Optimization and Multi-Site Feasibility of Yoga for Chronic Pain in People in YOGAMAT-II - Phase II: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Optimization and Multi-Site Feasibility of Yoga for Chronic Pain in People in YOGAMAT-II - Phase II (NCT04641221), a randomized na study. The primary outcome is Yoga Dosage.

2. Clinical Question

3. Trial Design

The registry describes a randomized, double-masked, factorial Phase II study. The title identifies it as the YOGAMAT-II optimization and multi-site feasibility study, while the registry phase field is listed as NA. Enrollment is listed as 192, and the study status is completed. The registry gives a start date of 2022-02-20 and a primary completion date of 2024-11-15.

The stated optimization aim was to evaluate four components alongside core yoga classes: personal practice videos, one-on-one instructor sessions, prompting text messages, and financial incentives for class attendance. The arm descriptions represent combinations of receiving or not receiving those components. There are 16 described combinations, and the primary outcome summaries list 12 participants per combination. The registry does not give further detail about the randomization procedure, allocation concealment, or what double masking meant in this behavioral intervention.

The study summary describes a multiphase optimization strategy and identifies three aims: estimate the impact of the components on yoga dosage, use Phase II results to choose an efficient combination with standard yoga classes, and examine hypothesized mechanisms. The posted primary outcome is total yoga minutes per week at 3 months. The registry record does not report the protocol for selecting the preferred combination, or a final selection among combinations.

The record lists weekly yoga classes as the behavioral intervention. It does not provide a separate, single control arm; the no-additional-component combination still receives the core yoga intervention. Thus the arm summaries describe combinations within a component-optimization experiment, not a comparison of yoga against no yoga.

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 16 groups are shown.

First 3 months

All 12 months

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.

Only the first 6 of 17 groups are shown.

6. Primary Endpoint

The registry lists the primary outcome as Yoga Dosage, measured over 3 months.

The primary outcome is yoga dosage, measured as total minutes per week of yoga practice. The registry definition includes practice during class and outside class, with weekly assessment by interview over 3 months. The arm summaries report means and 90% confidence intervals in minutes per week.

This measure matches the stated optimization question: whether the added support components are associated with the amount of yoga practice participants receive. It captures practice beyond attendance alone, because the definition includes both in-class and out-of-class minutes. However, the registry record does not specify how interview responses were collected, how missing weekly reports were handled, or how a participant’s repeated weekly measurements were combined to produce the arm summaries.

A mean describes the group’s average reported practice, while a confidence interval conveys uncertainty around that estimate under the method used to construct it. It is not a range expected to contain most individual participants’ values. The intervals are 90% confidence intervals; the record does not give the procedure used to calculate these intervals.

The registry also lists model-based results expressed as incidence rate ratios. Those results are a different scale from the arm-level minutes-per-week means. The record does not explain in detail how those model estimates map onto the endpoint definition, so the two types of estimates should not be treated as interchangeable.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The registry identifies a factorial randomized design with four intervention components, each present or absent across the described combinations. In general, a factorial design can evaluate components within one experiment rather than treating every combination as an unrelated intervention. It can also allow researchers to examine whether the association of one component depends on the presence of another. The registry summary says the aim was to evaluate each component’s impact, but the record does not specify the planned statistical contrasts or report component-specific effect estimates in a way that can be unambiguously identified.

What do the arm estimates describe?

The arm-level values are mean total yoga minutes per week, accompanied by 90% confidence intervals. A mean is a measure of the group’s average, not an account of the distribution of individual practice times. In general, a confidence interval represents uncertainty in an estimate under the assumptions of its method. It is not a range containing 90% of participants, and overlap between intervals is not a substitute for a formal comparison. The record does not state how these intervals were calculated.

The summaries list 12 participants for each combination. The registry enrollment is 192, but it does not explain whether the arm-summary counts reflect all enrolled participants, the analysis population, or a subset with outcome data. It also does not explain the handling of repeated weekly assessments or missing interview reports. Those details matter because the endpoint was assessed weekly over 3 months, while the displayed arm value is a single mean for each combination.

What is a zero-inflated negative-binomial model?

The reported analyses name a zero-inflated negative-binomial model, abbreviated ZINB in the registry record. In general, negative-binomial models can be useful for nonnegative count-like outcomes when observations vary more than a simple Poisson model allows. This extra variability is called over-dispersion. A zero-inflated model adds a way to represent a group of observations that are more likely to be zero than the usual count component alone would predict. These are general features of the model family, not confirmation that the endpoint was analyzed in a particular way beyond the model name reported by the registry.

The primary endpoint is expressed in minutes, while the model results are described as incidence rate ratios. The registry does not provide enough model specification to explain how the minute-based outcome was represented, whether and how zero values were handled, or what exposure or observation time was used. It also does not report the model’s covariates, link function, repeated-measures structure, or method for estimating uncertainty. Those omissions limit interpretation of the fitted ratios.

How should the three reported ratios be read?

The registry reports three ratios with 90% confidence intervals and p-values: 1.10 (1.02 to 1.18; p=0.028), 1.09 (1.02 to 1.17; p=0.039), and 0.86 (0.80 to 0.93; p=0.001). In general, a ratio of 1 represents equal rates under the numerator and denominator definitions; a value above or below 1 indicates a relative difference in the direction defined by the contrast. Here, the registry does not give the numerator, denominator, or exact contrast for each analysis. It also alternates between the labels “incidence rate ratio” and “incident rate ratio.” The ratios therefore cannot be assigned to a particular component or interpreted as a specific increase or decrease in yoga practice from the record alone.

A p-value measures how incompatible the observed result is with a stated null model, assuming the analysis and its assumptions are appropriate. It is not the probability that the null hypothesis is true, and it does not measure the size or importance of a difference. The posted p-values correspond to three separate analyses, but the registry record does not report whether these were planned as separate tests, whether multiplicity was addressed, or how the model tests were selected.

What does the factorial structure add?

Because each combination includes different presences and absences of videos, individual sessions, texts, and incentives, the design could in principle support estimates of component effects and interactions. An interaction means that a component’s association may differ depending on another component’s presence. In general, this is one reason factorial designs can be useful for optimization. The posted analysis summaries, however, group all 16 combinations and do not identify main effects, interactions, or specific contrasts. The record does not report an interaction model or results from a component-selection procedure.

The registry’s stated sample-size illustration is explicitly not the trial’s own design justification. It describes a planning approach based on an observed effect, a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. Such planning illustrations concern hypothetical future sample sizes; they are not observed power and do not establish the power of this completed experiment. The trial’s own sample-size justification is not reported in the registry record.

10. Analyses Reported by the Sponsor

The registry reports three sponsor-associated model analyses, all using the 16 listed groups. The results are 1.10 (90% CI 1.02 to 1.18; p=0.028), 1.09 (90% CI 1.02 to 1.17; p=0.039), and 0.86 (90% CI 0.80 to 0.93; p=0.001). They use zero-inflated negative-binomial methods, although the registry varies the wording of the model and ratio labels. The exact contrast for each estimate is not reported, which prevents a direct explanation of what each ratio compares.

The arm summaries are means in minutes per week, whereas the sponsor-reported results are ratios from fitted models. These are not equivalent estimands, and an arm mean should not be used to reconstruct or replace a model ratio without a defined method and contrast. A corresponding recalculated estimate for these primary comparisons is not reported in the registry record. The record does not flag a discrepancy between sponsor analysis and recalculation. In the absence of a documented discrepancy, the sponsor’s reported model results remain the available model-based findings, but their contrast definitions are needed for substantive interpretation.

11. Multiplicity, Missing Data and Analysis Populations

The registry reports three analyses of the primary outcome, each involving the same set of 16 groups. It does not report whether the analyses represent distinct prespecified contrasts, alternative model specifications, or another analytic sequence. A multiplicity adjustment or a rationale for treating the analyses as a family is not reported in the registry record. Consequently, the individual p-values should not be read as though the record establishes an overall error-rate procedure for all analyses.

The exact analysis populations are not reported in the registry record. Although enrollment is 192 and the arm summaries list 12 participants each, the record does not say whether those counts represent randomized participants, participants with endpoint data, or another analysis set. It does not report participant flow, withdrawals, exclusions, or the number contributing weekly interview observations.

Missing-data handling is also not reported. The endpoint was assessed weekly, but the registry record does not explain how absent interviews or incomplete practice reports were treated, whether any imputation was used, or whether the models included all available observations. Without these details, it is not possible to assess how missing observations may affect the estimates.

The record does not report interim analyses, stopping rules, or a prespecified statistical analysis plan. Nor does it report the trial’s own sample-size justification or the intended power for the factorial effects and interactions. These omissions do not prove that such methods were absent from study conduct; they mean that they cannot be evaluated from the registry record.

12. Safety Results

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

Deaths

Other (non-serious) adverse events

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

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: surgical and medical procedures - other, specify (16.7%), anemia (8.3%), anxiety (8.3%), arthralgia (8.3%), back pain (8.3%).

The safety summaries report serious adverse events by intervention combination. Each group lists 12 participants at risk, with serious events affecting between 1 and 4 participants in the individual groups. The registry reports zero deaths in every group. These are counts within the registry’s safety reporting, not evidence that the intervention components caused or prevented any event.

For other, non-serious adverse events, the group counts range from 5 to 10 affected participants, with 12 at risk in each group. The record lists “Surgical and medical procedures - Other, specify” at 16.7% among the top serious-event terms; anemia, anxiety, arthralgia, back pain, and fall are each listed at 8.3%. It does not provide enough detail to connect these terms to specific components or determine whether events were considered related to the intervention.

Event counts in groups of this size are descriptive and can be imprecise. The registry record does not give event-level narratives, timing, duration, severity grading beyond serious versus other, or comparative statistical analyses for safety. The figures should therefore be read as reported counts, not as a basis for comparing the safety of the combinations.

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

The results describe weekly yoga practice in a randomized factorial study where all groups received core yoga classes and differed in whether they received videos, one-on-one instructor sessions, prompting texts, and financial incentives. The arm means show that estimated practice minutes varied across combinations. The confidence intervals communicate uncertainty around those estimates, and the reported ZINB analyses provide model-based ratios with accompanying intervals and p-values.

These findings do not establish that any particular component increased or decreased practice. The registry does not identify the contrast represented by each ratio, nor does it present component-specific effects or interactions in a clearly defined form. The descriptive arm means cannot substitute for those contrasts. The registry also does not identify which combination was ultimately selected for the stated optimization aim.

The results do not show whether yoga changed chronic pain, opioid use, or another clinical outcome. The posted primary endpoint is yoga dosage, measured as minutes of practice per week. The registry lists chronic pain and opioid use as conditions, but the primary outcome reported is participation in yoga practice rather than a direct measure of either condition.

The results also do not determine whether the intervention is suitable for any individual, and they should not be used as medical advice. The record contains safety counts, but does not provide causal attribution or enough comparative detail to rank the combinations by safety. More complete analysis descriptions, including contrasts, missing-data methods, and the population analyzed, would be needed to judge the model results more fully.

In general, statistical significance and practical importance are different questions. The reported p-values alone do not establish the size or usefulness of an effect. The arm estimates are in minutes per week, but the registry does not state a threshold for a meaningful change in yoga dosage. Thus the posted results characterize practice patterns and model outputs, rather than settling the clinical or implementation value of any combination.

14. Why This Trial Matters Statistically

This study is statistically notable because it treats support for yoga practice as a set of components that can be varied in combination. A factorial design can be useful when the research question concerns multiple elements rather than a single package. In principle, this can help distinguish the contribution of videos, individual instructor sessions, texts, and incentives while also considering whether components work differently together.

The study’s optimization goal shifts attention from a simple treatment-versus-control question to choosing a combination that may efficiently support practice. That makes the definition of the target outcome and the contrast structure especially important. Here, yoga dosage is measured as weekly minutes, and the study summary explicitly describes maximizing dosage. But the registry does not report the component-specific estimates, interaction results, or selection rule needed to see how the optimization decision was reached.

The record also illustrates why statistical outputs need clear estimands. Arm-level means in minutes and model-based incidence rate ratios answer different questions. When the contrasts behind ratio estimates are not stated, even confidence intervals and p-values that appear precise can be difficult to connect to a practical decision. Clear descriptions of which groups are compared, how repeated observations are handled, and how multiplicity is addressed are essential to interpreting a factorial trial.

Finally, the modest per-combination summaries make uncertainty relevant to any apparent ordering of arm means. The confidence intervals convey that estimates are not exact, while the absence of a reported selection method limits claims about a best combination. The study provides a framework for component optimization, but the information in the registry record does not fully document how the statistical evidence informed that choice.

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 record restrict interpretation. First, the three reported model ratios lack named contrasts and reference categories. Their values include estimates above and below 1, but the record does not say which component or comparison each represents. Second, the model descriptions are sparse: covariates, repeated-measures handling, zero-inflation specification, exposure definition, and interval-estimation method are not reported.

The 16 arm summaries each list 12 participants, but the record does not clarify whether these counts include every randomized participant or only participants with endpoint data. Participant flow and missing weekly interview information are not reported. The reported endpoint is assessed weekly, but the process used to create the single arm mean is not specified.

The record gives no clear main-effect or interaction estimates for the four factorial components, and it does not describe the rule for selecting an efficient combination. It also does not report a multiplicity approach, a statistical analysis plan, interim analyses, or the trial’s own sample-size justification. The phase field is listed as NA even though the title identifies the study as Phase II.

Finally, the outcome measures yoga practice dosage, not clinical change in chronic pain or opioid use. Safety counts are reported descriptively, without event-level details or analyses comparing combinations. These limitations concern what can be concluded from the registry record; they do not establish what was or was not addressed in other study documents.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Optimization and Multi-Site Feasibility of Yoga for Chronic Pain in People in YOGAMAT-II - Phase II — NCT04641221.
• Registry record last updated 2025-11-20; status last verified 2025-10.
• Results first posted 2025-11-20.
• 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.

19. Record Summary

The completed randomized, double-masked factorial study evaluated weekly yoga classes alongside combinations of personal practice videos, one-on-one instructor sessions, prompting texts, and financial incentives. The registry lists enrollment of 192 and 16 combinations, with 12 participants represented in each primary-outcome summary.

The primary outcome was total yoga minutes per week, including in-class and out-of-class practice, assessed weekly by interview over 3 months. Arm means ranged from 84.5 to 185.6 minutes per week. Three zero-inflated negative-binomial analyses are reported, but their exact contrasts are not identified. Serious adverse events and other adverse events are summarized by group; no deaths are reported in any group.

The record does not report enough analytic detail to map the model ratios to specific components or determine how a preferred combination was chosen. The posted findings therefore describe practice estimates and model outputs, with important uncertainty about their contrasts, analysis populations, missing data, and multiplicity.