← Clinical Trials
Metastatic Colorectal Cancer Phase 3 Double-Blind NCT04607668

PRESERVE 1: Complete Statistical Analysis of Trilaciclib in Metastatic Colorectal Cancer

An independent statistical analysis of the randomized phase 3 PRESERVE 1 trial evaluating trilaciclib versus placebo in patients receiving FOLFOXIRI/bevacizumab for metastatic colorectal cancer, with primary analyses of severe neutropenia.

Trial status: Terminated  ·  Enrollment: 326  ·  Primary completion: February 13, 2023
Scope of this record

This page separates reported trial results from statistical interpretation. Numerical results are restricted to the information posted in the ClinicalTrials.gov record. The registry provides the official trial record.

Registry note: 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

PRESERVE 1 was a randomized, parallel, double-blind phase 3 trial of trilaciclib versus placebo in patients receiving FOLFOXIRI/bevacizumab for metastatic colorectal cancer. The registered primary endpoints focused on the duration and occurrence of severe neutropenia during induction.

326
Enrollment
Randomized trial
2
Arms
Trilaciclib and placebo
-1.2
DSN Mean Difference
95% CI -1.7 to -0.6
0.04
SN aRR
95% CI 0.01–0.19
FeaturePRESERVE 1
Trial namePRESERVE 1
ClinicalTrials.gov identifierNCT04607668
PhasePhase 3
StatusTerminated
StartJanuary 6, 2021
Primary completionFebruary 13, 2023
Lead sponsorG1 Therapeutics, Inc.
Sponsor typeIndustry
Enrollment326.0
AllocationRandomized
Design modelParallel
MaskingDouble
Primary purposeTreatment
InterventionsTrilaciclib; placebo
Results postedYes
Outcome measures posted15
Statistical analyses posted2

2. Clinical Question

The statistical question in PRESERVE 1 centers on whether trilaciclib changes severe neutropenia outcomes when patients receive FOLFOXIRI/bevacizumab for metastatic colorectal cancer. The two registered primary endpoints address both the duration of a severe neutropenia episode and the occurrence of severe neutropenia during the induction period.

Population

Patients receiving FOLFOXIRI/bevacizumab for metastatic colorectal cancer, with the trial also identifying myelosuppression in adults and chemotherapeutic toxicity among its conditions.

Intervention

Trilaciclib administered in the context of FOLFOXIRI/bevacizumab.

Comparator

Placebo administered in the context of FOLFOXIRI/bevacizumab.

Primary question

How does trilaciclib compare with placebo with respect to the duration and occurrence of severe neutropenia during the prespecified induction periods?

3. Trial Design

01
Randomize326 enrolled
02
Two armsTrilaciclib or placebo
03
Induction14-day cycles
04
Assess SNANC-based endpoint
05
AnalyzeANCOVA and Poisson regression
Allocation
Randomized
Participants were allocated in a randomized parallel-group design.
Masking
Double
The registry identifies the trial as double-blind.
Phase
Phase 3
The trial was designed as a phase 3 treatment study.
Primary purpose
Treatment
The registered primary purpose was treatment.
ARM A

Trilaciclib + FOLFOXIRI/Bevacizumab

  • Trilaciclib
  • FOLFOXIRI/bevacizumab chemotherapy regimen
  • Primary endpoint analyses included this treatment group
ARM B

Placebo + FOLFOXIRI/Bevacizumab

  • Placebo
  • FOLFOXIRI/bevacizumab chemotherapy regimen
  • Comparator arm for the randomized trial

4. Endpoints

Primary endpointRegistry definitionTime frame
Duration of Severe Neutropenia (DSN) The DSN was defined as the number of days for the first severe neutropenia (SN) event in Cycles 1, 2, 3, or 4 for participants who had at least one SN event in the first 4 cycles of Induction. It was calculated as the days from the date of the first absolute neutrophil count (ANC) value < 0.5 × 109/L to the date of the first ANC value ≥ 0.5 × 109/L where no additional ANC values < 0.5 × 1... Cycles 1 to 4 (14-day cycles up to 56 days)
Occurrence of Severe Neutropenia (SN) During Induction Severe neutropenia was defined as the absolute neutrophil count (ANC) laboratory value that met the Common Terminology Criteria for Adverse Events (CTCAE) criteria for ≥ Grade 4 toxicity, ie, ANC < 0.5 × 109/L in SI Unit. Induction Period, cycles 1-12 (14-day cycles up to 168 days)
Two different statistical questions: DSN is expressed in days and concerns how long the first qualifying severe-neutropenia episode lasted among participants meeting the endpoint's event criterion. Occurrence of SN is expressed as events per 100 cycles and addresses the frequency of severe-neutropenia events during induction. These are related but distinct outcomes.

5. Analysis Populations

Analysis populationRole in the posted analysis
Modified intent-to-treat (mITT) Used for the posted primary analyses. The registry states that the modification was intended to account for potential data-integrity issues resulting from the war in Ukraine.
Why this matters: An mITT population is not automatically equivalent to an all-randomized ITT population. Because the registry specifically identifies mITT as the analysis population, interpretation of the estimates should remain tied to that population rather than silently treating the results as though they were calculated in every randomized participant.

6. Statistical Methodology

ANCOVA for Duration of Severe Neutropenia

The posted analysis of DSN used analysis of covariance (ANCOVA) and reported a mean difference in final values. ANCOVA is a linear-model framework that compares an outcome between groups while allowing prespecified covariates to be incorporated into the model. The ClinicalTrials.gov record does not identify the individual covariates used in this analysis, so they are not inferred here.

Conceptual ANCOVA form
Y = β0 + β1Treatment + β2X1 + β3X2 + ε

The treatment coefficient represents the modeled difference between treatment groups after accounting for whatever covariates are included in the specified model. In PRESERVE 1, the registry reports the treatment effect as a mean difference in final DSN values.

Poisson regression for Severe Neutropenia Occurrence

The occurrence analysis used a modified Poisson model, normalized here as Poisson regression, and reported an adjusted rate ratio (aRR). The outcome unit was events per 100 cycles. Count and rate models are appropriate when the scientific question concerns how frequently events occur over an exposure or observation opportunity such as treatment cycles.

Conceptual rate comparison
Rate ratio = Event ratetreatment / Event ratecomparator

A rate ratio compares event rates rather than simply comparing the percentage of participants with an event. The registry analysis lists the trilaciclib + FOLFOXIRI/bevacizumab group under “Groups Compared”; it does not provide a separate comparator-group label in that field.

Confidence intervals

Both posted primary analyses include two-sided 95% confidence intervals. For DSN, the interval quantifies uncertainty around the estimated mean difference. For the occurrence analysis, it quantifies uncertainty around the reported aRR. A confidence interval is about uncertainty in the estimated parameter; it is not a range within which individual patient outcomes must fall.

Modified intention-to-treat analysis

7. Primary Results

ClinicalTrials.gov reports two formal statistical analyses corresponding to the two registered primary endpoints. Both include an estimate, a two-sided 95% confidence interval, and a P-value.

Duration of Severe Neutropenia

Mean difference in DSN

-1.2 days

95% CI: -1.7 to -0.6   ·   P < 0.001

Analysis method: ANCOVA  ·  Analysis population: mITT

The posted effect measure is Mean Difference (Final Values). The compared groups are listed as trilaciclib + FOLFOXIRI/bevacizumab versus placebo + FOLFOXIRI/bevacizumab. Thus, the negative mean difference indicates a lower estimated final DSN value for the first-listed group under the reported analysis.

EndpointEffect measureEstimate95% CIP-valueMethod
Duration of Severe Neutropenia Mean difference (final values) -1.2 days -1.7 to -0.6 <0.001 ANCOVA
Clinical Biostats interpretation

The estimate of -1.2 days represents a modeled difference in the duration of severe neutropenia, with the trilaciclib combination listed first and the placebo combination second. Because the estimate is negative, the direction favors a shorter estimated DSN in the first-listed group.

The estimate does not mean that every participant experienced exactly 1.2 fewer days of severe neutropenia. It is a group-level statistical estimate based on the specified mITT analysis population.

The two-sided 95% CI of -1.7 to -0.6 describes uncertainty around the estimated mean difference under the analysis framework. It does not describe the range of individual patient-level DSN values.

The P < 0.001 result addresses evidence against the relevant null hypothesis under the reported statistical test. It does not measure the size or clinical importance of the effect. Effect magnitude is conveyed by the -1.2-day estimate and its confidence interval.

Because the registry classifies DSN as a time-to-event endpoint but reports an ANCOVA analysis of final values, it is important not to substitute a Kaplan-Meier or Cox-model interpretation for the analysis actually posted. The reported result is an ANCOVA mean difference.

Occurrence of Severe Neutropenia During Induction

Adjusted rate ratio

0.04

95% CI: 0.01–0.19   ·   P < 0.001

Outcome unit: events per 100 cycles  ·  Analysis method: modified Poisson model

The posted analysis reports an aRR of 0.04 for the trilaciclib + FOLFOXIRI/bevacizumab group. The ClinicalTrials.gov record identifies the outcome as severe neutropenia during cycles 1–12 and gives the event unit as events per 100 cycles. Because the “Groups Compared” field in the ClinicalTrials.gov record lists only the trilaciclib group, this page does not assign an unreported comparator label to the aRR beyond the trial's randomized placebo arm.

EndpointEffect measureEstimate95% CIP-valueMethod
Occurrence of Severe Neutropenia During Induction aRR 0.04 0.01–0.19 <0.001 Modified Poisson model
Clinical Biostats interpretation

An aRR of 0.04 is a relative rate estimate. In the analysis as reported, it indicates a modeled rate ratio of 0.04 for the listed trilaciclib group relative to the model's reference rate. Expressed as a simple relative-rate interpretation, 0.04 corresponds to an estimated rate that is 4% of the reference rate, or a 96% lower estimated rate.

That interpretation does not mean that 96% of patients avoided severe neutropenia, nor does it mean that the probability of an individual patient's event was reduced by exactly 96%. A rate ratio concerns event frequency over the specified cycle-based exposure.

The 95% CI of 0.01–0.19 indicates substantial uncertainty in the precise magnitude of the rate ratio even though the entire interval is below 1.00. It should not be interpreted as a range of individual patient risks.

The P < 0.001 value provides evidence against the relevant null hypothesis for the reported model. It does not quantify the magnitude of the treatment effect; the estimate and confidence interval do that.

Because this endpoint is reported as events per 100 cycles, the distinction between an event rate and a participant-level incidence proportion is especially important. A participant can contribute exposure across multiple cycles, so the rate framework answers a different question from simply asking what percentage of participants experienced at least one event.

8. Safety Results

The ClinicalTrials.gov record reports serious adverse events by randomized arm. The reported figures are the number affected divided by the number at risk.

Safety measureTrilaciclib + FOLFOXIRI/BevacizumabPlacebo + FOLFOXIRI/Bevacizumab
Serious adverse events 47/159 47/160
Serious adverse events: affected / at risk
Trilaciclib combination
47/159
Placebo combination
47/160

The same number of affected participants, 47, is reported in both arms, while the denominators differ by one participant. These data should be interpreted as the registry-reported serious-adverse-event counts rather than as a formal hypothesis test, because no inferential safety analysis is reported in the ClinicalTrials.gov record.

Safety interpretation: the serious-adverse-event figures should not be combined with the efficacy estimates into a single numerical “benefit-risk” statistic. They describe a separate component of the trial evidence and use a different analysis question.

9. Statistical Methods Explained

Why was ANCOVA used for DSN?

ANCOVA is a linear-model approach for comparing a continuous outcome between treatment groups while allowing specified covariates to be included. In this trial, the posted DSN result is a mean difference in final values and the registry identifies ANCOVA as the method. The ClinicalTrials.gov record does not identify the specific covariates, so their effects cannot be described without going beyond the record.

What does a mean difference of -1.2 days mean?

A mean difference of -1.2 days means that the modeled mean DSN for the first-listed treatment group was 1.2 days lower than the comparator mean under the reported analysis. It is not a statement that every patient had a reduction of exactly 1.2 days.

Why does DSN have a time-to-event definition but use ANCOVA?

The registry classifies DSN as a time-to-event endpoint because it measures elapsed time from one ANC-defined state to another. However, the posted formal analysis uses ANCOVA and reports a mean difference in final values. Statistical interpretation should follow the actual registered analysis rather than replacing it with a more familiar survival-analysis method.

What does an adjusted rate ratio of 0.04 mean?

A rate ratio of 0.04 indicates an estimated event rate equal to 4% of the model's reference rate. Equivalently, the estimated rate is 96% lower relative to that reference. The “adjusted” designation indicates that the modified Poisson model accounts for model terms beyond a simple unadjusted rate ratio, although the ClinicalTrials.gov record does not specify all adjustment variables.

Why analyze events per 100 cycles rather than only participants?

Severe neutropenia during induction is reported in units of events per 100 cycles. A cycle-based rate incorporates the amount of treatment-cycle exposure into the outcome. This can provide information that a simple proportion of participants with at least one event would not capture.

What does the 95% confidence interval tell us?

The 95% confidence interval describes uncertainty around the estimated statistical parameter under the model and sampling framework. For DSN, the interval is -1.7 to -0.6; for the reported aRR, it is 0.01–0.19. Neither interval represents the range of responses that individual patients should be expected to experience.

Why does the P-value not measure effect size?

A P-value describes how compatible the observed result is with a specified null hypothesis under the statistical model. It depends on both the observed effect and the amount of information available. The effect estimate and confidence interval are therefore essential for understanding magnitude and precision; a small P-value by itself does not establish how large an effect is.

10. Intention-to-Treat and Randomization

Randomization is a central feature of PRESERVE 1: the registry identifies the allocation as randomized and the design as parallel. Randomization creates the framework for comparing treatment groups without intentionally selecting participants according to expected outcomes.

However, the posted primary analyses were conducted in a modified intent-to-treat population, rather than simply being described as analyses of all randomized participants. The registry explicitly states that the mITT population was used to account for potential data-integrity issues resulting from the war in Ukraine.

What randomization contributes

Randomization establishes the treatment-assignment mechanism and supports causal comparison between the randomized groups.

What mITT changes

The formal estimate is based on the specified modified population. Exact inclusion criteria matter when assessing how closely the analysis population corresponds to the randomized population.

11. Confidence Intervals and Statistical Precision

The two primary estimates have substantially different statistical forms, so their confidence intervals should be read on their respective scales.

EndpointEstimate95% CIWhat the interval describes
DSN -1.2 days -1.7 to -0.6 Uncertainty around the estimated mean difference in final DSN values.
SN occurrence aRR 0.04 0.01–0.19 Uncertainty around the estimated adjusted rate ratio.

The DSN interval is entirely below zero, consistent with the negative point estimate. The aRR interval is entirely below 1.00, consistent with the point estimate being below 1.00. These are descriptive properties of the reported estimates and intervals; they should not be confused with a statement about individual treatment response.

12. Primary Endpoint Comparison

The two primary endpoints describe different dimensions of severe neutropenia. Looking at them together provides a more complete statistical picture than either endpoint alone.

DimensionDSNSN occurrence
Question How long did the first qualifying severe-neutropenia episode last? How frequently did severe neutropenia occur during induction?
Time frame Cycles 1 to 4, up to 56 days Induction cycles 1-12, up to 168 days
Outcome unit Days Event per 100 cycles
Posted method ANCOVA Modified Poisson model
Effect measure Mean difference aRR
Estimate -1.2 0.04
95% CI -1.7 to -0.6 0.01–0.19
P-value <0.001 <0.001

The statistical scales should not be combined. A change in mean DSN cannot be directly added to, multiplied by, or otherwise converted into the reported rate ratio. They answer separate questions about severe neutropenia.

13. Time-to-Event Concepts in PRESERVE 1

DSN is classified in the ClinicalTrials.gov record as a time-to-event endpoint. That classification comes from the structure of the endpoint: the outcome is defined by the time between an ANC value meeting the severe-neutropenia threshold and a subsequent ANC value meeting the recovery threshold under the registry definition.

Time-to-event concept
Time to event = Event or recovery date − Starting date

For DSN, the registry describes the starting point as the date of the first ANC value < 0.5 × 109/L and the ending point as the date of the first ANC value ≥ 0.5 × 109/L subject to the registry's additional ANC condition.

In many clinical-trial settings, time-to-event outcomes are analyzed using Kaplan-Meier estimation and Cox regression because censoring and incomplete follow-up can be important. PRESERVE 1 is a useful reminder that an endpoint's scientific classification does not dictate a single mandatory statistical method: the actual posted analysis here was ANCOVA.

14. Why Poisson Regression Fits the Occurrence Endpoint

The occurrence endpoint is expressed as event per 100 cycles. That wording makes the exposure opportunity part of the statistical question. Poisson-type regression is designed for modeling counts or rates and can incorporate exposure information through the model specification.

Participant proportion

Answers a question such as how many participants experienced at least one event.

Cycle-based event rate

Answers a question about the frequency of events relative to treatment-cycle exposure.

Rate ratio

Compares event rates on a multiplicative scale.

Adjusted rate ratio

Represents the rate comparison from a model that includes adjustment terms.

The reported aRR of 0.04 therefore should be interpreted on a rate scale. It should not be relabeled as an odds ratio, risk ratio, or hazard ratio.

15. What the Primary Results Do — and Do Not — Establish

DSN result

The reported -1.2-day mean difference indicates a lower estimated DSN for the first-listed treatment group under the ANCOVA analysis. It does not establish that every patient experienced a 1.2-day reduction, and it does not describe the distribution of individual DSN values.

SN occurrence result

The reported aRR of 0.04 indicates a substantially lower modeled event rate relative to the model's reference rate. It does not mean that the probability of severe neutropenia for every individual was 96% lower, because the endpoint is a cycle-based event rate rather than simply an individual-level binary outcome.

Statistical significance

Both posted primary analyses have P < 0.001. This is evidence against the relevant null hypothesis under the reported analyses. It is not a measure of clinical importance, magnitude, or certainty about the experience of any individual participant.

16. Multiplicity, Interim Analysis, and Other Design Features

The ClinicalTrials.gov record identifies two primary endpoints and two statistical analyses, but it does not provide information about multiplicity adjustment, interim analyses, alpha spending, or a formal endpoint hierarchy. Those topics therefore cannot be assigned a specific procedure for PRESERVE 1 from the available record.

Design topicWhat the ClinicalTrials.gov record supports
Primary endpointsTwo registered primary endpoints.
Multiplicity procedureNot reported in the ClinicalTrials.gov record.
Interim analysisNot reported in the ClinicalTrials.gov record.
Alpha spendingNot reported in the ClinicalTrials.gov record.
Non-inferiority marginNot reported; the ClinicalTrials.gov record does not identify a non-inferiority design.
CrossoverNot reported in the ClinicalTrials.gov record.
Bayesian methodsNot reported in the ClinicalTrials.gov record.
Stratification factorsNot reported in the ClinicalTrials.gov record.
Missing-data/imputation procedureNot reported in the ClinicalTrials.gov record.
Why omission matters: the absence of a procedure in the ClinicalTrials.gov record is not evidence that the procedure was absent from the full protocol or statistical analysis plan. It means only that the ClinicalTrials.gov record does not support describing one.

17. Trial Timeline

January 6, 2021

Trial start

The registered trial start date was January 6, 2021.

February 13, 2023

Primary completion

The registered primary completion date was February 13, 2023.

Registry results

Statistical analyses posted

The ClinicalTrials.gov record reports 15 outcome measures and 2 statistical analyses, including both registered primary endpoints.

Current registry status

Terminated

the ClinicalTrials.gov record identifies the trial status as terminated.

18. Serious Adverse Events and Denominators

The serious-adverse-event data provide a useful statistical lesson about denominators. The two arms have the same reported number of affected participants, but their at-risk denominators differ.

ArmAffectedAt riskReported value
Trilaciclib + FOLFOXIRI/Bevacizumab 47 159 47/159
Placebo + FOLFOXIRI/Bevacizumab 47 160 47/160

A count alone is not sufficient for comparing event frequency when denominators differ. Conversely, the presence of similar counts does not by itself establish statistical equivalence or equivalence of risk. A formal comparison would require the appropriate safety-analysis definition and inferential method, neither of which is reported here.

19. Important Limitations and Interpretation Issues

20. Why This Trial Matters Statistically

PRESERVE 1 is a useful teaching example because its two primary endpoints require two different statistical perspectives. The first is a duration measure summarized with a mean difference from ANCOVA; the second is a cycle-based event-rate outcome analyzed with a modified Poisson model. The same randomized trial therefore illustrates why the statistical method must follow the structure of the endpoint rather than being selected solely from the disease area.

ConceptHow it appears in PRESERVE 1
RandomizationRandomized parallel-group phase 3 design.
BlindingDouble-blind trial.
Modified ITTmITT population used for both posted primary analyses.
ANCOVAUsed for the DSN mean-difference analysis.
Mean differenceDSN effect estimate of -1.2 days.
Confidence interval95% two-sided intervals reported for both primary analyses.
P-valueP < 0.001 for both posted primary analyses.
Poisson regressionModified Poisson model used for severe-neutropenia occurrence.
Rate ratioaRR of 0.04 for the posted occurrence analysis.
Exposure-based endpointSevere-neutropenia occurrence reported as events per 100 cycles.
Time-to-event endpointDSN is classified as time-to-event in the registry data.
Safety denominatorsSerious adverse events reported as affected / at risk by arm.

21. Statistical Interpretation in Context

The most important statistical feature of PRESERVE 1 is the distinction between effect magnitude, statistical uncertainty, and endpoint definition.

Magnitude

The DSN estimate is -1.2 days. The occurrence analysis reports an aRR of 0.04. These are different effect scales and should remain separate.

Precision

The DSN 95% CI is -1.7 to -0.6, while the occurrence-analysis 95% CI is 0.01–0.19. The width and scale of each interval provide information about uncertainty.

Evidence against the null

Both reported P-values are <0.001. This supports statistical evidence against the relevant null hypotheses under their respective models.

Clinical meaning

Statistical estimates describe group-level treatment effects. They do not determine the outcome for an individual participant or replace clinical judgment about the significance of an effect.

22. Related Tutorials

Learn more about the methods used in this trial:

23. Related Calculators

24. Sources

Continue through the Clinical Biostats statistical pathway

Explore the statistical methods behind randomized trials, continuous outcomes, count and rate data, confidence intervals, and time-to-event endpoints.

25. Record Summary

PRESERVE 1 provides a compact example of how statistical analysis should be matched to the structure of clinical-trial endpoints. The randomized, double-blind phase 3 design enrolled 326.0 participants and evaluated trilaciclib versus placebo in the setting of FOLFOXIRI/bevacizumab for metastatic colorectal cancer. Its two registered primary endpoints address different aspects of severe neutropenia: duration during cycles 1 to 4 and occurrence during induction cycles 1-12.

The posted DSN analysis used ANCOVA and reported a mean difference of -1.2 days with a two-sided 95% CI of -1.7 to -0.6 and P < 0.001. The posted severe-neutropenia occurrence analysis used a modified Poisson model and reported an aRR of 0.04 with a two-sided 95% CI of 0.01–0.19 and P < 0.001.

The statistical interpretation should remain disciplined: the DSN result is a mean difference, the occurrence result is a rate ratio, the confidence intervals describe uncertainty around their respective estimates, and the P-values do not measure effect size. The primary analyses were performed in a modified intent-to-treat population, and the ClinicalTrials.gov record does not provide enough information to reconstruct several other design or analysis procedures.

Clinical Biostats methodology: A trial-results page should not merely repeat reported numbers. The goal is to connect the endpoint definition, analysis population, statistical model, effect measure, confidence interval, and P-value so that the reader can understand exactly what each result does—and does not—show.