← Clinical Trials
Idiopathic Pulmonary Fibrosis Phase 3 Full Statistical Analysis NCT03733444

ISABELA 2: Complete Statistical Analysis of GLPG1690 in Idiopathic Pulmonary Fibrosis

An independent statistical review of the randomized phase 3 ISABELA 2 trial evaluating GLPG1690 used together with standard of care in participants with idiopathic pulmonary fibrosis (IPF), with emphasis on FVC decline, disease progression, hospitalization, patient-reported outcomes, time-to-event endpoints, and the statistical methods used to analyze them.

Trial start: November 5, 2018  ·  Primary completion: March 30, 2021  ·  Status: Terminated
Scope of this record

This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. Numerical trial results on this page are restricted to the registry-reported ISABELA 2 trial data.

1. Trial at a Glance

ISABELA 2 was a randomized, parallel-group, quadruple-masked phase 3 study evaluating GLPG1690 in participants with idiopathic pulmonary fibrosis when used together with standard of care. The registry reports an enrollment of 781 participants, three study arms, 35 posted outcome measures, and 26 posted statistical analyses.

781
Enrolled
Phase 3 study
3
Study arms
GLPG1690 600 mg, 200 mg, placebo
2
Primary comparisons
600 mg and 200 mg vs placebo
125
EoS week
For multiple secondary endpoints
FeatureISABELA 2
Trial nameISABELA 2
ClinicalTrials.gov identifierNCT03733444
PhasePhase 3
ConditionIdiopathic Pulmonary Fibrosis
AllocationRandomized
Design modelParallel
MaskingQuadruple
Primary purposeTreatment
Enrollment781
InterventionsGLPG1690 and placebo
Lead sponsorLakefront Biotherapeutics NV
Sponsor typeIndustry
StatusTerminated

2. Clinical Question

The central question was whether GLPG1690, administered at 600 mg or 200 mg and used together with standard of care, produced a favorable difference in the annual rate of decline in forced vital capacity (FVC) through week 52 compared with placebo.

Population

Participants with idiopathic pulmonary fibrosis enrolled in the phase 3 ISABELA 2 study.

Intervention

GLPG1690 at 600 mg or 200 mg.

Comparator

Placebo.

Primary question

Does either GLPG1690 dose change the annual rate of FVC decline from baseline through week 52 relative to placebo?

3. Trial Design

01
Randomize781 enrolled
02
Three armsTwo GLPG1690 doses + placebo
03
Quadruple maskMasked randomized comparison
04
Week 52Primary FVC endpoint
05
EoSSeveral outcomes through week 125
ACTIVE ARM · 600 mg

GLPG1690, 600 mg

  • GLPG1690
  • Used together with standard of care
  • Primary comparison against placebo
  • Serious adverse events: 64/259 affected/at risk
ACTIVE ARM · 200 mg

GLPG1690, 200 mg

  • GLPG1690
  • Used together with standard of care
  • Primary comparison against placebo
  • Serious adverse events: 63/260 affected/at risk
CONTROL ARM

Placebo

  • Placebo
  • Used with standard of care
  • Comparator for both GLPG1690 dose groups
  • Serious adverse events: 42/258 affected/at risk
DESIGN FEATURE

Parallel randomized comparison

  • Three-arm parallel design
  • Quadruple masking
  • Primary purpose: treatment
  • Hypothesis type for posted analyses: superiority

The ClinicalTrials.gov record identifies the study as randomized, parallel, and quadruple-masked. They do not provide a randomization ratio, a detailed stratification scheme, or a crossover procedure, so those features are not inferred here.

4. Primary Endpoint

EndpointRegistry definitionTime frame
Annual Rate of Decline in Forced Vital Capacity (FVC) up to Week 52 FVC (in mL) is the maximum amount of air exhaled from lungs by a participant after taking their deepest possible breath, as measured by spirometry. Baseline up to week 52

The endpoint is continuous and is expressed in mL/year. The statistical analyses posted on ClinicalTrials.gov use estimated slopes over time rather than simply comparing a single FVC measurement at week 52.

5. Secondary Endpoints and Analysis Framework

The posted analyses cover several distinct statistical structures. Continuous longitudinal outcomes were analyzed with mixed-effects models; binary progression outcomes were analyzed with logistic regression; and time-to-event outcomes were analyzed with Cox proportional-hazards models.

Endpoint familyExample posted endpointMethodEffect measure
Longitudinal continuousAnnual rate of FVC declineMixed-effects modelLS mean difference
Continuous patient-reported outcomeChange from baseline in SGRQ Total ScoreMixed-effects modelLS mean difference
BinaryPercentage with disease progressionLogistic regressionOdds ratio
Time-to-eventRespiratory-related hospitalizationCox proportional-hazards modelHazard ratio
Time-to-eventAcute IPF exacerbationCox proportional-hazards modelHazard ratio

6. Statistical Methodology

Mixed-effects modeling of FVC decline

The primary FVC analysis determined the treatment effect using estimated slopes for each treatment group on the basis of a time-by-treatment interaction term from a mixed model. This is an important distinction from a simple comparison of baseline-adjusted week-52 means.

Conceptual treatment effect
Treatment difference = estimated slope for GLPG1690 − estimated slope for placebo

The resulting least-squares mean difference is expressed in mL/year. A positive difference means the estimated annual rate under GLPG1690 was higher than the corresponding placebo rate; a negative difference would mean a lower estimated rate.

Because the endpoint is an annual rate of decline, the sign of the treatment difference must be interpreted in relation to the direction of the underlying FVC trajectory. The estimate alone should not be read as a percentage improvement.

Logistic regression

Disease progression through specified time frames was analyzed using logistic regression. The reported effect measure is an odds ratio comparing the odds of progression between GLPG1690 and placebo.

Odds ratio
OR = odds of progression in GLPG1690 group ÷ odds of progression in placebo group

An OR above 1 indicates higher estimated odds of the event in the GLPG1690 group relative to placebo; an OR below 1 indicates lower estimated odds.

Cox proportional-hazards model

Multiple time-to-event endpoints were analyzed with Cox proportional-hazards models. These included respiratory-related hospitalization, all-cause hospitalization, acute IPF exacerbation, and composite outcomes involving mortality and hospitalization.

Hazard ratio
HR = estimated instantaneous event rate in GLPG1690 ÷ estimated instantaneous event rate in placebo

An HR above 1 indicates a higher estimated instantaneous event rate in the GLPG1690 group; an HR below 1 indicates a lower estimated instantaneous event rate.

Intention-to-treat principle

The posted analyses identify the Full Analysis Set as the analysis population and identify intention-to-treat analysis as an additional concept in the analysis text. This means the interpretation should remain anchored to the randomized comparison rather than being reframed as an observational comparison among participants who completed treatment.

7. Primary Results: Annual Rate of FVC Decline Through Week 52

The registry reports two formal primary-endpoint analyses, both using the Full Analysis Set. The treatment effect was based on estimated slopes from the time-by-treatment interaction in a mixed model.

GLPG1690 600 mg vs Placebo

LS mean difference in annual FVC decline

2.8 mL/year

95% CI: -46.9 to 52.4   ·   P = 0.9123

Analysis population: Full Analysis Set

Clinical Biostats interpretation

The estimated LS mean difference was 2.8 mL/year for GLPG1690 600 mg versus placebo. As defined by the reported analysis, this is the difference between the estimated treatment-group slopes over time.

The estimate is not a 2.8% change, and it is not a statement that individual participants lost 2.8 mL of FVC per year. It is a between-group difference in estimated annual rates.

The 95% confidence interval extends from -46.9 to 52.4 mL/year. That relatively broad interval includes both negative and positive values, so the estimate is compatible with a range of possible treatment differences under the model and sampling framework.

The P = 0.9123 value addresses the statistical evidence against the tested null comparison; it does not measure the size, clinical importance, or probability of the treatment effect. The p-value should therefore be read together with the estimate and confidence interval.

The analysis is based on a mixed model and a time-by-treatment interaction. Interpretation depends on the modeling assumptions and the handling of repeated FVC observations; the ClinicalTrials.gov record does not provide enough detail to independently assess the covariance structure or missing-data assumptions.

GLPG1690 200 mg vs Placebo

LS mean difference in annual FVC decline

1.7 mL/year

95% CI: -47.4 to 50.8   ·   P = 0.9456

Analysis population: Full Analysis Set

Clinical Biostats interpretation

The estimated LS mean difference was 1.7 mL/year for GLPG1690 200 mg versus placebo. As with the 600 mg comparison, the estimate represents a difference between estimated longitudinal slopes rather than a direct percentage change in lung function.

The 95% confidence interval ranges from -47.4 to 50.8 mL/year. It spans zero and therefore leaves substantial uncertainty about the direction and magnitude of the underlying between-group difference.

The P = 0.9456 value is a hypothesis-testing quantity, not an effect-size measure. A p-value close to 1 does not establish that the two treatment effects are exactly identical; it indicates that the observed result provides little evidence against the null comparison under the specified analysis.

The comparison was a superiority analysis. It should therefore not be reinterpreted as a non-inferiority test merely because the confidence interval crosses zero. A superiority confidence interval and a non-inferiority margin answer different statistical questions.

Primary comparisonEstimate95% CIP-valueAnalysis
GLPG1690 600 mg vs placebo2.8 mL/year-46.9 to 52.40.9123Coefficient regression model; treatment effect from time-by-treatment interaction in mixed model
GLPG1690 200 mg vs placebo1.7 mL/year-47.4 to 50.80.9456Coefficient regression model; treatment effect from time-by-treatment interaction in mixed model

8. Secondary Results: Disease Progression Through Week 52

Disease progression up to week 52 was analyzed as a binary endpoint using logistic regression in the Full Analysis Set.

ComparisonOdds ratio95% CIP-value
GLPG1690 600 mg vs placebo1.150.76 to 1.740.5162
GLPG1690 200 mg vs placebo1.070.71 to 1.620.7566

The 600 mg odds ratio of 1.15 means the estimated odds of disease progression were 1.15 times those in the placebo group under the reported logistic regression. The 200 mg estimate of 1.07 similarly corresponds to estimated odds 1.07 times those in the placebo group. Neither confidence interval excludes 1.

Clinical Biostats interpretation

Odds ratios are not risk ratios. An OR of 1.15 cannot be translated directly into a 15% higher probability of progression without knowing the underlying event probability. The confidence intervals are also important: the interval for the 600 mg comparison is 0.76 to 1.74, while the interval for the 200 mg comparison is 0.71 to 1.62.

The p-values of 0.5162 and 0.7566 describe evidence against the corresponding null odds-ratio comparison. They do not indicate the probability that GLPG1690 is ineffective or the probability that the observed effect will replicate.

9. Secondary Results: Respiratory-Related Hospitalization

Respiratory-related hospitalization until end of study was analyzed as a time-to-event endpoint through week 125, with the registry-reported analysis describing time to first respiratory-related hospitalization.

ComparisonHazard ratio95% CIAnalysis
GLPG1690 600 mg vs placebo2.151.20 to 3.85Cox proportional-hazards model
GLPG1690 200 mg vs placebo1.690.93 to 3.10Cox proportional-hazards model

Respiratory-related hospitalization

HR 2.15

GLPG1690 600 mg vs placebo   ·   95% CI 1.20–3.85

Clinical Biostats interpretation

For the 600 mg comparison, an HR of 2.15 corresponds to an estimated instantaneous rate of first respiratory-related hospitalization that was 2.15 times the placebo rate under the Cox model. This is a relative time-to-event measure, not a statement that 2.15 times as many participants were hospitalized.

The 95% CI of 1.20 to 3.85 expresses uncertainty around the estimated hazard ratio. The interval lies above 1, but its width indicates that the magnitude of the estimated association is not known precisely.

For the 200 mg comparison, the HR was 1.69 with a 95% CI of 0.93 to 3.10. That interval includes 1. These results should be interpreted as the registry-reported model estimates rather than as evidence that one dose is definitively different from the other.

Because these are Cox-model estimates, interpretation also depends on the proportional-hazards assumption. The ClinicalTrials.gov record does not provide a diagnostic assessment of that assumption.

10. Secondary Results: SGRQ Total Score at Week 52

Change from baseline in the St. George's Respiratory Questionnaire (SGRQ) Total Score at week 52 was analyzed with a mixed-effects model in the Full Analysis Set.

ComparisonLS mean difference95% CIP-value
GLPG1690 600 mg vs placebo-0.1-3.2 to 3.00.9370
GLPG1690 200 mg vs placebo-0.4-3.4 to 2.70.8064

The treatment effect is reported as an LS mean difference in change from baseline. The 600 mg estimate was -0.1, with a 95% CI of -3.2 to 3.0; the 200 mg estimate was -0.4, with a 95% CI of -3.4 to 2.7.

11. Secondary Results: Annual Rate of FVC Decline Through End of Study

The registry also reports annualized FVC decline from baseline through end of study, defined as week 125 for this endpoint. The treatment effect was estimated from the time-by-treatment interaction in a mixed-effects model.

ComparisonLS mean difference95% CI
GLPG1690 600 mg vs placebo2.9 mL/year-41.1 to 46.8
GLPG1690 200 mg vs placebo8.0 mL/year-35.5 to 51.5

These estimates are distinct from the week-52 primary endpoint because the time frame extends to end of study. The 600 mg estimate was 2.9 mL/year, while the 200 mg estimate was 8.0 mL/year. Both confidence intervals include zero.

Interpretation caution: the direction of an annual FVC slope difference must be understood in the context of the underlying decline trajectory. A positive treatment-minus-placebo difference does not automatically mean clinical benefit or harm without specifying the absolute slopes for both groups. Those separate slopes are not reported in the ClinicalTrials.gov record.

12. Secondary Results: Disease Progression Through End of Study

Disease progression through end of study was analyzed with logistic regression.

ComparisonOdds ratio95% CI
GLPG1690 600 mg vs placebo1.320.90 to 1.94
GLPG1690 200 mg vs placebo1.060.72 to 1.56

The 600 mg OR of 1.32 corresponds to estimated odds of progression 1.32 times those in the placebo group, while the 200 mg OR of 1.06 corresponds to estimated odds 1.06 times those in the placebo group. Both confidence intervals include 1.

13. Secondary Results: SGRQ Total Score at Week 100

The registry reports change from baseline in SGRQ Total Score at week 100 using a mixed-effects model.

ComparisonLS mean difference95% CI
GLPG1690 600 mg vs placebo0.9-12.4 to 14.1
GLPG1690 200 mg vs placebo3.6-10.4 to 17.6

The reported treatment effects were 0.9 for 600 mg and 3.6 for 200 mg. Their respective 95% confidence intervals were -12.4 to 14.1 and -10.4 to 17.6.

14. Secondary Results: All-Cause Hospitalization

All-cause hospitalization until end of study was analyzed with a Cox proportional-hazards model for time to first all-cause hospitalization.

600 mg vs placebo

HR 1.54

95% CI: 1.01–2.35

ComparisonHazard ratio95% CI
GLPG1690 600 mg vs placebo1.541.01 to 2.35
GLPG1690 200 mg vs placebo1.400.91 to 2.16

The 600 mg HR of 1.54 indicates a higher estimated instantaneous rate of first all-cause hospitalization under the reported Cox model. The 200 mg HR was 1.40. The confidence intervals provide the corresponding uncertainty ranges.

15. Secondary Results: Acute IPF Exacerbation

Acute idiopathic pulmonary fibrosis exacerbation through end of study was analyzed as time to first acute IPF exacerbation using a Cox proportional-hazards model.

ComparisonHazard ratio95% CI
GLPG1690 600 mg vs placebo2.921.04 to 8.14
GLPG1690 200 mg vs placebo1.680.55 to 5.13
Clinical Biostats interpretation

The 600 mg estimate of 2.92 is substantially above 1, while its confidence interval of 1.04 to 8.14 is wide. The width is important: although the interval lies above 1, the precise magnitude of the estimated hazard ratio is uncertain.

The 200 mg estimate of 1.68 has a much wider interval, 0.55 to 5.13, which includes 1. This illustrates why the point estimate should not be interpreted independently of its precision.

No p-values for these Cox analyses are provided in the ClinicalTrials.gov record, so none are inferred from the confidence intervals.

16. Secondary Results: Mortality and Lung-Transplant Endpoints

The registry reports several composite time-to-event endpoints involving all-cause mortality and hospitalization for non-elective or qualifying lung transplant.

EndpointComparisonHR95% CI
All-cause mortality or hospitalization for non-elective lung transplant 600 mg vs placebo 2.27 1.06 to 4.82
All-cause mortality or hospitalization for non-elective lung transplant 200 mg vs placebo 1.87 0.86 to 4.06
All-cause mortality, hospitalization for non-elective lung transplant, or hospitalization for qualifying for lung transplant 600 mg vs placebo 2.27 1.06 to 4.82
All-cause mortality, hospitalization for non-elective lung transplant, or hospitalization for qualifying for lung transplant 200 mg vs placebo 1.87 0.86 to 4.06

These endpoints illustrate a central feature of composite time-to-event analysis: the hazard ratio describes the time until the first qualifying component of the composite, not each component separately. A composite HR should therefore not be interpreted as a mortality HR unless mortality itself is the endpoint.

17. Secondary Results: Mortality, FVC Decline, and Respiratory Hospitalization

EndpointComparisonHR95% CI
All-cause mortality or hospitalization meeting ≥10% absolute decline in %FVC or respiratory-related hospitalization 600 mg vs placebo 1.99 1.20 to 3.31
All-cause mortality or hospitalization meeting ≥10% absolute decline in %FVC or respiratory-related hospitalization 200 mg vs placebo 1.49 0.88 to 2.54
All-cause mortality or respiratory-related hospitalizations 600 mg vs placebo 1.99 1.20 to 3.31
All-cause mortality or respiratory-related hospitalizations 200 mg vs placebo 1.49 0.88 to 2.54

The two sets of estimates are numerically identical in the ClinicalTrials.gov record, although the endpoint definitions differ. This is a useful reminder that the endpoint definition must be read alongside the effect estimate: identical numerical HRs do not make two composite endpoints statistically interchangeable.

18. Secondary Results Summary

Endpoint600 mg vs placebo200 mg vs placebo
Disease progression, week 52OR 1.15 (95% CI 0.76–1.74), P = 0.5162OR 1.07 (95% CI 0.71–1.62), P = 0.7566
Respiratory-related hospitalization, EoSHR 2.15 (95% CI 1.20–3.85)HR 1.69 (95% CI 0.93–3.10)
SGRQ change, week 52LS mean difference -0.1 (95% CI -3.2 to 3.0), P = 0.9370LS mean difference -0.4 (95% CI -3.4 to 2.7), P = 0.8064
Annual FVC decline, EoS2.9 mL/year (95% CI -41.1 to 46.8)8.0 mL/year (95% CI -35.5 to 51.5)
Disease progression, EoSOR 1.32 (95% CI 0.90–1.94)OR 1.06 (95% CI 0.72–1.56)
SGRQ change, week 1000.9 (95% CI -12.4 to 14.1)3.6 (95% CI -10.4 to 17.6)
All-cause hospitalization, EoSHR 1.54 (95% CI 1.01–2.35)HR 1.40 (95% CI 0.91–2.16)
Acute IPF exacerbation, EoSHR 2.92 (95% CI 1.04–8.14)HR 1.68 (95% CI 0.55–5.13)
Mortality or non-elective lung transplant, EoSHR 2.27 (95% CI 1.06–4.82)HR 1.87 (95% CI 0.86–4.06)
Mortality/transplant/qualifying transplant hospitalization, EoSHR 2.27 (95% CI 1.06–4.82)HR 1.87 (95% CI 0.86–4.06)
Mortality or ≥10% absolute %FVC decline hospitalization or respiratory hospitalization, EoSHR 1.99 (95% CI 1.20–3.31)HR 1.49 (95% CI 0.88–2.54)
Mortality or respiratory-related hospitalization, EoSHR 1.99 (95% CI 1.20–3.31)HR 1.49 (95% CI 0.88–2.54)

19. Statistical Methods Explained

Why was a mixed-effects model used for the primary FVC endpoint?

The primary endpoint is an annual rate of FVC decline, so the analysis uses repeated FVC information over time rather than treating week 52 as the only relevant measurement. The registry analysis states that the treatment effect was determined from estimated slopes using a time-by-treatment interaction term from a mixed model. This approach allows the treatment comparison to be expressed through differences in longitudinal trajectories.

What does an LS mean difference of 2.8 mL/year mean?

It is the reported difference between the model-based estimated slopes for GLPG1690 600 mg and placebo. It is not a relative percentage and does not mean that every participant experienced a 2.8 mL/year difference. The corresponding 95% confidence interval, -46.9 to 52.4 mL/year, is essential for understanding the precision of that estimate.

What does an odds ratio of 1.15 mean?

An OR of 1.15 means the estimated odds of disease progression were 1.15 times the odds in the comparator group under the logistic regression model. It does not mean that the probability of progression was 15 percentage points higher, nor does it mean that 15% more participants progressed.

Why are hazard ratios used for hospitalization and exacerbation?

These endpoints concern when the first event occurs, not merely whether an event occurred by a fixed date. Cox regression uses the timing of events and accommodates right-censored participants. The resulting hazard ratio summarizes the relative instantaneous event rate under the model.

Why does a confidence interval matter more than a point estimate alone?

A point estimate is only one estimate of the treatment contrast. The confidence interval describes statistical uncertainty around that estimate. For example, the 600 mg acute IPF exacerbation HR is 2.92, but its 95% CI is 1.04 to 8.14. The wide interval shows that the precise magnitude is uncertain even though the interval lies above 1.

Why should p-values and effect estimates be reported together?

The p-value addresses evidence against a null hypothesis under the specified statistical framework. It does not tell the reader how large or clinically important the effect is. Effect estimates and confidence intervals provide the magnitude and precision; p-values provide a different piece of information about statistical evidence.

What does the Full Analysis Set imply?

The analyses posted on ClinicalTrials.gov identify the Full Analysis Set and also identify intention-to-treat analysis as a concept in the analysis text. The important statistical principle is that the randomized comparison should not be replaced by a comparison restricted only to participants who completed treatment or follow-up.

20. Confidence Intervals and Effect Size

The primary endpoint provides a clear example of why the point estimate and confidence interval must be read together.

ComparisonPoint estimate95% confidence intervalInterpretive feature
600 mg vs placebo2.8 mL/year-46.9 to 52.4Includes zero and spans both directions of effect
200 mg vs placebo1.7 mL/year-47.4 to 50.8Includes zero and spans both directions of effect

For a mean difference, zero is the natural null value. A confidence interval crossing zero therefore means that the data are compatible with both a negative and a positive treatment difference under the model. That is different from proving that the treatment effect is exactly zero.

21. Reading the Time-to-Event Results

Several secondary endpoints are time-to-event outcomes. Their HRs should be interpreted as relative event-rate measures rather than as simple comparisons of percentages.

HR above 1

The estimated instantaneous event rate is higher in the GLPG1690 group than in placebo under the Cox model.

HR below 1

The estimated instantaneous event rate is lower in the GLPG1690 group than in placebo under the Cox model.

CI crossing 1

The confidence interval includes the null hazard ratio of 1, so the direction of the underlying hazard difference remains uncertain at that confidence level.

Wide CI

A wide interval means that the numerical magnitude of the hazard ratio is estimated with limited precision, even when the point estimate appears large.

Proportional-hazards caution: the ClinicalTrials.gov record identifies Cox proportional-hazards models but do not report a formal assessment of the proportional-hazards assumption. A single HR should therefore be understood as a model-based summary rather than a complete description of how treatment effects behaved at every point in follow-up.

22. Multiplicity and Multiple Comparisons

The ClinicalTrials.gov record identifies 26 statistical analyses, including two primary-endpoint analyses and numerous secondary analyses. The primary comparisons use a superiority hypothesis framework.

Because the study contains two active dose comparisons and many secondary endpoints, the reader should distinguish the prespecified primary question from the broader collection of reported analyses. A nominal p-value from an individual secondary endpoint does not, by itself, establish control of the familywise type I error rate across all analyses.

FeatureWhat the ClinicalTrials.gov record establishesWhat is not reported
Primary endpointAnnual rate of FVC decline through week 52Detailed multiplicity procedure
Primary comparisons600 mg vs placebo and 200 mg vs placeboExact alpha allocation between dose comparisons
Hypothesis typeSuperiorityDetailed alpha-spending or hierarchical procedure
Secondary analysesMultiple logistic, mixed-model, and Cox analysesFormal multiplicity adjustment across all secondary endpoints

23. Interim Analysis, Crossover, Bayesian Methods, and Missing Data

The ClinicalTrials.gov record does not report a formal interim-analysis strategy, an alpha-spending procedure, a crossover scheme, a Bayesian analysis, or a specific missing-data imputation method. These topics are therefore not presented as features of the ISABELA 2 analysis.

Why this distinction matters: statistical trial pages should not infer design features from the existence of a particular endpoint or model. For example, the use of a mixed-effects model does not by itself establish a particular missing-data assumption, and the presence of several analyses does not establish an interim-monitoring plan.

24. Safety Results

The ClinicalTrials.gov record reports serious adverse events by treatment arm as affected participants over participants at risk.

ArmSerious adverse eventsAffected / at risk
PlaceboSerious adverse events42/258
GLPG1690, 600 mgSerious adverse events64/259
GLPG1690, 200 mgSerious adverse events63/260

The registry therefore reports serious adverse events in all three arms. The denominators in this safety measure differ from the overall enrolled population of 781, which is why the affected/at-risk counts should be reported exactly as provided rather than substituted with the enrollment total.

Placebo

42 of 258 participants at risk were reported as affected by serious adverse events.

GLPG1690 600 mg

64 of 259 participants at risk were reported as affected by serious adverse events.

GLPG1690 200 mg

63 of 260 participants at risk were reported as affected by serious adverse events.

Interpretation

The ClinicalTrials.gov record provides serious adverse-event counts by arm but do not provide a full adverse-event profile, severity breakdown, or formal safety comparison.

25. Trial Timeline

2018-11-05

Study start

ISABELA 2 began as a randomized phase 3 study in participants with idiopathic pulmonary fibrosis.

2021-03-30

Primary completion

The registry reports March 30, 2021 as the primary completion date.

Week 52

Primary endpoint window

The primary endpoint was the annual rate of decline in FVC from baseline up to week 52.

Week 125

End-of-study analyses

Multiple secondary endpoints were evaluated through end of study, defined as week 125 for the analyses posted on ClinicalTrials.gov.

26. Limitations

27. Why This Trial Matters Statistically

ISABELA 2 is a useful teaching case because it combines several important statistical structures within one randomized phase 3 study. The primary endpoint is longitudinal and continuous, while the secondary program includes binary outcomes, patient-reported continuous outcomes, and multiple time-to-event endpoints.

ConceptHow it appears in ISABELA 2
RandomizationRandomized three-arm parallel phase 3 design
BlindingQuadruple masking
Longitudinal analysisMixed-effects modeling of FVC trajectories and SGRQ outcomes
Time-by-treatment interactionUsed to determine the treatment effect on estimated FVC slopes
Odds ratioLogistic regression for disease progression
Hazard ratioCox proportional-hazards analyses of hospitalization, exacerbation, and composite endpoints
Confidence intervals95% two-sided intervals accompany the reported estimates
Intention-to-treat analysisIdentified in the primary and secondary analysis descriptions
Multiple endpoints35 outcome measures and 26 statistical analyses are posted
Safety analysisSerious adverse events reported by arm

28. Overall Statistical Interpretation

Primary endpoint

The two primary analyses report LS mean differences of 2.8 mL/year for 600 mg versus placebo and 1.7 mL/year for 200 mg versus placebo. Their 95% confidence intervals are -46.9 to 52.4 and -47.4 to 50.8, respectively, with p-values of 0.9123 and 0.9456. Both intervals include zero.

Secondary endpoints

The secondary analyses span several statistical measures. Some reported Cox hazard ratios are above 1, including the 600 mg estimates for respiratory-related hospitalization, all-cause hospitalization, acute IPF exacerbation, and several composite endpoints. The confidence intervals vary considerably in width, and some include 1. These estimates should be interpreted individually according to their endpoint definitions and analysis models.

What the data do not establish

The ClinicalTrials.gov record does not provide a basis for inferring a non-inferiority conclusion, a Bayesian conclusion, a formal subgroup-treatment interaction, a particular missing-data mechanism, or a specific interim-analysis procedure. Those conclusions would require information not contained in the ClinicalTrials.gov record.

29. Related Tutorials

Learn more about the methods used in this trial:

30. Related Statistical Calculators

31. Sources

Continue through the Clinical Biostats statistical pathway

Use the trial's endpoints as a starting point for deeper study of longitudinal models, survival analysis, regression, confidence intervals, and randomized clinical-trial methodology.

32. Record Summary

ISABELA 2 provides a compact example of how a modern phase 3 trial can require several statistical frameworks at once. Its primary endpoint was the annual rate of decline in FVC through week 52, analyzed using estimated slopes from a time-by-treatment interaction in a mixed model. Secondary analyses extended the statistical program to logistic regression for disease progression, mixed-effects models for SGRQ outcomes, and Cox proportional-hazards models for hospitalization, acute IPF exacerbation, mortality, transplant, and composite time-to-event endpoints.

The most important statistical reading principle is to keep the endpoint definition, effect measure, confidence interval, p-value when reported, and analysis population together. A hazard ratio is not an odds ratio; an odds ratio is not a risk ratio; an LS mean difference in annual FVC decline is not a percentage treatment effect; and a p-value does not quantify effect size.

Clinical Biostats methodology: A trial-results page should distinguish what the registry reports from what statistical reasoning can legitimately infer. For ISABELA 2, that means reporting the posted estimates exactly, explaining the models that generated them, and avoiding unsupported conclusions about missing-data procedures, multiplicity, interim monitoring, Bayesian methods, or other design features not contained in the ClinicalTrials.gov record.