CoChat
Sign Up
Cited by 51Diabetes Management and ResearchIntravenous Infusion Technology and SafetyDiabetes Management and Education

Early Detection of Infusion Set Failure During Insulin Pump Therapy in Type 1 Diabetes

Open access
Cescon, Marzia, DeSalvo, Daniel J., Ly, Trang T., Maahs, David M., Messer, Laurel H., Buckingham, Bruce A., Doyle, Francis J., Dassau, Eyal·2016·Journal of Diabetes Science and Technology
DOI 10.1177/1932296816663962PMC PMC5094340
Open this paper in CoChat — free accountOpen PDF

Abstract

Background: Insulin infusion set failure resulting in prolonged hyperglycemia or diabetic ketoacidosis can occur with pump therapy in type 1 diabetes. Set failures are frequently characterized by variable and unpredictable patterns of increasing glucose values despite increased insulin infusion. Early detection may minimize the risk of prolonged hyperglycemia, an important consideration for automated insulin delivery and closed-loop applications. Methods: A novel algorithm designed to alert the patient to the onset of infusion set failure was developed based upon continuous glucose sensor values and insulin delivered from an insulin pump. The method was calibrated on 12 weeks of infusion set wear without failures recorded by 4 patients in ambulatory conditions and prospectively validated on 18 weeks of infusion set wear with and without failures belonging to 9 other subjects in ambulatory conditions. Results: The algorithm, evaluated retrospectively, identified a failure 2.52 ± 1.91 days ahead of the actual event as recorded by the clinical team, corresponding to 50% sensitivity, 66% specificity and 55% accuracy. If set failure alarms had been activated in real time, the average time >180 mg/dl would be reduced from 82.7 ± 40.9 hours/week/subject (without alarm) to 58.8 ± 31.1 hours/week/subject (with alarm), corresponding to a potential 29% reduction in time spent >180mg/dl. Conclusion: The proposed method for early detection of infusion set failure based on glucose sensor and insulin data demonstrated favorable results on retrospective data and may be implemented as an additional safeguard in a future fully automated closed-loop system.

AI snapshot

Written from the paper’s full text.

TL;DR

The paper proposes a data-driven algorithm to anticipate infusion set failure during insulin pump therapy in type 1 diabetes using CGM, glucose trend, and insulin intake data. In retrospective validation, it detected failures 2.52 ± 1.91 days ahead but achieved 50% sensitivity, 66% specificity, and 55% accuracy.

Key findings

  • The algorithm achieved 50% sensitivity, 66% specificity, and 55% accuracy, with 6 true positives, 2 false positives, 6 false negatives, and 4 true negatives.
  • Failures were identified 2.52 ± 1.91 days ahead of the actual event recorded by the clinical team.
  • If the alarm had been activated in real time and patients reacted, the authors estimated a potential 29% reduction in time spent above 180 mg/dl, from 82.7 ± 40.9 hours/week/patient to 58.8 ± 31.1 hours/week/patient.
  • The method does not explicitly use a glucose regulatory model, meal announcement, or physical activity inputs, and aims to anticipate rather than detect faults; the authors state it is the first time such a method has been applied to actual outpatient data.

Methods

  • Twenty-three subjects with type 1 diabetes at 2 clinical sites wore teflon infusion sets for 7 days or until failure; CGM was measured with a Dexcom G4 Platinum sensor. Of 69 patient-weeks collected, 30 weeks from 15 subjects were suitable; 12 failure-free weeks from 4 patients were used for calibration, and 18 weeks (13 with verified failures, 5 without) from 9 patients were used for validation.
  • The algorithm used three components: CGM data segmentation via a time-varying autoregressive model with sum-of-norms regularization, linear trend slope assessment, and insulin counting comparing bolus number and total insulin infused in the past τ hours versus the same interval 24 hours earlier. User parameters were n = 3, λ = 0.007 λmax, α = 1.5, and τ = 4.
  • The algorithm was evaluated retrospectively and blind to failures; sensitivity, specificity, and accuracy were computed, and time spent above 180 mg/dl was compared with and without the alert.

Limitations

  • The paper states the method requires an uninterrupted stream of CGM data to detect an increasing glucose trend accurately; it is not suitable for abrupt failures such as tubing caught on a door handle or suddenly ripped off; patients changing the set and glucose sensor simultaneously at week 2 fed the algorithm a poorly calibrated data stream, negatively affecting performance, and no mitigation was taken.
  • The paper states long-term clinical studies are needed to minimize the false positive rate, and that additional data and clinical work may improve specificity and accuracy.
  • Evident from the text: validation was retrospective and small (18 weeks from 9 patients), and the 29% reduction in hyperglycemia was “potential,” assuming patients would acknowledge and react to alerts; real-time clinical utility remains to be established.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1177/1932296816663962

Cites

20 of 20 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

2 of 6 cited sources checked · 2 supported · 0 flagged

  • Premisecitation

    In all these cases, metabolic decompensation and development of ketoacidosis can occur due to insufficient insulin delivery.

    “In all these cases, metabolic decompensation and development of ketoacidosis can occur due to insufficient insulin delivery.5”
    Quote in the paperSubstantial claim1 of 1 sources checked
  • Findingnone given

    Infusion set failures are frequently characterized by variable and unpredictable patterns of increasing glucose values despite increased insulin infusion.

    “Set failures are frequently characterized by variable and unpredictable patterns of increasing glucose values despite increased insulin infusion.”
    Quote in the paperSubstantial claim
  • Findingown data

    Retrospective analysis of the collected data highlighted a persistent ramp-like increase in glycemia levels for weeks with an infusion set failure, as opposed to no increase for fault-free weeks.

    “Retrospective analysis of the collected data highlighted a persistent ramp-like increase in glycemia levels for those weeks presenting an infusion set failure, as opposed to no increase for weeks fault-free”
    Quote in the paperSubstantial claim
  • Findingnone given

    The main predictor of length of wear was the individual subject.

    “The main predictor of length of wear was the individual subject.”
    Quote in the paperNot a substantial claim
  • Findingown data

    Failures were identified 2.52 ± 1.91 days ahead of the actual event.

    “The failures were identified 2.52 ± 1.91 days ahead of the actual event.”
    Quote in the paperSubstantial claim
  • May combine two claims. 14
  • Classified as probably not a substantial claim. 11
  • The quote only partly states this claim. 1

  • May combine two claims.

    Premisecitation

    Infusion set failure can occur when there is local inflammation at the infusion site, when the catheter is occluded, kinked or has simply dislodged from the subcutaneous space.

    “Infusion set failure can occur when there is local inflammation at the infusion site, when the catheter is occluded, kinked or has simply dislodged from the subcutaneous space.3,4”
    Quote in the paperNot a substantial claim1 of 2 sources checked
    • cites [3] source
    • cites [4] source

  • May combine two claims.

    Findingnone given

    In case of a developing failure, the patient glucose-insulin dynamics is slowly changing, presenting variable patterns of increasing glucose levels, and more insulin is being infused by the patients outside of their ordinary therapy because insulin is not as effective in lowering blood glucose.

    “In case of a developing failure, the patient glucose-insulin dynamics is slowly changing, presenting variable patterns of increasing glucose levels.At the same time, more insulin is being infused by the patients outside of their ordinary therapy because insulin is not as effective in lowering blood glucose (see Figure 1 for a representative subject).”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    The algorithm achieved 50% sensitivity, 66% specificity and 55% accuracy, with 6 TP, 2 FP, 6 FN, 4 TN.

    “The algorithm achieved 50% sensitivity, 66% specificity and 55% accuracy, with 6 TP, 2 FP, 6 FN, 4 TN.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingown data

    In the validation subset, 17% of sets failed due to a failed correction dose, 17% failed because of serum ketones ≥0.6 mmol/l with hyperglycemia, 28% had erythema or induration >1 cm in diameter at the infusion site, 5% were removed because of bleeding and 5% were pulled out by accident.

    “With regard to the second subset, overall, 17% of the sets failed due to a failed correction dose, 17% failed because of the presence of serum ketones ≥0.6 mmol/l with hyperglycemia, 28% had erythema or induration > 1 cm in diameter at the infusion site, 5% were removed because of bleeding and 5% were pulled out by accident.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    The method was evaluated retrospectively on a validation data set collected from patients in ambulatory conditions and demonstrated favorable results detecting the onset of faulty events before they become severe with enough time to react to them.

    “The method was evaluated retrospectively on a validation data set collected from patients in ambulatory conditions and demonstrated favorable results detecting the onset of faulty events before they become severe with enough time to react to them.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Interpretationcitation

    Early detection of infusion set failure may minimize the risk of prolonged hyperglycemia, an important consideration for automated insulin delivery and closed-loop applications.

    “Early detection may minimize the risk of prolonged hyperglycemia, an important consideration for automated insulin delivery and closed-loop applications.1”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [1]

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Study design and conductnone given

    The method was developed and retrospectively validated on actual patients data collected in ambulatory conditions.

    “It was developed and retrospectively validated on actual patients data collected in ambulatory conditions.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Interpretationreasoning

    One single larger insulin injection may indicate the subject had a bigger meal, while several closely spaced small injections with total insulin the same as the larger dose are more likely to indicate that a failure is developing.

    “One single larger insulin injection may indicate the subject had a bigger meal, while several closelyspaced small injections with total insulin the same as the larger dose are more likely to indicate that a failure is developing.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    Overall, application of the method showed that it is possible to extend the length of wear of infusion sets beyond 3 days with associated benefits, without incurring an increase time spent in hyperglycemia.

    “Overall, application of the method showed that it is possible to extend the length of wear of infusion sets beyond 3 days with associated benefits, without incurring in an increase time spent in hyperglycemia.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: The patient would have acknowledged the alert and reacted taking appropriate actions.

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationreasoning

    Our method was designed to be proactive and to be able to deal with the harder constraint of slow occlusions in a predictive manner rather than to detect abrupt faults like catheter disconnections, which are unforeseen events and as such cannot be predicted.

    “Our method was designed to be proactive and to be able to deal with the harder constraint of slow occlusions in a predictive manner rather than to detect abrupt faults like catheter disconnections, which are unforeseen events and as such cannot be predicted.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    The proposed method is aiming at anticipating faults rather than recognizing them after they have happened.

    “Moreover, the proposed method is aiming at anticipating faults rather than recognizing them after they have happened.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    The results obtained with the proposed failure detection algorithm are independent of the type of insulin infusion system.

    “The results obtained with the proposed failure detection algorithm are independent of the type of insulin infusion system.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationreasoning

    Both the number of insulin boluses and the quantity of infused insulin need to be taken into account, under the assumption that neither the time of the meal nor the meal content are available.

    “We would like to highlight that both the number of insulin boluses and the quantity of infused insulin need to be taken into account, under the assumption that neither the time of the meal nor the meal content are available.”
    Quote in the paperSubstantial claim
    • The authors state this as an assumption: Neither the time of the meal nor the meal content are available.

  • Classified as probably not a substantial claim.

    Study design and conductown data

    The proposed algorithm was evaluated on 18 weeks of data from 9 patients in a retrospective and blind to the failures manner.

    “The proposed algorithm was evaluated on 18 weeks of data from 9 patients in a retrospective and blind to the failures manner.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Study design and conductnone given

    The 3-step procedure ensures an increased degree of robustness to the method.

    “The 3-step procedure ensures an increased degree of robustness to the method.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    Alerts on an infusion site failure will reduce the complications and glucose swings associated with this type of events.

    “However, the alerts on an infusion site failure will reduce the complications and glucose swings associated with this type of events.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Interpretationnone given

    To the best of the authors knowledge this is the first time that a method with these features has been applied to actual outpatient data.

    “To the best of the authors knowledge this is the first time that a method with these features has been applied to actual outpatient data.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims. The quote only partly states this claim.

    Interpretationreasoning

    The proposed method is robust and is able to deal with inter-current illness, medication, hormone cycles and physiological stress because its design was based on machine learning tools.

    “The proposed method is robust and is able to deal with this fact because its design was based on machine learning tools.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

    Interpretationreasoning

    Early detection and mitigation of such failures are critical safety issues not only for a closed-loop (artificial pancreas) system, but also for current open-loop systems.

    “Early detection and mitigation of such failures are critical safety issues not only for a closed-loop (artificial pancreas) system, but also for current open-loop systems.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationnone given

    The advantage of our algorithm compared to the cited works is that it does not require an explicit description of the glucose-insulin regulatory model, nor does it exploit user information such as meal announcement or physical activity.

    “The advantage of our algorithm compared to the above cited works is that it does not require an explicit description of the glucose-insulin regulatory model, nor does it exploit user information such as meal announcement or physical activity.”
    Quote in the paperSubstantial claim
20 of 20

  • The paper relies on this without support.

    Patients may only change the infusion set when they notice a persistent increase in their blood glucose level, meaning that the set is not properly working.

    “Patients may only change the infusion set when they notice a persistent increase in their blood glucose level meaning that the set is not properly working.”

  • The authors state this as an assumption.

    Neither the time of the meal nor the meal content are available.

    “under the assumption that neither the time of the meal nor the meal content are available.”

  • The authors state this as an assumption.

    The patient would have acknowledged the alert and reacted taking appropriate actions.

    “we assume the patient would have acknowledged the alert and reacted taking appropriate actions.”
3 of 3

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • A failure detection algorithm applied to 100 in silico pump patients with type 1 diabetes detected a total occlusion or dislocation in 75% of patients within 63 minutes with a 10% false positive rate.

    “A failure detection algorithm applied to 100 in silico pump patients with type 1 diabetes detected a total occlusion or dislocation in 75% of patients within 63 minutes with a 10% false positive rate.16”

    From Insulin Pump Occlusions: For Patients Who Have Been Around the (Infusion) Block (2017) · cites this paper as [16]

    The Performance Evaluation section says this paper tested its algorithm retrospectively on 18 weeks of data from 9 patients, not 100 simulated patients. The Results report 50% sensitivity and failures identified 2.52 ± 1.91 days before the recorded event, rather than 75% detected within 63 minutes; they do not report a 10% false-positive rate. The Discussion mentions a separate approach tested on 100 simulated patients, but does not give the claimed performance figures. (LLM, conf=0.89)

    Checked by a reasoning model — may be wrong.

    Experimental — accuracy not yet measured

  • IIS failures are the primary contributor to DKA development and the leading cause of DKA-related hospitalizations among insulin pump users.

    “Correspondingly, IIS failures are the primary contributor to DKA development and the leading cause of DKA-related hospitalizations among insulin pump users.[9][10][11][12]”

    From Frequency and Detection of Insulin Infusion Site Failure in the Type 1 Diabetes Exchange Online Community (2023) · cites this paper as [11]

    The Abstract and Conclusions state that infusion set failures can cause DKA in people with type 1 diabetes using pumps. They do not establish that these failures are the primary contributor to DKA or the leading cause of DKA-related hospitalizations among pump users. Page numbers are unavailable in our copy. (LLM, conf=0.87)

    Experimental — accuracy not yet measured

  • Occlusions of insulin infusion sets occur frequently.

    “Occlusions do occur frequently”

    From Insulin Pump Occlusions: For Patients Who Have Been Around the (Infusion) Block (2017) · cites this paper as [17]

    The Abstract (page unavailable) identifies catheter occlusion as a cause of infusion set failure, so it supports that occlusions occur. It does not say how frequently occlusions occur; its statement that failures are ‘frequently characterized’ by rising glucose describes their pattern, not their incidence. (LLM, conf=0.83)

    Experimental — accuracy not yet measured

  • Occlusions of insulin infusion sets often occur without an alarmed warning, or the warning is provided too late.

    “often without an alarmed warning or this provided too late”

    From Insulin Pump Occlusions: For Patients Who Have Been Around the (Infusion) Block (2017) · cites this paper as [17]

    The Discussion (page not available) says pump mechanisms may detect a full occlusion only after some time and do not detect infusion-site infections, supporting the concern about late or absent warnings for some failures. It does not establish that occlusions often occur without an alarm. (LLM, conf=0.84)

    Experimental — accuracy not yet measured

  • In a clinical trial with nine subjects, the Cescon et al algorithm achieved 50% sensitivity and 66% specificity for predicting IIS failure.

    “The algorithm achieved 50% sensitivity and 66% specificity.”

    From Insulin Pump Occlusions: For Patients Who Have Been Around the (Infusion) Block (2017) · cites this paper as [17]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Subject counts exceed the two subsets combined

    Minor

    The two subsets contain 4 and 9 patients, so they can include at most 13 distinct subjects, not the reported 15.

    Reported 15 · recomputed 13

  • Failure-reason codes conflict with their definitions

    Minor

    Table 1 defines A as a failed correction, B as hyperglycemia with ketones, and C as an infusion-site infection. Table 2 labels a failed correction B and a tubing accident C.

2 of 2

Cites

Flagged first, then by citations

Research works printed but not linked: 7

  1. Williams G, 2004, The Handbook of Diabetes
  2. Rojas R, 18th IFAC World Congress, 2011, 14127
  3. Rojas R, 2011, American Control Conference, 1952
  4. Cameron F, Diabetes Tech Meeting, 2012, A17
  5. Facchinetti A, 33rd IEEE EMBS, 2011, 4947
  6. Del Favero S, 2014, 19th IFAC World Congress, 1941
  7. Bergman RN, 1979, Am J Physiol, 236, E667

Cited by

Sorted by citations

Related

Authors
Cescon, M., DeSalvo, D. J., Ly, T. T., Maahs, D. M., Messer, L. H., Buckingham, B. A., Doyle, F. J., & Dassau, E.
Affiliations
  • University of California, Santa Barbara (US)
  • Baylor College of Medicine (US)
  • Stanford Medicine (US)
  • University of Colorado Denver (US)
  • Harvard University (US)
Corresponding author
Dassau, Eyal
ORCID
Venue
Journal of Diabetes Science and Technology
Publisher
SAGE Publications
Published
2016-09-25
Volume
10
Issue
6
Pages
1268-1276
ISSN
1932-2968 (ISSN-L)
DOI
10.1177/1932296816663962
PMC
PMC5094340
OpenAlex
W2519750058
Semantic Scholar
b01354bd500d2f07bf5246563b2e5c1b44f454d4
Cited by
51
References
16
Citations by year
2026: 2 · 2025: 3 · 2024: 7 · 2023: 8 · 2022: 4 · 2021: 5 · 2020: 5 · 2019: 5 · 2018: 8 · 2017: 3 · 2016: 1
FWCI
4.93
Type
journal article
Language
en
Open access
Licence
journals.sagepub.com/page/policies/text-and-data-mining-license (tdm)
Locations
  • Journal of Diabetes Science and Technology (publishedVersion)
  • PubMed (publishedVersion)
  • Europe PMC (PubMed Central) (submittedVersion)
  • PubMed Central (submittedVersion)
Funders
  • National Institutes of Health
  • National Institute of Diabetes and Digestive and Kidney Diseases (DP3DK 094331)
Topics
  • Health Sciences › Medicine › Endocrinology, Diabetes and Metabolism › Diabetes Management and Research
  • Physical Sciences › Engineering › Biomedical Engineering › Intravenous Infusion Technology and Safety
  • Health Sciences › Medicine › Endocrinology, Diabetes and Metabolism › Diabetes Management and Education
Keywords
type 1 diabetes, insulin pump therapy, continuous glucose monitoring, hyperglycemia, diabetic ketoacidosis, artificial pancreas, alarm system
MeSH terms
Algorithms, Blood Glucose, Diabetes Mellitus, Type 1, Equipment Failure, Female, Humans, Hyperglycemia, Hypoglycemic Agents, Insulin, Insulin Infusion Systems, Male, Sensitivity and Specificity, Blood Glucose Self-Monitoring, Clinical Alarms

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-13.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-13.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review