MyDiaText Text Messaging Intervention for Diabetes: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for MyDiaText Text Messaging Intervention for Diabetes (NCT02927639), a randomized na study. The primary outcome is Change in Self-Care Inventory (SCI) Score From Baseline to 6 Months.

2. Clinical Question

3. Trial Design

This was a randomized, single-masked, parallel-group trial, listed as completed. The phase is recorded as not applicable. The registry identifies Children's Hospital of Philadelphia as the sponsor. The study start date was 2018-07-31, and primary completion was 2019-10-01. Enrollment was 166.

Participants were assigned to MyDiaText plus usual standard care or to usual standard care alone. The intervention lasted six months and consisted of daily text messages sent to a personal mobile device, with a financial reward lottery intended to incentivize responding. The registry describes the messages as based on American Diabetes Association behavior goals. It does not identify who was masked, or provide further details about the usual standard care.

The primary outcome was change in Self-Care Inventory score from baseline to six months. The registry describes the analyzed outcome as a model-predicted mean score, using a multilevel mixed effects linear regression model. Participants were included in this outcome population if they completed at least one Self-Care Inventory survey. The posted outcome groups list 65 in the intervention arm and 70 in the control arm. The registry lists change in HbA1c from baseline to six months as a secondary outcome, but does not post its result in the record.

Although randomization and the parallel-group structure are recorded, the registry record does not report the randomization procedure, allocation concealment, masking details, or the trial’s original sample-size assumptions. Those details cannot be inferred from the posted analysis or from the sample-size planning illustration.

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

Analysis population for the primary outcome, as defined in the registry: Those analyzed were those who completed at least 1 self-care inventory survey. Analysis was preformed using multilevel mixed effects linear regression model. Outcomes are model-predicted.

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 Change in Self-Care Inventory (SCI) Score From Baseline to 6 Months, measured over 6 months.

The primary endpoint is the change in Self-Care Inventory score from baseline to six months. The inventory is a questionnaire measuring self-reported adherence to diabetes self-care behaviors. Its score is on a 1-5 scale, and higher scores are associated with more reported self-care behaviors. The endpoint therefore represents a continuous measure of reported behavior, rather than a count of clinical events.

The registry says that a response of N/A was excluded from calculation of the mean score for that question. It also states that the primary analysis used participants who completed at least one Self-Care Inventory survey, with outcomes estimated using a multilevel mixed effects linear regression model. The posted group values are model-predicted means at six months: 3.57 for the intervention group and 3.62 for the control group. The posted spread is a standard error for each mean.

A self-reported measure can capture behaviors that may not be directly represented by a laboratory measure, but it reflects participant reports and is subject to the way responses are collected and analyzed. The registry lists change in HbA1c as a secondary outcome, but does not report its result in this record. The original rationale for choosing the Self-Care Inventory as the primary endpoint is not reported in the registry record.

7. Results Posted in the Registry

• Intervention: mean 3.57 versus 3.62 on control; difference -0.05 (95% CI -0.24 to 0.14), standardized effect d = -0.09; z-test p = 0.614.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is continuous: it is a score on a scale, summarized as a mean. In general, a mean gives a measure of the center of a set of scores. The registry reports model-predicted means, not simply unadjusted observed averages. The reported group values are 3.57 for the intervention group and 3.62 for the control group, with a standard error of 0.07 in each group.

What does the mixed model do?

In general, a mixed effects model can account for data that have a grouped or repeated structure by representing different sources of variation in the model. The registry says the primary analysis used a multilevel mixed effects linear regression model, and calls the sponsor’s reported analysis a Mixed Models Analysis. However, the registry record does not report the model formula, the variables included, the covariance structure, or the exact way repeated surveys were handled. It is therefore not possible to reconstruct the sponsor’s model from the posted description alone.

The registry states that analyzed participants were those who completed at least one Self-Care Inventory survey, and that outcomes are model-predicted. These features matter: the reported values are tied to the analysis population and model rather than necessarily being raw group summaries. The record does not describe the number or timing of surveys each participant completed, or whether model predictions were adjusted for baseline scores or other variables.

What is the unadjusted re-calculation?

The separate re-calculation compares the posted intervention and control means using the posted summary statistics. It yields an intervention-minus-control mean difference of -0.05, a 95% confidence interval from -0.244 to 0.144, and an unadjusted p-value of 0.614. “Unadjusted” means this calculation does not reproduce a model-based adjustment or other features that may have been part of the sponsor’s analysis. It is a simple comparison based on posted summary data, not a replacement for the sponsor’s mixed-model result.

The registry’s reported sponsor p-value is 0.035. Since it differs from the re-calculated p-value, the sponsor’s own analysis takes precedence. The sponsor’s estimate and confidence interval are not reported in the registry record, so neither the size nor the precision of that modeled contrast can be independently checked from the posted summary. The reason for the discrepancy is also not reported; it would be inappropriate to attribute it to any particular modeling choice or data-handling decision.

Why does over-dispersion matter in general?

Over-dispersion is a general concern when a statistical model assumes less variability than the data actually contain. If variability is understated, standard errors can be too small and tests can appear more decisive than the data warrant. This concept is most often discussed for count outcomes and is not a result reported for this continuous score. The registry does not report an over-dispersion assessment, and there is no basis in the record to claim that it affected this analysis.

What does a p-value tell the reader?

In general, a p-value describes how unusual the observed data, or data more extreme, would be under a specified null model. It does not state the probability that the null hypothesis is true, nor does it quantify the effect’s importance. Interpretation also depends on the analysis that generated the value. Here, the sponsor’s p = 0.035 comes from the reported Mixed Models Analysis, while 0.614 comes from an unadjusted re-calculation. They should not be presented as though they came from the same analysis. The registry does not report the sponsor’s test statistic, degrees of freedom, or model-based effect estimate.

For continuous outcomes, the choice of scale and the meaning of a score help frame the comparison. A higher Self-Care Inventory score is associated with more self-care behaviors, but the difference between the group means alone does not establish why scores differ. The endpoint is self-reported, and the registry’s treatment of N/A responses means that a question could be omitted from the calculation of a participant’s mean score. The details of any further missing-data procedures are not reported in the registry record.

The posted results should therefore be read as a model-based sponsor analysis alongside a transparent but unadjusted recalculation. General statistical concepts can explain what each type of result is intended to convey, but they cannot fill in the sponsor model details or resolve the discrepancy without additional information.

10. Confidence Intervals and Precision

The unadjusted re-calculation estimates an intervention-minus-control mean difference of -0.05, with a 95% confidence interval from -0.244 to 0.144. In general, a confidence interval describes the uncertainty around an estimate under the assumptions of the method used to construct it. This interval includes zero, so the re-calculated summary-statistic comparison is compatible with no mean difference as well as with differences in either direction.

The interval belongs to the unadjusted re-calculation, not the sponsor’s Mixed Models Analysis. The registry reports the sponsor’s p-value as 0.035 but does not report a corresponding estimate or confidence interval. Consequently, the precision of the sponsor’s modeled contrast cannot be judged from the registry record. The sponsor result takes precedence over the discrepant re-calculation, but the absence of a reported interval limits assessment of its magnitude and uncertainty.

The group means are model-predicted and each has a reported standard error of 0.07. A standard error describes uncertainty in an estimated mean; it is not the spread of individual participant scores. The registry does not post confidence limits for the group means. Also, the unadjusted comparison does not capture any additional uncertainty or adjustment associated with the sponsor’s model.

Precision is distinct from importance. Even a narrow interval would not by itself determine whether an effect matters in practice, and a p-value does not supply that judgment. The registry record does not report a threshold for a meaningful change in the Self-Care Inventory score. The available interval therefore supports a description of the re-calculation’s uncertainty, but not a conclusion about the practical importance of a difference.

11. 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 planning figures are participants per arm, not total event counts. They illustrate sample sizes for detecting specified fractions of the observed effect; they do not describe the number enrolled in this trial or the trial’s original design target. The original design assumptions and sample-size justification are not reported in the registry record.

The planning calculation assumes a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. The effect to detect is set to the stated fraction of the observed effect. The rows describe planning for 100%, 75%, and 50% of the observed effect, with columns for 80% and 90% power. In plain language, 80% power is the probability of detecting an effect of the specified size under the planning assumptions if that effect is present; 90% power represents a higher target probability under those same assumptions. These are design probabilities, not guarantees for a future study.

Smaller effects need more participants to detect under otherwise comparable assumptions. That is why planning for a smaller fraction of the observed effect yields larger per-arm requirements. The observed effect used as the reference is especially uncertain because the re-calculated estimate is discrepant with the sponsor’s reported analysis, and the sponsor’s effect estimate is not reported. The planning figures should therefore be read as illustrations based on the stated observed-effect input, not as validated recommendations for a new study.

This is not observed power. It is a planning illustration based on assumptions stated for the calculation, and it is separate from the trial’s actual enrollment and analysis.

12. 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 = -0.050; pooled standard deviation σ = 0.576.
2. n = 2 × (1.96 + 0.842)² × 0.576² / -0.050² = 2,079.7, rounded up to 2,080 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 <- -0.0500000000000003                      # observed difference in means
sigma <- 0.5755089280939                     # 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(2927639)
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 = 2,080 (80% power), 2,785 (90% power)
 75% of observed effect: n per arm = 3,698 (80% power), 4,950 (90% power)
 50% of observed effect: n per arm = 8,319 (80% power), 11,137 (90% power)
power.t.test: 2080.7 
Simulated power at n = 2,080 per arm: 0.797 (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.

13. Analyses Reported by the Sponsor

The registry reports a sponsor Mixed Models Analysis for the intervention and control groups, with a superiority analysis type and p = 0.035. It does not report the sponsor’s numerical effect estimate or confidence interval. The record describes the primary outcome as model-predicted and says the analysis used a multilevel mixed effects linear regression model, but it does not give the model specification or enough detail to recreate that analysis.

A separate unadjusted re-calculation from the posted means and standard errors gives an intervention-minus-control difference of -0.05, a 95% confidence interval from -0.244 to 0.144, and an unadjusted p-value of 0.614. These results are discrepant from the sponsor’s reported p-value. The sponsor’s own analysis takes precedence, as it is the reported trial analysis; the re-calculation is a check based on summary statistics and should not be substituted for the sponsor’s model.

The available record does not identify why the results differ. It would be speculative to attribute the difference to adjustment, missing data, model structure, or any other particular factor. Because the sponsor estimate and interval are absent, readers cannot use the record to compare the estimated magnitude or precision of the two analyses. The discrepancy is best treated as an unresolved limitation of the posted statistical information.

14. Multiplicity, Missing Data and Analysis Populations

The registry lists two outcome measures: the primary change in Self-Care Inventory score from baseline to six months, and a secondary change in HbA1c from baseline to six months. It reports one analysis for the primary outcome. The HbA1c outcome result is not reported in the registry record. The record does not explain whether other outcomes were collected or analyzed beyond those listed.

Multiplicity arises in general when several statistical hypotheses are tested. Repeated testing can increase the chance of at least one apparently positive result, so plans may specify how tests are prioritized or adjusted. The registry record does not report a multiplicity strategy, a hierarchy of outcomes beyond identifying the primary and secondary outcomes, or whether any adjustment was used. No conclusion about multiplicity correction can be drawn from the posted p-value alone.

The primary outcome population included those who completed at least one Self-Care Inventory survey. The description also says that an N/A response was not included in the calculation of the mean score for that question. These statements identify some aspects of outcome availability and scoring, but do not provide a full missing-data plan. The registry record does not report how participants with no surveys were handled in the model, how missing survey items beyond N/A were treated, or whether sensitivity analyses were conducted.

The record does not say whether the sponsor used an intention-to-treat population, a per-protocol population, or another named analysis population. It reports only the at-least-one-survey criterion for the primary outcome analysis. The reasons participants may not have contributed outcome data, and the relationship between the randomized population and the analyzed population, are not reported in the registry record.

These omissions do not prove that the trial lacked a prespecified plan or additional analyses. They mean only that those details are not available in the registry record. Without them, readers cannot fully evaluate how missing outcomes, multiple tests, or analysis-population choices affected the reported result.

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

The registry describes a randomized comparison of daily MyDiaText messages plus usual standard care with usual standard care alone. Its primary outcome was change in a self-reported self-care score from baseline to six months. The posted six-month model-predicted means were 3.57 for the intervention group and 3.62 for the control group. The sponsor reports a Mixed Models Analysis p-value of 0.035, while an unadjusted re-calculation from the posted summary statistics gives a different result. The sponsor’s analysis takes precedence, but the record does not provide its effect estimate or confidence interval.

The result does not, on its own, establish the size or practical importance of a difference. The posted p-value is not a measure of effect magnitude, and the discrepancy means the unadjusted calculation cannot be used as if it were the sponsor’s model. The record does not report a threshold for a meaningful change on the Self-Care Inventory. It also does not post the result for the listed secondary HbA1c outcome.

The endpoint concerns self-reported adherence to diabetes self-care behaviors. It is not itself a direct measurement of every behavior or of a clinical outcome. The registry’s description indicates that higher scores are associated with more self-care behaviors, but the posted information cannot establish how participants interpreted or completed the questionnaire beyond the stated scoring rule for N/A responses.

The trial record does not provide enough information to assess the sponsor model in detail, determine the reason for the discrepancy, or evaluate all missing-data and multiplicity decisions. Nor does the record establish how well the findings generalize to populations or settings beyond the study’s described participants and care context. These are limits of what can be concluded from the registry information, not evidence that any unreported procedure was or was not performed.

16. Why This Trial Matters Statistically

This trial illustrates why an effect estimate, its uncertainty, and the analysis that produced it need to be considered together. The registry’s sponsor result is a mixed-model p-value, whereas the separate re-calculation is an unadjusted comparison of posted summary statistics. Those are not interchangeable analyses. Their discrepancy makes it particularly important to identify which result came from the trial’s own analysis and which is a secondary calculation.

It also highlights the value of reporting an estimated effect and confidence interval alongside a test result. The sponsor’s p-value is posted, but its corresponding estimated difference and interval are not. Without those values, the magnitude and precision of the sponsor’s modeled contrast cannot be assessed from the record. A p-value alone cannot describe the range of effects compatible with the analysis.

The endpoint adds another statistical consideration: the primary outcome is a continuous, self-reported score, and the posted values are model-predicted. Interpretation depends on the analysis population, scoring rules, and model details. The record identifies some of these elements but not enough to reconstruct the sponsor’s model or fully evaluate missing outcomes. The secondary HbA1c result is also not reported in the record.

Finally, sample-size planning illustrates the dependence of required enrollment on the effect a study aims to detect. The planning figures are explicitly based on fractions of the observed effect and assumptions for a planning calculation; they are not the trial’s original design rationale. This distinction prevents a retrospective illustration from being mistaken for evidence about how the trial was planned or how much power it actually had.

17. 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 principal limitation in interpreting the primary result is the discrepancy between the sponsor’s reported p = 0.035 and the unadjusted re-calculation’s p-value of 0.614. The sponsor’s own analysis takes precedence, but the registry does not provide its estimated contrast or confidence interval. It also does not provide enough model detail to explain or reproduce the reported result.

The primary outcome is self-reported, and the registry notes that N/A responses were excluded from calculation of the mean score for that question. Participants needed to complete at least one survey to be included in the stated analysis population. The record does not describe the amount or pattern of missing survey data, how participants with no surveys were treated, or whether sensitivity analyses addressed missingness.

The outcome group counts are 65 for the intervention and 70 for control, while registry enrollment is 166. These figures describe different aspects of the record; the registry does not provide a full accounting that reconciles enrollment, allocation, follow-up, and analysis. The record also does not give the sponsor’s model specification, masking details, allocation procedure, or original sample-size assumptions.

The registry lists change in HbA1c as a secondary outcome but does not report its result in the record. It does not report a multiplicity plan or a threshold for a meaningful change in the primary score. Adverse-event entries list zero serious events and zero other events for the posted at-risk groups, but the registry record does not provide further context for interpreting those entries. These reporting limits constrain assessment; they do not establish what may have been documented elsewhere.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: MyDiaText Text Messaging Intervention for Diabetes — NCT02927639.
• Registry record last updated 2021-06-09; status last verified 2021-05.
• Results first posted 2020-11-20.
• 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.

21. Record Summary

NCT02927639 is a completed, randomized, single-masked, parallel-group trial of MyDiaText for children aged 12 to 18 years with type 1 diabetes. Enrollment was 166. The intervention combined daily text messages and a financial reward lottery with usual standard care for six months; the control group received usual standard care.

The primary outcome was change in Self-Care Inventory score from baseline to six months. The posted model-predicted means were 3.57 in the intervention group and 3.62 in control. The sponsor reports a Mixed Models Analysis p-value of 0.035. An unadjusted re-calculation gives a mean difference of -0.05, with a 95% confidence interval from -0.244 to 0.144 and p = 0.614. These results are discrepant, and the sponsor’s own analysis takes precedence. The sponsor’s effect estimate and confidence interval, the result for the secondary HbA1c outcome, and several analysis details are not reported in the registry record.