DIABETES TECHN OLOGY & THERAPEUTICS
`Volume 3, Number 4, 2001
`© Mary Ann Liebert, Inc.
`
`Dose Accuracy Testing of the Humalog® /
`Humulin® Insulin Pen Device
`
`MICH AEL J. RO E, P.E., M .B.A., DEBRA IGNAUT, C.D.E., RN., TOM MIYAKAWA, M.S.,
`and CH ERYL HU LTMAN, Ph .D.
`
`ABSTRACT
`
`The primary purpose of the study was to determine whether pen users would challenge the
`insufficient remaining dose (IRD) stop mechanism with sufficient force to affect the dose ac(cid:173)
`curacy of the final dose. The secondary purpose was to determine the participant's positive
`and negative impressions of the Humalog®/ Humulin® pen and the likelihood of using the
`new prefilled pen. Three different modifications to the prefilled pen's IRD stop feature were
`made. These three pen models then underwent environmental dose accuracy tes ting at vari(cid:173)
`ous temperatures and humidities, and user dose accuracy testing by 64 patients with diabetes.
`Evaluation also involved challenging the IRD stop at various dialing torques. Thirty pens
`from each model were tested to failure of the IRD stop. A model of the prefilled pen was se(cid:173)
`lected for commercialization that met the dose accuracy targets of ± 1 unit (U) for insulin
`doses less than 20 U and ± 5% of dose volume for doses equal to or over 20 U. The selected
`pen model was superior at the minimum (1 unit), median (30 unit) and maximum (60 unit)
`dose volumes. Also 92% (n = 59) of patients interviewed felt that the stop mechanism for the
`final dose was clear. Extensive testing in the development of a prefilled insulin delivery de(cid:173)
`vice demons trates an accurate and reliable medical device.
`
`INTRODUCTION
`
`W HEN DEVELOPING NEW DRUGS, clinical trials
`
`are cond ucted to provide information on
`safety and efficacy in order to obtain regula(cid:173)
`tory approval. In gen eral, an insulin injector
`must be able to comply with global regulatory
`stan dards as well as U.S. Food an d Drug Ad(cid:173)
`ministra tion (FDA) regulations when these de(cid:173)
`vices are manufactured and distributed in the
`United States. In addition, all types of insulin
`delivery devices developed for distribution in
`Europe must comply with the Internation al Or(cid:173)
`ganiza tion of Standards (ISO). These stan dards
`are a series of laboratory-based tests that con(cid:173)
`firm the device will deliver an appropriate dose
`
`over a variety of environm ental conditions, for
`examp le, after being dropped from on e meter
`in a variety of positions. The ISO regulations
`con tain specific requirements regarding dose
`accuracy. The testing in th is article specifically
`addresses the dose accuracy requirements
`specified in the ISO guidan ce docum ent. For
`medical devices such as the insulin injector
`pens, there are no requirements to perform
`clinical trials in order to obtain European reg(cid:173)
`ulatory approval, termed the Conformity Eu(cid:173)
`ropean (CE) mark. However, the CE mark re(cid:173)
`q uires th at the pen incorporate a n umber of
`safety fea tures to ensure accurate dosage.
`The H umalog® / H um ulin® Disposable In(cid:173)
`sulin Injection pen was developed by Eli Lilly
`
`Lilly Research Laboratories, Eli Lilly and Company, Lilly Corporate Center, Indianapolis, Indiana.
`
`FRESENIUS EXHIBIT 1074
`Page 1 of 7
`
`

`

`624
`
`I.R.D. Key
`
`hook detail on Nut
`engages with hook
`on Screw key in
`self-locking mode
`
`FIG. 1. "Hook and Key" IRD stop.
`
`and Compan y (Indianapolis, IN) and first mar(cid:173)
`keted in Europe in 1998. It is capable of ad(cid:173)
`ministering Humalog® / H umulin® subcuta(cid:173)
`neously in accurate doses from 1 to 60 units, in
`1-unit increments. At the time, it was the only
`insulin pen on the market that allowed for sin(cid:173)
`gle-unit dose adjustmen t. The injector is de(cid:173)
`signed to meet the current draft of ISO stan(cid:173)
`dards1 (ISO/DIS 11608-1, 11608-2, 11608-3 TC
`84) governing design verification of pen injec(cid:173)
`tors for medical use.
`In the pen's development, it was importan t
`to establish that all doses delivered would have
`the same degree of accuracy when used by pa(cid:173)
`tien ts. This created a challen ge for Lilly's me(cid:173)
`chanical/ design engineers. The result was a
`prefilled pen that con tains an insufficient re(cid:173)
`maining dose (IRD) stop, which p revents the
`user from setting a dose greater than that re(cid:173)
`maining in the cartridge. This is don e by means
`of a hook and key detail on the nut an d screw,
`respectively. When the nut reaches the end of
`the screw, a stop in th e nut threads (hook) hits
`a stop at the end of the screw threads (key),
`thus preventing the user from dialing a dose
`greater than the amount remaining in the car(cid:173)
`tridge (Fig. 1). When this feature is engaged,
`the screw is prevented from turning by the anti(cid:173)
`backup device (ABD) fingers, w h ich are
`molded into the body halves (Fig. 2).
`In or der for the IRD stop feature to function
`properly, the hook an d key feature must be po(cid:173)
`sitioned such that the user feels an increase in
`dialing stiffness (the stop). This increase in stiff-
`
`ROE ET AL.
`
`ness should occur as the last true remaining
`dose starts to appear in the dose window. This
`indicates to the user that the number appear(cid:173)
`ing in the window is the last remaining dose,
`and signals him or her to stop turning the dial.
`In addition to the above, the IRD stop should
`be able to withstand the maximum dialing
`torque expected to be exerted by users while
`still delivering an accurate dose; otherwise, the
`user could exert enough force to override the
`stop feature, causing an under dosing of insulin
`on the final dose.
`This document will summarize a study of
`some of the torque and dose accuracy tests in(cid:173)
`volved in the development of the H uma(cid:173)
`log® /Humulin® Disposable Insulin Injection
`pen, along with the criteria used for selection
`of the appropriate model for launch-that is,
`production and distribution to health care
`providers and patients with diabetes. The pri(cid:173)
`mary purpose of the study was to determine
`whether pen users would challenge the IRD
`stop mechanism with sufficient force to affect
`the dose accuracy of the final dose. And the sec(cid:173)
`ondary purpose was to determine th e partici(cid:173)
`pant's positive an d negative impressions of the
`Humalog® /Humulin® pen an d the likelihood
`of using the new prefilled pen.
`
`MATERIALS AN D METHODS
`
`In order to evaluate the IRD stop feature and
`the IRD stop strength, three different modifi-
`
`- -
`
`·, ··,..___ -~
`
`....=A=n=ti-=Ba=c=ku=p==;-----,,
`Device
`(ABO)
`
`FIG. 2. ABO detail.
`
`FRESENIUS EXHIBIT 1074
`Page 2 of 7
`
`

`

`INSULIN DEVICE TESTING PROCESSES
`
`cations to the feature in question were evalu-
`ated and then tested by actual patients with di-
`abetes. These pen modifications will be con-
`sidered prototypes Pen1, Pen2, and Pen3. The
`modifications made were different phasings of
`the nut feature—that is, the position of the hook
`and key feature relative to the dial interface.
`
`Pens
`The Humalog®/Humulin® Disposable In-
`sulin Injection pen contains 3.0-mL cartridges
`that hold at least 300 units of Humalog®/Hu-
`mulin® or other insulin mixtures, and the pen
`is disposed of when expended. The maximum
`dose possible with this pen is 60 Units (U), or
`0.6 mL, and the smallest dose possible is 1 U,
`or 0.01 mL.
`
`Patients
`Sixty-four patients with diabetes who were
`currently taking insulin injections participated
`in the study. They completed questionnaires
`that contained both open-ended and closed-
`ended questions on new pen models and were
`also interviewed one-on-one at the end of the
`protocol exercises. The study was conducted in
`Columbus, Ohio, on April 8, 1998. Participants
`did not know that Lilly was sponsoring the
`study or that the Humalog®/Humulin® pre-
`filled pen was a Lilly product.
`
`Insufficient remaining dose accuracy
`Pens were tested for IRD dose accuracy by
`challenging the IRD stop at various torques,
`reading the indicated dose, and then expelling
`
`625
`
`the dose into a measuring device. From the
`weight and the specific gravity of the fluid, the
`exact volume expelled could be determined.
`At least 15 pens were tested at each torque.
`The challenge torque values were selected
`based on the specification for overall strength
`of the feature and evaluation of actual user
`test data.
`The graph in Figure 3 shows the results of
`the dose accuracy testing. The horizontal axis
`contains the challenge torque applied in inch-
`ounces, while the vertical axis contains the av-
`erage dose error expressed as a percentage of
`the indicated dose.
`
`User test insufficient remaining dose accuracy
`In this test, each individual was given one
`each of the three pen models, as well as a prac-
`tice pen to familiarize the individual with the
`device. The three different pens were tested
`using a Graeco-Latin square design. This de-
`sign allowed for randomization of the pen
`type as well as randomization of the dialing
`protocol sequence used. Dose sizes of 12, 18,
`28, and 41 units were chosen in order to eval-
`uate the feature under typical user conditions.
`Dose sequences were selected so that equal
`numbers of pens would hit the IRD stop at ap-
`proximately the four chosen doses. The user
`did not know when the IRD stop would be
`reached and so would adequately challenge
`the stop. Twenty-five actual uses at each
`unique dose were measured and the accuracy
`of the last dose was compared to the previous
`24 doses. The results of this testing are in-
`cluded in Figure 4.
`
`.. e .. w
`
`Cl) ,,,
`0
`0
`'#.
`
`---o- Pen 1
`_,.__Pen2
`
`-Pen3
`
`FIG. 3. Percent dose error versus patient challenge torque.
`
`
`
`FRESENIUS EXHIBIT 1074
`Page 3 of 7
`
`

`

`626
`
`ROE ET AL.
`
`.. e .. w
`
`QI .,
`0
`0
`c
`"' QI
`:Ii:
`
`□ Pen1 ■ Pen2
`
`□ Pen3
`
`- -~ -(cid:173)
`
`Absolute Difference
`
`Actual Difference
`
`FIG . 4. Patient dose accuracy tests on three different pen models.
`
`Insufficient remaining dose stop strength
`Thir ty pens from each build were tested to
`failure of the IRD stop, with the results detailed
`in the g raph in Figure 5.
`
`Level 1 and insufficient remaining dose
`accuracy simulation
`To simula te the performance of each config(cid:173)
`uration in real-world use, the user test dose ac(cid:173)
`curacy data for Pens 1, 2, and 3 were combined
`with w h at is termed level 1 dose accuracy da ta
`for con trol pens. That is, dose events a t the IRD
`stop were combined with a group of level 1
`dose accuracy data.
`In order to determ ine h ow man y IRD doses
`to include at each dose size, the following for(cid:173)
`mula was used:
`
`30-unit dose: [(30 units/ dose) x
`(60 doses/ test)]/(300 units/pen ) =
`6 pens per test
`
`ture [cool (5°C), standard (23°C with 50% rela(cid:173)
`tive h umidity), an d ho t (40°C with 50% relative
`humidity)], 1,000 trials were run in a simula(cid:173)
`tion program and the resulting K values (tol(cid:173)
`erance limit factors) were recorded. The results
`are expressed as th e probability of obtaining a
`K value high er than the required target K value.
`The target K value varied slightly for each
`group due to the different number of IRD doses
`includ ed at each dose size.
`These results are summarized in Figure 6. In
`each instance, a probability of less than 100%
`mean s that the IRD doses have an a d verse im(cid:173)
`pact on the likelihood of the prefilled pen pass(cid:173)
`ing dose accuracy standards. It is important to
`note that the representa tions in Figure 6 should
`not be confused w ith the con fidence interval or
`p con ten t of a dose accuracy test. They repre(cid:173)
`sen t the impact that the inclusion of the IRD
`dose has on th e overall dose accuracy of a cer(cid:173)
`tain body of data, with high er percen tages be(cid:173)
`ing more desirable than lower percentages.
`
`In the example above, on e would expect six
`pens to be used in th e test. At the end of each
`pen test, the IRD feature would be engaged, so
`six IRD doses are included a t the 30-unit level.
`
`60-unit dose: [(60 units/ dose) x
`(60 doses/ test)]/(300 units/pen ) =
`12 pens per test
`
`.c
`'c,
`C
`~
`ui
`0
`g;
`
`The IRD doses were combined with actual
`level 1 test data from a control model build. For
`each dose size (30 and 60 units) and tempera-
`
`Co ntrol
`
`Pen1
`
`Pen2
`
`Pen3
`
`FIG. 5. IRD stop strength versus pen type.
`
`FRESENIUS EXHIBIT 1074
`Page 4 of 7
`
`

`

`INSULIN DEVICE TESTING PROCESSES
`
`A
`
`%
`
`30 Unit K-Value
`Probabilities
`
`60 Unit K-Value
`Probabilities
`
`Pen 3
`
`%
`
`Pen 2
`
`627
`
`B
`
`Cool
`
`Std
`
`Hot
`
`Cool
`
`Std
`
`Testing Conditions
`
`Testing Conditions
`
`FIG. 6. 30U (A) and 60U (B) K value probabilities versus temperature.
`
`Pen 1
`
`Hot
`
`RESULTS
`
`Patients
`A total of 64 patients with diabetes were in(cid:173)
`volved in evaluating the three pen models (52
`patients produced a complete set of three test re(cid:173)
`sults). There were 26 males (40%) and 38 females
`(60%) included in the study; all were currently
`using insulin injections prior to the study. The
`majority of patients were using a needle and sy(cid:173)
`ringe (88%; n = 56), while 6%, or n = 4, were us(cid:173)
`ing the insulin pump, and 3%, or n = 2, were us(cid:173)
`ing the pen and 3%, or n = 2, were using both
`the pen and syringe and needles. The patient's
`age at diagnosis was 30 or younger for 30%, n =
`19, and over age 30 for 70%, n = 45.
`
`Analysis A: binomial method
`
`The absolute values of the differences from
`nominal were compared. This is the actual dif(cid:173)
`feren ce in the dose delivered by th e pen versus
`the dose indicated by th e dial on the pen at the
`beginning of the injection . This difference was
`calcula ted a t small, medium, and large dose
`sizes. Each pair-wise comparison was made
`and tested: Pen 1 versus Pen 2, Pen 1 versus
`Pen 3, Pen 2 versus Pen 3. For example, in the
`comparison of 1 versus 2, in 32 of 58 subjects,
`Pen 1 delivered a dose closer to the indicated
`value. Pen 2 was closer to indicated value in 26
`of 58 subjects. If the pens were identical, one
`would expect on the average a 29 /29 break(cid:173)
`down . Based on th e laws of probability of the
`
`binomial distribution, the p value for a 32/26
`breakdown is 0.5114, meaning that there is
`about a 50% chance of obtaining a result of
`32/26 or more extrem e (such as 34/24 or
`26/32), given the pens are n o different. There(cid:173)
`fore, 32/26 is not strong evidence to suggest
`that they are different with respect to deliver(cid:173)
`ing doses close to indicated value.
`
`Comparisons
`
`Closer to nominal
`
`Pen 1 vs. Pen 2
`Pen 1 vs. Pen 3
`Pen 2 vs. Pen 3
`
`Pen 1, 32; Pen 2, 26
`Pen 1, 18; Pen 3, 38
`Pen 2, 15; Pen 3, 42
`
`p value
`
`0.5114
`0.0111
`0.0006
`
`Analysis A results. Pens 1 an d 2 are n ot sig(cid:173)
`nificantly different at the 95% confidence level.
`Pen 3 is significantly better in dose delivery
`than Pen 1 and Pen 2.
`
`Analysis B: nonparametric test on absolute
`difference from indicated dose
`Again, the absolute value of the differential
`was calculated for each pen, each patient. The
`mean delta between the two p ens was graph ed
`and tested against 0 (zero). If a test shows a mean
`delta significantly different from 0, the devia(cid:173)
`tions (from indicated) on one pen are deemed
`larger than deviations from the other pen. Be(cid:173)
`cause a couple of the deltas were far askew from
`the other data points, the distributions in two of
`the three cases were not norm al and thus, a non(cid:173)
`parametric signed rank test was used.
`
`FRESENIUS EXHIBIT 1074
`Page 5 of 7
`
`

`

`628
`
`Analysis B results. Mean absolute value from
`indicated dose (in units) was as follows: pen 1,
`1.082; pen 2, 0.994; and pen 3, 0.354.
`
`Signed rank test results
`
`p value
`
`Pen 1 vs. Pen 2
`Pen 1 vs. Pen 3
`Pen 2 vs. Pen 3
`
`0.316
`0.000
`0.000
`
`Pen 3 is again significantly lower in devia-
`tion from the indicated dose, and no significant
`difference was found between Pen 1 and Pen 2
`(Fig. 4).
`
`Analysis C: actual differences from indicated
`The actual differences were calculated to as-
`certain the direction of bias from indicated, if
`a bias exists (Fig. 4). A negative value repre-
`sents a dose delivered which was less than the
`indicated value. The p value is asociated with
`the mean differential compared to zero (over-
`all, no bias).
`
`Analysis C results.
`
`Mean differential
`(difference in units
`from indicated dose)
`
`p value
`(difference from 0)
`
`Pen 1
`Pen 2
`Pen 3
`
`20.725
`20.895
`20.106
`
`.001
`.000
`.045
`
`As expected, all three pen means are nega-
`tive (delivered less than indicated). While all
`three means are different from zero, the bias of
`Pens 1 and 2 is almost a full magnitude larger
`than Pen 3.
`
`Insufficient remaining dose feature strength
`The graph in Figure 5 indicates that Pen
`2 and Pen 3 demonstrated higher IRD stop
`strengths, and Figure 6 indicates that Pen 3 pro-
`vides the higher percentage and therefore bet-
`ter results across all temperature conditions for
`both the 30- and the 60-U dosage amounts. Un-
`der the indicated conditions, Pen 3 was the
`more accurate of the three pen models tested.
`
`Patient results
`Patients rated the pens qualitatively on ease
`of use, accuracy, and the adequacy of the IRD
`
`ROE ET AL.
`
`stop. The majority (84%) of the patients felt that
`the pen was very easy to use, and 92% felt that
`the stop mechanism for the final dose was clear.
`Also, 77% of patients were confident in the ac-
`curacy of the pen. The majority of patients
`(70%) reported that they would be somewhat
`or very likely to use the new prefilled pen if it
`was on the market.
`The main patient concerns about using the
`new prefilled pen included concerns about not
`being able to mix different types of insulin and
`the cost/insurance issue, since, currently in the
`United States, the cost of insulin pen injectors
`is not covered by most types of medical insur-
`ance.
`
`DISCUSSION
`
`We believe that safety standards for insulin
`delivery systems should include both envi-
`ronmental and wet testing by actual patients
`with diabetes who used the insulin pens with-
`out any formal training. The results are pre-
`sented and summarized in Figures 3–6. In
`summary, the data indicate that Pen 3 dem-
`onstrated the best performance over the
`evaluation criteria, and therefore Pen 3 was
`considered the best candidate for commercial-
`ization efforts.
`The patients’ results indicated that most par-
`ticipants (84%) thought that the Humalog®/
`Humulin® pen was very easy to use. The stop
`mechanism for the final dose was rated “very
`clear” by 92%, or n 5 59, of the respondents,
`and 70% said they would be likely to use the
`new pen if it were on the market.
`These data have been corroborated by sub-
`sequent studies on the new prefilled pen’s
`functionality and patient acceptance of its use.
`In a study conducted in Spain on 42 patients
`with either type 1 or type 2 diabetes, 77% of pa-
`tients preferred the new prefilled pen over their
`previous pen, and 82% would recommend it to
`another patient with diabetes.2 Although stud-
`ies have shown that the use of insulin pens
`causes less perceived pain and results in more
`accurate insulin dosing and fewer missed in-
`jections that the use of insulin syringes,3,4 there
`are no or few clinical trials that prove they are
`associated with improved health outcomes.
`
`
`
`FRESENIUS EXHIBIT 1074
`Page 6 of 7
`
`

`

`INSULIN DEVICE TESTING PROCESSES
`
`629
`
`CONCLUSION
`
`Additional research involving prospective
`studies is needed to determine if these new pre-
`filled pens actually improve patient compli-
`ance and glycemic control, especially with type
`2 diabetes patients who may delay starting in-
`sulin therapy. If such research were to show
`that patient’s use of insulin pens improve
`health outcomes, it may be more cost effective
`for health insurers in the United States to cover
`the cost of these new medical devices.
`
`ACKNOWLEDGMENTS
`
`Appreciation is expressed to James Ander-
`son, Jr., John Lake, and Andrew Burroughs for
`technical assistance and to David C. Klonoff,
`M.D. for editorial assistance with the manu-
`script. Eli Lilly and Company sponsored this
`work.
`
`REFERENCES
`
`1. [ISO Standards] 11608-1: Pen-injectors for medical
`use—Part 1: Pen injectors—requirements and test
`
`methods. 11608-2: Pen injectors for medical use Part 2:
`Needles—requirements and test methods 11608-3 Pen
`injectors for medical use Part 3: Finished cartridges—
`requirements and test methods. Copyright by ISO, Au-
`gust 1999.
`2. Cobo A, Ignaut D, Hultman C, Reviriego J: Patients
`and health care professionals evaluate a new pre-filled
`insulin pen device. Avances en Diabetologia 2000;
`197–202.
`3. Kadiri A, Chraibi A, Marouan F, Ababou MR, el Guer-
`mai N, Wadjinny A, Kerfati A, Douiri M, Bensouda JD,
`Belkhadir J, Arvanitis Y: Comparison of NovoPen 3
`and syringes/vials in the acceptance of insulin therapy
`in NIDDM patients with secondary failure to oral hy-
`poglycaemic agents. Diabetes Res Clin Pract 1998;41:
`15–23.
`4. Graff MR, McClanahan MA: Assessment by patients
`with diabetes mellitus of two insulin pen delivery sys-
`tems versus a vial and syringe. Clin Ther 1998;20:486–
`496.
`
`Address reprint requests to:
`Michael J. Roe, P.E., M.B.A.
`Pharmaceutical Delivery Systems
`Eli Lilly and Company
`8770 Guion Road, Suite K
`Indianapolis, IN 46268
`
`E-mail: roe michael j@lilly.com
`
`
`
`FRESENIUS EXHIBIT 1074
`Page 7 of 7
`
`