Thiamine as a Metabolic Resuscitator After Cardiac Arrest: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Thiamine as a Metabolic Resuscitator After Cardiac Arrest (NCT03450707), a randomized phase 2 study. The primary outcome is Lactate.

2. Clinical Question

The clinical question concerns a biological and laboratory outcome after cardiac arrest, rather than a single measure of long-term recovery. The trial description says it was designed to investigate thiamine’s effects on lactate, cellular oxygen consumption, global oxygen consumption, and biomarkers of neurologic injury. Those stated aims describe what the study set out to investigate; they do not establish that any effect occurred.

The primary endpoint is a comparison at a specified timepoint, not a description of the entire post-arrest course. Lactate may change over time, so an observed value at 24 hours answers a narrower question than a trajectory across several measurements. The registry also identifies participants who had died before the primary timepoint, which affects whose lactate values contribute to the reported summaries. The estimates therefore need to be interpreted alongside the information about those missing measurements.

3. Trial Design

The registry record describes NCT03450707 as a phase 2, randomized, parallel-group trial with quadruple masking. Its brief summary calls the study randomized, double-blind, and placebo controlled. The planned enrollment was 93 participants. The listed status is terminated; the recorded start date is 2018-05-06, and the primary completion date is 2022-02-19. A reason for termination is not reported in the registry record.

Participants were assigned to thiamine or placebo. The thiamine arm was to receive 500 mg intravenously in 100 mL of normal saline every 12 hours for 5 doses. The placebo arm was to receive 100 mL of normal saline intravenously on the same schedule. The descriptions say other protocol aspects were the same across arms. The study also used a noninvasive monitor to measure global oxygen consumption; the thiamine-arm description specifies monitoring for at least 48 hours or until extubation. Further details about the monitoring procedures are not reported in the registry record.

The primary outcome was blood lactate at 24 hours, expressed in mmol/L. The registry reports medians and inter-quartile ranges for the two arms and also reports a mixed models analysis of final values. The primary summary identifies 17 participants who had died by the 24-hour timepoint and therefore lacked that lactate measurement. Secondary outcomes included global oxygen consumption, lactate-related metabolic measures, mortality, neurologic status, and organ-failure measures.

Enrollment and outcome-specific analysis counts are not interchangeable. The registry reports 93 planned participants, while the number contributing to a particular outcome can vary. For instance, global oxygen consumption and acute renal failure have their own reported arm counts. Detailed eligibility criteria, randomization procedures, allocation concealment, the full protocol, and the reason the trial was terminated are not reported in the registry record. The record also does not describe the intended analysis population or explain whether the study’s analysis followed an intention-to-treat approach.

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

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Seventeen patients (six in the placebo arm and eleven in the thiamine arm) missing 24-hour lactate as they had expired by this timepoint.

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 Lactate, measured over 24 hours.

The primary endpoint is blood lactate at 24 hours, measured in mmol/L. Lactate is a blood marker used to characterize metabolic state. In general, following a major event such as cardiac arrest, changes in lactate can be of interest as a signal of altered metabolism and the body’s response over time. That general rationale does not make lactate a direct measure of neurologic recovery or establish that a difference in lactate would translate into a difference in patient-centered outcomes.

The registry reports the endpoint as a median with an inter-quartile range. A median describes the midpoint of the observed values after ordering them; the inter-quartile range describes the central spread of those values. These summaries are useful when the distribution may not be well represented by a mean and standard deviation. The recorded median was 1.8 mmol/L in the thiamine arm and 2.2 mmol/L in the placebo arm. The respective inter-quartile ranges were 1.2 to 2.4 and 1.4 to 2.5.

The endpoint’s 24-hour timing gives the comparison a defined reference point, but the measurement was unavailable for 17 participants because they had died by then: six in the placebo arm and eleven in the thiamine arm. Thus, the reported lactate summaries describe participants with an available measurement, not all enrolled participants. Death is not simply an unobserved laboratory value; its relationship to the missing measurement matters when interpreting a comparison among observed values.

The registry also reports a mixed models analysis of final values. That model-based mean difference is a different summary from a direct comparison of the reported medians. The endpoint’s summaries and the model estimate therefore answer related, but not identical, statistical questions.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

Statistical methods help distinguish what was observed from what can be inferred. The registry reports two types of information for the primary outcome: arm-level summaries and a model-based comparison. The arm-level values are medians with inter-quartile ranges. The reported inferential analysis is a mixed models analysis of final values. These approaches describe the data in different ways, so the median values should not be substituted for the model estimate or used to reconstruct its confidence interval.

What does the confidence interval communicate?

A confidence interval expresses uncertainty around an estimated quantity. In general, a wider interval indicates less precision than a narrower interval. The primary model estimate is a mean difference of 0.34, with a 95% confidence interval from -1.82 to 2.50. Because the interval spans zero and extends in both directions, the estimate does not sharply distinguish between possible directions of a difference. The interval is not a range containing every plausible individual participant outcome, and it is not proof that the true difference is zero.

The reported p-value, 0.72, is a measure of how compatible the observed result is with a specified null hypothesis under the analysis model. A p-value is not the probability that the null hypothesis is true, nor does it quantify the probability that the result will replicate. It also does not tell readers whether an effect is clinically important. In this trial, the p-value should be read alongside the estimate, the confidence interval, and the endpoint’s missing measurements.

Why do the endpoint summaries and model estimate differ in form?

A median is a rank-based measure of location, whereas a mean difference compares averages. A mixed model can incorporate repeated observations or other modeled structure, but the registry does not describe the model specification, covariance assumptions, included measurements, or handling of missing values. Therefore, the exact relationship between the primary arm summaries and the reported final-value mean difference is not reported in the registry record. No reconstruction of the sponsor’s model is possible from the posted summary statistics alone.

The description of the primary outcome includes 24-hour lactate values and notes deaths before that timepoint. In general, missing data can affect interpretation when the reason a value is missing is related to the outcome or to the treatment group. Death prevents measurement of a later laboratory value and is itself an important outcome. A comparison restricted to participants with a measured lactate is therefore not automatically equivalent to a comparison of all randomized participants. The registry does not state how the primary model addressed those missing values.

How should secondary comparisons be read?

The continuous secondary outcome of global oxygen consumption has means and standard deviations by arm. A mean difference expresses the difference between group averages, while a confidence interval indicates uncertainty around that difference. The posted recalculation uses the arm summaries and is unadjusted; the registry also reports a linear regression analysis. Such simple recalculations do not reproduce a regression model that may use different modeling choices. The recalculated p-value is not adjusted for the other outcomes.

Acute renal failure is reported as a binary outcome, with participant counts in each arm. A risk difference describes the absolute difference between the proportions experiencing the outcome. Its confidence interval reflects uncertainty in that estimated difference. The recalculated comparison is unadjusted, and no sponsor-reported inferential analysis for this outcome appears in the registry record. When counts are limited, estimates can be imprecise; the confidence interval is more informative than treating the point estimate alone as definitive.

For continuous outcomes, the spread of the data matters because it affects precision. For binary outcomes, both event counts and the number assessed contribute to uncertainty. In general, over-dispersion matters when observed variability exceeds what a statistical model assumes; a model that ignores it can misstate uncertainty. The registry does not report a specific over-dispersion assessment for these outcomes, and no claim about such an assessment can be made from the posted record.

What do the planning illustrations mean?

The planning illustrations use observed effects to show hypothetical participants per arm for selected power targets. They are not the trial’s original sample-size justification and are not observed power. The stated assumptions are a two-sided significance level of 0.05, equal allocation, a normal approximation, and an effect equal to a stated fraction of the observed effect. For time-to-event outcomes, the stated convention is a total event count; for other outcomes, it is participants per arm. Because these calculations depend on observed effects, they are illustrations rather than evidence that the trial was adequately or inadequately designed. The trial’s own sample-size assumptions and justification are not reported in the registry record.

Interpretation also depends on multiplicity. Assessing many outcomes creates more opportunities for a nominally unusual result to occur by chance. The record lists 12 secondary outcomes, reports 9 analyses, and does not describe an adjustment strategy for multiple comparisons. In general, an unadjusted p-value answers a single-comparison question and does not account for the number of other outcomes examined. Secondary findings should therefore be considered exploratory unless supported by a prespecified analysis plan and appropriate control of multiplicity; those details are not reported in the registry record.

10. Analyses Reported by the Sponsor

The sponsor-reported primary analysis is a mixed models analysis of final values: mean difference 0.34, 95% confidence interval -1.82 to 2.50, and p = 0.72. This is the registry’s inferential analysis for the primary endpoint. The arm-level endpoint summaries are medians and inter-quartile ranges, so they cannot be used to independently reproduce the reported model-based mean difference. No separate recalculated primary inferential comparison is presented.

For global oxygen consumption, the registry reports a linear regression estimate of -0.38, with a 95% confidence interval from -1.34 to 0.58 and p = 0.43. The unadjusted recalculation from the posted arm summaries gives -0.41, with a 95% confidence interval from -1.388 to 0.568 and an unadjusted p-value of 0.411. These estimates are close but not identical. They arise from different approaches: one is the registry’s regression analysis, while the other is a simple recalculation from the summary statistics. The registry does not report a sponsor analysis for acute renal failure.

The discrepancy flag is false. The recalculations should be understood as checks based on posted summaries, not replacements for the sponsor’s models. Differences in estimates or precision can occur when a model-based analysis and a simple summary-statistic calculation use different methods. The model specifications and full analysis plan are not reported in the registry record, limiting further comparison.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 12 secondary outcomes and reports 9 analyses. These outcomes cover different measures and timeframes, and the record does not state how the analyses were selected or whether a hierarchy was established. A multiplicity adjustment plan is not reported in the registry record. In general, testing many outcomes without accounting for that process increases the chance that at least one result may appear unusual by chance. Accordingly, unadjusted secondary comparisons should not be treated as definitive evidence, particularly when considered separately from a prespecified plan.

Missingness is especially important for the primary endpoint. The registry states that 17 participants lacked 24-hour lactate because they had expired by that timepoint: six in placebo and eleven in thiamine. The primary arm summaries list 29 participants in the thiamine group and 30 in placebo. Since death precludes a 24-hour lactate measurement, the observed lactate values characterize participants with a measurement rather than the full randomized cohort. The record does not explain whether the mixed model incorporated earlier measurements, how it handled death, or what assumptions were used for missing observations.

Other outcomes also have outcome-specific numbers. Global oxygen consumption has reported groups of 12 and 18 participants, while acute renal failure has groups of 39 and 33. These differences mean that the amount of information available varies by outcome. The registry does not explain all reasons for the differing counts, or whether each analysis used all randomized participants with available outcome data.

The intended analysis population, use of intention-to-treat principles, per-protocol analyses, and missing-data sensitivity analyses are not reported in the registry record. Nor is a statistical analysis plan available in the record to clarify prespecified handling of missing values or multiplicity. These omissions restrict assessment of how robust the estimates are to analysis choices. The number enrolled or planned cannot by itself establish which participants contributed to any particular result.

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.

The safety record lists 40 participants at risk in the thiamine group and 36 in the placebo group for serious adverse events. It reports zero participants affected by serious adverse events in either group. For other adverse events, the at-risk counts are also 40 and 36, with zero affected in both arms. The record separately lists 30 deaths in the thiamine group and 24 in the placebo group.

These figures should be read as registry counts, not as a complete description of the timing, causes, severity, or relatedness of events. The definitions and observation period used for adverse-event reporting are not reported in the registry record. The absence of reported serious or other adverse events does not mean that no adverse events of any kind occurred; it means that none were recorded in those categories in this summary.

The death counts are distinct from the adverse-event counts and should not be combined with them or interpreted as evidence that treatment caused or prevented deaths. The registry does not provide a causal analysis of mortality in the safety summary. Differences in the listed at-risk counts and the numbers contributing to other endpoints also caution against treating all counts as if they came from one identical analysis population.

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

The posted results describe a randomized comparison of thiamine and placebo after out-of-hospital cardiac arrest, with 24-hour blood lactate as the primary endpoint. The registry reports arm-level medians and inter-quartile ranges and a model-based comparison of final values. The model estimate is uncertain: its confidence interval includes zero and spans values in both directions. The p-value does not establish equivalence, and the result does not demonstrate that a difference is absent.

The primary endpoint does not represent every dimension of recovery. Lactate is a metabolic measure; the secondary outcomes include other physiological measures, mortality, and neurologic status. The record does not show that a result for one of these measures can be translated into an effect on another. Nor do the data establish the clinical importance of any particular difference in lactate.

Missing 24-hour lactate measurements due to deaths are a central qualification. The reported endpoint summaries are based on participants with measurements, while the deaths occurred before the timepoint. The registry does not explain how this issue was handled in the mixed model. This limits conclusions about how the primary result represents all randomized participants.

The study is recorded as terminated, but the reason is not reported. The trial’s sample-size justification, full statistical plan, detailed model specifications, and analysis population are also not reported in the registry record. The available secondary analyses are limited and do not include a stated multiplicity strategy. These reporting gaps constrain both replication and assessment of how sensitive the results may be to analysis choices.

Accordingly, the results support a cautious description of the estimates and their uncertainty, not a claim that thiamine works or does not work. They also do not establish a treatment effect on mortality, neurologic recovery, or other outcomes merely because those endpoints appear in the registry. The findings are specific to the reported trial, outcomes, timepoints, and analyses.

14. Why This Trial Matters Statistically

This trial illustrates why outcome summaries, model estimates, and missingness need to be interpreted together. The primary endpoint is reported using medians and inter-quartile ranges, while the inferential result is a model-based mean difference. Those statistics describe different features of the data. Treating them as interchangeable could lead readers to infer more than the record supports.

The reason for missing primary measurements is also statistically consequential. The registry identifies deaths before the 24-hour endpoint rather than a generic failure to collect a laboratory sample. In general, when the event that prevents measurement is itself clinically important, an analysis of observed measurements may not answer the same question as an analysis of all randomized participants. Clear reporting of the estimand, missing-data assumptions, and sensitivity analyses is therefore important.

The trial also demonstrates the distinction between an analysis reported by the sponsor and a simple recalculation from arm summaries. Recalculations can help readers understand the scale and uncertainty of a comparison, but they do not reconstruct a regression or mixed model without its specifications and underlying data. Likewise, planning illustrations derived from observed effects are not evidence of observed power or a record of the original design rationale.

Finally, a broad set of secondary outcomes can offer useful context while increasing the importance of prespecification and multiplicity control. Without an available analysis plan, readers cannot determine how the secondary tests were prioritized. The registry’s status and outcome reporting make transparent documentation especially relevant: a complete account of analysis methods and termination circumstances would help distinguish planned inference from exploratory findings.

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 limitations are specific to the available record. The trial is listed as terminated, but no reason for termination is reported. The recorded enrollment is 93, while outcome-specific counts vary. That variation affects the information available for each comparison and means the enrollment figure should not be treated as the analysis count for every outcome.

The primary endpoint is missing for 17 participants because they had died by the 24-hour timepoint. The registry does not report the missing-data method used in the mixed model, the model’s assumptions, or sensitivity analyses addressing those deaths. The primary arm-level medians describe measured values and cannot by themselves characterize participants without a measurement.

The statistical plan is incompletely described. The model specifications, analysis population, sample-size justification, multiplicity approach, and detailed procedures for missing observations are not reported in the registry record. Those omissions limit independent reproduction and make it difficult to assess whether the analyses were prespecified. The secondary outcomes also have unequal counts, and some listed outcomes do not have a reported two-arm analysis.

The registry’s summary of the thiamine arm describes monitoring for at least 48 hours or until extubation, while protocol detail beyond the abbreviated arm descriptions is limited. The safety record gives counts but not event timing, causes, definitions, or relatedness. The record also does not explain the relationship between the different at-risk populations used for safety and the populations contributing to efficacy outcomes. These reporting limits should temper interpretation of both the outcome estimates and safety counts.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Thiamine as a Metabolic Resuscitator After Cardiac Arrest — NCT03450707.
• Registry record last updated 2023-08-24; status last verified 2023-08.
• Results first posted 2023-08-24.
• 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

NCT03450707 is a phase 2 randomized, parallel-group trial of intravenous thiamine versus placebo after out-of-hospital cardiac arrest. The registry lists quadruple masking and a planned enrollment of 93. The recorded status is terminated, with a primary completion date of 2022-02-19; the reason for termination is not reported in the registry record.

The primary endpoint was blood lactate at 24 hours. The registry reports medians and inter-quartile ranges by arm and a mixed models analysis of final values. Seventeen participants lacked a 24-hour lactate measurement because they had died by that timepoint. The reported model estimate is imprecise, and the missing measurements and unreported model details constrain interpretation.

Secondary outcomes cover metabolic measures, organ function, mortality, and neurologic status. The record lists 12 secondary outcomes and reports 9 analyses, without a reported multiplicity plan. Outcome counts differ, and the trial’s full analysis plan, sample-size justification, analysis population, and reason for termination are not reported. Overall, the posted record supports cautious description of the available estimates, not a definitive conclusion about treatment effects.