Abstract
Objective:
The aim of this study was to assess the effects of (a) auto-injector form factor on maximum applied force capability and (b) auto-injector design and instructions on force production and orientation.
Background:
Effective delivery of epinephrine through an auto-injector is the result of a multitude of design factors. At minimum, the design needs to allow the user to apply sufficient force for the needle to penetrate clothing and tissue.
Method:
Trainer devices for three commercially available epinephrine auto-injectors with different form factors (cylindrical, elliptical, prismatic) were tested in a laboratory-based repeated-measures experiment with 20 adults. Participants applied their maximum force onto a force plate positioned over their thigh and practiced an injection using the trainer device after viewing training videos. Participants also rated force confidence and preference.
Results:
The maximum force varied significantly across devices. The greatest force observed was 64 newtons with the elliptical device, and the lowest force was 61 newtons with the cylindrical device. Participants reported the highest force confidence when using the elliptical and cylindrical devices, ranking the elliptical as their preferred device.
Conclusion:
Force capability results for the elliptical device suggest that it may be more successful in achieving the necessary force for drug delivery in a larger set of adult users.
Application:
Results suggest that the auto-injector with the elliptical form may enable more successful drug delivery among a larger set of users.
Keywords
Introduction
Auto-injectors are devices designed for self-administration of a number of different drugs. Because they allow for quick access and delivery, auto-injectors are particularly important for the delivery of epinephrine (Simons, 2006) in cases of a life-threatening allergic reaction, or anaphylaxis (Sampson et al., 2006). Effective drug delivery of epinephrine through an auto-injector is an essential step in these cases (Joint Task Force on Practice Parameters for Allergy and Immunology, 2005) and is controlled by two parameters: the design of the mechanism that delivers the drug and the human factors associated with activating, positioning, and holding the auto-injector during drug administration (Dennerlein, 2014). For example, the depth to which the drug penetrates is dependent on the exposed needle length (Song, Nelson, Chang, Engler, & Chowdhury, 2005), the pressure at which the drug is injected (Lieberman et al., 2010), and the external force applied to the needle to overcome the resistance of the patient’s clothing and tissue (Jacobsen, Guess, & Burks, 2012; Schwirtz & Seeger, 2010, 2012).
Previously, various human factor metrics for evaluating auto-injectors have been used to investigate aspects such as preference (Camargo, Guana, Wang, & Simons, 2013), perceived required force (Schwarzenbach et al., 2014), and injection fluid volume and location (Berteau et al., 2010). These are important factors for effective drug delivery; however, there are limited data on interaction of auto-injector design and human capability. Design factors that affect human factor metrics include the device’s form factor (shape and dimensions), function (Cutkosky, 1989), and instructions given to the user. For example, previous studies of tool handles have shown a relationship between form factor and maximum force capability (Wells & Greig, 2001). Additionally, optimizing performance of any human–machine interface requires effective instructions (Amick et al., 2003), which are important to consider in device design.
The key element of successful drug delivery through an auto-injector is axial force (including direction, magnitude, and duration). The user needs to position the device approximately perpendicular to the leg to ensure the greatest penetration of the needle (direction) and apply enough force to overcome both the recoil and the mechanical activation trigger (magnitude) that releases the needle and administers the drug (Jacobsen et al., 2012). Next, the user needs to hold and maintain sufficient force as the drug is delivered (duration). The user applies this force by gripping the auto-injector and pressing or pushing its needle through clothing into the skin. Without sufficient force, drug delivery would not be successful.
Although there are a multitude of factors that affect the amount of force that can be applied, the three epinephrine auto-injectors (EAIs) commercially available in the United States differ mainly in size, shape, and perhaps to a lesser degree, friction. These factors are important in determining power grip performance as measured by maximum force production and transmission (Seo & Armstrong, 2011). Therefore, our objective was to examine the effect of shape on the maximum applied force capability of adults and the effects of device design and instructions on force production and orientation of an EAI. We hypothesized that in a given postural configuration, the physical form factor of EAIs and the device-specific delivery methods would affect performance during simulated drug delivery. To test this hypothesis, we completed a repeated-measures laboratory experiment in which participants performed two tasks with the three devices. The tasks were (a) a maximum force capability task in the required posture and (b) an application task, in which we examined the interaction between device and the force production and orientation of the device. For the first task, we hypothesized that the maximum axial force would vary across device form factors and that the grip effort would not differ between form factors. With respect to the application task, we hypothesized that performance metrics, such as time for applied force to reach device activation force, efficiency of device application, and variability in device positioning and orientation, would also vary across devices.
Method
Twenty adult participants (10 female, 10 male; mean age = 22.1 ± 2.3) volunteered for a repeated-measures (crossover) laboratory study in which they performed tasks using EAI trainer devices. All but one of the participants had no prior experience with EAIs. The one participant with experience had been prescribed an EpiPen Auto-Injector (Mylan Specialty L.P., Canonsburg, PA) and had been previously trained to use it. Four of the participants were left-handed; however, all protocols were completed with the right hand. The mean anthropometric measurements for the participants were typical of the average United States population (Table 1).
Mean Anthropometric Measures
Note. Standard deviations shown in parentheses.
The Northeastern University Institutional Review Board approved all protocols and informed consent forms.
Experimental Protocol and Tasks
Participants stood on a specially designed platform (Figure 1A) with their right leg against the middle rail, their feet parallel, and their right arm aligned with the force plate on the other side of the platform fence, thereby mimicking the potential standing posture adopted during EAI use. The force plate height was adjusted for each participant so that their fists aligned with the middle of the load cell while their arms were in the anatomical position (Figure 1B). This height was chosen to mimic the required injection site in the outer thigh, consistent with the instructed use of the devices.
The experimental setup mimicked injection to the outer thigh while standing. For the maximum force capability task, (A) participants stood upright on a platform and applied their maximum possible force for ~3 s onto a force torque load cell mounted on a fence positioned over their thigh. The force sensor was covered with a stiff foam rubber that roughly emulated the viscoelastic behavior of the thigh. For the application task, (B) participants used the specific injection method for each device after watching instructions. Device C, cylindrical; Device E, elliptical; Device P, prismatic.
Participants completed all tasks using three EAI trainer devices (Figure 2) that had different form factors: cylindrical (Device C), elliptical (Device E), and prismatic (Device P).
Epinephrine auto-injector classification. Epinephrine auto-injector devices were categorized by form factor as cylindrical for Device C (left), elliptical for Device E (middle), or prismatic for Device P (right). Views depicted include (A) top-view perspective for each device and (B) side view when in use.
The EAI trainer devices were reused among the participants. The EAI trainer devices differed from the active EAIs in that they did not contain needles or epinephrine but retained the same trigger mechanism. Device C was the Adrenaclick Auto-Injector (Amedra Pharmaceuticals LLC, Horsham PA), Device E was the EpiPen Auto-Injector, and Device P was the Auvi-Q Auto-Injector (Sanofi US LLC, Bridgewater, NJ).
Each participant performed two tasks: a maximum force capability task and an application task. For the maximum force capability task, participants were instructed to push into a force plate, ramp up to their maximum force, hold for approximately 3 s, and then release their force. This task was performed with safety caps on the devices so as to avoid the effects of the trigger action. For the application task, participants watched a training video for each device while holding the EAI trainer in their hand and then practiced the motion (training videos are available online; Mylan Specialty L.P., 2015; Amedra Pharmaceuticals LLC, 2015; AllergyGoAway.com, 2015). They were instructed to use the specific injection method for each device as explained in the respective video (Table 2), holding for 10 s after overcoming the initial trigger action. Participants were instructed to stand upright for all tasks.
Trainer Device Dimensions and Properties
For each task, a randomization procedure was used to minimize the effect of order and to vary the order in which devices were tested across participants. After watching each video, the participants performed the task twice, producing two trials. Participants were given 2-min breaks between devices to prevent fatigue.
Measurements and Signal Processing
A six-axis load cell (Delta, ATI, Apex, NC) attached rigidly to the middle rail of the testing platform measured the applied forces for both tasks. Mounted on the load cell was a metal plate with a foam and cloth layer that simulated the general properties of a clothed upper thigh. Load cell force and torque data were recorded at 100 Hz using LabVIEW (National Instruments, Austin, TX). The data were then digitally filtered with a 3 Hz Butterworth filter (filtfilt, MATLAB, MathWorks, Natick, MA). From these data, the magnitude of the resultant applied force, the magnitude of the force normal to the surface of the force sensor (the thigh), and the angle of the resultant force relative to the normal of the force sensor were calculated.
Surface electromyography (EMG) electrodes (DE-2.1 single differential electrode, Delsys, Natick, MA) recorded the activity of six forearm muscles (extensor digitorum, extensor carpi radialis, extensor carpi ulnaris, flexor digitorum superficialis, flexor carpi radialis, and flexor carpi ulnaris). Electrodes were placed in standard locations (Perotto, 1994). The amplified EMG signals were recorded at 1000 Hz then rectified and smoothed using a simple single-pole digital filter (filtfilt, MATLAB). To normalize results across participants, EMG data were collected during three 3-s maximum voluntary contractions (MVCs) for each muscle. For these MVCs, participants exerted their maximum effort using standard manual muscle testing procedures. The experimenters (AB and MJ) manually provided resistance for each forearm muscle in the direction previously defined (Buchanan, Lloyd, Manal, & Besier, 2005). Participants rested for 2 min between the same muscle contractions. The maximum value obtained during any of the three contractions was used as the MVC reference. The rectified and smoothed EMG signals were divided by this reference and multiplied by 100 to provide a percentage value. The signal levels for inactivity were negligible. The normalized EMG values averaged across the six muscles provided a proxy for grip effort (Duque, Masset, & Malchaire, 1995).
Two clusters of three infrared light-emitting diodes (IREDs), one placed on the platform and one on each EAI trainer device, along with a three-camera motion analysis system (Optotrak Certus, Northern Digital Inc., Waterloo, ON, Canada), measured the orientation and position of the device relative to the force plate. The system recorded the IRED x, y, and z trajectories at 100 Hz. The trajectories were digitally filtered using a fourth-order Butterworth filter with a 10 Hz cutoff frequency (filtfilt, MATLAB). Using the system’s digitizing probe, virtual markers were located on each of the four sides of the force plate (top, bottom, right, and left). Additionally, virtual markers were placed on the four corners of each EAI trainer device. The 3-D position and orientation of these landmarks were calculated on the basis of the position and orientation of their associated IRED cluster (Winter, 2005).
In addition to direct measurements, participants completed a brief survey in which they rated their force confidence and preference for each device and grip. Force confidence was assessed by asking the participants to “rate from 0 to 10 how confident you are in applying force with the different grips” (0 = least confident, 10 = most confident), with the scale presented on a modified 10-cm visual analog scale with marks at each centimeter. The participants were also asked to rank their preferences for each device, with 1 being the most preferred device and 3 the least preferred.
Analysis
The main dependent variables for the maximum force capability task were the following: applied resultant force (vector sum of three-dimensional components), applied normal force (the one-dimensional force component perpendicular to the surface of the force transducer), grip effort (average of the EMG amplitude from the six forearm muscles; Duque et al., 1995), applied force angle, and device angle. Additionally, since there is no established metric for evaluating efficiency of force transmission, two parameters were calculated: the ratio of applied force to grip effort and the difference between the angle of the device and the angle of the force vector. The more the radial direction effort is transferred axially, the more efficient the device is in transmitting force. A previous study on grip force coordination in hand tools (Lowe & Freivalds, 1999) used a similarly calculated force ratio to quantify efficiency. In addition, the alignment of the applied force vector and device is a metric of efficiency: The smaller the difference between the angles of the device and force vectors, the more efficiently the force is being transmitted. All of these values were calculated as the average value of the data collected across the 0.5 s before and after the time of peak resultant force detected during the task. The main independent variable for the maximum force capability task was the device.
The main dependent variables for the application task were the time to peak force and the mean and standard deviation of the following: applied resultant force, applied normal force, grip EMG effort, grip efficiency, applied force angle, and device angle. These values were calculated as the mean value of data across a 5-s window starting 1 s after the initial peak force. The instructions for the devices are to hold the force and position for a minimum of 5 s for Device P and 10 s for Devices E and C. The initial peak force was temporally associated with the activation of the devices’ trigger mechanism. Identification of the timing for the peak force used raw unfiltered data. The time to this initial peak force was the duration from the initial force application to this peak force. The main independent variable for the application task was the EAI.
The main dependent self-reported variables included force confidence and device preference, and the main independent variable was the device. Both force confidence and maximum force capability were assessed after the completion of all of the protocols.
Statistical analysis was performed using the JMP Pro 11 (SAS, Cary, NC) linear mixed model module, with the participant as the random effect and the independent variable as the fixed effect. Variation for each outcome measure across EAI trainer devices was tested using a one-way repeated-measures analysis of variance, with an alpha value of .05 as the level of significance. When a significant effect was found, a post hoc analysis with Tukey’s honest significance test was conducted across the three devices. A chi-square test was applied to compare preference by device. Individual analyses were run for each dependent variable.
Results
During the maximum force capability task, the resultant force differed significantly across the three devices (Table 3).
Across-Participant Averaged Values for the Dependent Variables for the 1-s Data Window During the Maximum Force Capability Task
Note. Standard errors shown in parentheses. MVC = maximum voluntary contraction.
For significant main effects, Tukey’s post hoc groupings are ranked such that A > B. Values with the same superscript letters indicate no significant difference. Significance defined as p < .05.
Maximum resultant force varied significantly among the three devices (p = .0387). Device E (elliptical form factor) elicited the highest resultant force. The difference in the resultant force between the prismatic and elliptical form factors was not statistically significant. There were significant differences in force angles (p < .0001) and normal force (p = .0083), with Device P (prismatic form factor) having the lowest maximum force of the devices tested. The highest grip EMG effort measured occurred with Device P. The ratio between grip effort and applied axial force (N/%MVC), a measure of efficiency of transmission of force, was the lowest on average (2.5 N/%MVC) for Device P, followed by Device C (cylindrical form factor) at 3.1 N/%MVC and Device E at 3.3 N/%MVC. The lowest device angle occurred with Device C, and both Devices C and E had the lowest force application angles. The difference between the device and force angles was lowest for Device E (<0). The difference for Devices C and P was 2°.
During the application task, all measurements varied significantly among the three devices for all of the dependent variables tested (Table 4).
Across-Participant Average Values for the Mean Dependent Variables for the 5-s Hold Window of the Application Task
Note. Standard errors shown in parentheses. MVC = maximum voluntary contraction.
For significant main effects, Tukey’s post hoc groupings are ranked such that A > B. Values with the same superscript letters indicate no significant difference. Significance defined as p < .05.
Similar to the differences in operating instructions (Table 2), Device E had the smallest (fastest) time to peak force. After reaching this force and during the 5-s hold period, there were small, but significant, differences between devices in average force magnitude and average grip EMG effort, with Device E having the highest applied force and Device P the highest grip effort. The ratios between these measures showed the same pattern as in the previous task, with Device E being the most efficient at 3.4 N/%MVC, followed by Device C at 3.3 N/%MVC and Device P, which was the lowest, at 2.7 N/%MVC. The angle of force application was the highest for Device P, and the device angle was the lowest for Device C. The difference between the device and force angles was the lowest for Device E at 1°; for Device C, it was 3°, and for Device P, it was 4°. While using Device P, participants exhibited a significantly higher hold-force standard deviation compared with the other two devices (Table 5).
Across-Participant Average Values for the Standard Deviation Dependent Variables for the 5-s Hold Window of the Application Task
Note. Standard errors shown in parentheses. MVC = maximum voluntary contraction.
For significant main effects, Tukey’s post hoc groupings are ranked such that A > B. Values with the same superscript letters indicate no significant difference. Significance defined as p < .05.
No significant differences were measured in the standard deviation during the 5-s hold window for force magnitude, grip effort, or device angle.
Participants exhibited the highest force confidence when using Devices C and E, both with a force confidence of 8.2 on a 10-point scale. Most participants (60%) ranked Device E as their preferred device, and 75% ranked Device P as their least preferred device (Table 6).
Mean Self-Reported Measures
Note. NA = not applicable.
For significant main effects, Tukey’s post hoc groupings are ranked such that A > B. Values with the same superscript letters indicate no significant difference.
Standard error was calculated for force confidence variable only and is indicated in parentheses.
1 = not confident, 10 = most confident.
1 = most preferred, 3 = least preferred.
Discussion
The purpose of this study was to determine the effect of device shape and size (form factor) on the maximum applied force capability and the effects of device design and instructions on force production and orientation in commercially available EAIs within a set of young adults. Results showed that the elliptical form factor had the highest value for maximum axial force, most normal orientation, least forearm muscle grip effort, highest grip efficiency (ratio between applied force and grip effort), and least amount of variability in its orientation. These results suggest that the design of the device affects these performance metrics and is important in effective drug delivery for most potential users.
Previous studies have also shown that the elliptical form factor is the most efficient in transmitting torque (Seo & Armstrong, 2011). Many hand tools are designed with elliptical handles to optimize this performance. In addition to shape, size may play a role in the differences found between the three commercially available EAIs evaluated. The width of the elliptical device at 28 mm is slightly below the range of diameters for maximum grip force. Maximum grip force in adult populations occurs for handle diameters between 35 and 40 mm (Edgren, Radwin, & Irwin, 2004; Kong, Freivalds, & Kim, 2004). The 15-mm diameter of the cylindrical device is well outside of this range, which would be associated with a lower grip force and hence can explain the lower maximum force and the smaller force confidence observed with this device in our study.
Because we believed that participants would maximize their effort in the maximum force task, we hypothesized that the grip EMG effort would not differ between form factors; however, it was lowest for the prismatic form factor, suggesting that it is more difficult to hold and thus requires more muscle effort to create similar amounts of applied force. This difference may be due to the form factor and the length of the device. The prismatic form factor elicits a different type of grip in that one long edge of the device abuts the palm of the hand between the first (thumb) and second (index finger) metacarpal bones and the fingers wrap around the other long side. Hence, the contact area between the fingers and hand is limited to these two regions. Additionally, the length of the gripping space for the prismatic device is 7 cm (8.5 cm in length minus 1.5 cm for the activation mechanism), whereas the average hand breadth in our test population was 8.9 mm, meaning that the contact area was further decreased in the prismatic device. In the cylindrical and particularly the elliptical form factors, a larger amount of surface area of the hand is in contact with the device.
The higher force confidence with the elliptical and cylindrical form factor devices correlated with the measured grip effort. The perception of needed force therefore was likely associated with the internal muscle efforts elicited by the participants. Furthermore, the device with the elliptical form factor was most preferred for both tasks. This preference contrasts with a previous study of device preference (Camargo et al., 2013) in which the authors “tested the preference . . . with regard to method of instruction, preference to carry, device size, and device shape” (p. 266). Thus, preference with regard to physical function was not tested in that study as was done in the experiment reported here. Additionally, the participants received no training and used each device only once, so the assessment of preference did not include functionality during drug delivery (Dennerlein, 2014).
Although this study did not measure drug delivery directly, it measured important human factors necessary for effective and easy drug delivery, namely, applied force, position, and time to achieve force. Effective drug delivery occurs when the needle reaches the muscle tissues of the thigh quickly, which requires applying enough force to reach the tissues and orienting the device most normal to the thigh surface. The fastest time to peak force for the device with the elliptical form factor was due to a difference in instructions, in which for the elliptical form factor, the participants were instructed to swing, and for the other two devices to push, into the thigh. A larger normal force (smaller angle of force application) will compress superficial tissues more to reach the muscle; the more normal the device is to the surface (smaller device angle), the deeper the needle can reach. The grip efficiency measured the ease by which a user can create the resultant applied force needed to compress the surface tissues so that the needle of the EAI can reach the muscle tissue. The maximum force capability demonstrated that similar to other studies, the elliptical form factor optimized the effort of the forearm muscles in creating the applied force to the thigh. This result was demonstrated through both the highest ratio of applied force to grip effort and the most normal orientation. With these logic and mechanics in mind, the implications and size of these differences and their clinical impact need to be further examined.
An important limitation of this study is its use of EAI trainer devices rather than active EAIs, which restricts the generalizability of the conclusions. Differences between EAI trainer devices and active EAIs include their weight and the lack of a needle and drug. In general, the EAI trainer devices weighed less than the active EAIs; however, these differences were similar for all of the EAIs tested.
There are many factors that affect the amount of force that can be applied through pushing that were not examined in this study. First, there is friction of the device. The three devices had similar surface materials, suggesting that the friction would be similar; however, we did not measure any specific friction metrics. Second, body posture factors, such as height and direction of force application and distance of force application from the body, will affect force production. The required location for drug delivery is the same across all devices: the outer thigh. Moreover, we examined participants in a standing posture to minimize the effects of varying body posture on our measurements of applied force, allowing us to focus this study on a design factor—device shape—rather than body posture. Our use of the standing posture is also consistent with training instructions, which demonstrate the injection method while standing.
Our approach is limited in that we used commercially available devices rather than exploring form factor design features through developing our own nonworking prototypes and developing a full factorial design. As a result, we cannot tease apart the effects of shape, size, and friction. Using the trainers did, however, provide a context for the task and did not require developing our own devices. This approach of using commercially available devices is similar to other studies examining the design of different tools, such as computer input devices (Rempel, Barr, Brafman, & Young, 2007; Lin, Young, & Dennerlein, 2015).
In addition, the conclusions here are limited solely to the factors associated with injection of the drug. There are, however, other factors important for ensuring that a device is available and successfully used. Examples include how size and shape affect carrying the device in pockets or handbags and clarity of instructions.
Furthermore, EAI trainer devices were used in a controlled laboratory setting, which does not perfectly simulate the emergency situation under which the device would be used. Additionally, during scenarios in which the drug needs to be administered, the stressful conditions may also add to the variability in the user’s performance.
In conclusion, this study was designed to test the effects of form factor on optimizing human capability in achieving effective drug delivery for EAIs. Standard maximum force protocols as well as simulated drug delivery training protocols were applied to examine these functional characteristics. The results suggest that the elliptical form factor may allow for the greatest force capability, and the elliptical device demonstrated the best performance as measured by time to peak force, efficiency of device application, and participant preference. In most individuals, a device with an elliptical form factor may be associated with more efficient drug delivery.
Key Points
Form factor and device-specific factors (mechanism of delivery and user instructions) each affect efficient drug delivery of an epinephrine auto-injector.
An elliptical form factor was associated with the greatest force capability and fastest time to peak force, and most participants (60%) ranked the elliptical form factor device as their preference.
In most individuals, a device with an elliptical form factor may be associated with more efficient drug delivery.
Footnotes
Acknowledgements
Funding for this study was provided by Mylan Specialty L.P. Editorial assistance was provided under the direction of the authors by MedThink SciCom with support from Mylan Specialty L.P. The authors have no other conflicts of interest.
Ana Barbir was a postdoctoral fellow at Northeastern University. She is currently a senior consultant at Rimkus Consulting Group, Inc. She received her PhD in mechanical engineering from the University of Vermont in 2010.
Mark V. Janelli is an undergraduate student in the Department of Mechanical and Industrial Engineering at Northeastern University.
Michael Y. Lin was a doctoral candidate at Harvard T. H. Chan School of Public Health. He received his ScD in ergonomics and human factors from Harvard University in 2015. He also holds a bachelor’s degree in materials engineering from the University of British Columbia. Michael is currently a user-experience researcher at Microsoft Inc. with a focus in product design.
Jack T. Dennerlein is a professor in the Bouvé College of Health Sciences at Northeastern University and an adjunct professor of ergonomics and safety at the Harvard T. H. Chan School of Public Health. He received his PhD in mechanical engineering from the University of California, Berkeley, in 1996.