IRIS: Complete Statistical Analysis of Follitropin Delta in Infertility

1. Trial at a Glance

IRIS is a completed phase 3 randomized trial evaluating the efficacy and safety of Follitropin Delta (FE 999049) in controlled ovarian stimulation in Indian women with infertility. The study compares Follitropin Delta with Follitropin Alfa (GONAL-F), with ongoing pregnancy rate specified as the primary endpoint.

2. Clinical Question

The central clinical question is whether Follitropin Delta (FE 999049) provides efficacy and safety information relevant to controlled ovarian stimulation in Indian women with infertility when compared with Follitropin Alfa (GONAL-F).

The registry identifies the study as a treatment trial rather than a diagnostic, prevention, or observational study. Its randomized allocation and parallel design establish the basic framework for comparing the two interventions while the single-masked structure affects how the trial should be interpreted.

3. Trial Design

The randomized parallel structure is statistically important because the principal comparison is between treatment groups formed before the outcome is assessed. Randomization is intended to reduce systematic differences between groups in factors that might otherwise influence the outcome. The registry, however, does not provide baseline characteristics or treatment-group sample sizes in the record summarized here, so the numerical balance of the groups cannot be evaluated from the registry information alone.

4. Trial Timeline

The distinction between study completion and availability of statistical results is important. A completed registry record establishes that the study has reached its listed completion status, but completion by itself does not establish what the observed treatment effect was.

5. Endpoints

The primary endpoint is a binary clinical outcome at a prespecified post-transfer time window. Each participant can therefore be classified according to whether the registry-defined condition for ongoing pregnancy is present at the endpoint assessment.

That endpoint differs from outcomes such as biochemical pregnancy, clinical pregnancy at an earlier assessment, live birth, or time to pregnancy. Those outcomes may be clinically related, but they are not interchangeable statistical endpoints. The registry definition should therefore remain the reference point when interpreting the primary endpoint.

Why the time frame matters

The assessment occurs 10–11 weeks after embryo transfer rather than at randomization. Statistically, that creates an important distinction between the trial's allocation point and its primary outcome assessment point. Randomization determines treatment assignment, while the endpoint is evaluated after the reproductive process has progressed to the registry-defined assessment window.

6. Planned Analysis

The ClinicalTrials.gov record identifies ongoing pregnancy rate as the primary endpoint but does not post statistical analyses or numerical efficacy estimates for that endpoint. No formal statistical analyses were posted to ClinicalTrials.gov.

How this endpoint is typically analysed

Ongoing pregnancy rate is a binary outcome. For each treatment group, the basic descriptive quantity is the proportion of participants meeting the registry-defined endpoint:

Common comparative measures for a binary endpoint include an absolute risk difference, a relative risk, or an odds ratio. The choice among them depends on the prespecified statistical analysis plan and the estimand of interest. The registry information summarized here does not identify which comparative effect measure was prespecified for the IRIS primary analysis.

Risk difference

An absolute risk difference compares the two proportions directly:

The major advantage of the risk difference is interpretability on an absolute scale. It directly describes the difference in the probability of meeting the endpoint. For a reproductive outcome, this scale can be particularly intuitive because it answers how much the observed event frequency differs between treatment groups.

Relative risk

A relative risk compares the probabilities multiplicatively:

Relative measures and absolute measures can tell different stories. A modest absolute difference can correspond to a larger relative difference when the baseline event probability is low, while the same relative ratio can correspond to a larger absolute difference when baseline probability is higher. For that reason, a complete statistical presentation would ideally report both absolute and relative measures when appropriate.

Confidence intervals

A confidence interval around a treatment effect quantifies statistical uncertainty associated with the estimated effect under the chosen statistical framework. For a risk difference, the interval is expressed on an absolute-proportion scale. For a relative risk or odds ratio, the interval is expressed on a ratio scale.

The confidence interval should not be interpreted as a range containing the individual outcomes of participants. It describes uncertainty around the estimated population-level effect. Its width is influenced by the amount of information available for the endpoint and by the number of outcome observations.

Hypothesis testing

A formal comparison of ongoing pregnancy rates could use a test for two independent proportions or a regression-based model for a binary outcome, depending on the prespecified analysis. A p-value from such a test would address evidence against a specified null hypothesis. It would not measure the size or clinical importance of the treatment effect.

7. Statistical Methodology

Randomization

Randomization is the central design feature supporting a causal comparison between Follitropin Delta and Follitropin Alfa. When participants are randomly allocated, treatment assignment is separated from prognostic characteristics in expectation. This reduces confounding compared with a nonrandomized comparison.

The statistical consequence is that the primary treatment comparison can be framed around the randomized groups rather than requiring statistical adjustment for every potential baseline characteristic. That does not mean the groups must be numerically identical at baseline. Randomization creates balance probabilistically, not deterministically.

Parallel-group analysis

The parallel design means participants remain associated with their randomized treatment group rather than switching between interventions as part of the study design. This is different from a crossover trial, where within-participant comparisons can play a central role.

For a binary endpoint such as ongoing pregnancy rate, the natural analysis is therefore a comparison of outcome proportions between the two randomized groups, subject to the analysis population and missing-data rules specified by the study's statistical plan.

Binary outcome analysis

The primary endpoint has a binary structure: participants either meet the registry definition of ongoing pregnancy at the specified assessment window or they do not. Binary outcomes are commonly summarized with event counts and proportions and compared with absolute or relative effect measures.

A simple two-group analysis can be represented as a two-by-two table:

The actual counts and treatment-group denominators are not reported in the registry information summarized here. The table therefore illustrates the statistical structure of the endpoint rather than presenting trial results.

Regression as an alternative framework

For a binary endpoint, logistic regression is one possible modelling framework. It estimates treatment effects through the odds of the event and can incorporate prespecified covariates when appropriate. Other regression approaches can directly model risks or risk differences.

Regression does not automatically improve a randomized comparison simply because more variables are included. Covariate adjustment should be connected to a prespecified analysis strategy and a clearly defined estimand. Without the trial's statistical analysis plan, the registry does not establish which regression model, if any, was used for the primary endpoint.

Analysis population

The interpretation of an ongoing pregnancy rate depends critically on its denominator. A rate among all randomized participants answers a different question from a rate among participants who received treatment or a rate among participants who reached embryo transfer.

This is one reason analysis populations matter. If participants are excluded after randomization because they do not reach a particular treatment or reproductive milestone, the resulting comparison may no longer represent the original randomized population in the same way. The registry information summarized here does not provide the analysis-population definitions or missing-data rules needed to determine the exact denominator used for a formal efficacy analysis.

8. Statistical Methods Explained

What makes ongoing pregnancy rate a binary endpoint?

The registry definition creates a yes-or-no outcome at the specified assessment window. A participant either has at least one intrauterine viable fetus 10–11 weeks after embryo transfer, or does not meet that definition. This allows the endpoint to be summarized as a proportion.

Why is the denominator important?

A percentage cannot be interpreted without knowing which participants are included in its denominator. For example, an ongoing pregnancy rate among all randomized participants and a rate among only participants reaching embryo transfer describe different populations. The denominator therefore forms part of the statistical definition of the result.

What does a risk difference tell us?

The risk difference puts the treatment comparison on an absolute scale. If one group has an ongoing pregnancy proportion of p₁ and the other has p₀, their difference is p₁ − p₀. It directly expresses how far apart the two observed proportions are.

What does a relative risk tell us?

The relative risk compares the two probabilities as a ratio. A relative risk of 1 would represent equal probabilities. A value above 1 would indicate a higher probability in the numerator group, while a value below 1 would indicate a lower probability. Relative risk does not tell us the absolute difference without knowing the underlying probabilities.

Why is a confidence interval useful?

An effect estimate is based on observed data and therefore has statistical uncertainty. A confidence interval provides a range of values compatible with the estimate under the specified statistical framework. A narrow interval indicates greater precision than a wide interval, all else being equal.

Why does a p-value not measure effect size?

A p-value addresses the compatibility of the observed data with a null hypothesis under a specified test. It depends on both the effect and the amount of information available. Two studies can therefore have similar effect estimates but different p-values, or similar p-values but different effect sizes. The treatment effect itself should be communicated using an interpretable effect estimate and its uncertainty.

Why does randomization matter for the analysis?

Randomization establishes treatment assignment before the outcome is observed and helps protect the comparison from systematic allocation differences. Statistical analysis then compares outcomes according to that randomized structure. Departures from the randomized population, such as exclusions or treatment-based analysis rules, can change the interpretation and therefore need to be specified carefully.

9. Missing Data and Endpoint Assessment

For a binary endpoint assessed after embryo transfer, missing endpoint assessments create a statistical issue because the pregnancy status is not observed for every participant who might otherwise contribute to the analysis.

Several approaches can be used in clinical-trial analysis, including prespecified definitions of the analysis population, conservative handling of missing outcomes, or model-based approaches. The appropriate choice depends on the estimand and the assumptions made about why observations are missing.

Missingness should not simply be treated as equivalent to failure without a prespecified rationale. Conversely, excluding participants with missing outcomes can alter the population being compared. The statistical analysis plan normally establishes how these cases are handled before unblinded outcome comparisons are made.

10. Single Masking and Statistical Interpretation

The ClinicalTrials.gov record identifies IRIS as single-masked. Masking can reduce the influence of treatment knowledge on participant behavior, investigator decisions, outcome assessment, or other aspects of trial conduct, depending on who is masked.

The registry information summarized here does not specify which party or parties were masked. That distinction matters because the statistical implications of masking depend on whether participants, investigators, outcome assessors, or another group were unaware of treatment assignment.

The primary endpoint is more objectively defined than a subjective symptom score because it uses the registry's specific clinical definition of an intrauterine viable fetus. Even so, endpoint ascertainment remains part of the trial's overall measurement process, and masking should be interpreted in the context of who performed the relevant assessments.

11. Multiplicity and Endpoint Hierarchy

The registry identifies one primary endpoint: ongoing pregnancy rate. This creates a comparatively simple primary-endpoint hierarchy within the information available from the record.

Multiplicity becomes particularly important when a trial evaluates many hypotheses. Testing multiple endpoints or multiple treatment comparisons without appropriate control of the overall type I error rate can increase the chance of at least one apparently positive result arising by chance.

For IRIS, the registry information summarized here identifies one primary endpoint but does not provide a multiplicity strategy for other potential analyses. The absence of a posted statistical analysis does not justify assuming that an unreported multiplicity procedure was or was not used.

12. Non-Inferiority, Superiority, and the Primary Question

The registry information identifies ongoing pregnancy rate as the primary endpoint but does not state a non-inferiority margin, superiority margin, or formal hypothesis-testing framework for that endpoint.

That distinction matters because a superiority trial and a non-inferiority trial answer different statistical questions. In a superiority framework, the analysis asks whether one treatment has evidence of a higher or lower outcome under the specified hypothesis. In a non-inferiority framework, the question is whether the new treatment is not worse than the comparator by more than a prespecified clinically acceptable margin.

A non-inferiority conclusion cannot be reconstructed merely from the absence of a statistically significant difference. The margin must be prespecified and the confidence interval interpreted relative to that margin. Because the registry information summarized here does not report such a margin, no non-inferiority interpretation can be assigned to the primary endpoint.

13. Safety Analysis

Safety is explicitly part of the official study title, which describes the efficacy and safety of FE 999049 in controlled ovarian stimulation. The registry identifies the two drug interventions and the study's treatment purpose.

However, the record summarized here does not provide numerical serious adverse-event counts by arm or other numerical safety results. Consequently, the statistical interpretation of safety must remain at the design level rather than treating the absence of posted event counts as evidence of equal safety.

How safety data are typically summarized

For randomized drug trials, adverse events are commonly summarized using participant-level event incidence, event counts, severity categories, seriousness, and treatment relationship when those variables are collected. For binary safety outcomes, treatment-group proportions can be compared using absolute and relative measures in much the same way as for an efficacy endpoint.

Serious adverse events require particular care because one participant can experience more than one event. A table of event counts is therefore not necessarily equivalent to a table of affected participants. The statistical unit should be made explicit whenever safety results are reported.

14. What a Complete Primary Endpoint Result Would Contain

A statistically informative report of the IRIS primary endpoint would contain more than a single percentage. At minimum, readers would need the treatment-group denominators, numbers meeting the ongoing pregnancy definition, the resulting proportions, an effect estimate comparing groups, and a measure of uncertainty.

This framework illustrates why an endpoint result should be interpreted as a complete statistical object rather than as a percentage in isolation. The percentage describes the event frequency; the effect measure describes the comparison; the confidence interval describes precision; and the p-value, when applicable, addresses a hypothesis-testing question.

15. Why Time From Embryo Transfer Matters

The registry defines the primary endpoint at 10–11 weeks after embryo transfer. This time origin is clinically meaningful because the endpoint is not simply a measurement of a laboratory parameter during ovarian stimulation. It represents a later reproductive outcome reached after several stages of the treatment pathway.

For statistical interpretation, this means that the endpoint combines information from the intervention period and the subsequent reproductive process leading to the assessment. Participants who do not reach embryo transfer, or whose endpoint cannot be assessed at the specified time, require explicit treatment in the analysis plan.

The time window also defines what the endpoint can and cannot establish. An ongoing pregnancy at 10–11 weeks after embryo transfer is a measure of pregnancy status at that point in time. It is not automatically a measure of outcomes occurring after the registry-defined assessment.

16. Statistical Interpretation vs Clinical Interpretation

These two perspectives should remain distinct. Statistical analysis asks how the randomized treatment groups compare and how precisely that comparison has been estimated. Clinical interpretation asks what the measured endpoint represents in the context of controlled ovarian stimulation and infertility treatment.

17. Important Limitations and Interpretation Issues

• No posted statistical analyses: the ClinicalTrials.gov record summarized here does not provide numerical primary-endpoint analyses, effect estimates, confidence intervals, or p-values.
• Primary endpoint only: the registry information identifies ongoing pregnancy rate as the primary endpoint, but does not provide additional endpoint results for interpretation.
• Analysis denominator: the registry information summarized here does not provide the treatment-group denominators used for the primary efficacy analysis.
• Missing-data rules: the record does not specify how participants without an evaluable ongoing pregnancy assessment were handled statistically.
• Masking details: the study is identified as single-masked, but the registry information summarized here does not specify who was masked.
• Baseline characteristics: the summarized registry information does not provide treatment-group baseline characteristics, so numerical comparability cannot be assessed.
• Statistical analysis plan: the registry information does not identify the formal effect measure, hypothesis test, regression model, or confidence-interval method used for the primary endpoint.
• Non-inferiority framework: no non-inferiority margin is reported in the registry information summarized here.
• Safety estimates: numerical serious adverse-event results by treatment arm are not reported in the information summarized here.
• Generalizability: the study concerns Indian women with infertility undergoing controlled ovarian stimulation; application of any future results to other populations would depend on their similarity to the enrolled population.

These limitations affect what can responsibly be concluded from the registry record. They do not change the fact that the study is a randomized phase 3 trial; rather, they define which statistical questions can be evaluated from the information currently reported in the record.

18. Why This Trial Matters Statistically

IRIS is a useful teaching case because it illustrates a common but important form of clinical-trial analysis: a randomized comparison with a binary clinical endpoint measured at a defined time after an intervention. The statistical challenge is not simply calculating two percentages. It is defining the estimand, the analysis population, the denominator, and the appropriate measure of treatment effect.

The deeper statistical lesson

A binary endpoint can look deceptively simple. In practice, its interpretation depends on a sequence of decisions: what constitutes an event, when the event is assessed, which participants form the denominator, how missing assessments are handled, which effect measure is chosen, and how uncertainty is quantified.

IRIS therefore provides a useful framework for understanding why trial statistics should be read as a linked chain of design and analysis decisions. Randomization establishes the comparison, the endpoint definition establishes what is measured, the analysis population establishes who contributes to the estimate, and the effect measure establishes how the treatment difference is communicated.

19. Sources

• ClinicalTrials.gov: IRIS — NCT04773353.

20. Record Summary

IRIS is a completed randomized phase 3 parallel-group trial enrolling 220 participants and evaluating Follitropin Delta (FE 999049) versus Follitropin Alfa (GONAL-F) for controlled ovarian stimulation in Indian women with infertility. The registry identifies ongoing pregnancy rate as the primary endpoint, measured 10–11 weeks after embryo transfer and defined as at least one intrauterine viable fetus at that time.

From a statistical perspective, the primary endpoint is a binary outcome for which treatment-group proportions can be compared using absolute and relative effect measures, accompanied by confidence intervals and, where prespecified, hypothesis testing. The most important interpretive issues concern the definition of the analysis denominator, the treatment of missing endpoint assessments, the specification of the treatment-effect measure, and the distinction between the endpoint itself and later reproductive outcomes.

The registry records randomized allocation, a parallel design, single masking, a treatment purpose, two drug interventions, and a total enrollment of 220. It does not provide the numerical primary efficacy analysis, formal statistical estimates, confidence intervals, p-values, treatment-group event counts, or numerical serious adverse-event results in the information summarized here. Consequently, the statistically appropriate reading of the record is centered on the trial design and the structure of the planned primary endpoint analysis rather than on an unreported treatment effect.