This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the results posted in the ClinicalTrials.gov record; calculations added by Clinical Biostats are labelled as such.
1. Trial at a Glance
This page covers the registry results for Neural-net Artificial Pancreas (NAP) (NCT05876273), a randomized na study. The primary outcome is Percent of Time-in-Range (TIR) on NAP Versus UMPC..
| Feature | Registry value |
|---|---|
| ClinicalTrials.gov identifier | NCT05876273 |
| Phase | NA |
| Condition | Type1 Diabetes |
| Allocation | Randomized |
| Design model | Crossover |
| Masking | None (open label) |
| Enrollment | 15 |
| Arms in the primary outcome | n/a |
| Lead sponsor | University of Virginia |
| Results posted | Yes |
2. Clinical Question
Population
The planned participants were adults with type 1 diabetes. The registry summary says investigators intended to consent up to 20 participants aged at least 18.0 years, with a goal of completing 15. The recorded enrollment is 15. The record does not report participant characteristics such as age distribution, diabetes history, or baseline glucose measures.
Intervention
The intervention was the Neural-net Artificial Pancreas (NAP), a neural-network implementation of the University of Virginia Model Predictive Control algorithm. In the crossover schedule, participants used NAP for 18 hours in one experimental period.
Comparator
The comparator was the University of Virginia Model Predictive Control (UMPC) algorithm. Participants used UMPC for 18 hours in the other experimental period. The randomized sequence was either NAP followed by UMPC or UMPC followed by NAP.
Primary question
The primary question was whether percent of time in the 70 to 180 mg/dL range differed between NAP and UMPC during the two 18-hour experiments. Numerical results for this comparison are not reported in the registry record.
This is a comparison of two automated insulin delivery control approaches under a short experimental schedule, rather than a comparison of treatment with no treatment. The primary measure is time-in-range (TIR), expressed as a percentage of time within the stated glucose range. The registry also lists percent of time in hyperglycemia, percent of time in hypoglycemia, system functionality, and participant feedback as secondary outcomes.
The crossover format means that the planned comparison involved both devices for each participant, with the sequence randomized. That can make a within-participant comparison relevant because measurements from the same person may be more alike than measurements from different people. However, the registry record does not report the primary outcome values, the estimated difference, or the statistical analysis used to compare periods. It therefore does not establish the size or precision of any observed difference.
3. Trial Design
The registry identifies this as a randomized, unmasked, crossover pilot study of NAP and UMPC. The record lists an enrollment of 15 and a completed status. Its summary says investigators intended to consent up to 20 adults and aimed to complete 15 participants. Enrollment and completion goals describe the study record; they do not establish how many participants contributed usable primary outcome data.
Each sequence consisted of two experimental periods, each lasting 18 hours. One sequence used NAP first and UMPC second; the other used UMPC first and NAP second. The primary endpoint timeframe is given as 36 hours across the two experiments. Randomizing the order helps avoid assigning every participant the same sequence, but the record does not report how randomization was generated, whether allocation was concealed, or whether any sequence-related procedures were used.
Masking is listed as none. The record does not explain whether participants, investigators, or outcome assessors could be blinded to device assignment. Because these are distinct device algorithms, lack of masking may be relevant to subjective measures such as participant feedback. It does not, by itself, tell readers whether the glucose measurements were collected or processed consistently.
The study began on 2023-05-30, reached primary completion on 2023-09-08, and is listed as completed. The sponsor is the University of Virginia. The phase is listed as not applicable. The registry identifies the devices as the two interventions and gives their sequence-specific schedules. It does not report details such as the study setting, eligibility criteria beyond the adult age criterion in the summary, protocol deviations, period-specific adherence, or how missing measurements were handled.
A crossover design is intended to compare conditions within participants, but interpretation depends on the actual analysis and the quality of measurements in both periods. The registry record does not report whether there was a washout interval, how potential carryover between periods was considered, or whether period and sequence effects were included in the analysis. These are important design and analysis details, not assumptions that can be filled in from the design label alone.
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
| Stage | NAP First, Then UMPC | UMPC First, Then NAP |
|---|---|---|
| STARTED | 8 | 7 |
| COMPLETED | 8 | 7 |
| NOT COMPLETED | 0 | 0 |
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.
| Characteristic | Unit | All Participants |
|---|---|---|
| Participants analysed | 15 | |
| Age, Categorical: <=18 years | Participants | 0 |
| Age, Categorical: Between 18 and 65 years | Participants | 13 |
| Age, Categorical: >=65 years | Participants | 2 |
| Age, Continuous | years | 47.9 (27.6 to 67.9) |
| Sex: Female, Male: Female | Participants | 10 |
| Sex: Female, Male: Male | Participants | 5 |
| Ethnicity (NIH/OMB): Hispanic or Latino | Participants | 0 |
| Ethnicity (NIH/OMB): Not Hispanic or Latino | Participants | 15 |
| Ethnicity (NIH/OMB): Unknown or Not Reported | Participants | 0 |
| Race (NIH/OMB): American Indian or Alaska Native | Participants | 0 |
| Race (NIH/OMB): Asian | Participants | 0 |
| Race (NIH/OMB): Native Hawaiian or Other Pacific Islander | Participants | 0 |
| Race (NIH/OMB): Black or African American | Participants | 2 |
| Race (NIH/OMB): White | Participants | 13 |
| Race (NIH/OMB): More than one race | Participants | 0 |
| Race (NIH/OMB): Unknown or Not Reported | Participants | 0 |
| Region of Enrollment: United States | participants | 15 |
| Percent Time in Range (70-180 mg/dL) | Percent of time between 70-180 mg/dl | 69.3 (SD 13.4) |
6. Primary Endpoint
The registry lists the primary outcome as Percent of Time-in-Range (TIR) on NAP Versus UMPC., measured over 36 hours (two 18-hour experiments).
| Element | Registry entry |
|---|---|
| Result type | Mean |
| Measure of spread | Standard Deviation |
| Unit | Percent of time between 70-180 mg/dl |
| Endpoint classification | Continuous Endpoint |
The primary endpoint is percent of time in the glucose range from 70 to 180 mg/dL, also called time-in-range or TIR. It measures the share of the recorded observation time during which glucose readings fall within that interval. The registry specifies a mean as the summary measure and standard deviation as the dispersion measure. Its timeframe is the two 18-hour experiments.
TIR is used as a way to describe glucose exposure across a period rather than relying on a single reading. As a general statistical point, a percentage over time can retain information about how readings are distributed across the observation period that would be missed by reporting only one measurement. Its interpretation still depends on the range definition, the completeness and quality of the underlying measurements, and the duration observed. Here, the registered range and observation timeframe are specified, but details of measurement frequency and data processing are not reported in the registry record.
The endpoint is a continuous measure, and the record identifies the planned summary as a mean. A mean describes the average value across the observations included in an analysis; a standard deviation describes variation among those values. Neither statistic is reported numerically for the primary outcome. The registry record also does not give a difference between NAP and UMPC, a confidence interval, or a test result. Consequently, the endpoint definition is available, but the magnitude and precision of the comparison cannot be assessed from the registry information.
The primary endpoint should be read as a measure of glucose time within the specified band during these short experiments. It is not, on its own, a complete account of all glycemic outcomes, device performance, or longer-term experience. The listed secondary measures address some other aspects, but their numerical findings are not reported in the registry record.
7. Why No Re-analysis Is Shown
8. Statistical Interpretation
The registry record defines a randomized crossover comparison and names percent of time in the 70 to 180 mg/dL range as its primary outcome. It reports a completed study with enrollment of 15. However, it does not report the primary endpoint mean for either device, the comparison between devices, or any measure of uncertainty. Without those results, the direction and size of the observed TIR comparison cannot be described.
A study status of completed indicates that the registry lists the study as completed; it does not substitute for outcome data. Likewise, the enrollment figure does not establish how many participants contributed to a particular analysis. The record does not report the number with usable data in each period, exclusions from analysis, or missingness for the primary endpoint. Readers therefore cannot infer the analysis population from enrollment alone.
The crossover design is potentially informative because participants used both NAP and UMPC, in randomized order. In general, within-participant comparisons can account for stable differences between people, provided the analysis uses the paired structure appropriately. The record does not report whether the analysis used paired observations or another approach, or whether it accounted for sequence and period. Nor does it describe a method for assessing carryover. These omissions limit the ability to judge how the design was translated into the reported analysis.
There are also no reported primary outcome estimates from which to evaluate precision. A confidence interval would describe the range of values reasonably compatible with the data under the analysis assumptions. No such interval is reported in the registry record. A p-value is also not reported, so there is no registry-based test result to interpret. Absence of these statistics does not mean the effects are equal; it means the registry record does not provide the evidence needed to quantify the comparison.
The record lists secondary outcomes covering time in hyperglycemia, time in hypoglycemia, system functionality, and participant feedback. It does not report their results. The registry lists five outcome measures and one analysis, but does not give further analysis details or explain how outcomes were prioritized beyond identifying the primary endpoint. Thus neither the additional measures nor their relationship to the primary comparison can be assessed from the record.
The sample-size illustration associated with this type of report should be distinguished from the trial's observed findings. The record describes a planning approach based on the observed effect, with a two-sided significance level of 0.05, equal allocation, and a normal approximation. It does not provide numerical sample-size estimates in the registry entries available for this account. Such illustrations are planning aids, not observed power and not evidence that the trial did or did not detect an effect.
Overall, the registry supports a description of the question, design, and endpoint, but not a statistical conclusion about the comparison. A careful interpretation should stop short of claiming that one algorithm had higher or lower TIR, or that the approaches were similar. Numerical estimates and details of the analysis are needed for those judgments.
9. Statistical Methods Explained
The registry characterizes the primary outcome as continuous, with the mean and standard deviation as the planned descriptive measures. The endpoint is percent of time within the 70 to 180 mg/dL glucose range. The record does not provide numerical means or standard deviations, nor does it report an estimated contrast between devices. The following explains general statistical ideas relevant to this design; it does not establish which procedures were used in this study.
What does a mean represent? In general, a mean is an average of measured values. For this endpoint, a mean would describe average time-in-range for whichever participants and periods were included in the calculation. The interpretation depends on the analysis unit: a mean could be calculated over participant-level summaries, but the registry does not specify the calculation details. The mean alone does not show whether values were tightly clustered or widely spread.
What does a standard deviation represent? In general, a standard deviation describes the spread of observations around their mean. It is not the same as the uncertainty in the estimated mean, and it is not a treatment contrast. A standard deviation for participant-level TIR values would describe between-participant variation; a standard deviation of within-participant differences would describe a different quantity. The registry identifies standard deviation as the dispersion measure, but does not report its value or clarify the exact calculation.
Why does crossover structure matter? In general, a crossover trial gives participants more than one study condition, so observations from one participant are related. Analyses can use the within-participant pairing: each participant's outcome under one condition is compared with that same participant's outcome under the other. Accounting for pairing can improve efficiency when measurements within a person are correlated. Treating all observations as independent when they are paired can misrepresent uncertainty. The registry does not report whether a paired analysis was performed, so no particular analysis can be attributed to this study.
Why can order and period matter? In a crossover, the conditions occur in a sequence. Outcomes can vary over time for reasons unrelated to the device, and the first period could influence the second. Randomized order helps distribute sequence assignment, but it does not by itself document how period effects or carryover were analyzed. As a general statistical matter, analyses may account for treatment, period, and sequence, and study protocols may specify whether a washout is needed. The registry record does not report these details or a carryover assessment.
What would a confidence interval add? In general, a confidence interval gives an interval estimate of the comparison and communicates its precision under the chosen model and assumptions. A narrow interval indicates greater precision than a wide interval, while the interval's location indicates which effect values are compatible with the data under that approach. It is not a guarantee that the true effect lies in any one calculated interval. No confidence interval for the primary comparison is reported in the registry record, so precision cannot be evaluated.
What would a hypothesis test add? In general, a test evaluates how compatible the data are with a specified null model, conditional on the test assumptions. A p-value is not the probability that the null hypothesis is true, nor does it measure clinical importance. A small p-value does not describe the size of an effect, and a larger p-value does not prove equivalence. The record does not report a p-value or identify a primary test. Therefore, there is no reported inferential result to interpret.
How should the planning illustration be understood? The registry describes a planning illustration based on the observed effect, a two-sided significance level of 0.05, equal allocation, and a normal approximation. It says that the effect to detect is a stated fraction of the observed effect. These are planning assumptions, not a report of the trial's analysis or a measure of observed power. The record does not report numerical estimates for the illustration in the registry record, and no sample-size value can be inferred. In general, planning calculations depend on assumptions about effect size, variability, design, and the chosen testing framework; they do not show that a study has demonstrated an effect.
What is not established by the registry description? It does not state whether the analysis was intention-to-treat, per-protocol, or based on another analysis population. It does not report model specification, handling of missing values, outlier rules, or adjustment for multiple outcomes. It also does not report an interim analysis plan or a sample-size justification for the trial itself. These are not details that can be recovered from the fact that the study is randomized, crossover, or completed.
For a continuous endpoint in a crossover design, the central statistical questions include how each participant's two periods were summarized, how within-participant dependence was handled, whether period or sequence was included, and what uncertainty measure accompanied the treatment comparison. The registry record does not provide those answers or the resulting numerical estimates. General descriptions of good statistical practice can clarify why these details matter, but cannot replace the missing trial-specific analysis information.
10. Multiplicity, Missing Data and Analysis Populations
The registry identifies one primary outcome, percent of time-in-range, and lists secondary measures for time in hyperglycemia, time in hypoglycemia, system functionality, and participant feedback. It lists five outcome measures and one analysis. This record does not explain which comparisons were included in that analysis, whether all listed measures were analyzed, or how the primary and secondary outcomes were ordered for inference.
As a general statistical principle, testing many outcomes or comparisons can increase the chance of obtaining at least one apparently unusual result by chance. Multiplicity procedures can address that issue, but the choice depends on the analysis plan and the intended interpretation. The registry record does not report a multiplicity adjustment or state that no adjustment was planned. It is therefore not possible to say how multiple outcomes were handled.
Missing-data details are also not reported in the registry record. The enrollment is 15, but the record does not say how many participants had usable measurements for each period or endpoint, whether any data were incomplete, or how missing measurements were treated. It does not describe imputation, exclusions, sensitivity analyses, or reasons for missingness. These details matter because an analysis restricted to available observations can represent a different population from the one originally enrolled.
The analysis population is not reported in the registry record. The record does not identify an intention-to-treat, per-protocol, or other population for the primary comparison, nor does it state how participants with protocol deviations were handled. Without that information, the results cannot be tied to a clearly defined analysis set. The registry also does not describe whether period-specific data or participant-level summaries were the unit of analysis.
The absence of these details should be read as a limitation of what the registry record reports, not as evidence that a particular procedure was or was not used. A full interpretation would need the analysis plan and outcome reporting, including the population analyzed, missingness, and any approach to multiple outcomes.
11. Safety Results
The registry reports serious adverse events as affected participants divided by participants at risk.
| Arm | Participants with serious adverse events | Participants at risk | Percentage |
|---|---|---|---|
| NAP First, Then UMPC | 0 | 8 | 0.0% |
| UMPC First, Then NAP | 0 | 7 | 0.0% |
Deaths
| Arm | Deaths (participants affected of at risk) |
|---|---|
| NAP First, Then UMPC | 0 of 8 |
| UMPC First, Then NAP | 0 of 7 |
Other (non-serious) adverse events
Participants with at least one non-serious event at or above the registry's 0% reporting threshold.
| Arm | Participants affected of at risk |
|---|---|
| NAP First, Then UMPC | 0 of 8 |
| UMPC First, Then NAP | 0 of 7 |
The adverse-event figures are reported separately for the two randomized sequences. For the NAP-first-then-UMPC group, the registry lists 8 at risk for serious adverse events and 0 affected; it also lists 8 at risk for other adverse events and 0 affected. For the UMPC-first-then-NAP group, it lists 7 at risk for serious adverse events and 0 affected; it also lists 7 at risk for other adverse events and 0 affected. The listed death count is 0 for each group.
These are registry-reported counts for the recorded groups and categories. They do not provide descriptions of event monitoring, severity definitions, collection procedures, or the time window used for safety reporting. The registry record does not list serious adverse-event terms. A zero in the affected column means no events are recorded in that category for the listed group; it should not be interpreted as proof that a device is free of risk.
The two groups represent randomized sequences in a crossover study, not separate long-term parallel treatment arms. The record does not report exposure-adjusted safety summaries or participant-level detail. These counts describe what the registry reports, but do not support broader conclusions about safety beyond the study record.
12. What the Result Does, and Does Not, Mean
The registry establishes that this completed pilot was designed to compare NAP with UMPC in a randomized crossover schedule. It identifies percent of time between 70 and 180 mg/dL as the primary outcome and specifies two 18-hour experimental periods. This is enough to understand the planned comparison, but not enough to assess its result.
The record does not report numerical primary outcome findings. It gives no mean TIR for either device, estimated difference, confidence interval, or p-value. Therefore, it cannot show from the registry information whether the observed TIR was higher under one approach, whether the estimates were close, or how precisely the comparison was measured. A completed status and an enrollment figure are not substitutes for these results.
The registry also does not report the findings for the listed secondary outcomes. Thus the record cannot establish how the algorithms compared on time in hyperglycemia, time in hypoglycemia, system functionality, or participant feedback. Nor does it describe the clinical importance threshold, if any, used to interpret changes in TIR. Statistical estimates and their context would be needed to assess the practical meaning of any observed difference.
In general, a study result is not just a direction of effect or a test result. Interpretation also depends on the size of the estimate, uncertainty, data completeness, analysis population, and design assumptions. Here, several of those elements are not reported in the registry record. The short experimental timeframe also confines the registered primary comparison to the two 18-hour experiments; it does not describe longer-term outcomes.
Accordingly, no claim that either algorithm performed better, worse, or equivalently is supported by the registry entries described. This is not a conclusion that there was no effect. It is a distinction between a study question and a reported numerical answer: the question and design are recorded, while the primary results and key analytical details are not reported in the registry record.
13. Why This Trial Matters Statistically
This trial is statistically notable for its crossover design and short, paired experimental periods. In general, comparing two conditions within the same participants can reduce the influence of stable differences between individuals. The design also introduces analysis considerations: the two measurements from a participant are related, and order or period effects may matter. The registry states that sequence was randomized, but does not report how the analysis addressed pairing, period, sequence, or possible carryover.
The endpoint is a continuous summary of time within a prespecified glucose band. Such an outcome represents a different aspect of glucose experience from a single reading, but its meaning depends on how measurements are collected and reduced to participant-level values. The registry supplies the measure and timeframe, yet not the data-processing details or numerical outcomes. This illustrates why clear endpoint definitions are important but do not by themselves provide an interpretable effect estimate.
The record also provides an opportunity to distinguish a planning calculation from an observed result. Its stated planning approach is based on an observed effect and describes assumptions for a sample-size illustration. A planning estimate is not a substitute for a confidence interval or a measure of how well the actual comparison was estimated. The available record does not report numerical values for that illustration.
Finally, the gap between a completed study record and a fully interpretable result is statistically consequential. Without the primary estimates, uncertainty, analysis population, and missing-data information, readers cannot evaluate effect size or robustness. The design can be described, but the evidence for the comparison cannot be quantified from the registry record alone.
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 registry record does not report the primary outcome results, including means for either device, a between-device estimate, a confidence interval, or a test result. This is the central limitation for interpreting the comparison: the study question is clear, but its numerical answer is unavailable in the record.
Although enrollment is listed as 15 and the study is marked completed, the number contributing usable primary endpoint data is not reported in the registry record. The record does not describe missing measurements, participant exclusions, protocol deviations, adherence to each period, or the analysis population. These omissions prevent evaluation of whether the analyzed data represent all enrolled participants or a selected subset.
The registry identifies a randomized crossover design but does not report the randomization procedure, allocation concealment, washout arrangements, or analysis of order, period, and carryover. Those details matter because outcomes in a crossover study can be affected by the sequence and timing of conditions. The record also does not describe measurement frequency, data processing, or criteria for valid glucose observations.
Masking is listed as none, but the record does not specify which parties were unmasked or how that might affect collection of subjective measures such as participant feedback. It lists secondary outcomes, but does not report their results or a multiplicity strategy. A sample-size justification for the trial itself and any interim analysis plan are not reported in the registry record.
Finally, the primary endpoint covers two 18-hour experiments. The record does not report longer-term outcomes or establish how these short periods relate to outcomes outside the study schedule. These limits restrict the conclusions available from the registry entry, independently of what the unreported numerical findings might show.
15. Related Tutorials
16. Related Statistical Calculators
17. Sources
- ClinicalTrials.gov: Neural-net Artificial Pancreas (NAP) — NCT05876273.
- Registry record last updated 2024-07-31; status last verified 2024-07.
- Results first posted 2024-07-31.
- 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.
Continue through the Clinical Biostats statistical pathway
Use this trial's endpoint as a starting point for deeper study of effect measures, confidence intervals, and sample-size planning.
18. Record Summary
The registry describes NCT05876273, Neural-net Artificial Pancreas (NAP), as a randomized, unmasked crossover pilot comparing NAP with the University of Virginia Model Predictive Control algorithm. The study is listed as completed, with enrollment of 15 and a target of completing 15 participants. Its recorded start date is 2023-05-30, and primary completion is dated 2023-09-08.
Participants were assigned to one of two sequences, using each device for 18 hours. The primary outcome was mean percent of time in the 70 to 180 mg/dL range across the 36-hour, two-experiment timeframe. The registry also lists secondary outcomes on hyperglycemia, hypoglycemia, system functionality, and participant feedback.
Numerical primary and secondary results, the analysis population, missing-data methods, and key crossover analysis details are not reported in the registry record. Safety counts list no serious or other adverse events and no deaths in either sequence group. The available information describes the study's question and structure, but does not permit assessment of the estimated device comparison or its precision.