Randomization and Trial Supply Management (RTSM) systems use telephone and Interactive Voice Response (IVR) technology to manage randomization and study medication dispensing and inventories in many clinical studies. They originated in the 1990s and according to one regulatory source are used in the majority of multicentre trials conducted by the pharmaceutical industry [1]. Complimentary systems that allow dual telephone and web access are now commonplace. In addition to managing randomization and trial supplies, these systems are useful to monitor real time recruitment, manage emergency treatment code-breaking and perform calculations to ensure accuracy of dosing [2]. Other complex applications have been described [3]. But to date nothing has been written about their use in the management of early phase cohort studies and this is the subject of the paper.

Phase I and early Phase II studies generally have the aim of establishing the safety, tolerability and pharmacokinetics of one or more doses and formulations. More specifically the goals of a Phase I study are to determine the maximum tolerated dose (MTD) of a drug for a specified mode of administration and to characterise the most frequent and dose limiting toxicities. Designs used include the 3+3 design, up and down designs and the continual reassessment method [4]. A characteristic of many such studies is that patients are entered in a staggered fashion to different cohorts where progression to subsequent cohorts is determined by the performance of the previous cohort. In some studies the dose progression is fixed in advance and the dose planned for each cohort is known. In other studies the dose step used for subsequent cohorts may be determined by the number of toxicities observed in the previous cohort and future doses are not known in advance. Often there is a pause between enrolments into successive cohorts as the performance of the dose used in the previous cohort is evaluated by a Data Monitoring Committee.

Traditionally many of these studies have tended to be conducted at a single site so that the randomization, dosing, dispensing, blood sampling procedures and progression between successive cohorts could be tightly managed. But given the need to accelerate drug development, there is a trend to sponsors using RTSM technology to enable these studies to be opened up to more centers; often these centers are located in different countries. Allowing competitive recruitment may be particularly important in patient studies in populations where it is difficult to recruit. RTSM technology allows management of the numbers recruited to each cohort which would be difficult, if not impossible, to coordinate with many centers. Similarly RTSM technology allows the management of progression between cohorts which reduces the administration and communication effort involved if sites were being managed manually via traditional methods such as fax and email.

This paper is split into three sections. We will detail some typical RTSM functionality that may be built as standard modules/functionality to aid the management of early phase cohort studies. We will then describe some advanced features that are used less regularly. Finally, a sample survey of the incidence and types of early phase cohort studies performed by Perceptive Informatics was undertaken. The results of this survey are described.

When considering how to manage an early phase cohort study, it might be initially thought that the flexibility required is too complex to allow production of a generic pre-validated RTSM system. However, it is possible to produce validated configurable modules which allow an RTSM system to handle the most commonly requested cohort functionality. At Perceptive Informatics there are standard modules available for: opening and closing of cohorts, recruitment caps and replacement randomizations. Hence management of an early phase cohort study through RTSM can be set up in a timely manner by parameterizing a pre-validated system. During the recruiting phase of a clinical trial, movement between cohort randomization schedules and their associated medication schemes can be controlled through simple IVR/IWR transactions made inside the RTSM system.

Early phase studies may typically require recruitment of a small number of subjects across many sites. An RTSM system can behave as the 'gate keeper' in studies with ‘stop – start’ recruitment. Recruitment into a fixed cohort of subjects can therefore be controlled centrally across multiple sites. Administrator IVR/Web transactions inside the RTSM system can be designed to open/close cohort recruitment, as well as set recruitment limits and define dose levels.

In a randomized cohort study, each cohort can be associated with a separate randomization schedule inside the RTSM system. Alternatively a single schedule stratified by cohort can be used with the treatments denoted by generic identifiers with no need to identify the dose e.g. “TEST” and “PLACEBO”. IVR/Web transactions can be tailored to a particular protocol. For example, if a study has 3 potential cohorts, then a way of activating or prematurely de-activating a cohort may be required.

In Fig. (1) an RTSM call flow (transaction map) could allow the study wide opening or closing of a cohort, as well as setting the dose level for a cohort. This is an example of how a simple management call could be placed to manage study wide recruitment into a cohort study. The subsequent randomization transaction made inside the active cohort will use the appropriate dose level and recruitment limit information. Additional functionality may be required by a particular sponsor and transactions can include extra functionality. For example, a pre-specified recruitment target may not be possible during the design phase, and a recruitment target may need to be specified during the activation of a cohort.

Fig. (1) Example of an IVR cohort activation call flow.

Each cohort may have a fixed recruitment target and an RTSM system can enforce that limit. But if a hard cohort recruitment limit is employed, then withdrawals can affect any subsequent data analysis. If a patient withdraws before completing all treatment visits (or before completing a desired follow-up period), then replacement functionality may be required to achieve treatment assignment balance within a cohort. The most common requirement is that patients are replaced on a “like for like” basis i.e. the replacement patient receives the same treatment as the patient who has withdrawn.

In randomized double-blind studies managed without automated methods, medication supplies are provided in kits which contain the patient number on the kit label. Replacement can thus be effected by having a set of corresponding parallel reserve kits with the same numbering system e.g. same numbering system suffixed by R. The investigator will have been instructed that if a patient is to be replaced, then he/she should allocate the patient the corresponding R-numbered pack. Notice that this can only be done once, it is wasteful of supplies and may have unbinding implications.

Alternatively, the benefit of RTSM is that automated replacement can be achieved in a blinded manner.

To allow for replacement randomizations we need to consider the randomization list design. Inside the randomization schedule, it is possible to have the spare randomization code below the main cohort randomization records that are reserved for replacement subject randomizations only. These records could be unavailable in the RTSM system at the start of the study and a replacement randomization entry in the spare records could be freed up each time a withdrawal occurs (Fig. 2). If a matching treatment group entry is freed up, then we can guarantee a like for like treatment group replacement will take place when the freed up entry is used. The option here could be to close cohort recruitment only when we have used all available randomization records. This will include the replacement records freed up during the study which will be used last.

Fig. (2) Example of a randomization list which can allow replacement.

Within the RTSM databases there would be a full audit trail and traceability so that it is clear which patient being replaced is responsible for the 'Unavailable' entry being freed-up. Stratification can also be accounted for in the replacement randomization list. However, if a stratum specific replacement is required, then this may delay completion of a cohort as we wait for an appropriate subject to enter the study.

Additionally, if it is expected that subjects may discontinue many times, a facility for multiple replacements can be considered. For example, two replacement lists could be employed to allow a randomization number to be replaced twice as a maximum (Table 1). In this case Record 1206 would be used to directly replace record 1106 if the subject receiving that entry withdraws. Likewise, Record 1106 would have been used as a Replacement for Record 1006.

Table 1

Example Randomization Number Format to Allow Replacement

The choice of whether to allow replacement can be left up to the study manager who simply uses the cohort management call to increase the recruitment target or it can be automated based on progress within the study e.g. an extra subject is needed if the investigator makes a withdrawal call.

In addition to the standard functionality previously discussed, it is always possible to incorporate other functionality on a bespoke programming basis. Advanced or less common functionality that Perceptive Informatics has experience with includes the following aspects of early phase studies:

RTSM functionality is increasingly used to manage studies which require control of entry into sequential cohorts. The electronic systems allow superior functionality to traditional non-automated methods which are restrictive and in particular limit the number of participating sites. Our survey of our database has shown that many cohort studies are conducted in multiple sites and countries with fairly low numbers of patients per site; this reflects the need for speed given that these studies are often on the critical path for drug development.

The authors are all employed by Perceptive Informatics, a provider of RTSM and clinical systems.