Evaluation of Insulin Pump Infusion Sites in Type 1 Diabetes: The DERMIS Study Free

Insulin pump (continuous subcutaneous insulin infusion [CSII]) therapy has been used for insulin delivery for >40 years (1). Despite vast improvements in the insulin and pumps, there has been little change in the infusion sets used to deliver the insulin, which require placement in the subcutaneous tissue and rotation to a different cutaneous site every 3 days. Infusion sets for traditional pumps use either Teflon catheters (which can be inserted at different angles) or steel needles. Insulin infusion sets have been called the Achilles heel of insulin pump therapy because of local skin reactions, infections, and variability in insulin absorption over time (2). A survey of 631 patients using insulin pumps and 76 caretakers of such patients found that 41.4% reported one or more infusion site failures per month (3). Although diabetic ketoacidosis was not reported in this survey study, it is a concerning consequence of insulin occlusion presumed to result from kinking of the infusion set. It has been reported that 55% of diabetic ketoacidosis cases in patients using pumps are due to pump/tubing malfunctions (4). Anecdotally, pump discontinuation is reported after 20 years because of site failures, but just how frequently this occurs is not known.

Little is understood about the cutaneous changes from chronic insulin infusion and how this may affect infusion site failure. Lipohypertrophy and lipoatrophy are well-described complications of insulin therapy that can affect insulin absorption. These two skin-related complications have decreased with improved insulin and with strict insulin injection site rotation (5). Irritant reactions from tape and allergic contact allergies have also been documented in patients using insulin pumps and glucose sensors. Implicated allergens are adhesives and acrylates used in manufacturing the pump and infusion set, which are found in plastics used for construction or at glued junctions between the infusion set and the catheter (6,7). To date, we have been largely dependent on animal models to better understand the skin changes from short-term insulin infusion therapy (8,9).

Optical coherence tomography (OCT) is a well-established noninvasive imaging modality that has several desirable attributes, including high-resolution surface and subsurface images and real-time information (10). OCT angiography (OCTA) is an extension of OCT that uses intrinsic blood flow movement to generate histologic-like angiographic images beneath the tissue surface without the need to inject contrast agent (10). Acute inflammatory infiltrates and edema lead to a decrease in light scattering in tissue, represented by reduced optical attenuation coefficient (OAC) (11,12), allowing OAC to serve as a surrogate method for assessing skin inflammation.

We performed the Defining, Reviewing, and Monitoring Skin Pathology in Type 1 Diabetes Study (DERMIS) to better understand the skin changes from chronic CSII use. In our investigation, we combined both noninvasive OCT and skin biopsies to investigate changes in the epidermal, dermal, and subcutaneous tissues.

For this cross-sectional study, participants were recruited from the University of Washington Diabetes Institute. Inclusion criteria included individuals with type 1 diabetes using CSII sets with vertically inserted cannulas for >20 or <10 years. The protocol design attempted to compare long- and short-term users, defined as those using the pump <10 years. Exclusion criteria included BMI ≥35 kg/m², use of any topical steroid or antifungal medication, insulin dose <0.3 units/kg per day or >1.0 units/kg per day, any active inflammatory skin disease, use of glycemic agents other than insulin, or suboptimal diabetes management (HbA_1c ≥9%). In addition to clinical data, blood was collected for insulin antibodies and HbA_1c measurement. Race, ethnicity, and sex were self-reported by study participants. As a novel study including biopsies, the sample size was based on accessibility in our clinic population.

The University of Washington Human Subjects Division reviewed and approved the study. The study followed procedures in accordance with the ethical standards of University of Washington institutional review boards. Because of the COVID-19 pandemic, minor modifications to the protocol became necessary, including remote visits for consenting and screening.

An endocrinologist and a dermatologist reviewed and selected, together with the participant, the sites to be used for subsequent OCT/OCTA images and biopsies. In this selection, the clinician ensured that the chosen areas did not present clinical lipohypertrophy. Three sites were selected for each participant: the current site, which was in active use with a CSII infusion set at the time of the procedures; the recovery site, which was used for CSII infusion 3–5 days before the procedures; and the control site located 15 cm inferior to the axilla on the flank, which was never used for CSII infusion. A schematic representation of study visits and procedures is shown in Supplementary Fig. 1.

The selected skin sites were imaged using OCT/OCTA immediately before skin biopsy. OCT/OCTA images were acquired using a custom swept-source OCT system operating at a 1,310-nm central wavelength with a bandwidth of 100 nm. The OCT system along with the handheld probe was specially designed for skin imaging (13). The system delivers a spatial resolution of 7 μm axially and 25 μm transversally. Before imaging, the skin was cleaned with alcohol, and then a drop of ultrasound gel was applied to the skin surface to reduce surface hyper-reflective artifacts before each scan. Each skin imaging volume covered a region of ∼8 × 8 × 1.5 mm (∼500 × 500 × 200 pixels) in the x (fast axis), y (slow axis), and z (penetration into the skin) directions, respectively. For each three-dimensional scan, the acquisition time was ∼15 s, and one imaging session took ∼10 min, including preparation, positioning, and OCT imaging. Before quantification of the vascular area density (VAD), acquired three-dimensional OCTA data were first segmented into a slab ∼805 μm thick. Maximum-intensity projection images were produced of the segmented slabs by mapping the maximal value in each A line (z direction) onto a two-dimensional map, termed an en face map. The en face maps were color coded by depth to show a vessel depth ranging from 35 to 840 μm (measured from the skin surface). Then, the en face image was processed into a binary image, and VAD was obtained by calculating the percentage of functional vascular area per image.

After obtaining the OCT/OCTA images, skin sites were photographed and biopsied. Skin punch biopsies (5 mm) were collected from the three imaged sites for each participant. Biopsies were bisected longitudinally, with half fixed in 10% formalin for histologic studies. The other half was flash frozen for additional studies.

Consecutive serial sections (5 mm) were cut from formalin-fixed paraffin-embedded tissue and stained with hematoxylin-eosin. The degrees of fibrosis, fibrin deposition, dermal inflammation, and fat necrosis were scored from 0 (none) to 3 (extensive/severe). Vasculature was scored from 0 (normal) to 3 (increased enlarged central dermal vessels). Eosinophils were scored as the number of cells per mm².

Immunohistochemical staining was performed for the following targets: IGF-I, IGF-I receptor, and transforming growth factor-β3 (TGF-β3; Abcam, Cambridge, MA). IGF-I staining was scored on a scale of 0–5 (14). The Allred method (range of 0–8) was used for IGF-I receptor, and TGF-β3 was scored using a scale of 0 (no staining), 1 (mild superficial perivascular and sparse interstitial fibroblast staining), 2 (moderately increased perivascular and interstitial fibroblast staining), or 3 (strong perivascular and interstitial fibroblast staining). All histopathology and immunohistochemistry were scored by a dermatopathologist who was blinded to the study participant and the site.

The data sets generated during and/or analyzed in the current study are available upon reasonable request from the corresponding author with permission of Helmsley. One of Helmsley’s goals in funding grants is to have all data and biosamples generated from funded research be made available in the manner most conducive to furthering scientific research. Requests will be considered on a case-by-case basis, and requestors will be asked to complete a data sharing agreement with the sponsor before data transfer. Study records will be stored for 25 years after the completion of the study before being destroyed.

Thirty participants were enrolled and included in analysis. Demographics and characteristics of the cohort are shown in Table 1. The mean age, duration of diabetes, and length of time using CSII (mean ± SD) were 48.3 ± 17.1, 30.4 ± 16.0, and 15.8 ± 11.9 years, respectively (Table 1). Skin symptoms at CSII sites were common, with 93.3% of participants reporting itchiness and 76.7% reporting skin redness. Visible skin changes (Supplementary Fig. 2) ranged from pinpoint redness at the catheter site to macular redness without scale that extended up to 1 cm around the insertion site. Only one participant had irritant or eczema-like skin changes. All 30 participants had skin biopsies from the selected sites, for a total of 89 skin biopsies. One recently used site was excluded when the participant did not mark the skin site after removing the infusion set, and we were unable to reliably identify the site for biopsy.

Twenty-five participants, with a total of 75 skin sites, were included in OCT imaging analysis (we excluded five participants because of poor imaging quality caused by motion artifacts) (Table 2). Figure 1 shows representative results of a 35-year-old woman. The photographs (Fig. 1A, E, and I) demonstrate a control site and two CSII sites with infusion set insertion point at the scan center. Scan sites were marked with black lines to ensure image acquisition and skin biopsy centered on the insertion site. The OCTA microvascular maps (Fig. 1B, F, and J) demonstrate increased blood vessels in both current and recovery CSII sites compared with the control site. The light scattering OAC maps (Fig. 1C, G, and K) demonstrate that light attenuation at the CSII sites was lower than that at the control site, and light attenuation at the current site was lower further compared with that at the recovery site. Cross-sectional B-frame images (Fig. 1D, H, and L) show skin structures with CSII insertion sites indicated by red arrows. Statistical analysis (Fig. 1M and N) shows a significant difference in VAD and OAC between CSII and control sites and between recovery and current CSII sites.

Histologic analysis (Table 2 and Supplementary Fig. 3) demonstrated that both current and recovery sites showed increased fibrosis, fibrin, inflammation, fat necrosis, vascularity, and eosinophils when compared with control sites. There were no significant differences seen between current and recovery sites.

There were no eosinophils observed in skin biopsies at control sites. Eosinophils were identified in 73% of skin biopsies from current sites (P < 0.01) and 75% of skin biopsies from recovery sites (P < 0.01). All study participants had eosinophils identified in at least one insulin infusion site, current and/or recovery. The eosinophils were located deep in the dermis near the interface with fat (Fig. 2). The number of eosinophils at pump sites ranged from zero to 31 per high-power field, with a median of four per high-power field. There was no significant association between type of insulin (insulin aspart and insulin lispro) or pump brand and number of eosinophils (Supplementary Fig. 4A). Higher eosinophil counts were seen in participants using pumps for <10 years compared with those using pumps >20 years (P = 0.02) (Supplementary Fig. 4B). Immunohistochemical staining also showed differences between current/recovery sites and the control site for IGF-I and TGF-β3 but no difference for IGF-I receptor.

All participants used the Dexcom G6 as part of usual care. Inflammation score was positively correlated with insulin dose (r = 0.48; P = 0.009) and negatively correlated with time in range on continuous glucose monitoring (r = −0.51; P = 0.01) (Supplementary Fig. 5).

There were no differences in either OCT or histologic findings based on duration of CSII use, previous use of animal insulin, or type of insulin.

Using both noninvasive OCT imaging and histology of skin punch biopsies, we present a first assessment of CSII skin sites in long-term insulin pump users. We demonstrate changes in vessel density and greater fibrosis, fibrinogen, inflammation (including eosinophils), IGF-I, TGF-β3, and fat necrosis at pump sites compared with control sites.

OCT allows for high-resolution cross-sectional tomographic imaging of internal microstructure in materials and biological systems by measuring scattered or reflected light (10,15). Although OCT is now used in diverse applications, it is most frequently used medically for retinal imaging (16). The use of OCT for the skin was first reported in 1997 for bullous skin disease and skin tumors (15).

Using OCT, we found increased vessel density (as measured by VAD) at pump sites when compared with control skin. This greater vessel density is likely an effect of injury and repair related to catheter insertion and could explain the Tamborlane effect (17), in which the longer an infusion site is inserted, the faster the kinetics of the insulin. Our histologic finding of increased vasculature at pump sites supports this as a possible cause of this real-life alteration in insulin kinetics.

We found a significant decrease in OAC and significant increase in inflammation by histologic analysis at pump sites, together supporting increased inflammation at pump sites compared with control. Although our data could not directly correlate the OAC measure of inflammation and inflammation scores assessed by histology, both methods identified the same significant result. Therefore, we propose that OCT provides a noninvasive means of surrogate assessment of inflammation at insulin pump sites. There are some important differences to note between OCT and histologic assessment. For example, several factors contribute to the light attenuation in tissue measured by OAC, including lymphatic perfusion, collagen remodeling, and cellular motility (11). Because of its noninvasive nature, future study of skin responses to insulin pump therapy should consider a systematic design involving OCT imaging to comprehensively assess tissue changes. Ultimately, using OCT to study infusion sites could provide important information about the skin response to these devices at different time points, including extended-wear devices.

The most surprising histologic finding in our study was the prevalence and degree of eosinophilic infiltration, which was seen in all study participants in at least one infusion site. Although there was a similar percentage of samples with eosinophils in both the current and recovery sites, the peak level of eosinophils was variable among the study participants. Eosinophils are not typically present as a component of resident inflammatory cells in the skin, and no eosinophils were identified in the control skin biopsies. Eosinophils may be occasionally present in normal wound healing; however, the absolute number and density of eosinophil in these samples support a delayed type hypersensitivity response, which is typically observed between 2 and 7 days after exposure to an allergen. Eosinophils are short-lived cells and are common in irritant or allergic reactions and host defense, and they have been shown to be involved in a wide range of fibrotic processes (18). Eosinophils are often correlated with symptoms of itchiness and likely explain the high percentage of participants who reported itchiness in this study (Table 1). The deeply situated nature of eosinophils in the skin biopsies and the lack of eczema-like changes observed on the surface of the skin suggest the allergen is delivered at the tip of the catheter rather than being a surface reaction to, for example, tape or adhesive. Identifying the source of potential allergens in medical devices is complicated by the limited information available from device manufacturers. Leading candidates for potential allergens include one or more of the preservatives of the insulin, materials used in plastics for the infusion set, or glue used in the assembly of the various components. This study was not designed to compare pump brands, and we saw eosinophilic inflammation in participants using a variety of devices, suggesting this is a common reaction pattern. The finding that long-term pump users (>20 years) had significantly fewer eosinophils may indicate that early users (<10 years) stop pump therapy because of allergic reactions or that tolerance develops over time. Although frequent site rotation and changing of infusion sets have been emphasized to reduce skin reactions, if the allergen is leaching out from device materials, extended wearing may lessen the allergic response.

The increase in fat necrosis at pump sites is also of interest. We believe that this fat necrosis is distinct from the lipoatrophy previously described in individuals with diabetes. Lipoatrophy is the loss of subcutaneous tissue at insulin injection sites, which can result in altered insulin absorption (19). Insulin-induced lipoatrophy was initially thought to be due to necrosis or apoptosis of adipocytes, but it is now hypothesized to be due to a reduction in adipocyte size (20). It is likely the fat necrosis we noted at pump sites is part of the wound healing process from the insulin catheter, but the relationship between this fat necrosis and any alteration in insulin absorption is unclear.

IGF-I and TGF-β3 are inflammatory and fibrogenic proteins that play key roles in wound healing (21,22), and our finding that both IGF-I and TGF-β3 were increased at pump sites was not surprising. Both of these proteins may not function normally in diabetes, thus prolonging wound repair processes (23). Eosinophilic inflammation has also been shown to slow wound healing and may be a source of TGF expression (24). TGF-β is also elevated in fibrotic skin conditions like scleroderma and keloid scarring (25), and one could hypothesize that alterations in wound healing and control of fibrosis at pump sites could affect insulin absorption over time.

Limitations of this study include a relatively restricted cohort of adult participants who were mostly White individuals with normal weight and excellent glycemic control. It is possible our findings may not be generalizable to other populations, particularly the pediatric population. Although our data help describe the inflammatory response and other architectural changes seen at infusion sites, we did not do testing to identify the allergens.

In conclusion, we assessed insulin pump sites both noninvasively and with skin biopsy and demonstrate higher vessel density and signals of inflammation by OCT, in addition to increased inflammation, fat necrosis, fibrosis, and eosinophilic infiltration by histopathology. This study significantly adds to growing evidence of allergic reactions to medical devices in diabetes. Additional studies, including patch testing, cooperation and transparency with device manufacturers, and interventions to promote allergen tolerance and determine the best timing of site rotation, would be helpful. We hope that these findings add information toward the development of solutions to minimize the infusion site failures commonly experienced by individuals using insulin pump therapy.

Acknowledgments. The authors thank the Experimental Histopathology services at the Fred Hutchinson Cancer Center for preparing and staining the skin specimens.

Funding. This study was funded by the Leona M. and Harry B. Helmsley Charitable Trust (grant G-192-03584). R.W. reports equipment from Carl Zeiss Meditec, Inc., and support from Washington Research Foundation, both to the institution and outside the submitted work.

The funders had no role in the preparation, review, or approval of the manuscript or the decision to submit the manuscript for publication. The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the US Department of Health and Human Services or other departments or agencies of the federal government.

Duality of Interest. R.W. reports grants from the National Institutes of Health, Carl Zeiss Meditec, Inc., Colgate-Palmolive Company, and Estee Lauder, Inc., outside the submitted work; consulting fees from Cyberdontics, Inc., and Carl Zeiss Meditec, Inc., outside the submitted work; and patents planned, issued, or pending (US8750586, US8180134, US9282905, US9759544, US10354378, and US1052906) outside the submitted work. I.B.H. reports grants and contracts from Insulet, Medtronic, and Dexcom, Inc., outside the submitted work; consulting fees from Abbott Diabetes Care, Lifescan, and Hagar outside the submitted work; and honoraria for lectures, presentations, speaker’s bureaus, manuscript writing, or educational events as section editor for UpToDate outside the submitted work. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. A.K., M.M.S., R.W., D.K., and I.B.H. contributed to the conception and design of the study and wrote the first draft of the manuscript. X.D. served as statistician and participated in the data analysis and interpretation of the results. A.K., M.M.S., J.D.B., D.K., J.L., and I.B.H. were involved in conducting the study. All authors participated in the data analysis and interpretation of the results and edited, reviewed, and approved the final version of the manuscript. I.B.H. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented at the 15th Conference on Advanced Technologies & Treatments for Diabetes, Barcelona, Spain, 27–30 April 2022; at the Society for Investigative Dermatology Annual Meeting, Portland, OR, 18–21 May 2022; and at the 82nd Scientific Sessions of the American Diabetes Association, New Orleans, LA, 3–7 June 2022.