018: A Novel Auto-Injector Design to Maintain Investigator Blinding in Randomised Controlled Trials

Statement of Purpose: The unintended unblinding of an investigator represents a critical risk of bias in an RCT study. The properties of injectable IMPs make them particularly high-risk for this as the differences between placebo and active injectates are readily discernible.

Current strategies to mitigate the risk of unblinding are imperfect and costly; tinted films can be applied to the syringe barrel to mask colour differences and additives may be used in placebos to match viscosity. These are only incremental measures, demonstrated by the decision of many trials to utilise an intentionally unblinded dosing investigator.

This research sought to establish an improved methodology for the blinded administration of injectables through the development of a generalisable, trial-adapted auto-injector.

Description of Methods & Materials: Testing was performed to compare manual, fixed-needle syringes (BD Micro-Fine 1ml) to identical syringes housed in an existing, commercially available auto-injector (OM AutoJect 2).

This testing was conducted by a staff of 10 Research Physicians (RPs), with the aim of identifying unblinding risks and the features required to maintain masking of two sample injectates of varying viscosity: model placebo - Normal Saline vs model active - 20% Glucose Solution. RPs performed serial injections into a human-tissue analogue and were asked to record when they felt able to identify the solution they had delivered and what had led to that conclusion.

Data & Results: When using the manual syringe, RPs were able to differentiate the model active injectate from the model placebo with a high degree of accuracy (n=95, 95.0%) - identifying the haptic feedback and visual appearance of the solutions as the primary unblinding causes.

In contrast, using the auto-injector, RPs were able to differentiate the two injectates on only two occasions (2.0%) - in these cases time to completion of injectate deliver was identified the cause of unblinding.

Interpretation, Conclusion or Significance: Initial testing demonstrated that the use of an auto-injector to conceal both haptic feedback and visual appearance during injectate delivery, resulted in a 93.0% reduction in the ability of investigators (RPs) to accurately distinguish a model active vs. placebo solution.

Where an RP was still able to identify the two solutions, this was attributed to the more viscid injectate (model active) taking an identifiably longer period to deliver.

As a consequence of this testing, a novel auto-injector has been designed with an integrated timer function which is dual activated alongside the actuation mechanism for injection delivery. By pre-setting this timer to a duration greater than the length of time to deliver the more viscid injectate, the investigator will be provided with a prompt to withdraw the device that does not unintentionally inform them of the injection duration (patent no. GB2217434.6).

The adoption of such a device will serve to fully mask the investigator from the potentially unblinding features of an injectate in a randomised controlled trial setting. This represents a significant step forward on current strategies to maintain blinding and removes the need for a costly, intentionally unblinded investigator.

Disclosures: The authors have no conflicts of interest to declare.