epiClinicalA fictional CRO

epiClinical is fictional. It is a portfolio demonstration by Adam Daniels. Its studies, sponsors, sites and enrolment figures are synthetic. The evidence behind its assumptions is real, and every number says how far it can be trusted.

EPI-214, Phase III. Recruiting

Heart failure with reduced ejection fraction

CTS-209 added to standard care in heart failure with reduced ejection fraction. Sponsor: Contoso Pharmaceuticals (fictional), Cardiovascular.

QTLs at a glance

Data to September 30th 2026. Each measure links to its chart below.

Stopped study treatment early

18.9% against 15%

Outside range Outside range

Ineligible patients randomised

2.5% against 5%

On course Within range

Missing primary endpoint data

3.3% against 15%

On course Within range

Lost to follow-up or withdrew consent

1.1% against 6%

On course Within range

Will the result hold?

Quality tolerance limits (acceptable ranges, in ICH E6(R3)), most important first. Priority is how likely this study's design makes a breach, times how strongly the evidence links the measure to a reliable result. Each has an early-warning line, a limit and a critical line where the study's power would fall to the escalation level.

RankQTLLikelihoodImpactWhy it is likely hereWatched at sites byWarningLimitCritical
1Stopped study treatment earlyEvidence grade B32
  • Placebo or no-added-treatment control arm (PD-05)From the design: What does the control arm receive? Placebo or no added treatment
12.5%15%22%
2Ineligible patients randomisedEvidence grade B22
  • Restrictive eligibility criteria and protocol complexity (EB-04)From the design: How many non-essential exclusions does the protocol carry? One or two
  • Eligibility criteria exclude patients (comorbidity, age, performance status, language) (PB-02)From the design: How many non-essential exclusions does the protocol carry? One or two
  • Restrictive eligibility criteria (organ function, brain metastases, prior or concurrent malignancy, HIV, laboratory thresholds, performance status) (PD-01)From the design: How many non-essential exclusions does the protocol carry? One or two
  • Number of eligibility criteria and complexity growth (PD-02)From the design: How many non-essential exclusions does the protocol carry? One or two
4%5%14.5%
3Missing primary endpoint dataEvidence grade A13None of its drivers applies to this design10.1%15%33.2%
4Lost to follow-up or withdrew consentEvidence grade A13None of its drivers applies to this design4.5%6%24.3%

Stopped study treatment early

Outside range% of patients; benchmark median 10.7%

Show data
Stopped study treatment early: running rate by month
MonthPatientsRate
Mar 202450%
Jun 2024200%
Sep 2024420%
Dec 2024891.1%
Mar 20251473.4%
Jun 20251947.2%
Sep 202524210.3%
Dec 202529214.4%
Mar 202635215.6%
Jun 202640017.8%
Sep 202645618.4%
Oct 202647218.9%

epiClinical assumption: above the Phase III benchmark of 11.5%; a breach dilutes the treatment effect and reduces power

Ineligible patients randomised

Within range% of patients; benchmark median 3.2%

Show data
Ineligible patients randomised: running rate by month
MonthPatientsRate
Mar 202450%
Jun 2024200%
Sep 2024422.4%
Dec 2024892.2%
Mar 20251472%
Jun 20251942.6%
Sep 20252422.9%
Dec 20252922.7%
Mar 20263522.3%
Jun 20264002.3%
Sep 20264562.2%
Oct 20264722.5%

epiClinical assumption: above the 3.2% median; eligibility errors were reported in 39% of 82 trials

Missing primary endpoint data

Within range% of patients; benchmark median 7.9%

Show data
Missing primary endpoint data: running rate by month
MonthPatientsRate
Sep 202450%
Dec 2024200%
Mar 2025420%
Jun 2025893.4%
Sep 20251464.1%
Dec 20251934.1%
Mar 20262413.3%
Jun 20262913.4%
Sep 20263503.4%
Oct 20263613.3%

The acceptable range is the drop-out the study's power calculation allowed for (planner parameters, D24); the early-warning line sits at two-thirds of it

Lost to follow-up or withdrew consent

Within range% of patients; benchmark median 3.7%

Show data
Lost to follow-up or withdrew consent: running rate by month
MonthPatientsRate
Mar 202450%
Jun 2024200%
Sep 2024420%
Dec 2024890%
Mar 20251470%
Jun 20251940%
Sep 20252420.4%
Dec 20252920.7%
Mar 20263520.6%
Jun 20264001%
Sep 20264561.1%
Oct 20264721.1%

epiClinical assumption: above the combined benchmark median of 3.7% and in line with published case studies that set limits around 4% for unknown survival status

Patients needed before a QTL is judged: 50. epiClinical assumption: QTL rates are unstable with few patients, and many early breaches in vendor data were noise (review note 01).

Breaches and decisions

Breach

Stopped study treatment early

Early warning November 2025; left the range February 2026
Status
Evaluated; actions proposed
What the evaluation found
Most patients who stopped did so after the dose step-up at week 12, reporting dizziness or low blood pressure; older patients were affected most. Sites that adjusted the dose as the protocol's guidance describes saw about half the stopping rate of those that did not.
Why it matters
Stopping treatment dilutes the measured treatment effect. The sample size allowed for 15% of patients stopping, so if the trend continues the study may lose the power to show a difference.

Proposed CAPAs

  1. Keep following every patient who stops treatment, so their outcome data are still collected.

    Protects the missing-data QTLs: outcomes collected after stopping treatment keep the result reliable (review note 01).

  2. Reissue the dose-adjustment guidance and run a short refresher with each site's investigator and study nurse.

    Evidence grade D. No quantitative evidence that site re-training changes outcomes; included because the finding points to how sites apply the guidance.

  3. Have the medical monitor review each stop with the investigator within a week, to see whether it could have been managed with a dose adjustment.

    A process step with no outcome evidence; it makes each decision visible and recorded.

A fictional scenario written to show how a breach is evaluated. No real patients or studies are described.

Critical lines assume each study was designed for 80% power at a two-sided 5% level, and mark where power would fall to 70%. Missing outcome data can bias a result well before it costs power, so its limit, not its critical line, is the one to act on.