This page separates reported trial results from statistical interpretation. Numerical results and trial characteristics on this page are restricted to the ClinicalTrials.gov record for NCT03711162 and the linked PubMed records identified in that data.
1. Trial at a Glance
ISABELA 1 was a randomized, parallel, quadruple-masked, phase 3 treatment study in idiopathic pulmonary fibrosis. The registry reports 525 enrolled participants and three study arms involving GLPG1690 600 mg, GLPG1690 200 mg, and placebo. The study was terminated prematurely based on recommendations of the Independent Data Monitoring Committee.
| Feature | ISABELA 1 |
|---|---|
| Trial name | ISABELA 1 |
| NCT identifier | NCT03711162 |
| Phase | Phase 3 |
| Condition | Idiopathic Pulmonary Fibrosis |
| Therapeutic area | Pulmonology |
| Allocation | Randomized |
| Design model | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Enrollment | 525 |
| Arms | 3 |
| Interventions | GLPG1690 and placebo |
| Status | Terminated |
| Lead sponsor | Lakefront Biotherapeutics NV |
| Sponsor type | Industry |
2. Clinical Question
The registry describes ISABELA 1 as a phase 3 study testing how effective and safe GLPG1690 is for subjects with idiopathic pulmonary fibrosis when used together with standard of care. The primary registered endpoint was the annual rate of decline in forced vital capacity (FVC) up to week 52.
Population
Subjects with idiopathic pulmonary fibrosis enrolled in the phase 3 ISABELA 1 study.
Intervention
GLPG1690, evaluated at 600 mg and 200 mg in the reported comparisons.
Comparator
Placebo.
Primary question
How does GLPG1690 compare with placebo for the annual rate of decline in FVC from baseline through week 52?
3. Trial Design
GLPG1690 600 mg
- GLPG1690
- Reported against placebo for the primary FVC analysis
- Reported for multiple secondary efficacy endpoints
GLPG1690 200 mg
- GLPG1690
- Reported against placebo for the primary FVC analysis
- Reported for multiple secondary efficacy endpoints
Comparator arm
The third arm was placebo. The ClinicalTrials.gov record compare each GLPG1690 dose with placebo rather than reporting a direct 600 mg versus 200 mg comparison.
The study was randomized and parallel, with quadruple masking. The registry reports that the study was terminated prematurely based on recommendations of the Independent Data Monitoring Committee. That fact is important when interpreting the maturity and precision of the reported estimates.
4. Trial Timing and Status
Study start
The registry lists November 28, 2018 as the study start date.
Primary completion
The registry lists March 30, 2021 as the primary completion date.
Terminated
The study was prematurely terminated based on recommendations of the Independent Data Monitoring Committee.
5. Primary Endpoint
| Endpoint | Registry definition | Time frame |
|---|---|---|
| Annual Rate of Decline in FVC up to Week 52 | FVC, in mL, is the maximum amount of air exhaled from the 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 the posted statistical analyses as mL/year. Rather than comparing only FVC at a single visit, the reported analysis estimated treatment-specific slopes over time and used the time-by-treatment interaction in a mixed model to determine the treatment effect.
The registry-reported analysis notes state that the treatment effect was determined from estimated slopes for each treatment group on the basis of the time-by-treatment interaction term from the mixed model.
6. Analysis Population
Several secondary analyses were also reported in the Full Analysis Set - Efficacy. For the composite endpoint involving mortality or non-elective lung transplantation, the registry-reported population description was FAS - EF with available data at specified time point.
7. Statistical Methodology
Mixed-effects modeling for longitudinal FVC
The primary FVC analysis was reported as a coefficient regression model. More importantly for interpretation, the analysis notes state that treatment effect was determined from estimated slopes using a time-by-treatment interaction term from a mixed model.
This structure is appropriate to the longitudinal nature of repeated FVC measurements because the question is fundamentally about how FVC changes over time. The treatment-by-time interaction asks whether the estimated rate of change differs between treatment groups.
A positive or negative mean difference therefore has to be interpreted in the context of the endpoint's definition: annual rate of decline in FVC, measured in mL/year.
Mixed-effects models for SGRQ
The secondary SGRQ analyses used mixed models. For the week 52 analyses, the registry-reported analysis notes specify treatment, categorical time, treatment-by-time interaction, stratum, and baseline SGRQ total score as fixed effects, with participant as a random effect.
This separates two sources of variation: systematic differences associated with treatment and time, and within-participant correlation created because the same participant contributes repeated observations.
Logistic regression
Disease progression was analyzed with logistic regression for the binary outcome "Percentage of Participants With Disease Progression." The reported effect measure was an odds ratio.
An odds ratio of 1 indicates equal odds under the fitted comparison. Values below 1 indicate lower estimated odds in the numerator group, while values above 1 indicate higher estimated odds. An odds ratio is not itself a risk ratio.
Cox proportional-hazards models
Time-to-event secondary endpoints were analyzed using Cox proportional-hazards models. The registry analysis text explicitly identifies Cox proportional hazards for respiratory-related hospitalization, all-cause hospitalization, acute IPF exacerbation, and several composite time-to-event outcomes.
The hazard ratio summarizes the estimated relative instantaneous event rate between groups under the fitted model. It does not directly report the difference in cumulative event probability at a particular time point.
Intention-to-treat as an associated analysis concept
The ClinicalTrials.gov record identifies intention-to-treat analysis as an associated concept for the primary and several secondary analyses. The actual analysis-population field, however, is the more specific source for determining who was included in each posted analysis.
8. Primary Results: Annual Rate of Decline in FVC Through Week 52
The registry posts two formal primary endpoint analyses: GLPG1690 600 mg versus placebo and GLPG1690 200 mg versus placebo. Both use a two-sided 95% confidence interval and a superiority hypothesis.
GLPG1690 600 mg vs Placebo
LS mean difference in annual FVC decline
95% CI: -52.3 to 97.6 mL/year · P = 0.5525
Estimated from the treatment-by-time interaction using the mixed-model slope approach.
The reported LS mean difference was 22.7 mL/year for GLPG1690 600 mg versus placebo. Because the endpoint is an annual rate of FVC decline, the sign should be interpreted according to the registry's coding and model definition rather than treated as a simple cross-sectional difference in FVC at week 52.
The 95% CI of -52.3 to 97.6 mL/year spans zero. In practical statistical terms, the registry-reported interval is compatible with treatment effects on either side of zero under the model and sampling framework. The interval also shows substantial uncertainty around the point estimate.
The P-value of 0.5525 is a measure of compatibility with the null hypothesis under the specified statistical test; it is not a measure of effect size. A P-value of 0.5525 does not mean that the treatment effect is "55.25% absent," nor does it quantify clinical importance.
Because the analysis is based on longitudinal slopes, interpretation also depends on the mixed-model specification and the handling of incomplete follow-up. The ClinicalTrials.gov record does not provide additional details about missing-data imputation or model diagnostics, so no stronger assumption should be attributed to the trial.
GLPG1690 200 mg vs Placebo
LS mean difference in annual FVC decline
95% CI: -100.5 to 47.1 mL/year · P = 0.4776
Estimated from the treatment-by-time interaction using the mixed-model slope approach.
The reported LS mean difference was -26.7 mL/year for GLPG1690 200 mg versus placebo. The negative point estimate is not sufficient by itself to establish a treatment effect because the corresponding 95% CI of -100.5 to 47.1 mL/year crosses zero.
The confidence interval is wide relative to the point estimate, indicating considerable statistical uncertainty. It is therefore important not to interpret the point estimate in isolation.
The P-value of 0.4776 does not measure the magnitude of the observed difference. It addresses the statistical evidence against the specified null under the reported analysis. It should be read together with the point estimate and confidence interval.
The two primary comparisons were specified as superiority analyses. Because both confidence intervals include zero and both reported P-values are substantially above conventional significance thresholds, the ClinicalTrials.gov record does not provide a formal statistical basis for claiming superiority on this endpoint.
| Primary comparison | Effect measure | Estimate | 95% CI | P-value |
|---|---|---|---|---|
| GLPG1690 600 mg vs Placebo | LS mean difference | 22.7 mL/year | -52.3 to 97.6 | 0.5525 |
| GLPG1690 200 mg vs Placebo | LS mean difference | -26.7 mL/year | -100.5 to 47.1 | 0.4776 |
9. Secondary Results: Disease Progression at Week 52
The registry reports disease progression through week 52 as a binary endpoint. Both dose comparisons used logistic regression and reported odds ratios with two-sided 95% confidence intervals.
GLPG1690 600 mg vs Placebo
Odds ratio for disease progression
95% CI: 0.56–1.74 · P = 0.9648
GLPG1690 200 mg vs Placebo
Odds ratio for disease progression
95% CI: 0.60–1.84 · P = 0.8530
| Comparison | OR | 95% CI | P-value | Method |
|---|---|---|---|---|
| 600 mg vs Placebo | 0.99 | 0.56–1.74 | 0.9648 | Logistic regression |
| 200 mg vs Placebo | 1.05 | 0.60–1.84 | 0.8530 | Logistic regression |
Both confidence intervals include 1.00, the null value for an odds ratio. The estimates therefore do not provide evidence of a clear difference in the odds of disease progression under the reported superiority analyses.
10. Secondary Results: Respiratory-Related Hospitalization
Respiratory-related hospitalization through the end of study was analyzed as a time-to-event endpoint. The registry describes the time frame as up to EoS, week 121, and reports hazard ratios from Cox proportional-hazards models.
GLPG1690 600 mg vs Placebo
Hazard ratio for respiratory-related hospitalization
95% CI: 0.50–2.05
GLPG1690 200 mg vs Placebo
Hazard ratio for respiratory-related hospitalization
95% CI: 0.47–1.88
A hazard ratio of 1.01 for the 600 mg comparison is very close to 1, while the 200 mg estimate of 0.93 is below 1. Neither point estimate should be interpreted without its confidence interval. The 95% CIs of 0.50–2.05 and 0.47–1.88 are both broad and include 1.
The registry does not post a P-value for these Cox analyses in the ClinicalTrials.gov record. Accordingly, no P-value-based conclusion should be added. The appropriate focus is the hazard-ratio estimate and its uncertainty.
Because these are time-to-event analyses, censoring and the proportional-hazards assumption are relevant. The ClinicalTrials.gov record identifies the Cox model but do not provide diagnostics for proportional hazards.
11. Secondary Results: SGRQ Total Score at Week 52
Change from baseline in St. George's Respiratory Questionnaire (SGRQ) total score at week 52 was analyzed using a mixed model. The registry reports LS mean differences with two-sided 95% confidence intervals and P-values.
| Comparison | LS mean difference | 95% CI | P-value |
|---|---|---|---|
| GLPG1690 600 mg vs Placebo | -0.5 | -4.4 to 3.3 | 0.7850 |
| GLPG1690 200 mg vs Placebo | 0.3 | -3.4 to 4.1 | 0.8617 |
The registry-reported model specification included treatment, categorical time, treatment-by-time interaction, stratum, and baseline SGRQ total score as fixed effects, with participant as a random effect. This is a repeated-measures structure: the model recognizes that observations from the same participant are correlated rather than treating them as independent observations.
12. Secondary Results: Annual Rate of FVC Decline Until EoS
The registry also reports annual rate of decline in FVC from baseline through EoS, defined as week 121. These analyses used a mixed-effects model and again derived treatment effects from estimated slopes based on the time-by-treatment interaction.
| Comparison | LS mean difference | 95% CI |
|---|---|---|
| GLPG1690 600 mg vs Placebo | 19.4 mL/year | -46.9 to 85.7 |
| GLPG1690 200 mg vs Placebo | -29.1 mL/year | -93.9 to 35.8 |
Both intervals include zero. These EoS slope estimates therefore show substantial uncertainty and should not be read as establishing superiority from the point estimates alone.
13. Secondary Results: Disease Progression Until EoS
Disease progression through EoS was analyzed using logistic regression.
| Comparison | Odds ratio | 95% CI |
|---|---|---|
| GLPG1690 600 mg vs Placebo | 1.15 | 0.68–1.95 |
| GLPG1690 200 mg vs Placebo | 1.25 | 0.74–2.09 |
Both confidence intervals include 1.00. The registry registry-reported no P-values for these EoS logistic-regression analyses, so the interpretation here is based on the odds-ratio estimates and their confidence intervals.
14. Secondary Results: SGRQ Total Score at Week 100
| Comparison | LS mean difference | 95% CI |
|---|---|---|
| GLPG1690 600 mg vs Placebo | 3.7 | -11.5 to 19.0 |
| GLPG1690 200 mg vs Placebo | 2.9 | -11.1 to 16.8 |
The registry-reported analysis notes state that these effects were estimated as the LS mean difference between each active treatment group and placebo from the mixed model. Both confidence intervals span zero, so the posted estimates are compatible with differences in either direction under the model.
15. Secondary Time-to-Event Results
The registry contains a substantial group of EoS time-to-event outcomes. These are particularly useful for understanding how different clinical definitions of progression or hospitalization can produce related but non-identical statistical estimands.
All-Cause Hospitalization
| Comparison | Hazard ratio | 95% CI |
|---|---|---|
| GLPG1690 600 mg vs Placebo | 1.08 | 0.65–1.78 |
| GLPG1690 200 mg vs Placebo | 1.23 | 0.76–1.98 |
The analysis was a Cox proportional-hazards model for time to first all-cause hospitalization. Both confidence intervals include 1.
Acute Idiopathic Pulmonary Fibrosis Exacerbation
| Comparison | Hazard ratio | 95% CI |
|---|---|---|
| GLPG1690 600 mg vs Placebo | 1.29 | 0.44–3.81 |
| GLPG1690 200 mg vs Placebo | 1.21 | 0.42–3.50 |
These intervals are notably broad, illustrating how relatively uncommon time-to-event outcomes can produce imprecise hazard-ratio estimates.
All-Cause Mortality or Hospitalization for Non-Elective Lung Transplant
| Comparison | Hazard ratio | 95% CI |
|---|---|---|
| GLPG1690 600 mg vs Placebo | 1.64 | 0.66–4.08 |
| GLPG1690 200 mg vs Placebo | 0.98 | 0.36–2.61 |
The analysis population for this endpoint was described as FAS-EF with available data at the specified time point. The wide intervals show substantial uncertainty, particularly for the 600 mg comparison.
All-Cause Mortality, Non-Elective Lung Transplant, or Hospitalization for Qualifying for Lung Transplant
| Comparison | Hazard ratio | 95% CI |
|---|---|---|
| GLPG1690 600 mg vs Placebo | 1.64 | 0.66–4.08 |
| GLPG1690 200 mg vs Placebo | 0.98 | 0.36–2.61 |
The registry values are identical to the preceding composite endpoint. This page reports them as separate registry outcomes rather than assuming that identical estimates represent a data-entry error.
All-Cause Mortality or Hospitalization Meeting a ≥10% Absolute Decline in %FVC or Respiratory-Related Hospitalization
| Comparison | Hazard ratio | 95% CI |
|---|---|---|
| GLPG1690 600 mg vs Placebo | 1.06 | 0.59–1.91 |
| GLPG1690 200 mg vs Placebo | 0.79 | 0.43–1.46 |
All-Cause Mortality or Respiratory-Related Hospitalization
| Comparison | Hazard ratio | 95% CI |
|---|---|---|
| GLPG1690 600 mg vs Placebo | 1.06 | 0.59–1.91 |
| GLPG1690 200 mg vs Placebo | 0.79 | 0.43–1.46 |
These composite endpoints illustrate an important statistical principle: combining several clinically meaningful events creates an estimand that is different from any single component. The hazard ratio describes time to the first qualifying event in the composite, not the treatment effect on each component separately.
16. Consolidated Secondary Results
| Endpoint | 600 mg vs Placebo | 200 mg vs Placebo |
|---|---|---|
| Disease progression to week 52 | OR 0.99 (95% CI 0.56–1.74); P = 0.9648 | OR 1.05 (95% CI 0.60–1.84); P = 0.8530 |
| Respiratory-related hospitalization to EoS | HR 1.01 (95% CI 0.50–2.05) | HR 0.93 (95% CI 0.47–1.88) |
| SGRQ change at week 52 | LS mean difference -0.5 (95% CI -4.4 to 3.3); P = 0.7850 | LS mean difference 0.3 (95% CI -3.4 to 4.1); P = 0.8617 |
| Annual FVC decline to EoS | LS mean difference 19.4 mL/year (95% CI -46.9 to 85.7) | LS mean difference -29.1 mL/year (95% CI -93.9 to 35.8) |
| Disease progression to EoS | OR 1.15 (95% CI 0.68–1.95) | OR 1.25 (95% CI 0.74–2.09) |
| SGRQ change at week 100 | LS mean difference 3.7 (95% CI -11.5 to 19.0) | LS mean difference 2.9 (95% CI -11.1 to 16.8) |
| All-cause hospitalization to EoS | HR 1.08 (95% CI 0.65–1.78) | HR 1.23 (95% CI 0.76–1.98) |
| Acute IPF exacerbation to EoS | HR 1.29 (95% CI 0.44–3.81) | HR 1.21 (95% CI 0.42–3.50) |
| Mortality or non-elective lung transplant to EoS | HR 1.64 (95% CI 0.66–4.08) | HR 0.98 (95% CI 0.36–2.61) |
| Mortality, non-elective transplant, or qualifying-for-transplant hospitalization to EoS | HR 1.64 (95% CI 0.66–4.08) | HR 0.98 (95% CI 0.36–2.61) |
| Mortality or hospitalization meeting ≥10% absolute decline in %FVC or respiratory-related hospitalization | HR 1.06 (95% CI 0.59–1.91) | HR 0.79 (95% CI 0.43–1.46) |
| Mortality or respiratory-related hospitalization to EoS | HR 1.06 (95% CI 0.59–1.91) | HR 0.79 (95% CI 0.43–1.46) |
17. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment arm. The figures are presented as affected participants divided by the number at risk.
| Study arm | Serious adverse events | Affected / at risk |
|---|---|---|
| GLPG1690 600 mg | Serious adverse events | 38 / 174 |
| GLPG1690 200 mg | Serious adverse events | 38 / 175 |
| Placebo | Serious adverse events | 36 / 174 |
The ClinicalTrials.gov record does not provide a statistical comparison or confidence interval for serious adverse events. The appropriate interpretation is therefore descriptive rather than a formal hypothesis test.
18. Statistical Methods Explained
Why was a mixed-effects model used for the primary FVC endpoint?
The primary endpoint is an annual rate of decline rather than a single measurement. Repeated FVC observations over time contain information about each participant's trajectory. A mixed-effects model can represent systematic treatment and time effects while accounting for repeated observations from the same participant. In this registry analysis, the treatment effect was specifically based on the time-by-treatment interaction and the resulting estimated slopes.
What does the FVC LS mean difference represent?
The LS mean difference is a model-based comparison of the estimated annual FVC decline between an active-treatment group and placebo. The 600 mg estimate was 22.7 mL/year, while the 200 mg estimate was -26.7 mL/year. These are slope differences, not differences between two raw FVC measurements at week 52.
Why are confidence intervals important for the primary endpoint?
The point estimate is only one possible summary of the data. The 95% confidence interval describes statistical uncertainty around that estimate under the model and sampling framework. For the 600 mg comparison, the interval was -52.3 to 97.6 mL/year; for 200 mg, it was -100.5 to 47.1 mL/year. Both include zero, the null value for a mean difference.
What does an odds ratio of 0.99 mean?
An odds ratio of 0.99 for disease progression in the 600 mg comparison means that the estimated odds ratio was very close to the null value of 1.00. It does not mean that the probability of progression was exactly 0.99 times the probability of progression, because odds and probabilities are different quantities.
Why was a Cox model used for hospitalization and exacerbation endpoints?
These outcomes are time-to-event endpoints: the analysis is concerned not only with whether an event occurred but also with when the first event occurred. Cox regression incorporates event times and right-censored observations and produces a hazard ratio as a relative measure of the instantaneous event rate.
What does a hazard ratio of 0.79 mean?
For the 200 mg comparison in the composite endpoint involving all-cause mortality or respiratory-related hospitalization, the reported hazard ratio was 0.79. Under the fitted Cox model, that corresponds to an estimated instantaneous event rate about 21% lower than placebo, because 1 - 0.79 = 0.21. This is a model-based relative statement; it does not mean that 21% of participants avoided the event or that cumulative event probability was reduced by exactly 21%.
Why should a hazard ratio not be interpreted as a constant risk ratio?
A hazard ratio concerns the relative instantaneous event rate at a given point in time under the model. A risk ratio compares cumulative probabilities over a specified period. They are mathematically different quantities. Cox interpretation also depends on the proportional-hazards framework, and the ClinicalTrials.gov record does not provide a diagnostic assessment of that assumption.
19. Primary Endpoint Interpretation in Detail
The primary 600 mg comparison produced an LS mean difference of 22.7 mL/year, while the 200 mg comparison produced -26.7 mL/year. Because the endpoint represents annual FVC decline and the treatment effect was estimated from longitudinal slopes, these values should be interpreted as differences between fitted trajectories rather than as simple end-of-study differences.
The 600 mg 95% CI was -52.3 to 97.6 mL/year; the 200 mg 95% CI was -100.5 to 47.1 mL/year. Both intervals span zero. The widths of these intervals also show that the point estimates are not highly precise summaries of a single underlying treatment effect.
The primary P-values were 0.5525 and 0.4776. A P-value is not an effect-size measure and should not be converted into a percentage benefit or percentage probability that the treatment works. Its role is to quantify the statistical evidence against the specified null under the reported test.
The registry identifies the primary hypothesis type as superiority. For a mean-difference analysis, the relevant null value is zero. The confidence intervals for both primary comparisons include zero, so the posted results do not establish statistical superiority on the registered primary endpoint.
20. Understanding the Time-to-Event Analyses
The EoS analyses use a common statistical framework but address different clinical events. The distinction is important because changing the event definition changes the estimand.
| Endpoint family | Statistical estimand | Model |
|---|---|---|
| Respiratory-related hospitalization | Time to respiratory-related hospitalization | Cox proportional-hazards model |
| All-cause hospitalization | Time to first all-cause hospitalization | Cox proportional-hazards model |
| Acute IPF exacerbation | Time to first acute IPF exacerbation | Cox proportional-hazards model |
| Mortality / transplant composites | Time to first qualifying component | Cox proportional-hazards model |
| Mortality / hospitalization composites | Time to first qualifying component | Cox proportional-hazards model |
A composite endpoint can increase the number of observed events, but it also combines events that may differ in clinical meaning. The reported hazard ratio therefore describes the composite endpoint as defined, not a generalized effect on every component.
21. Confidence Intervals and Null Values
Two null values recur throughout the ISABELA 1 statistical results:
Zero for mean differences
For LS mean differences in FVC decline and SGRQ change, zero represents no difference between treatment and placebo under the specified model.
One for ratios
For odds ratios and hazard ratios, one represents equal estimated odds or hazards between the compared groups.
This distinction is fundamental when reading the confidence intervals. A mean-difference interval that crosses zero and a ratio interval that crosses one are both compatible with their respective null values.
22. Multiplicity and Multiple Comparisons
The ClinicalTrials.gov record contains two primary treatment comparisons for the same primary endpoint and numerous secondary analyses across different outcome definitions and time frames. The ClinicalTrials.gov record identifies the hypotheses as superiority but do not provide an alpha-allocation or multiplicity-adjustment procedure.
This distinction is especially important when a trial contains multiple doses, multiple endpoints, repeated time points, and several related composite outcomes. Reporting all posted estimates is informative; it does not by itself establish that every individual comparison was confirmatory.
23. Interim Analysis and Early Termination
The registry caveat states that the study was prematurely terminated based on recommendations of the Independent Data Monitoring Committee.
This design feature affects how the final statistical record should be interpreted. An early-terminated trial may contain less accumulated information than originally anticipated. If the reason for stopping is not accompanied by the full stopping rule, information fraction, decision boundary, and statistical analysis plan, the registry result alone does not allow those design details to be reconstructed.
What is documented
The study was prematurely terminated following recommendations of the Independent Data Monitoring Committee.
What is not reported here
The ClinicalTrials.gov record does not specify an interim-analysis boundary, alpha-spending function, stopping threshold, or detailed missing-data procedure.
24. Missing Data and Model-Based Interpretation
The longitudinal FVC and SGRQ analyses rely on model-based estimates rather than simply calculating a complete-case arithmetic difference at one time point. That is a key statistical feature because longitudinal trials frequently have observations at some visits but not others.
However, the ClinicalTrials.gov record does not state a specific missing-data imputation method for the primary FVC analysis. This page therefore does not attribute a particular multiple-imputation, last-observation-carried-forward, or other missing-data procedure to ISABELA 1.
For the time-to-event analyses, participants who have not experienced the specified event by the end of their observed follow-up can contribute censored information. The ClinicalTrials.gov record identifies Cox modeling but does not provide detailed censoring rules beyond the endpoint definitions and analysis descriptions included in the data.
25. Dose Comparisons: Why the Two Estimates Should Be Read Separately
The primary results provide separate comparisons of 600 mg versus placebo and 200 mg versus placebo. They do not provide a formal 600 mg versus 200 mg treatment comparison in the statistical analyses posted on ClinicalTrials.gov.
| Feature | 600 mg vs Placebo | 200 mg vs Placebo |
|---|---|---|
| Primary LS mean difference | 22.7 mL/year | -26.7 mL/year |
| 95% CI | -52.3 to 97.6 | -100.5 to 47.1 |
| P-value | 0.5525 | 0.4776 |
It would be statistically incorrect to infer a dose-response relationship simply by comparing the two point estimates. A formal comparison between doses would require its own estimand and analysis.
26. Limitations
- Premature termination: the registry states that the study was prematurely terminated following recommendations of the Independent Data Monitoring Committee. This is an important feature of the evidence base and may affect information accumulation and precision.
- Incomplete design details in the ClinicalTrials.gov record: the ClinicalTrials.gov record does not specify the interim stopping boundary, alpha-spending method, sample-size assumptions, or detailed missing-data strategy.
- Analysis-population wording: the primary analyses are reported in the FAS-EF, which includes randomized participants who received at least one dose and has a site-related exclusion described only partially in the registry-reported text.
- Wide confidence intervals: several time-to-event endpoints, particularly acute IPF exacerbation and mortality/transplant composites, have broad confidence intervals.
- Multiplicity: the ClinicalTrials.gov record does not identify a multiplicity-adjustment strategy despite multiple doses, endpoints, and time frames.
- Cox model assumptions: hazard ratios rely on the Cox proportional-hazards framework. The ClinicalTrials.gov record does not provide proportional-hazards diagnostics.
- Composite endpoints: composite time-to-event outcomes combine multiple event types, so their hazard ratios should not be interpreted as effects on every component separately.
- Safety comparison: serious adverse events are reported descriptively by arm, without a formal statistical comparison or confidence interval.
- No subgroup results reported: the ClinicalTrials.gov record does not contain subgroup estimates, so no subgroup conclusions are drawn.
27. Why This Trial Matters Statistically
ISABELA 1 is a useful statistical teaching case because it combines several common clinical-trial estimands in one randomized phase 3 study. The primary endpoint is longitudinal and slope-based, while the secondary outcomes include binary, repeated-measures, and time-to-event endpoints.
| Concept | How it appears in ISABELA 1 |
|---|---|
| Randomization | Randomized parallel-group phase 3 design |
| Blinding | Quadruple masking |
| Longitudinal analysis | Annual rate of decline in FVC estimated from treatment-specific slopes |
| Mixed-effects model | Primary FVC analysis and SGRQ analyses |
| Time-by-treatment interaction | Basis for the estimated slope treatment effect in FVC analyses |
| Logistic regression | Disease progression endpoints |
| Odds ratio | Effect measure for binary disease-progression outcomes |
| Cox proportional-hazards model | Hospitalization, exacerbation, mortality, transplant, and composite time-to-event outcomes |
| Hazard ratio | Relative measure for time-to-event comparisons |
| Confidence intervals | Uncertainty around mean differences, odds ratios, and hazard ratios |
| Superiority testing | Primary and secondary hypothesis type reported as superiority |
| Early termination | Premature termination following Independent Data Monitoring Committee recommendations |
28. What the Primary Results Do and Do Not Establish
The reported primary LS mean differences were 22.7 mL/year for 600 mg versus placebo and -26.7 mL/year for 200 mg versus placebo, with 95% confidence intervals of -52.3 to 97.6 and -100.5 to 47.1, respectively.
Both intervals include zero, indicating that the ClinicalTrials.gov record is compatible with no mean difference as well as effects in either direction under the reported model.
The reported primary P-values were 0.5525 and 0.4776. These quantify statistical evidence against the relevant null hypothesis; they do not measure clinical magnitude, probability of treatment benefit, or the probability that the null hypothesis is true.
The registry-reported primary analyses do not establish superiority of either GLPG1690 dose over placebo for the registered annual rate of decline in FVC through week 52.
29. Secondary Endpoint Interpretation
The secondary analyses show the same need to keep effect size, precision, and endpoint definition separate.
Binary endpoints
Disease progression was analyzed with odds ratios. The null value is 1, and both week 52 confidence intervals include 1.
Longitudinal endpoints
SGRQ and longer-term FVC analyses use mixed-effects models, allowing treatment effects to be estimated in the presence of repeated observations.
Time-to-event endpoints
Hospitalization, exacerbation, mortality, transplant, and composite outcomes were analyzed with Cox proportional-hazards models.
Safety
Serious adverse events are reported descriptively by arm. The ClinicalTrials.gov record does not contain a formal safety hypothesis test.
30. Related Tutorials
Learn more about the methods used in this trial:
31. Related Calculators
Explore statistical tools corresponding to the main analysis concepts in this trial:
32. Sources
- ClinicalTrials.gov: NCT03711162 — ISABELA 1.
- PubMed record: PMID 38288082.
- PubMed record: PMID 37159034.
- PubMed record: PMID 31179008.
Continue with the statistical methods
Use the related tutorials and calculators to explore the longitudinal, categorical, and time-to-event methods represented in ISABELA 1.
33. Record Summary
ISABELA 1 provides a useful example of a phase 3 randomized trial in which the primary estimand is based on the longitudinal rate of FVC decline rather than a single cross-sectional measurement. The primary treatment effects were estimated from treatment-specific slopes using the time-by-treatment interaction in a mixed model, with LS mean differences of 22.7 mL/year for GLPG1690 600 mg versus placebo and -26.7 mL/year for GLPG1690 200 mg versus placebo. Their two-sided 95% confidence intervals both crossed zero, and the corresponding P-values were 0.5525 and 0.4776.
The secondary analyses broaden the statistical picture: disease progression was evaluated using logistic regression and odds ratios; SGRQ outcomes used mixed-effects models; and hospitalization, acute IPF exacerbation, mortality, transplant, and composite outcomes used Cox proportional-hazards models and hazard ratios. The ClinicalTrials.gov record reports serious adverse events in 38/174 participants in the 600 mg arm, 38/175 in the 200 mg arm, and 36/174 in the placebo arm.
The most important interpretive feature of the trial is that it was prematurely terminated based on recommendations of the Independent Data Monitoring Committee. That fact, together with the breadth of the confidence intervals for several secondary time-to-event endpoints and the absence of registry-reported details on multiplicity, interim boundaries, and missing-data procedures, means that the registry results should be read as a collection of prespecified and posted statistical estimates rather than as a single summary number.