Understanding the Blinded Read: Clinical Trial Imaging Series, Part 1

Joseph Pierro, M.D. and Brett A. Hoover | |

Imaging has changed beyond recognition in the last forty years, with the evolution of new imaging techniques and applications – all of which brings new design challenges to understand, risks to mitigate and opportunities to improve clinical trials. With the sophisticated and diverse range of imaging technology now in use often in combination and increasing importance of medical imaging endpoints in clinical design, a sponsor’s approach to imaging has never been more important.

This series on medical imaging trial design reviews the factors that sponsor should consider as they approach the collection, analysis, and submission of imaging data to support clinical trial endpoint evaluation. Here we present why the blinded read is so important, and how the various read categories and methods can impact your clinical trial imaging data.

Defining the Blinded Read

A blinded read defined as a Blinded Independent Central Review (BICR) is the practice of collecting all study images acquired at the investigative sites and presenting these data to separate groups of readers for evaluations. In most cases, end point clinical trial images are read by radiologists, but depending on the therapeutic area, and read requirements for subspecialty trained experts, the images may also be read by cardiologists, oncologists, orthopedic surgeons, nuclear medicine physicians or other experts as required. Hence the term ‘reader’ has become the all-encompassing term used to describe these individuals, and this is the term the FDA uses in their guidance documents.

Since all blinded reads are not created equally, it’s important to understand the nuances of each as they apply to your clinical trial designReads can be categorized into three categories (Table 1):

Eligibility Reads

These determine whether the study inclusion and exclusion criteria have been satisfied from an imaging standpoint, and the data supports a sponsor, site investigator or CRO’s ultimate decision to enroll or exclude a subject from the trial.

Efficacy Reads

These generate data in support of treatment efficacy endpoints. This is the most common category of read which may be performed at the end of the subject’s enrollment or at each imaging visit to monitor response or confirm progressive disease (PD).

Safety Reads

When required, these determine whether pre-defined deteriorations in body systems or disease activity have been met, and may or may not be related to the primary efficacy variable of the drug or medical device. Often safety reads assess for the presence or absence of a significant clinical findings that have been previously identified within a therapeutic drug category.

Three Categories of Blinded ReadsExamining Different Methods of Blinded Reads

Once it is determined what category of read a clinical trial will use, it’s key to understand how your data will be interpreted and which read methodology will be implemented as a consequence. Summarized here are the commonly used 7 BICR methodologies:

1. Consensus Read

Normally 3 readers possibly 4 or 5 in the same room, with same review monitors and lighting, sit and discuss the read and come to a general conclusion of result. Utilizing the internet and cloud based review platforms, consensus reads can also be achieved remotely via WebEx.

2. Paired Read, No Adjudication

Different reader results are averaged or two sets of results provided; often used in some musculoskeletal indications.

3. Paired Read, with Forced Adjudication

Reader 1 (R1) and Reader 2 (R2) each read 100% of the images, and Reader 3 (R3) adjudicates the differences. If there’s a difference between R1 and R2, the adjudicator tie-breaks by selecting the results he/she most agrees with, i.e. either R1 or R2.

4. Paired Read, with Open Adjudication

R1 and R2 each read 100% of the images, and R3 adjudicates. If there’s a difference between R1 and R2, the adjudicator independently reads the imaging exam, and their results are final (may or may not be the same as R1 and R2).

5. Paired Read, Re-Read, Then Adjudication

R1 and R2 reads are compared.  If different the first time the images are read, they are sent back to the readers. Only if different after a second review will the adjudicator be brought in to perform either open or forced adjudication, depending on the study read design.

6. Paired Read and Adjudication for Large Studies

This involves the creation of multiple pairs of independent readers to help with overall image evaluation throughput. If adjudication is required, the image goes to a single reader not associated with the reader pair.

7. Pseudo-Paired Read and Adjudication

For large studies, a pool of readers is employed who can be randomly assigned to either read or adjudicate, depending on the pairing. This “pooled” approach helps with overall image evaluation throughput.

As one can appreciate from this list, not all blinded reads are created equally.  It is important to understand the nuances of the various read categories and methods, and how they apply to your trial.  Some of the read paradigms are therapeutic area- or indication-specific, and your imaging experts should know the differences and how to best apply them relative to your study’s therapeutic area, indication, imaging modality and efficacy / safety endpoints. Your imaging experts will also be able to share their experience from dealing with regulatory agencies and the accepted reader methods.

Check back soon for the next installment to the series, which focuses on data management considerations in medical imaging trials.


Click here to learn how the right technology can significantly improve your clinical trial results.

Joe Pierro, M.D. is the Medical Director of Imaging at ERT and Brett A. Hoover is the Director of Imaging Product Management at ERT.