This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial facts on this page are restricted to the ClinicalTrials.gov record; the statistical explanations distinguish reported methods from general interpretation.
1. Trial at a Glance
PROMIS-I was a completed randomized, parallel, quadruple-masked phase 3 trial evaluating CMS versus placebo in non-cystic fibrosis bronchiectasis. The registry identifies prevention as the primary purpose and reports a 12-month primary endpoint measuring mean annual pulmonary exacerbation rate.
| Feature | PROMIS-I |
|---|---|
| Trial name | PROMIS-I |
| Brief title | Efficacy and Safety of Inhaled CMS in Bronchiectasis Subjects With Chronic P. Aeruginosa Infection. (PROMIS-I) |
| Phase | Phase 3 |
| Status | Completed |
| Therapeutic area | Pulmonology |
| Condition | Non Cystic Fibrosis Bronchiectasis |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Prevention |
| Enrollment | 377 |
| Interventions | CMS (drug); Placebo (other) |
| Lead sponsor | Zambon SpA |
| Sponsor type | Industry |
| ClinicalTrials.gov | NCT03093974 |
2. Clinical Question
The primary statistical question was whether CMS and placebo differed in the mean annual non-cystic fibrosis bronchiectasis pulmonary exacerbation rate over a 12-month time frame. The registered hypothesis type was superiority.
Population
Bronchiectasis subjects with non-cystic fibrosis bronchiectasis and chronic P. aeruginosa infection, as described in the trial's brief title and condition field.
Intervention
CMS (colistimethate sodium).
Comparator
Placebo.
Primary question
Does CMS produce a different mean annual pulmonary exacerbation rate from placebo over 12 months under the prespecified superiority analysis?
3. Trial Design
CMS
- CMS (Colistimethate Sodium)
- Drug intervention
Placebo
- Placebo
- Other intervention
4. Trial Timeline
Trial start
The PROMIS-I trial began on 2017-06-06.
Primary completion
The registry reports primary completion on 2021-04-09.
Results available
The ClinicalTrials.gov record identifies the study as completed and indicate that results and a statistical analysis were posted.
5. Primary Endpoint
| Endpoint | Time frame | Outcome unit | Registered description |
|---|---|---|---|
| Mean Annual Non-cystic Fibrosis Bronchiectasis (NCFB) Pulmonary Exacerbation Rate | 12 months | Number of Pulmonary Exacerbations | The primary efficacy assessment for an individual subject was the frequency of pulmonary exacerbations (exacerbation rate). A pulmonary exacerbation was defined as the presence concurrently of at least three of the following eight symptoms/signs for at least 24 hours: increased cough; increased sputum volume and/or consistency; increased sputum purulence; new or increased haemoptysis; and additional signs or symptoms in the registered definition. |
The registry therefore frames the primary outcome as a rate, rather than simply asking whether a subject experienced at least one exacerbation. That distinction is statistically important: a rate can incorporate the frequency of recurrent events over a period of observation.
6. Results: Mean Annual Pulmonary Exacerbation Rate
The registry contains one posted formal statistical analysis for the primary endpoint. The analysis compares CMS with placebo using a two-sided Wald chi-square test and reports an LS Mean rate ratio of 0.612 with a 95% two-sided confidence interval of 0.457–0.820 and a p-value of 0.00101.
LS Mean rate ratio
95% CI: 0.457–0.820 · P = 0.00101
CMS vs placebo · 12-month primary endpoint · Superiority hypothesis
The estimated rate ratio of 0.612 means that the model-based estimated exacerbation rate for CMS was 0.612 times the corresponding rate for placebo. Expressed as a simple relative interpretation, 0.612 corresponds to an estimated rate about 38.8% lower than the placebo rate.
This does not mean that 38.8% of subjects avoided an exacerbation, that each individual subject experienced exactly 38.8% fewer events, or that the absolute difference in the number of exacerbations was 38.8%. A rate ratio is a relative comparison of event rates.
The 95% confidence interval of 0.457–0.820 describes statistical uncertainty around the reported rate-ratio estimate under the analysis framework. It does not describe the range of effects experienced by individual subjects. Because the interval is below 1, every value in the reported interval corresponds to a lower estimated rate for CMS relative to placebo.
The p-value of 0.00101 addresses the statistical evidence against the null comparison used in the two-sided Wald chi-square test. It does not measure the magnitude of the treatment effect, the probability that CMS works, or the clinical importance of the observed rate ratio.
The registry reports a superiority hypothesis. The analysis also used a negative binomial model with treatment, pooled site (country), and baseline use of stable concomitant therapy with oral macrolides as fixed effects and log-time on treatment as an offset. Interpretation therefore belongs to that model-based framework rather than to an unadjusted comparison of raw event totals.
7. The Statistical Model Behind the Primary Result
The registry's analysis text states that the number of NCFB pulmonary exacerbations was compared between treatment groups using a negative binomial model. The fixed effects were treatment, pooled site (country), and baseline use of stable concomitant therapy with oral macrolides. The model used log-time on treatment as an offset.
The offset allows event counts to be interpreted in relation to the amount of time each subject contributes to observation. The registry analysis specifically identifies log-time on treatment as the offset.
Why a negative binomial model?
Recurrent pulmonary exacerbations are naturally represented as event counts over an observation period. A negative binomial model is commonly used for count outcomes when variability between subjects can exceed what a simple Poisson model would accommodate. The important point for this trial is not to substitute a different model: the registry explicitly reports a negative binomial model for the primary comparison.
Why include an offset?
Subjects may contribute different amounts of time on treatment. Modeling the logarithm of time as an offset allows the expected event count to be related to exposure time, which is appropriate when the estimand is an event rate rather than a simple event count.
Why adjust for pooled site and baseline macrolide use?
The reported model includes pooled site (country) and baseline use of stable concomitant therapy with oral macrolides as fixed effects. Including these prespecified model terms can account for variation associated with those factors while estimating the treatment comparison within the model.
8. Statistical Methodology
| Component | Reported method or role | Statistical interpretation |
|---|---|---|
| Primary outcome | Mean annual pulmonary exacerbation rate | A recurrent-event rate measured over 12 months |
| Primary formal test | Two-sided Wald chi-square test | Tests the treatment comparison within the reported model framework |
| Model | Negative binomial model | Models the recurrent event-count outcome while allowing rate-based inference |
| Effect measure | LS Mean rate ratio | Relative model-based rate comparison between CMS and placebo |
| Offset | Log-time on treatment | Accounts for the amount of treatment exposure time in the rate model |
| Fixed effects | Treatment, pooled site (country), baseline stable concomitant oral macrolide use | Adjustment variables specified in the reported negative binomial model |
| Analysis population | Modified ITT population | All subjects who provided informed consent, were randomized, and received at least 1 dose or partial dose of the IMP |
Modified intention-to-treat analysis population
The registry-reported analysis defines the Modified ITT (mITT) Population as all subjects who provided informed consent, were randomized, and received at least 1 dose or partial dose of the IMP. This differs from a definition that would include every randomized subject regardless of treatment exposure.
The distinction matters because the population actually analyzed determines the population to which the reported estimate directly applies. Randomization establishes the initial treatment comparison, while the mITT criterion adds a treatment-exposure requirement.
Wald chi-square testing
The formal test was a two-sided Wald chi-square test. In a model-based analysis, a Wald test evaluates whether the estimated treatment parameter differs from the null value relative to its estimated uncertainty. For a rate ratio, the natural null value is 1 rather than 0.
A rate ratio of 1 represents equal modeled rates. Values below 1 indicate a lower estimated rate in the numerator group; values above 1 indicate a higher estimated rate.
9. Analysis Population and Estimand Considerations
The primary analysis was performed in the registry-defined mITT population. This is an important detail when interpreting the estimate: the reported rate ratio is not simply an arithmetic comparison of all 377 enrolled subjects.
What the mITT definition includes
Subjects who provided informed consent, were randomized, and received at least 1 dose or partial dose of the IMP.
Why this matters
The analysis population determines which randomized subjects contribute to the formal treatment comparison.
Exposure time
The negative binomial analysis includes log-time on treatment as an offset, directly connecting the count outcome to treatment exposure time.
What is not reported
The ClinicalTrials.gov record do not give the mITT sample size by treatment arm, so this page does not derive or infer those numbers.
10. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment arm in the safety population. These are presented as affected subjects divided by subjects at risk.
| Safety population | Serious adverse events | Affected / at risk |
|---|---|---|
| CMS (Colistimethate Sodium) (SAF) | Serious adverse events | 31 / 176 |
| Placebo (SAF) | Serious adverse events | 46 / 197 |
The reported serious-adverse-event information is descriptive here. The statistical analyses posted on ClinicalTrials.gov do not provide a formal hypothesis test, confidence interval, or effect estimate for the serious-adverse-event comparison, so this page does not create one.
11. Statistical Methods Explained
Why was a negative binomial model used?
The primary endpoint is a number of pulmonary exacerbations measured as a rate over time. A negative binomial model is a count-data model that can accommodate greater variability in event counts than a basic Poisson model. Most importantly, this is the model explicitly reported for the PROMIS-I primary analysis.
What does a rate ratio of 0.612 mean?
A rate ratio compares the modeled event rate in the CMS group with the modeled event rate in the placebo group. A value of 0.612 means the estimated CMS rate was 0.612 times the placebo rate under the fitted model. It is not an individual-patient probability and is not an absolute reduction in the number of exacerbations.
Why is the null value 1 rather than 0?
Because the effect measure is a ratio. Equal rates produce a ratio of 1. A ratio below 1 indicates a lower rate in the CMS group, while a ratio above 1 indicates a higher rate.
Why use a log-time-on-treatment offset?
Counts are influenced by how long a subject is observed. The logarithm of time on treatment enters the model as an offset so that the analysis targets an event rate rather than treating different amounts of observation time as equivalent.
What does the 95% confidence interval of 0.457–0.820 tell us?
It describes uncertainty around the reported rate-ratio estimate under the model and statistical framework. The interval is entirely below 1, so its reported values all correspond to lower modeled exacerbation rates for CMS relative to placebo. It does not describe individual treatment responses.
What does the p-value of 0.00101 tell us?
The p-value quantifies the evidence against the null treatment comparison under the two-sided Wald chi-square test. It is not a measure of effect size. The size of the estimated effect is conveyed by the rate ratio, while the confidence interval communicates uncertainty around that estimate.
How should the mITT population be understood?
The registry defines the mITT population as subjects who provided informed consent, were randomized, and received at least 1 dose or partial dose of the IMP. This means the primary result applies directly to that analysis population rather than automatically to every enrolled subject.
12. Confidence Interval, Effect Size, and P-value
The PROMIS-I primary result illustrates why these three statistical quantities should be read together rather than interchangeably.
Effect estimate
Rate ratio 0.612 describes the direction and relative magnitude of the modeled treatment comparison.
Precision
95% CI 0.457–0.820 communicates uncertainty around the estimated rate ratio.
Statistical evidence
P = 0.00101 addresses the null hypothesis through the reported two-sided Wald chi-square test.
Clinical meaning
Clinical importance cannot be read from the p-value alone. The rate ratio must be considered alongside the endpoint definition, observation period, analysis population, and safety information.
The point estimate is below 1, the entire reported 95% confidence interval is below 1, and the two-sided p-value is 0.00101. Together, these describe a statistically supported lower modeled pulmonary-exacerbation rate for CMS relative to placebo in the reported mITT analysis. They do not establish that every individual subject benefits by the same amount or that the relative rate reduction is equivalent to an absolute reduction in exacerbations.
13. What the Rate Ratio Does — and Does Not — Mean
A rate ratio of 0.612 means that the estimated pulmonary exacerbation rate under the reported model was 0.612 times the corresponding placebo rate.
It does not mean that 0.612 is the probability of an exacerbation, that 61.2% of subjects experienced an event, or that every subject had the same proportional change in exacerbation frequency.
The 95% CI of 0.457–0.820 places an uncertainty range around the model-based treatment estimate. The interval is about the statistical estimate, not about the distribution of individual patient responses.
The p-value of 0.00101 measures evidence against the null comparison under the specified test. It does not tell us whether the treatment effect is large or small. That question is addressed by the rate ratio and its confidence interval.
14. Why the Analysis Uses a Rate Rather Than a Simple Proportion
A binary endpoint might classify each subject as having experienced an event or not. PROMIS-I instead uses the frequency of pulmonary exacerbations over a 12-month period. This distinction allows recurrent events to contribute to the endpoint rather than collapsing the experience of a subject into a single yes/no indicator.
The statistical consequence is substantial. A count-based rate analysis must account for the number of events and the amount of time contributing to the rate. The reported negative binomial model and log-time-on-treatment offset directly address this structure.
PROMIS-I's primary endpoint is explicitly a pulmonary exacerbation rate, so interpreting its rate ratio as though it were a binary risk ratio would change the meaning of the result.
15. Randomization and Blinding
The trial was randomized, parallel, and quadruple masked. Randomization is important because it provides the structural basis for comparing treatment groups while reducing systematic differences in treatment assignment that could otherwise confound the comparison.
Masking addresses a different problem. By keeping treatment assignment concealed from specified trial participants or personnel, masking can reduce the opportunity for knowledge of treatment assignment to influence behavior, assessment, reporting, or other aspects of trial conduct. The ClinicalTrials.gov record identifies the study as quadruple masked but do not specify the four masked roles.
Randomization
Controls treatment assignment through the trial's allocation process and supports causal comparison under the randomized design.
Masking
Reduces the possibility that knowledge of treatment assignment influences relevant aspects of trial conduct or assessment.
16. Multiplicity, Interim Analysis, and Other Design Features
The ClinicalTrials.gov record does not report a non-inferiority margin, crossover design, factorial design, multiplicity adjustment, or interim-analysis procedure. Because these features materially affect statistical interpretation when they are present, this page does not infer them from the existence of a phase 3 randomized trial.
| Design topic | Information in the ClinicalTrials.gov record | Interpretation |
|---|---|---|
| Non-inferiority margin | Not reported | The registered hypothesis is superiority, not non-inferiority. |
| Crossover | Not reported | No crossover analysis is described in the ClinicalTrials.gov record. |
| Factorial design | Not reported | The design model is parallel with 2 arms. |
| Multiplicity | Not reported | No multiplicity adjustment is specified in the ClinicalTrials.gov record. |
| Interim analysis | Not reported | No interim-analysis procedure is reported. |
| Bayesian methods | Not reported | The registry-reported analysis is frequentist, using a Wald chi-square test and confidence interval. |
| Missing-data/imputation method | Not reported | No imputation procedure is posted on ClinicalTrials.gov for the primary endpoint. |
| Stratification | Not reported | No randomization strata are specified in the ClinicalTrials.gov record. |
17. Why This Trial Matters Statistically
PROMIS-I is a useful teaching example because its primary endpoint is not a conventional time-to-first-event outcome. Instead, it asks how frequently pulmonary exacerbations occur over a defined period and analyzes that recurrent-event count using a negative binomial model with an exposure-time offset.
| Concept | How it appears in PROMIS-I |
|---|---|
| Randomization | The trial uses randomized allocation to CMS or placebo. |
| Parallel-group design | The registry identifies the design model as parallel. |
| Blinding | The trial is quadruple masked. |
| Modified ITT analysis | The primary analysis uses a registry-defined mITT population based on consent, randomization, and receipt of at least 1 dose or partial dose of the IMP. |
| Count outcomes | The primary endpoint measures the number of pulmonary exacerbations. |
| Rate modeling | The primary effect measure is a rate ratio. |
| Negative binomial regression | The reported model compares the number of exacerbations between treatment groups. |
| Offset | Log-time on treatment is included as an offset. |
| Covariate adjustment | Treatment, pooled site (country), and baseline stable concomitant oral macrolide use are fixed effects. |
| Wald testing | The formal comparison uses a two-sided Wald chi-square test. |
| Confidence intervals | The rate-ratio estimate is accompanied by a 95% two-sided confidence interval. |
| Safety analysis | Serious adverse events are reported separately for the CMS and placebo safety populations. |
18. Limitations
- Registry-level detail: the ClinicalTrials.gov record contains one formal primary statistical analysis but do not provide a complete statistical analysis plan.
- mITT population size by arm: the definition is provided, but the ClinicalTrials.gov record does not state the resulting mITT sample size by treatment group.
- No arm-specific primary event counts: the ClinicalTrials.gov record provides the rate-ratio estimate but not the underlying pulmonary exacerbation counts or modeled rates by arm.
- Model-based estimate: the rate ratio is an adjusted model-based effect measure rather than a simple arithmetic ratio of two raw event totals.
- Exposure time: the model uses log-time on treatment as an offset, so the result should be interpreted as a rate comparison rather than a comparison of raw event counts.
- Safety scope: the ClinicalTrials.gov record is limited to serious adverse events by arm; no formal statistical comparison is provided for this endpoint in the ClinicalTrials.gov record.
- Unreported design features: the ClinicalTrials.gov record does not specify multiplicity, interim monitoring, imputation, stratification, crossover, or a non-inferiority margin.
- Analysis population: because the primary analysis uses the registry-defined mITT population, the result should not automatically be described as an analysis of every enrolled subject.
19. Primary Result in Statistical Context
What was estimated?
An LS Mean rate ratio comparing the pulmonary exacerbation rate under CMS with the rate under placebo.
How was it estimated?
Using the reported negative binomial model with treatment, pooled site (country), and baseline stable concomitant oral macrolide use as fixed effects and log-time on treatment as an offset.
How was it tested?
With a two-sided Wald chi-square test under a superiority hypothesis.
How precise was it?
The reported 95% two-sided confidence interval was 0.457–0.820.
The PROMIS-I primary analysis reports an LS Mean rate ratio of 0.612 for CMS versus placebo, with a 95% CI of 0.457–0.820 and a two-sided p-value of 0.00101. The direction of the estimate is toward a lower modeled exacerbation rate with CMS. The inferential result comes from the specified negative binomial model and two-sided Wald chi-square test in the mITT population, rather than from an unadjusted comparison of raw event counts.
20. Related Tutorials
Learn more about the methods and statistical concepts used in this trial:
21. Related Statistical Calculators
22. Sources
- ClinicalTrials.gov: PROMIS-I, NCT03093974.
Continue through Clinical Biostats
Use the trial's statistical concepts as a starting point for deeper tutorials, calculators, and clinical-trial methodology resources.
23. Record Summary
PROMIS-I provides a useful example of how a randomized phase 3 trial can analyze a recurrent clinical event using a rate-based statistical model. The primary endpoint was the mean annual non-cystic fibrosis bronchiectasis pulmonary exacerbation rate over 12 months. The reported analysis used a negative binomial model with treatment, pooled site (country), and baseline stable concomitant oral macrolide use as fixed effects and log-time on treatment as an offset. The resulting LS Mean rate ratio for CMS versus placebo was 0.612, with a 95% two-sided confidence interval of 0.457–0.820 and a two-sided p-value of 0.00101.
The statistical interpretation depends on understanding the endpoint as a recurrent-event rate, the mITT analysis population, the model-based nature of the rate ratio, and the role of exposure time in the analysis. The available safety data separately report serious adverse events as 31/176 in the CMS safety population and 46/197 in the placebo safety population. Together, these elements illustrate why clinical-trial interpretation requires more than reading a p-value: the endpoint, estimand, analysis population, model, effect measure, uncertainty, and safety population all contribute to the statistical story.