Comparison of Contrast Agents in Liver Magnetic Resonance (MR) for the Detection of Hepatic Metastases: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Comparison of Contrast Agents in Liver Magnetic Resonance (MR) for the Detection of Hepatic Metastases (NCT04973007), a randomized phase 4 study. The primary outcome is Preference of Radiologists for the Hepatobiliary Phase Images Generated by MultiHance (amHBP) Versus Eovist (aeHBP)..

2. Clinical Question

This distinction matters when interpreting the result. A preference count captures readers’ relative image judgments under the evaluation described in the registry. It is not itself a measure of whether a lesion was correctly detected or excluded, and it does not establish that one imaging approach is clinically comparable to the other. The registry lists diagnostic performance, sensitivity and specificity, imaging time and cost among secondary outcome topics, but the primary result is radiologist preference.

3. Trial Design

The registry describes a Phase 4, randomized, single-masked, crossover trial, with enrollment of 10 and current status recorded as terminated. It lists the University of Colorado, Denver as sponsor. The start date is 2021-06-22 and the primary completion date is 2025-05-06. These record details describe the trial’s status and timeline, not the reason it was terminated; a reason is not reported in the registry record.

In the crossover design, each participant was to undergo two complete-protocol liver MR exams: one with gadoxetate disodium and one with gadobenate dimeglumine. The order was randomized. The two sequence groups are described as gadoxetate first followed by gadobenate, or gadobenate first followed by gadoxetate. The description for the reverse-order sequence specifies an interval of 3-20 days. The interval is not fully stated in the other sequence description.

A crossover comparison uses paired observations from the same participant, rather than treating the two imaging exams as unrelated groups. Pairing can help focus a comparison on within-participant differences, but order and timing may also matter. The registry’s primary outcome description says the image pairs were evaluated by three blinded radiologists. It does not give further detail about how reader-level and participant-level dependence was handled in the analysis results.

The posted primary result concerns 30 image evaluations from the described analysis population. The registry reports 25 evaluations preferring the Eovist images and 5 preferring the MultiHance images. These are evaluation counts, not counts of participants. The record does not report participant flow, the number completing both exams, reasons for missing evaluations, or the rationale for termination. It also does not report detailed eligibility criteria, masking procedures, or a full statistical analysis plan.

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: Up to 15 subjects who meet study eligibility criteria and who have exceptional circumstances will be evaluated as part of the PI Discretion Cohort. These subjects may bypass randomization and proceed with gadoxetate disodium scan followed by gadobenate dimeglumine to minimize clinical disruption.

Overall Study

Analysis population for the primary outcome, as defined in the registry: The analysis population included 10 participants with paired hepatobiliary phase liver MR image sets acquired using Eovist (gadoxetate disodium) and MultiHance (gadobenate dimeglumine). Three blinded radiologists. Independently evaluated each paired image set, resulting in 30 image evaluations; participant counts are reported separately.

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: 3 participants randomized to Gadobenate Dimeglumine exam first. 3 participants randomized to Gadobenate Dimeglumine exam first. 4 patients were not randomized due to clinical scheduling however underwent Gadoxetate disodium exam first

6. Primary Endpoint

The registry lists the primary outcome as Preference of Radiologists for the Hepatobiliary Phase Images Generated by MultiHance (amHBP) Versus Eovist (aeHBP)., measured over 1 month.

The primary endpoint is titled “Preference of Radiologists for the Hepatobiliary Phase Images Generated by MultiHance (amHBP) Versus Eovist (aeHBP).” Its unit is evaluations preferring an image, and its parameter is a number of evaluations. For each paired comparison, radiologists indicated which hepatobiliary-phase image they preferred. The analysis population is described as 10 participants with paired image sets and three blinded radiologists, who independently evaluated each paired set, resulting in 30 image evaluations.

This endpoint is used to summarize readers’ comparative judgments of the images. A preference measure can capture perceived image quality or usefulness as judged by readers, but the registry does not define preference as a diagnostic decision or report it as a direct measure of lesion detection. The record lists separate secondary topics for diagnostic performance, sensitivity and specificity, imaging time, and cost. Those concepts should not be substituted for the primary preference outcome.

The posted counts are 25 evaluations preferring Eovist images and 5 preferring MultiHance images. The registry gives a timeframe of 1 month. It does not report an effect estimate with a confidence interval for this endpoint. The description mentions ordinal response mixed effect models that included right/left image position as a covariate; the posted comparison is identified as Chi-squared. The record does not provide enough detail to reconcile these descriptions or establish that the reported count comparison represents the stated model.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The statistical information in the registry includes a description of an ordinal-response mixed effect model and a posted Chi-squared analysis. These methods answer related but not identical questions and rely on different assumptions. The record does not provide enough detail to determine how the stated model relates to the reported p-value, so they should be distinguished rather than treated as interchangeable.

What does the reported p-value mean?

The reported Chi-squared analysis has p = 0.005. In general, a p-value describes how unusual data at least as inconsistent with a specified null model would be if that model and the test assumptions held. It is not the probability that the null model is true, and it does not quantify the magnitude of the preference pattern. It also does not establish clinical relevance.

A Chi-squared test compares observed counts with counts expected under a null hypothesis. Its interpretation depends on the comparison being tested and on the assumptions used to calculate the test statistic. The registry identifies the method and p-value but does not state the null hypothesis in detail, give the test statistic, or report a confidence interval. It also does not explain how the paired image sets and repeated reader evaluations were represented in this test.

Why does pairing matter?

In general, a paired design compares measurements linked within the same experimental unit. Here, each participant is described as having images acquired with both contrast agents, and radiologists evaluated paired image sets. Such observations are not necessarily independent: two images from the same participant share characteristics, and repeated evaluations by the same radiologist may also be related. A method that accounts for those links can represent uncertainty differently from a method that treats every evaluation as independent.

The registry’s endpoint description says ordinal response mixed effect models were used and that right/left image position was included as a covariate. In general, mixed effect models can account for grouped or repeated observations by representing variation associated with clusters, such as readers or participants. A covariate for image position can address a systematic position effect when image placement might influence responses. The record does not give the model specification, random effects, coefficient estimates, uncertainty intervals, or details of how the model produced the posted result.

How should the endpoint scale be understood?

The endpoint is a count of evaluations preferring an image. The description also refers to ordinal responses, which ordinarily means responses have an order. The record does not list the full response scale or say whether responses beyond a preference for one image were possible. The reported values are counts: 25 evaluations preferring Eovist images and 5 preferring MultiHance images. These counts should not be mistaken for participant totals or diagnostic classifications.

As a general principle, an effect estimate describes the size and direction of a comparison, while a confidence interval describes the uncertainty around that estimate under a specified method. The record gives the counts and p-value but no confidence interval or model-based estimate. The precision of the result therefore cannot be evaluated from an interval. A p-value alone does not show the range of effect sizes compatible with the data.

What can be inferred from the analysis description?

The population description states that three blinded radiologists independently evaluated paired image sets from 10 participants, resulting in 30 image evaluations. It does not report whether the statistical analysis treated each reader evaluation as independent, modeled clustering by participant or reader, or used another approach. The contrast between the stated mixed effect model and the posted Chi-squared method is therefore an important reporting limitation. The registry does not report enough information to resolve it.

Multiplicity is another general consideration when a study assesses several outcomes or analyses. Testing multiple questions can increase the chance of at least one apparently notable result if no adjustment or prespecified hierarchy is used. The registry lists 9 outcome measures and 8 analyses, and lists several secondary outcome topics. It does not report whether an adjustment for multiple testing was used, whether the primary analysis was protected by a testing hierarchy, or how the reported p-value should be interpreted in relation to other analyses.

Missingness can also affect interpretation. A complete-pair analysis may differ from an analysis that includes incomplete records, depending on why observations are absent. The registry does not report missing image evaluations, participant disposition, or a missing-data method. It describes the analysis population but does not give a full account of how that population was formed.

The record includes sample-size planning assumptions in general terms: a two-sided significance level of 0.05, equal allocation, a normal approximation, and effects framed as fractions of the observed effect. Such assumptions describe planning illustrations, not observed power. No usable two-arm planning values are reported for this count endpoint because it is not reduced to a single comparison. In general, a planning illustration is not evidence that the trial had adequate power, nor is it a substitute for an uncertainty interval around the observed result.

10. Analyses Reported by the Sponsor

The registry reports a Chi-squared analysis for the primary preference outcome, with p = 0.005. The posted counts are 25 evaluations preferring Eovist hepatobiliary-phase images and 5 preferring MultiHance images. No recalculated effect estimate or confidence interval is available for this count endpoint. The registry’s analysis therefore remains the relevant reported inferential result; the count values can be described directly, but they do not support a separate recalculated estimate in the information recorded.

The endpoint description also refers to ordinal response mixed effect models, including right/left image position as a covariate. The record does not explain whether the reported Chi-squared result was derived from that model, was a separate analysis, or reflects a different analysis approach. Without model estimates, test details or an analysis plan, the relationship between the description and the reported result cannot be resolved.

The p-value should be interpreted as the registry’s reported test result, not as a measure of effect size or clinical importance. No confidence interval is reported. The observed preference counts concern reader evaluations and do not establish diagnostic performance, equivalence, cost savings or patient benefit.

11. Multiplicity, Missing Data and Analysis Populations

The record lists 9 outcome measures and 8 analyses, and names secondary topics including diagnostic performance, sensitivity and specificity, test validity, imaging time and cost. The primary result addressed is radiologist preference for paired hepatobiliary-phase images. The registry does not report whether the primary endpoint was assigned a formal testing hierarchy or whether the analyses of multiple outcomes were adjusted for multiplicity.

In general, when multiple outcomes or comparisons are tested, the chance of obtaining at least one small p-value can increase if the tests are treated as independent opportunities to reject a null hypothesis. Adjustment procedures or prespecified hierarchies can address that issue, but the record does not state whether either was used. The reported p = 0.005 should therefore be read as the result of the named Chi-squared analysis, without assuming an unreported multiplicity adjustment or a particular family of tests.

The endpoint description identifies an analysis population of 10 participants with paired image sets and three blinded radiologists, resulting in 30 evaluations. The record does not give a participant flow account, state whether all enrolled participants contributed complete paired images, or explain whether any evaluations were excluded. A missing-data strategy and reasons for missing observations are not reported in the registry record.

The posted results do not include an uncertainty interval for the preference counts. Nor does the registry report a detailed statistical analysis plan, a protocol for handling incomplete pairs, or a full account of how repeated evaluations by readers were modeled. These omissions limit assessment of the analysis population, the effects of missingness, and the reliability of the reported p-value under the design’s repeated-measure structure.

The record also does not report interim analyses or a sample-size justification. Their presence or absence should not be inferred from the listed enrollment or from the trial’s terminated status. The reason for termination and its relation, if any, to missing data or analysis decisions are not reported in the registry record.

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 registry lists serious adverse events as 0 in both randomized sequence groups, and deaths as 0 in both groups. For the gadoxetate-first, gadobenate-second sequence, the serious-event and other-event risk counts are each 7. For the gadobenate-first, gadoxetate-second sequence, both risk counts are 3. These are safety summaries by exam sequence, not separate estimates of risk attributable to each contrast agent.

The risk counts differ from the stated analysis population of 10 participants, and the record does not explain the difference. It does not report why participants were included in one safety denominator rather than another, or provide further detail about the safety observation period. No serious adverse-event terms are listed. The reported zeros mean that no events of those categories are recorded in these summaries; they do not establish that the agents have no risk.

Safety interpretation is limited by the small recorded risk groups and the absence of event descriptions or uncertainty estimates. These figures should be read as registry-reported counts for the listed sequence groups, not as a complete characterization of safety or as a comparison that isolates the effects of the contrast agents.

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

The primary result describes radiologists’ preferences among paired hepatobiliary-phase images. The record reports 25 evaluations preferring Eovist images and 5 preferring MultiHance images, with a Chi-squared p-value of 0.005. In this set of evaluations, the stated preferences were more often for the Eovist images. That is the direct meaning of the posted result.

The result does not directly answer whether either imaging method detects hepatic metastases more accurately. It does not provide sensitivity, specificity, diagnostic validity, or lesion-level results. Although those topics appear among the secondary outcomes, their results are not reported in the information described for this analysis. Image preference is not a substitute for diagnostic performance.

Nor does the result establish that the abbreviated gadobenate protocol is clinically comparable to standard-of-care gadoxetate imaging. A conclusion of comparability would require an appropriate comparison and a stated criterion for what degree of difference is acceptable. The record does not report such a criterion or a confidence interval for the primary result. A nonsignificant test, even if one had been reported, would not by itself establish equivalence; in this case, the reported p-value is not a test of equivalence.

The result also does not demonstrate savings in time, cost or patient effort. Those are motivations described in the brief summary, and imaging time and cost appear as secondary outcome topics. The record does not give their results in the registry record described for this analysis. No patient outcome benefit can be inferred from a count of reader preferences.

General interpretation of a p-value must remain separate from clinical importance. The reported p-value indicates compatibility with a null comparison under the named Chi-squared test and its assumptions; it is not the probability that the null is true and does not quantify the size of preference. The lack of a confidence interval and incomplete explanation of dependence among repeated evaluations limit assessment of precision.

The registry records a terminated trial with enrollment of 10. It does not report the reason for termination or whether the available results represent the planned analysis in full. The preference result is therefore best understood as a limited finding from the evaluated image pairs, rather than a broad conclusion about imaging practice or patient care.

14. Why This Trial Matters Statistically

This trial illustrates why the endpoint determines what a statistical result can support. The registry’s primary measure is the number of reader evaluations preferring each image set. That measure can describe a reader judgment, but it is not the same as diagnostic accuracy or clinical outcome. Keeping those concepts separate prevents a statistically notable preference result from being interpreted as proof of better detection or patient benefit.

The crossover structure also makes dependence important. Each participant is described as receiving both imaging agents, and multiple radiologists evaluated paired image sets. These linked observations can contain information that a simple count comparison does not fully express. The registry describes a mixed effect model accounting for image position, yet identifies the posted test as Chi-squared. The difference highlights the value of reporting precisely how pairing and repeated reader evaluations enter the analysis.

Counts and p-values answer different questions. The counts communicate the observed direction of stated preferences; the reported p-value summarizes compatibility with a null comparison under the named test. Neither alone gives a confidence interval or establishes clinical importance. The absence of an uncertainty interval, detailed model output and multiplicity information leaves key aspects of precision and inferential context unclear.

Finally, the study’s motivation concerns potential clinical comparability and possible reductions in time, cost and effort, while its primary result concerns image preference. The registry lists related secondary topics, but the record does not report those results. This makes the trial a useful example of the distinction between a study’s motivating claim and the specific endpoint for which a result is posted.

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 analysis population is described as 10 participants, with 30 image evaluations by three blinded radiologists. The trial is recorded as terminated. The registry does not report why it was terminated, whether all planned data collection was completed, or how many participants completed both exams. This makes it difficult to assess the completeness of the primary analysis from the record alone.

The primary endpoint description refers to ordinal response mixed effect models with image position as a covariate, whereas the posted analysis is identified as Chi-squared. The record does not explain the relation between these methods, provide model estimates, or describe how observations linked by participant and reader were handled. Because the same participants and readers contribute related evaluations, the missing detail is important for judging the reported uncertainty.

The result has no reported confidence interval, and the registry does not report a detailed analysis plan, missing-data method, multiplicity adjustment, or hierarchy across outcomes. It lists multiple outcomes and analyses, but the p-value’s status relative to those tests is not clarified. A sample-size justification is also not reported in the registry record.

The primary outcome measures image preference, not diagnostic accuracy. Results for the listed secondary topics of sensitivity, specificity, test validity, imaging time and cost are not included in the information described for this analysis. The record also gives limited details about eligibility criteria, image assessment procedures, and the clinical characteristics of participants. These limitations restrict both statistical interpretation and generalizability beyond the evaluated image sets.

Safety denominators are reported by randomized sequence and differ from the analysis population description. The record does not explain this difference or give detailed event observation information. The reason for termination, participant disposition, and the relationship of those issues to the reported data are not reported in the registry record.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Comparison of Contrast Agents in Liver Magnetic Resonance (MR) for the Detection of Hepatic Metastases — NCT04973007.
• Registry record last updated 2026-09-25; status last verified 2026-08.
• Results first posted 2026-09-25.
• 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

NCT04973007 is a Phase 4 randomized, single-masked crossover trial sponsored by the University of Colorado, Denver. It compares liver MR images acquired using gadoxetate disodium and gadobenate dimeglumine, with randomized exam order. The registry records enrollment of 10, status as terminated, a start date of 2021-06-22 and a primary completion date of 2025-05-06.

The primary endpoint counts radiologist evaluations preferring paired hepatobiliary-phase images. The record describes 10 participants and three blinded radiologists, resulting in 30 image evaluations. It reports 25 preferences for Eovist images and 5 for MultiHance images, with a Chi-squared p-value of 0.005. The endpoint description also mentions an ordinal response mixed effect model, but the relationship between that description and the reported analysis is not explained.

The result concerns reader preference, not diagnostic performance or patient outcomes. The record does not report a confidence interval, detailed multiplicity or missing-data procedures, the reason for termination, or results for the listed secondary outcome topics.