A Study to Evaluate the Efficacy and Safety of Tocilizumab in Hospitalized Participants With COVID-19 Pneumonia: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study to Evaluate the Efficacy and Safety of Tocilizumab in Hospitalized Participants With COVID-19 Pneumonia (NCT04372186), a randomized phase 3 study. The primary outcome is Cumulative Proportion of Participants Who Died or Required Mechanical Ventilation by Day 28.

2. Clinical Question

3. Trial Design

This was a Phase 3, randomized, double-blind, parallel-group, placebo-controlled study. It compared tocilizumab plus standard of care with placebo plus standard of care in hospitalized participants with COVID-19 pneumonia. The registry record describes one intravenous infusion for each assigned intervention, with up to one additional infusion permitted. Enrollment was 377. The study status is completed; the recorded start date is 2020-05-14 and the primary completion date is 2020-08-18.

The primary endpoint was the cumulative proportion of participants who died or required mechanical ventilation by Day 28. The primary analysis population is described as the mITT population: all randomized participants who received study treatment, analyzed according to treatment assignment at randomization. This population definition is important because it specifies both who contributed to the primary endpoint results and how participants were grouped.

The registry record also lists secondary outcomes concerning hospital discharge or readiness for discharge, improvement on a seven-category clinical-status scale, time to clinical failure, mortality by Day 28, clinical status at Day 28, and adverse events. It describes an optional long-term extension for eligible participants to explore sequelae after resolved COVID-19 pneumonia. The record does not report the study’s original sample-size justification or detailed statistical analysis plan, so the planning illustrations associated with the posted effect should not be read as the trial’s original design assumptions.

Randomization helps balance measured and unmeasured participant characteristics on average, while double masking is intended to reduce bias related to knowledge of assignment. Neither feature removes uncertainty in the outcome estimates. The confidence intervals remain important for understanding how precisely the primary contrast is estimated.

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: mITT population = all randomized participants who received study treatment, grouped according to the treatment assigned at randomization.

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: mITT population = all randomized participants who received study treatment, grouped according to the treatment assigned at randomization. Participants not receiving study treatment were excluded from analyses.

6. Primary Endpoint

The registry lists the primary outcome as Cumulative Proportion of Participants Who Died or Required Mechanical Ventilation by Day 28, measured over Up to Day 28.

The primary endpoint is the cumulative proportion of participants who died or required mechanical ventilation by Day 28. It combines two clinically serious outcomes into a binary event: a participant either met the endpoint during the stated period or did not. The registry describes the measure as a percentage of participants meeting the endpoint, with a 95% confidence interval.

A combined endpoint can capture more than one way that a participant may experience a serious outcome, and can therefore summarize the occurrence of either event in one measure. Its interpretation depends on the component events and the specified time window. A result for the combined endpoint does not show, by itself, whether any difference is mainly associated with death, mechanical ventilation, or both. The registry record does not provide separate primary-endpoint estimates for those components.

The Day 28 time limit defines the period over which the cumulative proportion is assessed. This differs from a time-to-event measure, which uses information about when an event occurred. The posted primary endpoint percentages describe whether participants met the endpoint within the stated period; they do not indicate the timing or duration of illness among participants who did not meet it.

Using the mITT population also gives the endpoint a specific analysis context: participants had to be randomized and receive study treatment, and they were grouped by their randomized assignment. The registry record’s definition should be kept in mind when interpreting the reported group estimates.

7. Results Posted in the Registry

• Tocilizumab: 12.0% (30/249) versus 19.5% (25/128) on control; risk difference -7.5 percentage points (95% CI -15.5 to 0.5), risk ratio 0.62 (0.38–1.00); unadjusted two-proportion z-test p = 0.051.

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Mortality Rate by Day 28

• Endpoint type: Binary: the percentage of participants with the event (counts rebuilt from the posted percentages and group sizes).
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Tocilizumab: 10.4% (26/249) versus 8.6% (11/128) on control; risk difference 1.8 percentage points (95% CI -4.3 to 8.0), risk ratio 1.22 (0.62–2.38); unadjusted two-proportion z-test p = 0.568.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Tocilizumab and the control. The interval gives the range of effects that remain compatible with the data.

The registry lists secondary outcomes spanning time to discharge or readiness for discharge, improvement on a clinical-status scale, clinical failure, mortality by Day 28, clinical status at Day 28, and adverse events. Together, these outcomes add perspectives on recovery, deterioration, survival, and safety that differ from the primary combined endpoint. The registry includes posted results for some secondary outcomes, but detailed analysis information is not reported in the registry record for each one.

Secondary outcomes can help place the primary result in context, but they do not replace it. Different endpoints measure different aspects of participants’ experience and may use different scales or time-to-event definitions. A secondary result that appears to point in a different direction is not necessarily inconsistent with the primary result, because it may measure a distinct outcome or use a different analysis. Conversely, several secondary measures do not independently confirm the primary comparison unless their analyses and interpretation support that conclusion.

Unadjusted secondary results generally carry less inferential weight when many outcomes are examined. General statistical explanation: testing multiple outcomes creates more opportunities for apparently notable results to occur by chance. A multiplicity procedure can address that issue, but the registry record does not report whether a procedure was planned or applied. This means the secondary findings should be treated as supportive or exploratory descriptions rather than as a collection of independent confirmatory conclusions.

The overall pattern is best assessed by considering what each endpoint measures, how much uncertainty surrounds its estimate, and whether the analysis was prespecified and adjusted for multiple testing. The registry record does not report enough detail to establish those features for every secondary analysis. Accordingly, secondary results provide additional descriptive context, while the primary endpoint remains the main registered efficacy question.

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

• Time to Hospital Discharge or "Ready for Discharge" (as Evidenced by Normal Body Temperature and Respiratory Rate, and Stable Oxygen Saturation on Ambient Air or >/= 2 Liters (L) Supplemental Oxygen)
• Time to Improvement of at Least 2 Categories Relative to Baseline on a 7-Category Ordinal Scale of Clinical Status
• Time to Clinical Failure, Defined as the Time to Death, Mechanical Ventilation, ICU Admission, or Withdrawal (Whichever Occurred First)
• Mortality Rate by Day 28
• Clinical Status on 7-Category Ordinal Scale at Day 28
• Percentage of Participants With Adverse Events

9. Statistical Interpretation

10. Statistical Methods Explained

What does a risk difference describe?

A risk difference compares the observed event proportions between two groups by subtracting one proportion from the other, using a stated direction for the comparison. In the posted primary comparison, the convention is tocilizumab relative to placebo. A negative estimate therefore corresponds to a lower observed event proportion in the tocilizumab group. The risk difference is expressed on the proportion scale, whereas the arm results are presented as percentages. These are two ways of describing the same broad contrast, but they answer different display needs: the arm percentages show each group’s result, and the risk difference directly expresses the between-group contrast.

The estimate is a summary of the observed trial data, not a guarantee that the same contrast would be observed in another study. Randomization supports a comparison of the assigned groups, but sampling variability remains. The confidence interval helps show the uncertainty associated with that variability.

How should the confidence interval be read?

A confidence interval gives a range of values that are compatible with the observed data under the method used to calculate it. It is not a range containing every individual participant’s outcome, and it is not a statement that the true value has a particular probability of lying in that specific interval. Its width conveys precision: a narrower interval indicates greater precision, while a wider interval indicates more uncertainty about the contrast.

For this primary comparison, the 95% confidence interval for the risk difference extends from -0.154 to 0.005. Because the interval crosses zero, it includes no difference as one value compatible with the estimate and its uncertainty. The interval also contains negative and positive values, so the result does not precisely distinguish between all contrasts represented by those values. The confidence interval should be interpreted together with the point estimate, endpoint definition, and study population rather than as a stand-alone verdict.

What does the p-value add?

The unadjusted p-value is 0.051. A p-value describes how unusual data at least as incompatible with a specified null hypothesis would be if that null hypothesis and the test’s assumptions held. It is not the probability that the null hypothesis is true, and it does not measure the size or clinical importance of an effect. The term “unadjusted” matters because the registry record does not report a multiplicity adjustment for this result. A threshold-based reading should not replace consideration of the effect estimate and confidence interval.

The numerical proximity of a p-value to a conventional cutoff does not create a sharp distinction between evidence and no evidence. Results just on either side of a threshold can have very similar uncertainty. Here, the confidence interval and p-value should be considered together: both indicate that the primary comparison remains uncertain rather than establishing a definitive direction.

Why can over-dispersion matter in other analyses?

General statistical explanation: over-dispersion occurs when outcome variability is greater than a model assumes. If a model underestimates variability, its standard errors can be too small and its confidence intervals too narrow, which can make evidence appear more precise than it is. The primary outcome posted for this trial is a binary percentage, and the registry record does not describe an over-dispersion adjustment for it. This concept is relevant to statistical modeling generally, but should not be assumed to have affected this trial’s analysis.

What does the analysis population specify?

The primary population is defined as randomized participants who received study treatment, analyzed according to the treatment assigned at randomization. This definition identifies the participants included in the primary results and preserves the original assignment grouping among those participants. It is not identical to a statement that every randomized participant contributed an outcome: the registry’s definition explicitly includes receipt of study treatment. The record does not report how missing endpoint observations were handled or whether additional analysis populations were used.

Different population definitions can lead to different interpretations because they determine which participants’ outcomes contribute to an estimate. Without details about missing data, exclusions, or sensitivity analyses, readers cannot assess how robust the result is to alternative handling of incomplete outcome information. Those methods are not reported in the registry record.

11. Confidence Intervals and Precision

The primary risk difference is -0.075, with a 95% confidence interval from -0.154 to 0.005. This interval is the main guide to the precision of the group contrast. It spans values below and above zero, so the posted data do not exclude no difference or every small positive contrast. The point estimate alone would conceal that uncertainty.

The confidence intervals for the group percentages also show uncertainty around each arm’s estimated cumulative proportion: 8.52% to 16.86% for tocilizumab and 13.34% to 27.36% for placebo. These intervals are not the same as the confidence interval for the risk difference. Each arm interval concerns an individual group estimate, whereas the risk-difference interval concerns the contrast between groups. The comparison interval is therefore the direct measure to use when considering uncertainty in the between-group result.

Confidence intervals are affected by the amount of information in the data and by the variability of the outcome. General statistical explanation: greater precision usually appears as a narrower interval, while less precision appears as a wider interval. A confidence interval is not a guarantee that the true value lies within its endpoints, and it should not be interpreted as a range of individual outcomes.

The posted interval does not establish whether a difference would be important in practice. The registry record does not report a threshold for an important difference on this endpoint. Precision and importance are related but distinct questions: the interval describes statistical uncertainty, while judgments about importance require a meaningful threshold and relevant context. The estimates should therefore be read as an uncertain comparison of the observed Day 28 proportions, not as a precise determination of what would happen in every setting.

12. Planning a Similar Trial: Sample Size

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

The planning illustrations describe participants per arm for 80% power and participants per arm for 90% power. They are not totals across both groups and are not observed group sizes. The displayed scenarios use 100%, 75%, and 50% of the observed effect. For the primary endpoint, the participants per arm for 80% power are 372, 693, and 1,630 across those effect scenarios; for 90% power, they are 497, 928, and 2,181.

The stated assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, a normal approximation, and an effect to detect defined as the stated fraction of the observed effect. The illustrations are computed from the observed effect and are not the trial’s original design assumptions; those original assumptions are not reported in the registry record. The note about total events applies to time-to-event planning displays; this primary endpoint display gives participants per arm.

In plain terms, 80% power means that, under the planning assumptions and if the effect used for planning were the true effect, the design would have an 80% chance of meeting its specified statistical criterion. The 90% column describes the corresponding 90% chance. These are planning probabilities, not certainty about any particular result. Smaller effects require more participants because they are harder to distinguish from random variation. This is a planning illustration, not observed power.

13. How the Sample Size Is Derived

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

Formula

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 = 0.195; treatment proportion p₁ = 0.120; average p̄ = 0.158.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.5157 = 1.0107.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.5130 = 0.4317.
4. n = (1.0107 + 0.4317)² / (-0.0748)² = 371.6, rounded up to 372 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 <- 0.1953125                          # control proportion
p1 <- 0.120481927710843                          # 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(4372186)
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 = 372 (80% power), 497 (90% power)
 75% of observed effect: n per arm = 693 (80% power), 928 (90% power)
 50% of observed effect: n per arm = 1,630 (80% power), 2,181 (90% power)
power.prop.test: 371.6 
Simulated power at n = 372 per arm: 0.806 (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.

14. Multiplicity, Missing Data and Analysis Populations

Multiplicity arises when several outcomes, time points, or statistical comparisons are examined. General statistical explanation: as the number of tests increases, the chance of at least one apparently unusual result can increase even when none of the underlying effects is real. Statistical plans sometimes address this with a prespecified testing sequence or an adjustment to the decision threshold. The registry record does not report a multiplicity strategy for the primary and secondary outcomes, so it is not possible to determine from the record whether any secondary findings were adjusted or intended to support confirmatory claims.

The record lists multiple secondary outcomes and reports an unadjusted p-value for the primary comparison. The term unadjusted describes the posted p-value; it does not establish whether other aspects of the analysis were adjusted. The statistical analysis plan, including any hierarchy among outcomes, is not reported in the registry record. The number or timing of interim analyses is also not reported in the registry record.

The primary analysis population is the mITT population: randomized participants who received study treatment, grouped according to treatment assigned at randomization. That definition is reported, but the registry record does not describe the handling of missing primary endpoint data, reasons for missingness, imputation methods, or sensitivity analyses. Missing-data assumptions can affect the interpretation of an endpoint estimate, especially when the outcome is not available for every participant.

The registry record does not report a sample-size justification or the trial’s original design assumptions. The posted planning illustration is based on the observed effect and stated assumptions, and should be kept distinct from a prospective design calculation. These omissions do not invalidate the posted results, but they limit the ability to reconstruct the full inferential framework or judge how planned analyses relate to the posted estimates.

15. 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: covid-19 pneumonia (2.4%), pneumonia (2.4%), septic shock (2.4%), acute kidney injury (2.4%), acute respiratory failure (2.4%).

The serious adverse event summary gives the number at risk and the number affected by serious adverse events in each group. In the tocilizumab group, 250 participants were at risk and 38 were affected; in the placebo group, 127 were at risk and 25 were affected. These are counts reported by the registry, not estimates of the primary endpoint. The record also lists deaths separately: 29 in the tocilizumab group and 15 in the placebo group. The registry record does not provide a causal interpretation for these counts.

The most frequently listed serious adverse event terms include COVID-19 pneumonia, pneumonia, septic shock, acute kidney injury, and acute respiratory failure, each reported as 2.4%; acute respiratory distress syndrome is reported as 2.0%. These are terms in the serious adverse event summary. The figures should be interpreted within the recorded study groups and the registry’s adverse-event definitions; they do not, on their own, show whether an event was caused by the assigned intervention.

Safety counts are descriptive and need to be considered alongside how events were collected, their timing, severity, and relationship assessment. Those details are not fully reported in the registry record. The counts should not be converted into a conclusion about comparative safety without that context.

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

The posted primary result describes the proportion of participants in each assigned group who died or required mechanical ventilation by Day 28. The tocilizumab group had a lower observed percentage than the placebo group, and the risk-difference estimate was negative. The 95% confidence interval includes zero, however, and the unadjusted p-value was 0.051. Together, these statistics indicate uncertainty about the magnitude and direction of the underlying difference. They do not establish a definitive benefit or absence of benefit.

The result applies to the study’s registered population and treatment comparison: hospitalized participants with COVID-19 pneumonia, assigned to tocilizumab plus standard of care or placebo plus standard of care, with the primary analysis conducted in the defined mITT population. It does not automatically answer what would occur in populations, settings, or treatment strategies that differ from those described in the registry record.

The combined endpoint does not show how many events arose from each component separately. It also does not describe the timing of events, longer-term outcomes, or the experience of participants who did not meet the endpoint. Secondary endpoints may add context, but their interpretation depends on their own definitions and analysis details, and the registry record does not report a multiplicity plan.

The posted safety counts and event terms are also descriptive. They should not be treated as proof that a particular event was caused by treatment or as a complete account of comparative harms. A full interpretation would depend on event ascertainment, timing, severity, and analysis methods, some of which are not reported in the registry record.

Statistically, the most defensible reading is narrow: the registry reports a lower observed primary endpoint percentage in the tocilizumab group, but the estimate remains imprecise enough to include no difference. The study’s results should be understood as estimates with uncertainty, not as a guarantee of outcomes for any individual participant.

17. Why This Trial Matters Statistically

This trial illustrates why a point estimate, confidence interval, and p-value should be read together. The primary estimate points toward a lower cumulative proportion in the tocilizumab group, while the 95% confidence interval crosses zero and the unadjusted p-value is 0.051. A single threshold-based label would obscure the range of contrasts compatible with the data. The interval makes the remaining uncertainty visible.

It also shows the importance of keeping endpoint definitions distinct. The primary outcome combines death and mechanical ventilation by Day 28. The registry separately lists mortality and other clinical outcomes, but those measures address different questions. A result for one endpoint cannot be substituted for another, and an aggregate result cannot reveal the separate contribution of its components without component-specific estimates.

The record highlights the value of transparent analysis details. The primary population is defined, and the trial’s randomized, double-blind design is recorded. But details about multiplicity, missing data, interim analyses, and original sample-size assumptions are not reported in the registry record. Such information helps readers understand how estimates were selected, how incomplete observations were treated, and how multiple results were intended to be interpreted.

Finally, sample-size planning illustrations based on an observed effect are not the same as a prospective design rationale or a measure of the trial’s observed power. Their assumptions and purpose need to be stated clearly. This distinction helps prevent retrospective calculations from being mistaken for evidence that a study was adequately or inadequately powered under its original plan.

18. 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 primary estimate has a confidence interval that includes zero, leaving uncertainty about the direction and magnitude of the group contrast. The registry record does not report separate primary estimates for death and mechanical ventilation, so the combined result cannot show how each component contributed to the observed pattern.

The primary analysis is defined using the mITT population of randomized participants who received study treatment. The registry record does not report how missing endpoint information was handled, why any observations might be missing, or whether sensitivity analyses examined alternative assumptions. This limits assessment of how incomplete outcome data could affect the estimate.

Although the registry lists several secondary outcomes, a multiplicity strategy is not reported in the registry record. The status of secondary analyses as confirmatory or exploratory, and any adjustment for examining multiple outcomes, therefore cannot be established from the record. The registry also does not report the original sample-size justification, detailed analysis plan, or interim-analysis information.

The serious adverse event summary gives group counts and selected event terms, but the registry record does not provide all details needed to interpret event timing, severity, or causal attribution. These figures alone do not establish that an event was caused by treatment. The optional long-term extension is described, but long-term results are not included among the posted results.

These limitations concern the amount of methodological and outcome detail available in the registry record. They do not change the reported estimates, but they constrain how completely their robustness, clinical meaning, and generalizability can be assessed.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: A Study to Evaluate the Efficacy and Safety of Tocilizumab in Hospitalized Participants With COVID-19 Pneumonia — NCT04372186.
• Registry record last updated 2023-02-10; status last verified 2023-02.
• Results first posted 2021-09-27.
• 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.

22. Record Summary

NCT04372186 is a completed Phase 3 randomized, double-blind, parallel-group study of tocilizumab plus standard of care compared with placebo plus standard of care in hospitalized participants with COVID-19 pneumonia. The recorded enrollment is 377. The primary endpoint is the cumulative proportion who died or required mechanical ventilation by Day 28, analyzed in the defined mITT population.

The posted primary percentages are 12.04% for tocilizumab and 19.26% for placebo. The risk difference is -0.075, with a 95% confidence interval from -0.154 to 0.005 and an unadjusted p-value of 0.051. The interval includes zero, so the result remains uncertain. The registry also records secondary outcomes and serious adverse event summaries, while several details needed to assess the full analysis plan are not reported in the registry record.