Adaptive experimental designs have become ubiquitous throughout the pharmaceutical industry, especially in therapeutic areas such as oncologyi,ii. Adaptions enable a sponsor to seamlessly move between development stages and/or tailor the study to those treatments that have positive patient clinical response. FDA’s Critical Path Initiative (FDA, 2004iii, 2006iv) is a bold supporter of these designs. This article will explore if our current data management systems support trial adaptions.
First, we should examine what data management systems do well. All, including Excel spreadsheets, allow data to be entered, held until needed, and then enable extractions. Advanced systems track who entered each data value and when it was entered, which are requirements of 21CRF11v. Evaluating the data against pre-defined ranges or a set of predefined responses, and providing queries when the ranges are exceeded, provides timely data cleaning opportunities.
Modern electronic data capture (EDC) systems contain web-based remote access which enables sites to enter their own data. As compared to classic paper-based studies with subsequent double-keyed data entry, web portals enable on-going data review, which timely meets the ICH E8 provisionvi that "emerging animal toxicological and clinical data should be reviewed and evaluated by qualified experts to assess their implications for the safety of the trial subjects."
Why allow a trial to adapt?
ICH E8 asks for data review and "in response to such findings, future studies and, when necessary, those in progress should be appropriately modified in a timely fashion to maintain the safety of trial participants." ICH focused on safety, but the FDA Critical Path Initiative and the EMEA Reflection Papervii moved trial adaption for efficacy into a mainstream regulatory accepted practice. In return, regulatory agencies expect controlled adaptions. For static protocols, this requires protocol amendments as opposed to protocol deviationsviii after observing something outside the trialix, such as conclusions from other studies with the same compound or public disclosure of competitive products.
Adaptive trials (FDA, 2010; EMEA, 2007) are designed to pre-plan alternative paths for the trial from the design stage. These may include gate-keeping steps for early stopping rules, sample size re-estimation, adding new geographic areas, or dropping one or more experimental arms. The protocol defines the requirements space such as the number of interim data looks, maintenance of blinding by using Independent Data Monitoring Committees, and the criteria to continue or adapt. These preplanned adaptions eliminate guess work and uses controls that are statistically prudent like group sequential or Bayesian methods.
Data management in adaptive trials
When incoming data from a trial is reviewed in aggregate up and beyond monitoring of Case Report Forms [CRF], it is considered an interim review by the guidelines. Clinical and Statistics are very dependent on Data Management to act in a manner that supports or enhances data collection and availability to enable a timely aggregate review as the first step in implementation of an adaption.
Key success criteria for a data management system to support adaptive trials include:
- Tools to rapidly collect clinical data
- Tools to display clinical data
- Standards to enable quick modification to protocol and database
- Implementation efficiency to enact preplanned adaptions
Consider a car’s GPS unit as an example of an adaptive trial. The starting point and ending point are defined by the user, and the best set of roads is pre-selected by the tool. During the drive, new data on the car’s position is constantly being collected, processed, and plotted. If the driver deviates from the GPS plan, the unit recalculates directions based on both the initial parameters and the new data, and adapts to a new route. Better units can gather external data like traffic, and propose adaptions before hitting the traffic jam. But, being warned just a mile early doesn’t help if the next exit is 4 miles ahead. Likewise, "temporary" speed zones are not considered (even if the road repair is years long) and the GPS may misleadingly provide the default speed limit, resulting in a higher likelihood of a traffic violation.
While a GPS can adapt on-the-fly, clinical trials are more restrictive and need layers of approval from the sponsor, Institutional Review Board (IRB), and possibly regulatory agencies. Gathering those approvals at trial start will allow the adaptions to occur when the criteria are met, without sequentially moving though the various gatekeepers.
Rapid data collection has not always been a forte of clinical trials, but EDC, with its remote access, greatly improves the timeliness of data entry by allowing an administrator at the clinical site to directly transcribe into the database. Conceptually, site administrators could enter data each day, but realistically, time lags of days or weeks occur. Consistency checks can now be directly sent to the site for resolution including "pop-up" queries issued to the site keyboarder immediately upon typing data outside of the predefined ranges.
While the current state of the art is quite an advance over paper based trials, it can use a few boosts to support adaptions.
Improvements to data collection and examples of field-level adaptions:
Most EDC are web-based systems, and theoretically could be used bed-side. This would completely erase the entry lag, and allow the pop-up queries to be addressed by the Investigator in real time. Data quality would improve since an unusual value for a vital sign could be flagged and corrected while the patient is still present.
What if the system also had an adaptive field-level operation? For example, if the systolic blood pressure of a controlled hypertensive patient was >160mmHg, instead of a simple query to confirm the high BP result, the system could instead adaptively ask the Investigator to collect multiple blood pressures matching the American Heart Association guidelinesx. This provides clinical information that was preplanned in the protocol, but only enacted when the relevant criteria was met. The database would have already been designed to handle the adaption, and the statistician and clinician already defined how to analyze this data.
Data management systems could be used to monitor a patient’s weight, and when it drops by 10% from baseline, enact the protocol adaption of a re-calculation of chemotherapy quantity. Time-to-event efficacy criteria, such as complex RECISTxi progression criteria, can be automatically implemented without the need for the physician to manually review the entire patient casebook.
Data managers with today’s systems often design 2 or more copies of the same CRF to handle one different data collection field. For example, at 0, 6, and 12 hours, vital signs plus temperature are collected, but at 3 and 24 hours only vital signs. To reduce complexity and costs, a single CRF could be designed with a field-level adaptive operation that displays the temperature field only on the correct visits, and either leaves the space blank or provides a grayed-out field [Fig. 1] so the Clinician explicitly knows at which time the temperature is to be collected.
Figure 1; Novel design of a CRF to detail requirements of certain visits
Laptops are portable but investigators rarely take them into the patient examination room, possibly due to time-related issues, patient objections and ease of use reasons.
One solution is to use mobile tools like tablets or smartphones. These tools are less intrusive than laptops and enable a clinician to enter data while maintaining their focus on the patient. However, small screen size and resolution means the sponsor and regulatory agencies need to downsize the number of variables.
Existing mobile tools like eDiaries and ePRO allow patients to collect their own data between visits. Upon entering the waiting room, this data can be loaded into the web system and accessed by the investigator during the visit. Even better would be a routine upload to the web database from the patient’s home before the visit, so the investigator can review the patient’s progress in real time, jump on adverse events, and have adapted treatment plans waiting for the patient on examination day. This changes the clinical care from reactive to proactive, with the hospital staff prepared to implement the adaption.
Even newer mobility tools are reaching the consumer market. Home blood pressure, fitness bands, portable pulse oximeters, EKG patches and electronic scales are all now available with Bluetooth connection to smartphones. Uploading to the EDC system would be a logical next step; however, these Apps and devices are not necessarily validated in a manner required by FDAxii or EMAxiii,xiv. While fine for home monitoring, they may not be sufficient for regulated clinical trials.
Patient-level adaption in data management systems
Data Management systems also can accommodate patient-level adaptions.
Patients are encouraged to complete a trial, but occasionally they withdraw. For paper studies, the unused CRF pages are either marked as not applicable or ignored after some type of early withdrawal form is completed. EDC systems often are a bit inflexible with edit checks expecting entry into fields after the withdrawal has occurred. Data managers manually remove expected eCRF pages or mark as not applicable within the EDC. Alternatively, building a single switch into the database to turn off all visits after the current day would be ideal, possibly auto linked to the disposition page.
Patient-level adaption could be triggered at preplanned phases during the study. For example, when a patient completes a specified number of treatment cycles or has a pre-planned clinical event, the EDC system could seamlessly move that patient into the next study epoch, such as invoking a new randomization (for a randomized withdrawal trial) or placing the patient into a treatment maintenance phase.
Adaption at the study arm level
Adaptive trials focus on moving seamlessly through development phases, dropping study arms in pick-the-winner designs, etc. These are triggered when a sufficient number of patients reach the criteria for the protocol adaption to occur. All previous discussion of the need for rapid data collection is amplified since these changes have a much higher magnitude of risk. If data collection is biased towards one region, for example, the action to close a treatment arm may be premature if there is an underlying regional effect.
From the EDC operational perspective, opening the next study phase or closing a treatment arm is minimally different than an early patient withdrawal. Gatekeepers, such as Independent Data Monitoring Committees (DMC)(FDA, 2006xv; EMEA, 2005xvi) and Sponsor Steering Committees, depend on delivery of sufficient efficacy and safety data from all treatment arms, and all regions, to render a well based decision. Rapid collection of high quality data is a must, which is aided by the field and patient-level adaptions discussed earlier.
One additional challenge for adaption at the study level is communication with the IWRS, drug inventory, and other systems. EDC adaption is a major step, but interoperable or integrated peripheral systems improve the likelihood that all systems can be designed to adapt simultaneously.
This article examined some key success elements for adaptive clinical trials by pre-planning EDC adaptions . Examples of field-level, patient-level, and study arm-level adaptions were shown, with increased risk associated with higher order adaptions. Early data retrieval at bedside and by mobile tools greatly enhances the data collection and quality. Using common database standards and "single-switches" improves data management’s implementation of changes. Ensuring uniform rapid data collection from all sites mitigates premature study-level decisions which could be inadvertently biased based on underlying regional differences.
The ability to design an EDC with all planned adaptions was shown to be implementable. A greater implementation challenge is how to link the EDC adaptions to the IWRS and other systems since study-level adaptions depend on the interoperability of multiple complex systems.
i) Berry, DA, Adaptive clinical trials in oncology, Nat Rev Clin Oncol. 2011 Nov 8;9(4):199-207. doi: 10.1038/nrclinonc.2011.165.
ii) Berry, Donald, Adaptive Clinical Trials: The Promise and the Caution, 2010, Journal of Clinical Oncology, http://jco.ascopubs.org/content/29/6/606
iii) FDA Critical Path Initiative (2004), http://www.fda.gov/ScienceResearch/SpecialTopics/CriticalPathInitiative/ucm076689.htm
iv) FDA Critical Path Opportunities List (2006): http://www.fda.gov/downloads/ScienceResearch/SpecialTopics/CriticalPathInitiative/CriticalPathOpportunitiesReports/UCM077258.pdf
v) FDA (2014), Part 11, Electronic Records; Electronic Signatures — Scope and Application, http://www.fda.gov/regulatoryinformation/guidances/ucm125067.htm
vi) International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceutical for Human Use, ICH Harmonised Tripartite Guideline, General Considerations for Clinical Trials E8, http://www.ich.org/fileadmin/Public_Web_Site/ICH_Products/Guidelines/Efficacy/E8/Step4/E8_Guideline.pdf
vii) EMEA, Committee for Medicinal Products for Human Use (CHMP), Reflection Paper on Methodological Issues in Confirmatory Clinical Trials with Flexible Design and Analysis Plan, 2007, CHMP/EWP/2459/02, http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2009/09/WC500003616.pdf
viii) Mehta M, Kurpanek, K, et al, The Life Cycle and Management of Protocol Deviations, Therapeutic Innovation & Regulatory Science, (2014), http://dij.sagepub.com/content/early/2014/04/17/2168479014530119.full.pdf+html
ix) FDA Guidance for Industry, 2010, Adaptive Design Clinical Trials for Drugs and Biologics, http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM201790.pdf
x) Pickering, T., Hall, JE, et al, Recommendations for Blood Pressure Measurement in Humans and Experimental Animals, Part 1: Blood Pressure Measurement in Humans, A Statement for Professionals From the Subcommittee of Professional and Public Education of the American Heart Association Council on High Blood Pressure Research, Circulation. 2005;111:697-716, http://circ.ahajournals.org/content/111/5/697
xi) Eisenhauer, E.A., Therasse, P., et al, New response evaluation criteria in solid tumours: Revised RECIST guideline (version 1.1), European Journal of Cancer 4 5 ( 2 0 0 9 ) 2 2 8 -2 4 7, https://www.eortc.be/Recist/documents/RECISTGuidelines.pdf
xii) Mobile Medical Applications, Guidance for Industry and Food and Drug Administration Staff, September 2013, http://www.fda.gov/downloads/MedicalDevices/…/UCM263366.pdf
xiii) EUROPEAN COMMISSION, DG HEALTH AND CONSUMER, Directorate B, Unit B2 "Health Technology and Cosmetics", Guidelines on the Qualification and Classification of Stand Alone Software Used in Healthcare with the Regulatory Framework of Medical Devicies, MEDDEV 2.1/6, January 2012
xiv) MHRA, Guidance on medical device stand-alone software (including apps), http://www.mhra.gov.uk/Howweregulate/Devices/Software/index.htm
xv) FDA, Guidance for Clinical Trial Sponsors, Establishment and Operation of Clinical Trial Data Monitoring Committees (March, 2006), http://www.fda.gov/downloads/regulatoryinformation/guidances/ucm127073.pdf
xvi) EMEA, COMMITTEE FOR MEDICINAL PRODUCTS FOR HUMAN USE
(CHMP), GUIDELINE ON DATA MONITORING COMMITTEES, EMEA/CHMP/EWP/5872/03 Corr, 27 July 2005