Effect of Gain on Closed-Loop Insulin: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Effect of Gain on Closed-Loop Insulin (NCT02065895), a randomized na study. The primary outcome is Glucose Area Under the Curve (AUC) Breakfast.

2. Clinical Question

The study focused on the system’s response to calibration error: the difference between the glucose value used by the closed-loop system and blood glucose. The registry also describes an interest in fasting target levels and breakfast meal response, but it does not post numerical fasting results among the outcomes. The primary outcome is specifically breakfast glucose AUC, measured from 8:00 AM to 2:00 PM on the day following admission. Samples were obtained every 10-15 minutes.

Because the design is crossover, the conditions were experienced on separate occasions rather than assigned as independent parallel groups. The order varied across participants. The registry record does not report enough detail to determine how the analysis handled within-participant correlation, order, or period effects. The posted condition-specific results therefore need to be read alongside that methodological uncertainty.

3. Trial Design

This was a randomized, open-label crossover study sponsored by Joslin Diabetes Center. The registry lists the phase as NA, enrollment as 8, and status as completed. The recorded study start was 2013-12 and primary completion was 2015-04. Masking was listed as none, so the registry does not indicate that participants or study personnel were blinded.

Participants underwent overnight and breakfast closed-loop glucose control on 3 occasions, with HIGH Error, NO Error and LOW Error conditions experienced in varying orders. The registry descriptions characterize HIGH Error as using a glucose value for control higher than blood glucose, LOW Error as using a value lower than blood glucose, and NO Error as using a value equal to blood glucose. The occasions could be 24 hours to 7 days apart and were completed within 6 week period, according to the primary outcome timeframe.

The crossover structure allows condition results to be obtained from participants exposed to multiple settings, but it makes the analysis dependent on how repeated observations are handled. A participant’s readings under different conditions are not necessarily statistically independent. The registry does not report whether its ANOVA accounted for repeated measurements, condition sequence, or time between occasions. It also does not report the planned sample-size justification, the randomization procedure in detail, or the analysis population definition.

The primary endpoint was assessed during the breakfast observation window, on the day following admission, separately for each calibration error value. The registry describes glucose sampling every 10-15 minutes during 8:00 AM to 2:00 PM. The listed outcome means have n=5.0 for each condition, while total enrollment was 8. The record does not explain the relationship between enrollment and the number of observations contributing to each condition-specific mean, so the reported n should not be interpreted as the number of independent participants in a parallel-group comparison.

The intervention is described as a device, and the trial examined an external ePID system. Detailed device settings beyond the calibration-error conditions, participant-level procedures, and protocol deviations are not reported in the registry record. The sequence descriptions identify differing condition orders, but the record does not provide a complete account of how sequence assignment or any missing occasion was handled.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Overall Study

Reasons for not completing this period:

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: One subject withdrew prior to visit 1; of the seven subjects admitted on 3 occasions, two had protocol violations on one occasion (one subject experienced technical problems; one subject completed the nighttime period but was called away on an emergency prior to breakfast); 5 subjects completed the study per protocol.

6. Primary Endpoint

The registry lists the primary outcome as Glucose Area Under the Curve (AUC) Breakfast, measured over On day #1, day #2 and day #3 (each day could be 24 hours to 7 days apart from prior one, and completed within 6 week period) 8:00 AM to 2:00 PM on day following admission, with samples obtained every 10-15 minutes, for each sequence of calibration errors.

The primary endpoint was Glucose Area Under the Curve (AUC) Breakfast, reported in mmol/l/min. AUC combines glucose measurements across a defined time window into a measure of overall exposure during that window. In this trial, the window was 8:00 AM to 2:00 PM on the day following admission, with samples obtained every 10-15 minutes. The registry says this period captures breakfast response and identifies breakfast as typically the most difficult meal to control.

The purpose of using AUC is to describe glucose exposure over time rather than relying on a single measurement. AUC can reflect both the level and duration of glucose values during the observation window. It does not, by itself, describe the timing or shape of the glucose response; for example, the same AUC could arise from different patterns of readings. The registry also lists peak and nadir postprandial glucose concentration as a secondary outcome, which concerns features of the glucose profile not fully expressed by AUC alone.

The registry states that lower AUC is desirable for this outcome. That statement describes the endpoint’s intended direction; it is not a clinical recommendation. Condition-specific reported means were 66.8 for HIGH Error, 48.8 for NO Error and 37.4 for LOW Error. Their reported 95% confidence intervals were 53.3 to 80.3, 41.1 to 56.4 and 33.4 to 41.5, respectively. The endpoint values summarize the observed condition results, while the confidence intervals convey uncertainty around the estimates.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does the mean describe?

For a continuous outcome, a mean gives a summary of the observed values in a condition. The registry identifies the primary outcome parameter as a mean and reports one mean for each calibration-error condition. These values describe the average AUC in the observations included in each condition’s summary. They do not show the full distribution of individual values, and they do not tell the reader whether every participant had a value close to the mean.

The condition means are 66.8 for HIGH Error, 48.8 for NO Error and 37.4 for LOW Error. The outcome unit is mmol/l/min. AUC aggregates glucose measurements over the specified breakfast window, so the mean summarizes overall glucose exposure in that window rather than a single glucose reading.

How should the confidence intervals be read?

The registry reports 95% confidence intervals for each condition mean: 53.3 to 80.3 for HIGH Error, 41.1 to 56.4 for NO Error and 33.4 to 41.5 for LOW Error. In general, a confidence interval expresses the uncertainty of an estimate under the assumptions used to construct it. Wider intervals indicate less precision than narrower intervals when the estimates and scales are comparable. An interval is not a description of the range of all individual observations, and it does not guarantee that the true value lies within that particular interval.

These intervals are attached to condition means. They are not themselves confidence intervals for differences between conditions. A difference in the estimated means and a statistical comparison of those means are distinct quantities. No condition-specific pairwise estimates or pairwise confidence intervals are reported in the registry record.

What does the ANOVA result tell the reader?

The registry identifies ANOVA as the analysis method and lists the 3 conditions as groups. In general, ANOVA is used to assess evidence of differences among group means within a specified model. It can provide an overall test without establishing which particular conditions differ. A p-value describes how compatible the data are with a null hypothesis under the model; it is not the probability that the null hypothesis is true, nor is it a measure of the size or clinical importance of an effect.

The reported p-value is 0.001. The analysis result is labeled “Mean Difference (Final Values)” and includes an estimate of 66.8 and a 95% confidence interval from 33.4 to 80.3. The registry does not describe what “Final Values” refers to, how the mean-difference estimate was defined for the 3 listed groups, or which contrast the interval represents. Those omissions make it difficult to connect the reported analysis estimate directly to the condition-specific descriptive means.

Why does the crossover design matter?

In a crossover study, observations from the same participant under different conditions can be related. General statistical practice accounts for this dependence when estimating uncertainty and testing condition differences. If repeated observations are instead treated as independent, the model may not represent the study design appropriately. The registry names ANOVA but does not report whether it used a repeated-measures formulation or another method that accounts for participant-level correlation.

Condition order can also matter in a crossover design. The registry describes randomized orders and occasions separated by 24 hours to 7 days, within 6 week period. It does not report whether the analysis adjusted for sequence or period, or whether effects carried over between occasions were assessed. These are questions about the analysis and design; the record does not permit a conclusion about how they were addressed.

What can be inferred from the posted precision?

The condition-specific confidence intervals provide a view of the precision of the mean estimates, under the method used to calculate them. The registry does not report the variance measure used, participant-level data, or the exact confidence-interval calculation. The reported n=5.0 for each condition and total enrollment of 8 also do not establish how many participants contributed complete data across all conditions. The analysis population and missing-data handling are not reported in the registry record.

For a crossover comparison, a confidence interval for the within-participant difference can differ from what would be inferred by considering the separate condition intervals alone. It depends on the pairing or repeated-measures information, which is not available in the registry summaries. Consequently, overlapping or non-overlapping condition intervals should not replace the stated statistical analysis, and no new comparison should be calculated from the marginal summaries alone.

What is not specified?

The registry does not report a sample-size justification, a prespecified statistical analysis plan, an interim-analysis plan, or a multiplicity adjustment. It also does not give sufficient detail to reconstruct the ANOVA or determine the population included in it. The sample-size illustrations associated with a study are planning concepts, not observed power, and no planning outputs are reported in the registry record. General statistical interpretation cannot fill in those missing protocol details.

10. Analyses Reported by the Sponsor

The registry reports an ANOVA analysis for the 3 conditions, with a stated value of 66.8, a 95% confidence interval from 33.4 to 80.3, and p = 0.001. The condition-specific means and confidence intervals are descriptive summaries; they are not a substitute for the analysis of the crossover data. The registry’s analysis is the formal inferential result posted for the primary endpoint.

The analysis label, “Mean Difference (Final Values),” does not specify which contrast or model quantity the estimate represents. Although the analysis lists HIGH Error, NO Error and LOW Error together, the record does not explain how one mean-difference value corresponds to all 3 groups. The interval also has endpoints that match the outer limits of the HIGH Error condition’s reported mean interval, but the registry does not explain whether that reflects the intended analysis quantity. No inference should be made from that resemblance alone.

The registry does not post a separate recalculated comparison from summary statistics, nor does it provide enough detail to independently reproduce the ANOVA. No discrepancy is identified in the registry record. Accordingly, the reported ANOVA remains the registry’s stated analysis result, while its exact estimand and model interpretation remain unclear. Its p-value should not be recast as a specific pairwise finding without a clearly identified contrast.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 3 outcome measures and 2 analyses. It identifies Glucose Area Under the Curve (AUC) Breakfast as the primary endpoint and lists Peak and Nadir Postprandial Glucose Concentration as a secondary outcome. Numerical results for the secondary outcome are not reported in the registry record. The record does not explain which outcomes correspond to the full count of outcome measures or how the 2 listed analyses relate to them.

Multiplicity arises when several outcomes, conditions, time points, or contrasts are tested. In general, conducting multiple tests can increase the chance of at least one apparently notable result under a global null hypothesis. A prespecified approach may control or otherwise address that issue, but the registry record does not report whether multiplicity was considered, whether any adjustment was used, or which comparisons were planned in advance. The ANOVA result does not by itself clarify whether follow-up comparisons were conducted.

Missing-data methods are not reported in the registry record. The registry lists enrollment of 8 and n=5.0 for each condition-specific mean, but does not explain whether those n values represent participants, observations, or another analysis count in the crossover structure. It does not state whether all enrolled participants completed every occasion, why any data might be absent, or whether imputation or another missing-data method was used. The analysis population, including any exclusions, is also not reported.

The handling of repeated observations, condition sequence, and period is not described. These omissions are relevant because the study used a crossover design with different condition orders. The record does not report an interim-analysis plan, stopping rule, or statistical analysis plan. Their absence from the registry description does not establish what may have been planned elsewhere; it means these details 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 safety record lists serious adverse events and other adverse events by sequence group. In every listed group, the number of affected participants is 0 for serious adverse events and 0 for other adverse events, and deaths are recorded as 0. The numbers at risk in the sequence-specific groups are 1 or 2. Those denominators are presented by order of conditions, not as separate parallel treatment arms.

These figures should be interpreted as the adverse-event counts recorded for the listed groups, rather than proof that no adverse events occurred under any circumstance. The registry does not list serious adverse event terms, and the list of top serious terms is empty. It does not explain the event-collection window, whether events were assessed by condition or over the full study, or how safety observations were attributed to a particular condition. The numbers at risk are small, so the zero affected counts provide limited information about uncommon events.

The registry describes six condition-order groups, with the sequence labels indicating the order in which HIGH Error, NO Error and LOW Error were experienced. The adverse-event results do not provide a basis for comparing safety between conditions, because the record does not report condition-specific event exposure or a statistical comparison. These data describe what was recorded in the registry and should not be interpreted as a general safety conclusion.

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

The posted results show different observed mean breakfast AUC values across the three calibration-error conditions: 66.8 for HIGH Error, 48.8 for NO Error and 37.4 for LOW Error. The registry reports 95% confidence intervals for those means and an ANOVA p-value of 0.001. The descriptive means are ordered, with the highest value under HIGH Error and the lowest under LOW Error. The registry’s stated direction is that lower AUC is desirable.

These findings describe this small randomized crossover study and its recorded endpoint. They do not establish the precise size of a within-participant difference between any particular pair of conditions, because pairwise estimates are not reported. The ANOVA result is not sufficiently detailed to identify which contrast its value and confidence interval represent. Nor do the results establish how other glucose outcomes behaved: numerical results for the listed peak and nadir postprandial glucose outcome are not reported in the registry record.

The results do not show whether the findings extend to people or settings beyond those represented in the trial. Enrollment was 8, and the registry does not report detailed participant characteristics. The condition-specific n values are 5.0, but the record does not explain how those counts relate to the full crossover sample. It also does not report the treatment of missing observations or the analysis population.

In general, a statistically small p-value does not by itself establish a meaningful effect, rule out bias, or demonstrate that a result will reproduce. Interpretation depends on the estimate being tested, the analysis assumptions, the design, and the completeness of reporting. Here, the reported ANOVA is evidence from the registry’s stated model, but the model details and the meaning of its estimate remain unclear. The findings therefore support a description of the posted glucose AUC results, not a broader claim about clinical benefit or treatment performance.

14. Why This Trial Matters Statistically

This trial raises a useful statistical issue: the glucose value used by a closed-loop system can be deliberately set higher than, equal to, or lower than blood glucose, and the study examines how breakfast glucose exposure is summarized across those settings. The registry’s condition means and confidence intervals make the distinction between descriptive summaries and inferential comparisons especially important. A reader can identify the observed ordering of the means without assuming that the registry has reported a fully specified pairwise analysis.

The crossover design is also statistically relevant. Participants experienced multiple conditions, and the order varied. In such designs, repeated observations from the same person are related, so an analysis needs to reflect that structure to estimate uncertainty appropriately. The registry identifies ANOVA but does not specify how repeated measurements or condition order were handled. This makes transparent reporting of the model and estimand particularly important for interpreting the p-value and interval.

The record also illustrates why a p-value should be read with the estimate, confidence interval, endpoint definition, and design. The p-value is reported as 0.001, but the analysis label does not identify a clear contrast among the 3 conditions. The condition-specific means are interpretable as summaries, while a specific claim about differences requires a well-defined comparison and analysis.

Finally, the trial’s enrollment of 8 highlights the distinction between a completed study and a precise or broadly generalizable estimate. Small studies can be informative about feasibility and patterns in observed data, but their estimates may be uncertain and sensitive to analysis choices. The registry record leaves important details about participant counts, missing data, and the analysis population unreported. Those gaps do not erase the posted results; they define the limits of what can be inferred from them.

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 trial enrolled 8 participants, and the primary outcome summaries report n=5.0 for each condition. The registry does not explain whether the condition-specific n values represent participants or observations, why they differ from total enrollment, or how many participants contributed measurements under all conditions. This prevents a clear assessment of the completeness of the crossover data.

The registry reports randomized allocation but no masking. It provides descriptions of differing condition orders, yet does not report the detailed randomization method, whether sequence was accounted for analytically, or whether period or carryover effects were evaluated. The statistical model is named as ANOVA, but the record does not establish whether it accounted for the repeated measurements expected in a crossover study.

The primary endpoint is described with a measurement window and sampling frequency, but the record does not provide participant-level data, a variance measure, or detail on how the confidence intervals were calculated. The analysis result’s “Mean Difference (Final Values)” label is not sufficiently specific to identify the quantity tested across the 3 groups. No pairwise comparisons are reported in the registry record.

The registry does not report the analysis population, missing-data approach, multiplicity strategy, sample-size justification, or a statistical analysis plan. Numerical results for the secondary outcome are also not reported. Participant characteristics beyond the listed condition are not described in the registry record in the registry record, limiting assessment of how well the trial population represents broader groups.

These limitations concern the scope and interpretability of the posted evidence. They do not establish that the study methods were inadequate; rather, they mark issues that cannot be resolved from the registry record. The reported means, confidence intervals and ANOVA should be read with those documentation gaps in mind.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Effect of Gain on Closed-Loop Insulin — NCT02065895.
• Registry record last updated 2018-05-17; status last verified 2018-05.
• Results first posted 2018-04-06.
• 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

NCT02065895, titled “Effect of Gain on Closed-Loop Insulin,” is a completed randomized, open-label crossover study sponsored by Joslin Diabetes Center. It lists Type 1 Diabetes, enrollment of 8, and study dates from 2013-12 to primary completion in 2015-04.

The primary endpoint was breakfast glucose AUC, measured from 8:00 AM to 2:00 PM on the day following admission. The reported means were 66.8 for HIGH Error, 48.8 for NO Error and 37.4 for LOW Error, with condition-specific 95% confidence intervals. The registry reports an ANOVA result with p = 0.001, but does not clearly identify the contrast or analysis model details.

Important details, including the handling of crossover correlation, missing data, analysis populations and multiplicity, are not reported in the registry record. The listed secondary outcome has no numerical result reported.