Safety and Pharmacology Study of Atezolizumab Alone and in Combination With Bacille Calmette-Guérin (BCG) in High-Risk Non-Muscle-Invasive Bladder Cancer (NMIBC) Participants: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Safety and Pharmacology Study of Atezolizumab Alone and in Combination With Bacille Calmette-Guérin (BCG) in High-Risk Non-Muscle-Invasive Bladder Cancer (NMIBC) Participants (NCT02792192), a non-randomized phase 1, phase 2 study. The primary outcome is Percentage of Participants With Adverse Events.

2. Clinical Question

3. Trial Design

This was a Phase Ib/II, open-label, non-randomized, parallel study of atezolizumab alone and in combination with BCG in high-risk non-muscle-invasive bladder cancer. The registry lists Hoffmann-La Roche as sponsor, an enrollment of 24, a start date of 2016-06-13, primary completion on 2020-09-29, and a terminated status. The reason for termination is not reported in the registry record.

The planned study described four cohorts. Cohort 1A was atezolizumab alone for BCG-unresponsive disease; Cohort 1B was atezolizumab plus BCG for BCG-unresponsive disease. Cohort 2 and Cohort 3 were combination cohorts for BCG-relapsing and BCG-naive disease, respectively. The posted record says participants were enrolled only in Cohort 1; no participants are listed for Cohorts 2 or 3.

Atezolizumab was planned at 1200 mg intravenously every 3 weeks, for a maximum of 96 weeks in the overall study description. For Cohort 1A, the arm description specifies a maximum of 32 doses or 96 weeks, whichever came first. The combination-cohort descriptions include an induction course with atezolizumab and weekly BCG. The study plan allowed BCG dose de-escalation, but the actual administered dose by participant is not reported in the registry record.

The open-label, non-randomized structure is important for interpretation: treatment assignment was not randomized and there was no masked assessment described. The registry lists safety, tolerability, pharmacokinetics, immunogenicity, patient-reported outcomes, and preliminary anti-tumor activity as study aims. The record does not report a randomized treatment-effect estimate.

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: All participants were included in the safety-evaluable population, defined as all participants treated with any amount of study drug. Participants were enrolled only in Cohort 1 of this study. The Sponsor decided not to enroll participants in Cohorts 2 and 3 after the enrollment of 24 patients in Cohort 1, as the study met its goals of providing preliminary safety and efficacy information for atezolizumab monotherapy alone and in combination with BCG in NMIBC.

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: Participants were enrolled only in Cohort 1 of this study. The Sponsor decided to stop further enrollment for Cohorts 2 and 3 after the enrollment of 24 participants in Cohort 1 (BCG-unresponsive cohort), as the study had met its goals of providing preliminary safety and efficacy information for atezolizumab monotherapy alone and in combination with BCG in NMIBC.

6. Primary Endpoint

The registry lists the primary outcome as Percentage of Participants With Adverse Events, measured over From Baseline up to end of study (up to approximately 4.3 years).

The primary endpoint is the percentage of participants with at least one adverse event during the study, from baseline through the end of study, up to approximately 4.3 years. Its description identifies participants treated with any amount of study drug as the safety-evaluable population. This endpoint is intended to capture whether participants experienced an adverse event during the specified observation period; it does not by itself describe event severity, timing, duration, or attribution to study treatment.

As a general statistical concept, a percentage with an event summarizes the proportion of an observed group meeting a defined event criterion. It is useful in safety monitoring because it presents the frequency of participants with events in relation to the group assessed. The interpretation depends on a clear outcome definition, the population included, and the length and completeness of follow-up.

The posted record does not state the primary-outcome percentage. It does provide cohort-level counts for serious and other adverse events, but these categories cannot simply be combined to reconstruct the percentage with any adverse event: participants may appear in both categories, and the record does not give overlap information. The primary endpoint therefore cannot be recovered from the posted counts alone. No confidence interval or statistical comparison for the primary endpoint is reported in the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The main outcome is binary at the participant level: a participant either had at least one adverse event during the defined study period or did not. As a general statistical concept, a proportion is formed by relating the number of participants meeting an outcome definition to the number assessed in the relevant population. The registry describes the safety-evaluable population as all participants treated with any amount of study drug. It does not report the primary-outcome percentage, so that estimate cannot be interpreted from the posted results.

What do the adverse-event counts describe?

The registry separates serious adverse events from other adverse events in its cohort entries. In Cohort 1A, the safety entries list 3 affected among 12 at risk for serious events and 11 among 12 for other events. In Cohort 1B, they list 3 among 12 for serious events and 12 among 12 for other events. The entries are descriptive counts, not a statistical test. They do not provide a measure of uncertainty or establish a difference between regimens.

Those categories should not be added together to derive the primary endpoint. A participant could have experienced both a serious and a non-serious event, and the registry does not report category overlap. The reported deaths are also listed separately, with 1 in Cohort 1A and 0 in Cohort 1B. The record does not indicate whether the death entry overlaps with any adverse-event category or provide a cause or attribution.

Why does the design matter?

The study is non-randomized, unmasked, and parallel. As a general statistical principle, random assignment helps balance measured and unmeasured participant characteristics across treatment groups on average. Without randomization, differences in outcomes between cohorts may be related to treatment, participant characteristics, clinical decisions, or other factors. Open-label conduct can also affect event reporting or assessment, depending on how outcomes are identified and recorded. The registry does not describe procedures that would quantify or eliminate these potential influences.

The two cohorts with posted safety entries are not identical regimens: Cohort 1A received atezolizumab, while Cohort 1B received atezolizumab with BCG. A direct numerical contrast would therefore be an unadjusted descriptive comparison, not a randomized treatment effect. The record does not report adjusted analyses or a formal test comparing these groups.

What is a confidence interval, generally?

As a general statistical explanation, a confidence interval gives a range of values compatible with the observed data under a specified method and its assumptions. It helps convey precision; a point estimate alone does not show how much sampling uncertainty remains. No confidence interval for the primary outcome or cohort difference is reported in the registry record. The sample-size assumptions describe planning illustrations based on an observed effect, with a two-sided significance level of 0.05, equal allocation, and a normal approximation. These are not observed power results, and no planning output values are reported in the record.

What about event frequency and follow-up?

For an event proportion, the observation window matters: participants followed for different lengths of time may have different opportunities for an adverse event to be observed. The endpoint timeframe extends from baseline through the end of study, up to approximately 4.3 years. The record does not give individual follow-up durations, completeness of follow-up, or details of how participants with incomplete observation were handled. It also does not say whether events were counted by participant, by event episode, or using another convention beyond the outcome description and posted categories.

As a general concept, over-dispersion is relevant when data vary more than a model assumes, often because observations are more heterogeneous or dependent than expected. It is not a reported feature of this binary participant-level outcome, and the registry does not identify a model or variance adjustment. The record does not report the statistical methods used for the primary endpoint, confidence intervals, or tests. No such method should be inferred from the counts alone.

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("Cohort 1A: Atezolizumab (BCG-unresponsive NMIBC)", "Cohort 1B: Atezolizumab + BCG (BCG-unresponsive NM")
n      <- c(12, 12)
events <- c(12, 12)
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  Cohort 1A: Atezolizumab (BCG-unresponsive NMIBC)  12  12  100.0  73.5  100.0  75.8  100.0  75.3  99.8
ARM  Cohort 1B: Atezolizumab + BCG (BCG-unresponsive NM  12  12  100.0  73.5  100.0  75.8  100.0  75.3  99.8

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 14 outcome measures, including adverse events, complete response, recurrence-free survival, disease-free survival, progression-free survival, cystectomy-free survival, overall survival, and atezolizumab maximum observed serum concentration. It reports 0 analysis entries. The existence of multiple listed outcomes raises a general multiplicity issue: when many outcomes or comparisons are tested, the chance of at least one apparently notable result can increase unless the analysis plan accounts for that. The registry record does not report a multiplicity adjustment, hierarchy of outcomes, or confirmatory testing strategy.

Missing-data handling is not reported in the registry record. The safety population is described as all participants treated with any amount of study drug, but no details are given about missing event assessments, withdrawal, incomplete follow-up, or how such cases were handled in the primary endpoint. The timeframe is from baseline through the end of study, up to approximately 4.3 years, but follow-up completeness by cohort is not stated.

The record also does not report imputation methods, censoring rules for time-to-event outcomes, or sensitivity analyses for missing information. These omissions matter because incomplete outcome ascertainment can affect estimates and their precision. No assumptions about missingness or analysis populations beyond the stated safety-evaluable definition can be made from the registry record.

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: haematuria (16.7%), blood loss anaemia (8.3%), atrial fibrillation (8.3%), complication associated with device (8.3%), pyrexia (8.3%).

The serious-adverse-event entries list 3 affected participants among 12 at risk in Cohort 1A and 3 among 12 in Cohort 1B. The entries for other adverse events list 11 among 12 in Cohort 1A and 12 among 12 in Cohort 1B. Deaths are listed as 1 and 0, respectively. These are the posted cohort-level figures; the record does not state that the deaths were caused by study treatment.

The listed serious-event terms are Haematuria at 16.7%, Blood loss anaemia at 8.3%, Atrial fibrillation at 8.3%, Complication associated with device at 8.3%, Pyrexia at 8.3%, and Coronavirus infection at 8.3%. The registry does not give enough detail to determine from these term percentages whether events occurred in both cohorts, whether individuals had multiple terms, or whether terms were considered treatment-related. Serious events, other events, and deaths are distinct reporting entries; they should not be summed to estimate the primary endpoint. The primary percentage with any adverse event is not reported in the registry record.

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

The posted results document safety-event counts for the two Cohort 1 groups with participants listed. They show that serious and other adverse events were recorded in both groups, with the cohort-specific counts stated in the registry. The reported serious-event term frequencies identify the terms recorded most prominently in the listed data. These observations describe what was recorded during the study; they do not, on their own, establish why an event occurred.

The registry’s primary question was the percentage of participants with at least one adverse event. That percentage is not reported, and the serious and other event categories may overlap. The available entries therefore cannot be used to reconstruct the primary outcome. No confidence interval or hypothesis test is reported, so the precision of the endpoint estimate and the uncertainty around a cohort contrast are not quantified.

The study was non-randomized and open-label. A comparison of Cohort 1A and Cohort 1B is not equivalent to a randomized comparison, because the treatment regimens differed and assignment was not random. The reported counts cannot establish that adding BCG caused a change in adverse events. Nor do they establish that either regimen is safer, less safe, effective, or ineffective.

The record planned to study additional cohorts and outcomes, but says participants were enrolled only in Cohort 1. No participant data for the planned BCG-relapsing or BCG-naive groups are posted. The listed safety information also does not answer the study’s pharmacokinetic, immunogenicity, patient-reported outcome, or anti-tumor activity questions. Those outcomes require their own results and analysis.

In general, a small descriptive study can identify events and inform further investigation, while leaving substantial uncertainty about comparative effects. The registry does not report a statistical plan, handling of incomplete follow-up, or an adjusted comparison. Interpretation should therefore remain limited to the recorded safety entries rather than extending to causal conclusions or broader patient groups.

14. Why This Trial Matters Statistically

This trial is statistically notable because it combines early-phase safety and pharmacology aims with preliminary activity outcomes in a non-randomized, open-label design. Such a design can provide descriptive information about events observed during treatment, but it does not offer the balance of a randomized comparison. That distinction is especially important when readers compare the atezolizumab-alone cohort with the atezolizumab-plus-BCG cohort.

The registry also illustrates why outcome definitions and analysis populations matter. The primary endpoint is the percentage with at least one adverse event, and the safety-evaluable population is defined as participants treated with any amount of study drug. Yet the posted entries separate serious and other adverse events without reporting their overlap or the primary percentage. A reader cannot reliably reconstruct a participant-level “any event” outcome by combining categories.

Another statistical lesson is the difference between event descriptions and estimates of treatment effect. Counts and term frequencies can characterize recorded safety events, but without a randomized design, an inferential comparison, and a measure of precision, they do not establish that treatment caused an observed difference. The registry lists 14 outcome measures but no analysis entries, leaving multiplicity and analytic choices unreported.

Finally, the record notes that enrollment was limited to Cohort 1, with no participants in the planned later cohorts. This limits the scope of conclusions that can be drawn about the broader planned population. The study’s structure, missing analysis details, and incomplete cohort coverage all shape what the safety data can support statistically.

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.

Several limitations are specific to this registry record. First, the primary endpoint’s definition is available, but its percentage is not reported. The posted serious and other adverse-event entries do not provide their overlap, so they cannot reconstruct the percentage with at least one event. The record also does not give confidence intervals or a statistical test for this endpoint.

Second, the study was non-randomized and open-label. Cohort 1A and Cohort 1B received different regimens, so their safety entries are not a randomized like-for-like comparison. No adjusted comparative analysis is reported. The record does not state whether event assessment was independently adjudicated or whether assessors were otherwise masked.

Third, the posted results cover Cohort 1 only. Although the study plan included Cohorts 2 and 3, the registry reports no participants in those cohorts. This restricts the relevance of the posted safety information to the planned BCG-relapsing and BCG-naive populations. The actual BCG dose administered to participants is also not reported in the available cohort descriptions.

Finally, the record does not report the reason for termination, individual follow-up, missing-data methods, or treatment attribution for the listed events. It lists multiple outcomes but no analysis entries. These omissions limit assessment of completeness, precision, multiplicity, and causal interpretation; they should not be filled in by assumption.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Safety and Pharmacology Study of Atezolizumab Alone and in Combination With Bacille Calmette-Guérin (BCG) in High-Risk Non-Muscle-Invasive Bladder Cancer (NMIBC) Participants — NCT02792192.
• Registry record last updated 2021-10-28; status last verified 2021-09.
• Results first posted 2021-10-28.
• 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

NCT02792192 was a Phase Ib/II, open-label, non-randomized, parallel study of atezolizumab alone and with BCG in high-risk non-muscle-invasive bladder cancer. The registry lists enrollment of 24, a start date of 2016-06-13, primary completion on 2020-09-29, and a terminated status.

The posted safety data concern Cohort 1A and Cohort 1B; participants were enrolled only in Cohort 1. The primary endpoint was the percentage of participants with at least one adverse event, but that percentage is not reported. Cohort-level serious and other event counts and serious-event term frequencies are listed, with no reported analysis entries or comparative estimate. The reason for termination, missing-data handling, and statistical analysis plan are not reported in the registry record.