QBSafe: a Randomized Trial of a Novel Intervention to Improve Care for People Living With Type 2 Diabetes.: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for QBSafe: a Randomized Trial of a Novel Intervention to Improve Care for People Living With Type 2 Diabetes. (NCT05553912), a randomized na study. The primary outcome is Illness Intrusiveness Measured by the Illness Intrusiveness Ratings Scale (IIRS) at Month 6.

2. Clinical Question

The trial’s motivating concern was that diabetes care can focus heavily on glycemic control, potentially leaving patients’ wider health and wellbeing less visible. QBSafe was designed to make patients’ own priorities part of a clinical conversation, rather than to test a medication or a change in glycemic targets. The registry describes the work as examining both the feasibility of research procedures and intervention effects on patient-reported outcomes.

This distinction matters when interpreting the question. The primary endpoint is a patient-reported measure of how illness and its treatment intrude on life domains; it is not a direct measure of glucose control, complications, or clinician behavior. The registry does not report a separate primary feasibility outcome or a detailed definition of feasibility. Its reported primary outcome is the month 6 IIRS score, with the analysis described as adjusted for the initial score and arm.

3. Trial Design

The registry describes QBSafe as a randomized, parallel trial with no masking. The listed intervention was behavioral, and the comparator was usual care. The study status is completed. Enrollment is reported as 154. The record lists Yale University as sponsor and gives the condition as type 2 diabetes.

The intervention combined participant-facing conversation cards with materials for clinicians. The cards were intended to help participants identify life concerns related to diabetes and communicate those concerns in clinical care. The clinician materials were intended to support responses to what participants raised. The registry gives no further detail about the contents of the materials, the training or support clinicians received, the frequency of intervention contacts, or how fidelity was assessed. Arm descriptions for QBSafe and usual care are blank in the record, so those details cannot be reconstructed from the arm fields.

The record gives a start date of 2022-11-02 and a primary completion date of 2025-02-24. It lists phase as NA. No randomization ratio or allocation procedure is reported in the registry record. Although the design fields describe a randomized parallel trial, the analysis notes identify the design as cluster-randomized and caution that simple individual-level recalculation would be misleading. The registry does not report the cluster unit, the number of clusters, or details of how clustering was handled in the sponsor’s models.

The primary endpoint was illness intrusiveness on the IIRS at month 6. The analysis description says the month 6 response would be assessed as the outcome, adjusting for arm and initial score. Secondary outcomes listed at month 6 were diabetes distress, measured with the DDS-17, and treatment burden, measured with the TBQ. The record lists 9 outcome measures and 11 analyses. The mapping of every analysis to an outcome is not reported in the registry record.

The record’s stated purpose includes feasibility of research procedures as well as patient-reported outcomes. It does not provide the protocol-level detail needed to assess recruitment targets, adherence, intervention exposure, or feasibility criteria. Those omissions limit how fully the trial’s operational aims can be evaluated from the registry entry 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

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 Illness Intrusiveness Measured by the Illness Intrusiveness Ratings Scale (IIRS) at Month 6, measured over Month 6.

The primary endpoint is illness intrusiveness at month 6, measured using the Illness Intrusiveness Ratings Scale. The scale asks participants how much their illness and/or its treatment interferes with life domains considered central to quality of life. It contains 13 items, and the overall score is formed by summing item responses, producing a possible score from 13 to 91. Higher scores indicate greater illness intrusiveness. The registry also describes subscale scores from 1-7, with higher scores representing more disruption or intrusiveness.

This endpoint is relevant to QBSafe’s stated aim because the intervention is intended to help patients bring concerns about living with diabetes into clinical conversations. A patient-reported measure of interference with daily life addresses a different aspect of care from a laboratory measure or a clinical event. It gives participants’ reported experience a direct role in evaluating the intervention.

The registry describes the primary analysis as using the month 6 response as the outcome and adjusting for treatment arm and the initial score. In general, accounting for a baseline score can help compare follow-up outcomes while recognizing where participants started. The record does not provide the observed arm-specific scores, estimated between-arm effect, confidence interval, or endpoint-specific p-value. Therefore, the scale’s meaning and planned analysis can be described, but the size and precision of the posted primary endpoint result cannot be established from the reported figures.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

What does a mixed model do?

The registry labels the reported comparisons as mixed models analysis. In general, a mixed model can represent relationships in outcome data while accounting for more than one source of variation. The exact structure depends on how observations were collected and on the trial’s design. The registry does not report the full model specification, the terms included beyond the stated adjustment for arm and initial score, or the estimation details. It is therefore not possible to determine from the record which particular mixed-model formulation was used for each outcome.

For the primary endpoint, the stated approach treats the month 6 response as the outcome and adjusts for treatment arm and the initial score. In general, baseline adjustment compares groups at follow-up while accounting for their starting scores. This can improve the relevance of the comparison when baseline outcome values are informative. The registry does not give the numerical baseline or follow-up summaries, so the extent of any baseline imbalance or its influence on the estimate is not reported.

Why does cluster randomization matter?

The analysis notes state that this was a cluster-randomized design and warn that a simple individual-level recalculation would be misleading. In a cluster-randomized trial, allocation occurs at the level of groups rather than independently for each person. Participants within the same group may share experiences or care settings, so their outcomes may be correlated. General statistical methods account for that dependence because treating correlated observations as fully independent can make uncertainty appear smaller than it is.

The registry’s general design fields also say randomized and parallel, but they do not describe the clusters. The cluster unit, number of clusters, allocation process, and model’s specific handling of within-cluster correlation are not reported in the registry record. Those omissions matter because readers cannot verify how the design was represented in the analysis. The record explicitly directs readers away from an individual-level recalculation and toward the sponsor’s own analysis.

What can p-values tell us?

The posted analysis information gives four p-values: 0.24, 0.44, 0.39, and 0.28. In general, a p-value assesses how surprising data would be under a statistical model and a stated null hypothesis. It does not state the size of an effect, its direction, or whether an effect matters to patients. Those interpretations require estimates and uncertainty intervals, along with an understanding of the outcome scale and the comparison being made.

Here, no estimate or confidence interval is attached to the listed analyses. Nor does the registry identify the outcome associated with each p-value. The result listing therefore cannot support a specific numerical interpretation of the primary endpoint. A confidence interval, when reported, describes a range of values compatible with the data and model at a stated confidence level; its width conveys precision. No interval is reported for these analyses, so their precision cannot be assessed from the p-values alone.

How should the outcome scale be understood?

The IIRS is a continuous measure of illness intrusiveness. Its overall score sums responses to 13 items, with possible values from 13 to 91; higher overall scores indicate increased intrusiveness. This direction helps interpret a reported score or contrast if one is available. But the registry does not give the observed scores or an estimated contrast, so the scale description cannot establish whether the intervention group had more or less reported intrusiveness at month 6.

General interpretation of a continuous outcome depends on both the estimated difference and its uncertainty. A p-value without the estimate does not show the scale-point difference, while an estimate without uncertainty would not show how precisely it was measured. Neither is available in the listed analysis fields. The endpoint’s patient-reported nature also means that its interpretation is about participants’ reported experience, not a direct measure of disease physiology or clinical events.

What do the planning illustrations mean?

Any sample-size figures on the page are planning illustrations computed from an observed effect under stated assumptions, not the trial’s original design justification. The stated illustration assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation; the effect to detect is expressed as a fraction of the observed effect. Such illustrations ask how many participants per arm would be needed under those assumptions for a selected planning target. They do not describe what the trial actually observed and are not observed power.

The registry record does not report the trial’s original sample-size justification or the assumptions used to plan enrollment. Since the analysis notes also caution against individual-level recalculation for a cluster-randomized design, illustrations based on a simplified model should not be mistaken for a design recommendation that accounts for clustering. The assumptions attached to an illustration must be read as assumptions, not as evidence that the original trial used them.

What methodological details remain unavailable?

The record does not report the full statistical analysis plan, model formula, missing-data methods, analysis populations, or how the 11 analyses relate to the 9 outcome measures. It also does not state whether interim analyses occurred or explain how multiplicity across outcomes and analyses was handled. These details affect how readers assess model choice, precision, and the risk of selectively emphasizing results. Their absence does not show that a method was or was not used; it means the registry entry does not document it.

The key methodological reading is consequently restrained. The registry reports mixed-model comparisons and gives p-values, while the analysis note rules out a simple individual-level recalculation. Without estimates, intervals, endpoint mapping, and fuller cluster-analysis details, the posted information does not permit a complete assessment of the primary result’s magnitude or precision.

10. Analyses Reported by the Sponsor

The record reports the sponsor’s own analyses as mixed models, with four p-values: 0.24, 0.44, 0.39, and 0.28. The analysis note says the trial was cluster-randomized and that a simple individual-level recalculation would be misleading. Accordingly, the page does not present an individual-level recalculation as an alternative estimate. For this design, the sponsor’s analysis takes precedence over a calculation that ignores the stated clustering concern.

There is no recalculated effect estimate to compare with the sponsor’s result. The registry provides no model estimates, confidence intervals, or mapping of each listed p-value to a specific endpoint. It is also not possible to verify the sponsor model’s exact handling of clusters from the available description. These limitations mean the sponsor’s analysis can be identified as the reported analysis, but its numerical effect and precision cannot be reconstructed from the listed p-values.

Any planning illustrations shown elsewhere on the page are not a reanalysis of the trial and are not observed power. They use stated assumptions and should not be read as replacing the sponsor’s analysis or as evidence about the trial’s original design calculations. The registry record does not report the original sample-size justification. The sound comparison is therefore between the reported sponsor analysis and the limits of the information available, not between competing effect estimates.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists 9 outcome measures and 11 analyses, but it does not report how the analyses map to the outcomes or which analysis corresponds to each endpoint. It also does not state whether the listed tests were all planned, whether any were exploratory, or whether an adjustment for multiple comparisons was used. When several outcomes or analyses are considered, multiplicity can increase the chance that at least one result appears notable by chance. In general, a multiplicity plan clarifies how to interpret a collection of tests; the registry record does not document one.

The primary endpoint is identified as the month 6 IIRS score, and the planned analysis is described as adjusting for arm and initial score. However, the results listing does not identify the primary endpoint’s p-value, and it gives no endpoint-specific estimate or confidence interval. The secondary outcomes listed are diabetes distress on the DDS-17 and treatment burden on the TBQ, both at month 6. The record does not provide enough information to assign the four reported p-values to these or other measures.

Missing-data methods are not reported in the registry record. The record does not say how absent month 6 assessments were handled, whether participants were analyzed according to their randomized group, or how analysis populations were defined. It also does not give outcome-specific participant counts. These omissions matter because the available participants and assumptions about unobserved responses can affect estimates and their uncertainty.

The total enrollment is reported as 154, while the safety groups list 69 participants at risk in usual care and 74 in QBSafe. The record does not explain how those safety counts relate to the analysis populations for each patient-reported outcome. They should not be used to infer completion or missingness for the IIRS, DDS-17, or TBQ.

The registry also does not report whether interim analyses occurred or how any such analyses would have been incorporated into the final interpretation. No inference should be made about these undisclosed procedures. Overall, the missing information limits assessment of multiplicity, attrition, and population definitions; it does not establish that any particular analysis practice was or was not followed.

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’s safety figures report no participants with a serious adverse event in either listed group: 0 of 69 at risk in usual care and 0 of 74 at risk in the QBSafe intervention group. The record lists 1 death for usual care and 2 for QBSafe. It does not provide further details about the deaths, their timing, causes, or relationship to the intervention. The listed death counts should therefore be described as recorded, not interpreted as caused by or unrelated to study participation.

Zero reported serious adverse events does not establish that no adverse experiences occurred. The record gives no terms for serious events in its top-term list and does not provide a fuller account of event types, duration of safety follow-up, or ascertainment procedures. It also does not state whether the reported at-risk counts cover everyone enrolled or only those included in the safety reporting set.

In general, safety counts are interpreted in the context of how events were defined, collected, and followed over time, as well as the number of participants observed. The registry’s summary does not supply those details. The figures are descriptive counts; they do not by themselves establish comparative safety or explain the deaths. They should be kept separate from the patient-reported primary endpoint and from the mixed-model p-values.

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

This trial addresses a meaningful statistical question about whether a behavioral intervention centered on patient priorities can be evaluated using patient-reported outcomes. Its primary endpoint, illness intrusiveness, reflects participants’ reported experience of how diabetes and its treatment interfere with life domains. That focus complements outcomes centered on glycemic control, consistent with the rationale stated in the registry. The listed secondary measures also concern participants’ experiences, including diabetes distress and treatment burden.

The record establishes that QBSafe was tested in a completed randomized parallel trial with no masking and enrollment of 154. It describes a month 6 primary endpoint and reports four mixed-model p-values. However, the endpoint-specific effect estimate and confidence interval are not reported in the analysis listing, and the p-value associated with the IIRS is not identified. The posted information therefore does not establish the direction, size, or precision of the primary outcome comparison.

These results do not, on their own, show that QBSafe improved or worsened illness intrusiveness. Nor do they establish that the intervention changed diabetes distress, treatment burden, glycemic control, or other outcomes. The registry record does not give enough outcome-specific estimates to support those claims. A p-value is not a substitute for an effect estimate, and the reported p-values cannot answer whether any observed difference would be important to participants.

The trial’s stated aim included feasibility, but the registry record does not provide feasibility criteria or a detailed set of feasibility results. The intervention’s design can be described, but the record does not tell readers how often participants used the cards, how clinicians responded, or how faithfully materials were delivered. Those are relevant to understanding how an intervention operates, not just whether a score differs.

Finally, the record’s cluster-randomization note makes the analysis structure important. It cautions against treating all participants as independently randomized, yet does not report enough cluster details to examine the analysis fully. The most defensible conclusion is limited: a randomized trial of QBSafe was completed, and mixed-model p-values were posted, but the available registry results do not provide the primary effect’s direction or precision. No medical recommendation follows from these registry data.

14. Why This Trial Matters Statistically

QBSafe is statistically notable because its stated rationale centers on a limitation of relying narrowly on glycemic control to represent the quality of diabetes care. The primary endpoint instead asks about illness intrusiveness: how much illness or treatment interferes with areas of life that matter to participants. This makes the choice of outcome part of the scientific question. A patient-reported score can capture an experience that is not interchangeable with a laboratory value or a count of clinical events.

The intervention also highlights the importance of separating an intervention’s intended mechanism from its measured endpoint. Conversation cards and clinician materials were intended to support discussion of participant priorities. The IIRS measures reported interference in life domains. A result on that scale would speak to the measured experience, but would not by itself establish whether conversations changed, whether clinicians acted differently, or whether other aspects of care changed. The registry does not report detailed process measures that would connect those steps.

The trial is also a reminder that the randomization unit shapes valid statistical interpretation. The registry’s analysis note identifies cluster randomization and cautions against simple individual-level recalculation. Accounting for dependence within clusters is a general requirement when allocation and shared context operate at the group level. Because cluster details are not reported, readers cannot fully inspect how that requirement was implemented.

Finally, the result record demonstrates why p-values alone are not enough. Without effect estimates and confidence intervals, the listed values do not show the magnitude, direction, or precision of a comparison. This is especially consequential for a patient-reported outcome, where the scale’s scoring direction is interpretable but the observed group contrast is absent from the analysis listing. Clear reporting of estimates, uncertainty, outcome mapping, and missing-data methods would make the findings easier to assess.

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.

Several limitations specific to this trial arise from gaps in the registry record. The primary outcome is clearly described, but its numerical result is not: no arm-specific IIRS summaries, model-based effect estimate, or confidence interval is listed. Four mixed-model p-values appear, but the record does not identify which belongs to the primary endpoint. As a result, the primary comparison’s direction, magnitude, and precision cannot be evaluated from the posted analysis information.

The record identifies the design as cluster-randomized in its analysis note, but does not state the cluster unit, number of clusters, or how the mixed models represented clustering. This prevents readers from independently assessing whether the analysis reflected the allocation structure. The note explicitly cautions that a simple individual-level recalculation would be misleading, so such a recalculation is not an appropriate substitute.

Other important analysis details are also absent. The registry does not report missing-data procedures, analysis populations, outcome-specific sample counts, the complete model specification, or whether results were adjusted for multiple comparisons. It lists 9 outcome measures and 11 analyses without mapping every analysis to an outcome. This limits assessment of how the results were selected and how the p-values should be interpreted together.

The intervention and comparator are described only at a broad level. The QBSafe concept includes participant conversation cards and clinician support materials, but the record does not describe delivery frequency, adherence, clinician training, or intervention fidelity. The usual-care arm description is blank, so readers cannot assess the content of the comparator. Feasibility is part of the stated trial purpose, but feasibility criteria and detailed results are not reported.

Safety reporting is also brief. Serious adverse event counts are zero in both groups, and deaths are listed, but event descriptions, causes, timing, and follow-up details are absent. These limitations do not show that procedures were not conducted or that omitted data favor either group. They identify what the registry record does not allow a reader to evaluate.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: QBSafe: a Randomized Trial of a Novel Intervention to Improve Care for People Living With Type 2 Diabetes. — NCT05553912.
• Registry record last updated 2026-08-06; status last verified 2026-07.
• Results first posted 2026-08-06.
• 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

QBSafe (NCT05553912) is a completed randomized, parallel behavioral trial sponsored by Yale University. The registry lists 154 participants enrolled, no masking, and type 2 diabetes as the condition. Its intervention combined conversation cards for participants with materials intended to help clinicians respond to their concerns; usual care was the comparator.

The primary endpoint was illness intrusiveness on the IIRS at month 6, with the analysis described as adjusting for arm and initial score. Diabetes distress on the DDS-17 and treatment burden on the TBQ were listed as secondary outcomes at month 6. The record lists 9 outcome measures and 11 analyses.

Four p-values from mixed-model analyses are reported, but the registry does not map them to specific outcomes or provide effect estimates and confidence intervals. An analysis note identifies cluster randomization and cautions against simple individual-level recalculation. The registry does not report key details about clustering, missing data, analysis populations, or multiplicity, limiting assessment of the primary result’s direction and precision.