Centralized versus Site-based Imaging Reads in Clinical Trials

Joseph Pierro and David Raunig | |

Curious about the differences between centralized and local site-based reads? Here we explore the basis, impact, and value of each.

Site-based Reads

In general, site-based reads are performed in the clinical institution where physicians have access to subjects’ clinical information and resources, which may include other consultants, past imaging exams, or additional clinical reports or tests. Reviews are conducted at each clinical site and performed by the radiologist on staff that day with readers having limited knowledge of the clinical protocol or read requirements. For example, the same reader typically will not read images for the entire study or may not even be the reader evaluating all of the time points for a single subject.

Typically in clinical trials, investigative sites contract with or receive imaging services from several different imaging groups in their geographic area.  When you consider the number of clinical sites, imaging groups and radiologists that may provide study interpretations, you’ll realize the staggering number of study readers possible, all with a wide variety of training, experience and opinions. This results in a higher level of endpoint variability and leaves no practical way to evaluate for bias or aberrant readers.

Centralized Reads

Centralized reads are performed by readers blinded to clinical information unrelated to the reads, thereby reducing bias and driving assessments based on image attributes required by the protocol. Reads are conducted within a more controlled environment and with systems designed to ensure study images comply with protocol requirements and are read using the same imaging software or equipment.

Additionally, centralized reads use a smaller number of readers who are selected based on training, experience, and knowledge of the specific subject population and who best align with study requirements. Given that readers – all readers – commit diagnostic errors approximately 30% of the time, it becomes important to be able to assess, monitor, limit and correct those errors when evaluating treatment efficacy or safety.1,2

Research Says…

Regardless of the therapeutic indication, the research indicates that site reads differ from central reads with differences typically in the range of 5-30%. However, this difference may be even greater for complex therapeutic areas. There are many reasons for this difference and the FDA has provided guidance documents, such as “Clinical Trial Imaging Endpoint Process Standards”, outlining approaches to optimize the quality and collection of clinical trial imaging data.3

It is well understood that sonography and echocardiography procedures are heavily operator dependent and that the person performing the exam and measurements may introduce a level of variability and human error.4 A large number of research articles exist which demonstrate site reader variability and the superior precision achieved with the use of echocardiography core labs (ECLs).

In a multicenter, serial echocardiographic study paper, Hole et al. examined the differences obtained between site and centralized assessments and reported that “only measurements in the core laboratory had significant prognostic value for subsequent clinical endpoints. These results and, importantly, the consensus among European and North American echocardiologists strongly support the use of a core laboratory in studies employing echocardiographic measurements”.5,6,7

What Regulators Think

Wide variability in both quantitative and qualitative echocardiography assessments exists among readers and is recognized in recent guidance publications offered by both The European Society of Cardiology (ESC) and The American Society of Echocardiography (ASE). These guidelines looked at sources of variability and identified quality control and training methods, specifically training consistency, as common factors influencing the overall quality of clinical data over time. They also point out that the ECL should be very active and involved in providing a set of standardized training sessions to sites including periodic and continual training follow-up sessions to avoid issues with reader performance or drift which may occur during the duration of the trial.

Per the ASE, “The main purposes of an ECL are to define and standardize processes for image acquisition and analysis, education and training of sonographers and over readers, and image and data management, as well as actually performing the image analysis itself. These activities should be part of a broad program of quality assessment and improvement for all aspects of ECL operations, as well as study-specific QA activities”.6

Let’s Talk Budget

Clinical study teams must weigh the financial impact of proceeding with local site versus centralized reads while acknowledging that operational implementation of the society recommendations may result in comparable or higher costs. The standardized processes, specific training and reader requirements typical when using an ECL result in lower variability that could be reflected in smaller studies when the different variabilities are considered for estimating subject sample sizes.

Cristofanailli et al. provided results from both site-based and centralized readers with supplemental data that demonstrated the consequences of the increased variability from sites.8 In turn, smaller sample sizes led to shorter subject recruitment periods, less error in determining the occurrence of an event and, in general, stronger results. An important factor to also consider is the ability to monitor readers for performance whereas any performance evaluation for site readers is neither practical nor possible.  

click here to learn how the right technology can improve your clinical trial results

Joseph Pierro is the Medical Director of Imaging at ERT and David Raunig is the Senior Principal Imaging Statistician at ERT.


  1. Kundel, H. L. “History of research in medical image perception.” J Am Coll Radiol. 2006; 3(6): 402-408.
  2. Busby, L. P., J. L. Courtier and C. M. Glastonbury. “Bias in Radiology: The How and Why of Misses and Misinterpretations.” RadioGraphics. 2018; 38(1): 236-247.
  3. FDA Guidance Document: Clinical Trial Imaging Endpoint Process Standards Guidance for Industry [https://www.fda.gov/regulatory-information/search-fda-guidance-documents/clinical-trial-imaging-endpoint-process-standards-guidance-industry]
  4. Oh J. Guest Editorial Is Core Laboratory Essential for Using Echocardiography in Clinical Trials? Controlled vs Random Error. Eur J Echocardiography (2002) 3, 245–247.
  5. Hole TOtterstad JE, et al. Differences between echocardiographic measurements of left ventricular dimensions and function by local investigators and a core laboratory in a 2-year follow-up study of patients with an acute myocardial infarction. Eur J Echocardiogr.2002 Dec;3(4):263-70.
  6. Douglas PS, et al. ASE Expert Consensus Statement Echocardiographic Imaging in Clinical Trials: American Society of Echocardiography Standards for Echocardiography Core Laboratories: endorsed by the American College of Cardiology Foundation. Journal of the American Society of Echocardiography. 2009 Jul 1;22(7):755-65.
  7. Galderisi, M. et al. “Recommendations of the European Association of Echocardiography How to use echo-Doppler in clinical trials: different modalities for different purposes.” European Journal of Echocardiography. 20111; 12(5): 339-353.
  8. Cristofanilli, M., et al. “Fulvestrant plus palbociclib versus fulvestrant plus placebo for treatment of hormone-receptor-positive, HER2-negative metastatic breast cancer that progressed on previous endocrine therapy (PALOMA-3): final analysis of the multicentre, double-blind, phase 3 randomised controlled trial.” The Lancet Oncology. 2016; 17(4): 425-439.