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
ATTAIN-1 was a randomized, double-blind, parallel-group phase 3 trial evaluating orforglipron in adults with obesity or overweight with weight-related comorbidities. The registry reports 3,127 enrolled participants across four arms, with placebo and three orforglipron dose groups.
| Feature | ATTAIN-1 |
|---|---|
| Trial name | ATTAIN-1 |
| ClinicalTrials.gov identifier | NCT05869903 |
| Phase | Phase 3 |
| Condition | Obesity; Overweight; Overweight or Obesity |
| Design | Randomized, double-blind, parallel-group |
| Primary purpose | Treatment |
| Enrollment | 3,127 |
| Interventions | Orforglipron and placebo |
| Lead sponsor | Eli Lilly and Company |
| Sponsor type | Industry |
| Status | Active, not recruiting |
| Primary completion | July 25, 2025 |
2. Clinical Question
The central statistical question was whether randomized assignment to orforglipron produced a different change from baseline in body weight at Week 72 compared with placebo during the primary treatment period.
Population
Adult participants with obesity or overweight with weight-related comorbidities.
Intervention
Orforglipron at 6 mg, 12 mg, or 36 mg.
Comparator
Placebo.
Primary question
How does the model-based change in body weight at Week 72 differ between each orforglipron dose and placebo?
3. Trial Design
Placebo
Placebo drug group. Serious adverse events: 46 of 948 participants at risk.
Orforglipron 6 mg
6 mg orforglipron group. Serious adverse events: 40 of 723 participants at risk.
Orforglipron 12 mg
12 mg orforglipron group. Serious adverse events: 39 of 724 participants at risk.
Orforglipron 36 mg
36 mg orforglipron group. Serious adverse events: 28 of 728 participants at risk.
The registry describes allocation as randomized, the design model as parallel, masking as double, and primary purpose as treatment. The four-arm structure permits separate placebo comparisons for the three orforglipron doses.
4. Endpoint Framework
The registered primary endpoint is percent change from baseline in body weight during the primary treatment period, measured from baseline to Week 72. The registry reports 12 outcome measures and 21 posted statistical analyses.
| Endpoint | Time frame | Role | Posted method |
|---|---|---|---|
| Percent Change From Baseline in Body Weight (Primary Treatment Period) | Baseline, Week 72 | Primary | Mixed Models Analysis |
| Change From Baseline in Waist Circumference (Primary Treatment Period) | Baseline, Week 72 | Secondary | Mixed Models Analysis |
| Change From Baseline in Systolic Blood Pressure (pooled doses) | Baseline, Week 72 | Secondary | Mixed Models Analysis |
| Percent Change From Baseline in Non-HDL Cholesterol (pooled doses) | Baseline, Week 72 | Secondary | Mixed Models Analysis |
| Percent Change From Baseline in Triglycerides (pooled doses) | Baseline, Week 72 | Secondary | Mixed Models Analysis |
| Change From Baseline in HbA1c | Baseline, Week 72 | Secondary | Mixed Models Analysis |
| Change From Baseline in Fasting Serum Glucose | Baseline, Week 72 | Secondary | Mixed Models Analysis |
| Percent Change From Baseline in Fasting Insulin | Baseline, Week 72 | Secondary | Mixed Models Analysis |
| Change From Baseline in Diastolic Blood Pressure (pooled doses) | Baseline, Week 72 | Secondary | Mixed Models Analysis |
| Change From Baseline in SF-36v2 Physical Component Score (pooled doses) | Baseline, Week 72 | Secondary | Mixed Models Analysis |
| Change From Baseline in SF-36v2 Mental Component Score (pooled doses) | Baseline, Week 72 | Secondary | Mixed Models Analysis |
5. Statistical Methodology
Mixed-model repeated measures
The primary body-weight endpoint was analyzed using a mixed-model repeated measures framework for post-baseline measures. The model incorporated geographic region, baseline by time by treatment, and strata by time by treatment. Strata were defined by the joint levels of sex and prediabetes status.
The registry reports an unstructured variance-covariance structure for change from baseline. The resulting LS Mean values are described as model-based estimates of the unconditional average treatment effect.
Why repeated measures matter
Body weight was evaluated longitudinally rather than only as a single independent observation. A mixed-effects repeated-measures approach uses the available post-baseline measurements within the prespecified treatment period and models their covariance rather than treating every observation as independent.
Stratification variables
The primary model used strata defined jointly by sex and prediabetes status, with the interaction of strata, time, and treatment included in the model. This allows the estimated treatment contrast to account for the prespecified stratification structure while modeling change over time.
Unstructured covariance
An unstructured variance-covariance structure allows the variances and covariances among repeated measurements to be estimated without imposing a simpler pattern such as compound symmetry. Statistically, this gives the model flexibility to represent different degrees of correlation among measurements at different time points.
Superiority testing
The posted primary comparisons use a superiority hypothesis. Each orforglipron dose was compared with placebo, and the registry reports two-sided 95% confidence intervals and P-values below .001 for all three primary comparisons.
6. Primary Results: Body Weight
The primary endpoint was percent change from baseline in body weight at Week 72. All three dose comparisons were analyzed with the mixed-effects model and reported as LS Mean Difference (Net).
6 mg Orforglipron vs Placebo
LS Mean Difference in percent change
95% CI: -7.71 to -6.18 · P < .001
Time frame: Baseline, Week 72
The estimated net difference in percent change from baseline in body weight was -6.95 percentage points for 6 mg orforglipron versus placebo under the posted mixed-effects analysis. The negative sign indicates that the modeled change was lower in the orforglipron group than in the placebo group for this endpoint.
The 95% confidence interval extends from -7.71 to -6.18. This interval describes uncertainty around the model-based treatment contrast; it does not describe the range of individual participant responses.
The P-value of < .001 addresses the statistical evidence against the superiority null hypothesis under the reported analysis. It does not measure the magnitude or clinical importance of the treatment effect. The effect estimate and its confidence interval provide that information.
Because the estimate comes from a longitudinal mixed model, its interpretation depends on the model specification, repeated-measures covariance structure, treatment-period definition, and analysis population. It should not be interpreted as a simple arithmetic difference between two raw Week 72 means.
12 mg Orforglipron vs Placebo
LS Mean Difference in percent change
95% CI: -9.22 to -7.58 · P < .001
Time frame: Baseline, Week 72
The estimated net difference in percent change from baseline in body weight was -8.40 percentage points for 12 mg orforglipron versus placebo. Relative to the 6 mg comparison, the posted estimate is numerically more negative, indicating a larger modeled between-group difference in this endpoint.
The 95% confidence interval of -9.22 to -7.58 quantifies uncertainty around that estimated treatment contrast. It does not mean that individual treatment effects must fall inside this interval.
The P-value of < .001 indicates strong statistical evidence under the reported superiority analysis, but a P-value is not a measure of effect size. The magnitude is communicated by the LS Mean Difference and its confidence interval.
As with the other primary comparisons, the estimate reflects the mixed-effects model and the registry-defined treatment period rather than an unadjusted comparison of observed Week 72 values alone.
36 mg Orforglipron vs Placebo
LS Mean Difference in percent change
95% CI: -12.35 to -10.61 · P < .001
Time frame: Baseline, Week 72
The estimated net difference in percent change from baseline in body weight was -11.48 percentage points for 36 mg orforglipron versus placebo. The point estimate is more negative than the posted estimates for 6 mg and 12 mg.
The 95% confidence interval, -12.35 to -10.61, indicates relatively tight statistical uncertainty around the estimated treatment contrast. As with all confidence intervals here, it concerns the estimated population-level model contrast rather than individual participant responses.
The P-value of < .001 provides evidence against the superiority null hypothesis under the specified analysis. It does not establish how important the effect is clinically and should not be interpreted as a probability that the null hypothesis is true.
The comparison remains a model-based longitudinal estimate. Interpretation therefore requires attention to the treatment-period definition and to the mixed-model assumptions rather than treating the estimate as a simple raw difference.
| Primary comparison | Estimate | 95% CI | P-value | Hypothesis |
|---|---|---|---|---|
| Placebo vs 6 mg Orforglipron | -6.95 | -7.71 to -6.18 | <.001 | Superiority |
| Placebo vs 12 mg Orforglipron | -8.40 | -9.22 to -7.58 | <.001 | Superiority |
| Placebo vs 36 mg Orforglipron | -11.48 | -12.35 to -10.61 | <.001 | Superiority |
7. Secondary Results: Waist Circumference
Change from baseline in waist circumference at Week 72 was analyzed using the same mixed-model framework, with separate comparisons of each orforglipron dose against placebo.
| Comparison | LS Mean Difference (cm) | 95% CI | P-value |
|---|---|---|---|
| Placebo vs 6 mg Orforglipron | -5.45 | -6.23 to -4.66 | <.001 |
| Placebo vs 12 mg Orforglipron | -6.90 | -7.71 to -6.10 | <.001 |
| Placebo vs 36 mg Orforglipron | -9.04 | -9.84 to -8.23 | <.001 |
All three posted waist-circumference contrasts are negative, with confidence intervals entirely below zero and P-values below .001. In this model, the negative LS Mean Difference represents a lower modeled change from baseline in the orforglipron group relative to placebo.
The three estimates range from -5.45 cm for 6 mg to -9.04 cm for 36 mg. The confidence intervals provide the statistical precision of each contrast; they do not establish that every participant experienced a change of that magnitude.
These are secondary endpoints, so their statistical interpretation should remain distinct from the primary body-weight endpoint. A small P-value alone does not determine clinical importance.
8. Secondary Results: Systolic and Diastolic Blood Pressure
The registry specifies pooled-dose analyses for blood pressure. The 6 mg, 12 mg, and 36 mg orforglipron groups were pooled for comparison with placebo.
| Endpoint | Comparison | LS Mean Difference | 95% CI | P-value |
|---|---|---|---|---|
| Systolic blood pressure | Placebo vs pooled 6 mg/12 mg/36 mg | -5.36 mmHg | -6.32 to -4.39 | <.001 |
| Diastolic blood pressure | Placebo vs pooled 6 mg/12 mg/36 mg | -1.60 mmHg | -2.26 to -0.94 | <.001 |
The pooled-dose analysis estimates a -5.36 mmHg net difference in systolic blood pressure and a -1.60 mmHg net difference in diastolic blood pressure, both with two-sided 95% confidence intervals entirely below zero.
Pooling changes the question being answered. These estimates describe the comparison of placebo with the combined 6 mg, 12 mg, and 36 mg groups rather than the treatment effect of any one dose separately.
The P-values below .001 indicate statistical evidence for a difference under the reported superiority analyses. They do not establish that the magnitude of either blood-pressure difference is clinically important for every participant.
9. Secondary Results: Lipids
Non-HDL cholesterol and triglycerides were analyzed as percent changes from baseline at Week 72 using pooled orforglipron doses.
| Endpoint | Comparison | Posted estimate | 95% CI | P-value |
|---|---|---|---|---|
| Percent change in non-HDL cholesterol | Placebo vs pooled 6 mg/12 mg/36 mg | -6.25 | -7.84 to -4.63 | <.001 |
| Percent change in triglycerides | Placebo vs pooled 6 mg/12 mg/36 mg | -12.2 | -14.8 to -9.5 | <.001 |
The registry reports these analyses using a geometric LS Mean Difference (Net), normalized in the statistical record as a ratio-of-geometric-means effect measure. The posted estimates themselves are -6.25 for non-HDL cholesterol and -12.2 for triglycerides.
Because the registry reports the effect estimates on this posted percent-change scale, they should be read as the reported net contrasts rather than converted into a different ratio or percentage without the underlying model output. The confidence intervals remain the appropriate measure of uncertainty around the posted estimates.
Both P-values are below .001. Statistical significance does not by itself establish the clinical relevance of the lipid changes, and the pooled-dose analysis does not provide dose-specific estimates for these endpoints.
10. Secondary Results: Glycemic Measures
HbA1c
| Comparison | LS Mean Difference | 95% CI | P-value |
|---|---|---|---|
| Placebo vs 6 mg Orforglipron | -0.28 percentage points | -0.31 to -0.25 | <.001 |
| Placebo vs 12 mg Orforglipron | -0.29 percentage points | -0.31 to -0.26 | <.001 |
| Placebo vs 36 mg Orforglipron | -0.35 percentage points | -0.38 to -0.32 | <.001 |
Fasting serum glucose
| Comparison | LS Mean Difference | 95% CI | P-value |
|---|---|---|---|
| Placebo vs 6 mg Orforglipron | -7.96 mg/dL | -9.01 to -6.91 | <.001 |
| Placebo vs 12 mg Orforglipron | -8.98 mg/dL | -10.02 to -7.94 | <.001 |
| Placebo vs 36 mg Orforglipron | -9.53 mg/dL | -10.62 to -8.44 | <.001 |
The HbA1c estimates are negative for all three doses, ranging from -0.28 to -0.35 percentage points, with relatively narrow 95% confidence intervals and P-values below .001. The fasting serum glucose estimates likewise favor the orforglipron groups numerically, ranging from -7.96 to -9.53 mg/dL.
These estimates describe model-based between-group contrasts at the Week 72 analysis time frame. They are not simply the average observed change within each arm and should not be interpreted as the percentage of participants achieving a prespecified glycemic threshold.
The P-values indicate statistical evidence under the reported superiority tests, whereas the effect estimates and confidence intervals describe magnitude and precision.
Fasting insulin
| Comparison | Posted estimate | 95% CI | P-value |
|---|---|---|---|
| Placebo vs 6 mg Orforglipron | -14.39 | -19.98 to -8.42 | <.001 |
| Placebo vs 12 mg Orforglipron | -17.88 | -23.46 to -11.89 | <.001 |
| Placebo vs 36 mg Orforglipron | -27.49 | -32.52 to -22.08 | <.001 |
Fasting insulin was analyzed using the geometric LS Mean Difference (Net), with the statistical record identifying the effect measure as a ratio of geometric means. The posted estimates are therefore retained exactly as reported rather than converted to an alternative scale.
11. Secondary Results: Health Status
The registry reports separate analyses for the physical and mental component scores of the Short Form-36 Health Survey Version 2 Acute Form, using pooled 6 mg, 12 mg, and 36 mg orforglipron doses versus placebo.
| SF-36v2 component | Comparison | LS Mean Difference | 95% CI | P-value |
|---|---|---|---|---|
| Physical Component Score | Placebo vs pooled 6 mg/12 mg/36 mg | 1.52 T-score | 1.00 to 2.04 | <.001 |
| Mental Component Score | Placebo vs pooled 6 mg/12 mg/36 mg | 0.44 T-score | -0.13 to 1.00 | 0.130 |
The physical component analysis produced a positive estimated difference of 1.52 T-score points, with a 95% confidence interval of 1.00 to 2.04 and P < .001. The mental component estimate was 0.44, but its 95% confidence interval, -0.13 to 1.00, crosses zero and the reported P-value is 0.130.
The contrast between these two results illustrates why confidence intervals and P-values should be considered together. A point estimate alone does not establish statistical evidence for a nonzero difference. For the mental component, the interval includes zero and the reported superiority test does not provide statistical evidence at conventional significance levels.
The physical and mental component scores were both analyzed using pooled doses, so neither result provides a separate estimate for 6 mg, 12 mg, or 36 mg.
12. Safety Results
The registry reports serious adverse events by treatment arm as the number affected divided by the number at risk.
| Arm | Serious adverse events | Affected / at risk |
|---|---|---|
| Placebo | Serious adverse events | 46 / 948 |
| 6 mg Orforglipron | Serious adverse events | 40 / 723 |
| 12 mg Orforglipron | Serious adverse events | 39 / 724 |
| 36 mg Orforglipron | Serious adverse events | 28 / 728 |
The registry's serious-adverse-event counts provide arm-specific affected and at-risk counts. These figures should be kept separate from the efficacy analyses because safety and efficacy use different estimands and address different questions.
The four arms have different numbers of participants at risk in the posted safety data, so the raw affected counts alone are not directly comparable. The affected/at-risk presentation provides the denominator needed to understand the frequency within each arm.
A serious-adverse-event frequency is also not a direct measure of causal risk without considering the trial's full safety definitions, exposure, follow-up, event timing, and appropriate comparative analysis.
13. Understanding the Primary Estimand
The primary analysis is more specific than the phrase "weight loss at Week 72" suggests. The registry defines the treatment-period data window as beginning at baseline and ending at the earliest date of study-drug discontinuation or initiation of prohibited weight-management treatments. Participants with evaluable data contribute to the endpoint-specific analysis.
The reported primary estimates are net model-based differences in percent change from baseline. A negative estimate indicates a lower modeled change in the orforglipron group relative to placebo.
This distinction matters because the reported LS Mean Difference is not necessarily equal to the difference between two simple observed means at Week 72. The mixed model incorporates repeated measurements, baseline information, time, treatment, geographic region, and the specified strata structure.
14. Statistical Methods Explained
Why was a mixed-effects model used?
The outcomes were measured repeatedly over time, and the registry specifies a mixed-model repeated-measures approach for post-baseline observations. This framework accounts for the longitudinal structure of the data and models within-participant covariance rather than treating repeated measurements as independent observations.
What does an LS Mean Difference represent here?
An LS Mean Difference is a model-based contrast between treatment groups after accounting for the variables included in the model. In ATTAIN-1, the registry describes the LS Mean as a model-based estimate of the unconditional average treatment effect. For the primary endpoint, the reported contrast is expressed as the net difference in percent change from baseline.
Why does the primary endpoint use percent change?
The registered primary endpoint is percent change from baseline in body weight, measured at baseline and Week 72. Expressing the outcome as percent change places the treatment contrast on a scale relative to baseline body weight rather than on an absolute weight scale.
What does the 95% confidence interval tell us?
The confidence interval quantifies statistical uncertainty around the model-based treatment contrast. For example, the 36 mg comparison has an estimate of -11.48 with a 95% CI of -12.35 to -10.61. The interval does not describe the distribution of individual participant responses and should not be interpreted as a range containing 95% of treatment effects in individual patients.
Why is a P-value different from an effect size?
A P-value describes the evidence against a specified null hypothesis under the statistical model. It does not tell us whether the estimated difference is small or large. That information comes from the treatment-effect estimate and its confidence interval. With a large randomized trial, even relatively modest differences can generate small P-values.
Why were some secondary outcomes pooled across doses?
The registry explicitly states that the systolic blood pressure, non-HDL cholesterol, triglyceride, diastolic blood pressure, and SF-36v2 component-score analyses used pooled data from the 6 mg, 12 mg, and 36 mg orforglipron groups. Pooling increases the amount of information contributing to a single treatment contrast but means that the resulting estimate does not distinguish among individual doses.
What is the importance of the unstructured covariance matrix?
Repeated observations from the same participant are correlated. An unstructured covariance matrix allows the model to estimate separate variances and covariances among the repeated measurements rather than imposing a predetermined correlation pattern. This flexibility can be useful when the relationship among measurements changes over time.
15. Dose-Specific Versus Pooled Analyses
One of the important statistical distinctions in ATTAIN-1 is that not all outcomes answer the same dose-response question. The primary body-weight endpoint and several metabolic outcomes have dose-specific comparisons, whereas several other secondary outcomes are prespecified as pooled-dose analyses.
| Analysis structure | Examples | Question answered |
|---|---|---|
| Dose-specific | Body weight, waist circumference, HbA1c, fasting serum glucose, fasting insulin | How does each specified dose compare with placebo? |
| Pooled-dose | Systolic blood pressure, non-HDL cholesterol, triglycerides, diastolic blood pressure, SF-36v2 | How does the combined orforglipron dose population compare with placebo? |
This distinction is important when interpreting the numerical estimates. A pooled estimate cannot be used to claim that each individual dose produced exactly the same effect.
16. Interpreting the Pattern of Primary Estimates
The three primary body-weight estimates are progressively more negative across the posted doses: -6.95 for 6 mg, -8.40 for 12 mg, and -11.48 for 36 mg, each compared with placebo.
| Dose | Estimated net difference | 95% CI | P-value |
|---|---|---|---|
| 6 mg | -6.95 | -7.71 to -6.18 | <.001 |
| 12 mg | -8.40 | -9.22 to -7.58 | <.001 |
| 36 mg | -11.48 | -12.35 to -10.61 | <.001 |
The point estimates form a numerically ordered pattern, with increasingly negative estimated differences at higher doses. That pattern is descriptively consistent with a larger modeled between-group difference at the higher posted dose levels.
However, a numerical ordering of point estimates is not itself a formal statistical test of dose-response. The posted analyses compare each dose with placebo; they do not, in the information reported here, provide a direct hypothesis test comparing 6 mg against 12 mg or 12 mg against 36 mg.
This is an important distinction between describing the observed pattern and making a formal inference about differences among doses.
17. Interpreting the Secondary Outcome Pattern
The posted secondary analyses cover several physiologic domains: anthropometric measures, blood pressure, lipids, glycemic measures, insulin, and health-status scores. Most reported estimates are statistically different from zero under their individual posted analyses, while the mental component of the SF-36v2 has a P-value of 0.130.
Anthropometry
Waist-circumference estimates were negative for all three dose-specific comparisons, with P-values below .001.
Blood pressure
Pooled-dose analyses produced negative estimates for both systolic and diastolic blood pressure, with P-values below .001.
Metabolic outcomes
HbA1c, fasting serum glucose, non-HDL cholesterol, triglycerides, and fasting insulin all had posted P-values below .001.
Health status
The physical component score had P < .001, while the mental component score had P = 0.130.
This pattern illustrates why a clinical-trial statistical review should examine the estimand, scale, confidence interval, and analysis population for each outcome rather than reducing the results to a single P-value.
18. Missing Data and Treatment-Period Handling
The registry specifies that all randomized participants with evaluable data for the specific outcome were included. The treatment-period definition ends at the earliest date of study-drug discontinuation or initiation of prohibited weight-management treatments.
The mixed-model approach uses the available post-baseline observations that meet the stated treatment-period definition. The registry describes the variance-covariance structure for the primary MMRM as unstructured.
19. Multiplicity and Multiple Comparisons
The registry reports three primary treatment comparisons for the single registered primary endpoint: placebo versus 6 mg, placebo versus 12 mg, and placebo versus 36 mg. It also reports numerous secondary analyses across different outcomes.
The posted P-values should therefore be understood in the context of the trial's prespecified statistical analysis plan and any multiplicity-control procedures that apply to the corresponding hypotheses. The registry information reported here identifies the hypothesis type as superiority and provides the individual two-sided P-values and 95% confidence intervals, but does not provide a detailed multiplicity hierarchy in the ClinicalTrials.gov record.
20. Confidence Intervals Versus P-values
ATTAIN-1 provides a useful illustration of why effect estimates, confidence intervals, and P-values should be read together.
| Endpoint | Estimate | 95% CI | P-value |
|---|---|---|---|
| Body weight, 6 mg | -6.95 | -7.71 to -6.18 | <.001 |
| Body weight, 12 mg | -8.40 | -9.22 to -7.58 | <.001 |
| Body weight, 36 mg | -11.48 | -12.35 to -10.61 | <.001 |
| SF-36v2 Mental Component | 0.44 | -0.13 to 1.00 | 0.130 |
The body-weight confidence intervals do not cross zero, while the mental-component confidence interval does. This corresponds to the very different reported P-values. The confidence interval additionally communicates the precision and plausible range of the estimated treatment contrast under the model.
21. Limitations and Interpretation Issues
- Model dependence: the primary and secondary efficacy estimates are model-based LS Mean differences from mixed-model analyses, so interpretation depends on the specified longitudinal model.
- Repeated-measures covariance: the primary MMRM uses an unstructured variance-covariance structure. Model-based inference depends in part on the adequacy of that covariance specification for the repeated observations.
- Evaluable-data population: each endpoint includes randomized participants with evaluable data for that specific outcome rather than a single identical analysis population for every endpoint.
- Treatment-period restriction: observations after the earliest discontinuation of study drug or initiation of prohibited weight-management treatments are outside the defined treatment-period analysis.
- Pooled versus dose-specific analyses: several secondary outcomes combine the three orforglipron doses, so their estimates cannot be interpreted as dose-specific treatment effects.
- Multiple endpoints: the trial has a large collection of posted analyses. The ClinicalTrials.gov record does not specify a complete multiplicity hierarchy for all posted secondary hypotheses.
- Effect-size interpretation: a statistically small P-value does not establish the clinical importance of an effect. Magnitude, precision, endpoint definition, and clinical context all matter.
- Generalizability: the registry describes a specific population of adult participants with obesity or overweight with weight-related comorbidities. Applying the estimates to populations outside the trial requires consideration of whether those populations are comparable.
22. Why This Trial Matters Statistically
ATTAIN-1 is a useful teaching example because it combines randomized parallel-group design with a longitudinal mixed-effects analysis and multiple types of clinical outcomes. The same core statistical framework is applied to body weight, waist circumference, blood pressure, laboratory measures, and patient-reported health-status scores, while the estimand and scale differ by endpoint.
| Statistical concept | How it appears in ATTAIN-1 |
|---|---|
| Randomization | Randomized allocation across four parallel treatment arms |
| Blinding | Double-masked trial design |
| Repeated measures | Post-baseline longitudinal observations analyzed with mixed models |
| Mixed-effects modeling | Primary body-weight and posted secondary analyses use Mixed Models Analysis |
| LS Mean Difference | Primary treatment effect reported as a model-based net difference |
| Confidence interval | Two-sided 95% intervals accompany the posted estimates |
| P-values | Superiority hypothesis tests accompany the posted estimates |
| Pooling | Several secondary outcomes pool the 6 mg, 12 mg, and 36 mg groups |
| Geometric modeling | Non-HDL cholesterol, triglycerides, and fasting insulin use geometric LS Mean Difference reporting |
| Safety denominators | Serious adverse events are reported as affected participants over participants at risk |
23. Record Summary
ATTAIN-1 provides a detailed example of longitudinal clinical-trial analysis in which the principal endpoint is a repeated-measures outcome summarized as a model-based LS Mean Difference. At Week 72, the registry reports primary body-weight estimates of -6.95, -8.40, and -11.48 for 6 mg, 12 mg, and 36 mg orforglipron versus placebo, respectively, each with a two-sided 95% confidence interval entirely below zero and P < .001.
The secondary results extend the same statistical framework to waist circumference, blood pressure, lipid measures, glycemic measures, fasting insulin, and SF-36v2 scores. The posted results also illustrate several important distinctions: dose-specific versus pooled analyses, arithmetic mean differences versus geometric LS Mean Difference reporting, statistical evidence versus effect magnitude, and the difference between a model-based population treatment contrast and an individual participant response.
24. Related Tutorials
Learn more about the methods used in this trial:
25. Related Calculators
26. Sources
- ClinicalTrials.gov: NCT05869903.
- PubMed: PMID 42750540.
- PubMed: PMID 42577069.
- PubMed: PMID 40960239.
Continue through Clinical Biostats
Connect trial endpoints and statistical methods with deeper biostatistics tutorials and practical statistical tools.