This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record.
1. Trial at a Glance
MATIC was a completed, randomized, double-blind phase 3 trial in children with house dust mite allergic rhinitis. The trial compared a sublingual allergy immunotherapy tablet containing 12 SQ-HDM with placebo and evaluated daily rhinitis symptoms and medication use during an 8-week primary efficacy assessment period.
| Feature | MATIC |
|---|---|
| Trial name | MATIC |
| ClinicalTrials.gov identifier | NCT04145219 |
| Phase | Phase 3 |
| Status | Completed |
| Therapeutic area | Respiratory |
| Conditions | Allergic rhinitis due to Dermatophagoides farinae; allergic rhinitis due to Dermatophagoides pteronyssinus; allergic rhinitis due to house dust mite |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Double |
| Primary purpose | Treatment |
| Enrollment | 1460 |
| Lead sponsor | ALK-Abelló A/S |
| Sponsor type | Industry |
| Start | 2019-10-12 |
| Primary completion | 2023-04-01 |
2. Clinical Question
The central question was whether treatment with a 12 SQ-HDM sublingual allergy immunotherapy tablet produced a different average daily total combined rhinitis symptom and medication score during the primary efficacy assessment period compared with placebo in children with house dust mite allergic rhinitis.
Population
Children enrolled in a phase 3 trial for allergic rhinitis due to house dust mite, including allergic rhinitis due to Dermatophagoides farinae and Dermatophagoides pteronyssinus.
Intervention
Sublingual allergy immunotherapy tablet containing 12 SQ-HDM.
Comparator
Placebo sublingual allergy immunotherapy tablet.
Primary question
Does the 12 SQ-HDM tablet produce a different average daily TCRS during the primary efficacy assessment period under a superiority framework?
3. Trial Design
HDM SLIT-tablet (12 SQ-HDM)
- Sublingual allergy immunotherapy tablet
- 12 SQ-HDM dose
- Evaluated against placebo under double masking
- Primary efficacy assessment began 44–49 weeks after initiation of investigational medicinal product
Placebo SLIT-tablet
- Placebo sublingual allergy immunotherapy tablet
- Parallel randomized comparison
- Double-masked design
- Primary efficacy assessment used the same 8-week period
The registry describes the study as randomized, parallel, double-masked, and treatment-focused. These design features are important statistically because randomization establishes the basis for a between-group causal comparison, while masking can reduce the influence of treatment knowledge on participant reporting and study conduct.
4. Timing of the Primary Efficacy Assessment
The primary endpoint was not a single visit measurement. It was the average of reported daily TCRS values over the 8-week primary efficacy assessment period.
This timing has an important statistical consequence. The analysis summarizes repeated daily observations over a defined assessment window rather than treating one isolated measurement as the endpoint. The registry therefore used mixed-effects methodology to account for the longitudinal structure of the data.
5. Primary Endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Average Daily Total Combined Rhinitis Symptom and Medication Score (TCRS) | The average daily total combined rhinitis symptoms and medication score. The score is on a scale of 0–24; higher scores indicate more severe symptoms and/or more use of rhinitis medication. The endpoint is the average of all reported daily values during the primary efficacy assessment period. | 8 weeks, beginning 44–49 weeks after initiation of IMP |
For example, the registry explains that if a participant reported 56 daily TCRS values, the primary endpoint would be calculated as the average of those daily values. Thus, the endpoint integrates symptom and medication information across the efficacy period rather than relying on a single day's score.
6. Statistical Methodology
Linear mixed-effects model for continuous outcomes
The primary TCRS analysis used a linear mixed-effect model with square-root transformation. The square root of the endpoint was modeled as the response variable. Treatment and cohort were included as fixed factors, the square root of baseline TCRS was included as a covariate, country/region within cohort was included as a random effect, and different residual errors were specified for each treatment.
The registry describes this as a linear mixed-effect model. The model adjusts for baseline TCRS and accounts for clustering through the country/region-within-cohort random effect.
Why transform the endpoint?
A square-root transformation changes the scale on which the model operates. Such transformations can be useful when the variability of a score is related to its magnitude or when the distribution is asymmetric. The important interpretive point is that the registry reports the treatment contrast on the final-value scale as a mean difference, while the model itself uses the square-root-transformed outcome.
Covariate adjustment
Baseline TCRS was included as a covariate. Adjusting for a baseline measurement can improve statistical efficiency when baseline severity is related to the subsequent outcome. The registry also included cohort and a country/region-within-cohort random effect, making the model more structured than a simple comparison of two unadjusted arithmetic means.
Missing data
The registry states that no missing-data approach was applied for the continuous mixed-model analyses. The analysis population was the full analysis set with observations in the relevant efficacy period. This distinction matters: a model that uses available observations is not equivalent to an analysis that explicitly replaces missing observations through multiple imputation.
Binary endpoints
Binary outcomes such as SABA-free days, rhinitis mild days, and rhinitis exacerbation days were analyzed with generalized linear mixed models using a logit link. These models incorporated treatment and cohort as fixed effects and subject-level and country/region-within-cohort random effects, with baseline covariates specified for the relevant endpoint.
7. Primary Result
Average Daily Total Combined Rhinitis Symptom and Medication Score
Mean difference in final values
95% CI: 0.5–1.4 · P < 0.0001
Two-sided confidence interval · Superiority hypothesis
| Primary endpoint | Active vs placebo | Effect measure | 95% CI | P-value |
|---|---|---|---|---|
| Average Daily TCRS during the primary efficacy assessment period | HDM SLIT-tablet (12 SQ-HDM) vs Placebo SLIT-tablet | Mean difference: 1.0 | 0.5–1.4 | <0.0001 |
The registry records the contrast as placebo versus 12 SQ-HDM. Because higher TCRS values indicate greater symptom severity and/or medication use, the positive mean difference is consistent with lower TCRS values for the 12 SQ-HDM group relative to placebo on the reported contrast scale.
What the estimate means: the registry reports a mean difference of 1.0 on the final-value scale, with placebo contrasted against 12 SQ-HDM. Because the endpoint is a measure of rhinitis symptoms and medication use and higher scores are worse, the direction of the contrast corresponds to lower average TCRS under 12 SQ-HDM.
What it does not mean: the estimate does not mean that every child experienced a 1.0-point reduction. It is a population-level treatment contrast produced by the specified statistical model.
What the confidence interval says: the 95% CI of 0.5–1.4 quantifies uncertainty around the estimated mean difference under the model and analysis framework. It is not a range containing the individual treatment effect for 95% of participants.
Why the P-value is different: P < 0.0001 addresses evidence against the null hypothesis under the specified statistical test. It does not measure the magnitude or clinical importance of the treatment effect. The estimate and confidence interval are needed to understand magnitude and precision.
Model caution: the primary result comes from a mixed-effects model with a square-root transformation, baseline adjustment, cohort effects, a country/region-within-cohort random effect, and treatment-specific residual errors. It should therefore be interpreted as a model-based comparison rather than an unadjusted difference between simple group means.
8. Secondary Efficacy Results: Rhinitis Symptom and Medication Components
The registry posted several secondary outcomes that dissect the combined rhinitis endpoint into symptom and medication components and extend the assessment to rhinoconjunctivitis outcomes.
| Secondary endpoint | Effect | 95% CI | P-value |
|---|---|---|---|
| Average Rhinitis Daily Symptom Score (DSS) | Mean difference: 0.4 | 0.2–0.6 | <0.0001 |
| Average Rhinitis Daily Medication Score (DMS) | Mean difference: 0.5 | 0.2–0.8 | 0.0016 |
| Average Daily Total Combined Rhinoconjunctivitis Symptom and Medication Score (TCS) | Mean difference: 1.1 | 0.6–1.7 | <0.0001 |
| Average Rhinoconjunctivitis Daily Symptom Score (DSS) | Mean difference: 0.5 | 0.3–0.7 | <0.0001 |
| Average Rhinoconjunctivitis Daily Medication Score (DMS) | Mean difference: 0.6 | 0.2–1.0 | 0.0018 |
All five endpoints were evaluated during the same 8-week primary efficacy assessment period. The registry used linear mixed-effect models for these continuous outcomes, generally with square-root transformation, baseline adjustment, cohort effects, country/region-within-cohort random effects, and treatment-specific residual errors.
The secondary continuous outcomes show the same statistical structure as the primary endpoint: the treatment contrast is expressed as a mean difference, with positive placebo-versus-12-SQ-HDM contrasts corresponding to lower scores for the active treatment when higher scores represent worse symptoms or greater medication use.
The confidence intervals provide a useful indication of precision. For example, the rhinitis DSS estimate is 0.4 with a 95% CI of 0.2–0.6, whereas the rhinoconjunctivitis DMS estimate is 0.6 with a 95% CI of 0.2–1.0. These intervals describe uncertainty around the corresponding model-based contrasts; they should not be treated as interchangeable evidence about the clinical importance of the endpoints.
The multiple secondary endpoints also raise a multiplicity issue. The registry identifies superiority hypotheses and reports individual P-values, but the ClinicalTrials.gov record does not specify an alpha-allocation hierarchy across these secondary outcomes. Therefore, the individual P-values should be interpreted as endpoint-specific statistical results rather than automatically as a single multiplicity-adjusted familywise conclusion.
9. Quality of Life and Asthma-Related Results
Paediatric Rhinoconjunctivitis Quality of Life Questionnaire
Overall PRQLQ score at end of trial
95% CI: 0.1–0.2 · P < 0.0001
Week leading up to visit 7, at the end of the primary efficacy assessment period, after approximately 52–57 weeks of treatment
The overall Paediatric Rhinoconjunctivitis Quality of Life Questionnaire score was analyzed using a linear mixed-effect model without the square-root transformation specified for several symptom and medication scores. The model included treatment and cohort as fixed factors, baseline overall PRQLQ score as a covariate, country/region within cohort as a random effect, and different residual errors for each treatment.
The reported mean difference is 0.2 with a 95% CI of 0.1–0.2. The registry reports this as a final-value mean difference rather than an odds ratio or time-to-event measure. The statistical evidence is strong under the reported two-sided test, but the P-value itself does not establish the magnitude of quality-of-life improvement; the effect estimate and its confidence interval provide the magnitude and precision.
Asthma daily symptom score
| Endpoint | Effect | 95% CI | P-value |
|---|---|---|---|
| Average Asthma Daily Symptom Score (DSS) | Mean difference: 0.1 | 0.0–0.2 | 0.0259 |
| Weekly number of puffs of as-needed SABA use | Mean difference: 0.4 | -0.1–1.0 | 0.1256 |
These analyses were restricted to participants with asthma in the full analysis set with observations during the primary efficacy period. The asthma DSS used a square-root-transformed linear mixed-effect model with baseline average asthma DSS as a covariate. Weekly SABA use used a linear mixed-effect model with the baseline value as a covariate.
The asthma DSS result has a P-value of 0.0259 and a 95% CI of 0.0–0.2. The registry's reported numerical precision is important here: the lower confidence limit is given as 0.0, so the result should not be described more precisely than the posted interval supports.
For weekly SABA puffs, the mean difference is 0.4 with a 95% CI of -0.1–1.0 and P = 0.1256. The interval includes zero, so the posted estimate is compatible with a range of differences that includes no mean difference on the reported scale. This illustrates why an effect estimate should be considered together with its confidence interval rather than relying only on whether a P-value crosses a conventional threshold.
10. Binary Secondary Outcomes
The registry used generalized linear mixed models with a logit link for binary daily outcomes. These models estimate treatment differences on the odds scale while incorporating the repeated daily structure through random effects.
| Binary endpoint | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| SABA Free Days | Odds ratio | 1.8 | 1.0–3.3 | 0.0527 |
| Rhinitis Mild Days | Odds ratio | 1.8 | 1.3–2.5 | 0.0008 |
| Rhinitis Exacerbation Days | Odds ratio | 0.6 | 0.4–0.7 | <0.0001 |
SABA-free days
The SABA-free-day analysis modeled whether each day was a SABA-free day, with treatment and cohort as fixed effects, baseline average asthma DSS as a covariate, and combined country/region within cohort and subject as random effects. The registry defines the odds ratio as the odds of a SABA-free day under 12 SQ-HDM divided by the odds under placebo.
An odds ratio of 1.8 means that the modeled odds of a SABA-free day were 1.8 times as high under 12 SQ-HDM as under placebo. An odds ratio is not the same as a risk ratio or a difference in percentages. The 95% CI of 1.0–3.3 indicates substantial uncertainty, and P = 0.0527 is close to, but above, 0.05. The result should therefore be reported using its exact estimate, confidence interval, and P-value rather than converted into a categorical claim based on a single threshold.
Rhinitis mild days
The odds ratio for a rhinitis mild day was 1.8 with a 95% CI of 1.3–2.5 and P = 0.0008. The registry defines the modeled probability as the probability of a day with no or mild rhinitis symptoms and defines the odds ratio as the odds under 12 SQ-HDM divided by the odds under placebo.
The point estimate of 1.8 indicates higher modeled odds of a rhinitis mild day with 12 SQ-HDM. Because the confidence interval remains above 1.0, the direction of the association is consistently above the null value across the posted 95% interval. The odds ratio still should not be interpreted as an 80% increase in probability; odds and probability are different quantities.
Rhinitis exacerbation days
The odds ratio for a rhinitis exacerbation day was 0.6 with a 95% CI of 0.4–0.7 and P < 0.0001. The registry defines the odds ratio as the odds under 12 SQ-HDM divided by the odds under placebo.
An odds ratio of 0.6 indicates lower modeled odds of a rhinitis exacerbation day under 12 SQ-HDM relative to placebo. Expressed purely as a relative odds comparison, 0.6 is 40% below 1.0. That statement applies to odds, not directly to the probability of an exacerbation day. The confidence interval of 0.4–0.7 provides the uncertainty around the odds ratio, while the P-value addresses evidence against the null odds ratio of 1.0.
11. Additional Continuous Rhinitis Outcomes
| Endpoint | Effect | 95% CI | P-value |
|---|---|---|---|
| Average Rhinitis Combined Symptom and Medication Score (CSMS) | Mean difference: 0.2 | 0.1–0.3 | <0.0001 |
| Average Rhinoconjunctivitis Combined Symptom and Medication Score (CSMS) | Mean difference: 0.2 | 0.1–0.3 | <0.0001 |
Both endpoints were analyzed with linear mixed-effect models using square-root transformation, treatment and cohort as fixed factors, the square root of baseline value as a covariate, country/region within cohort as a random effect, and treatment-specific residual errors.
The CSMS results reinforce an important feature of this trial's statistical architecture: multiple measures were used to characterize symptoms and medication use from somewhat different perspectives. Agreement across related endpoints can be informative descriptively, but each additional hypothesis also contributes to the broader multiplicity context.
12. Complete Posted Efficacy Results
| Endpoint | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|
| Primary: Average Daily TCRS | Linear mixed-effect model | Mean difference 1.0 | 0.5–1.4 | <0.0001 |
| Average Rhinitis DSS | Linear mixed-effect model | Mean difference 0.4 | 0.2–0.6 | <0.0001 |
| Average Rhinitis DMS | Linear mixed-effect model | Mean difference 0.5 | 0.2–0.8 | 0.0016 |
| Average Daily TCS | Linear mixed-effect model | Mean difference 1.1 | 0.6–1.7 | <0.0001 |
| Average Rhinoconjunctivitis DSS | Linear mixed-effect model | Mean difference 0.5 | 0.3–0.7 | <0.0001 |
| Average Rhinoconjunctivitis DMS | Linear mixed-effect model | Mean difference 0.6 | 0.2–1.0 | 0.0018 |
| Overall PRQLQ score | Linear mixed-effect model | Mean difference 0.2 | 0.1–0.2 | <0.0001 |
| Average Asthma DSS | Linear mixed-effect model | Mean difference 0.1 | 0.0–0.2 | 0.0259 |
| SABA Free Days | Generalised linear mixed model | OR 1.8 | 1.0–3.3 | 0.0527 |
| Weekly number of puffs of as-needed SABA | Linear mixed-effect model | Mean difference 0.4 | -0.1–1.0 | 0.1256 |
| Rhinitis Mild Days | Generalised linear mixed model | OR 1.8 | 1.3–2.5 | 0.0008 |
| Rhinitis Exacerbation Days | Generalised linear mixed model | OR 0.6 | 0.4–0.7 | <0.0001 |
| Average Rhinitis CSMS | Linear mixed-effect model | Mean difference 0.2 | 0.1–0.3 | <0.0001 |
| Average Rhinoconjunctivitis CSMS | Linear mixed-effect model | Mean difference 0.2 | 0.1–0.3 | <0.0001 |
This table illustrates the breadth of the registry's posted statistical analyses. The primary endpoint was continuous, while the secondary outcomes included continuous, binary, symptom, medication, quality-of-life, and asthma-related measures. The effect measure therefore appropriately changes with the statistical model and endpoint type.
13. Statistical Methods Explained
Why was a mixed-effects model used for the primary endpoint?
The primary efficacy measure was constructed from daily observations over an 8-week period. A mixed-effects model provides a framework for analyzing structured observations while accounting for clustering represented in the model through random effects. In MATIC, country/region within cohort was modeled as a random effect, while treatment and cohort were fixed factors.
Why was the square root of TCRS modeled?
The registry specifies a square-root transformation for TCRS and several related symptom and medication outcomes. Transformation changes the scale of the response used by the model and can alter the relationship between the mean and variance. The reported treatment result should therefore be understood in the context of the transformed model rather than as a simple unadjusted comparison of raw daily scores.
What does the mean difference of 1.0 mean?
The registry reports a final-value mean difference of 1.0 for the primary TCRS comparison, with placebo versus 12 SQ-HDM as the recorded contrast. Because higher TCRS means worse symptoms and/or greater medication use, the positive direction corresponds to lower modeled TCRS under 12 SQ-HDM.
What does an odds ratio of 1.8 mean?
For SABA-free days and rhinitis mild days, the registry defines the odds ratio as the odds under 12 SQ-HDM divided by the odds under placebo. An OR of 1.8 therefore means 1.8 times the odds, not 1.8 times the probability. The distinction becomes particularly important when an outcome is common rather than rare.
Why can the P-value not be used as the effect size?
A P-value summarizes evidence against a null hypothesis under a specified statistical model. It does not tell the reader how large the treatment difference is. The effect estimate describes magnitude, while the confidence interval describes statistical precision. In MATIC, for example, the primary result combines a mean difference of 1.0, a 95% CI of 0.5–1.4, and P < 0.0001.
What does the confidence interval contribute?
The 95% confidence interval communicates uncertainty around the estimated treatment contrast under the model and sampling framework. A narrower interval indicates greater statistical precision than a wider interval, all else being equal. It does not describe the range of individual responses among children in the trial.
Why does multiplicity matter here?
The registry reports one primary endpoint and numerous secondary outcomes, with individual statistical tests and P-values. When many hypotheses are evaluated, the chance of at least one apparently positive result can increase if no multiplicity control is used. The registry information posted on ClinicalTrials.gov for these results does not specify an overall multiplicity-adjustment procedure across the posted secondary analyses, so their P-values should be interpreted as individual reported analyses rather than automatically as a single familywise-error-controlled set.
14. Intention-to-Treat and Analysis Populations
The primary analysis population was the full analysis set with observations in the primary efficacy period. The registry explicitly identifies intention-to-treat analysis as a concept associated with the primary and several secondary analyses.
| Feature | Registry description |
|---|---|
| Primary efficacy population | Participants in the full analysis set with observations in the primary efficacy period |
| Asthma DSS population | Participants with asthma in the full analysis set with observations in the primary efficacy period |
| SABA-free-day population | Participants with asthma in the full analysis set with observations in the primary efficacy period |
| Quality-of-life population | Participants in the full analysis set with observations |
The intention-to-treat principle is important because randomized treatment assignment is the basis for the causal comparison. However, the registry's wording for individual analyses also specifies the presence of observations in the relevant analysis period. That qualification should be retained when describing the actual analysis population rather than simply labeling every analysis as an unrestricted analysis of all 1460 enrolled participants.
15. Randomization and Blinding
Randomization is the design mechanism that balances treatment assignment probabilistically with respect to baseline characteristics. Double masking can further reduce systematic differences in expectations, reporting, assessment, or treatment administration. These design features support interpretation of the observed between-group differences as treatment comparisons rather than simply comparisons between pre-existing groups.
16. Safety
The registry reports serious adverse events by randomized treatment arm using affected participants over the number at risk.
| Safety measure | HDM SLIT-tablet (12 SQ-HDM) | Placebo SLIT-tablet |
|---|---|---|
| Serious adverse events | 16/727 | 6/731 |
These figures describe serious adverse events by arm and should be kept separate from the efficacy analyses. No inferential comparison of serious adverse events is reported in the registry analyses, so the counts should not be converted into an unreported P-value or effect estimate.
The denominators differ between the two reported safety groups: 727 for the 12 SQ-HDM arm and 731 for placebo. The registry reports affected participants over those at-risk populations rather than a modeled treatment effect. Safety interpretation therefore begins with the observed counts and denominators, followed by consideration of the definition and ascertainment of serious adverse events.
17. Missing Data and Model-Based Interpretation
The registry repeatedly states that no missing-data approach was applied to the continuous mixed-model analyses. This is an important methodological detail because longitudinal clinical-trial analyses can use several different strategies for incomplete observations, including likelihood-based mixed models, multiple imputation, or other prespecified methods.
In MATIC, the registry instead describes analyses among participants with observations in the relevant efficacy period and explicitly states that no missing-data approach was applied. The resulting estimates therefore depend on the observed data included in the specified model and on the assumptions underlying that model.
18. Why the Primary Endpoint Is Statistically Interesting
The MATIC primary endpoint combines two clinically related dimensions: daily rhinitis symptoms and daily rhinitis medication use. That construction has both advantages and statistical consequences.
Symptom dimension
The symptom component captures the burden of rhinitis symptoms over repeated daily assessments.
Medication dimension
The medication component incorporates rhinitis medication use into the combined daily score.
Repeated measurement
The endpoint averages daily observations over an 8-week assessment period rather than relying on one visit.
Model adjustment
The primary model incorporates treatment, cohort, baseline TCRS, and a country/region-within-cohort random effect.
Because the primary endpoint is an average of daily values, the result represents a summary of the longitudinal experience during the prespecified efficacy period. It is not a statement that the treatment produced the same magnitude of change on every individual day.
19. Understanding the Primary Confidence Interval
Primary treatment contrast
95% CI: 0.5–1.4
The point estimate is the single best estimate produced by the reported analysis, while the confidence interval communicates its statistical precision. The interval from 0.5 to 1.4 is substantially narrower than an interval that would extend across many possible effect sizes, indicating that the estimated mean difference is statistically more precisely characterized than it would be with a very wide interval.
The interval also needs to be interpreted in the direction of the registry contrast. The registry records the continuous mean-difference analyses as placebo versus 12 SQ-HDM. Since higher rhinitis scores indicate worse symptoms and/or more medication use, a positive contrast in that direction corresponds to lower values in the active-treatment group.
The 95% CI is about uncertainty in the estimated population-level treatment contrast. It does not mean that 95% of individual children had treatment effects between 0.5 and 1.4. Individual treatment responses can vary considerably even when the estimated population-level effect is statistically precise.
20. Interpreting the Odds Ratios Correctly
The binary outcomes illustrate why the choice of effect measure matters. For MATIC, the registry uses odds ratios for daily binary outcomes and explicitly defines the direction of the odds ratios as 12 SQ-HDM relative to placebo.
| OR | Statistical interpretation |
|---|---|
| 1 | Equal modeled odds between treatment groups |
| >1.0 | Higher modeled odds under 12 SQ-HDM |
| <1.0 | Lower modeled odds under 12 SQ-HDM |
For SABA-free days and rhinitis mild days, values above 1 correspond to higher odds of the favorable daily state. For rhinitis exacerbation days, an odds ratio below 1 corresponds to lower odds of the unfavorable daily state.
This illustrates why the clinical meaning of an odds ratio cannot be separated from the endpoint definition. An OR of 1.8 for a favorable outcome and an OR of 0.6 for an unfavorable outcome can point in the same broad direction while representing different underlying events.
21. Limitations and Interpretation Issues
- Multiple endpoints: the registry reports one primary endpoint and numerous secondary analyses. The ClinicalTrials.gov record does not specify a multiplicity-adjustment strategy across the secondary analyses.
- Analysis population: the posted analyses generally use the full analysis set with observations in the relevant period rather than simply describing all 1460 enrolled participants as contributing to every endpoint.
- Missing data: the registry states that no missing-data approach was applied to the continuous mixed-model analyses.
- Model dependence: the continuous results depend on the specified transformations, covariate adjustments, fixed effects, random effects, and residual-error structure.
- Binary interpretation: odds ratios should not be interpreted as risk ratios or direct probability differences.
- Subgroup-specific asthma analyses: several asthma outcomes apply only to participants with asthma, so their interpretation is specific to that analysis population.
- Safety: serious adverse-event counts are reported descriptively by arm, without a posted inferential comparison in the analyses posted on ClinicalTrials.gov.
- Clinical magnitude: statistical evidence does not by itself establish whether a numerical difference is clinically meaningful. That interpretation depends on the endpoint scale and clinical context.
22. Why This Trial Matters Statistically
MATIC is a useful statistical teaching case because it combines randomized treatment allocation with repeated daily patient-reported or recorded outcomes. Rather than reducing the entire trial to a single P-value, the registry provides a layered structure of continuous and binary endpoints analyzed with related mixed-effects methods.
| Concept | How it appears in MATIC |
|---|---|
| Randomization | Randomized parallel-group comparison |
| Blinding | Double-masked design |
| Superiority testing | Primary and secondary analyses use a superiority hypothesis |
| Mixed-effects modeling | Primary and many secondary continuous outcomes use linear mixed-effect models |
| Generalized mixed models | Binary daily outcomes use generalized linear mixed models |
| Covariate adjustment | Baseline endpoint measurements are incorporated into many models |
| Random effects | Country/region within cohort is included as a random effect in several analyses |
| Odds ratios | Used for SABA-free days, rhinitis mild days, and rhinitis exacerbation days |
| Confidence intervals | Reported with the posted treatment estimates |
| Missing data | No missing-data approach is reported for the continuous mixed-model analyses |
23. Statistical Methods Explained in Practical Terms
Why not simply compare the average TCRS between the two arms?
The primary endpoint is derived from daily observations, and the registry specifies a model that incorporates baseline TCRS, cohort, and a country/region-within-cohort random effect. A simple unadjusted comparison would therefore not reproduce the registered analysis.
What is the role of the baseline TCRS covariate?
Baseline TCRS provides information about each participant's starting severity. Including the baseline value as a covariate adjusts the model for baseline differences and can improve the precision of the treatment comparison.
Why is country/region modeled as a random effect?
The registry specifies country/region within cohort as a random effect. This allows the model to account for clustering associated with those geographic/cohort units rather than assuming that all observations are statistically independent with respect to that structure.
Why are different residual errors specified for each treatment?
The registry explicitly specifies different residual errors for each treatment in the linear mixed-effect models. This permits the residual variability to differ between treatment groups rather than imposing a common residual variance.
Why use a GLMM for SABA-free days?
SABA-free day is binary: a given day is classified as SABA-free or not. A generalized linear mixed model with a logit link is designed for binary responses while allowing the model to incorporate the longitudinal and clustered structure represented by the specified random effects.
Does P < 0.0001 mean the treatment effect is large?
No. It indicates strong statistical evidence against the relevant null hypothesis under the specified analysis. Effect magnitude is described by the treatment estimate, while its precision is described by the confidence interval. A very small P-value can occur with a precisely estimated effect that is not necessarily large in clinical terms.
24. Overall Statistical Interpretation
The primary TCRS analysis reported a mean difference of 1.0 with a 95% CI of 0.5–1.4 and P < 0.0001. The registry's placebo-versus-12-SQ-HDM contrast and the endpoint's direction indicate lower modeled TCRS under the 12 SQ-HDM treatment.
Related rhinitis and rhinoconjunctivitis symptom and medication scores also produced posted mean differences in the direction of lower scores under 12 SQ-HDM, with confidence intervals and P-values reported for each endpoint.
The binary analyses provide a complementary view. The odds of SABA-free days and rhinitis mild days were estimated using odds ratios above 1, while the odds of a rhinitis exacerbation day were estimated using an odds ratio below 1.
The confidence intervals provide the principal measure of precision around the posted treatment estimates. Because many secondary outcomes were analyzed, individual P-values should not automatically be interpreted as a single multiplicity-controlled family of confirmatory tests without an explicitly reported adjustment strategy.
25. Related Tutorials
Learn more about the methods used in this trial:
26. Related Calculators
27. Sources
- ClinicalTrials.gov: NCT04145219.
- PubMed: PMID 39678704.
- PubMed: PMID 32548677.
Continue through Clinical Biostats
Explore the statistical methods underlying randomized trials, longitudinal analyses, binary outcomes, confidence intervals, and clinical-trial interpretation.
28. Record Summary
MATIC provides a useful example of how a randomized clinical trial can combine a longitudinal primary endpoint with multiple secondary symptom, medication, quality-of-life, and binary daily outcomes. The primary analysis used a linear mixed-effect model with square-root transformation, baseline adjustment, cohort effects, a country/region-within-cohort random effect, and treatment-specific residual errors. Binary outcomes were analyzed with generalized linear mixed models and odds ratios.
The primary TCRS result was a mean difference of 1.0 with a 95% CI of 0.5–1.4 and P < 0.0001. Related rhinitis and rhinoconjunctivitis measures generally showed mean differences in the same direction, while binary endpoints provided complementary odds-ratio estimates. The statistical interpretation is strongest when these effect estimates are considered together with their confidence intervals, endpoint definitions, analysis populations, model specifications, and the broader set of secondary analyses.