Antibiotic "Dysbiosis" in Preterm Infants: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Antibiotic "Dysbiosis" in Preterm Infants (NCT02784821), a randomized phase 2 study. The primary outcome is Number of Events of Composite Morbidities and Mortality, Including Necrotizing Enterocolitis (NEC), Late Onset Sepsis (LOS), Bronchopulmonary Dysplasia (BPD) and Death.

2. Clinical Question

The study asks how antibiotic exposure in preterm neonates relates to the developing microbiome and to major clinical outcomes. Its stated rationale is to evaluate risks and benefits of antibiotic use while considering infection and possible effects on the microbiome and subsequent morbidities. The posted results include counts for four groups, but do not report an inferential primary comparison between randomized groups. Therefore, the results support a descriptive reading of the recorded counts, not a definitive estimate of the primary treatment effect.

3. Trial Design

This was a Phase 2, randomized, unmasked, parallel-design study sponsored by the University of Florida. The registry lists enrollment as 98 and study status as completed. The record gives a start date of 2017-01-16 and a primary completion date of 2019-09-11. It describes randomization of preterm infants born at less than 33 weeks gestation to pre-emptive antibiotics or no pre-emptive antibiotics.

The posted results distinguish four groups. Group A had antibiotics indicated, and Group B had antibiotics not indicated. The other two groups are Group CI, randomized to antibiotics, and Group CII, randomized to no antibiotics. The group descriptions indicate that the first pair reflects clinical indication categories, while the latter pair is explicitly randomized. This distinction matters: the group A versus group B contrast is not described as randomized, whereas CI versus CII is the randomized comparison. The record does not report whether the indication groups were part of the same randomization process; their descriptions characterize them separately.

The listed interventions include antibiotics and sample collection or sample types including gastric fluid, breast milk and stool. The study aimed to analyze the microbiome, while the primary endpoint concerns a clinical composite. The registry lists several neonatal conditions and secondary clinical outcomes, including late-onset sepsis, bronchopulmonary dysplasia, necrotizing enterocolitis, deaths and length of stay.

The registry record does not report the allocation ratio, the details of sequence generation, concealment procedures, or a masking procedure beyond NONE. It also does not report a sample-size justification or the full statistical analysis plan. Posted results give group counts, but do not provide a reported primary hypothesis test or adjusted analysis.

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.

Before assignment: 98 infants were enrolled in the study and were included in the arms of the study to receive intervention.

Overall Study

Reasons for not completing this period:

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: x2 test or ANOVA

6. Primary Endpoint

The registry lists the primary outcome as Number of Events of Composite Morbidities and Mortality, Including Necrotizing Enterocolitis (NEC), Late Onset Sepsis (LOS), Bronchopulmonary Dysplasia (BPD) and Death, measured over Until discharge from the NICU, up to 1 year.

The primary endpoint is described as the number of events in a composite of morbidity and mortality, including necrotizing enterocolitis, late-onset sepsis, bronchopulmonary dysplasia and death. Its stated observation period is until discharge from the NICU, up to 1 year. The record says medical records would be reviewed to determine the number of patients with the composite outcome and the association between antibiotic administration and its components.

Composite endpoints combine several clinically relevant outcomes into one measure. In general, this can capture a broader range of events than any single component and can be useful when the study question concerns overall morbidity and mortality. Interpretation depends on knowing how the composite is defined, whether a participant can contribute more than one event, and whether the components are counted as events or as patients with at least one component. Those details are not fully described in the registry record.

There is a wording distinction in this record: the primary outcome title refers to a number of events, while its description refers to the number of patients with the composite outcome. The posted group results give counts, but the registry record does not supply event-level data or a separate definition clarifying whether those counts are unique patients, total events, or both. The group values can therefore be reported descriptively, but their exact unit should be interpreted with this documentation limitation in mind. No primary comparative estimate or test is reported.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

Several statistical ideas help distinguish what these results can establish from what they cannot. The registry posts counts by group for the primary composite and secondary outcomes. A count is descriptive: it records how many outcomes were posted for a group, but does not by itself quantify uncertainty or demonstrate that differences between groups are attributable to treatment.

What does a risk difference express?

For binary outcomes, a risk difference compares the proportion with the outcome in one group with the proportion in another. A negative value in the randomized comparisons means the observed proportion was lower in the named comparison group than in Group A, the listed control for the recalculations. A positive value means it was higher. This is a general explanation of the measure, not a claim that the difference is causal. Causal interpretation depends on the study design, analysis population, outcome definition and conduct of the trial.

The recalculated estimates use posted group counts and are simple, unadjusted comparisons. For example, the late-onset-sepsis randomized comparison has a risk difference of -0.009. The associated confidence interval runs from -0.205 to 0.187. A confidence interval describes a range of values reasonably compatible with the data under the statistical method used. When it spans values on both sides of zero, the data do not precisely distinguish a lower risk from a higher risk at the conventional two-sided significance threshold. This is a general interpretation of confidence intervals; it does not establish that either direction is true.

The bronchopulmonary-dysplasia comparison has a risk difference of -0.268 and a confidence interval from -0.471 to -0.065. Its unadjusted p-value is 0.017. The interval does not include zero under the recalculation, but this remains an unadjusted secondary result among several outcomes. The necrotizing-enterocolitis estimate is 0.040, with a confidence interval from -0.073 to 0.153 and an unadjusted p-value of 0.476. These values illustrate why both the estimated direction and its uncertainty matter; a p-value alone does not convey the range of plausible effects.

Why are some comparisons not available?

The primary measure is posted as a count-type composite, and the registry record does not provide a reported analysis that reduces it to a two-arm comparison. The record also does not provide enough detail to resolve the difference between the endpoint title's number of events and its description's number of patients with the composite. Accordingly, the primary group counts should not be converted into a newly calculated inferential result. The posted recalculations apply to the listed secondary binary outcomes only.

The trial's four groups also require care in comparison. Group CI and Group CII are identified as randomized to antibiotics and no antibiotics. Groups A and B are defined by clinical indication. Randomization, when properly implemented, is intended to balance both measured and unmeasured factors on average; indication-based group membership does not have that protection. A contrast between A and B can therefore reflect differences in underlying clinical circumstances as well as differences in antibiotic use.

What do the planning illustrations mean?

The registry includes planning illustrations for selected secondary comparisons. These are not observed trial power and are not the trial's reported design justification. They use the observed effect, or a stated fraction of it, with a two-sided significance level of 0.05, equal allocation and a normal approximation. For the randomized late-onset-sepsis comparison, planning for 100% of the observed effect lists 29,444 participants per arm for 80% power and 39,417 participants per arm for 90% power. Planning for 75% of the observed effect lists 52,593 participants per arm for 80% power and 70,407 participants per arm for 90% power; planning for 50% lists 118,890 participants per arm for 80% power and 159,160 participants per arm for 90% power.

These illustrations depend on the observed estimate, which can be unstable, especially when event counts are limited. They answer a hypothetical planning question under stated assumptions; they do not say what power the completed trial had. The registry record does not report the study's own sample-size justification or whether the planning assumptions match the original design.

The registry record does not report adjusted models, an analysis population, missing-data methods, an interim-analysis plan or a multiplicity strategy. These omissions limit reproducibility and make it difficult to know whether the posted counts represent all randomized participants or a defined subset. In general, adjusted analyses, prespecified outcome hierarchies and explicit missing-data rules can affect estimates and uncertainty. Without those details, the simplest reading is descriptive for the primary endpoint and exploratory for the recalculated secondary comparisons.

10. Multiplicity, Missing Data and Analysis Populations

The registry record does not report a multiplicity plan, such as a hierarchy of outcomes, a correction for testing multiple outcomes, or a distinction between confirmatory and exploratory analyses. This matters because the results include several secondary outcomes and more than one group contrast. In general, repeated testing can increase the likelihood that at least one result appears statistically notable by chance. The unadjusted p-values should therefore be read as individual exploratory comparisons, not as a family-wise conclusion.

Missing-data handling is not reported in the registry record. The posted results give group counts, but do not explain whether any participants had incomplete outcome information, how missing observations were handled, or whether the analysis used all randomized participants. The record also does not report an intention-to-treat or per-protocol analysis population. These omissions prevent readers from assessing whether exclusions or incomplete follow-up could have influenced the reported counts.

The observation period is stated as until discharge from the NICU, up to 1 year. The registry does not provide further detail on follow-up completeness or how differences in length of stay were accounted for. For the primary endpoint, the title refers to number of events and the description refers to number of patients with the composite; the exact counting rule is not clarified. No primary inferential analysis is reported, and no interim-analysis plan or full statistical analysis plan is reported.

These are reporting limitations, not evidence that a particular method was or was not used. The registry record does not report the information needed to determine the analysis population, the treatment of missing data, or how multiplicity was addressed. As a result, the posted counts and unadjusted recalculations should be interpreted cautiously.

11. 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 0% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: early onset sepsis (3.7%).

The adverse-event record lists serious events, other events and deaths separately for the four groups. Serious events affecting participants are recorded as 0 of 32 in Group A, 0 of 11 in Group B, 0 of 28 in Group CI and 1 of 27 in Group CII. The record also lists early onset sepsis as a serious adverse-event term at 3.7%. These are registry safety entries and should not be interpreted as a complete assessment of all possible harms.

Other adverse events affecting participants are recorded as 6 of 32 in Group A, 0 of 11 in Group B, 5 of 28 in Group CI and 4 of 27 in Group CII. Deaths are listed separately as 5, 0, 4 and 5, respectively. The registry does not explain in this safety listing how deaths relate to the primary composite counts, nor does it provide event-level timing or attribution. These categories should not be combined or treated as interchangeable.

In general, a safety count describes events recorded during the specified reporting period; it does not establish whether an intervention caused an event. The clinical-indication groups also differ from the randomized groups, so their counts are not direct randomized safety comparisons. The registry record does not report a comparative safety analysis or attribution details.

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

The posted results establish that the registry contains group-level counts for a primary composite and several secondary clinical outcomes. They also identify which groups were randomized to antibiotics or no antibiotics and which were classified by clinical indication. This provides a useful descriptive record of the study's reported outcomes and makes clear that the randomized comparison is distinct from the indication-based group comparison.

The results do not establish that antibiotics caused a change in the primary composite. No primary comparative estimate, confidence interval or p-value is reported. The primary outcome's wording also leaves uncertainty about whether the posted values represent patients with at least one composite component or total events. That distinction is important because patients and events are not necessarily the same unit of analysis.

The secondary recalculations can describe differences in posted proportions, but they are unadjusted and not accompanied by a reported multiplicity strategy. Their confidence intervals show varying levels of uncertainty. The bronchopulmonary-dysplasia randomized comparison has an unadjusted interval that excludes zero, but that finding alone is not a definitive conclusion across several outcomes. Other secondary comparisons have intervals spanning zero, indicating that the posted data do not precisely determine a direction of difference for those outcomes.

The planning illustrations are hypothetical sample-size calculations based on observed effects and stated assumptions. They are not observed power and should not be mistaken for evidence that the completed study was adequately or inadequately powered. The registry record does not report the original sample-size rationale.

In general, a randomized comparison can support causal conclusions when the design and analysis preserve the benefits of random assignment. The registry does not report enough detail about analysis populations, missing data or the primary statistical plan to evaluate that fully. The appropriate conclusion is therefore restrained: the posted results are descriptive, some recalculated secondary comparisons are exploratory, and the primary treatment effect remains unquantified in the registry record.

13. Why This Trial Matters Statistically

This study is statistically notable because its registry record combines randomized and indication-defined groups. Those structures answer different questions. The randomized antibiotic versus no-antibiotic groups are the relevant contrast for estimating the effect of assignment, while the clinically indicated and not-indicated groups may differ for reasons that led clinicians to consider antibiotics. Keeping these comparisons separate avoids attributing underlying clinical differences to treatment.

The record also illustrates why endpoint definition matters. A composite can bring several outcomes together, but the primary title describes a number of events while the endpoint description refers to patients with the composite. Without a clear counting rule and a reported comparative analysis, readers cannot tell how the primary values translate into a treatment-effect estimate. A count alone cannot communicate precision or uncertainty.

The secondary results show how much interpretation depends on both estimates and uncertainty. A point estimate is only one part of the evidence; confidence intervals indicate how imprecise that estimate may be. Unadjusted testing across multiple outcomes further limits the weight of isolated findings. The recalculated results are useful for understanding the posted counts, but they are not a substitute for prespecified sponsor analyses.

Finally, planning illustrations derived from observed effects can be informative as hypothetical design scenarios, but they do not report the study's original rationale or its observed power. Separating these planning calculations from the trial's actual design decisions is essential. The registry record's gaps on analysis populations, missing data and multiplicity show why statistical reporting is part of interpreting clinical evidence, not merely a technical appendix.

14. 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 is not reported with an inferential comparison, despite posted counts across the four groups. The outcome title and description also differ in whether they refer to events or patients with a composite. The record does not clarify the counting rule or provide component-level primary data that would resolve this ambiguity.

Second, the groups do not all represent randomized contrasts. Group A and Group B are organized according to antibiotic indication, while Group CI and Group CII are explicitly randomized. Comparisons involving the indication groups can reflect differences in clinical circumstances and should not be interpreted as randomized treatment effects.

Third, the secondary results are recalculated from posted summary counts and are unadjusted. The registry lists no reported analyses, no multiplicity procedure, and no prespecified analysis hierarchy. The analysis population, missing-data handling and any adjustments for baseline characteristics are not reported in the registry record. These gaps limit assessment of robustness and reproducibility.

Fourth, the registry record does not report the sample-size justification, detailed randomization procedures, or a full statistical analysis plan. The planning illustrations are based on observed effects and stated assumptions, not the original trial design. The record also does not describe the completeness of outcome ascertainment through the stated follow-up period.

Finally, group-level counts do not provide individual-level data, timing, or attribution. This limits assessment of recurrent events, relationships among composite components, and safety context. These limitations do not show that the trial was conducted improperly; they mark what cannot be evaluated from the registry record.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Antibiotic "Dysbiosis" in Preterm Infants — NCT02784821.
• Registry record last updated 2024-06-05; status last verified 2024-06.
• Results first posted 2024-06-05.
• 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.

18. Record Summary

NCT02784821 is a completed Phase 2, randomized, unmasked, parallel study sponsored by the University of Florida. The registry lists enrollment as 98 and describes preterm infants born at less than 33 weeks gestation. Its primary outcome is a composite including necrotizing enterocolitis, late-onset sepsis, bronchopulmonary dysplasia and death, assessed until discharge from the NICU, up to 1 year.

Results are posted for four groups: two defined by whether antibiotics were indicated and two described as randomized to antibiotics or no antibiotics. The registry provides primary group counts but no primary comparative analysis. Secondary outcomes include late-onset sepsis, bronchopulmonary dysplasia and necrotizing enterocolitis. Simple unadjusted recalculations are available for some secondary contrasts, but no multiplicity plan or reported analysis is listed. Important details about the primary outcome's counting rule, analysis population and missing-data methods are not reported in the registry record.