Prospective ED Headache Cohort Study: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Prospective ED Headache Cohort Study (NCT00313703), a registered phase not stated study. The primary outcome is Number of Participants Who Report Moderate or Severe Pain Within 24 Hours of Emergency Department(ED) Discharge.

2. Clinical Question

The record describes enrollment of 477 and a completed study. That enrollment figure does not establish how many participants contributed to the primary endpoint: the endpoint description says a small number were lost to follow-up and could not be included, without stating their number. Outcome counts, group-specific denominators, and prediction-performance results are not reported in the registry record. Therefore, the clinical question can be described, but its observed answer and the precision of any estimated group differences cannot be assessed from this record.

3. Trial Design

The study is titled Prospective ED Headache Cohort Study, and its official title concerns predicting poor pain and functional outcomes after discharge from the emergency department with a primary headache. The registry classifies its status as completed and lists enrollment of 477. The study start is recorded as 2004-03 and primary completion as 2005-08. Montefiore Medical Center is listed as the sponsor.

The title and stated purpose indicate a cohort investigation of headache outcomes after emergency-department discharge, rather than a randomized treatment trial. Allocation is not stated. The record lists Migraine, Tension-type Headache, Unclassifiable, and Other as groups, but does not describe how those categories were assigned, whether participants could change categories, or whether the groups were intended to be directly compared. They are diagnostic categories, not evidence of assigned treatments or a like-for-like control comparison.

The primary outcome measure is a count of participants reporting moderate or severe pain within 24 hours after discharge. The record also lists two outcome measures overall, but the other measure’s details are not included in the registry record described for this analysis. The registry records zero analyses. That means no statistical analyses are reported in the registry record; it does not establish that no analyses were performed outside the record.

The endpoint description notes that a small number of patients in each group were lost to follow-up and could not be included in this analysis. The exact number, reasons for loss, and handling of missing information beyond exclusion from this endpoint are not reported in the registry record. No randomization procedure, allocation concealment, blinding, recruitment setting details, or prespecified statistical analysis plan is reported. As a result, the design can be characterized as a prospective cohort from the title and purpose, but important details needed to assess selection, measurement, and analysis procedures remain unavailable.

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: 477 patients underwent a detailed interview in the ED. 309 of these patients were determined to have a primary headache disorder (a benign recurrent headache disease) and were entered into the post-ED follow-up portion of this cohort study.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: A small number of patients in each group were lost-to-follow-up and could not be included in this analysis.

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 Who Report Moderate or Severe Pain Within 24 Hours of Emergency Department(ED) Discharge, measured over 24 hours after Emergency Department (ED) discharge.

The primary endpoint counts participants who reported moderate or severe pain within 24 hours after emergency-department discharge. The description says that moderate and severe refer to the participant’s own description of pain. It is therefore a participant-reported measure of short-term pain after discharge, expressed as a number of participants rather than as a continuous pain score.

Such an endpoint can be useful for describing whether substantial pain persists or recurs soon after discharge. A binary classification of each participant as meeting or not meeting the stated severity threshold can also support estimation of the proportion with the outcome, if the relevant counts and denominators are available. The registry, however, specifies the measure as a number of participants and does not report the observed counts or denominators in the registry record for this analysis. A proportion, risk estimate, or uncertainty interval cannot therefore be described from this record.

The endpoint is limited to the stated assessment time. It does not by itself measure the full course of headache, longer-term function, or all aspects of continued suffering. The study purpose mentions prediction and recurrence, but the posted primary endpoint is not a prediction-performance statistic and does not, by itself, show whether headaches recurred. The record does not report the precise question used to elicit pain, how the severity categories were explained to participants, or whether pain was measured in other ways. Those details matter when interpreting a self-described outcome, because classification depends on the participant’s report and on how the measure was administered.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is described as a number of participants with moderate or severe pain by 24 hours after discharge. The record classifies the outcome type as binary and its meaning as participants with an event (counts). In a binary-outcome analysis, each assessed participant is classified according to whether the outcome occurred. The resulting count can be paired with the number assessed to estimate an event proportion. The registry does not report either the outcome count or the analysis denominator, so no event proportion can be stated.

When comparing binary outcomes between groups, common summaries include group-specific proportions and a contrast between those proportions. A confidence interval expresses the range of values reasonably compatible with the data and the analysis assumptions; its width conveys precision. A small count or a limited denominator can yield substantial uncertainty. These are general statistical principles, not methods reported as having been used in this study. No group estimates, confidence intervals, test results, or detailed analysis methods are reported in the registry record.

Because allocation is not stated and the study is not a randomized comparison, any group comparison would be observational. In general, random assignment helps balance measured and unmeasured participant characteristics on average; in a cohort comparison, differences in such characteristics may accompany differences in outcome. Statistical adjustment can address measured factors when they are recorded and modeled appropriately, but cannot automatically remove bias from unmeasured factors or poor measurement. The registry record does not report covariates, adjustment methods, or a causal analysis plan. No causal interpretation is warranted from the listed diagnostic groups alone.

The stated planning note describes a sample-size illustration computed from an observed effect, rather than the trial’s own design. Its stated assumptions are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation; the effect to detect is the stated fraction of the observed effect. The note says time-to-event tables give the total number of events using Schoenfeld, while other tables give participants per arm. These are planning conventions, not evidence of the study’s actual allocation, analysis, or achieved precision. The registry entries for this analysis do not include calculated sample-size values, and the record reports no analyses. The illustrations should not be read as observed power or as proof that the cohort was adequately sized for the endpoint.

For a binary outcome, the normal approximation is a general method for estimating uncertainty or planning sample size. Its suitability depends on the data and the assumptions behind the calculation. Other methods may be preferable in some settings, for example when event counts are sparse. The registry does not report which inferential method, if any, was used for the primary endpoint, so no method-specific conclusion can be attributed to this study. Likewise, over-dispersion is a concern primarily when count data vary more than a specified model allows; it is not something that can be diagnosed from the endpoint definition alone. The posted endpoint is a count of participants meeting a binary condition, but no data or fitted model are reported from which to evaluate dispersion.

Prediction requires a different assessment from simple outcome description. A prediction model uses participant characteristics to estimate an individual outcome risk. General evaluation considers how well predicted risks align with observed outcomes and how well the model distinguishes participants with and without the outcome; validation examines performance beyond the data used to develop the model. The study purpose mentions learning to predict continued suffering, but the registry entries do not identify predictors, model form, validation, calibration, discrimination, or prediction error. Thus, the stated objective should not be mistaken for a reported or validated prediction model.

Missing outcomes also affect interpretation. The endpoint description states that a small number of patients in each group were lost to follow-up and could not be included. If participants with missing outcomes differ systematically from those retained, an analysis limited to observed outcomes may not represent the full enrolled cohort. This is a general concern, not evidence that such bias occurred. The exact missing count, reasons for missingness, and any sensitivity analysis are not reported. No imputation or other missing-data procedure is described in the registry record.

Finally, the registry lists two outcome measures and zero analyses. An outcome measure is a planned or recorded measure; it is not itself a statistical result. A listed count of measures does not show that multiple comparisons were performed, and a listed count of analyses does not establish the complete analysis history outside the registry. The record does not state whether multiplicity was considered, whether outcomes were prioritized beyond the primary designation, or whether any analysis was adjusted. These omissions limit assessment of the statistical plan and the strength of any inferential claim.

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("Migraine", "Tension-type Headache", "Unclassifiable")
n      <- c(172, 32, 75)
events <- c(53, 6, 20)
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  Migraine  172  53  30.8  24.0  38.3  24.4  38.1  24.4  38.1
ARM  Tension-type Headache  32  6  18.8  7.2  36.4  8.9  35.3  9.0  35.5
ARM  Unclassifiable  75  20  26.7  17.1  38.1  18.0  37.6  18.0  37.7

11. Multiplicity, Missing Data and Analysis Populations

The registry lists two outcome measures and zero analyses. The primary measure is the count of participants reporting moderate or severe pain within 24 hours after discharge. The other measure’s details are not reported in the registry record described for this analysis. A count of outcome measures does not establish how many statistical tests were conducted. The record does not report a multiplicity plan, such as a procedure for controlling the chance of false-positive findings across multiple outcomes or comparisons. It also does not report whether any adjustment was used. With no analyses reported, there is no inferential result on which to assess the practical effect of multiplicity.

For general context, testing several hypotheses can increase the chance that at least one result appears statistically notable by chance, unless the analysis plan accounts for the number and structure of comparisons. This concern cannot be quantified for this study from the registry record. The listed diagnostic groups do not establish that every possible pairwise comparison was tested, and no such tests should be assumed.

For missing data, the primary outcome description states that a small number of patients in each group were lost to follow-up and could not be included. The number lost, reasons for loss, group-specific denominators, and characteristics of those with missing outcomes are not reported. The record does not describe imputation, weighting, sensitivity analyses, or assumptions about missingness. In general, excluding participants with unavailable outcomes can bias an estimate if the probability of missingness is related to the outcome or participant characteristics. That is a potential limitation, not a demonstrated bias in this cohort.

The record does not define an analysis population beyond noting that patients lost to follow-up were excluded from this endpoint analysis. It does not state whether the primary analysis included all other enrolled participants, whether any additional exclusions occurred, or how participants with incomplete information were handled. Enrollment of 477 should therefore not be read as the endpoint denominator or as the number analyzed. The available information permits identification of missing follow-up as a concern, but not an assessment of its magnitude or impact.

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

The registry record establishes that the study concerned headache after emergency-department discharge and specified moderate or severe pain within 24 hours as its primary endpoint. It does not report the endpoint count, analyzed denominator, or group-specific results. Consequently, it does not establish how common the measured pain outcome was in the cohort, how much outcomes differed across the listed diagnostic categories, or how precise any such difference might be.

The record also does not establish that one diagnosis caused more pain than another. The study is not a randomized comparison, allocation is not stated, and adjustment methods are not reported. Group differences in an observational cohort can reflect characteristics other than the group label. The diagnostic categories cannot be interpreted as assigned interventions or as randomized treatment and control groups.

The study purpose refers to prediction and continued suffering, but a stated purpose is not a reported prediction result. The registry entries do not identify a prediction model, its inputs, its validation, or its performance. The primary endpoint count alone would not demonstrate that a model can accurately estimate outcomes for individuals.

Missing follow-up further qualifies what can be inferred. The record says a small number in each group could not be included in the endpoint analysis, but does not state the number or describe their outcomes. Thus the analyzed participants may not fully represent everyone enrolled, and the direction or size of any resulting bias cannot be determined from the record.

These limits do not mean that the outcome was absent, that diagnostic groups were equivalent, or that the study had no value. They mean that the registry information available for this analysis is descriptive about the research question and endpoint, but insufficient for a numerical estimate or a conclusion about association, causation, or predictive performance. Any planning illustration based on an observed effect is a design aid, not evidence of the observed clinical result or achieved power.

13. Why This Trial Matters Statistically

This study illustrates the distinction between a clinically relevant question and a statistically interpretable result. The endpoint concerns participants’ reports of moderate or severe pain soon after emergency-department discharge, a clearly time-bounded outcome. Yet interpreting its frequency requires both the number meeting the endpoint and the number whose outcomes were assessed. The registry record does not report those values, so the endpoint’s magnitude cannot be quantified from the available information.

It also highlights the importance of design when comparing groups. Migraine, Tension-type Headache, Unclassifiable, and Other are listed as diagnostic groups, but allocation is not stated and the study is not a randomized comparison. A statistical contrast between observational groups would describe an association unless a stronger design and analysis justified causal interpretation. Group labels alone do not control for other differences between participants.

The distinction between outcome measurement and prediction is also important. Counting participants with pain at a specified time can describe an outcome, while prediction requires a model and evidence that its estimates perform adequately. Although prediction is part of the study’s stated purpose, the registry entries do not report model details or performance measures. This makes clear why a study objective cannot be substituted for evidence that the objective was achieved.

Finally, the record notes loss to follow-up without quantifying it and lists no analyses. That combination limits evaluation of missing-data effects, uncertainty, and multiplicity. The planning note’s sample-size assumptions are not a record of observed power or a substitute for results. Statistically, the case underscores why complete denominators, outcome counts, analysis methods, and uncertainty estimates are essential for turning a defined clinical question into an assessable result.

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.

The central limitation for interpretation is that the primary outcome result is not reported in the registry record. The endpoint definition is available, but its observed count, denominator, and group-specific values are not. No confidence interval or other measure of uncertainty is reported. This prevents assessment of outcome frequency, comparative magnitude, or precision.

The record also provides limited detail about the cohort. It identifies headache as the condition, gives a prospective cohort title, and lists diagnostic categories, but does not report eligibility criteria, recruitment procedures, how diagnoses were classified, or participant characteristics. These gaps make it difficult to assess who the results would represent, even if outcome estimates were available.

Loss to follow-up is acknowledged for each group, but its exact extent and reasons are not reported. The record does not describe the outcomes or characteristics of those lost, nor does it report sensitivity analyses for missing data. This leaves uncertainty about how exclusion may have affected the endpoint estimates.

The study’s stated prediction aim cannot be evaluated because the record does not report model predictors, model development, validation, or prediction metrics. Likewise, no statistical analyses are listed, and no adjustment strategy or multiplicity approach is described. The absence of these details limits assessment of both comparative and predictive claims.

Finally, the diagnostic groups do not constitute randomized treatment arms. Allocation is not stated, so direct treatment-effect interpretations would be inappropriate. The record’s enrollment figure of 477 is not an endpoint analysis denominator, particularly given the acknowledged losses to follow-up. These limitations are specific to the information recorded for this cohort; they do not establish that unreported analyses or procedures were not conducted elsewhere.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Prospective ED Headache Cohort Study — NCT00313703.
• Registry record last updated 2018-05-15; status last verified 2018-04.
• Results first posted 2011-05-09.
• 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

NCT00313703 is a completed prospective emergency-department headache cohort study sponsored by Montefiore Medical Center. The registry lists enrollment of 477, a start date of 2004-03, and primary completion in 2005-08.

The primary endpoint is the number of participants reporting moderate or severe pain within 24 hours after emergency-department discharge. The measure relies on participants’ descriptions of pain. The study purpose concerns headache recurrence and prediction of continued suffering after discharge.

The listed groups are Migraine, Tension-type Headache, Unclassifiable, and Other. The study is not a randomized comparison, and allocation is not stated. A small number of patients in each group were lost to follow-up and excluded from the endpoint analysis, but the exact number is not reported. The registry lists two outcome measures and zero analyses. The observed primary endpoint results, statistical methods, and prediction performance are not reported in the registry record.