Combined MRI and Optical Imaging to Improved Breast Cancer Diagnosis: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Combined MRI and Optical Imaging to Improved Breast Cancer Diagnosis (NCT00610467), a non-randomized na study. The primary outcome is Sensitivity, Specificity, and Overall Accuracy in Differentiating Between Benign and Malignant Lesions.

2. Clinical Question

3. Trial Design

NCT00610467, titled Combined MRI and Optical Imaging to Improved Breast Cancer Diagnosis, was sponsored by the University of California, Irvine. The registry lists the study phase as NA, the allocation as nonrandomized, masking as none, and the model as parallel. Enrollment is recorded as 2. The study started on 2006-10-01, had a primary completion date of 2016-10-30, and is listed as terminated.

The two named groups were the Disease Group and the Control Group. The Disease Group consisted of patients with suspicious breast diseases; the Control Group consisted of healthy volunteers for system testing. Because allocation was nonrandomized and the groups served different purposes, they should not be read as equivalent groups in a randomized comparison. The design does not support interpreting differences between them as effects caused by imaging or treatment.

The registry describes the study as terminated prematurely. It states that combined MRI and optical imaging had been performed on two subjects, that the sample size was too small to conclude findings, and that all files were destroyed per IRB requirement. No outcome measure data are available. The registry record does not report the enrollment target, reasons for termination beyond the stated description, or further details of the study procedures. Those details should not be inferred from the stated design or title.

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

Analysis population for the primary outcome, as defined in the registry: We had performed the combined MRI and optical imaging on two subjects. Due to the small sample size, there is low statistical power to conclude any findings. Study was terminated pre-maturely. All files have been destroyed per IRB requirement. No outcome measure data are available.

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 Sensitivity, Specificity, and Overall Accuracy in Differentiating Between Benign and Malignant Lesions, measured over After we meet the enrollment target.

The primary outcome was titled Sensitivity, Specificity, and Overall Accuracy in Differentiating Between Benign and Malignant Lesions. It was intended to assess how well imaging-derived measures classified lesions as benign or malignant. Sensitivity describes the ability of a diagnostic approach to identify lesions in the target category; specificity describes its ability to distinguish lesions outside that category; overall accuracy describes classification performance across the evaluated cases. These are general descriptions of the measures, not findings from this study.

The registry planned to use ROC curves obtained through a regression statistical approach and an artificial neural network. It described first evaluating optimized morphological and kinetic parameters measured by dynamic contrast-enhanced MRI, then comparing those curves with analyses that added choline measured by magnetic resonance spectroscopy, hemoglobin concentration and oxygen saturation measured by steady-state diffuse optical tomography, and transport rates in indocyanine green kinetics measured by dynamic diffuse optical tomography.

The stated timeframe was after the enrollment target was met. The enrollment target and the planned endpoint values are not reported in the registry record. Most importantly, the registry states that no outcome measure data are available. Consequently, sensitivity, specificity, overall accuracy, and ROC results cannot be reported as observed findings for this study.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The planned endpoint description names ROC curves, a regression statistical approach, and an artificial neural network. The registry does not report analysis results, model specifications, or fitted parameters. The concepts below explain the general methods named in the record; they do not imply that a completed analysis produced findings in this study.

What does an ROC curve describe?

In general, a receiver operating characteristic curve describes how a diagnostic rule’s sensitivity and specificity vary as its decision threshold changes. It can help assess how well a measure separates cases in the target category from those outside it, across possible thresholds. An ROC curve is not itself a single clinical decision rule. Interpretation depends on the intended use, the threshold selected, the accuracy of the underlying measurements, and the population in which performance is assessed. The registry planned ROC analyses for distinguishing benign from malignant lesions but reports no curves, thresholds, or performance summaries.

What do sensitivity, specificity, and accuracy contribute?

In general, sensitivity and specificity describe different aspects of classification. Sensitivity concerns detection of lesions in the designated target category; specificity concerns correct classification of lesions outside it. Overall accuracy combines classification outcomes, but its meaning can depend on the mix of cases in the evaluated population. A result for one measure does not automatically determine the others. These measures are meaningful only when the reference classification and analysis population are clear. The registry record does not report outcome data, a reference-standard definition, or estimates for any of these measures.

Why compare MRI alone with added measurements?

The planned sequence was to obtain ROC curves using optimized morphological and kinetic parameters measured by dynamic contrast-enhanced MRI, then compare them with curves using those MRI parameters plus additional measurements. The additions named in the record were choline from magnetic resonance spectroscopy; hemoglobin concentration and oxygen saturation from steady-state diffuse optical tomography; and transport rates in indocyanine green kinetics from dynamic diffuse optical tomography. In general, comparing nested sets of measurements can address whether additional information changes diagnostic performance. A valid interpretation would require reported results from the analyses and enough information about the participants, measurements, outcome classification, and model fitting. Those results and details are not available in the registry record.

What roles can regression and a neural network play?

In general, regression models can relate measured predictors to an outcome and produce estimates used for classification or prediction. An artificial neural network is a flexible modeling approach that can represent complex relationships between inputs and an output. Either approach requires specification, fitting, and evaluation. With limited data, flexible models may be especially difficult to estimate and assess reliably; this is a general statistical concern, not a quantified finding about a particular model in this study. The registry names both approaches but does not report model structure, predictor selection, training or validation procedures, or performance results. It is therefore not possible to evaluate how either method was implemented.

Why does sample size matter for diagnostic modeling?

In general, estimates based on few observations can be highly uncertain, and a model may not capture the range of variation present in the intended population. This matters for sensitivity and specificity because these depend on the cases in each outcome category, not merely the total enrollment. The record reports two subjects and explicitly says the sample size was too small to conclude findings. It gives no endpoint estimates or uncertainty intervals. The record’s statement that statistical power was low should not be converted into a claim that a particular effect was absent or that a specific analysis had a known probability of detection.

How should planning illustrations be read?

General sample-size planning asks how many participants might be needed under specified assumptions to detect an effect at a chosen significance level and power. Such calculations are planning illustrations, not observed results and not a measure of the study’s achieved power. The stated assumptions for the illustrations include a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation; the effect to detect is described as a stated fraction of the observed effect. The registry record does not report primary outcome data or an observed endpoint effect. Thus, the illustrations cannot be interpreted as evidence that this terminated study attained a particular power or as a substitute for its missing results. They also do not establish that these assumptions formed the trial’s own design, which is not reported in the registry record.

10. Multiplicity, Missing Data and Analysis Populations

The planned endpoint involved several performance measures and several sets of imaging inputs. It named sensitivity, specificity, and overall accuracy, and described comparing MRI parameters alone with analyses adding measurements from magnetic resonance spectroscopy and diffuse optical tomography. Multiple modeling approaches were also named: regression and an artificial neural network. In general, assessing multiple outcomes, models, or predictor combinations can raise multiplicity concerns because the chance of a seemingly notable result may increase when many analyses are considered. A prespecified analysis plan and clear reporting help readers understand which comparisons were primary and how results should be interpreted.

The registry record does not report a multiplicity adjustment, a hierarchy among outcome measures or imaging combinations, or enough detail to determine how analyses would have been selected or interpreted. It would be inappropriate to assume either that an adjustment was planned or that none was needed.

For missing data, the registry states that all files were destroyed per IRB requirement and that no outcome measure data are available. It does not report reasons for missingness by participant or outcome, a missing-data method, or an analysis of the effect of unavailable measurements. With no endpoint data, readers cannot assess whether missingness affected a particular estimate.

The record also does not report detailed analysis populations, exclusions, or handling of incomplete imaging. The small number of subjects and the absence of outcome data prevent an assessment of how many participants contributed to each planned analysis. These omissions should be treated as unreported information, not filled in through assumptions about standard practice.

11. 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 event figures separately for the Disease Group and Control Group. For the Disease Group, 1 participant was at risk for serious adverse events and 0 were affected; 1 was at risk for other adverse events and 0 were affected. Deaths are recorded as 0. For the Control Group, 1 participant was at risk for serious adverse events and 0 were affected; 1 was at risk for other adverse events and 0 were affected. Deaths are recorded as 0.

These are the figures recorded for this study, not evidence that an event could not occur or that the imaging procedures have established safety. The study involved very few subjects, was terminated prematurely, and has no outcome measure data available. The registry does not list serious adverse event terms. The absence of recorded affected participants in these entries should therefore be read narrowly: no affected participants are recorded in the listed event categories, rather than as a broad safety conclusion.

The group descriptions also matter. The Disease Group included patients with suspicious breast diseases, while the Control Group comprised healthy volunteers for system testing. The nonrandomized design and the registry’s warning that the arms are not a like-for-like control comparison mean the figures should not be used to compare event risk between groups.

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

The result of this registry record is primarily an account of a study that did not produce reportable endpoint findings. It documents a planned evaluation of MRI and added optical imaging measurements for differentiating benign from malignant lesions, but it does not provide sensitivity, specificity, overall accuracy, ROC curves, or model-performance estimates. It therefore cannot establish whether adding optical imaging measurements changed diagnostic performance.

The record states that imaging was performed on two subjects, that the study was terminated prematurely, and that the sample size was too small to conclude findings. It also states that all files were destroyed per IRB requirement and that no outcome measure data are available. These details explain why the diagnostic question remains unanswered in the posted results; they do not establish that the imaging approaches were effective, ineffective, equivalent, or inferior.

In general, statistical uncertainty is assessed using estimates and measures such as confidence intervals. Because the registry does not report outcome estimates or intervals, the degree of uncertainty around diagnostic performance cannot be quantified from this record. Nor can readers assess how performance might vary across participants or whether any proposed model would generalize beyond the people studied.

The adverse-event entries record no affected participants in the listed categories, but they should not be interpreted as definitive evidence about safety. The study size and early termination limit what those entries can establish, and the arms were not designed as a like-for-like comparison.

The appropriate conclusion is limited: the study had a stated diagnostic aim and planned analytic approach, but its posted record contains no outcome data with which to answer that aim. Any stronger conclusion would require information not reported in the registry record.

13. Why This Trial Matters Statistically

This study illustrates an important distinction in clinical research between a clearly stated scientific question and evidence that can answer it. The registry describes a diagnostic question, names the imaging measurements of interest, and outlines an intended ROC-based assessment. However, because no outcome measure data are available, the planned analysis cannot be translated into an estimate of diagnostic performance.

It also highlights why sample size is central to diagnostic studies. Measures such as sensitivity and specificity depend on the number and characteristics of participants in the relevant outcome categories. The record reports that combined MRI and optical imaging was performed on two subjects and states that the sample was too small to conclude findings. Without reported endpoint data, there is no basis for quantifying precision or assessing whether apparent model performance would be stable.

The study is also a reminder that comparison groups must match the question being asked. Here, patients with suspicious breast diseases and healthy volunteers for system testing had different stated roles, and the record explicitly warns that the arms are not a like-for-like control comparison. Their group labels should not be mistaken for a randomized diagnostic comparison.

Finally, planned methods and actual results are distinct. Naming regression, a neural network, and ROC curves does not show that the analyses were completed or validated. The record’s early termination and lack of outcome data make transparent reporting especially important: readers can identify the intended question and the limits of the available evidence without mistaking a protocol description for a result.

14. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The principal limitation is the absence of primary outcome data. The registry reports no sensitivity, specificity, accuracy, ROC curves, or analysis results. It states that all files were destroyed per IRB requirement, so the planned diagnostic comparisons cannot be independently interpreted from posted measurements.

The study was terminated prematurely and included two subjects who underwent combined MRI and optical imaging. The registry itself states that the sample was too small to conclude findings and that statistical power was low. No outcome estimates or uncertainty intervals are reported, leaving the magnitude and precision of diagnostic performance unknown.

The comparison structure also limits interpretation. Allocation was nonrandomized, masking was none, and the registry warns that the arms are not a like-for-like control comparison. The Disease Group comprised patients with suspicious breast diseases, while the Control Group comprised healthy volunteers for system testing. The record does not report a detailed analysis population or results that would allow readers to evaluate performance within an intended-use population.

The primary outcome description names regression and an artificial neural network, but model specifications, validation procedures, decision thresholds, and reference-standard details are not reported in the registry record. Nor are the enrollment target and the details of any multiplicity strategy reported. These gaps prevent assessment of how the proposed analyses would have been carried out and whether their performance could generalize.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Combined MRI and Optical Imaging to Improved Breast Cancer Diagnosis — NCT00610467.
• Registry record last updated 2020-04-20; status last verified 2020-04.
• Results first posted 2020-04-20.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

18. Record Summary

NCT00610467 was a nonrandomized, parallel study with no masking, sponsored by the University of California, Irvine. The registry lists enrollment as 2 and status as terminated. It planned to assess whether adding optical imaging measurements to dynamic contrast-enhanced MRI could improve differentiation between benign and malignant lesions, using sensitivity, specificity, overall accuracy, and ROC analyses.

The Disease Group included patients with suspicious breast diseases, and the Control Group included healthy volunteers for system testing. The registry reports that imaging was performed on two subjects, that the study was terminated prematurely, and that all files were destroyed per IRB requirement. No outcome measure data are available. The record therefore describes the intended diagnostic question but does not report findings that answer it.