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
SOLO3 was a randomized, open-label phase 3 treatment trial evaluating olaparib against single-agent chemotherapy in patients with relapsed ovarian cancer, BRCA mutation, and platinum sensitivity. The registry reports 266 participants, two treatment arms, a binary primary endpoint, and 17 posted statistical analyses.
| Feature | SOLO3 |
|---|---|
| Trial name | SOLO3 |
| ClinicalTrials.gov identifier | NCT02282020 |
| Phase | Phase 3 |
| Status | Completed |
| Therapeutic area | Oncology |
| Conditions | Relapsed Ovarian Cancer, BRCA Mutation, Platinum Sensitivity |
| Allocation | Randomized |
| Masking | None |
| Primary purpose | Treatment |
| Enrollment | 266 |
| Arms | 2 |
| Interventions | Olaparib; Single agent chemotherapy |
| Primary endpoint type | Binary |
| Results posted | Yes |
| Outcome measures posted | 21 |
| Statistical analyses posted | 17 |
| Lead sponsor | AstraZeneca |
2. Clinical Question
The central statistical question was whether treatment with olaparib produced a higher Objective Response Rate (ORR) than physician's choice of single-agent chemotherapy in the registered population of relapsed ovarian cancer patients with BRCA mutation and platinum sensitivity.
Population
Relapsed ovarian cancer patients with germline breast cancer susceptibility gene (BRCA) mutations who had progressed at least 6 months after last platinum treatment and had received at least 2 prior platinum treatments.
Intervention
Olaparib 300 mg twice daily.
Comparator
Single-agent chemotherapy, with analyses also referring to selected chemotherapy for specified endpoints.
Primary question
Does olaparib increase the odds of objective response relative to single-agent chemotherapy?
3. Trial Design
Olaparib 300 mg BID
- Olaparib administered at 300 mg twice daily.
- Randomized treatment comparison against single-agent chemotherapy.
- Registry analyses refer to this arm as olaparib 300 mg BID.
Single-agent chemotherapy
- Single-agent chemotherapy comparator.
- Some registry analyses identify the comparator as selected chemotherapy.
- Safety results are reported for this arm as a separate treatment group.
4. Trial Timeline
Trial start
The registry lists 2015-02-06 as the study start date.
Primary completion
The registry lists 2018-10-10 as the primary completion date.
Statistical results available
ClinicalTrials.gov reports 21 posted outcome measures and 17 posted statistical analyses, including one formal primary-endpoint analysis.
5. Primary Endpoint
| Endpoint | Registered definition / time frame | Analysis |
|---|---|---|
| Objective Response Rate (ORR) | RECIST follow-up assessments performed every 8 weeks (±1 week), up to 48 weeks, then every 12 weeks (±1 week) from randomisation. RECIST 1.1 criteria were used by a Blinded Independent Central Review (BICR) to assess participant response. The registry definition states that ORR is the number of participants with Complete Response (CR) or Partial Response (PR) in the Measurable Disease Analysis Set (MDAS). Complete response is declared when all lesions have disappeared or all lesions have disappeared and all nodal disease is \< 10 mm each. Partial response is declared when there is a decrease in sum of diameters of target lesions ≥ 30%. | Logistic regression |
6. Primary Result: Objective Response Rate
The registry reports a formal superiority analysis for ORR comparing olaparib 300 mg BID with single-agent chemotherapy. Logistic regression was used, with an adjustment factor for time to disease progression after the end of last platinum-based chemotherapy: 6–12 months versus >12 months.
Odds ratio for objective response
95% CI: 1.40–4.58 · P = 0.002
Two-sided superiority analysis; OR > 1 favours the olaparib arm.
| Primary endpoint | Population | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|---|
| Objective Response Rate (ORR) | Measurable Disease Analysis Set | Logistic regression | OR 2.53 | 1.40–4.58 | 0.002 |
An odds ratio of 2.53 means that the estimated odds of objective response were 2.53 times as high in the olaparib group as in the single-agent chemotherapy group under the reported logistic-regression model. Equivalently, the estimated odds were approximately 153% higher; this is a statement about odds, not about a 153% increase in response probability.
The odds ratio does not tell us the absolute response probability in either arm, the number of patients who responded, or the duration of response. Those quantities require corresponding response-rate or duration data.
The 95% confidence interval of 1.40–4.58 describes statistical uncertainty around the estimated odds ratio under the analysis framework. It is not an interval containing the individual treatment effects experienced by patients.
The p-value of 0.002 quantifies the evidence against the null hypothesis specified for this statistical comparison. It does not measure the magnitude or clinical importance of the treatment effect. The effect size is conveyed by the odds ratio and its confidence interval.
The analysis was performed in the MDAS rather than simply using every enrolled participant, because the registry defines this population around measurable disease at baseline. That distinction matters when considering what population the primary estimate directly describes.
7. Statistical Methodology
Logistic regression for the primary endpoint
ORR is a binary endpoint: each participant is classified according to whether the defined objective-response outcome occurred. Logistic regression models the probability of that binary outcome and expresses the treatment comparison through an odds ratio.
The exponentiated treatment coefficient, exp(β1), is interpreted as an odds ratio. In SOLO3, the reported model included a factor for the time to disease progression after the end of last platinum-based chemotherapy.
Odds ratio
The odds of response are defined as the probability of response divided by the probability of no response. The odds ratio compares those odds between the two treatment groups.
The registry reports OR = 2.53 and explicitly notes that an odds ratio greater than 1 favours the olaparib arm.
Log-rank testing for time-to-event endpoints
The registry reports log-rank tests for PFS, PFS2, OS, time to earliest progression or death, time to first subsequent therapy or death, time to second subsequent therapy or death, and time to treatment discontinuation or death. These endpoints are time-to-event outcomes, for which follow-up can be censored when the event has not occurred by the relevant observation time.
Mixed-effects modelling for longitudinal TOI
The change in Trial Outcome Index (TOI) score from baseline to Week 48 was analysed using a mixed-effects model. The reported model included patient, treatment, visit, treatment-by-visit interaction, baseline TOI score, and baseline TOI score-by-visit interaction.
Intention-to-treat principle
The registry descriptions of the Full Analysis Set state that it includes all randomized participants and compares treatment groups on the basis of randomized treatment, regardless of the treatment actually received. This preserves the treatment assignment created by randomization for the reported efficacy comparisons.
8. Secondary Efficacy Results
The registry contains a broad set of secondary analyses. Most are time-to-event endpoints analysed with the log-rank test and expressed as hazard ratios. The following results are reproduced exactly from the ClinicalTrials.gov record.
| Secondary endpoint | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|
| Progression Free Survival (PFS) | Log-rank test | HR 0.62 | 0.43–0.91 | 0.013 |
| Time From Randomisation to Second Progression (PFS2) | Log-rank test | HR 0.80 | 0.56–1.15 | 0.229 |
| Overall Survival (OS) | Log-rank test | HR 1.07 | 0.76–1.49 | 0.714 |
| Time to Earliest Progression by RECIST 1.1 or CA-125 or Death | Log-rank test | HR 0.59 | 0.41–0.85 | 0.005 |
| Time From Randomization to First Subsequent Therapy or Death (TFST) | Log-rank test | HR 0.49 | 0.35–0.69 | <0.001 |
| Time From Randomization to Second Subsequent Therapy or Death (TSST) | Log-rank test | HR 0.75 | 0.53–1.05 | 0.089 |
| Time From Randomization to Study Treatment Discontinuation or Death (TDT) | Log-rank test | HR 0.20 | 0.14–0.29 | <0.001 |
Progression Free Survival
PFS hazard ratio
95% CI: 0.43–0.91 · P = 0.013
A hazard ratio below 1 favours the olaparib arm.
The PFS hazard ratio of 0.62 indicates that the estimated instantaneous rate of progression or death was lower in the olaparib group than in the chemotherapy group under the reported time-to-event comparison. As a relative measure, it does not mean that 38% of patients avoided progression or that every patient experienced exactly a 38% reduction in risk.
The 95% CI of 0.43–0.91 represents uncertainty around the estimated hazard ratio. It does not provide a range of individual patient outcomes.
The p-value of 0.013 is evidence against the relevant null comparison under the reported statistical framework. It is not a measure of the size of the treatment effect; the HR and its confidence interval provide that information.
Because this is a time-to-event analysis, censoring and the shape of the event-time distributions matter. A single hazard ratio is a compact summary of a time-to-event comparison rather than a direct statement about absolute differences at every follow-up time.
Time From Randomisation to Second Progression
PFS2 hazard ratio
95% CI: 0.56–1.15 · P = 0.229
A hazard ratio below 1 favours the olaparib arm.
PFS2 extends the disease-control question beyond the first progression. The reported estimate is below 1, but the confidence interval includes 1.00. The registry's analysis is therefore best described directly through its estimate, interval, and p-value rather than treating the p-value as a measure of the magnitude of effect.
Overall Survival
OS hazard ratio
95% CI: 0.76–1.49 · P = 0.714
The registry analysis notes that an HR < 1 favours the olaparib arm.
The reported OS estimate is close to 1.00, with a 95% confidence interval extending below and above 1.00. This means the registry result does not provide a precise estimate of a directional OS difference from the reported comparison alone. Importantly, the OS hazard ratio should not be inferred from the PFS or response results.
Earliest Progression, CA-125 or Death
Time-to-earliest-event hazard ratio
95% CI: 0.41–0.85 · P = 0.005
The endpoint incorporates progression by RECIST 1.1, CA-125 progression, or death.
This composite time-to-event endpoint differs from PFS because its event definition includes progression by either RECIST 1.1 or CA-125, as well as death. The reported HR of 0.59 therefore describes the comparison for this particular event definition rather than automatically representing the treatment effect for every progression endpoint.
First Subsequent Therapy or Death
TFST hazard ratio
95% CI: 0.35–0.69 · P < 0.001
TFST captures the time from randomization to either first subsequent therapy or death. A lower hazard for this endpoint indicates a lower estimated rate of reaching that composite event under the reported analysis.
Second Subsequent Therapy or Death
TSST hazard ratio
95% CI: 0.53–1.05 · P = 0.089
TSST moves the treatment-pathway endpoint one step farther beyond the randomized treatment. Its interpretation therefore concerns time until second subsequent therapy or death, not simply tumor response or first progression.
Study Treatment Discontinuation or Death
TDT hazard ratio
95% CI: 0.14–0.29 · P < 0.001
The TDT endpoint concerns discontinuation of study treatment or death. The reported HR of 0.20 indicates a substantially lower estimated instantaneous rate of this composite event in the olaparib arm under the registry analysis. It should not be interpreted as an overall survival hazard ratio.
9. BRCA Gene Population Analyses
The registry also reports analyses specifically labelled as being in the BRCA gene population. These analyses largely reproduce the same statistical framework while using the corresponding BRCA-population outcome definitions.
| BRCA-population endpoint | Method | Effect | 95% CI | P-value |
|---|---|---|---|---|
| Objective Response Rate by BICR | Logistic regression | OR 2.40 | 1.32–4.39 | 0.004 |
| Participants with disease progression or death by BICR | Log-rank test | HR 0.62 | 0.42–0.91 | 0.014 |
| Second progression or death (PFS2) | Log-rank test | HR 0.80 | 0.56–1.14 | 0.213 |
| Overall Survival | Log-rank test | HR 1.07 | 0.76–1.51 | 0.699 |
| Discontinued study treatment or died | Log-rank test | HR 0.18 | 0.12–0.27 | <0.001 |
| Received subsequent chemotherapy or died | Log-rank test | HR 0.46 | 0.33–0.66 | <0.001 |
| Received second subsequent chemotherapy or died | Log-rank test | HR 0.72 | 0.51–1.01 | 0.055 |
BRCA-population response
Objective response odds ratio
95% CI: 1.32–4.39 · P = 0.004
An odds ratio greater than 1 favours the olaparib arm.
This estimate is similar in structure to the primary ORR analysis but represents the registry's specifically labelled BRCA-gene-population endpoint. The distinction between analysis populations is important: estimates should be interpreted within the population defined for that endpoint rather than treated as interchangeable results.
BRCA-population progression or death
Hazard ratio
95% CI: 0.42–0.91 · P = 0.014
BRCA-population PFS2
Hazard ratio
95% CI: 0.56–1.14 · P = 0.213
BRCA-population overall survival
Hazard ratio
95% CI: 0.76–1.51 · P = 0.699
The BRCA-population OS result has an estimate of 1.07 and a confidence interval of 0.76–1.51. As with the overall OS analysis, the interval spans 1.00, so the reported estimate should be read as an uncertain relative hazard rather than as evidence of a specific directional effect.
BRCA-population treatment discontinuation or death
Hazard ratio
95% CI: 0.12–0.27 · P < 0.001
BRCA-population subsequent chemotherapy or death
Hazard ratio
95% CI: 0.33–0.66 · P < 0.001
BRCA-population second subsequent chemotherapy or death
Hazard ratio
95% CI: 0.51–1.01 · P = 0.055
10. Patient-Reported Outcome: Trial Outcome Index
The registry includes a longitudinal analysis of mean change from baseline in Trial Outcome Index (TOI) score from Day 1 to Week 48 (±1 week), with a data cutoff of 10 October 2018.
Mean difference in final TOI values
95% CI: −0.5–5.5 · P = 0.108
Mixed-effects model; olaparib 300 mg BID versus selected chemotherapy.
The reported mean difference of 2.5 is the model-based difference in final TOI values between the treatment groups under the specified mixed-effects analysis. It is not a hazard ratio or odds ratio and should be interpreted on the TOI score scale.
The 95% CI of −0.5–5.5 spans zero, indicating uncertainty about the direction and magnitude of the between-group difference under this analysis.
The p-value of 0.108 describes evidence against the corresponding null comparison; it does not indicate that the treatment effect is exactly zero, nor does it measure the clinical importance of a TOI difference.
The model is longitudinal and includes patient, treatment, visit, treatment-by-visit interaction, baseline TOI score, and baseline TOI score-by-visit interaction. This allows the analysis to use the repeated-measure structure rather than treating the Week 48 observation as an isolated measurement.
Participants with an improvement in TOI score
Odds ratio for TOI improvement
95% CI: 0.88–6.86 · P = 0.092
Logistic regression; an odds ratio greater than 1 favours the olaparib arm.
This binary endpoint asks a different question from the continuous mean-change analysis. Instead of estimating the average difference in score, it compares the odds of meeting the registry's definition of improvement. The confidence interval is comparatively broad and crosses 1.00.
11. How the Statistical Analyses Fit Together
| Question | Endpoint type | Reported method | Effect measure |
|---|---|---|---|
| Did patients achieve objective response? | Binary | Logistic regression | Odds ratio |
| How long until progression or death? | Time-to-event | Log-rank test | Hazard ratio |
| How long until second progression or death? | Time-to-event | Log-rank test | Hazard ratio |
| How long until death? | Time-to-event | Log-rank test | Hazard ratio |
| How long until first subsequent therapy or death? | Time-to-event | Log-rank test | Hazard ratio |
| How long until treatment discontinuation or death? | Time-to-event | Log-rank test | Hazard ratio |
| Did TOI improve longitudinally? | Continuous | Mixed-effects model | Mean difference |
| How many participants improved in TOI? | Binary | Logistic regression | Odds ratio |
This is statistically coherent: the endpoint type drives the choice of model and effect measure. Binary outcomes naturally lead to odds ratios under logistic regression, while censored time-to-event outcomes are summarized with hazard ratios from survival-analysis comparisons. Repeated continuous measurements require a model that accounts for their longitudinal structure.
12. Statistical Methods Explained
Why was logistic regression used for ORR?
ORR is registered as a binary endpoint. Each participant contributes a response status rather than a continuous score or an event time. Logistic regression is therefore appropriate for modelling the probability of response and comparing the treatment groups through an odds ratio. In SOLO3, the reported model additionally included the time-to-progression factor after the last platinum-based chemotherapy.
What does an odds ratio of 2.53 mean?
An OR of 2.53 means that the estimated odds of response were 2.53 times as high with olaparib as with single-agent chemotherapy under the fitted model. Odds are not probabilities. For example, an odds ratio cannot be read as saying that the response probability increased by 153 percentage points or that the response rate was 2.53 times higher.
Why is the confidence interval important?
The 95% CI of 1.40–4.58 gives a range of values representing statistical uncertainty around the estimated odds ratio under the analysis framework. It conveys considerably more information than the point estimate alone. A wide interval indicates less precision than a narrow interval; it does not describe variation in treatment effect from one patient to another.
What does a hazard ratio of 0.62 mean?
For the PFS analysis, HR 0.62 means that the estimated instantaneous rate of progression or death was 62% of the corresponding rate in the chemotherapy group under the reported time-to-event analysis. The complement, 1 − 0.62 = 0.38, gives the derived statement that the estimated hazard was approximately 38% lower. This does not mean that 38% of patients were protected from progression or that each patient experienced a 38% reduction.
Why are PFS and OS not interchangeable?
PFS and OS have different event definitions. PFS concerns progression or death, whereas OS concerns death. A treatment can therefore produce different hazard-ratio estimates for the two endpoints. In SOLO3, the reported PFS HR was 0.62 while the reported OS HR was 1.07. Those estimates answer different questions and should not be averaged or interpreted as contradictory measurements of one endpoint.
Why was a mixed-effects model used for TOI?
The TOI endpoint was measured longitudinally from baseline through Week 48. The registry reports a mixed-effects model containing patient, treatment, visit, treatment-by-visit interaction, baseline TOI, and baseline TOI-by-visit interaction. Such a structure is designed to represent repeated observations over time and to distinguish treatment effects from changes associated with visit and baseline score.
13. Understanding the Primary ORR Result
Effect size
OR 2.53 indicates a 2.53-fold estimated odds comparison for objective response in favour of olaparib under the reported logistic model.
Precision
The 95% CI is 1.40–4.58. The interval expresses uncertainty around the estimated odds ratio.
Evidence
The reported two-sided p-value is 0.002. The p-value addresses the null hypothesis; it does not quantify effect size.
Population
The primary analysis was conducted in the MDAS, which the registry defines around baseline measurable disease.
The primary result is therefore best understood as a three-part statement: the estimated treatment effect was an odds ratio of 2.53, statistical uncertainty was represented by a 95% CI of 1.40–4.58, and the reported evidence against the null comparison was summarized by P = 0.002. None of these quantities alone describes absolute benefit.
14. Safety Results
The ClinicalTrials.gov record reports serious adverse events by treatment arm. These are counts of affected participants over the corresponding at-risk populations.
| Safety measure | Olaparib 300 mg BID | Single-agent chemotherapy |
|---|---|---|
| Participants with serious adverse events / participants at risk | 46/178 | 14/76 |
The reported safety information is distinct from the efficacy analyses. The serious-adverse-event counts describe the number affected and the corresponding number at risk in each arm; they do not provide a formal between-group risk ratio, odds ratio, or p-value in the ClinicalTrials.gov record.
15. Time-to-Event Endpoints: A Statistical Reading
Seven of the registry-reported secondary analyses use the log-rank test and hazard ratios. This is a recurring pattern in clinical trials because follow-up does not necessarily end at the same time for every participant, and some participants may be censored before experiencing the endpoint.
For endpoints where the registry explicitly states that HR < 1 favours olaparib, an HR below 1 represents a lower estimated instantaneous event rate under the reported comparison. The meaning of the event still depends entirely on the endpoint definition.
This last point is important in SOLO3. An HR of 0.20 for treatment discontinuation or death cannot be interpreted in the same way as an HR of 0.20 for overall survival because the events are different. Similarly, TFST and TSST describe treatment-pathway events rather than direct measures of tumor response.
16. Endpoint Definitions and Time Frames
| Endpoint | Registry time frame | Type |
|---|---|---|
| Objective Response Rate (ORR) | RECIST follow-up assessments every 8 weeks (±1 week), up to 48 weeks, then every 12 weeks (±1 week) from randomisation | Binary |
| Progression Free Survival (PFS) | RECIST follow-up assessments every 8 weeks (±1 week), up to 48 weeks, then every 12 weeks (±1 week) from randomisation | Time-to-event |
| Time From Randomisation to Second Progression (PFS2) | Visits every 12 weeks from the date of first progression | Time-to-event |
| Overall Survival (OS) | Visits every 12 weeks from the date of first progression | Time-to-event |
| Time to Earliest Progression by RECIST 1.1 or CA-125 or Death | RECIST and CA-125 follow-up assessments every 8 weeks (±1 week), up to 48 weeks, then every 12 weeks (±1 week) | Time-to-event |
| Time From Randomization to First Subsequent Therapy or Death (TFST) | Anti-cancer treatments initiated post discontinuation of study treatment and investigator's opinion of response and date of progression recorded, assessed from date of first patient randomised to data cut off: 16Apr2021 (approx. 6 years 2 months) | Time-to-event |
| Time From Randomization to Second Subsequent Therapy or Death (TSST) | Anti-cancer treatments initiated post discontinuation of study treatment and investigator's opinion of response and date of progression recorded, assessed from date of first patient randomised to data cut off: 16Apr2021 (approx. 6 years 2 months) | Time-to-event |
| Time From Randomization to Study Treatment Discontinuation or Death (TDT) | Patients randomised to olaparib administer tablets orally at 300 mg twice daily and continue olaparib until the registry-defined treatment endpoint | Time-to-event |
| Mean Change From Baseline in TOI Score | Baseline (Day 1) to Week 48 (±1 week). DCO: 10Oct2018 | Continuous |
| Number of Participants Who Show an Improvement in TOI Score | Baseline (Day 1) to Week 48 (±1 week). DCO: 10Oct2018 | Binary |
17. Analysis Populations
The registry descriptions distinguish the Full Analysis Set from the Measurable Disease Analysis Set.
| Population | Registry description and statistical role |
|---|---|
| Full Analysis Set (FAS) | Includes all randomized participants and compares treatment groups on the basis of randomized treatment, regardless of the treatment actually received. The FAS is used for the reported time-to-event and longitudinal secondary analyses. |
| Measurable Disease Analysis Set (MDAS) | Includes participants in the FAS with measurable disease at baseline according to RECIST 1.1, determined using blinded independent central review. This is the population specified for the primary ORR analysis. |
This distinction is statistically meaningful. Randomization establishes comparability in the randomized population, while restricting an analysis to participants with baseline measurable disease creates a defined analysis subset. The resulting estimate should be interpreted as applying to that analysis population.
18. Regression Adjustment in the Primary Analysis
The primary logistic-regression model includes a factor for the time to disease progression after the end of last platinum-based chemotherapy, divided into 6–12 months versus >12 months.
Why include the factor?
Including a prespecified disease-history factor can account for differences associated with the timing of progression after the previous platinum-based treatment.
What it does not do
The adjustment does not turn the OR into a risk ratio, and it does not remove uncertainty from the estimate.
Effect measure remains an OR
The reported treatment effect is still an odds ratio from logistic regression, not an adjusted probability difference.
Interpretation
The OR of 2.53 is conditional on the reported model specification, including this progression-timing factor.
19. P-values, Confidence Intervals, and Effect Sizes
SOLO3 provides useful examples of why these three statistical quantities should be read together rather than substituted for one another.
| Endpoint | Estimate | 95% CI | P-value | What the estimate measures |
|---|---|---|---|---|
| Primary ORR | OR 2.53 | 1.40–4.58 | 0.002 | Relative odds of objective response |
| PFS | HR 0.62 | 0.43–0.91 | 0.013 | Relative instantaneous rate of progression or death |
| PFS2 | HR 0.80 | 0.56–1.15 | 0.229 | Relative instantaneous rate of second progression or death |
| OS | HR 1.07 | 0.76–1.49 | 0.714 | Relative instantaneous rate of death |
| TOI mean difference | 2.5 | −0.5–5.5 | 0.108 | Difference in final TOI values |
The p-value answers a narrow hypothesis-testing question. The confidence interval provides information about precision. The effect estimate describes the scale and direction of the comparison. For educational interpretation, all three are needed.
20. Limitations and Interpretation Issues
- Open-label design: the registry specifies no masking. This can matter particularly for endpoints involving treatment experience or assessments, although the primary ORR assessment used blinded independent central review.
- Primary analysis population: the primary ORR estimate comes from the MDAS rather than simply all enrolled participants. Interpretation should therefore remain tied to the registry-defined measurable-disease population.
- Odds versus probabilities: an odds ratio of 2.53 cannot be translated directly into a 2.53-fold response probability without knowing the underlying response probabilities.
- Time-to-event interpretation: hazard ratios summarize relative event rates over follow-up and are not equivalent to absolute risk differences or differences in median survival.
- Different event definitions: PFS, PFS2, OS, TFST, TSST, and TDT measure different events. Their hazard ratios should not be treated as interchangeable.
- Multiple analyses: the registry reports 17 statistical analyses. The ClinicalTrials.gov record does not specify a multiplicity-adjustment strategy for these posted analyses, so individual p-values should be interpreted in the context of the full analysis set rather than treated as isolated numbers.
- Incomplete absolute outcome information: the ClinicalTrials.gov record provides effect estimates, confidence intervals, and p-values for the reported analyses but do not provide median survival values or complete arm-level response counts.
- Safety scope: the ClinicalTrials.gov record contains serious-adverse-event counts by arm but not a formal comparative safety analysis.
21. Why This Trial Matters Statistically
SOLO3 is a useful teaching case because it connects several core clinical-trial methods within a single randomized phase 3 study. The primary endpoint is binary and analysed with logistic regression, while many secondary outcomes are time-to-event endpoints analysed with log-rank testing. The trial also includes a longitudinal patient-reported outcome analysed with a mixed-effects model.
| Concept | How it appears in SOLO3 |
|---|---|
| Randomization | Randomized phase 3 treatment comparison with 266 enrolled participants. |
| Intention-to-treat analysis | The FAS is defined using all randomized participants and randomized treatment assignment. |
| Binary endpoint analysis | ORR and TOI improvement are analysed using logistic regression. |
| Odds ratio | Primary ORR effect measure is OR 2.53 with 95% CI 1.40–4.58. |
| Time-to-event analysis | PFS, PFS2, OS, TFST, TSST, TDT, and other progression endpoints use log-rank testing and hazard ratios. |
| Longitudinal analysis | TOI mean change uses a mixed-effects model with treatment, visit, interactions, baseline score, and patient terms. |
| Confidence intervals | Reported for the primary and secondary effect estimates. |
| P-values | Reported alongside effect estimates for the statistical analyses. |
| Central review | The primary ORR assessment uses blinded independent central review and RECIST 1.1. |
| Analysis populations | The primary ORR analysis uses MDAS, while many secondary analyses use FAS. |
22. Overall Statistical Interpretation
The primary ORR analysis reported an odds ratio of 2.53 with a 95% CI of 1.40–4.58 and a two-sided p-value of 0.002. The analysis used logistic regression in the MDAS and adjusted for the registry-specified timing of progression after the last platinum-based chemotherapy.
The reported PFS HR was 0.62 (95% CI 0.43–0.91; P = 0.013), while the HR for the RECIST/CA-125/death endpoint was 0.59 (95% CI 0.41–0.85; P = 0.005). These estimates describe different time-to-event definitions and should not be collapsed into a single treatment effect.
The reported OS HR was 1.07 (95% CI 0.76–1.49; P = 0.714). The ClinicalTrials.gov record does not provide median OS or absolute survival estimates, so the OS analysis should be interpreted using the reported hazard ratio, confidence interval, and p-value without adding unreported quantities.
The mixed-effects analysis of mean TOI change reported a mean difference of 2.5 (95% CI −0.5–5.5; P = 0.108). The corresponding binary improvement analysis reported OR 2.24 (95% CI 0.88–6.86; P = 0.092). These are different estimands: one concerns a continuous score difference and the other the odds of improvement.
23. Related Tutorials
Learn more about the methods used in this trial:
24. Related Calculators
25. Sources
- ClinicalTrials.gov: SOLO3, NCT02282020.
- Linked publication: PubMed record — PMID 39668137.
- Linked publication: PubMed record — PMID 32530768.
- Linked publication: PubMed record — PMID 32530766.
- Linked publication: PubMed record — PMID 32073956.
Continue through the Clinical Biostats statistical tutorials
Explore the statistical methods that connect randomized trials, binary endpoints, survival analysis, regression, longitudinal models, confidence intervals, and effect measures.
26. Record Summary
SOLO3 provides a compact example of how different clinical-trial questions require different statistical estimands. The primary ORR endpoint is binary and was analysed with logistic regression, producing an odds ratio of 2.53 with a 95% confidence interval of 1.40–4.58 and a two-sided p-value of 0.002. The major secondary progression and survival endpoints use log-rank testing with hazard ratios, while the TOI endpoint uses a mixed-effects model for longitudinal continuous data and logistic regression for the binary improvement analysis.
The most important statistical distinction is between the effect measure and the endpoint being measured. An odds ratio describes relative odds of a binary outcome; a hazard ratio describes a relative event rate for a time-to-event endpoint; and a mean difference describes a difference on a continuous scale. These quantities cannot be substituted for one another simply because they are all reported as treatment effects.