The Insulin-Only Bionic Pancreas Pivotal Trial: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for The Insulin-Only Bionic Pancreas Pivotal Trial (NCT04200313), a randomized na study. The primary outcome is HbA1c.

2. Clinical Question

The central question concerns average HbA1c at the end of the randomized treatment period, rather than a direct comparison between the two insulin formulations. The posted results give separate comparisons against usual care. They do not establish from these analyses whether the two Bionic Pancreas groups differ from one another.

3. Trial Design

This was a completed, randomized, parallel-group trial with no masking reported. The registry lists enrollment of 440 and describes a 13-week randomized comparison of insulin-only Bionic Pancreas use with usual care. The primary outcome was assessed at 13 weeks. The primary-completion date is 2021-10-30, and the recorded start date is 2020-03-31.

The randomized groups with posted primary HbA1c results were the Bionic Pancreas iLet with aspart/lispro insulin, the Bionic Pancreas iLet with Fiasp insulin, and usual care. The recorded group sizes were 219, 114, and 107, respectively. Usual care consisted of participants’ own insulin regimen plus continuous glucose monitoring. The aspart/lispro group included adults and pediatric participants; the Fiasp group description specifies adults. The record does not give further detail about the age composition of the primary-outcome analysis.

The trial description also mentions pathways outside the main randomized comparison. Participants could enter through screening or directly into the randomized protocol. An extension phase was planned in which the usual-care group used the Bionic Pancreas, followed by transition procedures involving either system guidance or pre-study insulin dosing. An optional ancillary study assessed using self-monitored blood glucose instead of continuous glucose monitoring as system input for approximately 48-60 hours. These descriptions do not supply results for those phases or establish that their outcomes belong to the primary randomized comparison.

The record identifies a total of 50 outcome measures and 16 analyses. It does not report the trial’s original sample-size justification, detailed randomization procedures, or the full statistical analysis plan in the information available in the registry record.

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.

Only the first 6 of 7 groups are shown.

RCT

BG Run Ancillary Study

Transition Phase

Extension Phase

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 HbA1c, measured over 13 weeks.

The primary endpoint was central-laboratory hemoglobin A1c, or HbA1c, at 13 weeks, reported as a mean in mmol/mol with standard deviations. HbA1c reflects the proportion of red blood cells that have glycated, meaning that sugar has bonded with them. It is used as a physiological marker of average blood glucose over the preceding period of roughly three months, as described in the registry.

The endpoint is relevant to this trial because the randomized treatment period lasted 13 weeks, so the measurement at its end captures a longer-term glycemic marker rather than a single glucose reading. The endpoint is continuous: each participant’s measurement can take a range of values, and group summaries can describe the mean and the spread of individual measurements. A difference in group means expresses the average separation between the groups, while the standard deviations describe how varied the measurements were within each group.

The registry specifies a superiority question for HbA1c. The posted arm means were 56 mmol/mol for aspart/lispro, 54 mmol/mol for Fiasp, and 61 mmol/mol for usual care. These descriptive summaries do not, by themselves, account for the trial’s mixed-model analysis. The registry record does not report a primary-outcome analysis population or further detail about how the central-laboratory measurements were handled when data were missing.

7. Results Posted in the Registry

• Bionic Pancreas iLet With Aspart/Lispro Insulin (BP-A/L): mean 56.00 versus 61.00 on control; difference -5.00 (95% CI -7.30 to -2.70), standardized effect d = -0.56; z-test p = <0.001.
• Bionic Pancreas iLet With Fiasp Insulin (BP-F): mean 54.00 versus 61.00 on control; difference -7.00 (95% CI -9.39 to -4.61), standardized effect d = -0.78; z-test p = <0.001.

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

CGM-measured Mean Glucose Level Over 13 Weeks

• Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
• Test: Unadjusted two-sample z-test for the difference in means, using the posted standard deviations.
• Result:
  • Bionic Pancreas iLet With Aspart/Lispro Insulin (BP-A/L): mean 164.00 versus 181.00 on control; difference -17.00 (95% CI -23.38 to -10.62), standardized effect d = -0.77; z-test p = <0.001.
  • Bionic Pancreas iLet With Fiasp Insulin (BP-F): mean 155.00 versus 181.00 on control; difference -26.00 (95% CI -32.39 to -19.61), standardized effect d = -1.10; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Bionic Pancreas iLet With Aspart/Lispro Insulin (BP-A/L) than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Bionic Pancreas iLet With Fiasp Insulin (BP-F) than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

CGM-measured Percentage Time 70-180 mg/dL Over 13 Weeks

• Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
• Test: Unadjusted two-sample z-test for the difference in means, using the posted standard deviations.
• Result:
  • Bionic Pancreas iLet With Aspart/Lispro Insulin (BP-A/L): mean 65.00 versus 54.00 on control; difference 11.00 (95% CI 7.57 to 14.43), standardized effect d = 0.90; z-test p = <0.001.
  • Bionic Pancreas iLet With Fiasp Insulin (BP-F): mean 71.00 versus 54.00 on control; difference 17.00 (95% CI 13.46 to 20.54), standardized effect d = 1.29; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically higher for Bionic Pancreas iLet With Aspart/Lispro Insulin (BP-A/L) than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.
  • The 95% confidence interval excludes no difference, so the mean is statistically higher for Bionic Pancreas iLet With Fiasp Insulin (BP-F) than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

CGM-measured Percentage Time >180 mg/dL Over 13 Weeks

• Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
• Test: Unadjusted two-sample z-test for the difference in means, using the posted standard deviations.
• Result:
  • Bionic Pancreas iLet With Aspart/Lispro Insulin (BP-A/L): mean 33.00 versus 44.00 on control; difference -11.00 (95% CI -14.61 to -7.39), standardized effect d = -0.87; z-test p = <0.001.
  • Bionic Pancreas iLet With Fiasp Insulin (BP-F): mean 27.00 versus 44.00 on control; difference -17.00 (95% CI -20.71 to -13.29), standardized effect d = -1.23; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Bionic Pancreas iLet With Aspart/Lispro Insulin (BP-A/L) than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for Bionic Pancreas iLet With Fiasp Insulin (BP-F) than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The posted secondary continuous outcomes show an overall pattern across the two Bionic Pancreas groups that is directionally consistent with the primary HbA1c comparisons. For the listed glucose-monitoring measures, the group summaries and sponsor-reported superiority tests indicate differences versus usual care in mean glucose, time in the 70-180 mg/dL range, and time above 180 mg/dL. The registry lists further glucose outcomes, including additional high- and low-range measures, among its secondary outcomes.

These outcomes provide related information about glucose levels and time spent in specified glucose ranges, whereas HbA1c captures a longer-term average glycemic marker. Agreement in direction across related measures can make the findings more coherent, but the endpoints are not identical and should not be treated as independent confirmations of the primary result. They also do not answer whether the intervention groups differ from one another.

The record includes 45 posted secondary outcomes and reports 16 analyses overall. It does not report whether the secondary analyses were adjusted for multiple comparisons. When many outcomes or tests are examined, the chance of at least one apparently notable result from random variation can increase. Without a stated multiplicity procedure or a clear hierarchy for interpreting secondary outcomes, individual unadjusted secondary p-values carry less confirmatory weight than the prespecified primary question.

The registry identifies non-inferiority for continuous glucose monitoring-measured time below 54 mg/dL as a key secondary endpoint, but the result details for that outcome are not among the secondary summaries described in the posted results. The non-inferiority margin and analysis details are not reported in the registry record. No conclusion about that endpoint should be inferred from the other glucose measures.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Non-inferiority for CGM-measured Time <54 mg/dL (Key Secondary Endpoint)
• CGM-measured Mean Glucose Level Over 13 Weeks
• CGM-measured Percentage Time 70-180 mg/dL Over 13 Weeks
• CGM-measured Percentage Time >180 mg/dL Over 13 Weeks
• CGM-measured Percentage Time >250 mg/dL Over 13 Weeks
• CGM-measured Glucose Standard Deviation (SD) mg/dL Over 13 Weeks
• CGM-Measured Percentage Time <70 mg/dL Over 13 Weeks
• CGM-measured Percentage Time <54 mg/dL Over 13 Weeks

9. Statistical Interpretation

10. Statistical Methods Explained

The primary outcome is continuous, so its group summaries are means and standard deviations. A mean gives the arithmetic average of participants’ measurements in a group. A standard deviation describes how dispersed measurements are around that group mean. It is not the uncertainty in the mean; that uncertainty depends both on the spread of observations and on the number of observations contributing to the estimate.

What does a mean difference express?

A mean difference compares the average outcome in one group with the average in another. For the recalculated HbA1c comparisons, the reference group is usual care. A negative estimate means the intervention-group mean is lower than the usual-care mean under that ordering. The estimate is expressed in the endpoint’s original unit, mmol/mol, which keeps the difference interpretable in the same measurement scale as HbA1c.

The recalculations use the posted arm means, standard deviations, and sample sizes to make simple unadjusted comparisons. They are not adjusted model estimates and do not reproduce the reported mixed-model analysis. The registry’s own primary analyses are described as mixed models and as superiority tests. The record does not report sufficient model details, such as the full covariate structure or treatment of repeated measurements, to recreate those analyses.

How should the p-values be read?

A p-value describes how compatible the observed data are with a specified null model, under the assumptions of the analysis. It is not the probability that the null hypothesis is true, nor is it a measure of the size or practical importance of a difference. For both primary comparisons, the sponsor-reported mixed-model p-value is < 0.001. The recalculated unadjusted p-values are also < 0.001. These should be distinguished from the confidence intervals around the recalculated estimates: the p-value addresses evidence against a null hypothesis, while an interval conveys a range of estimates compatible with the data and method.

In general, a mixed model can account for structured data such as measurements taken repeatedly from participants, depending on the model specification. That general capability does not establish which specific repeated measurements, covariates, or correlation structures were used in this trial. Those details are not reported in the registry record.

What does statistical power mean in planning?

Power is a design concept: under stated assumptions and for a specified effect, it is the chance that a planned statistical test will reject its null hypothesis. The sample-size illustrations use the observed effect as a starting point and consider smaller fractions of that effect. Smaller effects are harder to distinguish from ordinary sampling variation, so they generally require more participants for a given planned power. The illustrations use a two-sided significance level of 0.05, equal allocation between the two groups, and a normal approximation. They are not the trial’s documented design justification.

General statistical caution is needed when planning from an observed effect. An observed estimate is uncertain and can be larger or smaller than the underlying effect. A planning calculation based on it is therefore a conditional illustration, not a guarantee about future results. The registry’s original design assumptions and sample-size justification are not reported in the registry record.

Why can variation matter?

Greater variability within groups makes it harder to estimate a mean precisely, all else equal. The posted standard deviations differ across groups: for HbA1c they are 7.7 for aspart/lispro, 6.6 for Fiasp, and 10.9 for usual care. This variation is part of why precision depends on both the observed spread and the group sizes. The recalculated intervals incorporate those summary inputs under an unadjusted comparison; they do not express all uncertainty about model choice, missing data, or trial conduct.

11. Confidence Intervals and Precision

The recalculated HbA1c confidence interval for aspart/lispro versus usual care extends from -7.303 to -2.697 mmol/mol around a mean difference of -5.0 mmol/mol. For Fiasp versus usual care, the recalculated interval extends from -9.394 to -4.606 mmol/mol around a mean difference of -7.0 mmol/mol. Both intervals exclude zero, indicating that the unadjusted estimates are compatible with a difference below zero under the calculation used.

A confidence interval represents statistical uncertainty around an estimate under a particular method and its assumptions. It does not mean that the true effect has a specified probability of falling inside this particular interval. The interval communicates both direction and precision: a narrower interval indicates more precision on the measurement scale, while a wider interval leaves more plausible values in view. These are unadjusted intervals calculated from posted summary statistics, not intervals reported for the sponsor’s mixed-model estimates.

The two interval widths should not be interpreted as a direct comparison of treatment performance. The intervention groups have different sizes and spreads, and each estimate uses usual care as its comparator. The registry does not post a direct estimate or interval comparing the two intervention groups. In addition, the registry’s own reported analysis entries give p-values but not model-based effect estimates or confidence intervals, limiting direct assessment of the model-based precision.

Precision is also distinct from generalizability. An interval quantifies sampling uncertainty under its method, not uncertainty about whether the trial population or 13-week treatment period represents other settings. The record does not provide information on missing primary measurements or enough analysis detail to assess how those issues may affect precision.

12. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The sample-size illustrations are stated as participants per arm for the continuous outcomes in this trial. They are not total-event counts. The planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The calculations use a specified fraction of the observed effect as the effect to detect. The original trial design assumptions and its sample-size justification are not reported in the registry record.

The 80% power column describes the number of participants per arm that the illustration estimates would be needed for an 80% chance of detecting the chosen effect under those planning assumptions. The 90% power column describes the corresponding number per arm for a 90% chance. Power is a property of a planned test under its assumptions, not a statement about whether the observed trial result was likely or unlikely after the fact.

For example, for the primary aspart/lispro comparison, planning for 75% of the observed effect gives 88 participants per arm for 80% power and 118 participants per arm for 90% power. Planning for 50% of the observed effect gives 198 participants per arm for 80% power and 265 participants per arm for 90% power. The smaller target effect requires more participants because it is harder to distinguish from random variation. The 100% observed-effect scenario is more optimistic as a planning target, since it assumes the full observed effect will be the effect to detect.

These figures are planning illustrations based on the observed results, not the study’s original design calculations. They are not observed power.

13. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, Δ is the difference in means to detect and σ is the common standard deviation.

Worked calculation for this trial

1. Observed difference Δ = -5.000; pooled standard deviation σ = 8.875.
2. n = 2 × (1.96 + 0.842)² × 8.875² / -5.000² = 49.5, rounded up to 50 participants per arm for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
delta <- -5                      # observed difference in means
sigma <- 8.87482915850173                     # pooled standard deviation
n_per_arm <- function(power, frac = 1) {
  ceiling(round(2 * (z_a + qnorm(power))^2 * sigma^2 / (delta * frac)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (t-based, so very slightly larger)
cat("power.t.test:", round(power.t.test(delta = delta, sd = sigma, power = 0.80)$n, 1), "\n")

# Simulation: exact two-sample t-test using simulated means and variances
set.seed(4200313)
n    <- n_per_arm(0.80)
reps <- 20000
m_diff <- rnorm(reps, delta, sigma * sqrt(2 / n))             # difference in sample means
s2     <- sigma^2 * rchisq(reps, 2 * (n - 1)) / (2 * (n - 1))   # pooled sample variance
tstat  <- m_diff / sqrt(s2 * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(tstat) > qt(0.975, 2 * (n - 1))), reps))

Output from running the code

100% of observed effect: n per arm = 50 (80% power), 67 (90% power)
 75% of observed effect: n per arm = 88 (80% power), 118 (90% power)
 50% of observed effect: n per arm = 198 (80% power), 265 (90% power)
power.t.test: 50.4 
Simulated power at n = 50 per arm: 0.795 (20000 replicates)

The simulation draws many sets of group means and variances at the planned sample size and counts how often the two-sample t-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

14. Analyses Reported by the Sponsor

The registry reports mixed-model superiority analyses for both primary comparisons: aspart/lispro versus usual care and Fiasp versus usual care each have p < 0.001. The registry entries do not include a model-based effect estimate or confidence interval, so those entries communicate the reported test result but not the model-based magnitude and precision.

The simple recalculations use the posted arm means, standard deviations, and group sizes. They estimate mean differences of -5.0 mmol/mol for aspart/lispro versus usual care and -7.0 mmol/mol for Fiasp versus usual care, with confidence intervals from -7.303 to -2.697 and -9.394 to -4.606, respectively. Their unadjusted p-values are each < 0.001. The direction and broad statistical conclusion align with the registry’s reported analyses.

Alignment does not make the approaches interchangeable. Mixed models can incorporate features not represented in an unadjusted calculation from group summaries. The record does not give enough detail to reconstruct the sponsor’s model, and the recalculations should be treated as checks based on the posted summaries rather than replacement analyses. The discrepancy flag is false, so the posted summary results do not indicate a discrepancy requiring a different interpretation of the sponsor’s analysis.

15. Multiplicity, Missing Data and Analysis Populations

The registry identifies HbA1c at 13 weeks as the primary endpoint and reports superiority analyses for each Bionic Pancreas group against usual care. It also lists 50 outcome measures, 45 posted secondary outcomes, and 16 analyses. The registry record does not report a multiplicity-adjustment procedure, a full hierarchy of primary and secondary hypotheses, or whether the secondary analyses were intended to be confirmatory or exploratory.

Multiplicity matters because testing many outcomes can increase the chance of at least one apparently unusual result even when there are no underlying differences. Procedures such as adjusting significance thresholds or controlling a family-wise error rate can address that issue, but no such procedure should be assumed for this trial from the available record. Thus the reported secondary p-values should be read with more caution than the primary test, especially when considered individually.

The record describes mixed-model analyses for the primary comparisons but does not report the model specification, covariates, repeated-measure structure, or analysis population. It also does not state how missing outcome measurements were handled, whether imputation was used, or how participants were included in the analysis after randomization. These details affect interpretation because exclusion or handling of incomplete observations can influence both estimates and uncertainty.

The registry record does not report the original sample-size justification, interim analyses, or a detailed statistical analysis plan. The absence of these details in the registry record does not establish that such procedures did or did not occur; it means they cannot be assessed from the record’s posted information. For the key secondary non-inferiority outcome, the margin and analysis details are likewise not reported in the registry record.

16. 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.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: severe hypoglycemia (4.6%), diabetic ketoacidosis (1.8%), abdominal pain (1.1%), epiglottitis (0.9%), primary spontaneous pneumothorax (0.9%).

Safety figures are reported across several groups and phases, so their denominators should be read within the named group rather than treated as one common randomized comparison. In the primary randomized groups, serious adverse events were recorded for 13 of 219 participants in the aspart/lispro group, 5 of 114 in the Fiasp group, and 4 of 107 in usual care. The registry also reports other, non-serious adverse events for 119 of 219, 57 of 114, and 4 of 107, respectively. No deaths are listed in these groups.

Additional groups cover blood-glucose-measurement run-in, transition, and extension periods. Their serious-event counts range from 0 of 58 to 3 of 90, while the transition groups each list 0 serious events. These groups represent different phases or configurations and should not be combined casually with the main randomized groups.

The registry lists severe hypoglycemia at 4.6% among the top serious-event terms, followed by diabetic ketoacidosis at 1.8%. It also lists abdominal pain at 1.1%, epiglottitis at 0.9%, primary spontaneous pneumothorax at 0.9%, and attempted suicide at 0.9%. These summaries describe reported events; they do not establish causal attribution or permit conclusions about comparative risk without fuller event definitions and analysis details.

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

The primary results show that the posted mean HbA1c at 13 weeks was lower in each Bionic Pancreas group than in usual care. The registry’s mixed-model analyses report superiority for both comparisons, with p < 0.001. The simple recalculations from the posted summaries are consistent in direction and also yield intervals that exclude zero. Together, these statistics describe the observed group averages and the uncertainty around the unadjusted comparisons.

They do not mean that every participant experienced a lower HbA1c, nor do group means describe all individual responses. Statistical evidence against a null comparison does not establish the clinical importance of the difference; the record does not provide a threshold for a clinically important HbA1c change. The reported analyses also do not compare the two Bionic Pancreas configurations directly, so they do not show that one insulin formulation is preferable to the other.

The conclusions are limited to the trial’s recorded population, randomized groups, and 13-week primary endpoint. The registry describes adults and children in the trial overall, but the Fiasp arm description specifies adults and detailed age-specific outcome results are not reported in the registry record. The extension and transition descriptions are separate from the primary randomized comparisons, and their results should not be inferred from the HbA1c findings.

Safety counts and listed adverse-event terms are important context, but the posted summaries alone do not provide a complete comparative safety analysis. The registry record also does not report the missing-data methods, full analysis population, multiplicity plan, or model details needed for a fuller appraisal. The results therefore communicate evidence about particular posted comparisons rather than resolving every question about longer-term outcomes, subgroup effects, or comparative safety.

18. Why This Trial Matters Statistically

This trial illustrates why statistical interpretation benefits from separating a result’s direction, magnitude, precision, and analysis method. The posted group means establish the descriptive pattern, the recalculated mean differences express its size in the endpoint’s units, and the confidence intervals convey uncertainty under an unadjusted method. The sponsor-reported mixed-model p-values address the trial’s stated superiority comparisons but are not a substitute for an effect estimate and interval.

It also demonstrates the importance of matching the interpretation to the actual comparison. Each Bionic Pancreas configuration is compared with usual care; neither comparison directly tests one configuration against the other. That distinction prevents an apparent numerical difference between intervention-group means from being mistaken for a formal head-to-head result.

The many secondary outcomes show why a clear analysis hierarchy matters. Related measures can help describe a broader outcome pattern, but multiple testing and incomplete reporting of adjustment plans limit how strongly individual secondary results should be interpreted. The key secondary non-inferiority question requires its own margin and analysis details, which are not reported in the registry record.

Finally, planning illustrations based on observed effects are not the same as the original sample-size rationale or post-trial power. They show how a target effect and planning assumptions influence required sample sizes, while underscoring that smaller effects call for more participants. This distinction is useful when reading results without confusing retrospective planning scenarios with the design decisions made before the trial.

19. 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 is described as unmasked, which can matter for outcomes and care processes that participants or study personnel may influence. The primary endpoint is a laboratory measure, but the registry does not report enough detail to assess whether other aspects of outcome collection or follow-up could have been affected by the lack of masking.

The primary randomized period lasted 13 weeks. The posted primary endpoint therefore addresses the specified end-of-period HbA1c comparison, not long-term persistence of the observed group differences. The registry describes an extension and transition phase, but those descriptions do not provide results that can extend the primary comparison.

Usual care was participants’ own insulin regimen plus continuous glucose monitoring, and the record notes multiple daily injections or pump therapy. This makes usual care a clinically varied comparator rather than one uniform insulin regimen. The registry record does not report subgroup analyses by usual-care regimen or by age, and it does not provide detailed age composition for the primary-outcome groups.

The Fiasp group description specifies adults, while the aspart/lispro and usual-care descriptions include adults and pediatric participants. The record does not report enough detail to judge the comparability of age distributions or to interpret age-specific effects. Nor does it give a direct randomized comparison between the two intervention configurations.

Finally, the model specification, missing-data handling, analysis population, multiplicity approach, and original sample-size rationale are not reported in the registry record. These omissions constrain independent assessment of the sponsor’s model-based estimates and the strength of claims based on secondary analyses.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: The Insulin-Only Bionic Pancreas Pivotal Trial — NCT04200313.
• Registry record last updated 2025-02-20; status last verified 2025-02.
• Results first posted 2023-11-07.
• 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.

23. Record Summary

NCT04200313 records a completed randomized, parallel-group, unmasked trial of insulin-only iLet Bionic Pancreas configurations versus usual care in adults and children with type 1 diabetes. The primary outcome was central-laboratory HbA1c at 13 weeks. The posted means were 56 mmol/mol for aspart/lispro, 54 mmol/mol for Fiasp, and 61 mmol/mol for usual care; the sponsor reports mixed-model superiority results of p < 0.001 for each comparison with usual care.

Unadjusted recalculations from the posted summaries show mean differences of -5.0 mmol/mol and -7.0 mmol/mol, respectively, with confidence intervals excluding zero. These calculations are not replacements for the sponsor’s mixed models. The record also includes glucose-monitoring secondary outcomes, safety summaries, and descriptions of extension, transition, and ancillary study phases. Important statistical-plan details, including missing-data handling, multiplicity procedures, and the original sample-size justification, are not reported in the registry record.