By Tommy Jackson on September 1, 2015
Ensuring that patients are appropriately randomized to the correct treatment group is a critical factor for all clinical studies. Additionally, it is imperative that designated study personnel are masked to the assigned treatment group unless there is a medical necessity to disclose. Ideally, at the end of the study, there is a proportional number of cases in each treatment group as planned in the study design.
However, a variety of issues can occur that compromise the desired treatment group mix. For example, a case might be accidentally given the wrong treatment or a case that was randomized might be terminated early from the study. In some cases, in order to enroll enough patients, there is a need to allow certain sites to add additional patients. When a randomization table is used, even though there might be a few additional randomization lines available, more often than not, it is necessary to generate additional randomized treatment groups to support the site case enrollment expansion.
Furthermore, when either electronic or paper-based randomization tables are used, study personnel designated to be masked may accidentally become unblended due to simple human error, such as the randomization table mistakenly being left out on a table.
Additionally, sometimes the randomization strategy needs to be applied across the whole study rather than just at the site level. A study-wide randomization makes it more difficult for site-based personnel to try to pre-empt what treatment group the case might be assigned to and to start to notice trends that might allow them to start guessing what treatment each case is receiving. This is important as it is a natural behavior to try to guess and this inadvertently introduces unnecessary researcher bias.
In some cases, it may be ideal to apply multiple stratification criteria to assign cases to particular groups (for example, gender or age). When using paper-based or spreadsheet tables, adding stratification criteria is a complex proposition.
Prelude Dynamics was challenged to develop a smart electronic interface that will more easily perform randomization processes and allow for complex stratification criteria, mitigating these challenges.
Prelude Dynamics proposed to develop an integrated dynamic randomization module to meet the challenge and to respond to given specifications for future studies. The module will allow researchers to select whether the randomization is site- or study-based, input treatment group ratios and block size, and apply multiple stratification criteria. Further, the module will reassign the treatment group for a case that does not complete the study to new enrollees, ensuring treatment group minimums and ratios are met. The module would also be automated such that randomization would only be available when inclusion/exclusion criteria have been met.
The success of a clinical trial study relies on consistent randomization of patients into treatment groups and dependable masking of certain roles in order ensure biases is not introduced into the study. Generally, this means that at least one person at the site level, the treatment dispenser or coordinator, is knowledgeable about the treatment group to which each subject is assigned. Historically, this role had maintained a written table that indicated to which group the subject is assigned dependent on the enrollment order into the study.
Written randomization tables pose several challenges for the study overall. The unmasked role must ensure the randomization table is kept confidential and that information that could indicate the assigned treatment group is not accidentally verbally disclosed. Additionally, if there is inventory that needs to be kept up with, the inventory must be kept in such a manner that it does not inevitably disclose and relationship between subjects and treatment group.
When an individual is relied on to determine randomization of treatment groups, problems can occur due to human error. A subject being inadvertently given the wrong treatment is one of the biggest such problems. This would require making a decision about how to handle the next enrollee. The decision maker would have to evaluate if the next enrollee should be provided with the treatment this subject should have received, thereby skewing the order of randomization. Further, if the mistake is not discovered immediately, other subjects may have since been enrolled and the randomization ratios become harder to control and manage.
Subjects also sometimes withdraw from studies, which can cause a difference in the number of subjects per treatment group. Historically, it has been difficult or nearly impossible to guard the integrity of the study randomization and reassign the treatment group to another study. Unfortunately, subjects dropping out of the study for various reasons is a frequent and costly event.
Another problem that can be a consequence of study dropouts or inaccurate enrollment predictions is that more subjects need to be enrolled to satisfy the study design. When this occurs, there might be sites that have had more success enrolling the desired patient demographics than others, and these sites may be asked to expand the number of subjects enrolled. This presents a problem when the randomization table is exhausted, and makes it necessary for the study biostatistician to develop an appended randomization table.
Historically, it has been nearly impossible for studies to use complex stratification criteria to enroll subjects and assign treatment groups. These studies have to choose between either developing multiple randomization tables for each set of stratification criteria or simplifying the stratification criteria altogether. Additionally, if multiple randomization tables are used, the study incurs cost to train the treatment dispenser or unmasked role to use the more complicated tables. It has also historically been problematic to randomize across the study when it is conducted in different clinics.
Prelude Dynamics proposed to train users to utilize Prelude EDC’s randomization module, which allows for dynamic randomization and was set up to handle:
- Study-wide randomization across clinics
- Reassignment of dropouts
- Reassignment of accidental treatment assignment errors
- Just-in-Time randomization
- Dropping of treatment groups if deemed ineffective due to mid-study monitoring and statistical analysis
Results & Benefits
As a result of using Prelude EDC, treatment masking and assignment is completely secure and occurs in real-time. Prelude EDC eliminated the paper tables that could be left lying around or inadvertently unmask the wrong role. Additionally, since the treatment dispenser only sees one treatment group upon randomization, the subject is more likely to receive the correct treatment. If a subject is accidentally given the incorrect treatment, Prelude EDC mitigates the error by assigning the treatment again according to the randomization set up. When dropouts occur, Prelude EDC automatically knows what treatment group has been reduced and can automatically reassign it to new enrollees to maintain treatment group ratios.
As mid-study monitoring and data analysis occurs, if a treatment is deemed to be ineffective, Prelude EDC’s randomization module settings can quickly be changed to assign future subjects into the remaining treatment groups. This can save thousands of dollars on the study.
Prelude EDC makes it possible to handle study-wide randomization across multiple sites. Whenever a new subject is enrolled, regardless of the clinic where the subject is enrolled, Prelude EDC automatically uses the same formula that would have generated the paper randomization in order to designate the appropriate treatment group. Utilizing study-wide randomization makes it more difficult for any masked personnel to be able to predict treatment assignment.
Additionally, because Prelude EDC integrates with information located in other eCRFs, its randomization module settings can be used to enter stratification criteria. For example, a study could require each treatment group to be assigned to at least 30 subjects of each gender over 50 years old with diabetes and high blood pressure, 30 subjects of each gender under 50 years old with diabetes and high blood pressure, 30 subjects of each gender and age group with only diabetes, 30 subjects of each gender with only high blood pressure, and 30 subjects of each gender and age group that do not have either condition. Prelude EDC has the ability to handle these criteria by checking all the stratification criteria and then employing the formula to generate the correct treatment group. Only the treatment group assigned is displayed and all the dosing calculations are automatically done for that treatment. This reduces dosing errors and any other errors that might occur due to treatment group assignment.
Sponsors indicate they are excited about using Prelude EDC’s randomization functionality, and appreciate the cost savings and benefits that it can provide. The principal investigators have always found it a challenge to design studies without or with minimal stratification, and recognize that Prelude EDC allows them to stratify all the desired criteria. Statisticians recognize that they now have more control over the stratification criteria, which allows for more accurate comparisons between treatment groups. Statistically speaking, when stratification criteria is not applied correctly during randomization, one treatment group could contain insufficient subjects to meet the stratification criteria at the end of the study and could interfere will proper analysis, and therefore, study integrity.
About Prelude Dynamics & Prelude EDC
Prelude Dynamics is a global provider of customized web-based software systems for data collection, analysis and management of clinical trials, studies and registries. We streamline eClinical operations through our unique and innovative Prelude EDC software system, which allows us to rapidly configure data collection solutions for pharmaceutical, CRO, medical device, animal health, and university organizations.