timing of response, later progression, survival, or adverse events. The registry's primary endpoint timeframe is up to a maximum of 50 months, but the record does not report further details about response assessment timing or confirmation procedures beyond the stated confirmed-response definition.

7. Results Posted in the Registry

• Arm B: 33.3% (7/21) versus 69.6% (16/23) on control; risk difference -36.2 percentage points (95% CI -63.8 to -8.7), risk ratio 0.48 (0.25–0.93); unadjusted two-proportion z-test p = 0.016.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary endpoint is binary: a participant either had a confirmed complete or partial response, or did not meet that response definition. The registry reports the percentage with an event, where the event for this endpoint is achieving a response. Arm-specific percentages are therefore proportions of participants counted as responders among the denominators used for the response estimates. This is different from a time-to-event endpoint, which uses the timing of an event and may account for censoring.

What does the confidence interval describe?

A confidence interval expresses uncertainty around an estimate under a statistical procedure. In general, a 95% confidence interval is generated by a method that, over repeated applications under its assumptions, would cover the target quantity at the stated rate. It does not mean that there is a 95% probability that this particular interval contains the true response proportion. The registry gives an interval for each arm's response percentage, but the method used to calculate those intervals is not reported in the registry record.

The intervals are 47.0% to 87.0% for Arm A and 15.0% to 57.0% for Arm B. The width of an interval is informative about precision: wider intervals indicate less precision in the arm-specific estimate. With binary outcomes, precision depends in part on the amount of response information available and on the observed proportion. These intervals are uncertainty estimates, not predictions for individual participants and not a substitute for a direct interval for the between-arm contrast.

How should the arm comparison be read?

The registry posts an unadjusted p-value of 0.016 for the comparison. In general, a p-value describes how incompatible the observed data, or data more extreme, would be with a specified null model, assuming that model and the test's assumptions. It does not quantify the probability that the null hypothesis is true. It also does not measure clinical importance or show that the observed contrast will recur in another study.

The record includes a risk-difference comparison, but interpretation should retain the stated orientation: Arm B is compared with Arm A. The posted analysis is marked as unadjusted, and the registry does not report whether a particular test or model accounted for the trial's stratification factors. The sponsor's own analysis does not conflict with the posted results, so no discrepancy caveat applies. No additional adjusted comparison is reported in the registry record.

What do randomization and masking contribute?

In general, random assignment is intended to make treatment groups comparable on average at the point of assignment, including for factors that were not measured. It does not guarantee balance in a particular trial, especially when outcome information is limited. This trial is recorded as randomized, parallel and unmasked. The registry says stratification involved lymph node only and/or unresectable primary metastatic status, but gives no further method details.

Unmasked means that masking is recorded as none. The record does not state whether response reviewers were independently masked, or describe procedures that might limit assessment bias. These details are not reported in the registry record and should not be assumed. Because response is assessed against criteria, consistency of assessment matters, but the posted entry does not provide enough information to characterize assessment procedures beyond the stated RECIST 1.1 response definition.

What is not available about the analysis?

The registry lists five outcome measures and four secondary outcomes, but reports zero analyses in the entry. That does not establish that no analyses were conducted; it means analyses beyond the posted primary comparison are not reported in the registry record. The record does not provide a detailed statistical analysis plan, the original sample-size justification, handling rules for missing data, or details of interim analyses. None should be inferred from the observed response rates or from a planning illustration calculated from the observed effect.

For this binary endpoint, the primary interpretation is therefore based on the posted arm-specific response proportions, their confidence intervals, and the unadjusted p-value. The registry's evaluability note is important context for the Arm B denominator. It does not specify the statistical treatment of the two subjects who were not evaluable for best response. Without further reporting, the effect of that missing evaluability information on the estimate cannot be determined.

10. Confidence Intervals and Precision

The registry reports an objective response rate of 69.6% for Arm A, with a 95% confidence interval of 47.0% to 87.0%. For Arm B, it reports 33.3%, with a 95% confidence interval of 15.0% to 57.0%. The interval for each arm covers a substantial range of response proportions, so the point estimates should not be read as highly precise values.

As a general statistical explanation, a confidence interval represents uncertainty about an estimated quantity under a specified method and its assumptions. A wider interval signals less precision than a narrower one. It is not a guarantee that the true value falls inside the interval, and it does not mean every value in the interval is equally compatible with the data. The registry record does not report the method used to construct these intervals.

The two arm-specific intervals are not an interval for the difference between the arms. The record does post an unadjusted p-value of 0.016 for the arm comparison, but a p-value does not convey the precision of the contrast in the same way as a confidence interval for that contrast. No further adjusted comparison or contrast interval is reported in the registry record. The estimated response percentages and the comparison should therefore be considered alongside the uncertainty in each arm and the limited response information.

Precision is also affected by how participants are classified and included. The record reports 23 participants in Arm A and 21 in the response comparison for Arm B, while its population note says two out of 23 Arm B subjects were not evaluable for best response. The reason and handling of those cases are not reported in the registry record. This incomplete detail limits assessment of how evaluability may affect precision or interpretation.

11. Planning a Similar Trial: Sample Size

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

The sample-size figures are planning illustrations based on the observed effect, not the trial's original design assumptions. The registry record does not report the original sample-size justification. The planning assumptions stated for the illustration are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is set as a stated fraction of the observed effect.

The figures are participants per arm, rather than total participants across both arms. For planning for 100% of the observed effect, the illustration gives 29 participants per arm for 80% power and 38 participants per arm for 90% power. For planning for 75% of the observed effect, it gives 52 participants per arm for 80% power and 69 participants per arm for 90% power. For planning for 50% of the observed effect, it gives 114 participants per arm for 80% power and 151 participants per arm for 90% power.

In plain language, power is the chance that a study designed under specified assumptions will detect an effect of the chosen size using the stated test threshold. The 80% power column illustrates the per-arm sample size for that design target; the 90% power column uses a higher chance of detecting the same assumed effect. Smaller effects need more participants to detect under these assumptions, as reflected in the increasing planning figures for smaller fractions of the observed effect.

These are prospective planning illustrations, not observed power. They do not describe the trial's actual design process or establish whether its original planning was adequate. Because they use the observed effect as their starting point, they are sensitive to the uncertainty in that estimate.

12. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, p₀ and p₁ are the control and treatment proportions, and p̄ is their average.

Worked calculation for this trial

1. Control proportion p₀ = 0.696; treatment proportion p₁ = 0.333; average p̄ = 0.514.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.7068 = 1.3853.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.6587 = 0.5544.
4. n = (1.3853 + 0.5544)² / (-0.3623)² = 28.7, rounded up to 29 participants per arm 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
p0 <- 0.695652173913043                          # control proportion
p1 <- 0.333333333333333                          # treatment proportion
n_per_arm <- function(power, frac = 1) {
  q1 <- p0 + (p1 - p0) * frac               # smaller effect = fraction of the observed difference
  pbar <- (p0 + q1) / 2
  z_b <- qnorm(power)
  ceiling(round((z_a * sqrt(2 * pbar * (1 - pbar)) + z_b * sqrt(p0 * (1 - p0) + q1 * (1 - q1)))^2 / (q1 - p0)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (unrounded n per arm)
cat("power.prop.test:", round(power.prop.test(p1 = p0, p2 = p1, power = 0.80)$n, 1), "\n")

# Simulation: how often does the z-test reject at the 80%-power sample size?
set.seed(3451331)
n    <- n_per_arm(0.80)
reps <- 20000
x0 <- rbinom(reps, n, p0); x1 <- rbinom(reps, n, p1)
ph <- (x0 + x1) / (2 * n)
z  <- ((x1 - x0) / n) / sqrt(ph * (1 - ph) * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(z) > z_a, na.rm = TRUE), reps))

Output from running the code

100% of observed effect: n per arm = 29 (80% power), 38 (90% power)
 75% of observed effect: n per arm = 52 (80% power), 69 (90% power)
 50% of observed effect: n per arm = 114 (80% power), 151 (90% power)
power.prop.test: 28.7 
Simulated power at n = 29 per arm: 0.801 (20000 replicates)

The simulation draws many pairs of event counts at the planned sample size and counts how often the two-proportion z-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

13. Multiplicity, Missing Data and Analysis Populations

The registry lists objective response rate as the primary outcome and lists adverse events, duration of response, progression-free survival and overall survival as secondary outcomes. It also reports five outcome measures. However, the registry record does not provide results for the listed secondary outcomes in the information available for this comparison, and it lists zero analyses. Therefore, the multiplicity strategy across outcomes or analyses is not reported in the registry record.

In general, multiplicity arises when several hypotheses or outcomes are tested. If many tests are conducted without an adjustment or prespecified hierarchy, the chance of at least one apparently notable result can increase. Adjustment methods, gatekeeping rules or a clear distinction between confirmatory and exploratory analyses can address different aspects of this issue. The registry entry does not say whether such methods were planned or used, so no conclusion about multiplicity control is warranted.

Missing-data handling and analysis populations also affect interpretation. The registry says that two subjects among 23 in Arm B were not evaluable for best response, while the posted Arm B response comparison uses 21. It does not report the reasons, whether those subjects were excluded according to a prespecified rule, or whether any sensitivity analysis considered them. The record also does not describe a full intention-to-treat, per-protocol or other analysis population definition. The available denominator information should be read as reported, without assuming an unreported missing-data strategy.

The registry's randomization description includes stratification by lymph node only and/or unresectable primary metastatic status. It does not report whether the primary comparison adjusted for those factors, nor how the stratification was implemented in analysis. The posted p-value is unadjusted. More detailed information about multiplicity, missingness, analysis populations and model specification is not reported in the registry record.

14. 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 0% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: urinary tract infection (21.7%), platelet count decreased (13.0%), febrile neutropenia (8.7%), respiratory, thoracic and mediastinal disorders - other, specify (8.7%), skin and subcutaneous tissue disorders - other, specify (8.7%).

The serious adverse event figures report 15 affected participants out of 23 at risk in Arm A and 17 out of 23 at risk in Arm B. The figures for other adverse events report 22 out of 23 in Arm A and 23 out of 23 in Arm B. Death counts are 12 in Arm A and 16 in Arm B. These are posted counts, and the registry record does not provide the context needed to attribute deaths to treatment or to distinguish causes.

The listed serious adverse event terms and percentages are urinary tract infection, 21.7%; platelet count decreased, 13.0%; febrile neutropenia, 8.7%; respiratory, thoracic and mediastinal disorders—other, specify, 8.7%; skin and subcutaneous tissue disorders—other, specify, 8.7%; and skin infection, 8.7%. The entry does not map these listed terms to a particular arm in the information, so they should not be assigned to one arm.

As a general point, adverse event counts describe events or participants as defined by the registry reporting, and they are not by themselves evidence that an event was caused by treatment. The record does not report event timing, severity details beyond the serious-event category, exposure-adjusted rates, or attribution information for these figures. They should be read as descriptive safety information rather than as a complete comparative safety analysis.

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

The posted primary result is a comparison of objective response proportions in the two randomized arms. Arm A's reported response rate is 69.6% and Arm B's is 33.3%; the unadjusted p-value is 0.016. The confidence intervals, 47.0% to 87.0% for Arm A and 15.0% to 57.0% for Arm B, show that the estimates are not precise point values. Taken together, the posted numbers describe a higher observed response proportion in Arm A in this trial, while leaving uncertainty about the underlying proportions and the size of the contrast.

This result does not establish that the observed pattern will be the same in other patient groups or settings. The registry record describes cisplatin-ineligible patients with metastatic urothelial cancer, and further population details are not reported in the registry record. The result also does not answer whether responses lasted longer, whether progression-free or overall survival differed, or how the regimens compare across the full range of adverse events. Those outcomes are listed as secondary, but detailed results are not reported in the registry record.

A p-value is not a measure of clinical importance and does not state the probability that a regimen is beneficial. An observed response proportion is also not an individual prediction. Randomization supports a comparative trial interpretation, but it cannot remove uncertainty from limited outcome information or settle questions that the reported endpoint does not measure.

The record's note that two Arm B subjects were not evaluable for best response is relevant when considering the response denominator. The reasons for non-evaluability and the analysis rules are not reported in the registry record. The available result should therefore be interpreted as the posted analysis, with that limitation in view, rather than as a complete account of every randomized participant's response status.

16. Why This Trial Matters Statistically

Statistically, this trial is notable for comparing two active combination regimens in a randomized parallel design while measuring a clearly defined binary response endpoint. The primary outcome gives a directly interpretable proportion for each arm, and the registry also reports confidence intervals and an unadjusted comparison. This makes it possible to consider both the observed response levels and their uncertainty rather than relying on a single point estimate.

The result also illustrates why an effect estimate and its precision should be read together. The observed response proportions differ, while the arm-specific confidence intervals remain broad. The unadjusted p-value contributes information about the comparison under its test assumptions, but it does not replace a contrast estimate with an interval, nor does it describe longer-term outcomes or clinical importance.

The endpoint focuses on confirmed complete or partial response according to RECIST 1.1. That is a specific question, distinct from duration of response, progression-free survival, overall survival and safety. The registry lists those outcomes but does not report their results in the record described for this comparison. Keeping endpoints distinct prevents an ORR finding from being interpreted as evidence about outcomes it does not measure.

The record also underscores the importance of transparent reporting about analysis populations and missing outcome information. It reports non-evaluable subjects in Arm B and an outcome denominator that differs from the population count in the note. The registry does not explain the reasons or analysis rules. Such details are important for understanding the result's robustness, particularly when outcome information is limited. The posted data provide a focused comparison, while leaving several design and analysis questions unresolved.

17. 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.
• Participant counts were reconstructed from posted percentages and arm sizes, so small rounding differences are possible.

The registry record does not report the original sample-size justification, detailed statistical analysis plan, specific method for constructing the confidence intervals, or details of the test underlying the unadjusted p-value. It also does not say whether the primary comparison accounted for the stated randomization stratification. These omissions limit independent assessment of the planned analysis and its assumptions.

For response, the record reports 23 participants in Arm A and 21 in the Arm B comparison, while noting that two out of 23 Arm B subjects were not evaluable for best response. Reasons for non-evaluability, prespecified handling rules, sensitivity analyses and the complete analysis-population definition are not reported in the registry record. This makes it difficult to assess the possible effect of missing response information.

The record lists secondary outcomes but does not report their results in the comparison described. It also lists zero analyses. Thus, multiplicity procedures, additional adjusted or subgroup analyses, and interim analyses are not reported in the registry record. Their absence from the entry should not be taken as proof that they were or were not conducted.

The safety reporting is limited to posted counts and selected serious adverse event terms. The record does not provide attribution, timing, duration, exposure-adjusted comparisons or detailed causes of death for these figures. The listed serious-event terms are not assigned to an arm in the reported list. The trial is unmasked, and the registry does not describe whether outcome assessment had independent masking or other safeguards. These reporting gaps constrain interpretation of response and safety comparisons.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: Gemcitabine + Carboplatin + Nivolumab Versus Gemcitabine + Oxaliplatin + Nivolumab in Cisplatin-ineligible Patients With Metastatic Urothelial Cancer — NCT03451331.
• Registry record last updated 2024-05-16; status last verified 2024-04.
• Results first posted 2024-05-16.
• 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.

21. Record Summary

The registry describes NCT03451331 as a completed randomized phase 2 trial with a parallel design and no masking. It compares gemcitabine plus carboplatin plus nivolumab in Arm A with gemcitabine plus oxaliplatin plus nivolumab in Arm B for cisplatin-ineligible patients with metastatic urothelial cancer. The listed enrollment is 49.

The primary endpoint is objective response rate, defined as confirmed complete or partial response according to RECIST 1.1, with an outcome timeframe up to a maximum of 50 months. Posted response rates are 69.6% for Arm A and 33.3% for Arm B. The registry reports 95% confidence intervals of 47.0% to 87.0% and 15.0% to 57.0%, respectively, and an unadjusted p-value of 0.016.

The record lists adverse events, duration of response, progression-free survival and overall survival as secondary outcomes. Detailed results for those outcomes, the original sample-size justification, and further analysis details are not reported in the registry record.