DIABETES CARE, VOLUME 22, NUMBER 1, JANUARY 1999 137

Insulin is most commonly delivered toc h i l d ren with type 1 diabetes via peninjection devices and insulin syringes.

Little data exist concerning the accuracyand precision of these two methods ofinsulin delivery in children. Whereas sev-eral studies have compared glycosylatedhemoglobin levels (a measurement of dia-betes control) in patients using a pen versusa syringe (1–5), we are not aware of anystudy comparing the accuracy of these twoinsulin delivery systems in children. Theaim of this study was to compare the accu-racy and precision of insulin syringes andpen devices in children. We hypothesizedthat low insulin doses would be measure dless accurately than high insulin doses andthat a pen device may be more accurate in

delivering insulin than an insulin syringe.We also aimed to determine how fre q u e n t l ycontamination of NPH by regular insulinoccurs when children and caregivers aredrawing an insulin mixture and whetherany correlation exists between an individ-u a l ’s frequency of hypoglycemic events andthe accuracy and precision of insulin meas-u rement by the patient or care g i v e r.

RESEARCH DESIGN ANDM E T H O D S

SubjectsA total of 48 subjects, including 32 childre nwith type 1 diabetes and a parent of anadditional 16 children with type 1 diabetes,w e re re c ruited during a routine follow-up at

the Mayo pediatric outpatient diabetesclinic. Twenty-four children and adolescents(mean age of 14.1 years) were on the mul-tiple daily injection (MDI) program (oneinjection of regular insulin before each mealand ultralente at supper or bedtime) andknew how to use both pen injectors andinsulin syringes. There were 24 childre n(mean age 9.8 years) on the mixed-splitregimen (two doses of regular and NPHmixed) who re q u i red the mixing of re g u l a rand NPH and the use of insulin syringes.

Experimental methodsVials and/or cartridges were filled with a1 4C-radiolabeled solution by adding 0.0007µCi of [1 4C]glucose to 0.25 cm3 of a salinesolution. These were substituted for re g u l a rinsulin. In addition, vials were filled with a3H-radiolabeled solution by adding 0.001µCi of [3H]glucose to 0.25 cm3 of a salinesolution. These were substituted for NPHinsulin. Each drawn sample contained amaximum of 1,000 cpm. For each part i c i-pant, the amount of 1 4C and 3H present ineach vial or cartridge was determined inde-pendently by A.N.L. 3H and 1 4C concen-trations were measured by using a transferpipette (Pipetman; Rainin, Wo b u rn, MA).The coefficient of error of these standardp ro c e d u res was 1%. Dual counting of 1 4Cand 3H (with background subtraction) wasp e rf o rmed by scintillation spectro s c o p yusing a Wallac liquid scintillation counter.

ProceduresUnder supervision, subjects were i n s t ru c t e dto measure out morning insulin doses thre etimes from vials and/or cartridges contain-ing saline mixed with small amounts of[1 4C]glucose (solution used as re g u l a rinsulin) and [3H]glucose (solution used asNPH insulin). For 16 of the 24 children onthe mixed-split program, the child’s care-giver rather than the child was asked todraw up the insulin doses, because thec a regiver was routinely administering theinsulin doses. Children or caregivers withhand cuts or vision impairment and thosewho had an insulin dose that ro u t i n e l yexceeded 25 U were excluded. Subjectsregularly using a pen device (n = 24) were

F rom the Department of Pediatric Endocrinology, Mayo Clinic, Rochester, Minnesota.A d d ress correspondence and reprint requests to W. F. Schwenk, MD, Mayo Clinic, 200 First St. SW,

R o c h e s t e r, MN 55905. E-mail: schwenk.fre d e r i c k @ m a y o . e d u .Received for publication 11 June 1998 and accepted in revised form 1 October 1998.A b b re v i a t i o n s : C V, coefficient of variation; MDI, multiple daily injection.A table elsewhere in this issue shows conventional and Système International (SI) units and conversion

factors for many substances.

Accuracy of Pen Injectors Versus InsulinSyringes in Children With Type 1D i a b e t e s

O R I G I N A L A R T I C L E

O B J E C T I V E — To compare the accuracy and precision of insulin syringes and pen devicesused by children with type 1 diabetes and their pare n t s .

RESEARCH DESIGN AND METHODS — T h e re were 48 subjects (32 patients, a par-ent of an additional 16 patients) instructed to measure out morning insulin doses three timesf rom vials and/or cartridges containing saline mixed with small amounts of [1 4C]glucose (solu-tion used as regular insulin) and [3H]glucose (solution used as NPH insulin) and to dispensethe contents into a scintillation vial. Statistical analysis was used to determine the accuracy andp recision of both methods of insulin delivery.

R E S U LT S — The absolute error in measuring out doses of regular insulin 5 U was gre a t e rwith insulin syringes compared with pen injection devices (9.9 ± 2.4 vs. 4.9 ± 1.6%, re s p e c-tively). Both were comparable for regular insulin doses 5 U (3.2 ± 0.6 vs. 2.2 ± 0.4% forsyringes and pens, respectively). The accuracy in drawing up NPH doses was similar for lowand high insulin doses (mean percent error of 7.5 ± 1.5 vs. 5.6 ± 1.1%).

C O N C L U S I O N S — Pen devices are more accurate than insulin syringes in measuring outinsulin at low insulin doses. The accuracy of insulin syringes improves when higher doses ofregular insulin are measured out and becomes comparable to pen devices.

Diabetes Care22:137–140, 1999

AIDA N. LTEIF, MD

W. FREDERICK SCHWENK, MD

E m e r g i n g T r e a t m e n t s a n d T e c h n o l o g i e s

138 DIABETES CARE, VOLUME 22, NUMBER 1, JANUARY 1999

Pens versus syringes in IDDM children

asked to make measurements of re g u l a rinsulin using both the pen and the syringes.Subjects using an insulin mixture with-d rew the dose of regular insulin first andthen the dose of NPH. The content of thedrawn up samples was then dispensed intoscintillation vials. The NPH solution con-taining 3H was sampled to determine theamount of 1 4C contamination.

Becton Dickinson U100 syringes (0.3-and 0.5-ml syringes; Becton Dickinson,Franklin Lakes, NJ), Te rumo U100 syringes(0.25 ml in one-half U increments; Te ru m o ,Somerset, NJ), and pen devices, includingNovopen 1.5 (Novo Nordisk Pharm a c e u t i-cals, Princeton, NJ) and Autopen (OwenM u m f o rd, Marietta, GA), were used to exactlymatch the home pro c e d u re .

For each child, glycosylated hemoglo-bin levels were re c o rded at the time of theclinic visit. In addition, average blood glu-cose concentration and percentage of lowblood glucose measurements in the pre-ceding 250 tests re c o rded in the patient’sOne Touch glucose meter (LifeScan, Milpi-tas, CA) were determined using the OneTouch diabetes data management pro g r a m .

DefinitionsAccuracy (% error) =

expected insulin dose actual insulin dose 1 0 0

expected insulin dose

C o e fficient of variation (CV) =SD 1 0 0

mean insulin dose

Low blood glucose (hypoglycemia) = bloodglucose measurement that is below the goalrange for each individual patient ( 7 0mg/dl in our patients).

Statistical analysisData were expressed as means ± SE whenapplicable. Student’s t test was used for thecomparison of continuous variables.F i s h e r ’s exact test was used for categoricalvariables. Pearson’s correlation test and thestraight multiple re g ression analysis wereconducted to examine the re l a t i o n s h i pbetween selected variables. P 0.05 wasc o n s i d e red signific a n t .

R E S U LT S

Accuracy of pen injectors comparedwith insulin syringesWe compared the accuracy of pen injectorswith insulin syringes using the 24 subjectson the MDI program. For insulin doses

5 U (n = 9, 27 observations), the absolute

e rror in measuring out regular insulin viapen devices (mean insulin dose of 4.1 ± 0.4U) was less than the absolute error inmeasuring out regular insulin via insulinsyringes (P 0.01). However, both devicesw e re comparable for regular insulin doses

5 U (n = 15, 45 observations) (Fig. 1).

Dose-dependent accuracy of regularinsulinIn subjects using both syringes and pendevices (n = 24), the accuracy of pen injec-tors improved slightly but not signific a n t l yfor insulin doses 5 U (Fig. 1). However,syringes delivered insulin doses 5 U moreaccurately than insulin doses 5 U (Fig. 1).In terms of actual units, the absolute erro rin measuring out insulin was similar forlow and high doses of insulin whetherusing the pen or the syringe. The errors inpen measurements were 0.2 ± 0.05 and0.27 ± 0.06 U for insulin doses 5 and

5 U, re s p e c t i v e l y, while the errors insyringe measurements were 0.4 ± 0.1 and0.4 ± 0.07 U for insulin doses 5 and 5U, re s p e c t i v e l y. Participants on the mixed-split regimen also drew up regular insulindoses 5 U less accurately (mean perc e n te rror 13.4 ± 3.7 vs. 2 ± 0.6% for doses 5and 5 U, respectively) (Fig. 2). Compar-ing the dose-dependent accuracy of re g u l a rinsulin in all subjects (n = 48), we alsofound that insulin doses 5 U (meaninsulin dose of 3 ± 0.3 U) were measure dout by syringes less accurately than insulindoses 5 U (mean insulin dose of 11.1 ±0.8 U). All subjects included, the meanp e rcent error in measuring out re g u l a rinsulin via a syringe was 12.3 ± 2.6 vs. 2.9

F i g u re 1—Accuracy of pen injection devices vs. insulin syringes. The accuracy of both methods of deliv -ery was similar for insulin doses 5 U (n = 15, or 45 observations). Pen injectors were more accuratethan syringes for doses 5 U (n = 9, or 27 observations).

F i g u re 2—Accuracy of NPH measurements (error in regular insulin not taken into consideration). Theaccuracy in drawing up NPH was similar for low (n = 15, or 45 observations) and high (n = 9, or 27 obser -vations) insulin doses. For doses 5 U, regular insulin was measured out less accurately than was NPH.

DIABETES CARE, VOLUME 22, NUMBER 1, JANUARY 1999 139

Lteif and Schwenk

± 0.5% for insulin doses 5 and 5 U,re s p e c t i v e l y.

Dose-dependent accuracy of NPHAll 24 participants on the mixed-split pro-gram were included. The accuracy in meas-uring out NPH was similar for low (n = 15,45 observations) and high (n = 9, 27 obser-vations) insulin doses (Fig. 2). When thee rror in measuring out regular insulin(which is drawn up first) was taken intoconsideration, NPH was measured outm o re accurately than regular insulin fordoses 5 U (P 0.05) and less accuratelyfor doses 5 U (P 0.01) (Fig. 3).

Children compared with caregiversAll subjects using the pen devices werec h i l d ren and adolescents with diabetes.T h e re f o re, the accuracy of caregivers wasc o m p a red with the accuracy of childre nonly when an insulin mixture was used,and for regular insulin doses 5 U. Chil-d ren with diabetes (n = 5) were as accurateas their parents (n = 14) in measuring outthe insulin dose (mean percent error of14.5 ± 5.4 vs. 11.6 ± 4.2% for children andp a rents, re s p e c t i v e l y ) .

ContaminationOnly one patient contaminated the NPHvial with regular insulin during one of thet h ree measurements. We detected 8% ofthe total amount of regular insulin that wasdrawn in the NPH vial.

Precision of insulin measurementsIn subjects using both systems (n= 24), themean CV was 3.2% for pen injectors and

3.8% for syringes. In this group, the dosere p roducibility was better when insulindoses 5 U were measured out. Pen injec-tors had a CV of 4.7 ± 1 vs. 2.4 ± 0.4% forinsulin doses 5 and 5 U, respectively (P

0.05), while insulin syringes had a CV of5.3 ± 1.3 vs. 2.9 ± 0.4% for these samedoses (P 0 . 0 5 ) .

Correlation between hypoglycemicevents and percent errorR e g a rdless of insulin program, patients wereasked to test blood glucose four to five timesd a i l y. Computer analysis of re flectance glu-cose determinations was available in 41patients. The average number of tests store din the computer’s memory was four tests perd a y. When using an empirical cutoff for fre-quency of low blood glucose measure m e n t sof 20% (of total blood glucose tests in themeter), 10 of 11 children with lows thatexceeded 20% measured out more re g u l a rinsulin than expected (positive perc e n te rror). However, only 15 of 30 children withlows that did not exceed 20% drew up moreregular insulin than expected (P 0 . 0 5 ) .T h e re was a positive correlation betweenp e rcent lows and accuracy of regular insulindoses (r = 0.4, P 0.01). A positive corre-lation was also present between percent lowsand error in terms of actual units of re g u l a rinsulin (r = 0.4, P 0.01). However, boththese correlations were not significant whenc o rrected for insulin dosage and age aspotential confounding variables.

C O N C L U S I O N S — The accuracy andp recision of insulin delivery to childre nwith diabetes have rarely been assessed in

clinical practice. Weighing insulin dosesand -counting are the two techniques thathave been used for that purpose (6–11). Ina previous study, two disposable insulininjection devices (NovolinPen and Novolin-Set; Novo Nordisk Pharmaceuticals) (12)gave a remarkable degree of accuracy. Toour knowledge, no one has pre v i o u s l yc o m p a red the accuracy of pen injectorswith the accuracy of insulin syringes in thehands of children with type 1 diabetes ortheir care g i v e r s .

We found that pen injectors are moreaccurate than syringes when low doses ofinsulin are measured out. For doses 5 U,these two insulin delivery systems werecomparable. The lowest dose of insulinm e a s u red out by a pen device was 2 U.T h e re f o re, it seems that there is a clearadvantage in using pen devices for insulindoses between 2 and 5 U. For insulin dosesthat exceed 5 U, patients may choose eitherd e l i v e ry system and have similar re s u l t s .

Note that in this study, saline withadded 1 4C and 3H was used for insulinbecause of cost. We cannot rule out thepossibility that if actual insulin solutionhad been used, our results might have beend i ff e rent. For example, a pro t e i n - c o n t a i n-ing solution may affect the development ofair bubbles. However, we do not feel thatour overall results would be changed, sincethe same solution was used for both thepen injection devices and the insulins y r i n g e s .

Inaccuracies were found when twoinsulins were mixed in the syringe beforeinjection (10,11). Errors were most signifi-cant with low insulin doses (8,9,13). Pre v i-ous studies have shown that parents andpediatric nurses made large errors in draw-ing up insulin doses 2 U, and that care-givers of children with diabetes re c e i v i n glow insulin doses delivered amounts gre a t e rthan intended (7,8). Our study is consistentwith these results only for regular insulin (inthe mixed-split program), which is drawnup first. Syringes were similarly accurate forinsulin doses of NPH, whether they were

5 or 5 U. There f o re, we speculate thatthe greater inaccuracy in delivering lowdoses of regular insulin is due in part to thep resence of a small dead space and to tech-nical problems with air bubbles, whichmight be expected to have a greater effect onthe insulin that is drawn up fir s t .

Dose re p roducibility with pen injectorswas 3.2% and was comparable to are p o rted dose re p roducibility of 2% whennew Novopen injectors were used (6).

Figure 3—Accuracy of NPH measurements (error in regular insulin taken into consideration ). Theaccuracy in drawing up NPH was similar for low (n = 15) and high (n = 9) insulin doses. For doses

5 U, regular insulin was measured out more accurately than was NPH.

Dose re p roducibility with syringes in thesame subject group was 3.8%. Dose re p ro-ducibility was better for insulin doses 5 Uc o m p a red with doses 5 U. There f o re, weconclude that low doses of insulin arem e a s u red out with greater variability whenboth pen injector devices and insulinsyringes are used. For the same insulindoses, pen injectors and syringes are simi-larly re p ro d u c i b l e .

C h i l d ren (mean age of 11.5) were asaccurate as their parents in measuring outan insulin dose. In addition, re p o rted meanp e rcent error of older patients (14) usingthe syringe is similar to the mean perc e n te rror of children and adolescents who dre wup doses 5 U. There f o re, we feel thatyounger patients with type 1 diabetes canbe allowed to draw up and administer theirown dose of insulin.

It has been suggested that the higherrate of hypoglycemia in young childre nmay be due to the increased percent erro rin giving small amounts of insulin (13). Wefound that there was a simple corre l a t i o nbetween percent erro r, as well as actuale rro r, and frequency of hypoglycemia.When corrected for insulin dosage and ageas potential confounding variables, thesec o rrelations lost their signific a n c e .

One can only speculate on the clinicals i g n i ficance of this data. A large randomizedstudy in children on relatively low doses of

insulin using the insulin pen or syringeswould be needed to fully address this issue.The improved accuracy of measuring outinsulin with the pen device, however, mighthelp minimize one of the variables aff e c t i n gblood glucose concentrations.

We conclude that pen injectors areaccurate systems of insulin delivery. Theiruse should be encouraged when low dosesof insulin are being administered. Theaccuracy of insulin syringes improves andbecomes comparable to pen injectordevices when higher doses of re g u l a rinsulin are administere d .

R e f e re n c e s1 . Chase HP, Garg SK, Hoops SL, Harris S,

Wilcox W: Use of the pen delivery systemfor intensive insulin therapy in college-agestudents with type 1 diabetes. J AdolescH e a l t h12:373–376, 1991

2 . Dinneen SF, Cronin CC, O’Sullivan DJ:L o n g - t e rm efficacy of a pen injector. Ir JMed Sci160:286–287, 1991

3 . Andersson PO, Wikby A, Stenstrom U,H o rnquist JO: Pen injection and change inmetabolic control and quality of life ininsulin-dependent diabetes mellitus. D i a -betes Res Clin Pract36:169–172, 1997

4 . O’Hagan M, Greene SA: Pre-mixed insulind e l i v e red by disposable pen in the man-agement of children with diabetes. D i a b e tM e d10:972–975, 1993

5 . M u rr a y, DP, Kennan P, Gayer E, Dalmon P,

Tomkin GH, Dru ry ML, O’ Sullivan DJ: Arandomized trial of the efficacy and accept-ability of a pen injector. Diabet Med5:750–754, 1988

6 . G o rdon D, Wilson M, Paterson KR, SempleCG: An assessment of the accuracy ofNovopen 1 delivery after prolonged use.Diabet Med7:364–366, 1990

7 . Silva SR, Goodman SN, Clark L, PlotnickLP: Can caretakers of children with IDDMaccurately measure small insulin doses anddose changes? Diabetes Care19:56–59, 1996

8 . Casella SJ, Mongilio MK, Plotnick LP, Hest-b e rger MP, Long CA: Accuracy and pre c i-sion of low-dose insulin administration.P e d i a t r i c s91:1155–1157, 1993

9 . A rnott RD, Cameron MA, Stepanas TV,Cohen M: Insulin syringes: danger of deadspace. Med J Aust2:39–40, 1982

1 0 . Bell DS, Clements RS, Perentesis G, Rod-dam R, Wagenknecht L: Dosage accuracy ofself-mixed vs. premixed insulin. A rch InternM e d151:2265–2269, 1991

1 1 . B e rne C, Eriksson G, Lundgren P: Howaccurate are insulin mixtures pre p a red bythe patient? Diabetes Care9:23–26, 1986

1 2 . P e rentesis G, Clements RS: Accuracy of twonew insulin injection devices (Abstract). D i a b e t e s30:176A, 1990

1 3 . Bell DSH, Clements RS, Perentesis G: Hypo-glycemia in children with insulin-dependentdiabetes mellitus. J Pediatr340–341, 1990

1 4 . C o rcoran JS, Yudkin TS: How inaccurate isinsulin mixing? Patient variability andsyringe dead space effect. Diabet Med2:131–133, 1985

140 DIABETES CARE, VOLUME 22, NUMBER 1, JANUARY 1999

Pens versus syringes in IDDM children