Why Correcting Trial Imaging Errors at Central Review Is Too Late—and Too Expensive

01.10.25 12:43 PM

Article originally posted on The Clinical Trial Vanguard by Moe Alsumidaie on September 30, 2025

The patient drove two hours each way for the trial, which her oncologist and she believed was her best shot. After five infusions and 40 weeks, site radiology reports kept saying the same thing—“responding”—but the reads were using the wrong response criteria. On paper, the tumors looked smaller. In reality, the math was off.

When the CRO’s blinded central review re-read the scans with the correct criteria, those “responses” disappeared. The patient was devastated, removed from the protocol, and censored from the analysis, while the study absorbed the cost and delay. We’ve treated errors like this as the cost of doing business. They’re not.

This article discusses why central review is often too late for ethical patient care, how site-level imaging errors drive censoring and costly delays, provides real-world examples of the impact, and explains how moving quality upstream with technology that utilizes real-time, protocol-specific site reads enables central review to confirm—not rescue—the trial.

The Problem Is Already in the Trial When BICRs Begin

Most CROs’ imaging work is retrospective in the sense that the site assessment is often first and relied upon before the CRO results are in. In batched imaging CRO work, it’s even worse. All too often, images are read weeks or months after the patient visit, long after eligibility decisions have triggered changes in treatment, as happened to Sarah.

Multiple analyses have shown frequent discordance between local site reads and blinded central review, driven by baseline and target-lesion selection, measurement variability, new-lesion detection, and timing. These analyses also demonstrate that standardized reporting and RECIST training can improve agreement at the point of care. ^[1, 2, 3]

What that means in trials: a seemingly small site-level misapplication of RECIST (wrongly applied modifications to the tumor response criteria, inappropriate selection of target lesions, improper accounting of new lesions, and mistaken use of the wrong timepoint in comparison as a baseline) can set off chain reactions of undetected issues that central readers will later overturn. The database may eventually get “corrected.” The cancer patient’s experience and outcome, sadly, will not. These site quality issues quickly turn into patient safety issues and serious protocol violations.

The Hidden Cost of “Fixing It Later”

Central review of imaging, often referred to as Blinded Independent Central Review (BICR), is not free in money, time, or statistical side effects.

Cost & complexity: Dedicated double reads, adjudication, and central lab operations add significant budget and coordination overhead, which is why many sponsors reserve BICR for late-phase programs—leaving early-phase studies vulnerable to upstream errors that become embedded in early signals of treatment efficacy and trial go/no-go decisions
Informative censoring: When local investigators call progression, patients typically stop clinical trial protocol imaging steps. If BICR later disagrees with that progression at the same time point, those patients are censored in the BICR analysis, violating key survival-analysis assumptions and potentially inflating the median Progression Free Survival (PFS) statistic, which is often a primary endpoint on the trial.

And then there’s the operational math when you add up late corrections. Studies from NCI-designated centers reported 25%, 30%, and 50% site-level imaging error rates before introducing protocol-enforcing imaging informatics, dropping to <3% afterward. That upstream fix translates to fewer re-screens, fewer repeat scans, and fewer last-minute exclusions. Sponsors are increasingly moving directly from Phase 1 to Phase 3 in their trial designs, making the early site signals even more important.

A modeling exercise on trial budgets and cycle times—using those observed error rates—suggested potential savings on the order of $1.1 Million (~8%) and 57 days in Phase I, and $1.8 Million (~7%) and 192 days in Phase II when you prevent the errors instead of cleaning them at the end. It is intuitively evident that moving quality upstream to sites and being closer to the patient in real-time pays for itself many times over.

Two Moments That Change Everything

Below are real-world snapshots of how these problems unfold.

Early-phase Trials: A multicenter Phase I/II immunotherapy study screened 88 patients across 11 sites. With conventional workflows, 13 screen failures were discovered after randomization activities had begun—five were tied to baseline target-lesion selection errors, three to mismatched imaging windows, and the remainder to calculation drifts. Central review caught them, but too late: four patients had already received the first dose, six slots had to be reopened, and timelines slipped by more than seven weeks.

Late-phase Trials: In multiple programs, BICR assessments diverged from investigator assessments on calling PFS; when investigators declared progressive disease (PD), protocol imaging ceased. BICR later did not confirm PD at that time point yet those patients were now censored in the BICR dataset, which tilted the Kaplan–Meier curve and complicated interpretation and ongoing discussions with the FDA.

Both stories feature the current process and conclude in the same way: the database becomes “correct” over time, but only after the patient’s path, study schedule, and therapeutic efficacy have absorbed the impact of the protocol violations that occurred.

Things Only Accurate Site Imaging Can Do

The only way to avoid protocol violations, results delays, and the need for additional patient recruitment efforts is to understand the impact of accurate site imaging.

It ensures the correct patients are enrolled
It maintains patient eligibility in real-time
It eliminates the 10% patient censor rate in central review (BICR)
It protects patients’ interests and builds trust
It accelerates sponsor decision-making

Accurate site imaging is better, faster, and less costly for everyone.

Move Quality Upstream: Real-Time, Protocol-Specific Site Reads

In practice, moving quality upstream means enabling site radiologists and outsourced reading groups supporting sites, with tools that respect the protocol as they read, delivering:

Accuracy with trial-specific tumor response criteria that is enforced across all sites
Traceability linking every database value back to the images and workflow history
Direct imaging data access for sponsors/CROs to see real-time status and results

This is where Yunu fits. It has a mature and technologically advanced cloud imaging platform that eliminates disparate systems and orchestrates workflows across sites, CRO’s, and sponsors. Leading academic institutions utilize it to manage every trial at their site, as well as for sponsors to manage their trials across all participating sites, providing new workflow alternatives for central review processes. Whether it is used for site or central imaging, Yunu reduces study staff and project management time by 80% and reduces radiologist reading times by 50%. This ensures accuracy and data access for every patient, on every trial, running at every site.

Jeffrey Sorenson, CEO of Yunu

Central review should be used to validate, not rescue a trial. When every site read is protocol-guided and traceable down to the pixel, you stop losing patients and days to preventable errors.

Proof Where It Counts: Patients, Time, Budget

Moving quality upstream yields an impact in real-world examples.

Eligibility at screening: A centralized, protocol-guided screen read prevents baseline errors that force re-enrollment. This has an event rate some sponsors now admit is higher than expected and statistically painful.

PFS integrity in open-label trials: When site reads are standardized upfront, you reduce the local versus central divergence, ensuring the survival curve becomes clear of workflow-induced errors and is only reflective of biology and treatments.

Operational lift: The combination of embedded response criteria, structured lesion tracking, and live oversight significantly reduces the email-PDF-spreadsheet chaos that hinders imaging, making it the slowest and least traceable data stream in the study. Centers report shifting from months-late error detection and backlogged work to real-time reads with in-flight error checks.

Where BICR Still Belongs

After you move quality upstream, BICR becomes what it should have been all along:
A confirmatory layer on primary endpoints
A consistency check across sites and readers
A documented independent view for regulators

You keep the rigor. You drop the regrets.

And because site investigator vs. central review disagreement won’t disappear entirely (nor should it), central review remains essential for fairness and reproducibility. A single platform that can perform both harmoniously and accurately is a real advantage.

The Mandate

If your imaging process assumes late correction, you will keep paying late-correction prices—financially, operationally, and ethically.

The better path is now available: real-time, protocol-specific site reads with traceability back to the pixels, with central reviews playing a role that confirms, not salvages.

Sponsors that make this shift report fewer censored patients for preventable reasons, fewer repeat scans, fewer re-screenings, fewer inspection headaches, and fewer days lost to incidents that should never have occurred. That’s the difference between placing quality at the end of the process and placing it upstream where it matters most.

Moe Alsumidaie

Website | Moe Alsumidaie is Chief Editor of The Clinical Trial Vanguard. Moe holds decades of experience in the clinical trials industry. Moe also serves as Head of Research at CliniBiz and Chief Data Scientist at Annex Clinical Corporation.