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
PURSUE was a randomized, parallel-group, quadruple-masked phase 3 study evaluating a single injection of SCH 900962 against daily recFSH injections in women undergoing controlled ovarian stimulation. The primary endpoint was the percentage of participants with a vital pregnancy, analyzed as a non-inferiority comparison.
| Feature | PURSUE |
|---|---|
| Phase | Phase 3 |
| Condition | Infertility |
| Study name | PURSUE; Study P06029 |
| Enrollment | 1,424 |
| Allocation | Randomized |
| Design | Parallel |
| Masking | Quadruple |
| Primary purpose | Treatment |
| Study start | 2010-06 |
| Primary completion | 2011-07 |
| Lead sponsor | Organon and Co |
| ClinicalTrials.gov | NCT01144416 |
2. Clinical Question
The central statistical question was whether a single injection of 150 µg SCH 900962/MK-8962 produced a percentage of participants with a vital pregnancy that was not unacceptably lower than that observed with daily 300 IU recFSH.
Population
Women undergoing controlled ovarian stimulation in the PURSUE phase 3 study for infertility.
Intervention
Single injection of 150 µg SCH 900962/MK-8962, also identified in the record as corifollitropin alfa / Org 36286.
Comparator
Daily 300 IU recFSH, identified in the record as follitropin beta.
Primary question
Is the difference in percentage of participants with a vital pregnancy compatible with the pre-defined non-inferiority margin of −8%?
3. Trial Design
Single Injection of 150 µg SCH 900962/MK-8962
- Biological intervention
- Also identified as corifollitropin alfa / Org 36286
- Single injection
- Placebo for recFSH was also part of the masked design
Daily 300 IU recFSH
- Biological intervention
- Also identified as follitropin beta
- Daily injections
- Placebo for SCH 900962 was also part of the masked design
The record describes the study as randomized, parallel, and quadruple-masked. The use of placebo for both study treatments is consistent with maintaining masking despite the different treatment schedules.
4. Randomization, Analysis Populations, and Stratification
The primary efficacy analysis used an Intent-To-Treat Group consisting of all randomized participants who received one or more doses of SCH 900962 or recFSH. Participants were analyzed according to the treatment to which they were randomized.
| Population | Definition / role |
|---|---|
| Intent-To-Treat Group | All randomized participants who received 1 or more dose(s) of SCH 900962 or recFSH. Participants were grouped according to randomized treatment. |
| Primary endpoint handling | Participants without embryo transfer or lost to follow-up before the ultrasound assessment confirming vital pregnancy were counted as non-pregnant. |
| All-Subjects-Treated Group | All participants who received 1 or more dose(s) of SCH 900962 or recFSH; used for the reported safety-related categorical analyses. |
| AST treatment assignment | For the safety-related analyses, participants were grouped according to the active treatment they actually received. |
The primary vital-pregnancy analysis was adjusted for age class, stratified as ≤38 years versus >38 years. The analysis of number of oocytes retrieved was adjusted for age class and center. The registry also identifies stratified analysis and covariate adjustment as concepts in the reported analyses.
5. Endpoints
| Endpoint | Registry definition / time frame | Analysis |
|---|---|---|
| Percentage of Participants With a Vital Pregnancy | Vital pregnancy assessed by ultrasound at least 35 days after embryo transfer, with a timeframe of 35–42 days. Time from start of study treatment to embryo transfer was maximally 24 days. Vital pregnancy was defined as the presence of at least 1 fetus with heart activity at least 35 days (≥5 weeks) after embryo transfer in the controlled ovarian stimulation treatment cycle. | Generalized linear model; risk difference; non-inferiority |
| Number of Oocytes Retrieved Per Attempt | Maximally 21 days after the start of study treatment. | ANOVA; mean difference; non-inferiority |
| Live Birth Rate | Approximately nine months after embryo transfer. | Generalized linear model; difference in live birth rates; non-inferiority |
| Number of Participants With Moderate or Severe Ovarian Hyperstimulation Syndrome (OHSS) | Up to approximately 1 month after oocyte pick-up. | Fisher exact test; superiority |
| Number of Participants Who Cancelled the Cycle Due to a (Serious) Adverse Event | Up to time of embryo transfer, maximum of 24 days after start of study drug. | Fisher exact test; superiority |
6. Statistical Methodology
Primary endpoint: generalized linear model
The primary endpoint was binary: whether a participant had a vital pregnancy. The registry reports a generalized linear model for the comparison between the single SCH 900962 injection and daily recFSH.
The effect measure was the difference in percentage vital pregnancy, normalized here as a risk difference. The reported estimate was adjusted for age class, with participants stratified by ≤38 years versus >38 years.
A negative value indicates a lower observed percentage in the SCH 900962 group on this scale. The primary estimate was −3.0 percentage points.
Non-inferiority framework
The primary hypothesis was non-inferiority or equivalence, with a pre-defined non-inferiority margin of −8%. The relevant question is therefore not simply whether the estimated difference is above or below zero. The key question is whether the uncertainty interval remains above the pre-specified boundary of −8%.
For the primary analysis, the lower confidence-limit was −7.4%, which is above the −8% margin.
ANOVA for number of oocytes retrieved
The number of oocytes retrieved per attempt was analyzed using ANOVA. The analysis used the Intent-To-Treat Group and set the number of oocytes retrieved to zero for participants who did not have oocyte retrieval.
The reported mean difference was adjusted for age class and center. This combines a continuous endpoint with covariate adjustment while retaining the randomized treatment comparison as the primary grouping factor.
Fisher exact test for categorical safety outcomes
The number of participants with moderate or severe OHSS and the number who cancelled the cycle because of a serious adverse event were analyzed with Fisher exact tests in the All-Subjects-Treated Group.
Fisher's exact test is particularly useful for categorical comparisons when event counts may be small because it calculates the exact probability of the observed allocation of events under the null hypothesis rather than relying on a large-sample approximation.
Covariate adjustment and stratification
Covariate adjustment appears explicitly in the reported analyses. The primary vital-pregnancy comparison was adjusted for age class, while the oocyte analysis was adjusted for age class and center. The registry also identifies stratified analysis as a concept in the analysis text.
7. Primary Result: Vital Pregnancy
The primary endpoint was the percentage of participants with a vital pregnancy, assessed by ultrasound at least 35 days after embryo transfer within the specified 35–42 day window. The primary comparison was performed in the Intent-To-Treat Group.
Difference in percentage of participants with a vital pregnancy
95% CI: −7.4 to 1.4 · Two-sided 95% CI
Pre-defined non-inferiority margin: −8%
| Primary endpoint | Estimate | 95% CI | Hypothesis | Adjustment |
|---|---|---|---|---|
| Percentage of Participants With a Vital Pregnancy | −3.0 | −7.4 to 1.4 | Non-inferiority or equivalence | Age class: ≤38 yrs vs. >38 yrs |
The estimated difference of −3.0 percentage points means that the estimated percentage of participants with a vital pregnancy was 3.0 percentage points lower with the single SCH 900962 injection than with daily 300 IU recFSH on the reported risk-difference scale.
This estimate does not mean that SCH 900962 reduced the chance of pregnancy by 3.0% for every individual participant. It is a population-level treatment-group difference, adjusted for the specified age-class stratification.
The two-sided 95% confidence interval extends from −7.4 to 1.4 percentage points. Because the lower confidence limit of −7.4% remains above the pre-defined non-inferiority margin of −8%, the reported interval is compatible with the prespecified non-inferiority criterion.
The interval also includes 0. That is important because non-inferiority and superiority are different questions. Inclusion of 0 means the interval is compatible with no difference as well as with differences in either direction; it does not prevent the non-inferiority conclusion when the relevant non-inferiority boundary is −8%.
No p-value is needed to interpret the central non-inferiority question in the same way as a conventional superiority test. For a non-inferiority analysis, the relationship between the confidence interval and the pre-specified margin is central to interpretation.
8. Why the Non-Inferiority Margin Matters
The non-inferiority margin defines how much lower the new treatment's effect could be, at the relevant confidence level, while still satisfying the trial's pre-specified non-inferiority criterion. Here, that margin was −8% for the percentage of participants with a vital pregnancy.
The critical comparison is between the lower confidence limit and the margin. The lower limit is −7.4%, while the non-inferiority margin is −8%. Thus the entire reported 95% confidence interval lies above the non-inferiority boundary.
9. Secondary Efficacy Results
Number of Oocytes Retrieved Per Attempt
Mean difference in number of oocytes retrieved
95% CI: −0.2 to 1.2 · Two-sided 95% CI
Pre-defined non-inferiority margin: −3 oocytes
The analysis used ANOVA in the Intent-To-Treat Group. Participants who did not have oocyte retrieval were assigned a value of zero for this endpoint. The estimated difference was adjusted for age class and center.
The estimated mean difference was 0.5 oocytes, meaning the estimated average number of oocytes retrieved was 0.5 higher in the SCH 900962 group on the reported scale. The 95% confidence interval ranged from −0.2 to 1.2 oocytes.
The confidence interval includes 0, so the reported interval is compatible with no mean difference as well as with modest differences in either direction. For the non-inferiority question, however, the lower limit of −0.2 is well above the pre-defined margin of −3 oocytes.
Live Birth Rate
Difference in live birth rates
95% CI: −6.5 to 1.9 · Two-sided 95% CI
Pre-defined non-inferiority margin: −8%
Live birth rate was assessed approximately nine months after embryo transfer. The Intent-To-Treat Group was used, with participants analyzed according to randomized treatment assignment. The reported analysis used a generalized linear model and adjusted for age class.
The estimated difference of −2.3 percentage points corresponds to an estimated lower live birth rate in the SCH 900962 group on the reported difference scale. The 95% confidence interval extends from −6.5 to 1.9 percentage points.
The interval includes 0, so it does not establish a difference in either direction on a conventional two-sided confidence-interval interpretation. Its lower limit of −6.5% nevertheless remains above the pre-defined −8% non-inferiority margin.
10. Secondary Safety Results
Moderate or Severe Ovarian Hyperstimulation Syndrome
| Endpoint | Analysis | Hypothesis | P-value |
|---|---|---|---|
| Number of Participants With Moderate or Severe OHSS | Fisher exact test | Superiority | 0.30 |
| Time frame | Up to approximately 1 month after oocyte pick-up | ||
This endpoint was analyzed in the All-Subjects-Treated Group, with participants grouped according to the active treatment they actually received. The registry reports a Fisher exact test and a p-value of 0.30.
A p-value of 0.30 does not quantify the size of any treatment difference. It describes the statistical evidence against the specified null hypothesis under the test framework. The registry does not provide an effect estimate or confidence interval for this endpoint in the posted statistical analysis.
Cycle Cancellation Due to a (Serious) Adverse Event
| Endpoint | Analysis | Hypothesis | P-value |
|---|---|---|---|
| Number of Participants Who Cancelled the Cycle Due to a (Serious) Adverse Event | Fisher exact test | Superiority | >0.999 |
| Time frame | Up to time of embryo transfer, maximum of 24 days after start of study drug | ||
This analysis also used the All-Subjects-Treated Group and Fisher exact testing. The registry reports a p-value of >0.999. As with the OHSS analysis, the p-value is evidence under the specified test and is not itself an estimate of the magnitude of a treatment difference.
11. Serious Adverse Events by Arm
The registry reports serious adverse-event counts by treatment arm in two entries. The first comparison reports 4 affected participants among 692 at risk with the single 150 µg SCH 900962/MK treatment and 19 among 698 at risk with daily 300 IU recFSH. A separate follow-up entry reports 110 among 154 and 112 among 167, respectively.
| Treatment | Serious adverse events: affected / at risk |
|---|---|
| Single Injection of 150 µg SCH 900962/MK | 4 / 692 |
| Daily 300 IU recFSH | 19 / 698 |
| Single Injection of 150 µg SCH 900962/MK — follow-up entry | 110 / 154 |
| Daily 300 IU recFSH — Follow-up | 112 / 167 |
12. Statistical Methods Explained
Why was a generalized linear model used for vital pregnancy?
Vital pregnancy is a binary outcome: each participant is classified according to whether the endpoint occurred. The registry reports a generalized linear model for this endpoint and expresses the treatment effect as a difference in percentage of participants with a vital pregnancy. The model also incorporated age class as a stratification-related adjustment.
What does a risk difference of −3.0 mean?
A risk difference of −3.0 means that the estimated percentage of participants experiencing the endpoint was 3.0 percentage points lower in the SCH 900962 group than in the recFSH group on the reported scale. It is an absolute difference between groups, not a relative risk or a hazard ratio.
Why is the −8% margin more important than whether the estimate is below zero?
In a non-inferiority trial, a modest negative estimate can still satisfy the trial's non-inferiority criterion if the uncertainty interval excludes effects worse than the pre-defined margin. Here, the estimate was −3.0%, but the relevant lower confidence limit was −7.4%, which remained above the −8% boundary.
Why does the confidence interval include zero?
The interval from −7.4 to 1.4 includes zero because the data are compatible with a range of treatment differences that includes no difference. That does not by itself resolve the non-inferiority question, because non-inferiority is assessed against the −8% margin rather than against zero.
Why was ANOVA used for number of oocytes retrieved?
The number of oocytes retrieved is a quantitative endpoint. The registry reports ANOVA for the comparison and states that the estimated mean difference was adjusted for age class and center. Participants without oocyte retrieval were assigned zero, allowing the analysis to retain those participants in the stated Intent-To-Treat framework.
Why was Fisher exact testing used for the safety endpoints?
The OHSS and cycle-cancellation endpoints are binary categorical outcomes. Fisher exact testing provides an exact categorical comparison and is appropriate when event counts may be relatively small. The registry reports Fisher exact testing for both endpoints.
What does the p-value of 0.30 for OHSS tell us?
The reported p-value of 0.30 describes the statistical evidence from the Fisher exact test under its null hypothesis. It does not tell us the size of the treatment difference, the clinical importance of OHSS, or the probability that either treatment is safe.
13. Primary Endpoint Interpretation in Detail
The primary estimate was a −3.0 percentage-point difference in vital pregnancy. On the reported scale, the estimate favors the recFSH group numerically, but the magnitude of that estimated difference is smaller than the −8% non-inferiority margin.
The 95% confidence interval was −7.4 to 1.4 percentage points. The interval quantifies uncertainty around the estimated population-level treatment difference under the analysis framework. It does not describe the range of responses that an individual participant might experience.
The lower confidence limit of −7.4% lies above the pre-defined −8% margin. Therefore, the reported confidence interval is consistent with the trial's non-inferiority criterion.
The confidence interval crosses zero, so the primary result should not be described as evidence that SCH 900962 produces a higher vital-pregnancy percentage than daily recFSH. Non-inferiority and superiority address different hypotheses.
The primary analysis used the Intent-To-Treat Group and classified participants without embryo transfer or those lost before the confirming ultrasound as non-pregnant. This endpoint handling affects the estimand because missing confirmation was explicitly incorporated into the outcome definition used for analysis.
14. Relationship Between the Primary and Secondary Endpoints
| Endpoint | Effect measure | Estimate | 95% CI | Non-inferiority margin / test |
|---|---|---|---|---|
| Vital pregnancy | Risk difference | −3.0 | −7.4 to 1.4 | −8% |
| Number of oocytes retrieved | Mean difference | 0.5 | −0.2 to 1.2 | −3 oocytes |
| Live birth rate | Other difference | −2.3 | −6.5 to 1.9 | −8% |
| Moderate or severe OHSS | Fisher exact test | Not reported | Not reported | Superiority; P = 0.30 |
| Cycle cancellation due to a (serious) adverse event | Fisher exact test | Not reported | Not reported | Superiority; P >0.999 |
The endpoint-specific analyses use different estimands because the outcomes are different. Vital pregnancy and live birth are expressed as differences in percentages, oocyte retrieval as a mean difference, and the two categorical safety outcomes through Fisher exact testing. A p-value from one endpoint should therefore not be compared directly with an effect estimate from another endpoint.
15. Missing Data and Endpoint Assignment
The registry provides an explicit rule for the primary vital-pregnancy endpoint: participants who did not have embryo transfer or who were lost to follow-up before the ultrasound assessment confirming vital pregnancy were counted as non-pregnant.
This is statistically important because missingness is not being handled simply by excluding participants from the analysis. Instead, the endpoint definition assigns a non-pregnant status to these participants within the stated Intent-To-Treat analysis.
The oocyte endpoint similarly specifies that participants who did not have oocyte retrieval were assigned a value of zero. These rules preserve the stated analysis population while incorporating specified treatment-cycle outcomes into the endpoint itself.
16. Covariate Adjustment
The primary vital-pregnancy analysis adjusted for age class, defined as ≤38 years versus >38 years. This adjustment was tied to the stratified analysis described in the registry.
The number-of-oocytes analysis included two adjustment variables: age class and center. The registry therefore reports a different adjustment structure for the continuous oocyte endpoint than for the primary vital-pregnancy endpoint.
Age class
The primary vital-pregnancy estimate was adjusted for the two registered age classes: ≤38 years and >38 years.
Center
The oocyte-retrieval analysis was additionally adjusted for center.
Covariate adjustment can improve precision and account for pre-specified sources of variation. It does not turn an observational comparison into a randomized comparison; here, randomization provides the principal design basis for the treatment comparison.
17. Why the Trial's Masking Design Matters
The study was quadruple-masked and included placebo interventions for both SCH 900962 and recFSH. Masking is particularly relevant when treatment administration differs, because knowledge of assignment could otherwise influence participant behavior, investigator management, or assessment processes.
For a binary reproductive endpoint, masking can help reduce the possibility that post-randomization expectations affect clinical management or assessment. The value of masking is therefore complementary to randomization: randomization addresses allocation, while masking addresses knowledge of allocation.
18. Limitations and Interpretation Issues
- Non-inferiority margin: the interpretation depends on the pre-defined −8% margin for vital pregnancy and the −3 oocyte margin for oocyte retrieval. A non-inferiority conclusion is meaningful relative to the clinical and statistical justification for those margins.
- Confidence intervals: the primary confidence interval includes zero. Therefore, the analysis supports the non-inferiority comparison against −8%, but it does not establish superiority of SCH 900962.
- Endpoint assignment: participants without embryo transfer or without confirmation of vital pregnancy were counted as non-pregnant, while participants without oocyte retrieval were assigned zero oocytes. These rules are integral to interpretation of the reported estimates.
- Different analysis populations: efficacy analyses used the Intent-To-Treat Group, whereas the OHSS and cycle-cancellation analyses used the All-Subjects-Treated Group.
- Different treatment assignment rules: the ITT efficacy analyses grouped participants according to randomized treatment, while the safety-related analyses grouped participants according to active treatment actually received.
- Limited effect reporting for categorical safety endpoints: the posted statistical analyses provide p-values for OHSS and cycle cancellation but do not provide corresponding effect estimates or confidence intervals.
- Endpoint-specific methods: the trial uses different statistical methods because its endpoints have different data structures. Results should therefore be interpreted within the framework appropriate to each endpoint rather than as a single summary measure.
19. Why This Trial Matters Statistically
PURSUE is a useful teaching case because the central result illustrates an important distinction in clinical-trial statistics: non-inferiority is judged against a pre-specified clinically relevant margin rather than simply against zero.
| Concept | How it appears in PURSUE |
|---|---|
| Randomization | Participants were randomized to a single SCH 900962 injection or daily recFSH. |
| Parallel-group design | The registry identifies the design model as parallel. |
| Quadruple masking | The registry identifies quadruple masking and placebo interventions for both treatment schedules. |
| Intention-to-treat analysis | The primary efficacy analysis used all randomized participants who received at least one dose. |
| Risk difference | The primary effect measure was the difference in percentage of participants with a vital pregnancy. |
| Confidence interval | The primary estimate was accompanied by a two-sided 95% confidence interval. |
| Non-inferiority | The primary endpoint used a pre-defined −8% non-inferiority margin. |
| Covariate adjustment | Age class was used for the primary endpoint; age class and center were used for oocyte retrieval. |
| ANOVA | Number of oocytes retrieved was analyzed using ANOVA. |
| Fisher exact test | Moderate/severe OHSS and cycle cancellation due to a serious adverse event were analyzed with Fisher exact tests. |
| Missing/outcome assignment | Participants without embryo transfer or confirming ultrasound were counted as non-pregnant; no-retrieval participants were assigned zero oocytes. |
20. Statistical Takeaways
1. Read the margin first
For the primary non-inferiority analysis, the −8% margin defines the relevant boundary. The lower 95% confidence limit was −7.4%, above that boundary.
2. Do not confuse non-inferiority with superiority
The primary confidence interval includes zero. The result therefore should not be interpreted as establishing a higher vital-pregnancy percentage with SCH 900962.
3. Absolute differences are directly interpretable
The primary estimate is expressed in percentage points, making its scale different from a relative measure such as a risk ratio or hazard ratio.
4. Analysis populations matter
Efficacy analyses used randomized treatment assignment, while the reported safety analyses used active treatment actually received.
21. Related Tutorials
Learn more about the methods used in this trial:
22. Related Statistical Calculators
23. Sources
- ClinicalTrials.gov: NCT01144416.
- PubMed record: PMID 26003273. View PubMed record.
- PubMed record: PMID 27619773. View PubMed record.
- PubMed record: PMID 26991902. View PubMed record.
- PubMed record: PMID 26520396. View PubMed record.
Continue through the Clinical Biostats statistical library
Connect the methods used in PURSUE with deeper tutorials, statistical calculators, and additional clinical-trial analyses.
24. Record Summary
PURSUE provides a clear example of how a randomized phase 3 non-inferiority trial should be interpreted statistically. The primary endpoint was a binary reproductive outcome analyzed as a risk difference, with age-class adjustment and a pre-defined non-inferiority margin of −8%. The estimated difference in vital pregnancy was −3.0 percentage points with a two-sided 95% confidence interval of −7.4 to 1.4 percentage points. Because the lower confidence limit remained above the −8% margin, the reported interval is compatible with the prespecified non-inferiority criterion, while inclusion of zero means that superiority of SCH 900962 was not established by the reported interval.
The secondary analyses illustrate additional statistical structures: ANOVA for the number of oocytes retrieved, generalized linear modeling for live birth rate, and Fisher exact testing for categorical safety outcomes. Together, the results show why the interpretation of a clinical trial depends not only on the numerical estimate, but also on the endpoint definition, analysis population, covariate adjustment, confidence interval, and hypothesis framework.