This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
POSEIDON-2 is a clinical study focused on patient-reported outcomes, patient satisfaction, long-term outcomes, and burden after status epilepticus, with particular attention to health-related quality of life among patients managed in the intensive care unit.
| Feature | POSEIDON-2 |
|---|---|
| Study acronym | POSEIDON-2 |
| ClinicalTrials.gov identifier | NCT06100978 |
| Current status | Recruiting |
| Start date | 2024-05-25 |
| Primary completion date | 2027-03-25 |
| Enrollment | 145 |
| Lead sponsor | Versailles Hospital |
| Sponsor type | Other |
| Conditions / topics | Patient Reported Outcomes; Patient Satisfaction; Long Term Outcomes; Status Epilepticus; Intensive Care Unit; Burden; Relatives |
2. Clinical Question
The primary clinical question concerns the extent to which patients managed in the intensive care unit for or with status epilepticus experience global impairment in health-related quality of life at 3 months and 12 months.
Population
Patients managed in the ICU for or with status epilepticus.
Outcome
Global impairment of health-related quality of life, evaluated using the SF-36 summary scores.
Assessment times
The primary endpoint is assessed at 3 month and 12 month.
Primary question
What percentage of patients have global HRQoL impairment at the specified follow-up assessments?
3. Trial Design
The ClinicalTrials.gov record identifies POSEIDON-2 as a recruiting study with an enrollment of 145 and a primary completion date of 2027-03-25. The registry information identifies the study's focus on patient-reported outcomes and long-term outcomes after status epilepticus.
4. Endpoints
Primary endpoint
| Endpoint | Time frame | Definition |
|---|---|---|
| Percentage of global impairment of HRQoL | at 3 month and 12 month | Percentage of global impairment of HRQoL defined by one and/or other of the physical and mental impairments after dichotomization of the SF 36 summary scores compared to the general population in patients managed in the ICU for or with SE. |
The SF-36 is described in the registry as an HRQoL measure with a score ranging from 0 to 100, where a high score indicates better health status. The primary endpoint is therefore not simply the mean SF-36 score. It is a derived percentage based on whether the physical and/or mental summary scores meet the definition of impairment after dichotomization against the general population.
Why the endpoint is a binary HRQoL outcome
The statistical structure of the primary endpoint is important. The underlying SF-36 summary scores are continuous measurements, but the registered primary endpoint converts those measurements into impairment categories. The resulting analysis target is a percentage of patients classified as globally impaired, rather than an average score on the original 0–100 scale.
The registry's definition means that a patient can meet the global impairment criterion through one or both of the two component impairment domains.
5. Statistical Methodology
The registry specifies the endpoint definition and measurement framework but does not post a formal statistical analysis for the primary endpoint. For an endpoint expressed as the percentage of patients with global impairment, the natural statistical analysis is based on a binary outcome at each assessment time.
Describing the primary endpoint
At each time point, the fundamental quantity is the proportion of evaluable patients classified as globally impaired:
A confidence interval around this proportion would quantify uncertainty attributable to the observed sample. The appropriate interval method should reflect the binary nature of the endpoint and the observed sample size.
This distinction matters because the denominator should correspond to the patients who have an evaluable endpoint at the relevant assessment, while the registry's enrollment figure describes the planned study population rather than the number necessarily available for a particular follow-up analysis.
Analysis at 3 months and 12 months
The primary endpoint has two assessment times. A straightforward descriptive analysis would estimate the percentage with global impairment separately at 3 months and 12 months, together with confidence intervals.
If the scientific objective were to formally compare impairment between the two follow-up times within patients, the paired nature of repeated measurements would need to be taken into account. A simple comparison of two percentages that treats the observations as independent would not fully use the within-patient structure when the same patients contribute data at both time points.
Potential modeling framework
For repeated binary outcomes, a longitudinal logistic model or another method that accounts for within-patient correlation could be considered. Such a model would allow the probability of global impairment to be related to follow-up time while recognizing that measurements from the same patient are correlated.
Descriptive approach
Report the percentage of patients with global impairment separately at 3 months and 12 months, with uncertainty intervals.
Longitudinal approach
If change over time is a formal question, use a repeated-measures method that accounts for correlation between observations from the same patient.
Physical and mental summary scores
The endpoint begins with two SF-36 summary scores: physical and mental. Because the primary outcome is defined as impairment in one and/or the other, the two component domains should be preserved during analysis rather than collapsed into a single continuous score before applying the registered definition.
This also means that a patient with physical impairment alone, mental impairment alone, or both impairments can contribute to the global impairment percentage. The logical "and/or" structure should be implemented consistently at both follow-up assessments.
6. Statistical Methods Explained
Why is the SF-36 score dichotomized?
The registry defines global impairment after dichotomization of the SF-36 summary scores compared to the general population. Dichotomization changes the statistical question from "what is the average SF-36 score?" to "what percentage meets the predefined impairment criterion?" This produces a clinically interpretable classification, but it also discards some information contained in the original continuous scores.
What does a percentage of global impairment represent?
It represents the proportion of evaluated patients classified as having physical impairment and/or mental impairment according to the endpoint definition. It is not the average severity of impairment and does not indicate how far an individual patient's SF-36 score lies below the comparison threshold.
Why are 3-month and 12-month assessments different statistically?
They represent two distinct follow-up snapshots. If the same patients are assessed at both times, the observations are correlated. A longitudinal analysis can exploit that pairing; treating the two percentages as if they came from unrelated groups would ignore an important feature of the data.
Why can the original SF-36 scale and the primary endpoint tell different stories?
The SF-36 score ranges from 0 to 100, with higher scores indicating better health status. The primary endpoint, however, is based on a dichotomized impairment classification. Two patients can therefore have different SF-36 scores while being assigned to the same impairment category, and clinically meaningful changes that remain on the same side of the threshold may not change the binary endpoint.
Why are confidence intervals important for a percentage?
A percentage calculated from a finite number of patients is an estimate of the underlying proportion in the study population. A confidence interval communicates statistical uncertainty around that estimate. It should not be interpreted as saying that individual patients have a corresponding probability of belonging to the impairment category.
What would a regression model add?
A regression model could estimate associations between global impairment and explanatory variables while allowing adjustment for prespecified covariates. For repeated assessments, a longitudinal model could also account for the fact that multiple observations from the same patient are not independent. Any such model should be specified before being used for confirmatory inference.
7. Planned Analysis
The registry's primary endpoint will measure the percentage of global impairment of HRQoL at 3 month and 12 month. Global impairment is defined by one and/or other of the physical and mental impairments after dichotomization of the SF-36 summary scores compared to the general population.
For an endpoint of this type, the primary statistical description would normally consist of the observed proportion of patients meeting the global impairment criterion at each time point, accompanied by an appropriate confidence interval. If comparison between 3-month and 12-month impairment is a formal objective, a paired or longitudinal analysis would normally be preferable when the same patients are evaluated at both time points.
8. Interpreting the HRQoL Endpoint
The primary outcome summarizes whether each patient meets the registered definition of global HRQoL impairment. This is different from reporting a mean SF-36 score. The percentage therefore answers a prevalence-style question: what proportion of evaluated patients are classified as globally impaired at the specified follow-up?
Global impairment is defined by physical impairment and/or mental impairment. The two components therefore contribute jointly to the primary outcome. A global percentage can conceal whether impairment is driven predominantly by the physical domain, the mental domain, or both.
Because SF-36 summary scores are converted into categories, the primary endpoint does not preserve the full numerical distance between patients' scores. A continuous analysis of the SF-36 scores, if separately specified, could answer a different question about average or individual-level changes in HRQoL.
HRQoL at 3 months and HRQoL at 12 months represent different stages of recovery or long-term outcome. A difference between those assessments would describe a change in the observed burden over follow-up, not necessarily a causal effect of a particular treatment.
9. Missing Data and Follow-Up
Patient-reported outcomes create a particular missing-data issue: the patients who provide an SF-36 at follow-up may differ systematically from those who do not. For example, inability to complete a questionnaire can be related to health status itself. Consequently, the percentage among observed respondents may not necessarily represent the percentage that would have been observed if every enrolled patient had completed the assessment.
The registry information does not report a missing-data or imputation method for the primary endpoint. In a full statistical analysis plan, the handling of missing SF-36 assessments would therefore be an important prespecified component of the analysis.
Observed-case analysis
Uses patients with an evaluable outcome at the relevant assessment. Its interpretation depends on the extent and mechanism of missingness.
Imputation
Can be considered when justified by the analysis plan, but assumptions about the missing outcomes must be made explicit and evaluated.
10. Population, Measurement, and Generalizability
The registered population is specifically connected to intensive care management for or with status epilepticus. The primary outcome is consequently a measure of HRQoL within a clinically defined population rather than a general population survey.
The comparison with the general population is built into the endpoint definition through the dichotomization of SF-36 summary scores. This gives the endpoint a reference standard, but it also means that the interpretation depends on how the relevant general-population comparison is defined.
The ClinicalTrials.gov record does not report detailed baseline characteristics, subgroup definitions, or demographic distributions. Those characteristics would be important when determining how broadly the observed HRQoL findings could be generalized beyond the study population.
11. ClinicalTrials.gov Status and Study Timeline
Study start
The registered study start date is May 25, 2024.
Recruiting
The ClinicalTrials.gov record currently identifies POSEIDON-2 as recruiting.
Primary completion
The registered primary completion date is March 25, 2027.
The study is therefore positioned around prospective follow-up of patient-reported outcomes, with the primary HRQoL endpoint assessed at 3 month and 12 month.
12. Limitations
- No posted efficacy results: the ClinicalTrials.gov record does not currently report results for the primary HRQoL endpoint.
- No formal posted analysis: the registry does not post a formal statistical analysis for the primary endpoint.
- Dichotomization: converting continuous SF-36 summary scores into impairment categories can simplify interpretation while losing information about the magnitude of differences within categories.
- Composite structure: the global impairment endpoint combines physical and mental impairment through an "and/or" definition, so the single percentage may conceal differences between the two domains.
- Repeated assessments: measurements at 3 months and 12 months may be correlated when the same patients contribute data at both time points.
- Missing patient-reported outcomes: incomplete follow-up can affect the interpretation of observed percentages, particularly if missingness is related to health status.
- General-population comparison: the impairment classification depends on the reference framework used to dichotomize the SF-36 summary scores.
- Limited design detail: the registry information does not report a randomized intervention, comparator, allocation ratio, or treatment effect for the primary endpoint.
13. Why This Trial Matters Statistically
POSEIDON-2 is a useful statistical teaching case because its primary endpoint illustrates a common but consequential transformation in clinical research: a continuous patient-reported measurement is converted into a clinically defined binary outcome.
| Concept | How it appears in POSEIDON-2 |
|---|---|
| Patient-reported outcome | HRQoL is measured from the patient's perspective using the SF-36. |
| Continuous measurement | SF-36 summary scores range from 0 to 100, with higher scores indicating better health status. |
| Dichotomization | Physical and mental summary scores are converted into impairment classifications. |
| Composite endpoint | Global impairment is defined by physical impairment and/or mental impairment. |
| Binary outcome | The primary result is expressed as a percentage of patients meeting the global impairment definition. |
| Repeated follow-up | The primary endpoint is evaluated at 3 month and 12 month. |
| Reference population | The impairment definition compares SF-36 summary scores with the general population. |
| Missing-data considerations | Patient-reported follow-up can be incomplete, making the analysis population an important consideration. |
14. Statistical Interpretation of the Study Question
The central statistical object in POSEIDON-2 is not a hazard ratio or a treatment-group contrast. It is a proportion: the percentage of patients who satisfy a predefined definition of global HRQoL impairment at each specified follow-up time.
That distinction determines the appropriate analytical vocabulary. Kaplan-Meier methods and Cox proportional-hazards models are designed for time-to-event outcomes and are not naturally suited to this primary endpoint. Likewise, a treatment-effect measure such as a hazard ratio would not describe the registered HRQoL outcome.
Instead, the analysis begins with a patient-level classification derived from the two SF-36 summary domains. The resulting binary variable can then be summarized at each time point. If longitudinal change is of interest, the repeated observations should be analyzed with a method appropriate for correlated binary data.
Each step changes the statistical quantity being analyzed. The final primary endpoint is therefore not equivalent to the underlying continuous SF-36 score.
15. What the Primary Endpoint Does — and Does Not — Mean
The percentage represents the proportion of evaluated patients classified as having global HRQoL impairment according to the registered definition at 3 month or 12 month.
It does not provide a continuous measure of the severity of HRQoL impairment. Patients within the same impairment category may have substantially different underlying SF-36 scores.
A descriptive percentage at a follow-up time does not by itself establish that a particular exposure, treatment, ICU characteristic, or other factor caused the observed HRQoL outcome. Causal interpretation requires an appropriate study design and prespecified analysis.
Because global impairment can be triggered by either physical or mental impairment, the overall percentage should ideally be interpreted alongside the component domains when those results are available. The global endpoint answers whether impairment is present; the component scores help explain its domain structure.
16. Sources
- ClinicalTrials.gov: POSEIDON-2 (NCT06100978).
Continue through the Clinical Biostats knowledge graph
Explore additional statistical tutorials, calculators, and clinical-trial analyses covering the methods used to understand patient-reported outcomes and clinical research evidence.
17. Record Summary
POSEIDON-2 is a recruiting study of patient-reported outcomes and long-term outcomes after status epilepticus, with an enrollment of 145 and a primary completion date of 2027-03-25. Its primary endpoint is the percentage of global impairment of HRQoL at 3 month and 12 month, defined by physical and/or mental impairment after dichotomization of the SF-36 summary scores compared to the general population. The SF-36 ranges from 0 to 100, with higher scores indicating better health status.
Statistically, the study provides a clear example of how a continuous patient-reported measure can be transformed into a binary clinical outcome. The resulting percentage can be described with an appropriate confidence interval, while formal comparison of 3-month and 12-month outcomes would need to account for repeated measurements when the same patients contribute observations at both assessments. Interpretation also depends on the handling of missing patient-reported outcomes and on the reference framework used to define impairment.