Paracetamol With Caffeine to Treat Episodic Tension Type Headache: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Paracetamol With Caffeine to Treat Episodic Tension Type Headache (NCT01755702), a randomized phase 2, phase 3 study. The primary outcome is Time to First Perceptible Headache Relief.

2. Clinical Question

3. Trial Design

The registry describes a randomized, double-masked, crossover study in episodic tension-type headache. It lists Phase 2 and Phase 3 and reports enrollment of 66. The study compared an experimental paracetamol-plus-caffeine formulation with placebo, marketed paracetamol, and marketed ibuprofen. The stated purpose was to assess efficacy of paracetamol taken in combination with caffeine for this headache condition.

A crossover design means participants receive more than one study condition over time, so comparisons may use within-participant information rather than treating every treatment group as entirely separate sets of people. That is a general description of the design type, not a description of the trial’s exact sequence. The treatment order, number of treatment periods, washout procedures, and handling of period or sequence effects are not reported in the registry record. The available arm counts should therefore be read as reported outcome counts, not as a complete account of the crossover schedule.

The study began in 2009-07, and primary completion was recorded as 2010-03. Its status is terminated. The reason for termination, whether termination affected the planned conduct, and whether all intended comparisons were completed are not reported in the registry record. The registry lists nine secondary outcome measures, including headache relief scores, pain-relief summaries, time to rescue medication, complete headache relief, headache severity, and a global treatment assessment. It also reports six analyses for the primary outcome.

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: 66 participants were randomized. Each participant was expected to complete 1, 2, or 3 periods depending on number of headache episodes, for example participants experiencing 2 episodes would take the first 2 treatments, but would not get the 3rd treatment. Such participants would be considered as completer, not a withdrawal.

Period I

Wash Out Period I

Reasons for not completing this period:

Period II

Washout Period II

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Intention to treat (ITT) population: All randomized participants who received at least one study medication and who had at least one post-baseline efficacy assessment.

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 Time to First Perceptible Headache Relief, measured over Baseline to 6 hours.

The primary endpoint was time to first perceptible headache relief, measured in minutes from treatment to the participant’s first report of “a little” pain relief. The stated assessment window was baseline to 6 hours. If that time was unavailable in the electronic pad, the registry says the earliest time with a pain-relief score of 1 or greater was used instead.

This is a time-to-event outcome: the event is the first recorded perceptible relief, and the endpoint captures when it occurs rather than only whether a participant reports relief at a single fixed time. In general, time-to-event methods can incorporate information about event timing and account for observations without a recorded event during the assessment period, depending on the analysis and censoring rules. The precise censoring rules are not reported in the registry record.

The registry presents the outcome using medians and full ranges for each treatment condition, and compares conditions using Cox proportional hazards models. A median describes the central observed time; a full range describes the smallest and largest reported values, not the spread of most participants’ times. Those summaries alone do not show the distribution’s shape or establish how many participants had no recorded relief during the assessment window. The endpoint is useful for distinguishing timing, but interpretation depends on the event definition and on the model’s assumptions. Details of additional endpoint rules are not reported in the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry reports medians with full ranges for the primary time-to-relief endpoint and pairwise hazard ratios from Cox proportional hazards models. These methods answer related but different questions. The median and range describe the observed times within each condition. A Cox model compares the rate at which the defined event occurs over follow-up, without expressing the comparison directly in minutes.

What does a median convey?

In general, a median is the middle of an ordered set of observations. It is less affected by extreme values than a mean, but it does not show the whole distribution. Here, each reported median is 30 minutes. The full ranges differ across conditions, which indicates that the observed minimum and maximum times were not the same. A range is sensitive to extreme observations and does not indicate whether most observations cluster near the median or are widely dispersed. The registry does not report additional distribution summaries for this endpoint.

How should a hazard ratio be read?

In general, a hazard is the rate at which an event occurs among participants who have not yet experienced it. A hazard ratio compares that rate between conditions over the modeled follow-up. With first perceptible relief as the event, a value greater than 1 is generally consistent with a higher rate of reaching relief in the first-listed condition, while a value below 1 is generally consistent with a lower rate. This is not a direct ratio of median times, nor does it quantify a difference in minutes. A hazard ratio also does not by itself describe the absolute probability of relief at a particular time.

The Cox proportional hazards model is commonly interpreted under a proportional-hazards assumption: the relative hazards are treated as constant over the analysis period. In general, if the relative rate changes over time, one summary hazard ratio may not represent the pattern well. The registry does not report diagnostics for this assumption or details about any alternative modeling. It names the model but does not report the covariates, handling of ties, or censoring specifications.

What do the confidence intervals and p-values add?

In general, a confidence interval expresses uncertainty around an estimate under the model and its assumptions. A wider interval indicates less precision than a narrower one. The 95% intervals reported for all six comparisons include 1, the value corresponding to equal hazards. The intervals therefore do not rule out equal relative hazards, and they also include values on either side of 1. The p-values quantify how compatible the observed result is with a specified null model under the analysis assumptions; they are not the probability that a treatment is effective or ineffective. Each reported p-value exceeds 0.05, but that fact does not establish equivalence.

The registry calls the study a crossover trial, while the reported comparisons are presented as Cox-model hazard ratios. In general, crossover analyses may need to account for repeated observations from the same participant and for treatment period or sequence. The precise model structure and how these issues were handled are not reported in the registry record. Without those details, the model label alone does not establish whether the analysis accounted for all relevant features of the design.

Why can over-dispersion matter?

In general, over-dispersion means that observed variability is greater than a model’s assumed variability. It can make standard errors too small if ignored, producing intervals that are too narrow and tests that appear more decisive than warranted. The registry does not report an over-dispersion assessment, and the listed primary method is a Cox proportional hazards model. There is no basis in the registry record to claim that over-dispersion was present or that a particular adjustment was used.

The posted results also include a planning illustration based on a stated effect fraction, equal allocation, a two-sided significance level of 0.05, and a normal approximation. For time-to-event outcomes, the stated planning convention concerns total events; for other outcomes it concerns participants per arm. Such illustrations are hypothetical design calculations based on an observed effect, not observed trial results and not evidence of the probability that this completed study would reach a particular conclusion. No trial-specific sample-size justification is reported in the registry record.

10. Analyses Reported by the Sponsor

The registry identifies GlaxoSmithKline as sponsor and reports six Cox proportional hazards analyses for the primary endpoint. These are the posted comparisons: the hazard ratios range from 0.95 to 1.32, with 95% confidence intervals that include 1 and p-values from 0.201 to 0.819. The registry record does not identify a separate sponsor report or provide an alternate set of estimates for comparison.

No discrepancy is flagged between the posted analysis and a recalculation. A recalculation from summary statistics, when available, would generally be unadjusted and would not necessarily reproduce a model-based analysis, particularly in a crossover design. The reported results use a Cox model, and the registry does not provide enough detail to reconstruct its model specification or participant-level analysis. For that reason, the registry’s reported analysis is the relevant result to describe; no separate numerical recalculation is reported in the registry record.

The reported estimates should still be interpreted with their uncertainty and the limits of the available methods information. The registry does not report the protocol’s statistical analysis plan, model diagnostics, or a detailed account of how the crossover structure was handled. These omissions prevent a fuller appraisal of the sponsor’s analytic decisions, but they do not justify replacing the posted estimates with an assumed alternative analysis.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists six pairwise analyses of the primary time-to-first-perceptible-relief endpoint. It also lists nine secondary outcome measures, while eight secondary outcomes are posted. Multiple comparisons can increase the chance that at least one apparently noteworthy result occurs by chance when many tests are considered. In general, multiplicity procedures can control that risk, but the choice depends on the prespecified testing plan and the claims being evaluated. Adjustment for multiple comparisons is not reported in the registry record.

The six posted primary comparisons should therefore be read as a set of results rather than as isolated tests. All six reported p-values exceed 0.05, and all six 95% confidence intervals include 1. Those findings do not establish that the treatments are equivalent. They also do not reveal whether a particular comparison was designated as the main confirmatory contrast, whether the other comparisons were exploratory, or whether a correction was planned. Those details are not reported in the registry record.

The primary analysis population is described as randomized participants who received at least one study medication and had at least one post-baseline efficacy assessment. That definition is not identical to all randomized participants. Arm-level endpoint counts are reported as 47 for paracetamol/caffeine, 50 for ibuprofen, 45 for paracetamol, and 45 for placebo. The registry does not report a participant flow accounting for all randomized participants, the reasons for exclusions, the amount or pattern of missing endpoint data, or the exact handling of missing observations in the Cox models.

In general, missing data can affect estimates if the chance of an observation being missing is related to its outcome or treatment experience. The registry does not report sensitivity analyses for missing data or the assumptions used to handle censoring. It also does not report interim analyses, stopping rules, or a detailed statistical analysis plan. Because the study status is terminated, the reason for termination and its relationship to the planned analyses are not reported in the registry record. These gaps limit assessment of how the analyses were selected and how robust they are to alternative assumptions.

12. Safety Results

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

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 1% reporting threshold.

Most frequent non-serious adverse event terms

The registry reports zero serious adverse events for each of the four treatment conditions. The numbers at risk were 47 for paracetamol/caffeine, 50 for ibuprofen, 45 for paracetamol, and 45 for placebo. No serious adverse-event terms are listed. The record reports other adverse events for 2 participants receiving paracetamol/caffeine, 4 receiving ibuprofen, 0 receiving paracetamol, and 3 receiving placebo.

These are counts within the stated groups, not a complete safety comparison. A zero reported count means that no serious adverse events are recorded for that condition in this registry entry; it does not establish that the true risk is zero. In general, safety estimates from small groups are imprecise, and rare events may not occur in a study even when they can occur in broader use. The registry record does not provide event definitions, severity details for other adverse events, exposure duration by condition, or a fuller accounting of withdrawals due to adverse events. Death counts are not reported in the registry record. The available figures should be read as descriptive reporting, not as proof that one condition is safer than another.

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

The posted results describe the timing of first perceptible headache relief under four conditions in a randomized, double-masked crossover study. The median was 30 minutes in each condition. The Cox-model comparisons did not show a clear statistical distinction: every reported 95% confidence interval includes 1, and every p-value exceeds 0.05. At the same time, the intervals span a range of possible relative event rates, so the estimates do not establish that the conditions have identical effects.

These results do not demonstrate equivalence. Equivalence requires a prespecified margin and an analysis designed to determine whether uncertainty is sufficiently narrow to exclude differences considered important. The registry does not report an equivalence margin or an equivalence analysis. Nor do the findings establish that one treatment works or does not work. They describe the results recorded for a particular endpoint, population definition, and set of model comparisons.

The primary endpoint concerns time to first perceptible relief, not every aspect of headache experience. The registry lists other outcomes, including pain-relief scores, time to rescue medication, complete relief, headache severity, and global assessment, but the available primary results cannot stand in for those outcomes. The registry record does not provide results for every listed outcome in the information.

Interpretation is also limited by missing details about the crossover sequence, washout, model specification, censoring, missing data, multiplicity, and termination. Enrollment was 66, and the endpoint analysis counts are smaller or larger than enrollment depending on treatment condition; the registry does not provide a full flow explanation. The findings are therefore best treated as estimates with substantial uncertainty and incomplete information about the analysis context, rather than as definitive evidence about comparative treatment effects.

14. Why This Trial Matters Statistically

This trial illustrates how a clinical question about symptom relief can be framed as a time-to-event analysis. Rather than recording only whether relief occurred at a single assessment time, the primary endpoint records when first perceptible relief was reported. The registry pairs a simple descriptive summary, the median time, with Cox-model comparisons that use event timing across the observation period.

It also illustrates why a common median does not settle a comparison. All four posted medians are 30 minutes, but the ranges differ, and the hazard-ratio intervals allow uncertainty in both directions around equal hazards. In general, describing the data and estimating a between-condition contrast are distinct statistical tasks. Neither the median nor the hazard ratio alone gives a complete account of the response-time distribution.

The crossover structure makes reporting of design and analysis details especially important. General statistical considerations include repeated measurements within participants, treatment order, period effects, and the possibility of carryover. The registry calls the study a crossover trial, but the specific design details and their treatment in the model are not reported. Readers can therefore understand the broad design while recognizing that important analytic questions remain unanswered.

The record also shows why a non-significant result should not be mistaken for proof of no difference. The six confidence intervals include 1, but their breadth leaves uncertainty about the relative event rates. The registry does not report an equivalence framework, a detailed multiplicity plan, or a trial-specific sample-size justification. These limits matter when translating statistical results into claims: the posted analyses characterize uncertainty, but do not establish that the treatment conditions are identical or determine their clinical importance.

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.

Several trial-specific details needed to evaluate the primary analysis are not reported in the registry record. Although the design is identified as randomized, double-masked, and crossover, the treatment sequence, number and timing of periods, washout procedures, and approach to possible carryover or period effects are not described. The Cox model is named, but its covariates, censoring rules, handling of repeated observations, and diagnostic checks are not reported.

The primary endpoint population requires randomization, receipt of at least one study medication, and at least one post-baseline efficacy assessment. The registry does not provide a complete participant-flow account or explain why endpoint counts differ by condition. It does not report the amount, reasons, or pattern of missing outcome data, nor sensitivity analyses for missingness or censoring. Those omissions make it difficult to assess whether the available observations represent the intended analysis population.

The registry reports six primary-outcome comparisons and lists nine secondary outcome measures, with eight secondary outcomes posted. The multiplicity strategy and prespecified hierarchy of comparisons are not reported. The record also does not report the protocol’s sample-size justification, interim analyses, stopping rules, or detailed statistical analysis plan. Since status is terminated, the reason for termination and any consequences for planned enrollment or analysis are not reported.

Finally, the outcome is a participant-reported time to first perceptible relief, with a fallback rule based on a score of 1 or greater when the electronic-pad time was unavailable. The registry does not report how often the fallback rule was used or whether results differed under alternative endpoint definitions. The posted medians and hazard ratios consequently provide useful but incomplete information about the outcome and the analysis supporting the comparisons.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Paracetamol With Caffeine to Treat Episodic Tension Type Headache — NCT01755702.
• Registry record last updated 2015-03-27; status last verified 2014-11.
• Results first posted 2015-03-27.
• 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

The registry record describes NCT01755702, a Phase 2 and Phase 3 randomized, double-masked crossover study sponsored by GlaxoSmithKline. It enrolled 66 participants and compared an experimental paracetamol-plus-caffeine formulation with marketed ibuprofen, marketed paracetamol, and placebo for episodic tension-type headache. The recorded status is terminated; the reason is not reported in the registry record.

The primary endpoint was time to first perceptible headache relief from baseline to 6 hours. Each condition had a reported median of 30 minutes. Six Cox proportional hazards comparisons are posted; all confidence intervals include 1, and all p-values exceed 0.05. The registry also records zero serious adverse events in each condition. Important details about crossover procedures, missing data, multiplicity, and the statistical analysis plan are not reported, limiting how fully the estimates can be assessed.