Evaluation of TNF-α Blockade Effect in Patients With Severe Cutaneous Adverse Drug Reactions: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Evaluation of TNF-α Blockade Effect in Patients With Severe Cutaneous Adverse Drug Reactions (NCT01276314), a randomized na study. The primary outcome is Skin Healing Time.

2. Clinical Question

The registry describes a randomized, unmasked, parallel-group study with enrollment of 135. The primary outcome is separated by diagnosis rather than presented as one combined comparison: anti-TNF-α treatment versus control in SJS/TEN, and the same treatment comparison in DRESS. This distinction matters because the reported healing-time distributions differ substantially between the two diagnostic groups, and the study specifies different expected follow-up windows for them.

The study’s summary frames the clinical question as a comparison of systemic steroid treatment and anti-TNF-α treatment. Yet the record does not give a sufficiently detailed description of the control medication to establish exactly what participants in that group received. That gap limits interpretation of the treatment contrast. The record also describes molecular analyses as a study aim, but no secondary outcomes are posted.

3. Trial Design

The record describes a randomized, unmasked, parallel-group trial sponsored by Chang Gung Memorial Hospital. Its status is completed. The listed phase is NA. Enrollment is 135. The record gives a start date of 2009-01 and primary completion of 2015-05.

Participants were classified as having SJS/TEN or DRESS using criteria and histopathological examinations. The registry reports skin-healing results separately for each diagnosis and treatment group. For SJS/TEN, the anti-TNF-α group has 38 participants in the outcome listing and the control group has 33. For DRESS, the anti-TNF-α group has 18 and the control group has 17. The record does not explain how these outcome group counts relate to the enrollment figure or provide a flow of participants through enrollment, treatment, and analysis.

The anti-TNF-α arm description names etanercept. Both arm descriptions include case-report forms, blood testing, physiological assessment, measurement of TNF-α serum concentration, and peripheral blood mononuclear cell cDNA expression analysis. The control-arm description does not identify the administered drug. Prednisolone is listed among the interventions, and the study summary describes a comparison between systemic steroid and anti-TNF-α treatment, but the precise control regimen is not reported in the registry record.

Randomization can help make treatment groups comparable on average, while an unmasked design means participants and study personnel were not blinded according to the record. The method used to generate the allocation sequence, conceal allocation, or maintain treatment separation is not reported in the registry record. The record also does not report the planned allocation ratio, eligibility criteria in detail, or the statistical design assumptions. Its study aims extend beyond healing time to internal organ recovery, mortality, adverse events, and molecular mechanisms, but no secondary outcomes are posted.

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: We totally recruited 140 participants, but only 135 SCAR participants (including 91 SJS/TEN patients and 44 DRESS patients) were enrolled to analyze. 5 participants in the control group were not included to analyze (3 cases failed a screening with CD4\<200 cells/mm3 and 2 cases as diagnosis change).

Overall Study

Analysis population for the primary outcome, as defined in the registry: We determined if the enrolled participants had SJS/TEN and DRESS using the criteria and histopathological examinations.

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: A total of 135 SCAR patients (including 91 SJS/TEN cases and 44 DRESS cases) were enrolled to analyze.

6. Primary Endpoint

The registry lists the primary outcome as Skin Healing Time, measured over One to two months for SJS/TEN cases, and one to six months for DRESS cases..

The primary endpoint is skin healing time, measured in days. Healing is defined as complete re-epithelialization, or the complete absence of erosions. The study recorded the time taken for skin to heal. The stated timeframe is one to two months for SJS/TEN cases and one to six months for DRESS cases.

Time to healing captures not only whether an event occurs, but also when it occurs. A median gives the central point of the observed time distribution: it describes the time by which half of the relevant event-time distribution has been reached, subject to the analysis method and its handling of incomplete follow-up. A time-to-event approach can make use of participants whose healing time is not observed within the available follow-up, provided their follow-up and censoring are handled appropriately. The registry does not describe those details.

This endpoint is relevant to the study’s stated comparison because it represents the time until a defined skin-recovery milestone. It is distinct from mortality, adverse events, and internal organ recovery, which are also mentioned as study aims but are not posted as secondary outcomes in the registry record. The record gives medians and full ranges by group, along with Kaplan–Meier analyses, but does not give confidence intervals for the medians or a hazard-ratio estimate.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary endpoint is a time-to-event outcome: the event is complete skin healing, defined as complete re-epithelialization and absence of erosions. The registry gives medians and full ranges for each treatment group within SJS/TEN and DRESS. A median is a summary of the center of a distribution, while a full range gives the lowest and highest reported times. A range shows the spread of observed values but does not indicate where most participants’ times lie. It can also be sensitive to extreme observations. The record does not provide other distribution summaries such as quartiles.

The sponsor’s reported method is Kaplan–Meier analysis. In general, the Kaplan–Meier estimator describes the probability of remaining event-free over time while accounting for right-censored observations. Censoring occurs when an event time is not observed during a participant’s available follow-up, for example because observation ends first. The method can use the information available up to that point, rather than treating an incomplete event time as though the event had occurred at the end of follow-up. The registry does not report how many observations were censored, how censoring was defined, or whether follow-up differed between groups.

In general, a Kaplan–Meier comparison may be accompanied by a test of whether the event-time patterns differ between groups. The record reports p = 0.01 for the SJS/TEN comparison and p = 0.06 for the DRESS comparison. It labels the analyses Kaplan–Meier but does not identify the exact test, its assumptions, or any adjustment for other factors. It also does not report hazard ratios or confidence intervals. A p-value alone does not quantify the size of a time-to-event contrast.

A hazard describes the instantaneous event rate among people who have not yet experienced the event, conditional on remaining under observation. A hazard ratio compares hazards between groups. In general, interpretation of a single hazard ratio often relies on assumptions about how the relative hazards behave over time. The record does not post a hazard ratio for either comparison, so no hazard-ratio interpretation is available. A hazard ratio would also be distinct from comparing the group medians: the two summaries answer different questions and need not convey the same aspect of the event-time distributions.

Confidence intervals express statistical uncertainty around an estimate under the model and assumptions used. They can show a range of effect sizes compatible with the data and are useful for evaluating precision. No confidence intervals are reported for the posted medians or Kaplan–Meier comparisons. The full ranges are not confidence intervals; they describe the span of observed times. Without an uncertainty interval or the underlying event-time information, the precision of the medians and the comparative effect cannot be reconstructed from the registry record.

Randomization is intended to reduce systematic differences between groups at the point of assignment, on average. The record says that the study was randomized, but does not report the sequence-generation method, allocation concealment, or assignment ratio. It is unmasked, so the record does not indicate blinding of participants or study personnel. In general, awareness of assignment can affect care, follow-up, or assessments, depending on the outcome and study procedures. The record does not describe safeguards for outcome assessment.

The registry reports separate analyses for SJS/TEN and DRESS, which preserves the stated diagnostic groups and avoids treating them as one uniform population. Their follow-up windows differ: one to two months for SJS/TEN and one to six months for DRESS. Those windows describe the listed timeframe; the record does not explain actual follow-up duration, completeness, or how the endpoint was analyzed if healing was not observed within the timeframe.

In general, time-to-event analyses require clear definitions of the event, time origin, follow-up, censoring, and analysis population. The event definition is given, but the time origin and detailed follow-up procedures are not reported in the registry record. Nor does the record describe whether participants were analyzed according to randomized assignment, how treatment changes were handled, or how missing outcome information was addressed. These omissions limit independent assessment of the analysis.

The posted planning illustrations use assumptions stated in the registry materials, including a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. They are planning exercises based on a fraction of an observed effect, not evidence about the trial’s original design or its achieved power. The registry’s sample-size justification and any prospective power calculation are not reported in the registry record. These planning illustrations do not replace the observed results or provide an estimate of uncertainty in them.

10. Analyses Reported by the Sponsor

The registry reports the sponsor’s Kaplan–Meier analyses for the two diagnostic comparisons: p = 0.01 for SJS/TEN and p = 0.06 for DRESS. It does not report sponsor hazard ratios or confidence intervals. The group medians and full ranges provide descriptive summaries, but they do not replace a comparative effect estimate with an uncertainty interval.

No recalculated two-arm comparison is available for this median-time endpoint in the posted results. The data are marked as missing or not estimable for that kind of unadjusted recalculation. Consequently, there is no independent recalculated estimate to compare with the sponsor’s analysis. The record does not flag a discrepancy between sponsor and recalculated results.

When a discrepancy is flagged, the sponsor’s own analysis takes precedence over a simple unadjusted recalculation from summary statistics. That situation does not apply to the posted result for this endpoint. The appropriate reading is therefore to retain the sponsor-reported Kaplan–Meier p-values as reported, while recognizing that the registry does not give the associated effect estimates, confidence intervals, or enough analysis detail to assess the comparisons fully.

11. Multiplicity, Missing Data and Analysis Populations

The registry posts one primary outcome measure, skin healing time, and two reported analyses: one for SJS/TEN and one for DRESS. It reports no secondary outcomes. The study summary nevertheless describes aims involving beginning of re-epithelialization, internal organ recovery, mortality, adverse events, and molecular mechanisms. Whether these were prespecified endpoints, exploratory aims, or analyzed outcomes is not reported in the registry record.

Multiplicity arises when several outcomes, groups, time points, or statistical tests are examined. In general, testing more hypotheses can increase the chance of at least one apparently notable result under a global null model. Researchers may use a prespecified testing hierarchy or adjustment procedure to address this issue. The registry does not report whether the two diagnosis-specific analyses were treated as separate primary comparisons, whether any multiplicity adjustment was planned, or how the p-values should be interpreted in relation to other analyses.

Missing-data handling is also not reported in the registry record. The endpoint is time to healing, for which incomplete follow-up may be represented through censoring in a time-to-event analysis. The registry does not report censoring counts, reasons for loss to follow-up, missing outcome information, or sensitivity analyses. It therefore cannot be determined from the posted record how incomplete observation affected the estimates.

The analysis populations are not described. The record does not say whether analyses followed randomized assignment, included only participants who received treatment, or applied other eligibility rules after randomization. It also does not give participant flow, exclusions from analysis, or reasons for any differences between enrollment and outcome group counts. These details matter for understanding how the randomized design relates to the posted estimates.

The registry record does not report a statistical analysis plan, interim analyses, stopping rules, or sample-size justification. Their absence from the record should not be taken as evidence that such procedures were or were not used. It means the posted information is insufficient to assess prespecification and analysis choices beyond the outcomes and methods explicitly described.

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 0% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: respiratory failure (11.6%), sepsis (9.3%), upper gi hemorrhage (9.1%), hyperglycemia (4.5%), stridor, vocal cord palsy (2.3%).

The registry lists serious and other adverse-event counts by diagnostic and treatment group. For serious events, it reports 5 affected among 48 at risk in anti-TNF-α-treated SJS/TEN, 9 among 43 in control SJS/TEN, 2 among 22 in anti-TNF-α-treated DRESS, and 5 among 22 in control DRESS. These are the registry’s counts; they are not recalculated rates.

For other adverse events, the posted affected counts are 15 among 48, 19 among 43, 8 among 22, and 8 among 22 for those same groups, respectively. The registry does not explain why these safety denominators differ from the group counts listed for the healing-time outcome, nor does it report the safety analysis populations or follow-up period in detail. The figures should therefore be interpreted using their stated at-risk denominators, without assuming they represent the same analysis set as the primary endpoint.

The record lists respiratory failure (11.6%), sepsis (9.3%), upper GI hemorrhage (9.1%), hyperglycemia (4.5%), stridor, vocal cord palsy (2.3%), and bipolar disorder (2.1%) among the top serious terms. It does not provide event-level timing, severity grading, attribution, or a treatment-relatedness assessment for these terms. Death counts are not reported in the posted safety data. These counts and terms describe reported events; they do not establish that treatment caused them or permit a complete comparative safety assessment.

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

The posted results describe skin-healing times in two diagnostic groups, with separate treatment comparisons. In SJS/TEN, the anti-TNF-α group’s median is 13.75 days and the control group’s median is 19.0 days; the reported Kaplan–Meier p-value is 0.01. In DRESS, the medians are 40.5 days and 34.0 days, respectively, with a reported p-value of 0.06. The full ranges are broad, particularly in DRESS. These numbers summarize the registry record, not a complete account of each participant’s clinical course.

The results do not establish the size or precision of a treatment effect. The registry does not report hazard ratios or confidence intervals, and a p-value is not an effect-size estimate. The medians are descriptive summaries; they do not reveal the full shape of the time-to-healing distributions. Nor do they establish whether any observed difference is clinically important.

The findings should not be treated as evidence that treatment does or does not work. The record lacks important details about the control regimen, allocation procedures, analysis populations, missing data, and outcome follow-up. The study is unmasked, and the record does not describe how outcome assessment was protected from awareness of assignment. These limitations affect how confidently the reported comparisons can be interpreted.

The results also do not answer all the questions raised in the study summary. Although the summary mentions internal organ recovery, mortality, adverse events, and molecular mechanisms, no secondary outcomes are posted. The safety data offer counts and serious-event terms, but not enough detail to establish comparative risks or attribution. Death counts are not reported in the posted safety data.

The most defensible conclusion is descriptive: the registry reports different median healing times and Kaplan–Meier p-values for SJS/TEN and DRESS, with the comparisons pointing in different directions by median. Further interpretation requires effect estimates with uncertainty, details of the analysis plan and populations, and fuller reporting of follow-up and safety outcomes. These results alone do not provide medical advice.

14. Why This Trial Matters Statistically

This trial is statistically notable because it reports a time-to-event endpoint and analyzes it separately in SJS/TEN and DRESS. That structure highlights a key principle of clinical analysis: the population and outcome definition determine what a comparison means. A single pooled summary could obscure the different descriptive patterns across the two diagnoses, while the diagnosis-specific results remain limited by their available sample sizes and the lack of uncertainty intervals.

The record also illustrates why a p-value should not be read in isolation. The sponsor reports p = 0.01 for SJS/TEN and p = 0.06 for DRESS, but neither value communicates the magnitude or precision of the comparison. The group medians and full ranges provide useful descriptive context, yet no hazard ratios or confidence intervals are posted. Statistical evidence is better assessed through effect estimates, uncertainty, design, and prespecified analysis details together.

For time-to-event outcomes, the distinction between an observed healing time and incomplete follow-up is important. Kaplan–Meier methods can account for censoring under appropriate conditions, but the registry does not report censoring or follow-up details. This leaves readers unable to assess how much information contributed to each comparison or whether the analysis assumptions are plausible.

Finally, the record demonstrates the value of transparent reporting. Randomization and a defined healing endpoint are useful design features, but interpretation also depends on treatment descriptions, participant flow, analysis populations, multiplicity, missing-data procedures, and uncertainty estimates. Several of those details are not reported in the registry record. The statistical lesson is not a clinical verdict; it is that the strength of interpretation depends on more than a test result.

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 registry record does not fully specify the control treatment. Prednisolone is listed as an intervention and the study summary describes systemic steroid treatment, but the control-arm description identifies only “Drug admin.” This makes the exact treatment contrast difficult to characterize. The record also does not report the dose, duration, or administration details for the named drugs.

The trial is described as randomized and unmasked, but sequence generation, allocation concealment, and allocation ratio are not reported. The record does not describe blinding of outcome assessors or procedures to limit assessment bias. Because healing is a clinical endpoint, details about assessment timing and consistency would help readers understand how the endpoint was determined.

For the primary endpoint, the registry gives medians, full ranges, and Kaplan–Meier p-values but no hazard ratios or confidence intervals. It does not identify the exact statistical test or provide details of censoring, time origin, follow-up completeness, or handling of participants whose healing was not observed. Without these elements, the effect and its precision cannot be assessed comprehensively.

Participant flow and analysis populations are not reported. The record does not explain how outcome group counts relate to enrollment, whether participants were analyzed according to randomized assignment, or why safety at-risk counts differ from healing-outcome group counts. Missing-data methods, sensitivity analyses, and multiplicity procedures are also not reported in the registry record.

The registry describes additional study aims, including recovery outcomes, mortality, adverse events, and molecular mechanisms, but posts no secondary outcomes. Safety reporting includes counts and selected serious-event terms, without event timing, attribution, or death counts. The record does not provide the original sample-size justification, interim-analysis details, or statistical analysis plan. These limitations are specific to the available trial record and restrict conclusions that can be drawn from the posted results.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Evaluation of TNF-α Blockade Effect in Patients With Severe Cutaneous Adverse Drug Reactions — NCT01276314.
• Registry record last updated 2017-12-19; status last verified 2014-12.
• Results first posted 2017-12-19.
• 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

NCT01276314 is a completed, randomized, unmasked, parallel-group trial sponsored by Chang Gung Memorial Hospital. It enrolled 135 participants and lists Drug Hypersensitivity as the condition. The study compared anti-TNF-α treatment, with etanercept named in the arm description, against a control group. Prednisolone is listed as an intervention, but the specific control regimen is not reported in the registry record.

The primary outcome is skin healing time, defined as time to complete re-epithelialization with complete absence of erosions. Results are posted separately for SJS/TEN and DRESS. The registry reports group medians, full ranges, and Kaplan–Meier p-values, but no hazard ratios or confidence intervals. It posts no secondary outcomes.

The record gives a start date of 2009-01 and primary completion of 2015-05. Important details about allocation procedures, follow-up, analysis populations, missing data, multiplicity, and the statistical plan are not reported in the registry record. The posted results support a descriptive account of the reported groups, but the available record does not establish the magnitude or precision of a treatment effect.