Defining a glycemic persistence index (GPI) for continuous glucose monitoring

Defining a glycemic persistence index (GPI) for continuous glucose monitoring | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Method Article Defining a glycemic persistence index (GPI) for continuous glucose monitoring Ren Zhang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8841862/v2 This work is licensed under a CC BY 4.0 License Status: Posted Version 2 posted You are reading this latest preprint version Show more versions Abstract Sustained hyperglycemia is the core driver of diabetes complications. Continuous glucose monitoring (CGM) provides rich temporal glucose data, but effective summarization that integrates both magnitude and duration of sustained hyperglycemia into a single and paramter-free scalar remains challenging. We introduce the glycemic persistence index (GPI), a simple, threshold-free CGM-derived metric defined as the largest integer k such that at least k minutes are spent at glucose levels ≥ k mg/dL within a day. For example, a GPI of 125 means that glucose levels were ≥ 125 mg/dL for at least 125 minutes during that day. Geometrically, after ranking glucose values in decreasing order, GPI is given by the intersection at which glucose level and cumulative duration take the same value. Using CGM data from the International Diabetes Closed Loop randomized trial (168 adults with type 1 diabetes), GPI sensitively captured closed-loop treatment benefit with significantly lower values in the closed-loop group compared with sensor-augmented pump therapy (P = 0.0001). GPI correlated with HbA1c and daily mean glucose, while distinguishing patterns of sustained hyperglycemia among individuals with similar conventional metrics. As a simple, device-independent, and threshold-free scalar, GPI quantifies hyperglycemia by jointly capturing its magnitude and duration, enabling consistent and intuitive glycemic profiling accessible to both specialists and non-specialists. Continuous glucose monitoring CGM glycemic persistence index GPI hyperglycemia Figures Figure 1 Figure 2 Figure 3 Introduction Diabetes and dysglycemia are highly prevalent worldwide, spanning a continuum from prediabetes to established diabetes and affecting an increasing proportion of adults 1 , 2 . Across this spectrum, chronic exposure to elevated glucose is a central determinant of metabolic burden and long-term risk 3 , 4 . Continuous glucose monitoring (CGM) has transformed glycemic assessment by providing high-frequency measurements in daily life. These data capture rich temporal structure, including fluctuations, excursions, and prolonged periods of dysglycemia, but their volume and complexity necessitate effective summarization into concise, interpretable metrics. In both clinical practice and randomized trials, CGM-derived summaries are routinely used to evaluate glycemic control and therapeutic efficacy 5 , 6 . Commonly reported metrics include daily mean glucose, measures of glycemic variability, and threshold-based summaries such as time-in-range (TIR) 7 – 11 . For example, TIR served as the primary efficacy endpoint in the International Diabetes Closed Loop (DCLP3) randomized trial, in which closed-loop therapy significantly improved TIR compared with control 12 . Threshold-based metrics, however, require prespecified glucose cutoffs. These cut points may differ across populations, including type 1 diabetes, type 2 diabetes, and prediabetes, and results can vary depending on the selected thresholds. Once defined, such metrics provide no further gradation above the chosen boundary. Moreover, sustained hyperglycemia reflects not only how long glucose remains elevated but also how high it rises. Mean glucose summarizes overall exposure, and variability measures describe dispersion, but neither directly integrates magnitude and duration into a unified representation of persistence. A metric that captures glycemic persistence should therefore encode both dimensions within a single, interpretable scalar while remaining independent of arbitrary cutoffs. To address these challenges, we introduce the glycemic persistence index (GPI), a simple, nonparametric metric derived from CGM data that integrates glucose magnitude and duration into a single scalar expressed in minutes. GPI is immediately interpretable and does not rely on predefined glucose thresholds or modeling assumptions. Using CGM data from the DCLP3 randomized trial, we evaluate GPI and examine its relationship to existing CGM summaries. Materials and methods Continuous glucose monitoring data CGM data were derived from the International Diabetes Closed Loop (iDCL) randomized trial, which enrolled 168 participants with type 1 diabetes (112 assigned to closed-loop therapy and 56 to sensor-augmented pump therapy) 12 . The study duration was 26 weeks. Glucose measurements were obtained using the Dexcom G6 continuous glucose monitoring system (Dexcom), which records interstitial glucose values every 5 minutes (up to 288 measurements per day). Raw timestamped CGM data were analyzed without smoothing or interpolation. Timestamps were converted from SAS datetime format to standard datetime objects. For each participant, glucose values were aligned relative to the randomization date. Baseline CGM data were defined as recordings obtained prior to randomization during the run-in period. Glucose measurements were grouped by participant and calendar date to generate daily glucose profiles. Days with fewer than 200 valid CGM readings were excluded to ensure adequate data completeness. All analyses were conducted at the subject-day level. Definition of the glycemic persistence index (GPI) For each subject-day, CGM measurements were used to compute the GPI. Glucose values within a day were represented as a time series with corresponding timestamps and ranked in decreasing order of glucose concentration. Cumulative time was computed based on the native sampling interval, yielding the total duration (in minutes) associated with glucose values exceeding a given level. Let \({G}_{t}\) denote the glucose concentration (mg/dL) measured at time \(t\) , and let \({\Delta}{t}_{t}\) denote the duration (in minutes) represented by that CGM measurement, determined by the native sampling interval. The GPI was defined as $$\text{G}\text{P}\text{I}=\text{m}\text{a}\text{x}\left\{k\in\mathbb{N}:\sum_{\left\{t:{G}_{t}\gek\right\}}{\Delta}{t}_{t}\gek\right\}.$$ This definition identifies the largest integer value k for which the cumulative time spent at glucose values greater than or equal to k mg/dL is at least k minutes. Thus, GPI jointly encodes glucose magnitude and duration on a common numerical scale, increasing when higher glucose levels persist for longer periods. GPI can be computed by iterating over integer glucose levels k, summing CGM durations at or above k, and identifying the maximum k satisfying the persistence condition. GPI is expressed in units of time (minutes) and is evaluated over integer values of \(k\) , ensuring that it is independent of CGM sampling frequency and directly comparable across devices and study settings. Statistical analysis, software, and visualization Daily GPI values were computed for each subject-day using quality-controlled CGM data. For longitudinal treatment analyses, daily GPI values were averaged within each participant during the baseline period and during the post-randomization period (weeks 1–26). Between-group comparisons of post-randomization GPI were performed using analysis of covariance (ANCOVA), with treatment group included as a fixed effect and baseline mean GPI as a covariate. For subject-day-level analyses, daily mean glucose and daily glucose variance were calculated using all available CGM measurements within each 24-hour period. Associations between daily GPI and conventional glycemic metrics were assessed using Pearson correlation coefficients. All statistical tests were two-sided, and P values < 0.05 were considered statistically significant. Data processing and statistical analyses were performed using Python 3 (statsmodels). Data manipulation was conducted using standard scientific computing libraries, and figures were generated using matplotlib. Results Geometric definition of the glycemic persistence index Geometrically, GPI is defined from ranked daily CGM data (Fig. 1 A). For each subject-day, glucose measurements were ordered in descending magnitude and plotted against cumulative duration, yielding a monotonic decreasing curve that jointly encodes glucose intensity and exposure time. GPI is the maximal value k such that glucose levels ≥ k mg/dL are sustained for ≥ k minutes within a 24-hour period. This value corresponds to the intersection between the ranked glucose curve and the identity line (y = x), where glucose magnitude and duration are expressed on a common numerical scale. The horizontal and vertical projections of this intersection define the shared scalar k (Fig. 1 A). By jointly encoding severity and persistence in a unified geometric framework, GPI provides an interpretable summary of daily hyperglycemic burden. Larger GPI values reflect days characterized by sustained exposure to elevated glucose, whereas lower values indicate either shorter durations or lower magnitudes of hyperglycemia. For example, GPI for subject 85 on January 24, 2018, was 280, indicating that the subject spent at least 280 minutes with glucose levels ≥ 280 mg/dL during the 24-hour period (Fig. 1 B). This geometric construction yields a single scalar measure that integrates intensity and persistence of hyperglycemia. GPI reflects treatment effects during randomized closed-loop therapy The International Diabetes Closed Loop (iDCL) study was a 26-week, multicenter, randomized controlled trial involving 168 participants with type 1 diabetes assigned in a 2:1 ratio to closed-loop insulin delivery or sensor-augmented pump therapy. In that trial, the primary efficacy endpoint was the percentage of time that glucose levels were within the target range of 70–180 mg/dL (time-in-range, TIR), with closed-loop therapy demonstrating a significant improvement in TIR 12 . To determine whether GPI reflects clinically relevant treatment effects in a randomized controlled setting, we analyzed GPI using the same CGM dataset from the iDCL study. Specifically, we examined (1) whether GPI differs between treatment groups over 26 weeks of follow-up and (2) how GPI relates to TIR, the prespecified primary endpoint of the parent trial. At baseline, mean daily GPI values were similar between the closed-loop and control groups. After randomization, a consistent separation between groups became evident. In participants assigned to closed-loop therapy, mean GPI declined during the first post-randomization interval and remained lower than baseline across all subsequent 4-week periods through week 26. In contrast, the control group showed minimal change over time, with mean GPI remaining near baseline levels (Fig. 2 A). The between-group difference was maintained across the entire 26-week period without attenuation. Variability, reflected by standard error bars, remained relatively stable in both groups over time, suggesting that the observed difference was not driven by increasing dispersion. After adjustment for baseline GPI, mean post-randomization GPI was significantly lower in the closed-loop group than in the control group ( P = 0.0001). Because higher GPI values indicate longer duration at higher glucose levels, the reduction observed in the closed-loop group reflects decreased persistence of hyperglycemia during treatment. Given that TIR was the primary endpoint of the parent randomized trial, we examined daily GPI values stratified by 5–percentage-point TIR categories during weeks 1–26. Median GPI decreased progressively across higher TIR strata, consistent with an inverse association between time spent within 70–180 mg/dL and glycemic persistence. However, substantial dispersion in GPI was observed within individual TIR categories. Even among days with similar TIR percentages, GPI values spanned a broad range, indicating heterogeneity in the magnitude and duration of hyperglycemia not fully characterized by TIR alone. To illustrate this distinction, CGM traces from two participants with identical TIR values (68%) but markedly different GPI values are shown (Fig. 2 C–D). Subject 164 exhibited prolonged hyperglycemia with sustained glucose levels exceeding 300 mg/dL, resulting in a GPI of 311. In contrast, Subject 20 had shorter and lower-amplitude excursions, with glucose rarely exceeding approximately 220 mg/dL, yielding a GPI of 155. Despite identical TIR, the 156-point difference in GPI (311 vs. 155) reflects substantially greater persistence of severe hyperglycemia in Subject 164, corresponding to 156 additional minutes during which very high glucose levels were sustained. GPI in relation to HbA1c and daily mean glucose To examine how GPI relates to established glycemic metrics, we analyzed baseline GPI in relation to HbA1c and daily mean glucose. Baseline GPI was positively correlated with baseline HbA1c across 168 participants (Pearson’s r = 0.72, P < 0.001; Fig. 3 A). Higher HbA1c values were generally associated with higher GPI values. However, substantial dispersion in GPI was observed at comparable HbA1c levels, indicating heterogeneity in glycemic persistence among individuals with similar chronic glycemic exposure. At the subject-day level (n = 2,162 baseline days), daily mean glucose was grouped into 10 mg/dL categories. Median GPI increased progressively across higher mean glucose strata (Fig. 3 B), consistent with greater persistence of hyperglycemia at higher average glucose levels. Despite this monotonic trend, wide within-bin variability in GPI was observed across all mean glucose categories. Even among days with similar mean glucose values, GPI values spanned broad ranges, indicating that mean glucose does not uniquely determine the magnitude–duration structure of hyperglycemia. To further illustrate this distinction, we examined two representative baseline subject-days with identical mean glucose (179 mg/dL) but differing GPI values (Fig. 3 C–D). Subject 78 exhibited sustained high-amplitude hyperglycemia early in the day, resulting in a GPI of 315. In contrast, Subject 143 demonstrated shorter and lower-magnitude excursions, yielding a GPI of 210 despite the same daily mean glucose. The 105-point difference in GPI reflects 105 additional minutes of sustained severe hyperglycemia in Subject 78. These examples suggest that GPI differentiates glycemic patterns that are not fully characterized by HbA1c or daily mean glucose alone. Discussion In this study, we introduce GPI as a simple, nonparametric CGM-derived metric that integrates glucose magnitude and duration into a single scalar expressed in minutes. In a randomized clinical trial dataset, GPI was responsive to treatment and correlated strongly with mean glucose and glycemic variability, yet remained non-redundant with either measure, indicating that it captures an additional dimension of glycemic burden. Conceptually, GPI reflects the persistence of elevated glucose by jointly encoding how high glucose rises and how long it remains elevated. By emphasizing sustained hyperglycemia rather than brief excursions or overall dispersion, GPI provides a concise representation of prolonged glycemic exposure. The observed relationships between GPI and established CGM summaries help position it within the existing metric framework. Mean glucose reflects overall exposure, variability measures describe dispersion, and threshold-based metrics quantify time within or outside prespecified ranges. In contrast, GPI characterizes the magnitude–duration structure of elevated glucose. By jointly encoding how high glucose rises and how long it remains elevated, GPI captures sustained hyperglycemic burden in a way that is mathematically distinct from both average exposure and dispersion-based measures. Accordingly, GPI is considered complementary to existing summaries. GPI offers several practical advantages. 1) GPI does not rely on predefined glucose cutoffs. In clinical practice and trials, metrics such as TIR or TAR require prespecified thresholds, which may differ across populations and study designs, potentially yielding different results depending on the selected cut points. Once a threshold is crossed, such metrics provide no further gradation of severity. In contrast, GPI requires no externally imposed ranges and adapts continuously to the observed glucose profile, thereby avoiding sensitivity to chosen cutoffs. 2) GPI also condenses two-dimensional information, magnitude and duration, into a single scalar. Metrics such as mean glucose or standard deviation capture only one dimension of glycemia, whereas GPI jointly encodes how high glucose rises and how long it remains elevated. This integration allows sustained hyperglycemia to be summarized in a concise and unified manner. 3) GPI is computationally simple and assumption-free. It can be derived directly from CGM data without modeling, smoothing, or distributional assumptions and is directly comparable across devices and study settings. 4) An important advantage is its intuitive interpretability. Because GPI is expressed in minutes, its meaning is immediate: for example, GPI = 120 indicates that at least 120 minutes were spent at glucose levels ≥ 120 mg/dL. This direct magnitude–duration correspondence facilitates interpretation by both specialists and non-specialists. Several limitations should be acknowledged. Analyses were based on CGM data from a single randomized clinical trial, and associations with long-term clinical outcomes were not evaluated. Accordingly, the present study focuses on defining and characterizing GPI rather than establishing its prognostic or therapeutic implications. Future studies should assess GPI in broader and more diverse populations, including different diabetes subtypes and varying degrees of dysglycemia, and examine its relationship to complications, treatment response, and longitudinal risk markers. Integration of GPI into standardized CGM reporting frameworks may further clarify its potential role in clinical assessment. In summary, GPI quantifies glycemic persistence by integrating glucose magnitude and duration into a single scalar expressed. It is threshold-free, device-independent, and can be derived directly from CGM data without modeling assumptions. Because its value corresponds directly to matched units of glucose and time, GPI is immediately interpretable, providing a concise and intuitive representation of sustained hyperglycemic burden. Declarations Conflicts of Interest The authors declare no competing interests. Funding This work was supported in part by the National Institutes of Health grant DK132065 (to R.Z.). Author Contributions R.Z. conceptualized, designed the study, analyzed the data, wrote the paper, and obtained the funding. Data Availability Statement The CGM dataset analyzed in this study is publicly available from the International Diabetes Closed Loop (DCLP3) trial repository. The Python scripts used for analysis are available from the corresponding author upon reasonable request. References Roglic G (2016) WHO Global report on diabetes: A summary. Int J Noncommunicable Dis 1(1):3–8 Zhou B, Rayner AW, Gregg EW et al (2024) Worldwide trends in diabetes prevalence and treatment from 1990 to 2022: a pooled analysis of 1108 population-representative studies with 141 million participants. Lancet 404(10467):2077–2093 Giugliano D, Ceriello A, Esposito K (2008) Glucose metabolism and hyperglycemia. Am J Clin Nutr 87(1):217S–222S Zimmet PZ, Magliano DJ, Herman WH et al (2014) Diabetes: a 21st century challenge. lancet Diabetes Endocrinol 2(1):56–64 Rodbard D (2016) Continuous Glucose Monitoring: A Review of Successes, Challenges, and Opportunities. Diabetes Technol Ther 18(Suppl 2):S3–S13. 10.1089/dia.2015.0417 Klonoff DC, Ahn D, Drincic A (2017) Continuous glucose monitoring: A review of the technology and clinical use. Diabetes Res Clin Pract 133:178–192. 10.1016/j.diabres.2017.08.005 Battelino T, Alexander CM, Amiel SA et al (2023) Continuous glucose monitoring and metrics for clinical trials: an international consensus statement. Lancet Diabetes Endocrinol 11(1):42–57. 10.1016/S2213-8587(22)00319-9 Kovatchev BP (2017) Metrics for glycaemic control - from HbA(1c) to continuous glucose monitoring. Nat Rev Endocrinol 13(7):425–436. 10.1038/nrendo.2017.3 Nguyen M, Han J, Spanakis EK et al (2020) A Review of Continuous Glucose Monitoring-Based Composite Metrics for Glycemic Control. Diabetes Technol Ther 22(8):613–622. 10.1089/dia.2019.0434 Marling CR, Shubrook JH, Vernier SJ et al (2011) Characterizing blood glucose variability using new metrics with continuous glucose monitoring data. J Diabetes Sci Technol 5(4):871–878. 10.1177/193229681100500408 Zahalka SJ, Galindo RJ, Shah VN et al (2024) Continuous Glucose Monitoring for Prediabetes: What Are the Best Metrics? J Diabetes Sci Technol 18(4):835–846. 10.1177/19322968241242487 Brown SA, Kovatchev BP, Raghinaru D et al (2019) Six-Month Randomized, Multicenter Trial of Closed-Loop Control in Type 1 Diabetes. N Engl J Med 381(18):1707–1717. 10.1056/NEJMoa1907863 Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 2 posted You are reading this latest preprint version Show more versions Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8841862","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Method Article","associatedPublications":[],"authors":[{"id":589459137,"identity":"7ac41809-6dbe-4433-8c8d-d80ff2c7bd4a","order_by":0,"name":"Ren Zhang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6UlEQVRIiWNgGAWjYFACHhBhA8QJcCEDYrSkka7lMAlazPvPHvxc8Ot8ND97ApvEz7Y7cgzszdsk8GmROXAuWXpm3+3cmT0P2CR7254ZM/AcK8OrRYKxx0Cat+d27oYbCWw3eNsOJzZI5Jjh18LMY/ybt+dc7n6glpt/2w7XN8i/IaCFjcdMmufHgdwNEglst4G2JDBI8BDQwsNjZs3bkJw748zD9t8y5w4btvGkFVvg1cJ/xvg2zx+73P725MOGb8oOy/OzH954A58WMGBsA5MNDIxsDAxsBJWDwR8MxigYBaNgFIwCBAAAr8tI9aEOUUsAAAAASUVORK5CYII=","orcid":"","institution":"","correspondingAuthor":true,"prefix":"","firstName":"Ren","middleName":"","lastName":"Zhang","suffix":""}],"badges":[],"createdAt":"2026-02-10 14:03:20","currentVersionCode":2,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8841862/v2","doiUrl":"https://doi.org/10.21203/rs.3.rs-8841862/v2","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105003312,"identity":"70b75b76-6167-4b82-9fd5-806132472f39","added_by":"auto","created_at":"2026-03-19 17:36:31","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":23632,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGeometric definition of the glycemic persistence index (GPI). \u003c/strong\u003eA) Glucose measurements from a single day are ranked in descending order and plotted against time (minutes). The dashed identity line (k) represents the index-defining condition in which the same scalar k is applied to both duration and glucose magnitude. The GPI is defined as the largest value k for which at least k minutes exhibit glucose concentrations ≥ k mg/dL, corresponding to the intersection of the ranked glucose curve with the identity line. B) Representative CGM example. Ranked 24-hour CGM glucose values (5-minute sampling) from a representative participant (Subject 85, January 24, 2018). A GPI of 280 indicates sustained hyperglycemia ≥280 mg/dL for at least 280 minutes.\u003c/p\u003e","description":"","filename":"f1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8841862/v2/9b74fc19adce71a7573589a8.jpg"},{"id":105003314,"identity":"11d120ad-c57a-4f18-b6ce-d84fb15d472b","added_by":"auto","created_at":"2026-03-19 17:36:32","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":63064,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGlycemic persistence index (GPI) captures treatment effects and sustained hyperglycemia. \u003c/strong\u003eA) Longitudinal changes in daily GPI over 26 weeks of treatment. Mean GPI at baseline and across 4-week intervals through 26 weeks in the closed-loop group (red; N = 112) and sensor-augmented pump (control) group (blue; N = 56). Points show subject-weighted means; error bars indicate standard errors. Lower GPI reflects reduced sustained hyperglycemia. During 26 weeks, mean daily GPI was significantly lower in the closed-loop group after adjustment for baseline GPI (\u003cem\u003eP\u003c/em\u003e = 0.0001, ANCOVA). B) GPI According to TIR Bands. Daily GPI values during weeks 1–26 are shown within 5–percentage-point categories of TIR (70–180 mg/dL). Boxes represent interquartile ranges, horizontal lines medians, whiskers full ranges, and circles means. Although GPI inversely correlates with TIR, substantial variability within TIR bands indicates that glycemic persistence is not fully captured by TIR. C) CGM traces from two subjects with comparable TIR (70–180 mg/dL) but marked differences in glycemic persistence. (D) Higher GPI reflects more sustained hyperglycemia in the same subjects.\u003c/p\u003e","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8841862/v2/b2ee2ba1f79ef331ccfdab5a.jpg"},{"id":105003313,"identity":"25fbaa27-b7f9-4d13-a297-b49f5a16ea12","added_by":"auto","created_at":"2026-03-19 17:36:32","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":50283,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eRelationship between baseline GPI, HbA1c, and mean glucose. \u003c/strong\u003eA) Baseline GPI correlated with HbA1c (n = 168; Pearson’s \u003cem\u003er\u003c/em\u003e = 0.72, \u003cem\u003eP\u003c/em\u003e\u0026lt; 0.001), yet substantial variability in GPI at comparable HbA1c levels indicates that hyperglycemia persistence is not fully captured by HbA1c. B) Baseline CGM data were analyzed at the subject-day level (n = 2,162 baseline days from 168 participants). Daily mean glucose values were grouped into 10 mg/dL bins. Within each bin, box plots display the distribution of daily GPI values (median, interquartile range, and whiskers representing the full range excluding outliers). Although GPI increases progressively with higher mean glucose, substantial dispersion in GPI is observed within individual mean glucose strata. (C–D) Representative baseline subject-days with identical mean glucose but different GPI values. Despite the same average glucose (179 mg/dL), Subject 78 (GPI = 315) exhibited substantially greater persistence of hyperglycemia than Subject 143 (GPI = 210).\u003c/p\u003e","description":"","filename":"f3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-8841862/v2/87fe2e0818a68f62eb978500.jpg"},{"id":105035199,"identity":"08907d24-6ae8-4b37-bc0d-38fbdd798a7a","added_by":"auto","created_at":"2026-03-20 07:25:40","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":699239,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8841862/v2/f90d9356-6adc-46e9-ae4e-219c90c361ef.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"Defining a glycemic persistence index (GPI) for continuous glucose monitoring","fulltext":[{"header":"Introduction","content":"\u003cp\u003eDiabetes and dysglycemia are highly prevalent worldwide, spanning a continuum from prediabetes to established diabetes and affecting an increasing proportion of adults\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Across this spectrum, chronic exposure to elevated glucose is a central determinant of metabolic burden and long-term risk\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eContinuous glucose monitoring (CGM) has transformed glycemic assessment by providing high-frequency measurements in daily life. These data capture rich temporal structure, including fluctuations, excursions, and prolonged periods of dysglycemia, but their volume and complexity necessitate effective summarization into concise, interpretable metrics. In both clinical practice and randomized trials, CGM-derived summaries are routinely used to evaluate glycemic control and therapeutic efficacy\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e,\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e. Commonly reported metrics include daily mean glucose, measures of glycemic variability, and threshold-based summaries such as time-in-range (TIR) \u003csup\u003e\u003cspan additionalcitationids=\"CR8 CR9 CR10\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e. For example, TIR served as the primary efficacy endpoint in the International Diabetes Closed Loop (DCLP3) randomized trial, in which closed-loop therapy significantly improved TIR compared with control\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThreshold-based metrics, however, require prespecified glucose cutoffs. These cut points may differ across populations, including type 1 diabetes, type 2 diabetes, and prediabetes, and results can vary depending on the selected thresholds. Once defined, such metrics provide no further gradation above the chosen boundary. Moreover, sustained hyperglycemia reflects not only how long glucose remains elevated but also how high it rises. Mean glucose summarizes overall exposure, and variability measures describe dispersion, but neither directly integrates magnitude and duration into a unified representation of persistence. A metric that captures glycemic persistence should therefore encode both dimensions within a single, interpretable scalar while remaining independent of arbitrary cutoffs.\u003c/p\u003e \u003cp\u003eTo address these challenges, we introduce the glycemic persistence index (GPI), a simple, nonparametric metric derived from CGM data that integrates glucose magnitude and duration into a single scalar expressed in minutes. GPI is immediately interpretable and does not rely on predefined glucose thresholds or modeling assumptions. Using CGM data from the DCLP3 randomized trial, we evaluate GPI and examine its relationship to existing CGM summaries.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eContinuous glucose monitoring data\u003c/h2\u003e \u003cp\u003eCGM data were derived from the International Diabetes Closed Loop (iDCL) randomized trial, which enrolled 168 participants with type 1 diabetes (112 assigned to closed-loop therapy and 56 to sensor-augmented pump therapy)\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. The study duration was 26 weeks. Glucose measurements were obtained using the Dexcom G6 continuous glucose monitoring system (Dexcom), which records interstitial glucose values every 5 minutes (up to 288 measurements per day). Raw timestamped CGM data were analyzed without smoothing or interpolation. Timestamps were converted from SAS datetime format to standard datetime objects. For each participant, glucose values were aligned relative to the randomization date. Baseline CGM data were defined as recordings obtained prior to randomization during the run-in period. Glucose measurements were grouped by participant and calendar date to generate daily glucose profiles. Days with fewer than 200 valid CGM readings were excluded to ensure adequate data completeness. All analyses were conducted at the subject-day level.\u003c/p\u003e \u003c/div\u003e"},{"header":"Definition of the glycemic persistence index (GPI)","content":"\u003cp\u003eFor each subject-day, CGM measurements were used to compute the GPI. Glucose values within a day were represented as a time series with corresponding timestamps and ranked in decreasing order of glucose concentration. Cumulative time was computed based on the native sampling interval, yielding the total duration (in minutes) associated with glucose values exceeding a given level.\u003c/p\u003e \u003cp\u003eLet \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({G}_{t}\\)\u003c/span\u003e\u003c/span\u003e denote the glucose concentration (mg/dL) measured at time \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(t\\)\u003c/span\u003e\u003c/span\u003e, and let \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\({\\Delta}{t}_{t}\\)\u003c/span\u003e\u003c/span\u003e denote the duration (in minutes) represented by that CGM measurement, determined by the native sampling interval. The GPI was defined as\u003cdiv id=\"Equa\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equa\" name=\"EquationSource\"\u003e\n$$\\text{G}\\text{P}\\text{I}=\\text{m}\\text{a}\\text{x}\\left\\{k\\in\\mathbb{N}:\\sum_{\\left\\{t:{G}_{t}\\gek\\right\\}}{\\Delta}{t}_{t}\\gek\\right\\}.$$\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eThis definition identifies the largest integer value \u003cem\u003ek\u003c/em\u003e for which the cumulative time spent at glucose values greater than or equal to \u003cem\u003ek\u003c/em\u003e mg/dL is at least \u003cem\u003ek\u003c/em\u003e minutes. Thus, GPI jointly encodes glucose magnitude and duration on a common numerical scale, increasing when higher glucose levels persist for longer periods. GPI can be computed by iterating over integer glucose levels k, summing CGM durations at or above k, and identifying the maximum k satisfying the persistence condition. GPI is expressed in units of time (minutes) and is evaluated over integer values of \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(k\\)\u003c/span\u003e\u003c/span\u003e, ensuring that it is independent of CGM sampling frequency and directly comparable across devices and study settings.\u003c/p\u003e"},{"header":"Statistical analysis, software, and visualization","content":"\u003cp\u003eDaily GPI values were computed for each subject-day using quality-controlled CGM data. For longitudinal treatment analyses, daily GPI values were averaged within each participant during the baseline period and during the post-randomization period (weeks 1\u0026ndash;26). Between-group comparisons of post-randomization GPI were performed using analysis of covariance (ANCOVA), with treatment group included as a fixed effect and baseline mean GPI as a covariate. For subject-day-level analyses, daily mean glucose and daily glucose variance were calculated using all available CGM measurements within each 24-hour period. Associations between daily GPI and conventional glycemic metrics were assessed using Pearson correlation coefficients. All statistical tests were two-sided, and \u003cem\u003eP\u003c/em\u003e values\u0026thinsp;\u0026lt;\u0026thinsp;0.05 were considered statistically significant. Data processing and statistical analyses were performed using Python 3 (statsmodels). Data manipulation was conducted using standard scientific computing libraries, and figures were generated using matplotlib.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eGeometric definition of the glycemic persistence index\u003c/h2\u003e \u003cp\u003eGeometrically, GPI is defined from ranked daily CGM data (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA). For each subject-day, glucose measurements were ordered in descending magnitude and plotted against cumulative duration, yielding a monotonic decreasing curve that jointly encodes glucose intensity and exposure time. GPI is the maximal value k such that glucose levels\u0026thinsp;\u0026ge;\u0026thinsp;k mg/dL are sustained for \u0026ge;\u0026thinsp;k minutes within a 24-hour period. This value corresponds to the intersection between the ranked glucose curve and the identity line (y\u0026thinsp;=\u0026thinsp;x), where glucose magnitude and duration are expressed on a common numerical scale. The horizontal and vertical projections of this intersection define the shared scalar k (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eA). By jointly encoding severity and persistence in a unified geometric framework, GPI provides an interpretable summary of daily hyperglycemic burden. Larger GPI values reflect days characterized by sustained exposure to elevated glucose, whereas lower values indicate either shorter durations or lower magnitudes of hyperglycemia. For example, GPI for subject 85 on January 24, 2018, was 280, indicating that the subject spent at least 280 minutes with glucose levels\u0026thinsp;\u0026ge;\u0026thinsp;280 mg/dL during the 24-hour period (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eB). This geometric construction yields a single scalar measure that integrates intensity and persistence of hyperglycemia.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eGPI reflects treatment effects during randomized closed-loop therapy\u003c/h2\u003e \u003cp\u003eThe International Diabetes Closed Loop (iDCL) study was a 26-week, multicenter, randomized controlled trial involving 168 participants with type 1 diabetes assigned in a 2:1 ratio to closed-loop insulin delivery or sensor-augmented pump therapy. In that trial, the primary efficacy endpoint was the percentage of time that glucose levels were within the target range of 70\u0026ndash;180 mg/dL (time-in-range, TIR), with closed-loop therapy demonstrating a significant improvement in TIR \u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e. To determine whether GPI reflects clinically relevant treatment effects in a randomized controlled setting, we analyzed GPI using the same CGM dataset from the iDCL study. Specifically, we examined (1) whether GPI differs between treatment groups over 26 weeks of follow-up and (2) how GPI relates to TIR, the prespecified primary endpoint of the parent trial.\u003c/p\u003e \u003cp\u003eAt baseline, mean daily GPI values were similar between the closed-loop and control groups. After randomization, a consistent separation between groups became evident. In participants assigned to closed-loop therapy, mean GPI declined during the first post-randomization interval and remained lower than baseline across all subsequent 4-week periods through week 26. In contrast, the control group showed minimal change over time, with mean GPI remaining near baseline levels (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eA). The between-group difference was maintained across the entire 26-week period without attenuation. Variability, reflected by standard error bars, remained relatively stable in both groups over time, suggesting that the observed difference was not driven by increasing dispersion. After adjustment for baseline GPI, mean post-randomization GPI was significantly lower in the closed-loop group than in the control group (\u003cem\u003eP\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.0001). Because higher GPI values indicate longer duration at higher glucose levels, the reduction observed in the closed-loop group reflects decreased persistence of hyperglycemia during treatment.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eGiven that TIR was the primary endpoint of the parent randomized trial, we examined daily GPI values stratified by 5\u0026ndash;percentage-point TIR categories during weeks 1\u0026ndash;26. Median GPI decreased progressively across higher TIR strata, consistent with an inverse association between time spent within 70\u0026ndash;180 mg/dL and glycemic persistence. However, substantial dispersion in GPI was observed within individual TIR categories. Even among days with similar TIR percentages, GPI values spanned a broad range, indicating heterogeneity in the magnitude and duration of hyperglycemia not fully characterized by TIR alone. To illustrate this distinction, CGM traces from two participants with identical TIR values (68%) but markedly different GPI values are shown (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eC\u0026ndash;D). Subject 164 exhibited prolonged hyperglycemia with sustained glucose levels exceeding 300 mg/dL, resulting in a GPI of 311. In contrast, Subject 20 had shorter and lower-amplitude excursions, with glucose rarely exceeding approximately 220 mg/dL, yielding a GPI of 155. Despite identical TIR, the 156-point difference in GPI (311 vs. 155) reflects substantially greater persistence of severe hyperglycemia in Subject 164, corresponding to 156 additional minutes during which very high glucose levels were sustained.\u003c/p\u003e \u003c/div\u003e"},{"header":"GPI in relation to HbA1c and daily mean glucose","content":"\u003cp\u003eTo examine how GPI relates to established glycemic metrics, we analyzed baseline GPI in relation to HbA1c and daily mean glucose. Baseline GPI was positively correlated with baseline HbA1c across 168 participants (Pearson\u0026rsquo;s \u003cem\u003er\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.72, \u003cem\u003eP\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001; Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eA). Higher HbA1c values were generally associated with higher GPI values. However, substantial dispersion in GPI was observed at comparable HbA1c levels, indicating heterogeneity in glycemic persistence among individuals with similar chronic glycemic exposure. At the subject-day level (n\u0026thinsp;=\u0026thinsp;2,162 baseline days), daily mean glucose was grouped into 10 mg/dL categories. Median GPI increased progressively across higher mean glucose strata (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eB), consistent with greater persistence of hyperglycemia at higher average glucose levels. Despite this monotonic trend, wide within-bin variability in GPI was observed across all mean glucose categories. Even among days with similar mean glucose values, GPI values spanned broad ranges, indicating that mean glucose does not uniquely determine the magnitude\u0026ndash;duration structure of hyperglycemia. To further illustrate this distinction, we examined two representative baseline subject-days with identical mean glucose (179 mg/dL) but differing GPI values (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eC\u0026ndash;D). Subject 78 exhibited sustained high-amplitude hyperglycemia early in the day, resulting in a GPI of 315. In contrast, Subject 143 demonstrated shorter and lower-magnitude excursions, yielding a GPI of 210 despite the same daily mean glucose. The 105-point difference in GPI reflects 105 additional minutes of sustained severe hyperglycemia in Subject 78. These examples suggest that GPI differentiates glycemic patterns that are not fully characterized by HbA1c or daily mean glucose alone.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, we introduce GPI as a simple, nonparametric CGM-derived metric that integrates glucose magnitude and duration into a single scalar expressed in minutes. In a randomized clinical trial dataset, GPI was responsive to treatment and correlated strongly with mean glucose and glycemic variability, yet remained non-redundant with either measure, indicating that it captures an additional dimension of glycemic burden. Conceptually, GPI reflects the persistence of elevated glucose by jointly encoding how high glucose rises and how long it remains elevated. By emphasizing sustained hyperglycemia rather than brief excursions or overall dispersion, GPI provides a concise representation of prolonged glycemic exposure.\u003c/p\u003e \u003cp\u003eThe observed relationships between GPI and established CGM summaries help position it within the existing metric framework. Mean glucose reflects overall exposure, variability measures describe dispersion, and threshold-based metrics quantify time within or outside prespecified ranges. In contrast, GPI characterizes the magnitude\u0026ndash;duration structure of elevated glucose. By jointly encoding how high glucose rises and how long it remains elevated, GPI captures sustained hyperglycemic burden in a way that is mathematically distinct from both average exposure and dispersion-based measures. Accordingly, GPI is considered complementary to existing summaries.\u003c/p\u003e \u003cp\u003eGPI offers several practical advantages. 1) GPI does not rely on predefined glucose cutoffs. In clinical practice and trials, metrics such as TIR or TAR require prespecified thresholds, which may differ across populations and study designs, potentially yielding different results depending on the selected cut points. Once a threshold is crossed, such metrics provide no further gradation of severity. In contrast, GPI requires no externally imposed ranges and adapts continuously to the observed glucose profile, thereby avoiding sensitivity to chosen cutoffs. 2) GPI also condenses two-dimensional information, magnitude and duration, into a single scalar. Metrics such as mean glucose or standard deviation capture only one dimension of glycemia, whereas GPI jointly encodes how high glucose rises and how long it remains elevated. This integration allows sustained hyperglycemia to be summarized in a concise and unified manner. 3) GPI is computationally simple and assumption-free. It can be derived directly from CGM data without modeling, smoothing, or distributional assumptions and is directly comparable across devices and study settings. 4) An important advantage is its intuitive interpretability. Because GPI is expressed in minutes, its meaning is immediate: for example, GPI\u0026thinsp;=\u0026thinsp;120 indicates that at least 120 minutes were spent at glucose levels\u0026thinsp;\u0026ge;\u0026thinsp;120 mg/dL. This direct magnitude\u0026ndash;duration correspondence facilitates interpretation by both specialists and non-specialists.\u003c/p\u003e \u003cp\u003eSeveral limitations should be acknowledged. Analyses were based on CGM data from a single randomized clinical trial, and associations with long-term clinical outcomes were not evaluated. Accordingly, the present study focuses on defining and characterizing GPI rather than establishing its prognostic or therapeutic implications. Future studies should assess GPI in broader and more diverse populations, including different diabetes subtypes and varying degrees of dysglycemia, and examine its relationship to complications, treatment response, and longitudinal risk markers. Integration of GPI into standardized CGM reporting frameworks may further clarify its potential role in clinical assessment.\u003c/p\u003e \u003cp\u003eIn summary, GPI quantifies glycemic persistence by integrating glucose magnitude and duration into a single scalar expressed. It is threshold-free, device-independent, and can be derived directly from CGM data without modeling assumptions. Because its value corresponds directly to matched units of glucose and time, GPI is immediately interpretable, providing a concise and intuitive representation of sustained hyperglycemic burden.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eConflicts of Interest\u003c/h2\u003e \u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding\u003c/h2\u003e \u003cp\u003eThis work was supported in part by the National Institutes of Health grant DK132065 (to R.Z.).\u003c/p\u003e\u003ch2\u003eAuthor Contributions\u003c/h2\u003e \u003cp\u003eR.Z. conceptualized, designed the study, analyzed the data, wrote the paper, and obtained the funding.\u003c/p\u003e\u003ch2\u003eData Availability Statement\u003c/h2\u003e \u003cp\u003eThe CGM dataset analyzed in this study is publicly available from the International Diabetes Closed Loop (DCLP3) trial repository. The Python scripts used for analysis are available from the corresponding author upon reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eRoglic G (2016) WHO Global report on diabetes: A summary. Int J Noncommunicable Dis 1(1):3\u0026ndash;8\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou B, Rayner AW, Gregg EW et al (2024) Worldwide trends in diabetes prevalence and treatment from 1990 to 2022: a pooled analysis of 1108 population-representative studies with 141 million participants. Lancet 404(10467):2077\u0026ndash;2093\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGiugliano D, Ceriello A, Esposito K (2008) Glucose metabolism and hyperglycemia. Am J Clin Nutr 87(1):217S\u0026ndash;222S\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZimmet PZ, Magliano DJ, Herman WH et al (2014) Diabetes: a 21st century challenge. lancet Diabetes Endocrinol 2(1):56\u0026ndash;64\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRodbard D (2016) Continuous Glucose Monitoring: A Review of Successes, Challenges, and Opportunities. Diabetes Technol Ther 18(Suppl 2):S3\u0026ndash;S13. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1089/dia.2015.0417\u003c/span\u003e\u003cspan address=\"10.1089/dia.2015.0417\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKlonoff DC, Ahn D, Drincic A (2017) Continuous glucose monitoring: A review of the technology and clinical use. Diabetes Res Clin Pract 133:178\u0026ndash;192. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.diabres.2017.08.005\u003c/span\u003e\u003cspan address=\"10.1016/j.diabres.2017.08.005\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBattelino T, Alexander CM, Amiel SA et al (2023) Continuous glucose monitoring and metrics for clinical trials: an international consensus statement. Lancet Diabetes Endocrinol 11(1):42\u0026ndash;57. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S2213-8587(22)00319-9\u003c/span\u003e\u003cspan address=\"10.1016/S2213-8587(22)00319-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKovatchev BP (2017) Metrics for glycaemic control - from HbA(1c) to continuous glucose monitoring. Nat Rev Endocrinol 13(7):425\u0026ndash;436. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/nrendo.2017.3\u003c/span\u003e\u003cspan address=\"10.1038/nrendo.2017.3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNguyen M, Han J, Spanakis EK et al (2020) A Review of Continuous Glucose Monitoring-Based Composite Metrics for Glycemic Control. Diabetes Technol Ther 22(8):613\u0026ndash;622. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1089/dia.2019.0434\u003c/span\u003e\u003cspan address=\"10.1089/dia.2019.0434\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarling CR, Shubrook JH, Vernier SJ et al (2011) Characterizing blood glucose variability using new metrics with continuous glucose monitoring data. J Diabetes Sci Technol 5(4):871\u0026ndash;878. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/193229681100500408\u003c/span\u003e\u003cspan address=\"10.1177/193229681100500408\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZahalka SJ, Galindo RJ, Shah VN et al (2024) Continuous Glucose Monitoring for Prediabetes: What Are the Best Metrics? J Diabetes Sci Technol 18(4):835\u0026ndash;846. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1177/19322968241242487\u003c/span\u003e\u003cspan address=\"10.1177/19322968241242487\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrown SA, Kovatchev BP, Raghinaru D et al (2019) Six-Month Randomized, Multicenter Trial of Closed-Loop Control in Type 1 Diabetes. N Engl J Med 381(18):1707\u0026ndash;1717. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1056/NEJMoa1907863\u003c/span\u003e\u003cspan address=\"10.1056/NEJMoa1907863\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Continuous glucose monitoring, CGM, glycemic persistence index, GPI, hyperglycemia","lastPublishedDoi":"10.21203/rs.3.rs-8841862/v2","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8841862/v2","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eSustained hyperglycemia is the core driver of diabetes complications. Continuous glucose monitoring (CGM) provides rich temporal glucose data, but effective summarization that integrates both magnitude and duration of sustained hyperglycemia into a single and paramter-free scalar remains challenging. We introduce the glycemic persistence index (GPI), a simple, threshold-free CGM-derived metric defined as the largest integer k such that at least k minutes are spent at glucose levels\u0026thinsp;\u0026ge;\u0026thinsp;k mg/dL within a day. For example, a GPI of 125 means that glucose levels were \u0026ge;\u0026thinsp;125 mg/dL for at least 125 minutes during that day. Geometrically, after ranking glucose values in decreasing order, GPI is given by the intersection at which glucose level and cumulative duration take the same value. Using CGM data from the International Diabetes Closed Loop randomized trial (168 adults with type 1 diabetes), GPI sensitively captured closed-loop treatment benefit with significantly lower values in the closed-loop group compared with sensor-augmented pump therapy (P\u0026thinsp;=\u0026thinsp;0.0001). GPI correlated with HbA1c and daily mean glucose, while distinguishing patterns of sustained hyperglycemia among individuals with similar conventional metrics. As a simple, device-independent, and threshold-free scalar, GPI quantifies hyperglycemia by jointly capturing its magnitude and duration, enabling consistent and intuitive glycemic profiling accessible to both specialists and non-specialists.\u003c/p\u003e","manuscriptTitle":"Defining a glycemic persistence index (GPI) for continuous glucose monitoring","msid":"","msnumber":"","nonDraftVersions":[{"code":2,"date":"2026-03-19 17:36:27","doi":"10.21203/rs.3.rs-8841862/v2","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}},{"code":1,"date":"2026-02-11 08:30:01","doi":"10.21203/rs.3.rs-8841862/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"52dc5582-0998-4be6-b19a-f892c2af2831","owner":[],"postedDate":"March 19th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-02-11T08:30:01+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-19 17:36:27","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v2","identity":"rs-8841862","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8841862","identity":"rs-8841862","version":["v2"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}