A Study of Atezolizumab in Locally Advanced or Metastatic Urothelial or Non-Urothelial Carcinoma of the Urinary Tract: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study of Atezolizumab in Locally Advanced or Metastatic Urothelial or Non-Urothelial Carcinoma of the Urinary Tract (NCT02928406), a na phase 3 study. The primary outcome is Percentage of Participants With Adverse Events (AEs).

2. Clinical Question

In statistical terms, the central question is descriptive: what adverse events were recorded among participants treated in this study, and how many participants were affected? A single-arm safety summary can characterize events observed during treatment and follow-up. By itself, it cannot establish how those outcomes compare with another treatment or determine whether an event was caused by atezolizumab. That distinction matters because the registry’s adverse-event definition includes events whether or not investigators considered them related to the medicinal product.

3. Trial Design

The registry identifies this as a Phase III study, while its study summary describes it as a Phase IIIb, multicenter safety study. It was open-label and single-arm. Participants received atezolizumab; there was no randomized allocation to a comparison treatment. The record lists Hoffmann-La Roche as the sponsor and urinary tract cancer as the condition.

The study concerned locally advanced or metastatic urothelial or non-urothelial carcinoma of the urinary tract. The stated treatment setting was second- to fourth-line therapy. Atezolizumab was administered every 3 weeks, with treatment continuing until loss of clinical benefit as assessed by the investigator, unacceptable toxicity, a decision to withdraw by the investigator or participant, or death, whichever occurred first. The primary safety timeframe ran from baseline to the end of the study, described as up to approximately 6 years.

The registry lists a start date of 2016-11-30, primary completion on 2022-12-12, and status as completed. The registered enrollment was 1004. The safety population included enrolled participants who received at least one dose of study medication; the adverse-event summary lists 997 participants at risk for each of its reported categories. Those figures describe different elements of the record and should not be treated as interchangeable.

Several secondary outcomes are listed, including overall survival, progression-free survival under RECIST v1.1 and modified RECIST, best overall response, disease control and duration of response. However, numerical results for these efficacy outcomes are not reported in the registry record. The record’s design information also does not report a detailed statistical analysis plan, a sample-size justification, or whether interim analyses were planned.

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: The safety population included all enrolled participants who received at least one dose of study medication.

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: The safety population included all enrolled participants who had received at least 1 dose of atezolizumab. The safety population was used for all baseline and safety analyses.

6. Primary Endpoint

The registry lists the primary outcome as Percentage of Participants With Adverse Events (AEs), measured over Baseline up to end of study (up to approximately 6 years).

The primary endpoint is the percentage of participants with adverse events, measured from baseline through the end of the study, up to approximately 6 years. The registry defines an adverse event as an untoward medical occurrence after administration of a pharmaceutical product, regardless of whether it is attributed to that product. The definition includes signs, symptoms, diseases, worsening or recurrence of conditions, certain laboratory or clinical-test changes, and events related to protocol-mandated interventions.

Although the endpoint is expressed as a percentage, the posted safety summary gives participant counts for the reported adverse-event categories. It lists 380 participants affected by serious adverse events and 888 affected by other adverse events, each among 997 at risk. It also lists 780 deaths. These counts describe what was recorded in the study; they do not, by themselves, establish cause or treatment attribution.

General statistical context: an adverse-event endpoint is used to describe safety experience over a defined observation period. A participant count answers how many people experienced an event category, while an event count would instead count occurrences and could include more than one occurrence per person. The record’s measure is participant-based. The registry does not report a numerical percentage for the primary endpoint in the posted adverse-event summary, nor does it report confidence intervals for these counts.

Because the study has no control group, the endpoint supports a description of events among treated participants, not a direct estimate of how event frequency differs from an alternative treatment or from the underlying disease course.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does a single-arm design estimate?

General statistical context: a single-arm study describes outcomes within one treated group. For a binary outcome such as whether a participant experienced an event, the basic data are the number of participants with the event and the number evaluated or at risk. The registry reports participant counts for the safety categories and an at-risk count of 997 for each. The record does not provide a control-group estimate, so a treatment-versus-control contrast cannot be calculated from these results.

A within-group event frequency and a comparative treatment effect answer different questions. The former describes the study participants’ observed experience. The latter requires a valid comparison and depends on how groups were formed, how outcomes were measured and how differences were analyzed. Randomization can help make comparison groups similar in expectation, but this study has no randomized comparator. Comparisons with other studies would also need careful consideration of differences in eligibility, follow-up, outcome definitions and clinical practice; such external comparisons are not reported in the registry record.

Why distinguish participants from events?

General statistical context: a participant-based count records whether each person experienced an outcome, generally counting a person once for that category even if the event recurred. An event-based count instead records every occurrence and can therefore exceed the number of participants affected. The primary endpoint is expressed as the percentage of participants with adverse events, and the safety summary gives affected-participant counts. It does not provide recurrence counts or a detailed rule for handling multiple events of the same type.

Categories may overlap. A participant with a serious adverse event could also have another adverse event or die during the study. Unless categories are explicitly defined as mutually exclusive, their counts should not be added to create a total number of affected participants. The registry record does not state that the categories in the summary are mutually exclusive.

What do uncertainty measures add?

General statistical context: a confidence interval gives a range of values compatible with the observed data under a specified statistical method and its assumptions. It conveys estimation uncertainty, not the probability that a particular treatment caused an event. For event counts or proportions, interval estimates can help readers judge precision, especially when the number of observed events is limited. The registry record does not report confidence intervals for the adverse-event counts or the listed serious-event terms, so their precision cannot be assessed from the reported figures alone.

Statistical methods also depend on the observation period. Participants may have different durations of follow-up, and a simple participant count does not by itself account for time at risk. The primary safety timeframe is baseline to the end of study, up to approximately 6 years. A time-to-event analysis can incorporate follow-up time and censoring, but the record does not report such an analysis for the primary safety endpoint. It also does not report how incomplete follow-up was handled for the adverse-event summaries.

How should the planning illustrations be read?

General statistical context: a sample-size calculation asks how many observations might be needed to detect a chosen effect under specified assumptions. The planning illustrations associated with the posted results use the observed effect as a starting point, vary the effect to be detected by stated fractions, and assume a two-sided significance level of 0.05, equal allocation between two arms and a normal approximation. These calculations are planning illustrations, not evidence that the study had a particular probability of detecting an effect.

Those assumptions do not describe this trial’s design: this study is single-arm and has no equal allocation between two treatment arms. The illustrations cannot be read as the study’s original sample-size plan or as observed power. The registry record does not report the trial’s sample-size justification, planned target effect, statistical power, or a corresponding two-arm design. No unreported planning values should be inferred from the study’s enrollment.

What can be said about analysis choices?

The registry lists the primary safety endpoint and secondary efficacy outcomes, but does not report a detailed statistical analysis plan for the results. The analysis method for the primary endpoint, confidence-interval method, handling of repeated events and treatment of missing outcome data are not reported in the registry record. For the listed efficacy outcomes, the registry names measures such as RECIST v1.1 and modified RECIST, but numerical results and analysis details are not reported in the registry record.

General statistical context: a prespecified analysis plan helps readers distinguish planned analyses from choices made after observing the data. It can specify which population is analyzed, how outcomes are defined, how missing observations are handled and how multiple outcomes are interpreted. Without these details, the numerical summaries can still be described, but the assumptions behind more elaborate inferential conclusions cannot be evaluated from the registry record.

10. Exact Estimates for Each Arm

No two-arm comparison is shown for this trial, but each arm's result can still be summarized with exact interval estimates computed from the posted counts. The Clopper-Pearson interval is exact and conservative; the Wilson interval is a common closer-to-nominal alternative; the Bayesian credible interval uses a uniform Beta(1, 1) prior. For small groups the three can differ noticeably.

R code and output

arm    <- c("Atezolizumab")
n      <- c(997)
events <- c(900)
for (i in seq_along(n)) {
  ci <- binom.test(events[i], n[i])$conf.int          # Clopper-Pearson exact interval
  cr <- qbeta(c(0.025, 0.975), 1 + events[i], 1 + n[i] - events[i])   # Bayesian, uniform Beta(1,1) prior
  wi <- (function(x, n) { p <- x / n; z <- qnorm(0.975); d <- 1 + z^2 / n
    c((p + z^2 / (2 * n) - z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d, (p + z^2 / (2 * n) + z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d) })(events[i], n[i])
  cat(sprintf("ARM|%s|%d|%d|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f\n", arm[i], n[i], events[i], 100 * events[i] / n[i],
              100 * ci[1], 100 * ci[2], 100 * wi[1], 100 * wi[2], 100 * cr[1], 100 * cr[2]))
}

ARM  Atezolizumab  997  900  90.3  88.3  92.0  88.3  92.0  88.3  92.0

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 12 outcome measures, including the primary adverse-event endpoint and secondary measures of survival, progression, response and disease control. It records 0 analyses. Numerical efficacy results and a statistical analysis plan are not reported in the registry record. The record therefore does not explain whether, or how, the many listed outcomes were prioritized in the analysis or whether any adjustment for multiple comparisons was planned.

General statistical context: when several outcomes or statistical tests are considered, the chance of at least one apparently unusual result can increase if every result is interpreted without accounting for multiplicity. Adjustments or a prespecified hierarchy can address particular multiplicity concerns, but neither should be assumed without documentation. In a single-arm study, the lack of a randomized comparator remains a separate limitation that multiplicity procedures cannot resolve.

The registry defines the safety population as enrolled participants who received at least one dose of study medication. The adverse-event summary lists 997 at risk. The record does not report a participant-flow account explaining the difference between that safety count and the registered enrollment of 1004, nor does it detail missingness by outcome. It also does not report the handling of incomplete follow-up, participants who discontinued treatment, or missing efficacy assessments.

These omissions matter because analysis populations and missing-data methods affect the interpretation of a result. General statistical context: excluding participants with incomplete data can change a summary if missingness is related to health status or outcome experience. Sensitivity analyses can examine how conclusions depend on assumptions about missing observations, but such analyses are not reported in the registry record. The registry entries support describing the listed safety counts; they do not support claims about unreported missing-data procedures or multiplicity control.

12. 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: urinary tract infection (5.9%), urosepsis (1.9%), haematuria (1.9%), pneumonia (1.7%), intestinal obstruction (1.6%).

The serious adverse-event summary lists 380 affected participants among 997 at risk. The separate other-adverse-event summary lists 888 affected participants among 997 at risk, and the death count is 780. These are participant counts, not a set of categories that can safely be added together. The registry record does not state that the categories are mutually exclusive, and it does not provide a causal attribution for the deaths.

The listed serious-event terms are urinary tract infection, 5.9%; urosepsis, 1.9%; haematuria, 1.9%; pneumonia, 1.7%; intestinal obstruction, 1.6%; and acute kidney injury, 1.6%. These percentages are registry figures. The record does not provide confidence intervals for them or enough detail to reconstruct their analysis.

General statistical context: “serious” is a regulatory classification of an adverse event and is not simply a synonym for “severe.” An adverse event can be serious because of its outcome or consequences, while severity describes intensity. The registry’s general adverse-event definition includes events regardless of causal attribution. Therefore, the counts should be read as events recorded during the study, not as events proven to have been caused by atezolizumab.

The study has no control group, so the counts do not establish whether event frequency was higher or lower than it would have been under another treatment. The safety population included participants who received at least one dose; details about follow-up duration for individual participants and the handling of missing safety information are not reported in the registry record.

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

The results establish that the registry records substantial numbers of participants with adverse events during this single-arm atezolizumab study. Specifically, it reports 380 participants with serious adverse events and 888 with other adverse events, each among 997 at risk, along with 780 deaths. It also lists selected serious adverse-event terms with their registry percentages. These figures describe the recorded safety experience; they do not establish why the events occurred.

The results do not establish that atezolizumab caused any particular event, because the adverse-event definition includes events regardless of causal attribution and the study lacks a control group. Nor do the counts establish how safety compares with another treatment or with the expected course of urinary tract cancer. Differences in disease status, prior treatment, follow-up and other factors cannot be separated from treatment effects using a single treated group alone.

The registry names efficacy outcomes such as overall survival, progression-free survival, response and disease control. However, numerical efficacy results are not reported in the registry record. The safety counts cannot be used as a substitute for those outcomes, and the record does not support a conclusion about the efficacy of atezolizumab.

General statistical context: a study result is most informative when the outcome definition, analysis population, follow-up, uncertainty and comparison are all clear. The record defines the safety population and describes the adverse-event endpoint, but does not report confidence intervals, a detailed missing-data approach or a full statistical analysis plan. These gaps limit assessment of precision and analytic choices.

Accordingly, the defensible interpretation is descriptive: the posted safety summary characterizes events among treated participants in a completed, open-label, single-arm study. It is not a randomized estimate of comparative risk, a causal assessment of individual events, or a complete account of the study’s efficacy findings.

14. Why This Trial Matters Statistically

This trial is statistically important to read as a safety description rather than as a comparative treatment experiment. Its open-label, single-arm design permits reporting what happened among participants who received atezolizumab, but it does not supply a randomized control group. That distinction sets a boundary on interpretation: descriptive counts are available, while comparative causal effects are not established by this design.

The primary endpoint also illustrates why outcome definitions matter. The endpoint is the percentage of participants with adverse events, while the posted summary gives counts of affected participants and uses a broad adverse-event definition that does not depend on attribution. A reader should distinguish the occurrence of an event from a judgment about its cause, and distinguish participant counts from counts of repeated event occurrences.

The record lists a broad set of secondary outcomes, including survival and response measures, but numerical results for these outcomes are not reported in the registry record. This highlights the difference between an outcome being listed and a result being available for interpretation. The absence of reported estimates, uncertainty measures and analysis details constrains what can be concluded from the registry.

Finally, the planning illustrations associated with the posted results must be kept separate from the trial’s design. Their assumptions describe hypothetical planning scenarios, including equal allocation between two arms, whereas this study was single-arm. Such illustrations are not observed power and do not establish the study’s original sample-size rationale. The registry record does not report that rationale. Recognizing these distinctions helps prevent descriptive safety data and hypothetical planning calculations from being mistaken for comparative or causal evidence.

15. 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 single-arm design is the central limitation for interpretation. With no randomized comparator, the registry’s adverse-event counts cannot show whether the observed events were more or less frequent than under another treatment. The open-label design also means treatment masking was not used; the registry record does not describe any procedures intended to reduce assessment bias.

The safety findings apply to the safety population, defined as enrolled participants who received at least one dose. The registry lists enrollment of 1004 and 997 at risk in the adverse-event summary, but a detailed participant-flow account is not reported in the registry record. The record also does not provide individual follow-up durations or explain how incomplete safety follow-up was handled, despite the primary timeframe extending from baseline to study end, up to approximately 6 years.

The adverse-event definition includes events regardless of causal attribution. The reported counts therefore cannot be treated as confirmed treatment-related harms. The record does not provide a causal breakdown for the listed events or deaths. It also does not report confidence intervals for the safety summaries, limiting assessment of statistical precision.

Although overall survival, progression-free survival, response and disease-control outcomes are listed, their numerical results are not reported in the registry record. The statistical analysis plan, missing-data procedures, multiplicity approach and sample-size justification are also not reported. These limitations leave important questions about analysis choices and outcome interpretation unanswered. The record’s stated enrollment and safety counts remain useful descriptive information, but they do not resolve those gaps or support comparative conclusions.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: A Study of Atezolizumab in Locally Advanced or Metastatic Urothelial or Non-Urothelial Carcinoma of the Urinary Tract — NCT02928406.
• Registry record last updated 2024-04-04; status last verified 2024-03.
• Results first posted 2024-04-04.
• 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.

19. Record Summary

This completed registry record concerns a Phase III study described in its summary as a Phase IIIb, multicenter safety study of atezolizumab for locally advanced or metastatic urothelial or non-urothelial carcinoma of the urinary tract. It was open-label and single-arm, with treatment every 3 weeks and no randomized comparator. The stated treatment setting was second- to fourth-line therapy.

The primary endpoint was the percentage of participants with adverse events from baseline through study end, up to approximately 6 years. The safety population included enrolled participants who received at least one dose. The registry reports 380 participants with serious adverse events, 888 with other adverse events, and 780 deaths; the adverse-event categories each list 997 at risk. The registry also lists selected serious-event terms and percentages.

Numerical efficacy results, confidence intervals, detailed analysis methods, missing-data procedures and a sample-size justification are not reported in the registry record. The safety findings are descriptive and do not establish comparative risk or causal attribution.