ROMAN: Phase 3 Trial Investigating the Effects of GC4419 on Radiation Induced Oral Mucositis in Head/Neck Cancer Patients: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for ROMAN: Phase 3 Trial Investigating the Effects of GC4419 on Radiation Induced Oral Mucositis in Head/Neck Cancer Patients (NCT03689712), a randomized phase 3 study. The primary outcome is Cumulative Incidence of Severe Oral Mucositis.

2. Clinical Question

ROMAN was a randomized phase 3 trial of GC4419 (avasopasem manganese) 90 mg versus placebo in patients receiving chemoradiotherapy for locally advanced, non-metastatic head and neck cancer. The primary question was whether the proportion of participants with any severe oral mucositis during the study treatment period differed between the randomized groups.

The outcome covered the period from the first IMRT fraction through the end of study treatment, including completion of study drug administration, IMRT, and chemotherapy. The registry estimates this treatment period to be 7 weeks. The endpoint therefore concerns whether a participant experienced severe mucositis at any time in that period, rather than the average severity or duration of symptoms.

3. Trial Design

ROMAN is registered as a randomized, quadruple-masked, parallel-group phase 3 trial. The intervention groups were GC4419 (avasopasem manganese) 90 mg and placebo. The registry lists an enrollment of 455. The posted primary outcome groups contain 241 participants in the GC4419 group and 166 in the placebo group; the record does not explain in the posted results why those group counts differ from the enrollment figure.

The trial concerned patients receiving chemoradiotherapy for locally advanced, non-metastatic head and neck cancer. The brief study description says GC4419 was administered prior to intensity-modulated radiation therapy (IMRT). The primary endpoint followed participants from the first IMRT fraction through the end of the study treatment period, which included study drug administration, IMRT, and chemotherapy and was estimated to be 7 weeks.

The primary outcome population was the intent-to-treat population as defined in the record: randomized participants who received at least one dose of avasopasem manganese or placebo, except those whose treatment course was discontinued because of the sponsor's voluntary suspension of dosing. This is the registry's stated definition; it is not simply a statement that every randomized participant was included in the posted primary outcome.

The record marks the study as completed. It lists a start date of 2018-10-03 and a primary completion date of 2021-08-13. The registry record does not report further details of the randomization procedure, the masking procedures, or the original design assumptions for the trial's sample size. The number of primary outcome measures listed is one, and no secondary outcome results are posted in the record.

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 Intent to treat population consisted of all randomized subjects who received at least 1 dose of avasopasem manganese or placebo (with the exception of subjects whose treatment course was discontinued due to the sponsor's voluntary suspension of dosing)

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: Intent-to-Treat (ITT) population includes all randomized subjects who received at least one dose of GC4419/Placebo, with the exception of the 28 subjects whose treatment course was discontinued during the Sponsor's voluntary suspension of dosing.

6. Primary Endpoint

The registry lists the primary outcome as Cumulative Incidence of Severe Oral Mucositis, measured over From the first IMRT fraction through the end of the study treatment period (completion of study drug administration, IMRT and chemotherapy), which is estimated to be 7 weeks..

The primary endpoint was the cumulative incidence of severe oral mucositis, reported as a percentage of patients. It was defined as the proportion of participants with any occurrence of WHO Grade 3 or 4 oral mucositis during the study treatment period. In the registry definition, an occurrence counted if it happened after the start of IMRT and through the last IMRT fraction within the treatment period. The observation window began with the first IMRT fraction and extended through completion of study drug administration, IMRT, and chemotherapy; this period was estimated to be 7 weeks.

This is a binary, participant-level endpoint: a participant either had at least one qualifying occurrence during the defined period or did not. It does not describe the number of episodes, the duration of mucositis, or the highest grade beyond the threshold needed to qualify. Cumulative incidence is useful for a question about how many participants experience an event over a specified period, because it incorporates whether the event occurred at any point in that window.

The registry notes that multiple imputation was used for participants with incomplete follow-up for severe mucositis through 60 Gy who had not experienced severe mucositis before the last IMRT fraction. Imputation is a way to account for missing outcome information rather than automatically treating incomplete follow-up as evidence that no event occurred. The record does not report the imputation model or its detailed assumptions.

7. Results Posted in the Registry

• GC4419 (Avasopasem Manganese) 90mg: 53.9% (130/241) versus 63.9% (106/166) on control; risk difference -9.9 percentage points (95% CI -19.6 to -0.3), risk ratio 0.84 (0.72–0.99); unadjusted two-proportion z-test p = 0.046.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is binary: each participant is classified according to whether any severe oral mucositis occurred during the specified treatment period. A proportion is the number of participants meeting that event definition relative to the relevant group size. This presentation is different from a continuous outcome, such as a mean symptom score, and different from a time-to-event outcome, which would use the timing of an event. Here, the registry focuses on whether an event occurred within the treatment window.

What the risk difference expresses

The risk difference compares the event proportions by subtracting the placebo proportion from the GC4419 proportion. In the posted comparison, the risk difference is -0.099. Its negative direction indicates that the observed proportion was lower in the GC4419 group. The risk difference is an absolute contrast on the proportion scale, not a relative percentage reduction. It is often straightforward to interpret because it describes the distance between the two event proportions, but the estimate remains subject to sampling uncertainty and the design and analysis limitations of the study.

The posted comparison also gives a confidence interval and an unadjusted p-value. These are complementary summaries, not competing answers. The estimate describes the observed contrast; the confidence interval describes the range of values compatible with the data under the interval method; and the p-value summarizes how unusual results at least this discrepant would be under a specified null hypothesis and test procedure. A p-value does not give the probability that the null hypothesis is true. The record labels the p-value unadjusted, and no multiplicity adjustment is reported.

How to understand the endpoint window

Cumulative incidence counts participants who experienced at least one qualifying event during a specified period. For this outcome, the qualifying event was any WHO Grade 3 or 4 oral mucositis. The registry describes the window as beginning at the first IMRT fraction and continuing through the study treatment period, estimated to be 7 weeks. Because the measure records any occurrence, it does not distinguish a single episode from multiple episodes or convey how long an episode lasted. Nor does it show how severity changed over time.

The endpoint definition includes imputation for certain incomplete observations. Participants with incomplete follow-up for severe mucositis through 60 Gy and no qualifying event before the last IMRT fraction were handled using multiple imputation. In general, multiple imputation replaces missing values with a set of plausible values generated under a statistical model, then combines the resulting analyses to reflect uncertainty about those missing values. This approach can preserve information better than simply excluding every participant with incomplete follow-up, but its validity depends on the assumptions and model being appropriate. The detailed model, imputation variables, number of imputations, and sensitivity analyses are not reported in the registry record.

Analysis population and randomization

The registry defines the primary outcome population as randomized participants who received at least one dose of avasopasem manganese or placebo, with an exception for participants whose treatment course was discontinued because of the sponsor's voluntary suspension of dosing. In general, an analysis population determines which participants contribute to an estimate. Randomization is intended to support fair comparisons between groups, but the population used for analysis and the handling of post-randomization events are important to interpretation. The registry's stated definition should be kept in view rather than assuming that the posted result includes every randomized participant.

The record describes the study as randomized, quadruple-masked, and parallel-group. Randomization can reduce systematic differences between groups at assignment, while masking can reduce some forms of bias in treatment delivery, outcome assessment, and reporting. The registry does not report the specific randomization method or explain which roles were masked. Those procedural details are not reported in the registry record.

Planning calculations and their limits

The planning illustrations use the observed effect as a reference and consider 100%, 75%, and 50% of that effect. They assume a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The displayed requirements are participants per arm for this binary outcome. For the time-to-event setting, the record's general note says planning figures would instead represent total events, but the primary outcome in ROMAN is not a time-to-event endpoint.

These calculations are illustrations based on the observed effect, not the trial's original design calculation. The registry does not report the trial's own sample-size justification or original planning assumptions. The illustrations should not be interpreted as observed power, nor as evidence that the trial was designed around the effect ultimately observed. They show how a planning requirement changes under specified assumptions and effect targets.

Over-dispersion is a concern in some types of data when variability exceeds what a simple model assumes, but the primary outcome is reported as a binary percentage, and over-dispersion is not a separate reported feature of this endpoint. More generally, inference depends on the outcome definition, the analysis population, missing-data assumptions, and the statistical procedure. The registry does not report a full statistical analysis plan or details of any adjusted model.

10. Confidence Intervals and Precision

The posted risk-difference estimate is -0.099, and its 95% confidence interval runs from -0.196 to -0.003. This interval describes uncertainty around the estimated contrast. In general, a confidence interval is constructed by a method that, under its assumptions, captures the underlying parameter at a stated frequency across repeated applications. It is not a range containing a known proportion of individual participant outcomes, and it does not assign a probability to the true effect after the data have been observed.

Both endpoints of the reported interval are negative, but the upper endpoint is close to zero. The interval therefore indicates that the posted data are compatible with a comparatively small negative difference as well as with a larger negative difference. Its width is important: the point estimate alone does not convey the uncertainty. The interval does not establish that every value within it is equally plausible, nor does it account for sources of uncertainty that were not included in the calculation, such as uncertainty about the analysis choices or missing-data assumptions.

The registry also reports event proportions of 53.8% for GC4419 and 64.0% for placebo. Those percentages are descriptive estimates for the posted groups. The confidence interval applies to the comparative risk-difference estimate, not separately to each arm's percentage. The registry does not post arm-specific confidence intervals for these percentages.

The unadjusted p-value is 0.046. It provides a test-based summary under a null hypothesis and the stated analysis, while the interval conveys the magnitude and precision of the estimated difference. Neither statistic indicates clinical importance on its own. The record does not report a clinically meaningful threshold against which the estimate or interval could be compared.

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 planning figures are participants per arm for 80% power and participants per arm for 90% power. The record's planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The calculations use the observed effect as a reference and examine planning for 100%, 75%, and 50% of that observed effect. These are planning illustrations, not observed power.

For planning for 100% of the observed effect, the figures are 386 participants per arm for 80% power and 516 participants per arm for 90% power. For planning for 75% of the observed effect, they are 680 participants per arm for 80% power and 910 participants per arm for 90% power. For planning for 50% of the observed effect, they are 1,514 participants per arm for 80% power and 2,026 participants per arm for 90% power.

In plain terms, 80% power is the chance, under the planning assumptions and target effect, of detecting that effect at the stated significance level if it is truly present. The 90% power column expresses a higher planned chance of detection under the same framework. A higher power target generally requires a larger planned sample. Smaller effects also require more participants to distinguish them from random variation, which is why planning for a smaller fraction of the observed effect gives larger requirements.

The figures are not the trial's reported enrollment and should not be read as a retrospective judgment about whether the completed trial was adequate. The original sample-size justification and design assumptions are not reported in the registry record. For this binary endpoint, the figures represent participants per arm; they are not total events.

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 = 0.639; treatment proportion p₁ = 0.539; average p̄ = 0.589.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.6958 = 1.3638.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.6923 = 0.5826.
4. n = (1.3638 + 0.5826)² / (-0.0991)² = 385.5, rounded up to 386 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 <- 0.63855421686747                          # control proportion
p1 <- 0.539419087136929                          # 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(3689712)
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 = 386 (80% power), 516 (90% power)
 75% of observed effect: n per arm = 680 (80% power), 910 (90% power)
 50% of observed effect: n per arm = 1,514 (80% power), 2,026 (90% power)
power.prop.test: 385.5 
Simulated power at n = 386 per arm: 0.807 (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 one primary outcome measure and posts no secondary outcome results. It also lists zero analyses. The record therefore does not provide a broader set of posted efficacy analyses with which to assess whether multiple comparisons were made or how they were handled. The primary comparison's p-value is explicitly unadjusted. A multiplicity adjustment, hierarchical testing procedure, or other approach to controlling error across multiple tests is not reported in the registry record.

Multiplicity matters when multiple outcomes, treatment contrasts, time points, or analysis choices are tested. If many tests are conducted, the chance of at least one apparently notable result can increase unless the analysis plan accounts for that testing. This is a general statistical consideration, not evidence that any unreported testing took place in ROMAN. The registry record does not report an analysis plan that would resolve this question.

Missing outcome information was addressed using multiple imputation for participants with incomplete severe-mucositis follow-up through 60 Gy who had no occurrence before the last IMRT fraction. The record identifies this procedure but does not report the model, variables, assumptions, or sensitivity analyses. In general, imputation relies on assumptions about how missingness relates to observed and unobserved information. Without the detailed specification, readers cannot independently assess whether those assumptions are plausible or how robust the result is to alternative missing-data approaches.

The stated primary outcome population includes randomized participants who received at least one dose, except participants whose treatment course was discontinued because of the sponsor's voluntary suspension of dosing. The record does not provide a separate posted analysis including all randomized participants or a sensitivity analysis using a different population definition. The primary outcome group counts are 241 for GC4419 and 166 for placebo; the registry's enrollment figure is 455. The registry does not explain the relationship between those counts in the posted results.

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 3% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: febrile neutropenia (6.2%), failure to thrive (3.6%), vomiting (3.6%), acute kidney injury (3.0%), dehydration (2.9%).

The serious adverse-event summaries list 81 participants affected among 241 at risk in the GC4419 (avasopasem manganese) 90 mg group, and 52 among 166 at risk in the placebo group. The record also lists other adverse events affecting 240 of 241 participants in the GC4419 group and 166 of 166 in the placebo group. These are participant counts in the posted safety summaries; they are not the primary mucositis outcome.

The registry lists 20 deaths in the GC4419 group and 10 in the placebo group. It also identifies the following top serious adverse-event terms with their registry percentages: Febrile Neutropenia, 6.2%; Failure to Thrive, 3.6%; Vomiting, 3.6%; Acute Kidney Injury, 3.0%; Dehydration, 2.9%; and Pulmonary Embolism, 2.9%. The record does not provide counts for those listed terms in this summary.

Safety counts and percentages describe events recorded in the study groups. They do not establish that an event was caused by a study intervention, and the registry summary alone does not account for differences in exposure, duration, or participant characteristics. These figures should be read as descriptive safety information rather than as a stand-alone causal comparison.

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

The posted results show that severe oral mucositis was reported in 53.8% of the GC4419 (avasopasem manganese) 90 mg group and 64.0% of the placebo group during the defined study treatment period. The posted unadjusted risk difference is -0.099, with a 95% confidence interval from -0.196 to -0.003 and an unadjusted p-value of 0.046. These statistics describe the observed primary outcome contrast and its uncertainty in the registry's stated analysis population.

The result does not by itself establish the clinical importance of the difference. The registry record does not report a threshold for a clinically meaningful change. The confidence interval includes estimates close to zero as well as more negative values, so the magnitude is not determined with high precision. A p-value is not the probability that the treatment has no effect, and it does not indicate whether a difference matters to patients.

The endpoint counts any occurrence of WHO Grade 3 or 4 oral mucositis during the treatment window. It does not summarize the duration of symptoms, timing of onset, recurrence, or experiences below the severe threshold. The registry record does not post secondary outcome analyses that might address other dimensions of benefit or burden.

The result also should not be generalized beyond the studied context without additional evidence. It concerns patients with locally advanced, non-metastatic squamous cell carcinoma of the head and neck receiving chemoradiotherapy, and the outcome window is tied to IMRT and the study treatment period. Details of the original sample-size justification, full analysis plan, and missing-data model are not reported in the registry record. The safety summaries are descriptive and do not establish causality.

16. Why This Trial Matters Statistically

This trial illustrates how a binary clinical endpoint can be summarized using both group-specific event proportions and a comparative risk difference. The proportions show how common the defined event was in each posted group; the risk difference gives the absolute contrast between groups. Reading these together helps distinguish the observed group outcomes from the comparative estimate and its uncertainty.

The endpoint definition is central to the statistical meaning of the result. Severe oral mucositis was counted if it occurred at least once during the specified treatment period. That definition makes cumulative incidence directly relevant to the question of whether participants experienced severe mucositis, while leaving unanswered questions about duration, timing, and symptom course. Clear endpoint timing and grading criteria are necessary for interpreting any percentage as a clinical outcome.

The registry's use of multiple imputation for specified incomplete follow-up also highlights that missing data are part of the analysis, not merely a reporting detail. The method is named, but its model and assumptions are not reported in the registry record. That limits independent evaluation of how missing observations may influence the estimate.

Finally, the sample-size illustrations show how planning depends on the effect size a study aims to detect and the desired power. Smaller target effects call for larger planned groups under the stated assumptions. These planning calculations are distinct from observed power and from the trial's original design rationale, which is not reported. Keeping those distinctions clear prevents a retrospective calculation from being mistaken for the study's actual design plan.

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 leaves several trial-specific uncertainties. Although it identifies the study as randomized, quadruple-masked, and parallel-group, it does not report the randomization method or the roles covered by masking. The original sample-size justification, original design assumptions, and full statistical analysis plan are not reported in the registry record.

The primary outcome population is defined as randomized participants who received at least one dose, with an exception for participants whose treatment course was discontinued because of the sponsor's voluntary suspension of dosing. The record lists enrollment of 455 and primary outcome group counts of 241 and 166, but does not explain their relationship in the posted results. No alternative population analysis is posted.

Multiple imputation was used for a specified group with incomplete follow-up, but details needed to assess that procedure are not reported in the registry record. The number and pattern of missing observations, the imputation model, its assumptions, and sensitivity analyses are not reported. The registry also does not provide secondary outcome analyses or a reported multiplicity strategy.

The primary endpoint captures any occurrence of severe oral mucositis during the treatment period, not its duration, onset, recurrence, or severity trajectory. The record does not report a threshold for clinical importance. The safety summary provides participant counts and selected serious-event terms, but does not by itself establish causality or account for exposure duration. These limitations constrain interpretation of both the treatment comparison and the broader clinical context.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: ROMAN: Phase 3 Trial Investigating the Effects of GC4419 on Radiation Induced Oral Mucositis in Head/Neck Cancer Patients — NCT03689712.
• Registry record last updated 2023-10-24; status last verified 2023-10.
• Results first posted 2023-04-13.
• 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.

21. Record Summary

ROMAN (NCT03689712) is a completed randomized, quadruple-masked, parallel-group phase 3 trial sponsored by Galera Therapeutics, Inc. It compared GC4419 (avasopasem manganese) 90 mg with placebo in patients receiving chemoradiotherapy for locally advanced, non-metastatic head and neck cancer. Registry enrollment was 455.

The primary outcome was cumulative incidence of severe oral mucositis, defined as any WHO Grade 3 or 4 occurrence during the study treatment period, estimated to be 7 weeks. The posted group percentages were 53.8% for GC4419 and 64.0% for placebo. The posted unadjusted risk difference was -0.099, with a 95% confidence interval from -0.196 to -0.003 and an unadjusted p-value of 0.046.

The registry describes multiple imputation for specified incomplete follow-up and defines the primary outcome population with an exception related to voluntary suspension of dosing. No secondary outcome results are posted. The detailed imputation model, full statistical analysis plan, and original sample-size assumptions are not reported in the registry record.