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
DUET-2 was a phase 1 treatment study of XmAb20717 in subjects with selected advanced solid tumors. The registry describes a sequential, non-masked study with one intervention arm and an enrollment of 150 subjects.
| Feature | DUET-2 |
|---|---|
| Study acronym | DUET-2 |
| Phase | Phase 1 |
| Status | COMPLETED |
| Design model | SEQUENTIAL |
| Allocation | NA |
| Masking | NONE |
| Primary purpose | TREATMENT |
| Enrollment | 150 |
| Number of arms | 1 |
| Intervention | XmAb20717 (biological) |
| Lead sponsor | Xencor, Inc. |
| Sponsor type | INDUSTRY |
| Start date | 2018-07-10 |
| Primary completion date | 2022-06-01 |
| ClinicalTrials.gov | NCT03517488 |
2. Clinical Question
The registry describes DUET-2 as a treatment study of XmAb20717 in subjects with selected advanced solid tumors. The primary clinical question represented by the registered endpoint is whether treatment can be characterized with respect to safety and tolerability during the specified 56-day assessment period.
Population
Subjects with selected advanced solid tumors represented by the registry's listed conditions and tumor categories.
Intervention
XmAb20717, classified in the registry as a biological intervention.
Comparator
The registry identifies one intervention arm and does not identify a comparator arm.
Primary question
What is the safety and tolerability profile of XmAb20717, based on treatment-related adverse events assessed by CTCAE v4.03 over 56 days?
3. Trial Design
The combination of a phase 1 designation, one intervention arm, no masking, and a sequential design has direct statistical consequences. This is not a conventional parallel-group randomized comparison. There is no randomized control group in the registered design against which an efficacy or comparative safety effect can be estimated.
4. Disease and Eligibility Context
The registry lists a broad group of selected advanced solid tumors and related tumor categories. These include melanoma; breast carcinoma; hepatocellular carcinoma; urothelial carcinoma; squamous cell carcinoma of the head and neck; renal cell carcinoma; colorectal carcinoma; non-small cell lung carcinoma; gastric or gastroesophageal junction adenocarcinoma; endometrial carcinoma; mesothelioma; neuroendocrine carcinoma; cervical cancer; small cell lung carcinoma; squamous cell carcinoma of the anus; castration-resistant prostate carcinoma; nasopharyngeal carcinoma; cholangiocarcinoma; basal cell carcinoma; ovarian carcinoma; fallopian tube carcinoma; thymoma; thymic carcinoma; squamous cell carcinoma of the penis; vulvar carcinoma; solid tumors with published evidence of anti-tumor activity with anti-PD1/PDL1 and/or anti-CTLA4-directed therapy; malignant adnexal neoplasms; and non-squamous cell salivary gland carcinoma.
This breadth is statistically important. A population assembled across many tumor types can be useful for early-phase safety characterization, but the resulting population is heterogeneous with respect to disease biology, prognosis, prior treatment, and expected clinical course. Without a randomized comparator, these factors cannot be separated from treatment effects using a simple between-arm comparison.
5. Endpoints
| Endpoint | Time frame | Registry definition |
|---|---|---|
| Primary: Determine the safety and tolerability profile of XmAb20717 | 56 Days | Treatment-related adverse events as assessed by CTCAE v4.03 |
The registered primary endpoint is a safety and tolerability assessment, not a time-to-event efficacy endpoint. The endpoint is defined through treatment-related adverse events and their assessment under CTCAE v4.03 during the 56-day time frame.
6. Planned Analysis
No formal statistical analyses are posted to ClinicalTrials.gov for the registered primary endpoint. The registry reports the endpoint that was to be measured, but it does not provide posted statistical estimates, confidence intervals, p-values, or event counts for that endpoint.
What would normally be measured?
For a one-arm phase 1 safety study, the natural statistical description would focus on the frequency and severity of treatment-related adverse events. Events would typically be summarized using counts and percentages, with adverse-event severity classified according to the specified CTCAE framework.
Why a one-arm analysis is different
Because the registered study has one intervention arm and no comparator arm, the principal descriptive question is the observed safety experience among treated subjects. A treatment-versus-control hazard ratio, risk ratio, risk difference, or between-arm p-value is not part of the registered design.
For an adverse-event endpoint, a count and percentage can describe how commonly an event occurred. The interpretation depends on the analysis population, exposure, event definition, severity classification, and observation period.
Confidence intervals for one-arm event rates
If event counts were available, an adverse-event proportion could be accompanied by a confidence interval. For relatively uncommon events, exact binomial methods are often preferable to simple large-sample approximations because the sampling distribution can be asymmetric and bounded between 0 and 1.
Here, x is the number of subjects experiencing a specified event and n is the number of subjects in the relevant analysis population. A confidence interval quantifies uncertainty about the underlying event probability; it does not describe the range of outcomes for individual subjects.
The registry does not report the event counts needed to produce such numerical summaries. Consequently, the statistical interpretation remains at the level of the planned endpoint and the appropriate analysis framework rather than an observed treatment effect.
7. Statistical Methodology
Descriptive safety analysis
The registered primary endpoint is naturally suited to descriptive statistics. Treatment-related adverse events can be tabulated by event category and severity, with the number of affected subjects and the corresponding percentage reported for the relevant population.
CTCAE v4.03 assessment
The registry specifies CTCAE v4.03 as the assessment framework for treatment-related adverse events. This is important because adverse-event analysis depends not only on whether an event occurred, but also on how its severity and relationship to treatment are classified.
One-arm estimation
In a one-arm study, the estimated proportion of subjects experiencing an adverse event is a direct description of the observed study population. It is not automatically a causal treatment effect because there is no randomized comparator providing a counterfactual estimate of what would have happened without XmAb20717.
Time-window interpretation
The primary endpoint has a 56-day time frame. A safety percentage calculated within that window would therefore describe the experience observed during the specified assessment period. It should not automatically be interpreted as a lifetime incidence or as the incidence over an unspecified longer period.
Population denominators
Safety percentages are meaningful only when the denominator is defined. In a phase 1 study, the number enrolled, the number treated, and the number with sufficient observation can differ. The registry provides an enrollment of 150, but it does not provide posted adverse-event counts or a separate safety denominator in the information available here.
8. Statistical Methods Explained
Why is this primarily a descriptive analysis?
The registered primary endpoint concerns the safety and tolerability profile of a single biological intervention. With one intervention arm and no comparator arm, descriptive estimation of adverse-event frequency and severity is more directly aligned with the design than a randomized treatment-effect test.
What does an adverse-event percentage mean?
If, for example, a specified adverse event were observed in a proportion of the evaluated subjects, that percentage would describe how common the event was in the study population during the defined observation period. It would not by itself establish that XmAb20717 caused the event or that the event would have been absent without treatment.
Why does treatment-relatedness matter?
The registered endpoint specifically concerns treatment-related adverse events. This makes attribution part of the endpoint definition. Statistical summaries of all adverse events and summaries restricted to treatment-related events answer different questions and should not be treated as interchangeable.
Why is CTCAE v4.03 important?
A standardized toxicity framework allows adverse events to be classified consistently by severity. Without a defined grading framework, a numerical adverse-event rate would provide less information about the clinical seriousness of the observed events.
Why would confidence intervals be useful?
A percentage calculated from a finite sample is an estimate rather than an exact population parameter. A confidence interval communicates sampling uncertainty around that estimate. Wider intervals generally reflect less statistical precision, particularly when the number of observed events is small.
Why is there no hazard ratio here?
A hazard ratio compares event hazards between groups or covariate-defined groups in a time-to-event model. DUET-2 is registered with one intervention arm, and its primary endpoint is an adverse-event safety assessment over 56 days rather than a comparative time-to-event endpoint. A hazard ratio therefore does not naturally represent the primary analysis.
9. The Importance of the 56-Day Time Frame
The primary endpoint is explicitly defined over 56 Days. This time boundary is statistically important because adverse-event incidence depends on follow-up duration. A longer observation period can identify events that would not be observed during an earlier window, while a shorter window may be more closely aligned with early treatment tolerability.
What the window establishes
It defines the registered time frame for assessing the primary safety endpoint.
What it does not establish
A 56-day assessment should not automatically be interpreted as a complete account of all possible later safety events.
Why denominators matter
An event percentage depends on how many subjects were actually evaluable during the relevant period.
Why timing matters
Safety is longitudinal: the probability of observing an event depends partly on how long subjects are observed.
10. Safety and Tolerability as Statistical Endpoints
Safety analysis differs from efficacy analysis because the objective is usually to characterize the distribution of unwanted events rather than to estimate a single treatment-effect parameter. A useful safety analysis can therefore involve several dimensions simultaneously: whether an event occurred, its severity, its relationship to treatment, and its timing.
| Statistical question | Typical measure | Interpretive purpose |
|---|---|---|
| How many subjects experienced an event? | Count | Magnitude of the observed safety experience |
| How common was an event? | Percentage | Frequency relative to the analysis population |
| How severe was the event? | CTCAE grade | Characterization of toxicity severity |
| Was the event considered treatment-related? | Treatment-related classification | Attribution within the registered endpoint |
| Over what period was it assessed? | 56-day time frame | Defines the primary endpoint window |
These measures are complementary. A low overall event percentage can coexist with clinically important severe events, while a relatively common mild event may have a different interpretation. Statistical reporting should therefore preserve the structure of the safety endpoint rather than compressing it into a single number.
11. Why the One-Arm Design Matters
The one-arm design is one of the most important statistical features of DUET-2. Randomization is the mechanism that creates comparability between treatment groups in a conventional controlled trial. DUET-2 does not have that treatment-versus-control structure in its registered design.
| Question | One-arm interpretation |
|---|---|
| What can be directly described? | The observed safety experience associated with the intervention arm |
| What cannot be obtained from a between-arm comparison? | A randomized treatment-versus-control effect estimate |
| What is the main role of percentages? | Describing the frequency of observed adverse events |
| What is required for causal comparison? | An appropriate counterfactual or comparator framework |
This distinction is especially important for adverse events because some events can occur as part of the underlying disease, prior treatment exposure, background medications, or other clinical circumstances. Without a concurrent randomized comparator, the observed event rate cannot by itself isolate the incremental effect of XmAb20717.
12. Sequential Design
The registry identifies the design model as SEQUENTIAL. Sequential designs can be useful in early-phase development because information accumulates as subjects enter and are treated, allowing the study process to proceed through successive stages rather than requiring a conventional fixed parallel-group comparison.
Statistically, a sequential design makes the timing and decision rules important. If repeated assessments are used to guide continuation, expansion, or modification of treatment cohorts, those decision rules should be prespecified because repeatedly examining accumulating information can affect the operating characteristics of a study.
13. What a Formal Safety Table Would Tell the Reader
A complete safety analysis would normally organize adverse events so that the reader can distinguish frequency from severity and attribution. The most informative presentation would preserve the denominator and the event definition for each summary.
| Potential reporting element | Statistical interpretation |
|---|---|
| Number of subjects with treatment-related adverse events | Observed frequency in subjects meeting the relevant safety definition |
| Percentage with treatment-related adverse events | Observed event proportion in the relevant denominator |
| CTCAE v4.03 severity | Distribution of adverse events by toxicity grade |
| Event-specific counts | Identifies which adverse events contributed to the overall profile |
| 56-day assessment | Defines the primary endpoint observation period |
The registry does not report the corresponding posted numerical safety results. As a result, the statistical story available from the registry is primarily about the design and endpoint rather than an observed adverse-event distribution.
14. Limitations
- No randomized comparator: the registered study has one intervention arm, so observed safety frequencies cannot be converted into randomized treatment-effect estimates.
- Broad disease population: the registry lists many selected advanced solid tumor categories. Heterogeneity across tumor types can affect the interpretation and generalizability of observed safety findings.
- Limited primary endpoint window: the registered primary safety endpoint is assessed over 56 days, so the endpoint is specifically tied to that time frame.
- Attribution: the primary endpoint concerns treatment-related adverse events, making assessment of treatment relationship part of the outcome definition.
- Denominator dependence: safety percentages require a clearly defined analysis population. Enrollment of 150 does not by itself establish the denominator for every safety analysis.
- No posted statistical estimates: the registry does not provide posted numerical results for the primary endpoint, so estimates, confidence intervals, and p-values cannot be interpreted from the registry record.
- Sequential design: a sequential designation does not by itself specify interim boundaries, stopping rules, or multiplicity adjustments.
- Early-phase interpretation: phase 1 safety characterization generally emphasizes describing observed toxicities and tolerability rather than establishing comparative efficacy.
15. Why This Trial Matters Statistically
DUET-2 illustrates a different statistical problem from the large randomized efficacy trials that often dominate clinical-trial reporting. Its central question is safety characterization in an early-phase, one-arm setting. That makes the study useful for understanding why trial design determines which statistical quantities are meaningful.
| Concept | How it appears in DUET-2 |
|---|---|
| Phase 1 development | The registry identifies the study as phase 1. |
| One-arm design | The registry identifies one intervention arm. |
| Sequential design | The registry identifies the design model as SEQUENTIAL. |
| Safety endpoint | The primary endpoint concerns the safety and tolerability profile of XmAb20717. |
| Adverse-event analysis | Treatment-related adverse events are the operational measure. |
| Standardized toxicity assessment | Events are assessed using CTCAE v4.03. |
| Fixed observation period | The primary endpoint has a 56-day time frame. |
| Descriptive inference | Counts, percentages, and uncertainty intervals are natural tools for a one-arm safety assessment. |
| Causal interpretation | Without a comparator, observed event rates do not directly quantify a randomized treatment effect. |
16. Statistical Interpretation of Safety Rates
An adverse-event rate is a descriptive estimate of how frequently an event occurred among subjects in the relevant analysis population during the defined observation period. It is not automatically a measure of excess risk attributable to treatment.
A confidence interval around a one-arm event rate describes statistical uncertainty associated with estimating an underlying event probability from the observed sample. It does not describe the range of event rates that individual subjects might experience.
Without a randomized control arm, a percentage such as an adverse-event incidence cannot be interpreted as a relative increase or decrease versus an untreated or alternative-treatment population. Such comparisons require an appropriate external or concurrent reference framework.
A p-value, when used in a hypothesis test, measures the compatibility of observed data with a specified null hypothesis under the assumptions of that test. It does not measure the magnitude, clinical importance, or probability that an adverse event was caused by treatment.
17. Population Heterogeneity and Safety Interpretation
The breadth of the listed conditions creates an important statistical consideration. When subjects with different advanced solid tumors are combined, the overall safety profile represents the aggregate study population. That aggregate can be useful for characterizing treatment exposure, but it can also conceal differences in the underlying clinical context of individual tumor populations.
For example, an overall adverse-event percentage is a weighted summary across the subjects included in the analysis. The overall percentage therefore depends on the composition of the study population as well as the event frequencies within its component groups.
The aggregate rate reflects both event frequency and the relative representation of the different clinical populations contributing subjects to the denominator.
This is one reason why subgroup analyses, when prespecified and sufficiently informative, can complement an overall safety summary. The registry information available here does not provide posted subgroup safety results.
18. Trial Timeline
Study start
The registry lists 2018-07-10 as the study start date.
Early-phase treatment study
The study is registered as phase 1, with XmAb20717 as the biological intervention and treatment as the primary purpose.
Primary endpoint window
The registered primary endpoint evaluates treatment-related adverse events over 56 days using CTCAE v4.03.
Primary completion
The registry lists 2022-06-01 as the primary completion date.
Study status
The registry lists the study status as COMPLETED.
19. What This Design Can and Cannot Establish
Can describe
The observed treatment-related adverse-event experience in the study population during the registered assessment period.
Can characterize
The safety and tolerability profile of XmAb20717 using the registered CTCAE v4.03 assessment framework.
Cannot directly compare
XmAb20717 with a randomized concurrent control treatment because no comparator arm is registered.
Cannot infer causality from frequency alone
An observed adverse-event rate alone does not establish how much of the event burden was caused by the intervention.
This distinction is fundamental to clinical-trial statistics. The same numerical event rate can have a different interpretation depending on whether it comes from a randomized controlled trial, a single-arm study, an external-control analysis, or another design. The design supplies the counterfactual structure needed to determine what a statistical comparison means.
20. Statistical Concepts in This Trial
The statistical interpretation of DUET-2 centers on early-phase safety estimation, standardized adverse-event assessment, one-arm study design, and the importance of clearly defining the observation period and analysis denominator.
| Concept | Application to DUET-2 |
|---|---|
| Descriptive statistics | Natural framework for summarizing adverse-event frequency and severity in a one-arm study. |
| Proportion estimation | Adverse-event percentages describe observed frequency in the relevant population. |
| Confidence intervals | Can quantify uncertainty around one-arm event-rate estimates when event counts are available. |
| CTCAE v4.03 | Provides the registered framework for adverse-event assessment. |
| One-arm clinical trial | Provides no randomized concurrent comparator for treatment-effect estimation. |
| Sequential design | Identified in the registry and potentially relevant to how information accumulates during the study. |
| Observation window | The registered primary endpoint is assessed over 56 days. |
| Population heterogeneity | The registry includes numerous selected advanced solid tumor categories. |
21. Sources
- ClinicalTrials.gov: DUET-2 (NCT03517488), the official registry record for the study.
Continue through the Clinical Biostats knowledge graph
Explore statistical concepts, clinical-trial methods, and analysis workflows that provide additional context for interpreting clinical research.
22. Record Summary
DUET-2 is registered as a completed phase 1 treatment study of XmAb20717 in subjects with selected advanced solid tumors. The study has a sequential design, no masking, one intervention arm, and an enrollment of 150. Its registered primary endpoint is to determine the safety and tolerability profile of XmAb20717 through treatment-related adverse events assessed by CTCAE v4.03 over 56 days.
Statistically, the key lesson is that study design determines the appropriate estimand. In this one-arm setting, the central quantities are descriptive safety measures such as event counts and percentages, potentially accompanied by confidence intervals. A randomized treatment-effect estimate requires a comparator framework that is not present in the registered design.