Advances in Insulin Infusion Set in the New Era of Automated Insulin Delivery: A Systematic Review

Abstract

Introduction:

Automated insulin delivery (AID) has become a well-known research topic devoted to achieving better glycemic outcomes. AID systems consist primarily of three components: the continuous glucose monitoring system, the insulin delivery system, either tethered or patch pump, and the control system (algorithm). A key component in the tethered pump AID system is the insulin infusion set (IIS). This Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) study was conducted to evaluate the IIS evolution in the era of AID and to provide future perspectives of IIS clinical use.

Methods:

Literature searches for articles published from January 2016 to July 2022 were performed in Embase/Medline and PubMed. Data were extracted following PRISMA guidelines. Primary meta-analysis outcomes were IIS wear duration, total daily dose of insulin, and IIS failure reasons/modes.

Results:

We identified 387 publications, of which 15 eligible studies compared various IISs comprising over 1400 participants and >53 000 wears. Half of the studies published in 2022 were focused on extended IISs designed for wear durations of seven days or more. Three clinical trials have demonstrated the safe use of extended IISs to seven days of wear in individuals with type 1 diabetes, and two also demonstrated good glycemic control throughout the seven-day use.

Conclusions:

Research in insulin infusion technology has increased in the last six years, and extended IISs have demonstrated improved overall performance, particularly in duration of wear. Paths for future products are discussed with an emphasis on understanding the existing barriers related to both technical and nontechnical issues.

Introduction

Type 1 diabetes (T1D) is an autoimmune condition affecting pancreatic beta cells, leading to insulin deficiency and hyperglycemia. A person with T1D is often required to make complex dosing decisions to maintain glucose levels in the desired range while avoiding hypoglycemia. The potential consequences of suboptimal glycemic control are multisystemic, leading to substantial morbidity and increased mortality. ¹ Multiple factors influence glucose levels and insulin sensitivity, causing considerable day-to-day variability of glucose levels as well as insulin requirements.^2,3 Automated insulin delivery (AID) mitigates some of these issues resulting in improved glycemic outcomes, while also decreasing the disease burden.

The concept of closed-loop glucose control has been present since the 1960s. ⁴ The availability of advanced insulin pumps, the emergence of accurate and reliable continuous glucose monitoring (CGM) systems, the access to secure and safe wireless communication technologies, and the development of safe control algorithms enabled the development and approval of closed-loop systems. The commercialization of the first hybrid closed-loop system (MiniMed 670G insulin pump system; Medtronic, Northridge, CA) in 2016 for use by people living with T1D reflects these advancements, heralding the new era of diabetes management. Since then, the Cambridge CamAPS FX system (Camdiab Ltd., Cambridge, UK) was approved in United Kingdom and Tandem/Dexcom Control-IQ system (Tandem Diabetes Care, Inc., San Diego, CA, USA) was approved by Food and Drug Administration in 2019.

Publications on technology advancements in the AID system since January 2016 to July 2022 were reviewed. There have been many systematic and meta-analysis reviews focused on the CGM sensors and insulin pumps.^{5 -10} Insulin infusion sets (IISs), an integral part of several key commercial AID systems, have been understudied and require greater attention (Figure 1), as IISs are regarded the “weakest link” in tethered pump insulin delivery. ¹¹ Based on the insertion angle and cannula (the small tube inserted in the subcutaneous tissue) material, regular IISs currently on the market may be categorized into 3 types: 90° Teflon, angled Teflon, and 90° steel, which are approved for wear durations up to three days, three days, and two days, respectively. One extended IIS (the Medtronic Extended with 90° Teflon) has received regulatory clearance for wear for up to seven days in the United States and Europe.

Figure 1.

The proportion of articles returned from the Embase/Medline (left) and PubMed (right) database searches (2016-2022) related to continuous glucose monitoring, insulin pump, and infusion set technologies.

In this review, we focus on all the IIS-centered clinical studies and conducted a systematic review, quantitative meta-analysis, and narrative synthesis to evaluate state-of-the-art insulin infusion technologies and to provide future perspectives of IIS clinical use in closed-loop systems.

Materials and Methods

Data Sources and Searches

The search using Embase/Medline and PubMed focused on publications from January 2016 to July 2022. We used a simple search term (“infusion set”) and searched titles and abstracts to generate a more comprehensive library and mitigate potential misses.

Study Selection

Title, abstract, and then full text of relevant studies were evaluated independently for inclusion and exclusion criteria by two authors (E.R.-A. and G.Z.), respectively. Any discrepancy was resolved by discussion. Results were screened according to Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines to remove duplicates, and titles/abstracts were examined to exclude reviews, case studies, symposia, retrospective analyses, animal studies, surveys, and letters. The remaining articles were subjected to full-text reviews to include only IIS clinical studies. All study designs were included in the review, and no demographic or geographic limitations were applied to the analysis.

Data Extraction

Two authors (E.R.-A. and T.K.) independently examined the included articles, reviewing the main reports and supplementary materials and extracting relevant summary estimates. Studies were included based on the availability of information for relevant IIS characteristics (ie, cannula length and angle) and key performance indicators including mean wear duration, total daily dose of insulin (TDD), and IIS failure mode/reason analysis. Failure mode/reason analyses included multiple categories that were nonexclusive, that is, unexplained hyperglycemia, suspected and confirmed IIS occlusion, site reactions (pain, erythema, edema), set failures (pullout, leakage, adhesive failures, etc), infection, and others/unknown (reservoir out of insulin, inaccurate time calculation, personal preference/cosmetic reasons, etc). Data were expressed in mean ± standard deviation. Statistical analysis was performed with Minitab Version 20.1.3 software (Minitab, LLC, State College, PA, USA).

Results

Study Selection

The PRISMA-based study selection process is shown in Figure 2 including justification for exclusion. A total of 387 related articles were identified during the initial search. After title and abstract screening, 20 articles met eligibility criteria and were assessed for full-text eligibility. Fourteen articles,^{12 -25} containing 15 studies with over 1400 participants and >53 000 wears, were included in this analysis: One of the articles contained data from two studies.

Figure 2.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram.

The characteristics of the included studies are shown in Table 1. Five of the 15 studies tested the IISs for seven days or more or until device failure. Overall, the mean age of participants in 12 of the 15 studies was in the range of 39 to 49 years,^{12,14 -24} with the mean age in three studies being below 30 years.^13,25 Many of these IIS studies were conducted with sensor-augmented pump (SAP) therapy, with a few studies conducted in an AID system. Therefore, safety and effectiveness of each IIS was assessed and controlled by the subject based on sensor glucose values using the SAP or AID system and confirmed by self-monitored blood glucose (SMBG) values. The mean baseline HbA1C in all studies was in the range of 7.0% to 7.6%, as is typical in most diabetes technology studies involving insulin pumps.^2,3

Table 1.

Characteristics of the Included Studies.

Publication date	Type of study	Pump system	IIS	Approved/planned IIS wear time	Participants (M/F)	Wears	Mean age (SD), years	Baseline HbA1c (SD), %	Registration at clinicaltrial.gov	Reference
2022 June	Prospective, nonrandomized, home-use feasibility study of IIS	Tandem t:slim X2 (30%), Medtronic 640G (55%), Medtronic 670G (15%); Dexcom G5	Capillary Biomedical Extended Wear (CBX)	7/7 days	20 (10/10)	41	44.1 (14.0)	7.2 (0.7)	NCT04503564	Kastner et al 12
2022 June	Two studies: pivotal trial; randomized trial of pump system	Tandem t:slim X2, Dexcom G6	Tandem all types & unknown	2-3/3 days	263 (132/131)	25 154	25 (17)	7.4 (0.9)	NCT03563313	Kanapka et al 13
2022 March	Single-arm, IIS nonrandomized trial	Medtronic 670G, Guardian Sensor 3	Medtronic Extended	7/7 days	259 (125/134)	3041	45.0 (14.1)	7.2 (0.6)	NCT04113694	Brazg et al 14
2022 January	Post-hoc analysis of Onset 5 data (insulin Faster Aspart study)	Medtronic 530G, Paradigm, Veo, or Revel. Dexcom G4	Medtronic Quick-set, Silhouette, Mio, Sure-T	2-3/3 days	472 (203/269)	19 013	43.5 (14.7)	7.5 (0.5)	NCT02825251	Gorst-Rasmussen et al 15 & Zijlstra et al 23
2021 December	Double-blind, crossover study of URLi & Lispro	Medtronic 670G with GS3 Sensor	Medtronic Mio	3/3 days	42 (15/27)	784	47.8 (13.8)	7.07 (0.47)	NCT03760640	Bode et al 16
2021 June	Open-label, single-center, single-arm, controlled pilot study	Medtronic (10), Animas (10), Tandem (2), Dexcom G6	ConvaTec Extended	7/10 days	22 (12/10)	22	40.1 (14.3)	7.1 (0.8)	NCT03819634	Lal et al 17
2021 January	Phase 3, double-blind, crossover study of URLi & Lispro	Medtronic 530G, 630G, 640G, Dexcom G5	Medtronic Mio	3/3 days	49 (23/26)	1358	39.6 (11.7)	7.06 (0.68)	Not registered	Bode et al 18
2020 October	Interventional, open-label, randomized IIS crossover study	Ypsomed mylife YpsoPump, Dexcom G5	Ypsomed Orbit Micro and Soft	2-3/7 days	40 (16/24)	160	49.2 (12.5)	7.3 (0.7)	NCT03683368	Waldenmaier et al 19
2020 September	Randomized, open-label, 2 × 4-week crossover study of Aspart & Biosimilar	Animas (4) Medtronic (41)	Not reported	2-3/3 days	45 (17/28)	860	43.1 (13.6)	<8.0 (44); ≥8.0 (1)	NCT03436498	Thrasher et al 20
2019 January	Open, mono-center, single-arm study of YpsoPump system	Ypsomed YpsoPump	Ypsomed Orbit Soft	3/3 days	35 (13/22)	406	44.8 (12.3)	Not reported	Not registered DRKS00011838	Waldenmaier et al 21
2018 May	Randomized, open-label, 2 × 4-week crossover study of Lispro & Biosimilar	Animas (5), Medtronic (22)	Not reported	2-3/3 days	27 (8/19)	520	42.2 (14.6)	<8.5%	NCT02603510	Thrasher et al 22
2018 January	Randomized, double-blind, parallel-group study of Faster Aspart & Aspart	Medtronic Paradigm Pump	Medtronic Quick-Set or Silhouette	3/3 days	37 (22/15)	553	44.3 (12.8)	7.4 (0.7)	NCT02825251	Zijlstra et al 23
2017 March	Randomized, cross-over study of two types of IISs	Not reported	Roche FlexLink Plus & FlexLink	3/3 days	80 (47/33)	1971	46.5 (12.9)	7.6 (1.0)	NCT02103595	Freckmann et al 24
2016 July	Multicenter cross-over trial of IIS in LH & non-LH areas	Personal Pump, Dexcom G4 Platinum	Medtronic Silhouette or Comfort 13 mm	3/7 days	20 (N/A)	80	28.1 (9.0)	7.3 (0.8)	Not registered	Karlin et al 25
				Overall	1411	53 963

Abbreviations: F, female; HbA1c, Hemoglobin A1C level; IIS, insulin infusion set; LH, lipohypertrophy; M, male; N/A, not available; SD, standard deviation; URLi, ultra-rapid lispro.

Assessment of Potential Bias

None of the trials masked the participants to the IISs. The risk of selection bias, including random sequence generation and allocation concealment, was considered low. There is also a low risk of attrition and reporting bias. Other than masking, most of the studies had low risk or unclear (if the information was not available) risk of bias. The studies with a low risk for each component other than masking were considered. A funnel plot of the primary outcomes showed no evidence of publication bias visually (Figure 3).

Figure 3.

Funnel plots of primary outcomes: wear duration (n = 23, see Figure 5) and TDD (n = 6 IISs). Abbreviation: IISs, insulin infusion set; TDD, total daily dose of insulin.

Outcomes

In the past six years, there have been 14 published articles assessing IIS performance in the insulin pump delivery system:

• Two post-hoc analyses of three clinical studies with regular IISs^13,15;

• Five cross-over trials of regular IIS compatibility with two insulin formulations—Aspart and Faster Aspart, ²³ Lispro and ultrarapid lispro (URLi),^16,18 Aspart and SAR341402 (biosimilar) ²⁰ ; Lispro and SAR342434 (biosimilar) ²² ;

• Two studies of regular IIS use for seven days^19,25;

• One regular IIS performance study in the mylife YpsoPump system (Ypsomed, Burgdorf, Switzerland) ²¹ ;

• One cross-over study of two different regular IISs ²⁴ ; and

• Three single-arm studies of extended IIS use for seven days or more.^12,14,17

Figure 4 shows the map of IIS comparisons for the outcomes of wear duration, TDD, and IIS failure reasons/modes. Meta-analysis results were based on these outcomes and are presented in the following paragraphs.

Figure 4.

A map of insulin infusion set comparisons for the outcomes: The size of each circle and the width of each line are proportional to the number of participants to each insulin infusion set combination and the number of trials comparing each insulin infusion set combination, respectively.

Wear duration

Comparisons of IIS wear durations from all 15 studies were pooled and presented in Figure 5. In recent years, new insulins (ultra-fast formulations and biosimilars) have been included in IIS studies. From all the comparative insulin study results, the impact of insulin formulation on wear duration of regular IISs was negligible.

Figure 5.

Interval plot for the IIS wear duration with standard deviation comprising the same within-study cointerventions. Missing standard deviations were imputed consistent with Cochrane guidelines. Abbreviations: HCL, hybrid closed-loop; IIS, insulin infusion set; LH, lipohypertrophy; SAR-Asp, SAR341402 biosimilar insulin aspart; SAR-Lis, SAR342434 biosimilar insulin lispro; URLi, ultra-rapid lispro.

For the regular IISs approved for two to three days or with planned set change at three days,^{13,15,16,18,20 -24} the overall grand mean wear duration was calculated to be 2.9 ± 1.0 days. For the regular IISs approved for two to three days but prespecified to be removed at seven days,^19,25 the grand mean wear duration was 5.5 ± 1.5 days. For the three extended IISs with a designed/planned set change at seven days or more,^12,14,17 the grand mean wear duration was 7.0 ± 1.9 days. It should be noted that the grand mean was a mathematical calculation weighted by the number of wears in the averaged studies, which did not reflect the glycemic control complications or other issues associated with the wear duration. Interestingly, contrary to the approved device wear time, the wear durations for 90° steel IISs (approved for two-day use) and 90° Teflon IISs (approved for three-day use) were not statistically different during reported clinical studies.^13,19

Figure 6 shows the results of mean differences in IIS wear duration between various paired IIS groups as interval plots. Based on the limited data sets, there was no significant difference between the angled and 90° extended sets. Although the mean difference in IIS wear duration favored those IISs with a longer designed/planned wear time (seven days or more), the wear duration did not demonstrate only the significant performance benefit.

Figure 6.

Interval plots of mean differences in insulin infusion set wear duration between various paired groups. Study results for devices falling into the same categories were combined into a group by a weighted average based on the number of device wears. The combined results were compared to each other using the two-sample t test with a two-sided 95% CI using the Minitab statistical software. Mean difference results were plotted in the figure below.

Total daily dose of insulin

Daily insulin requirement is a good indicator for IIS wear performance over time. Comparisons from IIS studies with reported TDD values were pooled to analyze normalized TDD, assuming the first reported TDD value at 100% (Figure 7). In three of the four studies tested to seven days or more,^14,17,19 TDD remained stable across the seven days of IIS wear. In one study (with Capillary Biomedical Extended Wear IIS), the TDD increased significantly with wear time (P = .018), although this difference did not reach statistical significance when comparing day 1.5 and 6.5 (43.5 ± 5.7 units vs 54.5 ± 14.1 units; P = .054). ¹²

Figure 7.

TDD versus seven-day wear time for IISs. The Ypsomed sample size estimated by 62% survival multiplied by the number of participants. Abbreviations: IIS, insulin infusion set; TDD, total daily dose of insulin.

IIS failure reasons/modes

During the clinical studies, participants were required to routinely change their infusion sets at a predetermined wear time, for example, 72 ± 4 hours, ²⁰ 174 hours, ¹⁴ or 10 days. ¹⁷ The date and time of all infusion set changes (scheduled or premature) and the reason for any premature infusion set changes (also called IIS failures to meet the planned wear duration) were recorded. Set failure reasons/modes (see details in the Data Extraction section) for all 15 studies were examined and are presented in Table 2.

Table 2.

Failure Modes for Various Infusion Sets in the Respective Clinical Studies.

Approved/tested IIS wear time	Categorized IIS	Reference	Time of failure evaluatedup to	Subjects in study	Total IIS wears	% Wears a
						Unexplained hyperglycemia & occlusion	Discomfort (pain, erythema, edema) at infusion site	Set related: pull-outs, leakage, adhesive failures	Infection	Others, unknown b
7/7 days	Capillary Biomedical Extended Angled Teflon	Kastner et al 12	7 days	20	41	9.7	7.3	2.4	4.8	0.0
7/7 days	Medtronic Extended 90° Teflon	Brazg et al 14	7 days	259	3041	1.3	1.6	15.7	0.5	3.9
7/10 days	ConvaTec Extended 90° Teflon	Lal et al 17 c	7 days	22	22	18.2	NR	4.5	NR	NR
2-3/7 days	Ypsomed Regular 90° Steel	Waldenmaier et al 19	7 days	40	80	18	1	11	NR	1
	Ypsomed Regular 90° Teflon		7 days		80	20	3	9	NR	5
3/7 days	Medtronic Regular Angled Teflon	Karlin et al 25	7 days	20	40	23	8	23	3	5
	Medtronic Regular Angled Teflon (LH)		7 days		40	35	5	18	8	10
2-3/3 days	Tandem Regular 90° Teflon	Kanapka et al 13	3 days	263	12372	3.6	NR	NR	NR	NR
	Tandem Regular 90° Steel		3 days		4676	1.8	NR	NR	NR	NR
	Tandem Regular Angled Teflon		3 days		3429	1.2	NR	NR	NR	NR
	Regular Unknown Sets		3 days		4677	3.8	NR	NR	NR	NR
2-3/3 days	Medtronic Regular Sets Asp	Gorst-Rasmussen et al 15	3 days	472	9360	NR
	Medtronic Regular Sets Faster Asp		3 days		9653	NR
3/3 days	Medtronic Regular 90° Teflon Lispro	Bode et al 16 d	3 days	42	392	2.3	0	3.3	NR	NR
	Medtronic 90° Regular Teflon URLi		3 days		392	2.9	1.3	2.0	NR	NR
3/3 days	Medtronic 90° Regular Teflon Lispro	Bode et al 18 e	3 days	49	672	1.6	0.1	1.6	NR	0.9
	Medtronic 90° Regular Teflon URLi		3 days		686	2.5	1.7	1.7	NR	0.3
2-3/3 days	Medtronic Regular Sets—Aspart	Thrasher et al 20 f	3 days	45	430	3.5	0.9	NR	NR	1.9
	Medtronic Regular Sets—SAR-Asp		3 days		430	4.7	0.2	NR	NR	3.7
2-3/3 days	Medtronic Regular Sets—Lispro	Thrasher et al 22 g	3 days	27	270	3.3	NR	NR	NR	NR
	Medtronic Regular Sets—SAR-Lis		3 days		250	5.6	NR	NR	NR	NR
2-3/3 days	Medtronic Regular Sets (Asp)	Zijlstra et al 23	3 days	12	174	4.6	NR	NR	NR	NR
	Medtronic Regular Sets (Faster Asp)		3 days	25	379	4.7	NR	NR	NR	NR
3/3 days	Ypsomed Regular 90° Teflon	Waldenmaier et al 21	3 days	35	406	3.9	3.7	10.6	NR	0.5
3/3 days	Roche Regular 90° Teflon FL+	Freckmann et al 24	3 days	80	895	8.4	2.4	4.0	NR	5.1
	Roche Regular 90° Teflon FL		3 days		854	6.7	3.1	2.6	NR	3.7

Abbreviations: FL, FlexLink infusion set; IIS, insulin infusion set; LH, lipohypertrophy; NR, Not Reported in this table; SAR-Asp, SAR341402 biosimilar insulin aspart; SAR-Lis, SAR342434 biosimilar insulin lispro; URLi, ultra-rapid lispro.

^aMajority of studies allow for selection of ≥1 failure reasons.

^bOthers or unknown included failures such as mechanical failures, kinked cannula, and personal decision of removal when BG was higher than their comfort level.

^cThe failure rates were re-calculated based on 7-day failures instead of 10-day data in the publication.

^dThe failure rates were re-calculated based on four-week total wears instead of subjects used in Table 3 of the original reference: 4*7*42/3 = 392 for the lispro group, 4*7*42/3 = 392 for the URLi group.

^eThe failure rates were re-calculated based on six-week total wears instead of subjects used in Table S3 of the original reference: 48*14 = 672 for the lispro group, 49*14 = 686 for the URLi group.

^fThe failure rates were re-calculated based on 30-day total wears instead of subjects used in Table 3 of the original reference: 43*10 = 430 for the aspart group, 43*10 = 430 for the biosimilar group.

^gThe failure rates were re-calculated based on 30-day total wears instead of subjects used in Table 2 of the original reference: 27*10 = 270 for the lispro group, 25*10 = 250 for the biosimilar group.

Unexplained hyperglycemia (UH, also named “suspected occlusion” or “prolonged hyperglycemia”) was a major efficacy endpoint in 14 of the 15 IIS studies.^{12 -14,16 -22} Although the details in the UH definition were slightly different in each study, UH may be described as a high blood glucose event that could not be explained by a missed prior bolus, dietary indiscretion, rebound or treatment of hypoglycemia, a pump failure, an empty pump reservoir, an infusion set complication (e.g., kinked, pullout, leakage), or an infusion site complication (e.g., pain, redness). The UH event was normally detected by CGM and confirmed by SMBG >13.9 mmol/L (250 mg/dL) that did not decrease within 1-2 h after a correction bolus delivered via the pump. The rate of UH, in combination with the rate of occlusion (detected or suspected), was evaluated across all studies and plotted in Figure 8. When the regular IISs were used for three days as approved, the reported rates of UH and occlusion during the clinical studies were generally low (≤ 9%). When the regular IISs were used for seven days, the rate of UH and occlusion increased significantly, up to 35% in subjects with lipohypertrophy.^19,25 Even for one of the extended IISs designed for seven days, the seven-day rate of UH and occlusion was still high at 18%. ¹⁷

Figure 8.

IIS removals due to unexplained hyperglycemia and occlusion for all the referenced studies. Abbreviations: FL, FlexLink infusion set; HCL, hybrid closed-loop; IIS, insulin infusion set; LH, lipohypertrophy; SAR-Asp, SAR341402 biosimilar insulin aspart; SAR-Lis, SAR342434 biosimilar insulin lispro; URLi, ultra-rapid lispro.

IIS planned change rate or survival rate

The three-day planned IIS change rates were reported in half (5) of the regular IIS studies and the seven-day IIS survival rates were reported in all the IIS studies with wear time of seven days or more. The relevant data were analyzed and presented in Table 3. The three-day planned change rates for regular sets varied from 65.9 to 95.3% in various studies, which were consistent with observations from other earlier studies: Patel et al. ²⁶ Three-day survival rate of 85% for regular 90° steel sets and 77% for regular 90° Teflon sets; Schmid et al ²⁷ three-day survival rate of 75% for regular Angled Teflon sets. The seven-day survival rates for the regular IISs worn for seven days varied from 35 to 66%, which were slightly higher than the seven-day survival rates observed by Patel et al ²⁶ : 32% for regular 90° steel sets and 33% for regular 90° Teflon sets. On the other hand, the seven-day survival rates for the three extended IISs were all ≥75%.

Table 3.

Summary of Planned IIS Change Rate for 2/3-Day Sets or 3/7-Day Survival Rate for IISs (Regular and Extended) Tested to Seven Days.

Publication date	IIS	Total wears	Planned 2/3-day IIS removal (%) a	3-day survival (%)	7-day survival		Reference
					Rate (%)	95% CI (%)
2022 June	Capillary Biomedical Extended Angled Teflon	41	N/A	95	88	±10	Kastner et al 12
2022 March	Medtronic Extended 90° Teflon	3041		95	78	±1.5	Brazg et al 14
2022 January	Medtronic Regular Sets	19 013	65.9	N/A			Gorst-Rasmussen et al 15
2021 June	ConvaTec Extended 90° Teflon	22	N/A	100	77	±18	Lal et al 17
2020 October	Ypsomed 90° Steel	80		93	69	±10	Waldenmaier et al 19
	Ypsomed 90° Teflon	80		95	63	±11
2019 January	Ypsomed Regular 90° Teflon	406	81.0	N/A			Waldenmaier et al 21
2018 January	Medtronic Regular Sets	553	95.3				Zijlstra et al 23
2017 March	Roche Regular 90° Teflon FL+	895	78.5				Freckmann et al 24
	Roche Regular 90° Teflon FL	854	82.0
2016 July	Medtronic Regular Angled Teflon	40	N/A	85	43	±15	Karlin et al 25
	Medtronic Regular Angled Teflon (LH)	40		85	24	±13

Abbreviations: CI, confidence interval; FL, FlexLink infuion set; IIS, insulin infusion set; LH, lipohypertrophy; N/A, not applicable.

^aPlanned removals may be less than three days.

Discussion

To our knowledge, this is the first meta-analysis of IIS wear duration, TDD, and IIS failure reasons/modes. We found a small number of clinical trials that investigated extended IISs. It is clear from these studies that wear duration of IISs has improved over the past few years.

There were three IISs designed to extend wear duration: the ConvaTec Extended 90° Teflon set, the Medtronic Extended 90° Teflon set, and the CBX Extended-Wear Angled Teflon set. They were all modified upon standard IISs to address root cause failure modes. Both the ConvaTec and the CBX sets are purported to have a novel cannula: the former with additional slitted openings designed to reduce the risk of occlusion and a modified coating to reduce inflammation ¹⁷ and the latter designed with an angled cannula with significantly improved immunological tolerability at the infusion site. ¹² The Medtronic set is claimed to reduce site inflammation and pain through reducing particulates/aggregates and hydrophobic molecules in the infused insulin. ²⁸ Detailed IIS failures for the three extended sets are presented in Table 2 and further discussed in later sections.

Interestingly, some of the aforementioned studies reported that standard IIS survival appeared to depend heavily on the individual user. In the study reported by Patel et al, ²⁶ 21% of users had a mean length of infusion set wear of less than four days, and 42% had a mean length of infusion set wear of greater than six days. Waldenmaier and colleagues found that 75% of participants wore at least one standard IIS for seven days. ¹⁹ Therefore, in a pilot study with fewer participants, results from one or two subjects may significantly affect the study outcomes. In the larger-scale study conducted for the Medtronic Extended 90° Teflon sets, 99.6% of the participants were able to wear at least one IIS for seven days. Therefore, more participants may be needed for future IIS studies.

Since 2016, IIS studies have investigated IIS compatibility with new insulins and extending IIS wear duration. This systematic review suggests that there is limited impact of new insulins (including ultrarapid and biosimilar formulations of insulin lispro and aspart) on the performance of regular IISs for two- to three-day use, although site reaction (pain) was observed at higher rate for URLi than for control.^16,18 Because the major concern of site reaction (pain, UH, or occlusion) might be mitigated by the extended wear set design, whether or not the observations from the use of ultrarapid insulins with regular IISs apply to the extended IISs remain to be determined in future clinical trials.

Among all the reviewed IIS studies, only several studies were conducted with the AID systems used (Table 1).^{12 -14,16} Based on the limited data, the impact of AID system on IIS wear duration (Figure 5), the rate of UH and occlusion (Table 2), or the IIS survival rate (Table 3) was not significant for both the regular^13,16 and extended^12,14 infusion sets.

Total daily utilization of insulin can be used as an indirect measure for infusion site performance. Increased TDD with wear time may be partially attributed to diminishing site patency. TDD ranged widely in various studies, reflecting the wide TDD distribution in pump users. For better visual comparison, TDD was normalized in Figure 7. In the three studies with 90° sets (regular and extended) with planned wear time of seven days or more, TDD remained stable across the wear time.^14,17,19 The only TDD increase with wear time was observed in the CBX angled Teflon set, which suggests a deterioration of insulin absorption with time. ¹² However, TDD data in three of the four studies were from small exploratory trials with only 20, 22, and 40 participants,^12,17,19 and the mean TDD on day seven only contained data from 15, 19, and 25 participants. As such, data from these small study groups might have been heavily influenced by a few subjects. A larger-scale study would be needed to evaluate TDD change with wear time as part of the IIS performance assessment.

One factor which may limit the wide adoption of an extended IIS is the size of the insulin reservoir. As seen in the data from Brazg et al ¹⁴ that included 259 participants, while the mean TDD was relatively unchanged (54.3 ± 28.6 U, 53.5 ± 28.5 U, and 53.6 ± 28.0 U on days one, four, and seven, respectively) the range of TDDs varied from 10.4 U/day to 168.7 U/day. Thus, a user with a high TDD (eg, 168.7 U/day) must change reservoirs (with a max fill volume of 300 U) daily, whereas the user with a TDD of 10 U/day did not need to change the reservoir until the end of seven days. With the overall mean TDD of ~50 U/day, more than half of the users had to change reservoirs during the seven-day wear. To fully improve the usability of seven-day IISs, a larger reservoir may be desirable. But given the limitations of the available space in insulin pumps, this is unlikely to occur. Using more concentrated insulins (eg, U-200 or U-400) with extended wear IISs is another option, yet no studies have reported with these insulins.

Understanding the reasons for IIS failures is an important aspect of assessing IIS clinical safety and usefulness. Thus, the reasons/modes of IIS failures (also called premature removals or unplanned removals) were segregated into five different categories, and the rate of each categorized failure was examined for all IISs. For the regular IISs used as approved, the rates of UH and occlusion ranged from 1.3% to 8.4% of wear. When the regular IISs were used for seven days, the rate of UH and occlusion increased to the range of 18% to 35%.^17,19 The IISs had different primary failure modes during the clinical studies, for example, the main failure reasons for Medtronic Extended 90° Teflon sets (15.7%) were inadvertent pull-outs, leakage, and adhesive failures. Despite the commercialization of Medtronic Extended IIS in Europe in 2021, more research on the adhesive patch is needed to fully utilize the extended IIS technology for improving the quality of life for IIS users. For the other two extended IISs (CBX Extended Angled set and ConvaTec Extended 90° Teflon set), the rate of UH and occlusion may be further improved.

Currently, reservoir filling and change is a complex task for pump users. Using prefilled cartridges to accelerate the reservoir change process may be beneficial for wider adoption of the extended IIS technology. Moreover, as revealed in these published studies, possible confounding factors, such as the benefit of extending the wear duration and the safety risk of increased UH and occlusion or site-/set-related failures associated with longer wear time, should be further evaluated.

Finally, the current literature review clearly indicates that further studies are needed to assess patient-reported outcomes, potential increase of insulin requirements, reservoir change frequency, and skin reactions associated with extending IIS wear duration. In addition, economic analyses are necessary to assess the cost-effectiveness of extended IISs and new insulins combined in an AID system.

Conclusion

IIS technology advancements have included not only extending the duration of IIS wear but also maintaining glycemic control, while mitigating adverse events across wear duration. This meta-analysis provides the most comprehensive analyses of the available clinical evidence on IIS technology to date. It offers a roadmap for further improvement on a few remaining issues of insulin infusion devices—both technical (eg, improving adhesive performance to reduce set failures) and nontechnical (eg, increasing reservoir size or using prefilled cartridge to improve user experience).