The Set-Point Study for Type 2 Diabetes: Evaluating the Use of an Insulin Only Bionic Pancreas System in Type 2 Diabetes: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for The Set-Point Study for Type 2 Diabetes: Evaluating the Use of an Insulin Only Bionic Pancreas System in Type 2 Diabetes (NCT03552523), a randomized na study. The primary outcome is Mean Dexcom CGM Glucose (CGMG) Level.

2. Clinical Question

3. Trial Design

The Set-Point Study was a randomized, open-label crossover study sponsored by Massachusetts General Hospital. The registry lists the study as completed. Its start date was 2017-07-28, and its primary completion date was 2018-03-29. The phase is listed as NA. The randomized order refers to the sequence in which participants received the study conditions; crossover means that participants completing both arms were observed under both conditions. The registry record does not describe a masking procedure, and masking is listed as none.

The two conditions were an insulin-only bionic pancreas and usual care. In the bionic-pancreas arm, the device was used with a rapid-acting insulin analog, with a set-point of 100 mg/dl stated in the brief summary. In usual care, participants continued their prescribed home insulin regimen, which could be an insulin pump or multiple daily injections. Continuous glucose monitoring and a study-provided glucose meter were described for the conditions. These features define the comparison, but they do not establish that the same procedures or adherence occurred for every participant.

The registry lists enrollment as 16, while the primary-result description states that 7 subjects started participating and 5 completed both study arms; 2 withdrew during the first study arm. The primary endpoint population is described as only completed study arms. Thus, the posted primary results represent the completed crossover experience rather than everyone counted in the enrollment field. The record does not explain the discrepancy between the enrollment field and the participation counts in the primary-result description.

The primary completion date marks the registry’s stated completion milestone, but the registry entries do not provide a detailed schedule for each participant or each arm. The registry record also does not report the rationale for the planned enrollment, the sequence-generation procedure, allocation concealment, or a detailed statistical analysis plan. These details should not be inferred from the randomized crossover 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.

Before assignment: 16 subjects consented and were screened for eligibility. 7 subjects were ineligible. Of the 9 that were eligible, 2 subjects were eligible but not able to schedule their study participation. The other 7 eligible subjects were randomized and started participating in the trial. 5 of those 7 subjects completed both arms of the study.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Only completed study arms are analyzed. 7 subjects started participating, but only 5 subjects completed both study arms. 2 subjects withdrew during the first study arm.

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: 16 subjects were enrolled. 7 subjects participated. 5 subjects completed both arms of the study.

6. Primary Endpoint

The registry lists the primary outcome as Mean Dexcom CGM Glucose (CGMG) Level, measured over days 2-7.

The primary endpoint was Mean Dexcom CGM Glucose (CGMG) Level, expressed in mg/dl. It is the average glucose value from information collected by the Dexcom continuous glucose monitoring device during each control and experimental arm. The stated timeframe was days 2-7. The registry does not further explain why this particular window was chosen or whether other periods were assessed as outcomes.

A mean summarizes the central level of a continuous measurement over the stated monitoring period. For glucose monitoring, it condenses many recorded observations into a value that can be compared across study conditions. That summary is useful for describing the overall recorded level, but it does not show the full pattern of readings, such as variation over time or the frequency of readings at particular levels. Those aspects cannot be inferred from the mean alone.

The posted endpoint is accompanied by a standard deviation for each arm. A standard deviation describes the spread of participant-level values around the mean in the analyzed data; it is not a confidence interval and does not directly quantify the precision of the mean or the treatment contrast. The posted results give means of 152.8 mg/dl with a standard deviation of 21.3 for usual care, and 137.1 mg/dl with a standard deviation of 7.5 for the bionic pancreas. Each arm reports n=5. The registry does not post an inferential comparison or confidence interval for the difference between conditions.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

A randomized crossover design assigns the order of conditions and observes participants under each condition. Because each participant can contribute observations under both conditions, a comparison may use within-participant contrasts. This can account for stable participant characteristics that would otherwise contribute to differences between separate groups. The design does not eliminate all sources of uncertainty: condition order, changes over time, carryover, incomplete participation, and variation in measurements can still matter. The registry identifies the study as randomized and crossover, but the statistical plan and detailed design procedures are not reported in the registry record.

The posted primary outcome is continuous: a mean glucose measurement in mg/dl. The reported arm-level summaries are a mean and standard deviation. A mean represents the center of the observed values, while a standard deviation reflects their spread. These answer different descriptive questions. Similar means can occur with different spreads, and similar spreads can accompany different means. The standard deviation is not the uncertainty interval around a mean or a contrast; those require an inferential calculation and an analysis specification.

For a crossover comparison, arm-level means alone are not enough to reconstruct the uncertainty of the paired contrast. The association between a participant’s measurements under the conditions affects how variable the within-person differences are. The registry does not post the individual paired measurements, the mean paired difference, its standard deviation, a confidence interval, or a test statistic. The primary endpoint therefore can be described by condition, but the precision of the direct comparison cannot be assessed from the posted summaries.

A confidence interval, in general, is a range calculated from data and a statistical model to express uncertainty around an estimate. It conveys which values remain reasonably compatible with the observed data under the method used. An interval for an arm-to-arm contrast would help readers assess both the direction and precision of the estimated difference. No such interval is reported for this outcome. A p-value, in general, assesses how unusual data at least as incompatible with a specified null model would be if that model were true. It does not measure the probability that the null hypothesis is true, nor does it describe the magnitude or clinical importance of a difference. No p-value for the primary comparison is posted.

Standard deviations can be useful for describing participant-level variability, but they do not show the distribution’s shape. A small dataset may be especially difficult to characterize fully with only a mean and standard deviation. Potential skew, outlying values, or differences in the amount of monitoring data per participant are not described in the posted summary. The registry record does not report participant-level distributions or diagnostics, so those features cannot be assessed.

Over-dispersion is a general concern when data vary more than a statistical model assumes. If a method underestimates variability, its uncertainty intervals may be too narrow and tests may be too optimistic. Conversely, a method that accounts for the structure and variability of the data can better reflect uncertainty. This trial’s record does not report the analysis model or assumptions used for the primary endpoint, and no analysis results are listed. It would be inappropriate to infer whether over-dispersion was evaluated or how it was handled.

Repeated glucose readings also have a structure: observations collected from the same participant are related rather than independent in the ordinary sense. A participant-level mean can reduce a stream of readings to a single summary for a defined period, but it does not remove the need to specify how participants and paired conditions are compared. The registry describes the endpoint as an average from the CGM information during days 2-7. It does not explain the construction of each participant’s summary, the treatment of missing readings, or the model used to compare conditions.

The posted arm summaries report n=5 for usual care and n=5 for the bionic pancreas. These values describe the reported outcome rows; the population description states that 5 subjects completed both study arms. The record does not report a separate inferential analysis population or explain whether any outcome-specific exclusions occurred. Although the numerical n is the same for each condition, that does not supply the paired differences or the uncertainty of the crossover contrast.

The registry’s sample-size assumptions describe a planning illustration based on a stated fraction of an observed effect, a two-sided significance level of 0.05, equal allocation between arms, and a normal approximation. The record also says that a missing comparison for a continuous endpoint prevents values from being estimated. No usable planning values are reported for this comparison. Such illustrations, when available, are planning calculations rather than observed power or evidence that the study had a particular probability of detecting an effect. The record does not report the trial’s own sample-size justification, and no such justification should be inferred from a planning illustration.

Randomization and masking are distinct design features. Randomization concerns assignment order; masking concerns whether participants or other parties are kept unaware of condition. The record identifies random allocation and no masking. In an unmasked study, awareness of the condition can potentially affect behavior or measurement processes, depending on the outcome and procedures. The glucose endpoint is collected by a monitoring device, but the record does not discuss possible behavioral effects or specify whether outcome processing was blinded. No conclusion about such effects can be drawn from the registry entries.

Finally, the registry lists zero analyses. This means the record does not present a statistical analysis of the primary comparison beyond the arm-level descriptive summaries. The analytical methods for missing outcomes, crossover sequence, period effects, and multiplicity are not reported in the registry record. Readers should distinguish these absent inferential details from the descriptive values that are posted.

10. Confidence Intervals for Each Arm

No two-arm comparison is shown for this trial, but each arm's mean can be given with a t-based 95% confidence interval computed from the posted mean, spread and number of participants.

R code and output

arm <- c("Usual Care", "Bionic Pancreas")
n   <- c(5, 5)
m   <- c(152.8, 137.1)
sp  <- c(21.3, 7.5)
sd  <- sp      # posted standard deviations
for (i in seq_along(n)) {
  se <- sd[i] / sqrt(n[i]); h <- qt(0.975, n[i] - 1) * se
  cat(sprintf("CARM|%s|%d|%.3f|%.3f|%.3f|%.3f\n", arm[i], n[i], m[i], sd[i], m[i] - h, m[i] + h))
}

CARM  Usual Care  5  152.800  21.300  126.353  179.247
CARM  Bionic Pancreas  5  137.100  7.500  127.788  146.412

11. Multiplicity, Missing Data and Analysis Populations

The registry lists two outcome measures but no posted secondary results. The primary endpoint is identified as mean Dexcom CGM glucose level. No multiplicity adjustment or strategy for interpreting several outcome tests is reported in the registry record. Since no secondary results are posted and no analysis results are listed, the primary arm summaries should not be treated as one result selected from a documented set of inferential tests.

The analysis population is stated explicitly: only completed study arms were analyzed. The description says 7 subjects started participating, 5 completed both study arms, and 2 withdrew during the first study arm. No results for the withdrawals are reported, and the registry does not state how missing outcome data were handled. It does not describe imputation, a sensitivity analysis, or an analysis that includes participants with incomplete crossover data. Those methods cannot be assumed.

Restricting the primary summary to completed arms means the results describe participants with data from both conditions, as stated in the registry. If completion is related to participants’ outcomes or experiences, completer summaries may not represent all who started. Whether that concern applies cannot be determined from this record because reasons for withdrawal and relevant outcome information are not reported. The enrollment field lists 16, while the outcome description gives the participation and completion counts; the record does not reconcile these figures.

The registry record does not report a protocol for handling incomplete CGM recordings within the days 2-7 timeframe. It also does not state whether every participant contributed the same amount of monitoring data or how a participant-level mean was formed when readings were unavailable. These omissions limit evaluation of potential missingness within otherwise completed arms, separately from the reported withdrawals.

12. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

The registry reports serious adverse event figures separately for the bionic-pancreas and usual-care groups. For each group, 7 participants were listed as at risk, and zero serious adverse events were reported as affecting participants. The record also reports zero other adverse events affecting participants in each group and zero deaths in each group. No serious adverse event terms are listed.

These are descriptive safety counts in the registry record. They do not establish that no adverse events of any kind occurred, because the posted figures specify the reported categories and the record gives no further event details. Zero reported serious events is not proof that the underlying risk is zero. A small observed group and limited follow-up can provide limited information about uncommon events; the registry entries do not quantify the uncertainty around these counts.

The adverse-event groups each list 7 participants at risk, whereas the primary endpoint description reports that 5 subjects completed both study arms. These figures refer to different posted summaries and should not be treated as interchangeable analysis populations. The record does not explain the relationship between the adverse-event risk counts and the primary endpoint completer population, or provide a detailed account of adverse-event collection and attribution. The zero counts should therefore be read as the registry’s reported figures, not as a comprehensive estimate of safety in a broader population.

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

The results describe mean CGM glucose levels and their standard deviations during the specified study conditions among the completed study arms. Usual care is reported with a mean of 152.8 mg/dl and standard deviation of 21.3; the bionic-pancreas arm is reported with a mean of 137.1 mg/dl and standard deviation of 7.5. The registry reports n=5 for each arm. These statistics provide a descriptive account of the observed endpoint in this study.

They do not, on their own, establish a statistically supported difference between the conditions. The registry does not report a comparison estimate, confidence interval, or p-value, and arm-level summaries do not reveal the paired participant-level contrasts that are relevant to a crossover study. Without those details, the precision of the comparison cannot be evaluated from the posted results.

The results also do not establish what would occur in a different population, under different usual-care regimens, with another device configuration, or over a different duration. The intervention was specifically described as an insulin-only bionic pancreas using a rapid-acting insulin analog and a set-point of 100 mg/dl in the brief summary. The endpoint covered days 2-7. No longer-term endpoint result is reported in the registry record for this analysis.

The primary summaries apply to completed study arms. The record states that 7 subjects started participating, 5 completed both arms, and 2 withdrew during the first study arm. The available results do not show how outcomes among those who withdrew might compare with outcomes among completers. Nor do the posted summaries show the full distribution of CGM readings or participant-level values.

Thus, the appropriate interpretation is descriptive and bounded: the registry reports the observed means and spreads in the analyzed participants, but does not provide inferential evidence or enough detail to quantify uncertainty in the direct comparison. The results should not be framed as a definitive conclusion that the device does or does not change glucose outcomes.

14. Why This Trial Matters Statistically

This study raises statistical issues that are important when evaluating small crossover trials. A crossover design can make each participant their own comparator, which may help distinguish condition-related variation from stable differences between people. But the value of that structure depends on analyzing paired observations and describing uncertainty in the within-participant comparison. Here, the registry posts arm-specific means and standard deviations but no paired contrast or inferential analysis.

The enrollment and completion information also makes the analysis population central to interpretation. The registry lists enrollment as 16, while the primary-result description says 7 subjects started participating, 5 completed both arms, and 2 withdrew during the first study arm. The endpoint population is restricted to completed study arms. This distinction matters because a descriptive result among completers does not automatically characterize everyone enrolled or everyone who began participation.

The record also illustrates why descriptive results and statistical evidence should be kept separate. Means and standard deviations communicate the level and spread of observed values, while confidence intervals and a prespecified comparison analysis address uncertainty in a contrast. The registry does not report those inferential results. It is therefore possible to describe the posted values without claiming that the observed difference is precise or generalizable.

Finally, the study’s scope is specific: an insulin-only bionic-pancreas configuration, usual-care insulin regimens, and a defined CGM outcome timeframe. Recognizing that scope prevents the result from being extended to unreported configurations, outcomes, or longer periods. The record’s limited reported analysis is itself a statistical consideration, not a basis for filling in unreported methods or conclusions.

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 main limitation in interpreting the primary result is the absence of an inferential comparison. The registry reports arm means and standard deviations but no estimate and uncertainty interval for the paired contrast, and it lists no analyses. This leaves the precision of the direct crossover comparison unquantified. The arm-specific summaries cannot substitute for a paired analysis because they do not show how outcomes from the same participant relate across conditions.

The endpoint population is limited to completed study arms. The record says 7 subjects started participating, 5 completed both arms, and 2 withdrew during the first study arm. It does not report withdrawal reasons or outcomes for those who did not complete both arms. This prevents assessment of whether the completers’ summaries may differ from the experience of those who withdrew. The enrollment field lists 16, and the registry does not explain how that figure relates to the participation counts in the outcome description.

The record does not describe missing CGM observations within the days 2-7 timeframe, how participant-level means were calculated, or how incomplete readings were handled. It also does not report the statistical model, treatment of period or sequence effects, or procedures addressing multiplicity. These methods are not reported in the registry record and should not be inferred.

The study was unmasked, and the registry entries do not describe blinded outcome processing or behavioral effects. The reported standard deviations provide some information about spread, but no participant-level data or distributional summaries are available. Finally, the results concern a specific outpatient insulin-only configuration and a short stated endpoint timeframe; broader settings and longer-term outcomes are not represented by the posted primary result.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: The Set-Point Study for Type 2 Diabetes: Evaluating the Use of an Insulin Only Bionic Pancreas System in Type 2 Diabetes — NCT03552523.
• Registry record last updated 2019-09-10; status last verified 2019-08.
• Results first posted 2019-09-10.
• 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

NCT03552523 is a completed randomized, open-label crossover study sponsored by Massachusetts General Hospital. It compared an insulin-only bionic pancreas with usual care for adults with type 2 diabetes. The primary endpoint was mean Dexcom CGM glucose level in mg/dl during days 2-7.

The registry reports n=5 and a mean of 152.8 mg/dl with standard deviation 21.3 for usual care, and n=5 and a mean of 137.1 mg/dl with standard deviation 7.5 for the bionic-pancreas arm. The primary-result description states that 5 subjects completed both study arms, after 7 started and 2 withdrew during the first arm. Only completed study arms were analyzed.

No inferential analysis of the primary comparison, confidence interval, or p-value is reported. The registry also lists no secondary results. Safety figures report zero serious adverse events, zero other adverse events, and zero deaths in each listed group. These summaries describe the posted record but do not establish the precision of a treatment contrast or broader safety conclusions.