Original Article

Efficacy and Safety of AIR Inhaled Insulin Comparedto Insulin Lispro in Patients with Type 1 Diabetes Mellitus

in a 6-Month, Randomized, Noninferiority Trial

Angel L. Comulada, M.D., F.A.C.E.,1 Eric Renard, M.D., Ph.D.,2 Masako Nakano, M.D., Ph.D.,3

Nadeem Rais, M.D.,4 Xuejing Mao, M.S.,5 David M. Webb, Ph.D.,5 and Zvonko Milicevic, M.D., Ph.D.6

Abstract

Background: Patients with type 1 diabetes may prefer features of AIR� inhaled insulin (developed by Alkermes,Inc. [Cambridge, MA] and Eli Lilly and Company [Indianapolis, IN]; AIR is a registered trademark of Alkermes,Inc.) over insulin injection, but the two methods need to be compared for efficacy and safety.Methods: This multicenter, 6-month, parallel-group, noninferiority trial had 500 patients with type 1 diabetesrandomized to morning doses of basal insulin glargine plus either preprandial injectable insulin lispro orpreprandial AIR insulin. We hypothesized that AIR insulin is noninferior (upper bound of the 95% confidenceinterval � 0.4%) to insulin lispro for change-from-baseline hemoglobin A1C (A1C).Results: Baseline A1C was 7.95� 0.08% for both groups. At end point, A1C was lower with insulin lispro thanwith AIR insulin by 0.27% (95% confidence interval 0.11, 0.43; P< 0.001). Noninferiority of AIR insulin to insulinlispro was not demonstrated, but similar percentages of patients in each group achieved A1C <7.0% (P¼ 0.448).Overall daily blood glucose was similar between groups at baseline (P¼ 0.879) and end point (P¼ 0.161). Two-hour postprandial blood glucose change from baseline was significantly (P< 0.001) higher with AIR insulin(20.77� 4.33 mg=dL at 3 months and 15.85� 3.08 mg=dL at end point) than with insulin lispro (3.29� 4.14 mg=dLat 3 months and 1.67� 2.91 mg=dL at end point). Overall hypoglycemia was similar between treatment groups(P¼ 0.355). The AIR insulin group had greater decrease in diffusing capacity of the lung for carbon monoxide atend point (P¼ 0.020) and greater incidence of cough (P¼ 0.024) and dyspnea (P¼ 0.030). Body weight decreased inthe AIR insulin group and increased in the insulin lispro group.Conclusions: Insulin lispro provided lower A1C than AIR insulin, but the difference may not be clinicallyrelevant.

Introduction

Patients with type 1 diabetes mellitus require multipledaily doses of insulin for the maintenance of glucose ho-

meostasis. At present, insulin is administered by subcutane-ous injection, which for some patients can induce anxiety1 andother negative emotions.2 Therefore, the availability of novelnoninjectable routes of insulin administration may lead toincreased treatment compliance, satisfaction, and quality of

life for many patients with diabetes and facilitate optimizationof therapy.

Delivery of insulin through the lung by inhalation is onepotential alternative to injections.3 The lung is an attractiveroute for drug administration because of its large surface area(about 143 m2) and thin barrier of epithelium cells (about0.61 mm) for absorption into the blood.4 The potential utility ofpulmonary insulin therapy has recently been reviewed,5,6 anda number of studies have demonstrated proof of principle for

1Instituto de Endocrinologıa, Diabetes & Metabolismo, Toa Baja, Puerto Rico.2Endocrinology Department, Centre Hospitalier Universitaire and University of Montpellier, Montpellier, France.3Eli Lilly Japan K.K., Kobe, Japan.4Chowpatty Medical Center, Mumbai, India.5Eli Lilly and Co., Indianapolis, Indiana.6Eli Lilly Regional, Vienna, Austria.This study is registered with Clinical Trial Registry Number NCT00356109 (Evaluate the Efficacy of Insulin Patients With Type 1 Diabetes)

at http:==ClinicalTrials.gov.

DIABETES TECHNOLOGY & THERAPEUTICSVolume 11, Supplement 2, 2009ª Mary Ann Liebert, Inc.DOI: 10.1089=dia.2009.0041

S-17

inhaled insulin in humans.7–16 However, there remains un-certainty regarding the safety of its long-term use.

The AIR� Inhaled Insulin System (developed by Alkermes,Inc. [Cambridge, MA] and Eli Lilly and Co. [Indianapolis, IN];AIR is a registered trademark of Alkermes, Inc.) delivers drypowdered aerosols composed of large, low-density particlesto the deep lung.17–19 The particles, which contain humaninsulin and excipient, are generated by a spray-drying tech-nology.20 Because of the low cohesive forces associated withlow mass density and high geometric-to-aerodynamic diam-eter, the particles are dispersed and aerosolized by breathactivation.

Clinical pharmacology studies demonstrated equivalentdoses of AIR insulin and subcutaneous (SC) insulin lisproprovide similar overall pharmacokinetic exposure and phar-macodynamic effect.21,22 A Phase 2 study in patients withtype 1 diabetes showed treatment with AIR insulin had sim-ilar effect on hemoglobin (Hb) A1C (A1C) compared with SCinsulin after 12 weeks of therapy. Although similar A1C lev-els were obtained, the risk for nocturnal hypoglycemia wasincreased in the AIR insulin group, possibly because thepatients had therapy of injectable mealtime insulin beforerandomization and lacked prior dosing experience with AIRinsulin.22

The primary objective of this Phase 3 study was to compareAIR insulin with SC insulin lispro in patients with type 1diabetes for up to 6 months of treatment. The trial design wasbased on the hypothesis that the two insulins have similareffect (noninferiority) on blood glucose. Other standardmeasures of insulin efficacy and safety were also evaluated.

Subjects and Methods

Study design and participants

This was a randomized, multicenter, open-label, active-comparator, two-arm, parallel-group, 6-month study (Fig. 1)conducted by physicians at 49 centers in the United States,India, Mexico, France, Germany, Argentina, Belgium, Italy,and Puerto Rico. A total of 500 patients with type 1 diabetesfor�24 months, two or more insulin injections per day for�2months, A1C <11.0%, �18 years old, and nonsmokers for �6months before study entry were enrolled. Patients who werebreastfeeding, pregnant, or intended to become pregnantduring the study period were excluded. All patients wererequired to undergo pulmonary function tests (PFTs) ac-cording to American Thoracic Society standards23,24 and havediffusing capacity of the lung for carbon monoxide (DLCO)>70% of predicted, forced expiratory volume in 1 s (FEV1)=forced vital capacity (FVC) greater than the lower limit ofnormal, and FEV1 >70% of predicted. Patients were excludedfrom this study if they had a history of a clinically relevantpulmonary disease or lung cancer or obvious clinical signs orsymptoms of liver, kidney, or heart disease within 6 months ofstudy entry. Patients signed informed consent before partici-pating in the study. This study was conducted in accordancewith the Declaration of Helsinki and the International Councilon Harmonization Guidelines to Good Clinical Practice.

Treatments

This study consisted of a 2–4-week lead-in period, 6 monthsof study treatments, and 2 months of a follow-up safety as-

sessment. During the lead-in period, all patients optimizedtheir daily dosage schedule of prandial insulin lispro (Eli Lillyand Co.) plus a morning dose of insulin glargine (SanofiAventis, Paris, France). Patients were then randomly assignedto one of two treatment groups: prandial AIR insulinþ basalinsulin glargine or prandial injectable insulin lisproþ basalinsulin glargine. Patients were monitored for the efficacy andsafety of these treatments at 1-, 3-, and 6-month visits. Duringthe 2-month follow-up, all patients in both treatment groupswere treated with a regimen of insulin lisproþ insulin glar-gine and were assessed for safety at the 8-month visit (Fig. 1).

The AIR Inhaled Insulin System in this study consistedof the Lilly=Alkermes insulin inhaler and human insulin in-halation powder loaded into a cellulose capsule, as well asthe directions for use. The capsules were available in twostrengths: low (2 U equivalent; 0.9 mg of human insulin) andmid (6 U equivalent; 2.6 mg of human insulin). Insulin glargineand insulin lispro were injected with a syringe or reusable pen.The prandial insulins, AIR insulin and insulin lispro, wereadministered within 15 min before the start of a meal. Thebasal insulin, insulin glargine, was administered in the morn-ing to minimize the risk of nocturnal hypoglycemia. Insulindoses were adjusted for patients to target A1C <7% and totarget daily blood glucose values based on the recommenda-tions of the American Association of Clinical Endocrinol-ogists25 converted to equivalent plasma values based on theInternational Federation of Clinical Chemistry and LaboratoryMedicine (IFCC) recommendations.26,27 The IFCC-equivalenttargets were 75 mg=dL (4.2 mmol=L) to 115 mg=dL (6.3mmol=L) for fasting and preprandial blood glucose concen-trations and � 145 mg=dL (8.0 mmol=L) for 2-h postprandialblood glucose (PPBG) concentrations.

Efficacy measurements

During office visits that occurred at randomization,3 months, and 6 months, A1C levels were assessed by high-performance liquid chromatography (VARIANT� II TURBOHemoglobin A1C Program, Bio-Rad Laboratories, Inc., Her-cules, CA) at regional laboratories operated by QuintilesTransnational Corp. (Research Triangle Park, NC).

FIG. 1. Study design showing times of measurements.Antibodies are cross-reactive, human insulin-specific, andinsulin lispro-specific anti-insulin antibodies. SMBG repre-sents eight-point SMBG measurements.

S-18 COMULADA ET AL.

Patients obtained eight-point self-monitored blood glu-cose (SMBG) profiles on three separate days during the 2weeks before visits (Fig. 1). All glucose measurements werenormalized to equivalent IFCC plasma glucose values as de-scribed by Mathieu et al.27 in an accompanying article. Eachprofile consisted of blood glucose measurements taken be-fore the start of morning, midday, and evening meals, 2 h(�30 min) after the beginning of the morning, midday, andevening meals, at bedtime, and at 3 a.m. (�1 h).

Safety measurements

All PFTs, including FEV1, FVC, total lung capacity (TLC),and DLCO tests, were performed during scheduled officevisits (Fig. 1) using criteria that met or exceeded the guidelinesestablished by the American Thoracic Society.23,24 All testresults were assessed for quality at a central quality assur-ance center, which was blinded to the identity of treatmentgroups. The local testing service provided DLCO uncorrectedvalues based on a standardized Hb value for all patients,with some services using gender-specific values. These DLCO

values were corrected for Hb as follows: for men, DLCO cor-rected¼DLCO uncorrected�(10.22þHb)=1.7 Hb; and forwomen, DLCO corrected¼DLCO uncorrected�(9.38þHb)=1.7 Hb, where Hb is the Hb value (in g=dL) of each patientsample assayed by the central laboratory. As described byGarg et al.,28 further evaluation of pulmonary status wasperformed if any prespecified for-cause (reduced lung func-tion) triggers occurred in a study patient.

A hypoglycemic episode was defined as any time a patientfelt a sign or symptom of hypoglycemia or had an IFCC bloodglucose level <68 mg=dL (3.8 mmol=L) with or without signs,symptoms, or treatment of hypoglycemia. Severe hypogly-cemia was defined as an episode when the patient requiredassistance of another person and had either a blood glucoselevel <55 mg=dL (3.1 mmol=L) or symptoms that ceased aftertreatment with oral carbohydrate, glucagon, or intravenousglucose. Nocturnal hypoglycemia was defined as any eventthat occurred after bedtime and before breakfast.

Whole blood samples were collected from patients at ran-domization, at 1 and 6 months of study treatment, and atfollow-up 2 months after completion of study treatment.From these samples, insulin-specific, insulin lispro-specific,and cross-reactive antibodies were determined by a vali-dated quasi-quantitative, radioligand-binding assay (Milli-pore Corp., Regulatory Compliant Lab, St. Charles, MO)similar to previously reported methods.29–31 Serum sampleswere classified as positive if binding levels were � 1.0% forinsulin-specific antibodies, �0.8% for insulin lispro-specificantibodies, and �1.3% for cross-reactive antibodies.

Each patient had blood chemistry and hematology mea-sured at baseline and 6-month end point as well as Hb mea-sured at each visit. Other safety assessments includedreporting adverse events and measuring vital signs and bodyweight at each visit.

Statistical analyses

An analysis of covariance model was used to establish thenoninferiority of AIR insulin to insulin lispro. Treatmentnoninferiority was established if the upper bound of the two-sided 95% confidence interval (CI) of the difference in A1Cchange was �0.4%. Anticipating a 15% dropout rate, an in-

tent-to-treat (ITT) analysis with 442 completers (221 pertreatment group) would have 96% power to show non-inferiority if AIR insulin was truly noninferior to the com-parator. The method of last observation carried forward(LOCF) was used for patients who dropped out of the studyor had missing data at the final treatment period visit.

To evaluate patient characteristics, a two-sample t test wasused for continuous measures, and Pearson’s w2 test was usedfor categorical measures. If test conditions were not met,Fisher’s exact test was used. To evaluate hypoglycemia rateand insulin antibodies, a nonparametric analysis was used.To evaluate other continuous variables, a model similar to theprimary efficacy analysis was used. To evaluate the propor-tion of patients with A1C <7.0% and �6.5%, a logistic re-gression was used. To evaluate other dichotomous variables,Pearson’s w2 test was used. If test conditions were not met,Fisher’s exact test was used.

The ITT population included all randomized patients whohad one or more measurements after baseline. A per-protocolpopulation was defined as ITT patients who sufficiently com-plied with randomly assigned treatment, completed the finaltreatment period visit, and did not have any major protocolviolations. The safety population was defined as all ran-domized patients who received one or more doses.

Results

Of 807 patients with type 1 diabetes screened, 500 wererandomized to treatment with either AIR insulin or insulinlispro. Details of patient disposition after randomization areshown in Figure 2. More patients discontinued from the AIRinsulin group (57 patients) than from the insulin lispro group(34 patients) (P¼ 0.008). Discontinuations because of adverseevents were few in both groups, whereas discontinuationsdue to patient decision and ‘‘other’’ reasons were more fre-quent in the AIR insulin group than insulin lispro group. Nopatient deaths were reported during the study period orfollow-up. A total of eight patients were discontinued fromstudy because they did not adhere to the protocol. The AIRinsulin and insulin lispro treatment groups were similar for allpatient demographics and baseline characteristics (Table 1).

FIG. 2. Patient disposition by number (%) of patients.Other reasons for discontinuation were protocol violation,physician decision, and sponsor decision.

AIR INHALED INSULIN IN TYPE 1 DIABETES S-19

Glycemic control

A1C. The A1C values (least squares mean [LSM]�standard error [SE]) for both treatment groups were identicalat baseline (7.95� 0.08%), whereas at end point, A1C was8.11� 0.06% for the AIR insulin group and 7.85� 0.06% forthe insulin lispro group. The change in A1C from baseline to 3months was similar between treatment groups (P¼ 0.183).However, at LOCF 6-month end point, A1C increased for AIRinsulin (0.17� 0.06%) and decreased for insulin lispro(�0.10� 0.06%) from baseline for a significant difference of0.27% (95% CI 0.11, 0.43) between the treatment groups(P< 0.001). The 0.43% upper limit of this 95% CI was greaterthan the 0.4% maximum difference established a priori fornoninferiority; thus, this test was inconclusive for nonin-feriority.32 Consequently, AIR insulin failed to achieve non-inferiority for A1C compared with insulin lispro aftertreatments for 6 months in these patients with type 1 diabetes.Treatment effect on change in A1C from baseline to end pointwas not significantly different among the nine countrieswhere the study was conducted (P¼ 0.639 for treatment bycountry interaction).

The percentages of patients achieving an A1C goal of<7.0% at end point were similar between treatment groups,with 22.8% (54 of 237) in the AIR insulin group and 27.0% (65of 241) in the insulin lispro group (P¼ 0.448). The percentagesof patients achieving an A1C goal of�6.5% at end point werealso similar, with 13.9% (33 of 237) of patients in the AIRinsulin group and 17.0% (41 of 241) in the insulin lispro group(P¼ 0.404).

SMBG. At baseline, SMBG values at eight daily timepoints were similar between the AIR insulin and insulin lisprogroups. At the 6-month end point, the AIR insulin grouphad significantly lower morning-fasting SMBG, significantly

higher 2-h postprandial and bedtime SMBG values, andsimilar midday preprandial, evening preprandial, and 3 a.m.SMBG values compared with those for the insulin lisprogroup (Fig. 3). Changes in overall daily SMBG values were

Table 1. Demographic and Baseline Characteristics

Treatment group

Insulin lispro (n¼ 251) AIR insulin (n¼ 249)

Age (years) 39.7 (13.5) 38.8 (13.0)Gender [n (%)]

Female 104 (41.4) 111 (44.6)Male 147 (58.6) 138 (55.4)

Origin [n (%)]Caucasian 156 (62.2) 156 (62.7)African 4 (1.6) 6 (2.4)Hispanic 44 (17.5) 40 (16.1)Other 47 (18.7) 47 (18.9)

Height (cm) 167.60 (10.01) 167.93 (10.51)Weight (kg) 72.04 (17.38) 72.11 (17.13)BMI (kg=m2) 25.32 (4.66) 25.15 (4.29)Duration of diabetes (years)a 16.60 (10.88) 17.16 (11.08)Insulin and glucose measures

Daily prandial insulin (U=kg of body weight) 0.39 (0.02)b 0.37 (0.02)b

Daily total insulin (U=kg of body weight) 0.80 (0.02)b 0.76 (0.02)b

A1C�8.5% [n (%)] 174 (69.3) 179 (71.9)A1C>8.5% [n (%)] 77 (30.7) 70 (28.1)Daily LSM 8-point blood glucose (mg=dL) 166.28 (2.50)b 166.76 (2.52)b

Data are given as mean (SD) except where noted. BMI, body mass index.aNumber of patients for AIR insulin treatment group was 248.bMean (SE).

FIG. 3. Averages of three daily eight-point SMBG values(LSM� SE) and the daily insulin doses collected at the LOCFend point for treatment with prandial AIR insulin or insulinlispro, plus basal insulin glargine. Postprandial is 2 h afterthe start of a meal. The P values statistically compare thechange from baseline to end point (data not shown) betweenthe treatment groups for blood glucose at each daily timepoint.

S-20 COMULADA ET AL.

similar between the groups at 3-month and 6-month endpoints; however, the overall 2-h PPBG (three post-mealtimescombined) (LSM� SE) increased in both groups, but thechange was significantly (P< 0.001) greater in the AIR insulingroup than in the insulin lispro group at 3 months (AIR in-sulin, 20.77� 4.33 mg=dL; insulin lispro, 3.29� 4.14 mg=dL)and end point (AIR insulin, 15.85� 3.08 mg=dL; insulin lispro,1.67� 2.91 mg=dL).

Insulin doses. At end point, the total daily prandial in-sulin doses (LSM� SE, adjusted for body weight) were similarbetween the AIR insulin (0.39� 0.01 U=kg) and insulin lispro(0.37� 0.01 U=kg) groups (P¼ 0.270). The total daily insulindoses (including AIR insulin or insulin lispro, and insulinglargine) were also similar between the AIR insulin (0.79�0.02 U=kg) and insulin lispro (0.79� 0.02 U=kg) groups (P¼0.979). Mean prandial and basal insulin doses by time of dayare shown in Figure 3.

Safety

Hypoglycemia. The overall incidence of hypoglycemia inthe safety population was similar between the AIR insulingroup (88.2%, 217 of 246) and insulin lispro group (85.0%, 210of 247) (P¼ 0.355). At 1 month, the rate of hypoglycemia(events=patient=30 days) was significantly higher in the AIRinsulin group (rate¼ 8.00) than in the insulin lispro group(rate¼ 6.46) (P¼ 0.013). The rates were similar between thegroups at 3 months (AIR insulin rate¼ 5.20; insulin lisprorate¼ 5.43; P¼ 0.825) and end point (AIR insulin rate¼ 4.27;insulin lispro rate¼ 4.79; P¼ 0.651). The rates of hypoglyce-mia were highest during the first month for both groups.

The rates of nocturnal hypoglycemia at 1 and 3 monthswere significantly higher in the AIR insulin group than in

the insulin lispro group (P< 0.01); however, at end point thisdifference was smaller and not significant (P¼ 0.071). Themost significant difference in the rates of nocturnal hypogly-cemia occurred in the first month between the AIR insulin(rate¼ 2.91) and insulin lispro (rate¼ 1.68) groups (P< 0.001).

The incidences (percentages of patients) of severe hypo-glycemia between AIR insulin and insulin lispro treatmentgroups were similar overall between AIR insulin (19.9% [49 of246]) and insulin lispro (17.8% [44 of 247]), and similar duringany time interval of the study.

Pulmonary function and ‘‘for-cause’’ evaluations. Base-line PFT values and changes from baseline to end point arepresented by treatment group in Table 2. The PFT values forboth treatment groups were similar at baseline. At end point,DLCO decreased significantly more in the AIR insulin group(�1.104 mL=min=torr) than in the insulin lispro group(�0.467 mL=min=torr) (P¼ 0.020). At follow-up, 2 monthsafter discontinuation of AIR insulin, change from baseline inDLCO for the AIR insulin group (�1.047 mL=min=torr) wasstill significantly lower than for the insulin lispro group(�0.240 mL=min=torr) (P¼ 0.011). The changes from baselineat end point and follow-up for FEV1, FVC, FEV1=FVC, andTLC were similar between the AIR insulin and insulin lisprogroups.

For-cause evaluations are summarized in Table 3. Of all110 initial for-cause triggers, a decrease in DLCO of �3.5 mL=min=torr from baseline was the most common for bothtreatment groups, occurring in 64 patients in the AIR insulingroup and 36 in the insulin lispro group. Additionally, of all18 first-repeat for-cause triggers, a decrease in DLCO was alsothe most common, occurring in nine patients in the AIR in-sulin group and six patients in the insulin lispro group. Of all128 initial and first-repeat for-cause triggers, 46 (36%) were

Table 2. PFTs by Treatment: LSM (SE) at Baseline and Change from Baseline to End Point

and 2-Month Follow-Up

BaselineChange from baselineto LOCF end point Change from baseline to follow-up

n LSM (SE) n LSM (SE) n LSM (SE)

FEV1 (L)Insulin lispro 251 3.262 (0.031) 244 �0.031 (0.015) 214 �0.034 (0.016)AIR insulin 249 3.270 (0.032) 243 �0.047 (0.015) 191 �0.014 (0.017)

P value for difference 0.842 0.405 0.344FVC (L)

Insulin lispro 251 4.022 (0.040) 244 �0.031 (0.016) 212 �0.052 (0.019)AIR insulin 249 4.038 (0.040) 243 �0.031 (0.017) 191 �0.017 (�0.020)

P value for difference 0.745 0.988 0.151FEV1=FVC

Insulin lispro 251 0.814 (0.004) 244 �0.001 (0.002) 212 0.004 (0.002)AIR insulin 249 0.814 (0.004) 243 �0.005 (0.002) 191 0.000 (0.003)

P value for difference 0.847 0.065 0.237DLCO (mL=min=torr)

Insulin lispro 251 26.045 (0.323) 243 �0.467 (0.219) 206 �0.240 (0.247)AIR insulin 249 26.528 (0.327) 243 �1.104 (0.220) 184 �1.047 (0.261)

P value for difference 0.234 0.020 0.011TLC (L)

Insulin lispro 233 5.640 (0.056) 185 �0.042 (0.042) Not measuredAIR insulin 226 5.620 (0.057) 185 �0.067 (0.044) Not measured

P value for difference 0.771 0.641 —

AIR INHALED INSULIN IN TYPE 1 DIABETES S-21

confirmed, and 10 (eight patients taking AIR insulin and twotaking insulin lispro; 8% total) had a new, significant abnor-mal finding on high-resolution computed tomography scan-ning. Nine of these 10 abnormal findings were not consideredclinically significant after additional evaluation by the inves-tigator. One finding in the AIR insulin group was reported asallergic alveolitis, which was later considered to be a seriousadverse event.

Anti-insulin antibodies. Table 4 compares the bindinglevels for cross-reactive, insulin-specific, and lispro-specificantibodies between the two treatment groups at baseline,change-from-baseline at end point, and follow-up. For allthree antibody types, the binding levels among treatmentgroups were similar at baseline, but in the AIR insulin grouplevels increased significantly and in the insulin lispro group

levels changed little from baseline to end point to follow-up.At 2-month follow-up, mean antibody levels were lower thanat end point for the AIR insulin group but did not return tobaseline levels.

Adverse events. Overall, 73.1% (182 of 249) of patients inthe AIR insulin group and 64.9% (163 of 251) of patients in theinsulin lispro group had one or more treatment-emergentadverse events (TEAEs) (P¼ 0.053 for treatment difference).The most frequent TEAEs were headache, nasopharyngitis,cough, and pharyngolaryngeal pain. Cough occurred in 28patients taking AIR insulin and 14 patients taking insulinlispro (P¼ 0.024). Dyspnea occurred in five patients takingAIR insulin and none of the patients taking insulin lispro(P¼ 0.030). No overall between-group differences were ob-served in mean change from baseline to end point in diastolicand systolic blood pressures, pulse rate, respiratory rate, andbody temperature.

Body weight. At both 3 and 6 months after baseline, bodyweight (LSM� SE) decreased in the AIR insulin group andincreased in the insulin lispro group. At 3 months, the changein body weight was significantly lower for the AIR insulingroup (�0.28� 0.17 kg) than the insulin lispro group (0.37�0.16 kg) (P¼ 0.002); at the 6-month end point, it was signifi-cantly lower for the AIR insulin group (�0.31� 0.19 kg) thanthe insulin lispro group (0.70� 0.19 kg) (P< 0.001).

Discussion

In this study, noninferiority of prandial AIR insulin to SCinsulin lispro was not demonstrated in the mean change inA1C from baseline in patients with type 1 diabetes. Aftertreatment for 6 months, A1C change from baseline was sig-nificantly lower in the insulin lispro group than in the AIRinsulin group (AIR insulin minus insulin lispro¼ 0.27% A1C).

Table 3. For-Cause Evaluations by Number

and Percentage of Randomized Patients

(Safety Population)

Treatment group [n (%)]

Insulin lispro AIR insulin

Randomized to treatment 251 (100) 249 (100)Initial for-cause trigger 43 (17) 67 (27)First-repeat for-cause

trigger8 (3) 10 (4)

Confirmed change frombaseline PFTa,b

13 (5) 33 (13)

HRCT new significantabnormalityb

2 (0.8) 8 (3)

HRCT, high-resolution computed tomography.aConfirmed by repeated PFTs.bTotals are initial and first-repeat for-cause triggers combined.

Table 4. Binding Levels for Cross-Reactive, Insulin-Specific, and Lispro-Specific Antibodies

at Baseline and Change-from-Baseline at End Point and Follow-Up

Percentage by treatment group

Insulin lispro AIR insulinn Mean (SD) n Mean (SD) P value

Cross-reactiveBaseline 241 6.33 (9.52) 241 6.29 (9.76) 0.742Change from baseline

End pointa 234 �0.88 (6.41) 238 14.46 (15.44) <0.001Follow-upb 176 �1.41 (7.05) 155 11.03 (13.70) <0.001

Insulin-specificBaseline 241 0.40 (1.27) 240 0.36 (1.11) 0.401Change from baseline

End pointa 234 0.01 (0.61) 237 0.55 (1.58) <0.001Follow-upb 176 �0.02 (0.68) 154 0.41 (1.17) 0.005

Lispro-specificBaseline 241 0.21 (0.38) 240 0.15 (0.20) 0.269Change from baseline

End pointa 234 0.01 (0.62) 237 0.23 (0.72) <0.001Follow-upb 176 0.03 (0.49) 154 0.16 (0.50) 0.005

aLOCF.b2 months after study end point.

S-22 COMULADA ET AL.

Although this difference was statistically significant, the up-per bound of the 95% CI (0.11, 0.43) was greater than thepredefined clinically relevant difference of 0.4% for non-inferiority, such that noninferiority between these treatmentswas inconclusive. With this CI, it is possible the true treatmentdifference in A1C is >0.4% (detailed explanation in extensionto the CONSORT statement32).

Two other noninferiority studies22,28 also compared AIRinsulin with SC injected prandial insulins (insulin lispro orregular human insulin) in patients with type 1 diabetes. Asseen in the present study, Garg et al.28 found the SC insulingroup had significantly lower A1C change from baseline at 24months than did the AIR insulin group (AIR insulin minusinsulin lispro¼ 0.44% A1C), yet because the upper bound ofthe 95% CI (0.24, 0.63) was greater than the predefined 0.4%margin for noninferiority, determination of noninferioritybetween the two treatments was inconclusive. In contrast, the2006 Phase 2, crossover study by Garg et al.22 showed AIRinsulin to be noninferior to SC insulin for mean A1C values at3 months of treatment, a different result that may be related tothe shorter length of the study. At 3 months, we too observedno significant difference in A1C between AIR insulin and in-sulin lispro (data not shown). The overall higher A1C found inthe AIR insulin group compared with SC insulin group at6 and 24 months, but not at 3 months, suggests any differencebetween these insulin treatments on overall glycemic controloccurs gradually in patients with type 1 diabetes. Still, allthree studies found no significant difference between AIRinsulin and SC insulin groups in the proportion of total pa-tients who achieved the goal of A1C <7.0% at the treatmentend points. The absence of a significant difference suggestsmotivated patients who prefer to inhale, rather than inject,insulin can successfully achieve their glycemic goals.

The higher PPBG in patients taking AIR insulin than inthose taking insulin lispro may explain the difference in A1Cbetween the treatment groups. This daily blood glucose pat-tern is consistent with pharmacokinetic results showing AIRinsulin has equivalent time to maximum concentration butlower maximum concentration and a greater duration of ac-tion compared with equivalent doses of insulin lispro.21

Therefore, AIR insulin lowers PPBG more slowly, whichcould explain why AIR insulin resulted in higher blood glu-cose during the 1.5–2.5-h time of PPBG measurement butcomparable preprandial blood glucose and lower fastingblood glucose relative to insulin lispro. Patients using AIRinsulin may have also been reluctant to increase their insulindose to lower 2-h PPBG further because it could increase therisk of hypoglycemia later in the postprandial period. Finally,lack of experience with inhaled insulin could have negativelyaffected patients’ ability to appropriately adjust the dose ofAIR insulin. This notion is supported by the results of Garget al.,22 who showed that AIR inhaled insulin and injectableprandial insulin yield similar PPBG control in patients withtype 1 diabetes. The PPBG accounts for about 50% of A1C inpatients with type 2 diabetes who have A1C of 7.3–8.4%33

(baseline A1C in this trial was 7.95� 0.08%). Although treat-ment with AIR insulin was associated with significantly lowermorning fasting blood glucose, it appears that higher PPBG inthis treatment group influenced A1C more than did improvedfasting blood glucose.

AIR insulin, as compared with insulin lispro, had a signif-icantly higher rate of all hypoglycemia at 1 month but not at

3 and 6 months of treatment, and a higher rate of noctur-nal hypoglycemia at 1 and 3 months but not at 6 months oftreatment. In 2006, Garg et al.22 also found the rate of allhypoglycemia to be significantly higher in patients taking AIRinsulin at 1 month but not at 3 months; additionally, the rate ofnocturnal hypoglycemia was significantly higher in patientstaking AIR insulin than in those taking SC insulin for theentire 3-month study. In particular, they found the rate ofnocturnal hypoglycemia to be significantly higher with AIRinsulin treatment in the subgroup of patients previouslytreated with insulin lispro. The investigators attributed thisfinding to patients who were newly taking AIR insulin beingunaccustomed to its longer pharmacokinetic duration ex-tending into the night. The lower rate of hypoglycemia withAIR insulin during the later periods of these studies indicatespatients learned from experience to reduce hypoglycemia.

Among the PFTs in this study, only DLco demonstratedsignificantly lower values in the AIR Insulin group than in theinsulin lispro group at end point and 2-month follow-up. Theother studies with AIR insulin in patients with type 1 diabetesrevealed similar PFT results; however, DLCO values eitherfully recovered to those of SC insulin 2 months after patientscompleted the 24-month study treatments28 or recovered tonear that of baseline after patients crossed over to taking SCinsulin for 3 months.22 Another inhaled insulin, Exubera�

(Pfizer Inc. [New York, NY] and Nektar Therapeutics [SanCarlos, CA]), compared with SC insulin in patients with type1 diabetes, had a small (<2%) but statistically significant de-crease in DLCO within 3 months of initiation, and the differ-ence between treatment groups lasted 2 years without furtherprogression.34 Within 1 month of the subsequent washoutperiod, DLCO recovered in the Exubera group to the valuesrealized in the SC insulin group.35 For reasons not clear in ourstudy, the 2-month follow-up period did not show the samedegree of DLCO recovery demonstrated in these other studies.In total, the mean loss of DLCO associated with inhaled insulinoccurs early during the course of treatment, is thereafter notprogressive relative to treatment with SC insulin, is partiallyor fully reversible, and is not a clinically relevant change forhealthy, nonsmoking adults.36

Insulin-specific, lispro-specific, and cross-reactive antibodyformation increased significantly in the AIR insulin group atend point and did not return completely to baseline levels atthe 2-month follow-up (Table 4). These results are consistentwith findings during the first 6 months of treatment and later2-month follow-up of the 24-month study of AIR insulinversus SC insulin in patients with type 1 diabetes.28 A meta-analysis of Exubera studies by Fineberg et al.37 demonstratedthat anti-insulin antibodies develop substantially higher titerswith use of inhaled insulin than with SC insulin. When usinginhaled insulin, patients with type 1 diabetes develop sub-stantially higher titers than patients with type 2 diabetes. In 2-year studies, anti-insulin antibody levels increased during thefirst 12–15 months and decreased thereafter.28,35 No correla-tion between anti-insulin antibodies and A1C has been found.We also observed no association between anti-insulin anti-bodies and treatment safety or efficacy.

AIR insulin provided significantly lower change-from-baseline body weight at 3 and 6 months compared to insulinlispro. Less weight gain was also found with inhaled insulincompared to SC insulin in a meta-analysis of 6-month endpoints from five pooled studies of Exubera in type 1 and type 2

AIR INHALED INSULIN IN TYPE 1 DIABETES S-23

patients38 and longer-term at the 2-year end points for type 1patients in studies of AIR insulin28 and Exubera.34 This dif-ference in body weight between inhaled and SC insulins hasnot been consistently associated with insulin dosage or gly-cemic control, and the therapeutic or physiological basis isunclear.

Potential limitations of this and other studies comparinginhaled insulin and SC insulin are the mechanical and for-mulary differences between the respective delivery systems.Because of the mechanical difference in drug delivery, thetreatments were unblinded. The AIR insulin System deliversdoses in estimated increments of 2 U bioequivalent to SC in-sulin, whereas SC insulin systems administer doses in incre-ments of �1 units. This difference in dose capacity was notapparent in the mean prandial insulin dosages between thetreatment groups, perhaps because the numerous decisionspatients made to take either more or less AIR insulin whensingle-unit increments would be preferred offset each other inboth directions. On the other hand, the higher A1C, higherPPBG, and lower weight gain in the AIR insulin group mightindicate a tendency for those patients to underdose, fearful ofhypoglycemia that could occur if they chose to take the ad-ditional 2 U of insulin. If so, the similar doses found betweentreatment groups in this study might indicate the bioequiva-lence of AIR insulin relative to insulin lispro is somewhatoverestimated. Another limitation to this study is that patientsin the AIR insulin group had no prior experience with inhaledinsulin, whereas the insulin lispro group did have prior ex-perience with their insulin system. These issues should beconsiderations for future research, development, and under-standing of insulin delivery systems.

In summary, AIR insulin did not demonstrate non-inferiority to insulin lispro for a predefined minimal, clinicallyrelevant difference of 0.4% A1C change from baseline duringup to 6 months of treatment in patients with type 1 diabetes.Although insulin lispro provided a statistically significant0.27% lower A1C for glycemic control, the clinical relevanceof this difference is uncertain because a similar proportion ofpatients reached their glycemic targets in the two groups andnoninferiority was inconclusive. With the exception of re-duced DLCO and increased cough and dyspnea, the safetyprofile of AIR insulin was comparable to that of insulin lispro.These results are consistent with those of other inhaled-insulinstudies in patients with type 1 diabetes and are importantconsiderations for the future development and use of AIRinsulin and other insulin delivery systems.

Acknowledgments

This work was sponsored by Eli Lilly and Company and isrelated to study protocol H7U-MC-IDAV. We acknowledgethe writing assistance provided by Keyra D. Martinez, M.D.,of Primo Scientific, Inc., Panama City, Panama and technicalassistance on antibody assays provided by Mark O’Dell, B.A.,and Ronald R. Bowsher, Ph.D., of Millipore BioScience Divi-sion, St. Charles, MO.

Author Disclosure Statement

A.L.C., E.R., and N.R. are consultants to Eli Lilly andCompany and have also received research grants from thecompany. Z.M., X.M., M.N., and D.M.W. are employees andshareholders of Eli Lilly and Company.

References

1. Zambanini A, Newson RB, Maisey M, Feher MD: Injectionrelated anxiety in insulin-treated diabetes. Diabetes Res ClinPract 1999;46:239–246.

2. Hayes RP, Bowman L, Monahan PO, Marrero DG, McHor-ney CA: Understanding diabetes medications from the per-spective of patients with type 2 diabetes: prerequisite tomedication concordance. Diabetes Educ 2006;32:404–414.

3. Hoffman A, Ziv E: Pharmacokinetic considerations of newinsulin formulations and routes of administration. ClinPharmacokinet 1997;33:285–301.

4. Gehr P, Bachofen M, Weibel ER: The normal human lung:ultrastructure and morphometric estimation of diffusioncapacity. Respir Physiol 1978;32:121–140.

5. Heinemann L, Pfutzner A, Heise T: Alternative routes ofadministration as an approach to improve insulin therapy:update on dermal, oral, nasal and pulmonary insulin de-livery. Curr Pharm Des 2001;7:1327–1351.

6. Owens DR: New horizons—alternative routes for insulintherapy. Nat Rev Drug Discov 2002;1:529–540.

7. Cefalu WT, Skyler JS, Kourides IA, Landschulz WH, Ba-lagtas CC, Cheng S, Gelfand RA: Inhaled human insulintreatment in patients with type 2 diabetes mellitus. AnnIntern Med 2001;134:203–207.

8. Elliott RB, Edgar BW, Pilcher CC, Quested C, McMaster J:Parenteral absorption of insulin from the lung in diabeticchildren. Aust Paediatr J 1987;23:293–297.

9. Heinemann L, Traut T, Heise T: Time-action profile of in-haled insulin. Diabet Med 1997;14:63–72.

10. Hermansen K, Ronnemaa T, Petersen AH, Bellaire S,Adamson U: Intensive therapy with inhaled insulin via theAERx insulin diabetes management system: a 12-weekproof-of-concept trial in patients with type 2 diabetes. Dia-betes Care 2004;27:162–167.

11. Jendle JH, Karlberg BE: Effects of intrapulmonary insulin inpatients with non-insulin-dependent diabetes. Scand J ClinLab Invest 1996;56:555–561.

12. Jendle JH, Karlberg BE: Intrapulmonary administrationof insulin to healthy volunteers. J Intern Med 1996;240:93–98.

13. Laube BL, Georgopoulos A, Adams GK III: Preliminarystudy of the efficacy of insulin aerosol delivered by oralinhalation in diabetic patients. JAMA 1993;269:2106–2109.

14. Rave K, Nosek L, Heinemann L, Gonzales C, Ernest CS2,Chien J, Muchmore D: Inhaled micronized crystalline hu-man insulin using a dry powder inhaler: dose-response andtime-action profiles. Diabet Med 2004;21:763–768.

15. Skyler JS, Weinstock RS, Raskin P, Yale JF, Barrett E, Gerich JE,Gerstein HC: Use of inhaled insulin in a basal=bolus insulinregimen in type 1 diabetic subjects: a 6-month, randomized,comparative trial. Diabetes Care 2005;28:1630–1635.

16. Weiss SR, Cheng SL, Kourides IA, Gelfand RA, LandschulzWH: Inhaled insulin provides improved glycemic control inpatients with type 2 diabetes mellitus inadequately con-trolled with oral agents: a randomized controlled trial. ArchIntern Med 2003;163:2277–2282.

17. Dunbar C, Scheuch G, Sommerer K, DeLong M, Verma A,Batycky R: In vitro and in vivo dose delivery characteristicsof large porous particles for inhalation. Int J Pharm2002;245:179–189.

18. Edwards DA, Hanes J, Caponetti G, Hrkach J, Ben-Jebria A,Eskew ML, Mintzes J, Deaver D, Lotan N, Langer R: Largeporous particles for pulmonary drug delivery. Science 1997;276:1868–1871.

S-24 COMULADA ET AL.

19. Valente AX, Langer R, Stone HA, Edwards DA: Recent ad-vances in the development of an inhaled insulin product.BioDrugs 2003;17:9–17.

20. Vanbever R, Mintzes JD, Wang J, Nice J, Chen D, Batycky R,Langer R, Edwards DA: Formulation and physical charac-terization of large porous particles for inhalation. Pharm Res1999;16:1735–1742.

21. Rave KM, Nosek L, de la Pena A, Seger M, Ernest CS,Heinemann L, Batycky RP, Muchmore DB: Dose response ofinhaled dry-powder insulin and dose equivalence to sub-cutaneous insulin lispro. Diabetes Care 2005;28:2400–2405.

22. Garg S, Rosenstock J, Silverman BL, Sun B, Konkoy CS, de laPena A, Muchmore DB: Efficacy and safety of preprandialhuman insulin inhalation powder versus injectable insulin inpatients with type 1 diabetes. Diabetologia 2006;49:891–899.

23. American Thoracic Society: Single-breath carbon monoxidediffusing capacity (transfer factor). Recommendations for astandard technique—1995 update. Am J Respir Crit CareMed 1995;152:2185–2198.

24. American Thoracic Society: Standardization of spirometry—1994 update. Am J Respir Crit Care Med 1995;152:1107–1136.

25. AACE Diabetes Mellitus Clinical Practice Guidelines TaskForce: American Association of Clinical Endocrinologistsmedical guidelines for clinical practice for the managementof diabetes mellitus. Endocr Pract 2007;13(Suppl 1):3–68.

26. D’Orazio PD, Burnett RW, Fogh-Andersen N, Jacobs E,Kuwa K, Kulpmann WR, Larsson L, Lewenstam A, MaasAHJ, Mager G, Naskalski JW, Okorodudu AO: ApprovedIFCC recommendation on reporting results for blood glu-cose. Clin Chem Lab Med 2006;44:1486–1490.

27. Mathieu C, Cuddihy R, Arakaki R, Belin RM, Planquois J-M,Lyons JN, Heilmann CR: A comparison between simplifiedand intensive dose-titration algorithms using AIR inhaledinsulin for insulin-naive patients with type 2 diabetes in arandomized noninferiority trial. Diabetes Technol Ther 2009;11(Suppl 2):S-53–S-61.

28. Garg SK, Mathieu C, Rais N, Gao H, Tobian JA, Gates JR,Ferguson JA, Webb DM, Berclaz P-Y: Two-year efficacy andsafety of AIR inhaled insulin in patients with type 1 diabetes:an open-label randomized controlled trial. Diabetes TechnolTher 2009;11(Suppl 2):S-5–S-16.

29. Fineberg NS, Fineberg SE, Anderson JH, Birkett MA, GibsonRG, Hufferd S: Immunologic effects of insulin lispro [Lys(B28), Pro (B29) human insulin] in IDDM and NIDDMpatients previously treated with insulin. Diabetes 2008;45:1750–1754.

30. Heise T, Bott S, Tusek C, Stephan JA, Kawabata T, Finco-Kent D, Liu C, Krasner A: The effect of insulin antibodies onthe metabolic action of inhaled and subcutaneous insulin: a

prospective randomized pharmacodynamic study. DiabetesCare 2005;28:2161–2169.

31. Moxness M, Foey J, Finco-Kent D, Bedian V, Krasner A,Kawabata T: Development and validation of radioligandbinding assays to measure total, IgA, IgE, IgG, and IgMinsulin antibodies in human serum. Ann N Y Acad Sci2003;1005:265–268.

32. Piaggio G, Elbourne DR, Altman DG, Pocock SJ, Evans SJW:Reporting of noninferiority and equivalence randomizedtrials: an extension of the CONSORT statement. JAMA 2006;295:1152–1160.

33. Monnier L, Lapinski H, Colette C: Contributions of fastingand postprandial plasma glucose increments to the overalldiurnal hyperglycemia of type 2 diabetic patients: variationswith increasing levels of HbA1c. Diabetes Care 2003;26:881–885.

34. Skyler JS, Jovanovic L, Klioze S, Reis J, Duggan W: Two-yearsafety and efficacy of inhaled human insulin (Exubera) inadult patients with type 1 diabetes. Diabetes Care 2007;30:579–585.

35. Skyler JS, Hollander PA, Jovanovic L, Klioze S, Krasner A,Riese RJ, Reis J, Schwartz P, Duggan W: Safety and efficacyof inhaled human insulin (Exubera�) during discontinuationand readministration of therapy in adults with type 1 dia-betes: a 3-year randomized controlled trial. Diabetes ResClin Pract 2008;82:238–246.

36. Knudson RJ, Kaltenborn WT, Knudson DE, Burrows B: Thesingle-breath carbon monoxide diffusing capacity: referenceequations derived from a healthy nonsmoking populationand effects of hematocrit. Am Rev Respir Dis 1987;135:805–811.

37. Fineberg SE, Kawabata T, Finco-Kent D, Liu C, Krasner A:Antibody response to inhaled insulin in patients with type 1or type 2 diabetes. An analysis of initial phase II and IIIinhaled insulin (Exubera) trials and a two-year extensiontrial. J Clin Endocrinol Metab 2005;90:3287–3294.

38. Hollander PA, Krasner A, Klioze S, Schwartz P, Duggan W:Body weight changes associated with insulin therapy: aretrospective pooled analysis of inhaled human insulin(Exubera) versus subcutaneous insulin in five controlledphase III trials. Diabetologia 2007;30:2508–2510.

Address correspondence to:Zvonko Milicevic, M.D., Ph.D.

Eli Lilly RegionalKolblgasse 8–10

1030 Vienna, Austria

E-mail: milicevic_zvonko@lilly.com

AIR INHALED INSULIN IN TYPE 1 DIABETES S-25

This article has been cited by:

1. Marshall Grant, Andrea Leone-Bay. 2012. Peptide therapeutics: it’s all in the delivery. Therapeutic Delivery 3:8, 981-996.[CrossRef]

2. R.C.L. PageInsulin, other hypoglycemic drugs, and glucagon 34, 685-702. [CrossRef]