Evaluation of a Computerized Complex Instrumental Activities of Daily Living Marker (NMI): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Evaluation of a Computerized Complex Instrumental Activities of Daily Living Marker (NMI) (NCT02843529), a randomized na study. The primary outcome is Change in Diagnostic Area Under the Receiver Operating Characteristic Curve (ROC-AUC).

2. Clinical Question

3. Trial Design

NCT02843529 is a completed study sponsored by Altoida. The registry classifies the allocation as randomized, the masking as double, and the model as parallel. It lists the phase as not applicable. The study description characterizes the project as a prospective cohort intended to evaluate the Altoida system in real-world clinical settings. The registry record does not explain the randomization process, what was masked, or how those design labels relate to the cohort description.

The listed enrollment is 548. In the posted primary results, the MCI group has 213 participants and the cognitively normal group has 283. The registry record does not reconcile the enrollment figure with the group counts. The arm descriptions state that neurological examinations and neuropsychological assessments were planned at baseline and every 6 months. Alzheimer disease biomarker measurements were described at inclusion and at the end of the study, although the available arm descriptions are truncated. The primary outcome timeframe is approximately 40 months of follow-up.

The study groups were MCI, including prodromal Alzheimer disease, and cognitively normal, including preclinical Alzheimer disease. The intervention entry names neuropsychological, MRI, EEG and CSF biomarker assessments. The primary outcome description also identifies digital inputs such as voice, hand movements, posture, eye tracking, navigation and gait. The study aimed to assess diagnostic classification between groups and prediction of cognitive decline within the MCI group.

The registry lists 30 outcome measures and 29 posted secondary outcomes. It lists 0 analyses. Detailed eligibility criteria, recruitment procedures, randomization methods, masking procedures, protocol deviations, and the analysis populations are not reported in the registry record. The record gives a start date of 2016-10-17, a primary completion date of 2020-02-18, and a completed status.

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

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 Change in Diagnostic Area Under the Receiver Operating Characteristic Curve (ROC-AUC), measured over approximately 40 months follow up.

The primary endpoint is titled “Change in Diagnostic Area Under the Receiver Operating Characteristic Curve (ROC-AUC).” ROC-AUC summarizes how well a diagnostic model ranks people with one classification relative to people with another across possible decision thresholds. A higher AUC generally indicates better discrimination in that evaluation; it does not by itself establish that a model is calibrated, clinically useful, or suitable for stand-alone diagnosis.

The endpoint description says that models using voice, hand micromovements and micro-errors, posture changes, eye tracking, visuospatial navigation micro-errors and spatio-temporal gait parameters would be tested. It identifies differential diagnosis between the study groups and prediction of cognitive decline in the MCI group, with decline assessed using a neuropsychological test battery. The listed unit is probability, the summary parameter is a geometric least-squares mean, and uncertainty is represented by a 95% confidence interval.

The timeframe is approximately 40 months of follow-up. The posted results report an estimate of 0.94 (95% CI 0.93 to 0.96) for MCI, including prodromal Alzheimer disease, and 0.91 (95% CI 0.90 to 0.95) for cognitively normal participants, including preclinical Alzheimer disease. These are group estimates; the registry record does not report the numerical change from baseline or a direct comparison of the group estimates. The record also does not specify the decision threshold, the reference standard used for classification, or whether the reported AUCs refer to a particular follow-up point.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does ROC-AUC measure?

ROC-AUC is an overall measure of a classifier’s ability to distinguish between two labeled groups across decision thresholds. In general, it can be interpreted as the probability that a randomly selected member of one group receives a higher model score than a randomly selected member of the other group, subject to the scoring convention. It summarizes ranking discrimination; it does not identify a single operating threshold. AUC alone does not provide sensitivity, specificity, predictive values, calibration, or the consequences of false classifications at a chosen threshold.

The registry reports AUC estimates for two diagnostic groups: 0.94 for MCI, including prodromal Alzheimer disease, and 0.91 for cognitively normal participants, including preclinical Alzheimer disease. The record does not make clear whether these values represent separate classification tasks against a shared reference, or a different analysis for each group. It does not report a direct contrast between the estimates. Consequently, the values can be described separately, but the registry record does not support a claim that one group’s estimate is statistically higher than the other’s.

How should the confidence intervals be read?

A confidence interval expresses statistical uncertainty around an estimate under the model and assumptions used in the analysis. The reported 95% confidence interval for the MCI estimate is 0.93 to 0.96; the interval for the cognitively normal estimate is 0.90 to 0.95. A narrower interval generally indicates greater precision, but a confidence interval does not account for every source of uncertainty. For example, uncertainty from participant selection, measurement choices, model development, or departures from analysis assumptions may not be represented by the interval.

The registry identifies the summary parameter as a geometric least-squares mean and the dispersion as a 95% confidence interval. It does not describe the fitted model, transformation, covariate adjustment, or construction of the interval. Those analytic details are not reported in the registry record, so the estimates should be interpreted using the reported labels rather than assumptions about how the calculation was performed.

What does “change” add to the endpoint?

A change endpoint ordinarily compares a measure at a later assessment with a baseline measure, or otherwise estimates how the measure evolves over time. Such an analysis requires a defined time contrast and an analysis that accounts for repeated observations when relevant. The registry title refers to change in diagnostic AUC and gives an approximately 40-month follow-up timeframe. However, the posted results contain group-specific AUC estimates, not a stated baseline value, follow-up value, change, or interval for change. The record does not report how repeated measurements contributed to the posted estimates.

What can the planning illustration tell readers?

The planning illustration uses the observed effect as its starting point and assumes a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. It considers detecting a stated fraction of the observed effect. This type of calculation describes hypothetical sample-size requirements under its assumptions; it is not an observed result and is not observed power. A planning illustration cannot establish what power the trial actually had, because the trial’s design assumptions, target effect, and sample-size justification are not reported in the registry record.

In general, sample-size requirements depend on the effect chosen for detection, variability, allocation, significance level, and the analysis method. Changing these assumptions changes the planning result. The illustration should therefore be understood as conditional planning information, not as a retrospective statement about whether the study succeeded or whether a result is reliable.

Why do comparisons and model validation matter?

Separate estimates do not answer whether two estimates differ. A formal comparison would require an estimated contrast and an uncertainty measure or test for that contrast. Neither is reported in the registry record. Similarly, a high discrimination estimate in an evaluation sample does not alone establish performance in new participants. General statistical practice distinguishes model development from validation, because evaluating a model on data that contributed to its development can yield an overly optimistic estimate. The registry description says the models were developed for the system and tested in the prospective cohort, but does not report the specific validation strategy or whether the evaluation data were independent of model development.

The record’s stated aim includes sensitivity, specificity, accuracy, and prediction of cognitive decline. These are related but distinct measures or tasks. AUC does not determine their values at a selected threshold, and the registry record does not report numerical estimates for those analyses. Interpretation should remain with the posted AUC estimates and their confidence intervals rather than filling in unreported performance quantities.

10. Multiplicity, Missing Data and Analysis Populations

The registry lists 30 outcome measures and 29 posted secondary outcomes. The primary endpoint concerns change in diagnostic ROC-AUC; the secondary outcomes include changes in clinical and cognitive measures. When many outcomes or analyses are considered, the chance of at least one apparently notable result can increase if no multiplicity strategy is used. General approaches include pre-specifying a hierarchy, adjusting inference, or distinguishing confirmatory from exploratory analyses. The registry record does not report a multiplicity plan, an outcome hierarchy, or which outcomes were treated as confirmatory.

The registry lists 0 analyses. It does not report p-values or a formal test of the primary endpoint, and it does not provide estimates for the posted secondary outcomes in the registry record by the record. A reader therefore cannot determine which secondary analyses were conducted, how results were selected for reporting, or whether any multiplicity adjustment was applied. The absence of those details is not evidence that a particular analysis or adjustment was or was not performed.

Missing-data procedures and analysis populations are not reported in the registry record. The posted primary results identify 213 participants in the MCI group and 283 in the cognitively normal group, while the listed enrollment is 548; the record does not explain the relationship between these counts. It also does not state the number with complete follow-up, the handling of withdrawals or unavailable assessments, or whether all randomized, enrolled, or evaluable participants were included in the estimates.

In general, missing measurements can affect both precision and the validity of an estimate, depending on why they are missing and how the analysis handles them. Without information about missingness and the analysis population, the confidence intervals do not resolve those questions. The registry record also does not report interim analyses or a statistical analysis plan.

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 posted safety figures list 0 serious adverse events and 0 other adverse events in each group. The serious-event risk sets are 213 for MCI, including prodromal Alzheimer disease, and 283 for cognitively normal participants, including preclinical Alzheimer disease. The same group sizes are listed for the other-adverse-event risk sets. The record also lists 0 deaths in each group, and no serious-event terms are listed.

These are counts recorded in the registry, not proof that no adverse experiences occurred. Interpretation depends on how events were defined, collected, monitored, and attributed, as well as the period covered by safety follow-up. Those details are not reported in the registry record. The record does not specify whether the counts include all enrolled participants, all participants with follow-up, or a different safety population.

Zero recorded events in the posted results should therefore be read narrowly: the registry reports no serious or other adverse events and no deaths for the stated group risk sets. The figures do not establish that every participant was observed for the same duration or that the system has no risks. The study description presents an assessment system and diagnostic-marker evaluation; the safety counts should not be interpreted as evidence about treatment benefit or as a general safety guarantee.

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

The result is a pair of reported diagnostic discrimination estimates with uncertainty intervals: 0.94 (95% CI 0.93 to 0.96) for the MCI group, including prodromal Alzheimer disease, and 0.91 (95% CI 0.90 to 0.95) for the cognitively normal group, including preclinical Alzheimer disease. These estimates describe ROC-AUC as reported in the registry. They indicate how the models were reported to discriminate within the study’s group classifications, subject to the design and analysis details available in the record.

The result does not establish a difference between the groups. The registry record does not report a direct contrast or its confidence interval. It also does not report the stated primary change as a numerical baseline-to-follow-up estimate. The two values should not be treated as measurements of change, or as evidence that performance improved or declined over time.

The posted estimates do not establish sensitivity, specificity, accuracy at a particular threshold, calibration, or individual-level predictive performance. The primary outcome description says those types of model performance and cognitive-decline prediction were to be tested, but their numerical results are not reported in the registry record. Nor do the estimates show that use of the system changes clinical decisions, improves patient outcomes, or can replace physician assessment. The registry description explicitly presents the information as an additional aid and says findings are not stand-alone or diagnostic per se.

General interpretation of a diagnostic model also depends on the population, diagnostic reference standards, measurement procedures, handling of missing observations, and validation design. The registry record does not report enough detail on these matters to determine how well the estimates would transfer to settings or participants beyond the study. The values are therefore best treated as reported estimates for the studied groups, not as universal performance guarantees.

Finally, the planning illustration is not evidence of observed power or a result of the clinical evaluation. It is a conditional sample-size exercise based on assumptions. The primary findings remain the posted AUC estimates and their confidence intervals, alongside the important absence of a reported direct comparison, change estimate, and numerical secondary analyses.

13. Why This Trial Matters Statistically

This study is statistically notable because it evaluates a diagnostic-marker concept rather than a treatment effect. That distinction shapes what the estimates can answer. ROC-AUC addresses discrimination between labeled groups; it does not by itself establish whether a marker is useful at a clinical decision threshold or whether using it improves decisions. Keeping that distinction clear prevents an overall discrimination statistic from being interpreted as a complete assessment of a diagnostic tool.

The record also illustrates why the endpoint wording and the posted result need to be read together. The primary outcome is titled as a change in AUC over approximately 40 months, yet the posted results report group-level estimates and confidence intervals without a change estimate. The record’s stated question and the available numerical result are therefore not identical in detail. Recognizing that gap is important when judging whether the primary question has been answered.

Multiple outcome measures add another statistical consideration. The registry lists 30 outcome measures and 29 posted secondary outcomes, but does not report an analysis hierarchy or multiplicity strategy. This makes it difficult to tell how the primary outcome was distinguished from other analyses or how any secondary findings should be weighed. The record also does not report a formal comparison of the two group estimates.

Finally, transparent reporting of model development, validation, diagnostic reference standards, and missing-data handling is central to evaluating a classifier. Those details are not reported in the registry record. The posted confidence intervals communicate uncertainty under the reported analysis, but cannot replace information about study conduct or external validation. The result is informative as a record of reported discrimination estimates, while leaving key questions about their comparison, longitudinal interpretation, and applicability unresolved.

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 in interpreting these results is that the primary outcome is described as change in ROC-AUC, but the posted results provide one AUC estimate for each group without a stated change from baseline. The record does not identify the assessment time corresponding to these values or report a formal test of change. It also does not report a direct contrast between the MCI and cognitively normal estimates.

The registry does not describe the diagnostic reference standard, the threshold used to classify participants, or the details of model training and validation. The primary outcome description names several digital inputs and says models would be tested in the prospective cohort, but the record does not establish whether model development and evaluation used independent data. Without those details, the posted AUC estimates cannot be fully assessed for potential optimism or reproducibility.

The listed enrollment of 548 is not reconciled with the posted group counts of 213 and 283. The registry record does not report the analysis population, the number completing follow-up, missing-data methods, or how withdrawals and unavailable measures were handled. It also does not report participant eligibility details or how the groups were recruited and characterized.

Although 30 outcome measures and 29 secondary outcomes are listed, the registry lists 0 analyses and does not report a multiplicity plan or detailed results for the secondary measures. Numerical results for sensitivity, specificity, accuracy, and cognitive-decline prediction are not reported in the registry record. The arm descriptions are truncated, leaving some assessment and follow-up details incomplete. Together, these omissions limit interpretation of the AUC estimates, the stated longitudinal objective, and the extent to which the results apply beyond the evaluated groups.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Evaluation of a Computerized Complex Instrumental Activities of Daily Living Marker (NMI) — NCT02843529.
• Registry record last updated 2024-08-23; status last verified 2024-07.
• Results first posted 2020-12-16.
• 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

NCT02843529, sponsored by Altoida, is listed as completed. Its title concerns evaluation of a computerized complex instrumental activities of daily living marker, and its official title describes the intended use of the Altoida Neuro Motor Index in preclinical or prodromal Alzheimer disease. The registry lists a randomized, double-masked, parallel design and a phase of not applicable; detailed procedures underlying those classifications are not reported in the registry record.

The study description presents a prospective evaluation of digital measures and biomarker assessments for diagnostic classification and prediction of cognitive decline. The primary outcome is change in diagnostic ROC-AUC over approximately 40 months. Posted estimates are 0.94 (95% CI 0.93 to 0.96) for the MCI group and 0.91 (95% CI 0.90 to 0.95) for the cognitively normal group. A direct contrast, a change estimate, and numerical results for the stated prediction analyses are not reported in the registry record.

The registry lists enrollment of 548, 30 outcome measures, 29 posted secondary outcomes, and 0 analyses. The listed study start date is 2016-10-17, and the primary completion date is 2020-02-18.