AIR Inhaled Insulin in Subjects With Chronic Obstructive Pulmonary Disease: Pharmacokinetics, glucodynamics, safety, and tolerability

A total of 45 subjects, comprised of both healthy subjects and subjects diagnosed with moderate COPD class II (forced expiratory volume in 1 s [FEV₁] 30 to 79% of predicted value, respectively) (9), were enrolled in this study. Subjects between 18 and 65 years of age, recruited from a number of nearby pulmonology practices, were considered eligible for the screening phase if they presented without a history of diabetes, BMI ≤35 kg/m², fasting blood glucose ≤117 mg/dl (6.5 mmol/l), and the ability to perform spirometry according to American Thoracic Society criteria (10). Subjects with COPD were subcategorized to either a chronic bronchitis group if they had a history of chronic productive cough (productive cough for at least 3 months in each of the preceding 2 years) or to a pulmonary emphysema group if they lacked such cough history. Exclusion criteria included active smoking during the previous 6 months or serum cotinine levels ≥20 ng/ml at the screening visit; history of asthma, recent upper respiratory or lower respiratory tract infection; or acute exacerbation of COPD within the last 2 months. Individuals using β-blockers, thiazides, or systemic corticosteroids within the past 3 months were also excluded.

A local ethics committee approved the protocol. This phase I trial was conducted in agreement with the principles outlined in the Declaration of Helsinki and the International Conference on Harmonization Guideline to Good Clinical Practice. All subjects were given a full explanation of study procedures at the screening visit, and written informed consent was obtained before participation in the study.

This phase I, open-label, randomized, three-way, crossover study was conducted at a single investigative site (Profil Institute for Metabolic Research, Neuss, Germany). Eligible subjects entered a screening phase, which included clinical laboratory tests and an evaluation of a subject's ability to achieve suitable inhalation flow and volume targets (30 L/min for 3–5 s) using the AIR insulin inhaler connected in series with a spirometer.

Qualified participants were randomly assigned to one of three possible treatment sequences, wherein each subject was administered insulin: twice as AIR insulin and once as subcutaneous insulin lispro. AIR insulin was supplied as 6 units equivalent capsules (2.6 mg insulin/capsule) packaged in foil blister cards. The capsule contents were aerosolized for deep lung delivery by means of a hand-held, breath-actuated, mechanical inhaler device. Insulin lispro (Humalog; Eli Lilly, Indianapolis, IN) was supplied in 10-ml vials (100 unit/ml).

Euglycemic clamp procedures were separated by an interval of 5–18 days. At each clamp visit, subjects were administered (∼90 min after the start of the baseline clamp procedure) either 12 units (two inhalations of 6 units equivalent capsule strength) AIR insulin or 12 units subcutaneous insulin injected into the abdominal region, according to the treatment sequence. The dose equivalence of AIR insulin relative to subcutaneous insulin lispro was determined in a previous study in healthy volunteers (11). Blood samples were collected at multiple time points (up to 10 h) for free serum immunoreactive insulin (IRI) determination. Routine physical examinations, vital signs, electrocardiograms, and pulmonary function tests (PFTs) were performed at screening and at the final visit. Observational data regarding adverse reactions were collected throughout the study.

Complete PFTs, which included flow volume spirometry (FEV₁ and forced vital capacity [FVC]), carbon monoxide diffusing capacity (DL_CO), total lung capacity (TLC), and residual volume measurements, were performed before and after (within 48 h) each clamp procedure.

Additionally, bedside spirometry (FEV₁ and FVC) was performed at predose, ∼30 min postdose, halfway through the clamp, and immediately at the end of the clamp with the subject in a semirecumbent position. The protocol specified that any subject who experienced ≥20% decline in FEV₁ during the clamp, following AIR insulin exposure (confirmed by laboratory PFT), would be excluded from further exposure to inhaled insulin.

Following administration of AIR insulin and subcutaneous insulin lispro, IRI concentrations were measured using a conventional competitive radioimmunoassay method validated over the range of 20–2,500 pmol/l, with 100% cross-reactivity for regular human insulin and insulin lispro (MDS Pharma Services, St. Laurent, Quebec City, Canada).

WinNonlin software (professional edition 4.1; Pharsight, Mountain View, CA) was used to compute noncompartmental pharmacokinetic parameters, including maximum IRI concentration (C_max), time of maximum IRI concentration, and area under the serum insulin concentration curve from time 0 until time of return to baseline (AUC_0–t′). The relative bioavailability (F) of AIR insulin to that of subcutaneous insulin lispro was calculated according to the equation

where D is the dose of insulin (in mg) for each route of administration (11).

Glucodynamic parameters were derived from the glucose clamp data. Glucose infusion rate (GIR) data were individually fitted using a LOESS (locally weighted linear regression) smoothing function (12). Parameters determined from the individual LOESS prediction included the maximum GIR (R_max) and the time of maximum GIR (t_Rmax). The total amount of glucose infused from time 0 to 10 h (G_tot) was calculated from the observed GIR. The biopotency for AIR insulin relative to subcutaneous insulin lispro (F′, analogous to bioavailability) was individually assessed, based on the glucodynamic parameter GIR_tot, as previously described (11).

All statistical analyses were performed using SAS (version 8.2; SAS Institute, Cary, NC). Data from all subjects with stage IIA and IIB severity were pooled in the chronic bronchitis and emphysema groups, respectively, for all analyses. The pharmacokinetic parameters (AUC_0–t′ and C_max) were log transformed. A linear mixed-effect model was applied using the subject group (healthy, chronic bronchitis, and emphysema), visit, and treatment sequence as fixed effects and subject-nested within-group sequence as a random effect. The log-transformed results were transformed to the original scale by exponentiation to obtain geometric least-squares means and treatment ratios. Glucodynamic parameters were analyzed similarly.

The effect of PFT parameters on pharmacokinetic and glucodynamic parameters was explored using a linear mixed regression model that included all available observations. The model was fitted for each PFT parameter with treatment, group, pharmacokinetic or glucodynamic parameter, and interactions between the terms, if deemed appropriate. Parameters were log transformed before the analysis. The model considered random variations due to subject.

PFT variables (FEV₁, FVC, TLC, residual volume, and DL_CO) and their change from baseline and at visits were analyzed to detect group and treatment differences. Each pre- and postclamp observation was analyzed separately. A linear mixed-effects model was fitted with group, treatment, group by treatment, visit, baseline, and sequence as fixed effects and subject (group sequence) as a random effect. Group and treatment differences in pulmonary function response variables were examined.

Of the 45 subjects who were assigned to treatment and received at least one dose of the study drugs, 42 completed the study. One subject in the emphysema group discontinued the study because of an adverse event (malaise, nausea, and dry and productive cough). Two other subjects (both from the chronic bronchitis group) were discontinued for safety reasons after experiencing >20% decrease in FEV₁ following AIR insulin administration. Neither of the subjects experienced symptoms associated with decline in pulmonary function.

At baseline, mean body weight was 82.8 ± 18.8, 88.9 ± 19.4, and 75.8 ± 14.1 kg for healthy subjects, those with chronic bronchitis, and those with emphysema, respectively. The mean age of healthy subjects was 38.3 ± 12.7 years, a lower mean age than that for subjects with chronic bronchitis and emphysema (53.4 ± 9.3 and 57.5 ± 5.6 years, respectively). Baseline mean FEV₁ was 4.06 ± 1.04 L/s in the healthy group, 2.14 ± 0.60 L/s in the chronic bronchitis group, and 1.67 ± 0.61 L/s in the emphysema group. Neither pulmonary nor demographic characteristics differed significantly between both COPD groups.

Subcutaneous insulin lispro administration resulted in time-exposure profiles (Fig. 1A) and total insulin exposures (Table 1) that were indistinguishable among groups. Following the inhalation of AIR insulin, subjects with chronic bronchitis demonstrated significantly reduced maximal insulin levels (C_max), 59.3% of those found in healthy subjects (P = 0.002), with a total exposure (AUC_0–t′) of 55.7% of that of healthy subjects (P = 0.001) (Fig. 1B). Subjects with pulmonary emphysema had reduced maximal insulin levels (C_max) and reduced total exposure (AUC_0–t′) (respectively, 75.8 and 78.5% of the levels in healthy subjects; not significantly different at α = 0.05). AIR insulin bioavailability compared with subcutaneous insulin lispro was 6.2% in healthy subjects, 3.4% in subjects with bronchitis, and 4.9% in subjects with pulmonary emphysema. Intrasubject coefficient of variation (CV) for AUC_0–t′ ranged from 28.0% (subjects with emphysema and healthy subjects) to 52% (subjects with chronic bronchitis).

Glucodynamic response to AIR insulin administration in subjects with COPD was 35% lower than that in healthy subjects, whereas all three groups had comparable glucodynamic responses to subcutaneous insulin lispro (Fig. 1C and D). Subjects with chronic bronchitis exhibited a total insulin effect (GIR_tot) that was 60.4% of that seen in healthy subjects (P < 0.01). Similarly, subjects with pulmonary emphysema exhibited a total insulin effect of 67.1% of that seen in healthy subjects (P < 0.01). Healthy subjects had a higher relative biopotency (Table 1) of 9.3% compared with subjects with COPD (chronic bronchitis 5.2% and pulmonary emphysema 6.3%). Intrasubject CV estimates ranged from 16.8 to 43.0% for the different glucodynamic parameters (Table 1).

A statistically significant relationship was found between DL_CO and pharmacokinetic responses to AIR insulin, including AUC_0–t′ (Fig. 2A) and C_max (P < 0.01). However, the correlation coefficient associated with this relationship was low (r = 0.00148). No other significant relationships were observed between any of the pharmacokinetic or glucodynamic variables and PFT parameters (P > 0.2) using the described models (Fig. 2 B–D).

Review of the laboratory, vital signs, and electrocardiogram data found no safety issues of clinical relevance (data not shown). Pulmonary function tests did not generally suggest any safety concerns, although bedside spirometry showed clinically minor but consistent declines (preclamp to end clamp) in subjects with COPD that were statistically significant (P < 0.001), regardless of the treatment administered. In subjects with chronic bronchitis, bedside FEV₁ declined −0.24 L with AIR insulin administration and −0.33 L with subcutaneous insulin lispro. In subjects with pulmonary emphysema, FEV₁ reductions were −0.24 L and −0.16 L, respectively. At the end of the clamp, no significant between-treatment differences in FEV₁ were observed. No clinically relevant changes in FEV₁ were noted in healthy subjects.

Following AIR insulin administration, 27 subjects reported a total of 15 adverse events considered by the investigator to be related to the study drug. Events were mild to moderate in intensity and included headache, cough, abdominal pain, rhinitis, and palmar erythema. One subject in the pulmonary emphysema group discontinued the study because of adverse events (malaise, cough, and nausea). Two subjects in the chronic bronchitis group were discontinued from the study by investigator decision. According to protocol-specified procedures, subjects who experienced a >20% decline in FEV₁ (following AIR insulin dosing) were to be discontinued from further exposure if this decline was confirmed at the next scheduled laboratory PFT. One subject experienced a 31% decline in FEV₁ (with no pulmonary symptoms) during an AIR insulin clamp, but follow-up PFT only showed an 18% decline compared with the preclamp laboratory-based PFT. Another subject experienced a 38% decline in FEV₁ (also without symptoms) during an AIR insulin clamp. However, it was not possible to accurately assess this subject's pre- to postclamp FEV₁ change, since the preclamp PFT was given an “unacceptable” quality score of “F.” Although neither subject completely fulfilled protocol-defined criteria for study discontinuation, both were withdrawn by investigator decision based on these PFT results.

This study is the first published clinical trial to report the impact of COPD on pharmacokinetic and glucodynamic responses to inhaled insulin versus subcutaneous insulin. Based on its panoply of physiologic and anatomic abnormalities (13,14), COPD might be expected to affect the size of the absorptive surface available and/or the distribution and deposition of inhaled particles in the lung, potentially reducing the amount of insulin available for systemic absorption. The results in this study support this hypothesis in that although pharmacokinetic and glucodynamic responses to subcutaneous insulin lispro were comparable in healthy subjects and in COPD, the administration of AIR insulin resulted in a significant reduction in insulin exposure and metabolic effects in subjects with emphysema and chronic bronchitis compared with healthy subjects. Similar bioavailability and biopotency values for AIR insulin (relative to insulin lispro) in healthy subjects have been previously reported (11) and have been corroborated by clinical experience in patients with diabetes (15). Further, the intrasubject variability values that were observed for AIR insulin—an important consideration in the context of an insulin's clinically reproducible metabolic effect (16,17)—are compatible with the potential use of this product in the treatment of diabetes.

In contrast with our results, studies in active smokers have shown that the absorption of pulmonary insulin of various formulations is actually increased (18–21), although this is partially reversible upon smoking cessation. Since the subjects in this study were only included if they had ceased smoking for at least 6 months, acute increased pharmacokinetic and glucodynamic responses would have been expected to abate. Thus, we hypothesize that the reduced pharmacokinetic and glucodynamic responses observed in our study are likely attributable to lung tissue damage associated with COPD.

The significant relationship observed between DL_CO and pharmacokinetic parameters could provide a partial explanation for the reduced bioavailability of inhaled insulin observed in the COPD groups. DL_CO is an index of a key aspect of lung function and may be altered by a variety of pathophysiologic and anatomic processes, including effective diffusion barrier thickness, pulmonary capillary blood flow, and size of the total absorptive surface available. COPD may affect any of these parameters. Our results are the first demonstration of a relationship between a pulmonary function variable and pharmacokinetic parameters for a systemically delivered inhaled drug. Although there is certainly a significant difference in FEV₁ across the study groups and a comparable difference is seen between study groups in both pharmacokinetic and glucodynamic responses, the lack of a significant relationship between FEV₁ and the pharmacokinetic/glucodynamic parameters suggests that the mechanisms responsible for airflow obstruction are not responsible for these observed pharmacokinetic and glucodynamic differences. In other words, breathing mechanics seem to be less important than factors related to gas transfer mechanisms in modulating the absorption of inhaled insulin among patients with COPD.

No serious adverse events and no clinically significant changes in safety-related laboratory parameters, vital signs, or electrocardiograms were reported in this study. Adverse events reported as possibly related to AIR insulin were mild or moderate in intensity and consistent with previous AIR insulin studies in nonsmoking, healthy subjects or insulin studies involving glucose clamp procedures (11,20). Inhaled insulin formulations have also demonstrated safety profiles similar to the comparator regimens (22–24).

In our study, a decline in bedside spirometric pulmonary function was demonstrated in subjects with COPD during the prolonged bed rest of the clamp procedure, following both AIR insulin and subcutaneous insulin lispro treatments. However, there was no difference between laboratory PFTs performed on the days before and after clamps. Healthy subjects did not demonstrate a decline in FEV₁ during clamps with either insulin treatment. It thus appears that the decline in FEV₁ during the clamps in most of the COPD subjects was probably not therapy related but rather a function of the disease state and study procedures. However, two subjects with chronic bronchitis were withdrawn from further inhaled insulin exposure after experiencing >20% reductions in FEV₁ following AIR insulin administration. Neither of these cases definitely fulfilled protocol-based PFT criteria for discontinuation from study, and neither subject had any symptoms associated with the acute declines in FEV₁. PFTs are intrinsically more variable in subjects with COPD compared with healthy subjects, making it more difficult to identify changes in PFTs in the subject population. Given these constraints, the two subjects were discontinued from further study based on investigator decision.

Single doses of AIR insulin (12 units equivalent or 5.2 mg) were well tolerated in both healthy subjects and in subjects with moderate COPD. AIR insulin was absorbed as rapidly as subcutaneous insulin lispro irrespective of the disease state. For healthy subjects, the bioavailability and biopotency of AIR insulin relative to subcutaneous insulin lispro, as well as the overall pharmacokinetic and PD profiles, appear to be consistent with a previous study (11). This profile has been shown to be suitable for treatment of patients with diabetes (15).

In summary, this clinical experimental study suggests that AIR insulin bioavailability and biopotency are reduced in subjects with moderate COPD, relative to healthy subjects. Because of the reduced pulmonary absorption, an individual adaptation of insulin doses (as usual in diabetes care) is essential. In the case of patients with chronic bronchitis, the observation of increased intrasubject variability may also be clinically important because this may lead to greater fluctuations in glycemic control. Overall, AIR insulin was generally well tolerated. The results of this study do not, however, provide a basis for dosing guidance of AIR insulin in patients with COPD, since longer-term experience will be required to more fully characterize both the effects of COPD on inhaled insulin dosing and the effects of inhaled insulin on the COPD disease process. Such experience will provide the basis for recommendations regarding inhaled insulin dosing and pulmonary safety monitoring that may be specific to patients with COPD. The results do, however, support the conduct of long-term studies with AIR insulin in patients with COPD and diabetes to evaluate these questions. Until such data are available, it would be advisable for patients with concomitant diabetes and COPD to use conventional approaches to diabetes management.

This work was sponsored by Eli Lilly.

We thank Hollins Showalter for statistical analyses and Jeffrey R. Gates for manuscript preparation. Data in this manuscript are from study H7U-MC-IDAN. The following abstract by Rave et al. has been previously published: Diabetes 55 (Suppl 1):A26, 2006