Timothy Callahan, PhD

	In 1949, Norman J. Holter and Joseph A. Generelli, building on previous work, described a “radiocardiogram” and a “radio-encephalograph” in the 

, thus introducing a new technology to medical science. The goal of these recordings was to “….inquire into how various bodily functions are modified by exercise, excitement, emotion, or other activity….”

	Since then, the “Holter” recording methods have been improved by advances in miniaturization and recording technologies. The state-of-the-art in recording media has gone from magnetic tape to digital storage, with some recordings being transmitted to central processing centers using Bluetooth technologies, making almost real-time analysis possible. And, in the last decade, technology has advanced to the point where multiple-day recordings are possible.

	Use of single-day recordings reached its zenith prior to the publication of the Cardiac Arrhythmia Suppression Trial (CAST) findings. CAST tested the hypothesis that suppression of ventricular arrhythmias as seen on Holter recordings in patients with myocardial infarction would decrease mortality. The CAST study showed an increased mortality with treatment as opposed to placebo.

 After the CAST findings were announced, however, the use of Holter recordings was dramatically reduced, as the need for finding simple ventricular arrhythmias was attenuated.

	Up to this point, single-day data yielded a false sense of safety or concern, possibly because, as DiMarco noted, of the considerable day-to-day variability of arrhythmia and myocardial ischemia.

 The evanescent nature of arrhythmias was described by Corday, who compared a 10-hour Holter recording with a “conventional ECG”.

 What we have come to know about “normal” has evolved over the years as recording times have increased.

	The traditional way to describe “normal” on Holter recordings is to explain the abnormalities encountered. This includes any single or run of ectopic beats, rhythms, and sometimes included an ST-segment analysis or, more rarely, a QT-interval analysis. Judgments were made based on these ectopic counts and waveform descriptions. Additionally, normal healthy subjects might manifest benign arrhythmias not seen in other populations. These confounders make any description of “normal” difficult.

	As technology improved, recording periods were sometimes prolonged to two or three days.  Now, technology has advanced even further and recordings can be taken over weeks or months. It is, once again, time to rethink “what we know.”

	One of the challenges for those who write clinical trial protocols is to collect enough data to ensure that evanescent arrhythmias are not considered “drug-related.” This will most likely be done in two ways: 1) over-collection of screening and baseline data, 2) redevelopment of normal ranges.

	Due to the expense of clinical trials, cost-cutting has become very important in the drug-development process. The risk of cutting costs is that we collect homogenous data, and infrequent, otherwise unimportant abnormalities might take on an unwarranted significance. Indeed, the goal is to avoid “problematic” data, or data that might be considered an adverse event. An alternative to collecting limited data is to over-collect baseline data. Certainly, if we have 30 days of Holter data at baseline, we might be in a better position to determine a subject’s “normal” profile.

	From long-term data sets, new normal ranges can be developed. If it’s known that sinus pause of three seconds or 5-beat runs of VT are not uncommon, perhaps seeing these events in the drug-development process would not be considered an adverse event warranting further, expensive follow-up. The logical extreme is that some good drugs might not be killed in the development process.

	To conclude, technology has advanced enough that Holter’s original vision can be expanded to include time as a new variable. As technology changes, so must the prism of our analysis. What we thought was abnormal might now be considered normal or at least a normal variant. We need to collect more robust data to answer these questions.

, PhD, is Chief Scientific Officer, Biomedical Systems; 

, MD, is Chief Medical Officer, Biomedical Systems

	1. Holter NJ, Generelli JA. Remote Recording of Physiologic Data by Radio. 

	2. The Cardiac Arrhythmia Suppression Trial (CAST) Investigators. Preliminary Report: Effect of Encainide and Flecainide on Mortality in a Randomized Trial of Arrhythmia Suppression after Myocardial Infarction. 

	3. DiMarco JP, Philbrick JT. Use of ambulatory electrocardiographic (Holter) monitoring. 

	4. Corday E, Bazika V, Lang TW, Pappelbaum S, Gold H, Bernstein H. Detection of Phantom Arrhythmias and Evanescent Electrocardiographic Abnormalities. 

Articles

With significant technology advances in ECG recording through the years, the need to expand on traditional monitoring baselines-and include new variables such as time when designing clinical trial protocols-is important.

Long-term Electrocardiographic Monitoring: Time to Rethink What We Know?

	After nearly a decade of implementation and three rounds of “Question and Answer” documents to clarify procedural and scientific questions surrounding the ICH E14, the US FDA, in open meetings, stated the Thorough QT (TQT) trial “must” be replaced. This marks an extraordinary reversal of regulatory opinion. It can be estimated that over 250 TQT trials have been conducted costing drug developers many millions of dollars.

	While the TQT “process” (including pre-clinical investigations) has, unquestionably, kept QT prolonging drugs off the market, the process has tended to treat all non-antiarrhythmic drugs the same. It is well known, however, that not all QT-prolongation is pro-arrhythmic. Once a compound is shown to prolong the QT-interval drug developers have a very difficult task proving their QT-prolonging compound is safe.

	This review covers both the problems with the TQT trial and what is being considered for the future of cardiac safety.

	When administering a new chemical entity (NCE) to human beings and subsequently assessing the pro-arrhythmic potential, it is important to understand whether the problem resides with the parent drug, the metabolite, or a drug-drug interaction. The problem might also be a combination of these factors. For example, terfenidine was a non-sedating antihistamine that caused Torsade de Pointes (TdP), a sometimes fatal polymorphic ventricular tachycardia. Investigations revealed that the concentration of terfenidine (Seldane) in the blood increased when given along with ketoconazole, an anti-fungal that is metabolized through the CYP450 pathway.

 The active metabolite, fexofenadine, did not itself prolong the QT interval or cause TdP.

	Since the early 1990’s, TdP increased in the consciousness of industry and the regulators.  A simple MEDLINE search shows the number of articles between 1990 and 1999 averaged less than 100 per year, but increase to an average of over 150 articles per year from 2000 through 2013.

	During the mid-to-late 1990s, as scientific awareness was raised, so was the concern of the regulators. The International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) has introduced a number of guidance documents that detailed both clinical and preclinical testing for the potential of the new chemical entity to be pro-arrhythmic. These include:

          Safety Pharmacology for Human Pharmaceuticals

         The Non-clinical Evaluation of the Potential for Delayed Ventricular Repolarization (QT Interval Prolongation) by Human Pharmaceuticals

          The Clinical Evaluation of QT/QTc Interval prolongation and Proarrhythmic Potential for Non-Antiarrhythmic Drugs

	While the ICH S7A discusses more than just the cardiovascular system, it reminds us that cardiac safety involves more than the QT interval. It recommends that heart rate and blood pressure be tested with possible follow-up studies involving indices such as cardiac output, ventricular contractility, and vascular resistance.

	The ICH S7B is the preclinical correlate of the ICH E14. The S7B reviews the non-clinical evaluation for delayed ventricular repolarization potential. It calls for both in-vitro and in-vivo investigations. The in-vitro investigations are mainly the hERG channel assay and Purkinje fibers assessment. The acronym hERG stands for “human ether-a-go-go related gene.”  The gene is responsible for encoding the inward potassium rectifying channel (Ikr). Blockage of this channel is correlated with the prolongation of the QT-intervals.

	The in-vivo assessments are focused on the ECG from intact animals such as the dog and monkey, as well as some others. Heart rates and other parameters can also be evaluated from the animal model.

	The thorough QT trial is covered in the ICH E14. The E14 was discussed for a number of years before it was endorsed in 2005. Subsequent updates to regulatory thought have been published in question-and-answer documents. The first such document was released in 2008. The third document (r2) was released in 2014.

 These documents have clarified issues such as automated measures, baselines, gender analyses, as well as others.

	The E14 is focused on ventricular repolarization in human trials. It calls for a dedicated trial and uses the corrected QT-interval (QTc) as a surrogate marker for the risk of TdP. It is important to note that the E14 does not specifically say that the QT interval causes TdP, only that the “delay in cardiac repolarization creates an electrophysiological environment that favors the development of cardiac arrhythmias…”

	The FDA went a step further and formed the QT Interdisciplinary Review Team (QTIRT) to review both protocols and data around the QT issue. This group has been instrumental in giving pre-study advice and reviewing thorough QT data.

	However, it was recognized early in the ICH E14 era that NCE’s can have an effect on more than just the QT-interval on the electrocardiogram. The thorough and intensive QT trial also reports data regarding the QRS duration and the PR interval. We know that these intervals are encoded by a series of sodium and calcium channels, respectively.

	Twelve-lead electrocardiograms are a wealth of information regarding the functioning of the heart. Features, both waveforms and intervals on the ECG correlate to the structure and function of different segments of the myocardium.  The table below shows some of these structures and functions and the main abnormalities seen in clinical trials.

	As we have come to know the basics of the TQT trial, most of these trials fall into a fairly narrow set of parameters. The TQT trial is usually a crossover or parallel study with four dosing levels including a therapeutic dose, a supra-therapeutic does, placebo, and a positive control.  Intrinsic variability of the QT-interval is controlled by “the collection of multiple ECGs at baseline and during the study.”

 For most TQT studies, this has come to mean “triplicate” ECGs at each time-point. Depending on the study design, there can be as few as approximately 40 completed subjects for a crossover design or more than 180 completed subjects for parallel design. It is worth noting that most TQT trials use normal healthy volunteers.

	Efforts to reduce the sample size for TQT trials have included collecting more ECGs and increasing data quality by standardizing on-site procedures. The outcome of these efforts is to reduce the standard deviation associated with the QT-interval. While these efforts are important, there is a practical need to ensure a minimum number of subject be recruited in a TQT trial; the positive control. Every TQT trial needs to show assay sensitivity. If too few subjects are recruited, the statistical test for assay sensitivity might not be met, thus invalidating the TQT trial.

	Statistically the thorough QT trial seeks to exclude the upper bound 95% confidence interval of 10 ms, which is done by comparing the on-treatment QTc against the baseline and placebo corrected QTc (ΔΔQTc) at each study time point. In addition the plasma concentration is compared and modeled against the QTc.

	While the thorough QT trial is mainly interested in the corrected QT interval, the FDA generally requests data regarding other ECG measurements. Because of multiple ion channel involvement the ECG study report should include data from PR interval and QRS duration measurements. In addition the heart rate is measured and analyzed.

	In many instances it is unclear the role of multiple ion channel blockade plays in the safety analysis. For example, both verapamil and ranolazine are multiple ion channel blockers that have been deemed safe from a ventricular repolarization standpoint. Conversely the multiple ion channel blocker propoxyphene has been removed from the market because of changes to the ECG including prolonged PR interval widened QRS complex in prolonged QT interval.

	As mentioned previously, the exposure-QTc (ER) relationship is also a large part of the thorough QT trial. This analysis was proposed by the FDA and has been adopted by industry. The standard approach is outlined by Garnett.

	This approach takes a mixed effects modeling paradigm to plot the concentration of the drug versus the QTc. All data are pooled across dosing groups. A regression line is then plotted against the ΔΔQTc to see whether or not there is an increase in the QTc change from baseline and placebo as the concentration of the drug increases.

	An example of the method is shown in the figure, below. The plasma concentration is shown on the x-axis versus the ΔΔQTcF on the y-axis. In this case, the compound being tested is moxifloxacin. As is expected and shown on the figure the slope of the regression line is positive meaning that as the concentration of moxifloxacin increases the QTcF change from baseline and placebo also increases.

	When used in early phase development, the ER model can provide data on a wider range of exposures than is tested in the TQT trial. This allows for better modeling and can give the drug-developer early information on electrocardiographic responses to the compound.

	However, the thorough QT study is not without problems. One major issue that has been known is that prolongation of the QT interval does not necessarily mean a higher risk of TdP. This means that some good drugs that prolong the QT interval might be abandoned or given a warning label reducing their use because of benign QT prolongation.

	Another problem is the expense of the thorough QT trial. While the cost of the thorough QT trial might be less than 1% of the overall cost to bring a drug to market it is still a major expense to many companies including small bio-techs who don't have the resources to spend. So, drugs are being abandoned because of hERG blockade or QT prolongation, or they can be continued with an added expense of intensive testing and large Phase III trials, limiting the return-on-investment.

	Although bad drugs are not getting on the market, and thus the TQT trial has been successful from that point of view, the TQT trial has not given us the information we hoped; namely, a clear link from prolonged QT to TdP. So we've learned that the hERG assay is a very good biomarker for the prolongation of the QT interval but the QT interval is a very bad biomarker for finding TdP.

	Therefore, the regulators have decided it's time for a new paradigm. First, we must replace the thorough QT trial. Ongoing efforts involve: preclinical review of multiple ion channels, collecting ECGs in early phases, and placing a premium on ER relationships.

	A new preclinical paradigm called CIPA (comprehensive in-vitro pro arrhythmic analysis) will include multiple ion channel evaluations (MICE), isolated cardiac myocytes, and computer modeling.

	Both verapamil and ranolazine are hERG blockers without causing TdP. These drugs also block more than just the hERG channel. It is postulated that the multiple ion channel effect impart a degree of cardiac safety not seen in other drugs that cause TdP.

	To determine whether or not studying addition ion-channels could help discriminate safe and non-safe compounds Kramer et al. studied 32 torsadogenic and 23 non-torsadogenic drugs.

 Assays for the Cav1.2 and Nav1.5 channel currents were conducted in addition to the hERG assay. The details of the study are beyond the scope of this discussion, but the authors state that the MICE model is better at finding drugs that cause TdP than the single hERG assay.

	Clinically, more emphasis is being placed on exposure-response (ER) modeling. The Cardiac Safety Research Consortium (CSRC) and IQ-PHRMA have undertaken a study to determine whether or not ER modeling can replace the TQT in clinical development.

 The study includes both QT prolonging drugs and a negative control.  The results are scheduled to be presented in December, 2014.

	One item of interest is the use of a positive control. In TQT studies, a positive control has been used to ensure the ability of the process to find a small increase in the QT interval if one is present. The most widely used positive control is moxifloxacin. In early phase studies, using a positive control might not be statistically sound. Instead, Malik proposed use of statistical methods based on the variances of the data to ensure data quality.

 This, or other measures might be useful to replace the positive control requirement. By adopting a statistical model for data quality, a high-quality study can be ensured without unnecessarily exposing subjects to pharmaceutical compounds.

	In the pre-E14 era, ECGs were recorded and measured on paper. The quality of the ECG was suspect as collection procedures, equipment, storage, and timing of the ECGs were not standardized. The focus on the QT-interval caused the industry to re-evaluate the importance of standardization. 

	Current “best-practices” include the recognition that ECG intervals are both time- and heart- rate dependent. Collecting ECGs for measurement of the intervals involves having a subject rest for a period of time (usually 5 or 10 minutes) and comparing ECGs in such a way as to minimize the effects of circadian rhythms.
	
	Measurement and reading of ECGs are now done on a computer screen to maximize precision.  Before digital ECG readings, the precision of the measurement was based on paper-speed and line-width, current precision is based on sampling rates and on-screen pixel resolution.

	Finally, the digital age has seen open-architecture ECG storage. The XML ECG can be read by any number of ECG viewers and no longer subject to the manufacturers’ proprietary storage schema. This allows for a level of transparency whereby different reviewer can see and measure the same ECG with very few technical hurdles.

	In a little over a decade, the QT interval has come full-circle, from a regulatory “darling” to “just another feature of the ECG.” However, before we bury the QT interval in the historical archives, it is worth noting a few of the positive results of the analysis of this measurement.

	Clinically, the QT interval is now in the consciousness of clinicians. The American College of Cardiology and the American Heart Association released a position paper on preventing TdP in the hospital setting.

 Taking this paradigm further, hospitals such as the Indiana University Health Methodist Hospital have implemented a computerized clinical decision support system (CDSS) to help reduce the prescribing QT-prolonging medications to those patients with known risk factors of TdP.

 They’ve shown that this type of system can influence prescribing patterns in the hospital.

	To conclude, the era of widespread use of the TQT trial appears to be ending. While it is difficult to foresee regulatory agencies totally abandoning a paradigm that has met with some success, shifting the burden of proof from a TQT trial to a combination of pre-clinical data and early phase human data might be beneficial.  It is anticipated that the new paradigm will be less costly for the sponsor while yielding more scientifically valid data regarding proarrhythmic risk. However, it is important to note that 12-lead electrocardiograms yield more information, including all-cause mortality, important to drug development than just the QT-interval.

	The author would like acknowledge Richard Kovacs, MD, for his guidance and review

	Tim Callahan, Chief Scientific Officer, Biomedical Systems

		Honig PK, Wortham DC, Zamani K, Conner DP, Mullin JC, Cantilena LR (1993). Terfenadine-ketoconazole interaction. Pharmacokinetic and electrocardiographic consequences. JAMA 269(12):1513-1518	

		Honig PK, Wortham DC, Lazarev A, Cantilena LR (1996). Grapefruit juice alters the systemic bioavailability and cardiac repolarization of terfenadine in poor metabolizers of terfenadine. J Clin Pharmacol  36(4):345-351	

		ICH S7a. (2000) Safety Pharmacology for Human Pharmaceuticals.	

		ICH S7b. (2005) The Non-clinical Evaluation of the Potential for Delayed Ventricular Repolarization (QT Interval Prolongation) by Human Pharmaceuticals	

		ICH E14. (2005) The Clinical Evaluation of QT/QTc Interval prolongation and Proarrhythmic Potential for Non-Antiarrhythmic Drugs.	

		Brown AM. (2005). HERG block, QT liability and sudden cardiac death. Novartis Found Symp  266:118-31.	

		ICH E14 Q&As R2 (2014). Questions & Answers: The Clinical Evaluation of QT/QTc Interval prolongation and Proarrhythmic Potential for Non-Antiarrhythmic Drugs.	

		http://www.fda.gov/Safety/MedWatch/SafetyInformation/SafetyAlertsforHumanMedicalProducts/ucm2 34389.htm	

		Garnett CE, Beasley N, Bhattaram VA, Jadhav PR, Madabushi R, Stockbridge N, Tornoe CW, Wang Y, Zhu H, Gobburu JV. 2008. Concentration-QT relationship plays a key role in the evaluation of proarrhythmic risk during regulatory review. J Clin Pharmacol 48:13-18.

		Kramer J, Obejero-Paz CA, Myatt G, Kuryshev YA, Bruening-Wright A, Verducci JS, Brown AM. 2013. MICE Models: Superior to the hERG Model in Predicting Torsades de Pointes. 3:1-7.	

		Darpo B, Sarapa N, Garnett C, Benson C, Dota C, Ferber G, Jarugula V, Johannesen L, Keirns J,  Krudys K, Ortemann-Renon C,  Riley S, Rogers-Subramaniam D, Stockbridge S. 2014. Ann.Noninv.Card.  The IQ-CSRC prospective clinical Phase 1 study: "Can early QT assessment using exposure response analysis replace the thorough QT study?". 19(1):70-81	

		Malik M, Zhang J, Johannesen L, Hnatkova K, Garnett, C. 2011. Assessing electrocardiographic data quality and possible replacement of pharmacologic positive control in thorough QT/QTc studies by investigators of drug-free QTc stability. Heart Rhythm. 8(11):1777-1785.

		Drew BJ, Ackerman MJ, Funk M, Gibler WB, Kligfield P, Menon V, Philippides GJ, Roden DM, Zareba W. 2010. American Heart Association Acute Cardiac Care Committee of the Council on Clinical Cardiology; Council on Cardiovascular Nursing; American College of Cardiology Foundation. Prevention of torsade de pointes in hospital settings: a scientific statement from the American Heart Association and the American College of Cardiology Foundation. JACC; 55:934-947.	

		Tisdale JE, Jaynes HA, Kingery JR, Overholser BR, Mourad NA, Trujillo TN, Kovacs RJ. 2014. Effectiveness of a clinical decision support system for reducing the risk of QT interval prolongation in hospitalized patients. Circ Cardiovasc Qual Outcomes. 7:381-390.

The US FDA, in open meetings, stated the Thorough QT trial "must" be replaced.

The Future of QT in Clinical Trials

The sponsor of a recent Phase II global study, conducted at more than 50 sites, invested in a centralized digital pathology core lab to ensure a homogenous and representative study population. For this study, the lab, powered by Biomedical Systems, worked to confirm the pathological diagnoses of patients' disease states.  

The investment proved advantageous, and the findings were startling: 30% of patients submitted for central review did not meet the inclusion criteria. Failure to properly classify its patient population can leave sponsors with a significantly underpowered study—rendering the drug manufacturer with inaccurate efficacy results, a failed study, and a heavy price tag. 

To understand how the study sites, which relied on patient medical histories combined with site-prepared histology biopsy slides, approved patients who did not meet the inclusion criteria requires an understanding of the digital pathology evolution. 

Before the advent of slide digitization, pathologists reviewed specimens on glass slides under microscopes. To secure multiple pathologists' interpretations, labs would prepare several sets of slides from the patient's tissue block—each slide containing a different level of tissue shipped to the reviewer. Variance in tissue levels triggered variability among inter-reader interpretations of the specimens. Subsequently, sponsors began leveraging digital pathology to provide reviewers with identical digital images of patients' tissues, effectively reducing variability. 

In the referenced Phase II study, variability could explain the 30% discrepancy between internal site review and the central review process. With the study, pathology slides were reviewed at more than 50 sites; each of the sites employs its own pathologists. In addition, each site offers its own equipment and methods for slide collection and patient history documentation. 

With centralized digital pathology, the study and its readers must follow rigorous, standardized processes focused solely on inclusion criteria. Pathologists may only assess the morphology depicted in the digital slide image sets, which reduce variability and bias.  

For the Phase II global study, readers were trained on standardized equipment, scoring methodology, and the electronic case report form application. Each reader in the study viewed the exact same digital slide image sets. Readers followed proven project management practices and systems to ensure their assessment results were directly linked to the evaluation. The electronic case report form guided each of the readers through the same protocol-specific evaluation and grading scale.  

After a reader completed each required field, he or she would enter a given password to apply a digital signature and date/time stamp to the completed record. The completed record was sequestered, preventing the reader from returning to the record or the digital slide image set. These processes effectively reduced variability, promoted homogenous analysis methodology across all readers, and ensured complete records were provided in a format conducive to statistical analysis. 

			 By centralizing a study, sponsors and its study partners can typically decrease the number of pathologists needed to evaluate patients for inclusion. This reduction in pathologists lessens the variability associated with the inclusion assessments. In the study, the centralized lab effectively worked to significantly decrease the number of pathologists evaluating patients. Reviewer reduction was coupled with standardization—providing uniform training, equipment, and review processes to improve patient inclusion percentages and reduce variability. 

 In the study, the independent review, which is inherently part of the centralized review process, proved effective in ensuring only patients who met the protocol inclusion criteria were enrolled in the study.  

In cases of site-selected inclusion, site pathologists review patients' history, test results, and histology primarily for patient care. Study inclusion criteria often serve as the site's secondary goal. The pathologists' expanded patient knowledge, which is important for patient care, can potentially hinder the clinical trial process by injecting unintended bias into the patient inclusion evaluation.  

The readers were required to follow  a centralized digital pathology process, and only focused on the morphology represented within the digital slide images. For the study, readers were blind to all extraneous patient information.  

The impact of an underpowered study can cost drug manufacturers millions of dollars and their reputations. Centralized digital pathology, in this example, reduced bias and variability while increasing the probability of accurately assessing efficacy and statistically relevant data. 

Tim Callahan, PhD, Chief Scientific Officer, Biomedical Systems. Member of Applied Clinical Trials' Editorial Review Board.

The sponsor of a recent Phase II global study, conducted at more than 50 sites, invested in a centralized digital pathology core lab to ensure a homogenous and representative study population.

Central Lab Review Prevent Underpowered Studies

The sponsor of a recent Phase 2 global study, conducted at more than 50 sites, invested in a centralized digital pathology core lab to ensure a homogenous and representative study population.  For this study, the lab, powered by Biomedical Systems, worked to confirm the pathological diagnoses of patients’ disease states.   

The investment proved advantageous, and Biomedical Systems’ findings were startling: 
  30 percent of patients submitted for central review did not meet the inclusion criteria.  Failure to properly classify its patient population can leave sponsors with a significantly underpowered study—rendering the drug manufacturer with inaccurate efficacy results, a failed study and heavy price tag.

To understand how the study sites, which relied on patient medical histories combined with site-prepared histology biopsy slides, approved patients who did not meet the inclusion criteria requires an understanding of the digital pathology evolution.

Before the advent of slide digitization, pathologists reviewed specimens on glass slides under their microscopes.  To secure multiple pathologists’ interpretations, labs would prepare several sets of slides from the patient’s tissue block—each slide containing a different level of tissue shipped to the reviewer.  Variance in tissue levels triggered variability among inter-reader interpretations of the specimens.  Subsequently, sponsors began leveraging digital pathology to provide reviewers with identical digital images of patients’ tissues, effectively reducing variability.

In the referenced Phase 2 study, variability could explain the 30 percent discrepancy between internal site review and the central review process.  With the study, pathology slides were reviewed at more than 50 sites; each of the sites employs its own pathologists, who innately apply their schooling, training and personal experiences to the review process.  In addition, each site offers its own equipment and methods for slide collection and patient history documentation.

With centralized digital pathology, the study and its readers must follow rigorous, standardized processes focused solely on inclusion criteria.  Pathologists may only assess the morphology depicted in the digital slide image sets, which reduce variability and bias.  

For the Phase 2 global study, Biomedical Systems relied on Aperio equipment and its applications, in combination with its own validated proprietary systems, to manage digital slides and facilitate simultaneous independent reviews from anywhere in the world.  Biomedical Systems trained readers on the standardized equipment, scoring methodology and the Electronic Case Report Form application.  Each reader in the study viewed the exact same digital slide image sets from their various locations.  Readers followed proven project management practices and systems to ensure their assessment results were directly linked to the evaluation.  The Electronic Case Report Form guided each of the readers through the same protocol-specific evaluation and grading scale.  

After a reader completed each required field in the Electronic Case Report Form, he or she would enter a given password to apply a digital signature and date/time stamp to the completed record. The Electronic Case Report Form application sequestered the completed record, preventing the reader from returning to the record or the digital slide image set.  These processes effectively reduced variability, promoted homogenous analysis methodology across all readers and ensured complete records were provided in a format conducive to statistical analysis.


By centralizing a study, sponsors and its study partners can typically decrease the number of pathologists needed to evaluate patients for inclusion. This reduction in pathologists lessens the variability associated with the inclusion assessments.  In the Phase 2 study, the centralized lab effectively worked to significantly decrease the number of pathologists evaluating patients.  Biomedical Systems coupled the reviewer reduction with standardization—providing uniform training, equipment and review processes to improve patient inclusion percentages and reduce variability.


According to the U.S. Food and Drug Administration[1], independent reviews can reduce bias in clinical trials.  In this study, the independent review, which is inherently part of the centralized review process, proved effective in ensuring only patients who met the protocol inclusion criteria were enrolled in the study.  

In cases of site-selected inclusion, site pathologists review patients’ history, test results and histology primarily for patient care.  Study inclusion criteria often serve as the site’s secondary goal.  The pathologists’ expanded patient knowledge, which is important for patient care, can potentially hinder the clinical trial process by injecting unintended bias into the patient inclusion evaluation.  For example, during the slide interpretation process, pathologists could be influenced by their knowledge of the patient’s history and his or her symptoms that seemingly match the inclusion criteria.   

The Phase 2 study readers, required to follow Biomedical Systems’ centralized digital pathology process, were only focused on the morphology represented within the digital slide images.  For the study, Biomedical Systems blinded its readers to all extraneous patient information.  Readers’ evaluations and diagnoses were made solely on the pathology employing a standardized process, void of patient information or variability often inherent at a study site.


The impact of an underpowered study can cost drug manufacturers millions of dollars and their reputations.  Centralized digital pathology helps reduce bias and variability while increasing the probability of accurately assessing efficacy and statistically relevant data in clinical trials.

[1] Food and Drug Administration. (2011). Guidance for Industry Standards for Clinical Trial Imaging Endpoints (DRAFT). Washington, DC:  Government Printing Office  

Author | Timothy Callahan, PhD