Personalised Risk Assessment in Febrile Illness to Optimise Real-life Management Across the European Union (PERFORM): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Personalised Risk Assessment in Febrile Illness to Optimise Real-life Management Across the European Union (PERFORM) (NCT03502993), a registered phase not stated study. The primary outcome is Number of Participants With Clinically-assigned Retrospective Phenotype.

2. Clinical Question

3. Trial Design

PERFORM is a prospective, hospital-based study of febrile children, with recruitment of controls as well. Its stated aim is to develop and validate ways of distinguishing bacterial from viral disease using biomarkers in combination with clinical phenotypic markers and host genetic markers. The registered intervention is validation of a biomarker; the record does not describe a randomized treatment assignment or a trial comparing two clinical management strategies.

The registry reports an enrollment of 7247. The planned study period was described as 5 years. The record lists a start date of 2016-06-02, a primary completion date of 2021-01-25, and a status of COMPLETED. It describes BIVA-ED recruitment to capture children with confirmed bacterial infection, additional recruitment of children with less common febrile illnesses, and afebrile controls. The arm descriptions do not establish that these groups were intended to be exchangeable, and the registry explicitly identifies them as not a like-for-like control comparison.

Participants were monitored for the primary outcome throughout their hospital stay and received follow-up review at 10 days. The record does not report details of blinding, the process for assigning retrospective phenotypes, the number of sites, or the exact recruitment and follow-up procedures beyond that summary. It also does not report a protocol-level statistical plan or a sample-size justification for the study itself. The design is best read as prospective data and sample collection for diagnostic validation, with the posted primary outcome represented as a participant count, rather than as a randomized efficacy trial.

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

Analysis population for the primary outcome, as defined in the registry: Number of children with definite bacterial illness assessed by positive blood culture and/or positive culture of sample from sterile site

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.

6. Primary Endpoint

The registry lists the primary outcome as Number of Participants With Clinically-assigned Retrospective Phenotype, measured over Participants were monitored for outcome throughout their stay in hospital, and received follow-up review at 10 days.

The primary endpoint is the number of participants with a clinically assigned retrospective phenotype, classified according to the cause of illness. The record specifies a population of children with definite bacterial illness assessed by a positive blood culture and/or a positive culture of a sample from a sterile site. The outcome description therefore connects a clinical classification with microbiological evidence in the stated population.

A participant count can describe how many people met a defined outcome classification in a study. It does not, on its own, show whether a diagnostic test correctly distinguishes disease categories. To assess diagnostic performance, analyses would generally need to compare a test result with a defined reference classification and report measures such as sensitivity, specificity, or predictive values. Those estimates are not reported in the registry record.

Retrospective phenotype assignment may support characterization of the illness after clinical information has accumulated, but the record does not specify the adjudication process, the timing of test interpretation relative to assignment, or whether assessors were blinded to biomarker results. Those details matter when judging whether the phenotype can serve as an independent reference for validating a marker. The endpoint as registered is a count, not a direct measure of treatment benefit, clinical utility, or the effect of a diagnostic strategy.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

Statistical interpretation begins by identifying what was measured and how groups were formed. In PERFORM, the registered primary outcome is a participant count, while the study is described as prospective biomarker validation. The registry does not describe randomized allocation. It also identifies the listed arms as not a like-for-like control comparison. These features make it important not to read the study as a conventional randomized comparison of treatments.

What does a participant count convey?

In general, a count tells how many participants met an outcome definition in the population being described. Counts are useful for reporting the scale of an observed category, but a count alone does not show how frequently the outcome occurred unless its denominator and analysis population are clearly identified. It also does not establish a difference between groups. The primary outcome is described as a count of children with definite bacterial illness assessed by positive culture, but the registry record does not report the primary outcome count or a corresponding analysis.

The outcome is described as a binary kind, meaning participants with an event or without an event. In general, a binary outcome can be summarized as counts in each category and analyzed using methods designed for event data. The choice of method depends on the study design, the sampling process, the outcome definition, and whether observations are independent. The registry record does not identify a statistical model for this endpoint. No model, adjustment set, or estimate of uncertainty is reported for the primary outcome.

Why does the comparison structure matter?

In general, randomized assignment can make comparison groups more comparable at baseline, subject to chance variation and the trial's conduct. In an observational or nonrandomized study, participants may differ for reasons related to their clinical presentation or how they entered the study. Here, the registry says allocation is not stated and cautions that the arms are not a like-for-like control comparison. A simple contrast between febrile participants and afebrile controls would therefore be descriptive unless the analysis and assumptions justified a stronger interpretation. The registry does not report a causal comparison or an adjustment strategy.

For diagnostic validation, an important general issue is the reference standard used to decide whether a condition is present. The primary outcome description refers to culture evidence and retrospective clinical phenotyping. Interpreting test accuracy would require details about how biomarker results were compared with the reference classification, including the timing of measurements and whether classification was independent of the marker under evaluation. The registry record does not report those details or diagnostic accuracy estimates. Consequently, sensitivity, specificity, predictive values, and discrimination cannot be assessed from the posted primary outcome description alone.

What do uncertainty measures add?

In general, a confidence interval expresses the range of estimates compatible with the data and the assumptions of a statistical method. Its width indicates precision: wider intervals represent greater uncertainty, while narrower intervals represent greater precision. A p-value addresses compatibility with a specified null hypothesis under a statistical model; it does not measure the size or clinical importance of an effect. Neither type of result is reported for the primary endpoint in this registry record. Without an estimate and its uncertainty, the magnitude and precision of any primary-outcome contrast cannot be assessed.

The record mentions planning illustrations based on a fraction of an observed effect, with a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. These are planning assumptions, not the study's own design justification. Such illustrations ask, in general, how many participants per arm might be needed under specified assumptions to detect a target effect. They do not represent observed power, and they do not establish that the study was designed to detect that effect. The registry record does not report the observed effect used as the basis for the illustrations or the original study's sample-size justification.

For count or event outcomes, methods may also need to account for how variability relates to the mean. In general, over-dispersion occurs when observed variability is greater than a model assumes; ignoring it can make standard errors too small and confidence intervals too narrow. This is a general consideration, not a reported feature or analysis of PERFORM. The registry record does not provide the outcome distribution or model diagnostics needed to judge whether over-dispersion applied.

What can be concluded from the posted analyses?

No statistical analyses are listed for the primary outcome. That means the registry record does not supply estimates from which to assess the magnitude, precision, or robustness of a primary result. It also does not report subgroup analyses, sensitivity analyses, or a strategy for handling clustering or repeated measurements. These are not claims that such work did not occur outside the registry; they are details not reported in the registry record. Any interpretation should therefore distinguish the study's stated diagnostic aims and collected outcomes from analyses that are actually posted.

10. Exact Estimates for Each Arm

No two-arm comparison is shown for this trial, but each arm's result can still be summarized with exact interval estimates computed from the posted counts. The Clopper-Pearson interval is exact and conservative; the Wilson interval is a common closer-to-nominal alternative; the Bayesian credible interval uses a uniform Beta(1, 1) prior. For small groups the three can differ noticeably.

R code and output

arm    <- c("BIVA Studies", "Controls for BIVA Studies")
n      <- c(5952, 1290)
events <- c(655, 0)
for (i in seq_along(n)) {
  ci <- binom.test(events[i], n[i])$conf.int          # Clopper-Pearson exact interval
  cr <- qbeta(c(0.025, 0.975), 1 + events[i], 1 + n[i] - events[i])   # Bayesian, uniform Beta(1,1) prior
  wi <- (function(x, n) { p <- x / n; z <- qnorm(0.975); d <- 1 + z^2 / n
    c((p + z^2 / (2 * n) - z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d, (p + z^2 / (2 * n) + z * sqrt(p * (1 - p) / n + z^2 / (4 * n^2))) / d) })(events[i], n[i])
  cat(sprintf("ARM|%s|%d|%d|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f|%.1f\n", arm[i], n[i], events[i], 100 * events[i] / n[i],
              100 * ci[1], 100 * ci[2], 100 * wi[1], 100 * wi[2], 100 * cr[1], 100 * cr[2]))
}

ARM  BIVA Studies  5952  655  11.0  10.2  11.8  10.2  11.8  10.2  11.8
ARM  Controls for BIVA Studies  1290  0  0.0  0.0  0.3  0.0  0.3  0.0  0.3

11. Multiplicity, Missing Data and Analysis Populations

The registry record lists one primary outcome measure and no statistical analyses. It does not report a statistical analysis plan, the number of hypotheses tested, or any procedure for controlling multiplicity. In general, testing several outcomes, markers, thresholds, or subgroups can increase the chance of finding an apparently notable result by chance. Prespecified outcome hierarchies or multiplicity adjustments can address some of that concern, but the registry does not describe whether such methods were used. No multiplicity-adjusted or unadjusted primary result is reported.

Missing-data handling is also not reported in the registry record. The record does not give the amount or pattern of missing outcome data, the number of participants with evaluable samples, or any imputation or complete-case strategy. These details are relevant because missingness can affect both the representativeness of an analysis population and the precision of estimates. Without those details, the impact of missing information on the primary outcome cannot be assessed.

The registry does not define analysis populations such as an intention-to-treat group or a per-protocol group. For a diagnostic-validation study, the analysis population may depend on sample availability, the timing and quality of test results, and whether a reference classification can be assigned. The record does not explain how participants lacking a usable marker result or definitive phenotype were handled.

These omissions limit assessment of selection, multiplicity, and robustness. They should not be filled in by assuming a particular protocol or analytical convention. The appropriate conclusion is that the registry record does not report the relevant methods or results.

12. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

Most frequent serious adverse event terms

Serious events with the highest arm-level rate: sepsis syndrome (2.0%).

The registry lists serious adverse event figures separately for BIVA Studies and Healthy Controls. For BIVA Studies, the record reports 5957 participants at risk, 120 with a serious adverse event, 0 with another adverse event, and 50 deaths. For Healthy Controls, it reports 1290 participants at risk, 0 with a serious adverse event, 0 with another adverse event, and 0 deaths. The listed top serious term is Sepsis syndrome at 2.0%.

These are descriptive safety entries, not a randomized comparison. The registry record does not report whether an event or death was considered related to study participation, how seriousness was defined, or the precise period over which each safety measure was collected. The primary outcome follow-up description refers to monitoring during the hospital stay and review at 10 days, but the record does not state that this is the safety follow-up window.

The counts should not be added together or treated as mutually exclusive categories. The registry does not explain overlap between serious events and deaths or provide participant-level detail. The Sepsis syndrome percentage is reported as a top term; its denominator and relationship to the other safety counts are not specified in the record. The figures cannot establish that the study caused, prevented, or changed the risk of these events.

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

The registry record supports a limited but useful description of PERFORM: a completed prospective study focused on validating biomarkers and related clinical and host markers in febrile children, with afebrile controls also recruited. It reports enrollment of 7247 and a primary outcome defined as the number of participants assigned a retrospective clinical phenotype, with a stated bacterial-illness population assessed using culture evidence. It also reports group-level safety counts.

Those details do not amount to a reported test-validation result. The record does not provide diagnostic accuracy estimates, marker-specific results, uncertainty intervals, or analyses linking test information to treatment decisions. It therefore does not show whether a marker distinguished bacterial from viral illness, how precisely it did so, or whether using it improved management. These questions require results and methods that are not reported in the registry record.

The study should not be interpreted as a randomized treatment trial. The registered intervention is biomarker validation, allocation is not stated, and the listed arms are explicitly not a like-for-like control comparison. As a result, differences between groups cannot be attributed to a diagnostic test or treatment on the basis of the registry description alone. The stated research rationale concerns better distinction of bacterial and viral disease; a rationale is not itself evidence of diagnostic performance or clinical benefit.

The safety counts also require restraint. They describe reported events in the listed groups but do not establish causation or provide a controlled estimate of risk. The record does not report event attribution or sufficient detail to assess whether the groups had comparable safety surveillance.

In short, the posted information describes the study's scope, endpoint, and selected safety counts, but does not report the statistical results needed to judge the primary diagnostic question. That distinction preserves what the registry can establish without extending its findings beyond the analyses it reports.

14. Why This Trial Matters Statistically

PERFORM addresses a setting in which the underlying clinical question is diagnostic: how to distinguish bacterial illness from other causes of fever while evaluating children in hospital. Statistically, that makes the definition of the reference classification, the timing of tests, and the relationship between the marker and clinical assessment central to interpreting any validation result. The registry identifies culture evidence and retrospective phenotyping in the primary outcome description, but does not report the detailed validation methods or accuracy estimates.

The study also illustrates why a large enrollment does not, by itself, establish a precise or clinically useful result. Precision depends on the number of participants contributing to a particular analysis, the number and distribution of outcome categories, and the statistical method. The registry reports enrollment of 7247, but does not report the primary outcome count or the number with evaluable results for particular markers. The overall enrollment figure therefore cannot substitute for endpoint-specific results.

Its group structure is also statistically important. Febrile participants and afebrile controls may help characterize different clinical contexts, but the registry explicitly warns that the arms are not a like-for-like comparison. Such a comparison is not equivalent to randomized evidence that a diagnostic strategy changes care. Making that distinction helps avoid causal interpretations of descriptive group data.

Finally, the record separates a compelling research aim from the evidence available to evaluate it. Biomarker validation can require multiple candidate markers, thresholds, and subgroups, making prespecification and multiplicity control important general concerns. The registry does not report those methods or statistical analyses. PERFORM therefore highlights both the potential complexity of diagnostic research and the need for transparent endpoint definitions, analysis populations, uncertainty estimates, and validation results when drawing statistical conclusions.

15. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The main limitation for interpretation is that the registry record does not report statistical analyses of the primary outcome. It gives no primary-outcome count, comparative estimate, confidence interval, p-value, or diagnostic-accuracy measure. Thus, the central stated aim of distinguishing bacterial from viral disease cannot be evaluated from the posted primary results described in the record.

The endpoint is defined as a count of participants with a retrospective clinical phenotype, but the record does not explain the adjudication process or how culture findings and clinical information were combined. It also does not report whether those assigning phenotypes were blinded to biomarker results. These omissions limit assessment of the independence and reproducibility of the reference classification.

The record describes febrile participants and afebrile controls, while cautioning that the arms are not a like-for-like comparison. Allocation is not stated. The registry does not report an adjusted comparison or details that would allow assessment of confounding or selection. The safety entries also lack reported event-attribution methods and a clearly specified safety follow-up period.

Finally, the record does not report missing-data procedures, analysis populations, multiplicity handling, or the original sample-size justification. Planning illustrations based on an observed effect are not a substitute for those design details and do not indicate observed power. These limitations concern what is reported in the registry record; they do not establish what analyses or procedures may have been documented elsewhere.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Personalised Risk Assessment in Febrile Illness to Optimise Real-life Management Across the European Union (PERFORM) — NCT03502993.
• Registry record last updated 2023-10-13; status last verified 2023-10.
• Results first posted 2023-10-13.
• 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.

19. Record Summary

PERFORM (NCT03502993), sponsored by Imperial College London, is listed as COMPLETED. The registry reports a start date of 2016-06-02, a primary completion date of 2021-01-25, and enrollment of 7247. The study prospectively recruited febrile children presenting to hospital and included afebrile controls, with a stated focus on validating biomarkers alongside clinical phenotypic and host genetic markers.

The primary outcome is the number of participants with a clinically assigned retrospective phenotype according to cause of illness. The stated bacterial-illness population is assessed by positive blood culture and/or positive culture from a sterile site. The record also lists serious adverse event counts for BIVA Studies and Healthy Controls. It does not report statistical analyses of the primary outcome, diagnostic-accuracy estimates, or a comparative treatment effect. The listed groups are not a like-for-like control comparison, and allocation is not stated.