The Relationship Between Continuous Glucose Monitoring and OGTT in Youth and Young Adults With Cystic Fibrosis

Abstract

Context

Early glucose abnormalities in people with cystic fibrosis (PwCF) are commonly detected by continuous glucose monitoring (CGM). Relationships between these CGM abnormalities and oral glucose tolerance testing (OGTT) in PwCF have not been fully characterized.

Objective

This work aimed to determine the relationship between CGM and common OGTT-derived estimates of β-cell function, including C-peptide index and oral disposition index (oDI) and to explore whether CGM can be used to screen for OGTT-defined prediabetes and cystic fibrosis–related diabetes (CFRD).

Methods

PwCF not on insulin and healthy controls aged 6 to 25 years were enrolled in a prospective study collecting OGTT and CGM. A subset underwent frequently sampled OGTTs (fsOGTT) with 7-point glucose, insulin, and C-peptide measurements. Pearson correlation coefficient was used to test the association between select CGM and fsOGTT measures. Receiver operating curve (ROC) analysis was applied to CGM variables to determine the cutoff optimizing sensitivity and specificity for detecting prediabetes and CFRD.

Results

A total of 120 participants (controls = 35, CF = 85), including 69 with fsOGTTs, were included. CGM coefficient of variation correlated inversely with C-peptide index (Cpeptide₃₀-Cpeptide₀/Glucose₃₀-Glucose₀) (r = –0.45, P < .001) and oDIcpeptide (C-peptide index)(1/cpep₀) (r = –0.48, P < .0001). In PwCF, CGM variables had ROC – areas under the curve ranging from 0.43 to 0.57 for prediabetes and 0.47 to 0.6 for CFRD.

Conclusion

Greater glycemic variability on CGM correlated with reduced β-cell function. However, CGM performed poorly at discriminating individuals with and without OGTT-defined CFRD and prediabetes. Prospective studies are now needed to determine how well the different tests predict clinically relevant nonglycemic outcomes in PwCF.

Cystic fibrosis–related diabetes (CFRD) is prevalent in the CF population (1); up to 20% of adolescents and 30% to 50% of adults with CF develop CFRD (1). However, the onset of CFRD is often insidious and preceded by a long period of progressive insulin insufficiency that may itself have clinical effects. Dysglycemia progression in CF, from normal glucose tolerance (NGT) to CFRD, is influenced by multiple factors—genetic predisposition (2), early structural islet abnormalities (3), collateral β-cell damage from pancreatic exocrine disease (1), and inflammation (4, 5)—although the complete pathogenesis of CFRD remains incompletely understood. Given the high prevalence of CFRD, annual screening with an oral glucose tolerance test (OGTT) is recommended starting at age 10 years. CFRD is defined by a 2-hour glucose (2hG) greater than or equal to 200 mg/dL (11.1 mmol/L) on OGTT, and prediabetes or impaired glucose tolerance is defined by a 2hG on OGTT of greater than or equal to 140 mg/dL (7.8mmol/L). Glucose elevations greater than or equal to 200 mg/dL (11.1 mmol/L) on intermediate OGTT glucose time points are designated “indeterminate” and considered a precursor to prediabetes and diabetes (6).

These OGTT-based criteria for diagnosing diabetes in the CF population have been extrapolated from populations at risk for type 2 diabetes (T2D) (7) and were developed to identify individuals at risk for microvascular complications of T2D. In the CF population, these criteria may not be sensitive enough to detect early dysglycemia or insulin insufficiency associated with clinically relevant declines in body composition, muscle mass, and/or lung function. Furthermore, adherence to annual OGTTs as recommended by CF screening guidelines is poor (8), and continuous glucose monitoring (CGM), a commonly used tool for monitoring of glycemic control in patients with known diabetes, has been increasingly used for detection of early glucose abnormalities in individuals with CF (9-12). These CGM-identified abnormalities have been associated with worse lung function and weight (11-14), and small studies have suggested that early insulin intervention for CGM-identified dysglycemia may improve clinical outcomes (15, 16). However, how glucose abnormalities on CGM relate to OGTT values and how CGM variables perform as screening measures for prediabetes or diabetes have not been fully explored.

The objectives of this study were 1) to examine the associations between CGM variables and common OGTT-derived estimates of β-cell function, and 2) to explore how well specific CGM variables predict CFRD and prediabetes as defined by current OGTT criteria.

Materials and Methods

Study Population

Participants with and without CF aged 6 to 25 years were enrolled as part of GlycEmic Monitoring in CF (GEM-CF, NCT02211235), a study of early glucose abnormalities in youth with CF. Participants with CF were recruited from pulmonary and diabetes clinics at Children’s Hospital Colorado. Inclusion criteria for participants with CF included a diagnosis of CF by newborn screen, sweat chloride testing, or genetic testing. CF patients with known glucose abnormalities along the entire glycemic spectrum were included. Exclusion criteria for participants with CF included known type 1 or T2D, use of medications affecting glucose (eg, insulin, systemic steroids) in the prior 3 months, hospitalization in the prior 6 weeks, or pregnancy. Youth without CF were identified using recruitment flyers and emails at the University of Colorado Anschutz Medical Campus. Exclusion criteria for healthy controls (HCs) included diagnoses of diabetes or prediabetes, overweight or body mass index (BMI) greater than or equal to the 85th percentile as defined by the Centers for Disease Control and Prevention BMI percentile growth charts in youth (17), chronic disease, acute illness, or pregnancy. This study was approved by the Colorado Multiple Institutional Review Board (Aurora, Colorado) and appropriate consent and assent were obtained.

Study Visits

Study visits took place in the Clinical Translational Research Center at Children’s Hospital Colorado between October 2014 and May 2018. Height and weight were measured, BMI z scores calculated (17), and physical exam and pubertal Tanner staging were completed by a single pediatric endocrinologist (C.L.C.). CF genotype, presence of pancreatic insufficiency, gastrostomy tube feedings, and lung function data from the most recent pulmonary clinic visit were collected via review of electronic medical records.

Laboratory Procedures

Participants arrived to the outpatient research center between 8 a.m. and 10 a.m. after a minimum of 8 hours of fasting.

Oral glucose tolerance testing

All participants underwent an OGTT and collection of glycated hemoglobin A_1c (HbA_1c). Glucola was administered at a dose of 1.75 g/kg (maximum 75 g). From October 2014 through September 2015, participants underwent an OGTT with collection of plasma glucose at 0, 60, and 120 minutes. The study protocol was subsequently amended, and participants studied between October 2015 through May 2018 underwent a frequently sampled OGTT (fsOGTT) with measurements of glucose, insulin, and C peptide at 0, 10, 20, 30, 60, 90, and 120 minutes.

HbA_1c was measured on a DCA Vantage Analyzer (Siemens), an instrument aligned to Diabetes Control and Complications Trial standards, with an interday coefficient of variation (CV) of 2.8%. Additional laboratory assays were performed by the Clinical Translational Research Center core laboratories. Plasma glucose was analyzed by hexokinase with an intra-assay variability of 0.67%, an interassay variability of 1.44%, and a sensitivity of 10 mg/dL (Beckman Coulter); insulin was analyzed by chemiluminescent immunoassay with an intra-assay variability of 1.60%, an interassay variability of 2.80%, and a sensitivity of 0.5 μIU/mL (Beckman Coulter); and C peptide was analyzed by enzyme-linked immunosorbent assay with an intra-assay variability of 4.80%, an interassay variability of 4.80%, and a sensitivity of 0.1 ng/mL (Mercodia).

Continuous glucose monitoring

All participants wore a blinded iPro2 CGM (Medtronic Inc) for up to 7 days (minimum 3). They were provided a glucometer (OneTouch; LifeScan) and trained to collect capillary blood glucoses 4 times daily—before meals and at bedtime, as the iPro2 CGM requires a minimum of 3 capillary blood glucose readings daily for calibration of sensor data. Participants were asked to maintain a food log during the period of CGM wear. The CGM was placed within 1 week of the OGTT (the same day as the OGTT in HCs, to minimize the burden associated with coming in for multiple study visits; and 1 week before the OGTT in participants with CF, for the purpose of collecting CGM data before venipuncture for laboratory values related to the primary analysis, as previously described) (18). The following CGM variables were analyzed (19): average, minimum, and maximum sensor glucose; percentage of time greater than 140 mg/dL (7.8 mmol/L), percentage of time greater than 180 mg/dL (>10.0 mmol/L), percentage of time greater than 200 mg/dL (11.1 mmol/L), percentage of time 70 to 140 mg/dL (3.9-7.8 mmol/L), and measures of glycemic variability (SD, CV, and mean amplitude of glycemic variability [MAGE]). CGM thresholds for hyperglycemia were determined based on historical references in CF (11, 20-22) and international consensus metrics for CGM analysis and reporting (23).

HbA_1c and OGTT results were used to exclude prediabetes and diabetes in HC participants and OGTT results were used to classify participants with CF into categories based on glycemic status: NGT (fasting plasma glucose [FPG] < 100 mg/dL [5.6 mmol/L], 1hG < 200 mg/dL [11.1 mmol/L], and 2hG < 140 mg/dL [7.8 mmol/L]), abnormal glucose tolerance (AGT, FPG 100-125 mg/dL [5.6-6.9 mmol/L], 1hG ≥ 200 mg/dL [11.1 mmol/L], and/or 2hG 140-199 mg/dL [7.8-11.0 mmol/L]), and CFRD (FPG ≥ 126 mg/dL [7.0 mmol/L] and/or 2hG ≥ 200 mg/dL [11.1 mmol/L]).

Calculations

OGTT-derived estimates of β-cell function were determined with the following equations: insulinogenic index (IGI) = (insulin₃₀ – insulin₀)/(glucose₃₀ – glucose₀) and C-peptide index (IGI_Cpep) = (C-peptide30 – C-peptide₀)/(glucose₃₀ – glucose₀) (24); integrated AUC (iAUC) for glucose, insulin, and C peptide calculated as the AUC above the fasting value using the trapezoidal method; and total insulin secretion as iAUC-insulin/iAUC-glucose and iAUC-cpeptide/iAUC-glucose over the entirety of the OGTT curve. Insulin sensitivity was estimated as 1/insulin₀ and modeled with the Matsuda equation (10 000/√[fasting glucose*fasting insulin*mean glucose*mean insulin]) (25). oDI, a measure of β-cell function accounting for insulin sensitivity, was calculated with insulin as oDI = IGI*1/insulin₀ (26) and with cpeptide oDIcpep = IGI_cpep*1/cpeptide₀.

Insulin clearance was estimated as cpeptide₀/insulin₀ (representing fasting insulin clearance), iAUC-cpeptide_0-30/iAUC-insulin_0-30 (representing early-phase insulin clearance), and iAUC-cpeptide/iAUC-insulin over the entire OGTT (27).

Analysis Plan

The distributions of all variables were examined before analysis. Descriptive statistics were calculated by cohort. Groups were compared using analysis of variance or the Wilcoxon test for continuous variables, and the chi-square or Fisher exact test for categorical variables.

In individuals who had undergone an fsOGTT, Pearson correlation coefficient was used to test the association between CGM variables and fsOGTT outcomes.

Receiver operating curve (ROC) analysis was used to examine the CGM variables associated with prediabetes/diabetes and determine the cut point that optimized the Youden Index (28) (sensitivity + specificity–1) for CFRD and CF prediabetes. The distribution of 2hG above and below the cut point of CGM variables identified to optimize sensitivity in order to minimize false negatives was visualized using box plots. We also identified cut points that optimized sensitivity, by maintaining the cost of a false negative at 30 times that of a false positive.

All analyses were conducted using R version 4.0 (R Core Team).

Results

There were 120 participants included in the total analysis and 69 underwent an fsOGTT. Table 1 shows descriptive statistics by glycemic grouping for all participants. There were no differences in age, sex, race, Tanner stage, forced expiratory volume in 1 second, or forced vital capacity among glycemic groups. HbA_1c was higher in all CF groups compared to HCs and also higher in CF AGT vs CF NGT. Participants wore CGM for a mean ± SD of 5.2 ± 1 days and obtained 4.0 ± 1 glucometer readings per day for calibration. CGM outcomes by glycemic category are presented in Table 1. Notably, with the exception of minimum glucose, participants with CF differed from HCs for most CGM measures. CF NGT did not differ statistically from HCs for average sensor glucose nor percentage of time greater than 140 mg/dL (7.8 mmol/L).

Table 1.

Demographics and clinical data for overall group (n = 120)

	Healthy controls	CF NGT	CF AGT	CFRD	P
	(n = 35)	(n = 27)	(n = 45)	(n = 13)
Age, y	14.3 ± 4.0	13.1 ± 3.4	13.7 ± 3.8	13.4 ± 2.8	.61
Sex, female	19 (54%)	16 (59%)	23 (51%)	9 (69%)	.68
Race					.26
White	29 (83%)	25 (93%)	40 (89%)	10 (77%)
Black/African-American	1 (3%)	1 (4%)	0 (0%)	0 (0%)
Asian	2 (6%)	0 (0%)	0 (0%)	0 (0%)
Hispanic	2 (6%)	1 (4%)	5 (11%)	2 (15%)
Other/Multiple	1 (3%)	0 (0%)	0 (0%)	1 (8%)
Tanner stage					.61
I	11 (31%)	10 (37%)	18 (40%)	1 (8%)
II	4 (11%)	3 (11%)	2 (4%)	2 (15%)
III	5 (14%)	4 (15%)	4 (9%)	3 (23%)
IV	1 (3%)	1 (4%)	2 (4%)	1 (8%)
V	14 (40%)	9 (33%)	19 (42%)	6 (46%)
HbA1c %, (mmol/mol)	5.2 ± 0.2 (33.3 ± 2.2)	5.4 ± 0.3 (35.5 ± 3.3)	5.6 ± 0.3 (37.7 ± 3.3)	5.6 ± 0.3 (37.7 ± 3.3)	< .001a
FEV1%	–	92.4 ± 15.4	90.2 ± 16.2	93.9 ± 12.9	0.70
FVC%	–	100.2 ± 13.3	98.8 ± 15.4	102.6 ± 11.5	0.69
Gtube feeds	0 (0%)	5 (19%)	7 (16%)	4 (31%)	.44
Pancreatic insufficiency	0 (0%)	24 (89%)	43 (96%)	13 (100%)	.44
BMI z score	–0.22 ± 0.68	–0.28 ± 0.72	0.21 ± 0.82	–0.27 ± 0.77	.03
Mutation class					.10
Class I-III	–	20 (74%)	39 (93%)	12 (92%)
Class IV-V	–	5 (19%)	1 (2%)	0 (0%)
Unidentified	–	2 (7%)	2 (5%)	1 (8%)
Fasting glucose, mg/dL (mmol/L)	89 ± 7 (4.9 ± 0.4)	88 ± 7 (4.9 ± 0.4]	95 ± 9 (5.3 ± 0.5]	100 ± 11 (5.6 ± 0.6]	< .001b
1-h glucose, mg/dL (mmol/L)	109 ± 24 (6.1 ± 1.3)	145 ± 31 (8.0 ± 1.7]	203 ± 45 (11.3 ± 2.5]	254 ± 60 (14.1 ± 3.3]	< .001c
2-h glucose, mg/dL (mmol/L)	107 ± 22 [5.9 ± 1.2)	108 ± 18 (6.0 ± 1.0)	143 ± 32 (7.9 ± 1.8)	216 ± 19 (12.0 ± 1.1)	< .001d
CGM variables
Average glucose, mg/dL (mmol/L)	102 ± 10 (5.7 ± 0.6)	109 ± 12 (6.1 ± 0.7)	114 ± 15 (6.3 ± 0.8)	116 ± 13 (6.4 ± 0.7)	< .001e
Min glucose, mg/dL (mmol/L)	67 ± 11 (3.7 ± 0.6)	61 ± 15 (3.4 ± 0.8)	65 ± 14 (3.6 ± 0.8)	66 ± 12 (3.7 ± 0.7)	.41
Max glucose, mg/dL (mmol/L)	150 ± 23 (8.3 ± 1.3)	204 ± 35 (11.3 ± 1.9)	218 ± 60 (12.1 ± 3.3)	220 ± 41 (12.2 ± 2.3)	< .001f
% time > 140 mg/dL (> 7.8 mmol/L)	2.3 ± 5.0	10.3 ± 7.2	14.6 ± 15.1	15.1 ± 10.7	< .001e
% time > 180 mg/dL (> 10.0 mmol/L)	0.0 (0.0-0.0)	0.6 (0.1-1.8)	0.8 (0-4.6)	1.9 (0.4-5.3)	< .001g
% time > 200 mg/dL (> 11.1 mmol/L)	0 (0-0)	0 (0-0.7)	0.1 (0-1.9)	0.5 (0.1-1.9)	< .001g
% time = 70-140 mg/dL (3.9-7.8 mmol/L)	96.1 ± 5.1	85.7 ± 7.6	82.9 ± 15.0	82.4 ± 9.8	< .001f
SD, mg/dL (mmol/L)	12.9 ± 3.1 (0.7 ± 0.2)	22.0 ± 5.4 (1.2 ± 0.3)	23.8 ± 10.5 (1.3 ± 0.6)	24.3 ± 7.8 (1.3 ± 0.4)	< .001f
Coefficient of variation, %	12.7 ± 2.4	20.4 ± 5.1	20.6 ± 7.3	20.9 ± 5.7	< .001f
MAGE, mg/dL (mmol/L)	61.6 ± 37.7 (3.4 ± 2.1)	72.5 ± 34.1 (4.0 ± 1.9)	68.0 ± 33.2 (3.8 ± 1.8)	82.4 ± 36.9 (4.6 ± 2.0)	.34

Statistics given are No. (%), mean ± SD, or median (25th percentile-75th percentile).

Abbreviations: AGT, abnormal glucose tolerance; BMI, body mass index; CF, cystic fibrosis; CFRD, cystic fibrosis–related diabetes; CFTR, cystic fibrosis transmembrane conductance regulator; CGM, continuous glucose monitoring; Gtube, gastrostomy tube; HbA_1c, glycated hemoglobin A_1c; FEV1, forced expiratory volume in 1 second; FVC, forced vital capacity; Max, maximum; Min, minimum; MAGE, mean amplitude of glycemic excursions; NGT, normal glucose tolerance.

^a All CF different from controls; CF AGT different from CF NGT.

^b CF AGT and CFRD different from CF NGT and controls.

^c All pairwise comparisons differ.

^d All pairwise comparisons differ except CF NGT vs controls.

^e CF AGT and CFRD different from controls.

^f All CF different from controls.

^g CF AGT different from controls.

Table 2 presents descriptive statistics, OGTT data, and CGM variables for the subgroup who underwent an fsOGTT (n = 69). Baseline demographics were similar among groups, with no differences in age, sex, forced expiratory volume in 1 second, forced vital capacity, nor BMI z score. Similar to the overall group, race, Tanner stage, and CF transmembrane conductance regulator genotype were no different among glycemic groups (not shown). HbA_1c was highest in the CF AGT group. IGI was reduced in all patients with CF, including those with CF NGT, compared to HCs. IAUCinsulin:iAUCglucose and iAUCcpeptide:iAUCglucose, although numerically lower in individuals with CF relative to HCs, were not statistically different among any of the groups. Insulin sensitivity by Matsuda was higher in patients with CF NGT compared to HCs, CF AGT, and CFRD. ODI was reduced in all CF participants compared to HCs; even CF patients with NGT had an oDI approximately 50% of that seen in HCs. IGI and oDI declined in the CF participants as glycemic tolerance worsened; however, there were no statistically significant differences in IGI nor oDI among the 3 groups of CF participants. There were no differences in insulin clearance estimates across the groups (data not shown), although iAUCcpep:iAUCinsulin in female participants with CF trended lower compared to male participants with CF (0.16 ± 0.05 vs 0.19 ± 0.07, P = .05).

Table 2.

Subgroup with frequently sampled oral glucose tolerance test descriptive statistics by glycemic grouping (n = 69)

	Healthy controls (n = 16)	CF NGT (n = 11)	CF AGT (n = 34)	CFRD (n = 8)	P
Age, y	13.3 ± 3.7	13.1 ± 4.0	13.1 ± 3.9	14.3 ± 3.2	.90
Sex, female	8 (50%)	7 (64%)	21 (62%)	6 (75%)	.69
HbA1c %, mmol/mol	5.3 ± 0.2 (34.4 ± 2.2)	5.6 ± 0.3 (37.7 ± 3.3)	5.7 ± 0.3 (38.8 ± 3.3)	5.6 ± 0.3 (37.7 ± 3.3)	< .001a
FEV1%	NA	93.5 ± 11.4	89.1 ± 17.3	93.5 ± 14.8	.63
FVC%	NA	100.2 ± 11.7	98.6 ± 16.2	104.6 ± 12.7	.59
Gtube feeds	0 (0%)	1 (9%)	6 (18%)	2 (25%)	.66
Pancreatic insufficiency	0 (0%)	9 (82%)	32 (94%)	8 (100%)	.28
BMI z score	–0.2 ± 0.7	–0.2 ± 0.5	0.1 ± 0.9	–0.7 ± 0.8	.20
fsOGTT variables
Fasting glucose, mg/dL (mmol/L)	92 ± 7 (5.1 ± 0.4)	89 ± 4 (4.9 ± 0.2)	96 ± 8 (5.3 ± 0.4)	103 ± 13 (5.7 ± 0.7)	.002b
1-h glucose, mg/dL (mmol/L)	111 ± 23 (6.2 ± 1.3)	147 ± 30 (8.2 ± 1.7)	209 ± 46 (11.6 ± 2.6)	273 ± 60 (15.2 ± 3.3)	< .001c
2-h glucose, mg/dL (mmol/L)	105 ± 21 (5.8 ± 1.2)	113 ± 17 (6.3 ± 0.9)	143 ± 30 (7.9 ± 1.7)	221 ± 20 (12.3 ± 1.1)	< .001c
Fasting insulin, μIU/mL (pmol/L)	4.4 ± 2.4 (26.4 ± 14.4)	2.7 ± 1.1 (16.2 ± 6.6)	3.8 ± 2.5 (22.8 ± 15)	4.6 ± 3.0 (27.6 ± 18)	.24
Matsuda	12.60 ± 6.65	22.72 ± 12.56	14.03 ± 7.36	10.19 ± 9.00	.009d
IGI insulin	1.03 ± 0.58	0.35 ± 0.26	0.28 ± 0.17	0.23 ± 0.15	< .001e
IGI C peptide	0.09 ± 0.04	0.04 ± 0.02	0.03 ± 0.02	0.03 ± 0.01	< .001e
Oral DI insulin	0.26 ± 0.13	0.13 ± 0.08	0.10 ± 0.06	0.06 ± 0.04	< .001e
Oral DI C peptide	0.08 ± 0.02	0.04 ± 0.02	0.04 ± 0.02	0.02 ± 0.01	< .001e
iAUCins:iAUCglc	2.42 ± 5.17	0.84 ± 0.69	0.76 ± 1.01	0.37 ± 0.16	.17
iAUCcpep:iAUCglc	0.34 ± 0.72	0.14 ± 0.11	0.12 ± 0.20	0.05 ± 0.02	.24
CGM variables
Average glucose, mg/dL (mmol/L)	104 ± 12 (5.8 ± 0.7)	108 ± 14 (6.0 ± 0.8)	115 ± 17 (6.4 ± 0.9)	112 ± 16 (6.2 ± 0.9)	0.17
Min glucose, mg/dL (mmol/L)	68 ± 14 (3.8 ± 0.8)	58 ± 14 (3.2 ± 0.8)	65 ± 14 (3.6 ± 0.8)	61 ± 14 (3.4 ± 0.8)	.31
Max glucose, mg/dL (mmol/L)	156 ± 25 (8.7 ± 1.4)	216 ± 36 (12.0 ± 2.0)	222 ± 63 (12.3 ± 3.5)	224 ± 51 (12.4 ± 2.8)	.002e
% time > 140 mg/dL (> 7.8 mmol/L)	3.8 ± 7.1	10.4 ± 6.4	16.2 ± 16.4	14.5 ± 12.1	.04f
% time < 60 mg/dL (< 3.3 mmol/L)	0.4 ± 0.7	2.6 ± 5.5	1.1 ± 2.2	1.3 ± 2.8	.30
% time > 180 mg/dL (> 10.0 mmol/L)	0.2 ± 0.6	1.4 ± 1.4	3.8 ± 6.0	3.9 ± 4.5	.08
% time > 200 mg/dL (> 11.1 mmol/L)	0.0 ± 0.0	0.5 ± 0.6	2.1 ± 4.2	1.9 ± 2.6	.17
% time = 70-140 mg/dL (3.9-7.8 mmol/L)	94.6 ± 6.6	84.6 ± 5.6	81.1 ± 16.0	81.5 ± 10.2	.01e
SD, mg/dL (mmol/L)	13.2 ± 3.4 (0.7 ± 0.2)	22.8 ± 3.4 (1.3 ± 0.2)	24.5 ± 10.6 (1.4 ± 0.6)	26.1 ± 9.6 (1.4 ± 0.5)	.001e
Coefficient of variation (%)	12.5 ± 2.3	21.3 ± 3.0	21.0 ± 7.00	23.0 ± 6.5	< .001e
MAGE mg/d (mmol/L)	26.6 ± 7.2 (1.5 ± 0.4)	52.7 ± 12.7 (2.9 ± 0.7)	60.1 ± 27.4 (3.3 ± 1.5)	64.8 ± 29.7 (3.6 ± 1.6)	< .001e

Statistics given are No. (%) or mean ± SD.

Abbreviations: AGT, abnormal glucose tolerance; BMI, body mass index; CF, cystic fibrosis; CFRD, cystic fibrosis–related diabetes; CFTR, cystic fibrosis transmembrane conductance regulator; CGM, continuous glucose monitoring; DI, disposition index; FEV1, forced expiratory volume in 1 second; fsOGTT, frequently sampled oral glucose tolerance test; FVC, forced vital capacity; Gtube, gastrostomy tube; HbA_1c, glycated hemoglobin A_1c; HCs, healthy controls; iAUC, incremental area under the curve; IGI, insulinogenic index; Max, maximum; Min, minimum; MAGE, mean amplitude of glycemic excursions; NA, not available; NGT, normal glucose tolerance.

^a CF AGT different from HCs.

^b CFRD different from all other groups, except CF AGT.

^c HCs no different from CF NGT; all other groups different from one another.

^d CF NGT different from HCs, CF AGT, and CFRD.

^e HCs different from all CF groups.

^f HCs different from CF AGT and CFRD.

Table 3 presents Pearson correlation coefficients between CGM and fsOGTT measures. Multiple CGM measures correlated with fsOGTT measures. Correlations were greatest for measures of glycemic variability, specifically SD, CV, and MAGE, a measure of glycemic variability accounting for excursions exceeding 1 SD of the average sensor glucose) with IGIcpeptide and oDIcpeptide. No CGM measures correlated with iAUCcpeptide:iAUCglucose nor the Matsuda index. Average sensor glucose and percentage of time less than 60 mg/dL did not correlate with IGI nor oDI. Table 3 also presents Pearson correlation coefficients between CGM and FPG, 1hG, and 2hG on OGTT. Average glucose correlated with FPG and 2hG. Other CGM measures including peak glucose and percentage of time in hyperglycemia as well as measures of glycemic variability also correlated with FPG, 1hG, and 2hG.

Table 3.

Pearson correlation coefficients between continuous glucose monitoring and frequently sampled oral glucose tolerance test measures

	IGIcpep r P	iAUCcpep:iAUCglc r P	Matsuda r P	oDIcpep r P	FPG r P	1hG r P	2hG r P
Average sensor glucose	–0.16 .22	–0.08 .55	–0.05 .69	–0.13 .30	0.22 .02	0.18 .11	0.35 < .001
Maximum sensor glucose	–0.40 .002	–0.08 .52	–0.07 .59	–0.42 < .001	0.29 .002	0.32 .003	0.44 < .001
% time > 140 mg/dL (> 7.8 mmol/L)	–0.28 .03	–0.07 .59	–0.05 .68	–0.25 < .05	0.27 .004	0.21 .05	0.36 < .001
% time < 60 mg/dL (< 3.3 mmol/L)	–0.08 .54	–0.01 0.95	0.003 .98	–0.08 .54	–0.03 .78	–0.08 .46	0.03 .73
% time > 180 mg/dL (> 10.0 mmol/L)	–0.28 .03	–0.06 .65	–0.07 .57	–0.30 .02	0.25 .01	0.22 < .05	0.39 < .001
% time > 200 mg/dL (> 11.1 mmol/L)	–0.27 .04	–0.05 .69	–0.01 .96	–0.27 .04	0.21 .03	0.17 .12	0.34 .001
% time = 70-140 mg/dL (3.9-7.8 mmol/L)	0.33 .01	0.10 .46	0.05 .63	0.31 .01	–0.30 .002	–0.20 .06	–0.38 < .001
SD	–0.41 .001	–0.11 .41	–0.04 .79	–0.43 < .001	0.29 .002	0.27 .01	0.47 < .001
Coefficient of variation	–0.45 < .001	–0.11 .40	–0.03 .80	–0.48 < .001	0.28 .003	0.27 .01	0.44 < .001
MAGE	–0.49 < .001	–0.07 .57	–0.06 .64	–0.51 < .001	–0.02 .87	0.25 .02	0.21 .03

Abbreviations: 1hG, 1-hour glucose; 2hG, 2-hour glucose; FPG, fasting plasma glucose; iAUC, incremental area under the curve; IGI, insulinogenic index; MAGE, mean amplitude of glycemic excursions; oDI, oral disposition index.

Next, we explored how well CGM variables in our CF participants could distinguish individuals with OGTT-defined diabetes or prediabetes from those with CF NGT. ROC analysis was applied to mean sensor glucose, peak glucose, percentage of time greater than 140 mg/dL (7.8 mmol/L), percentage of time greater than 180mg/dL (10.0 mmol/L), percentage of time greater than 200 mg/dL (11.1 mmol/L), SD, CV, and MAGE. These CGM measures were selected given their availability on standardized CGM glucose profiles and/or historical associations with clinically relevant outcomes in CF (11, 13). ROC AUCs ranged from 0.47 to 0.6 for detecting CFRD and 0.43 to 0.57 for detecting either abnormal glucose tolerance or CFRD (Supplementary Table 4) (29). Selected ROC curves are presented in Fig. 1. For example, percentage of time greater than 140 mg/dL (7.8 mmol/L) at an optimal threshold of 9.7% had a sensitivity of 75% and specificity of 55% for detecting CFRD (ROC AUC 0.60) (see Supplementary Table 4) (29).

Figure 1.

Select receiver operating characteristic (ROC) curves for continuous glucose monitoring–based detection of oral glucose tolerance testing–defined cystic fibrosis–related diabetes.

Last, we examined specific CGM thresholds and the 2hG distributions of individuals above and below the identified CGM thresholds that optimized sensitivity for detecting CFRD, while maintaining the cost of a false negative at 30 times that of a false positive (Fig. 2). The thresholds identified, in order not to miss an individual with 2hG greater than or equal to 200 mg/dL (11.1 mmol/L), were notably close to or within the normal range, and there was significant overlap in 2hG values between individuals who had CGM measures above and below the thresholds identified.

Figure 2.

A to D, Two-hour glucose distributions for individuals below and above select continuous glucose monitoring thresholds identified to optimize sensitivity (100%) and minimize false negatives for detecting cystic fibrosis–related diabetes.

Discussion

This is one of the first studies to examine the relationship between CGM and OGTT estimates of β-cell function and to explore CGM as a screening tool for OGTT-defined diabetes and prediabetes in CF youth and young adults, a population at high risk for diabetes. As previously demonstrated by our group and others (30, 31), abnormalities of glucose metabolism detectable by CGM and fsOGTT (including reduced IGI and oDI) are common in individuals with CF when compared to age-matched controls, even in those with CF NGT. Here, we found that specific CGM measures, CV, SD, and MAGE, correlated inversely with estimates of β-cell function, so that reduced insulin secretion manifested as greater glycemic variability on CGM. A growing body of evidence links early insulin insufficiency and dysglycemia detected by CGM (12-14) and by elevations in intermediate OGTT glucoses (11, 13, 32, 33) with pulmonary function decline and poor weight gain in CF. Small studies reporting an introduction of insulin therapy in CF individuals at the stage of early dysglycemia identified by CGM and/or OGTT, but not frank diabetes, have demonstrated improvements in weight and pulmonary function (15, 34, 35). However, other studies of early insulin introduction have not demonstrated clear benefits (36, 37). Given the early mortality seen in individuals with CF (8) and the increased mortality risk conferred by CFRD (38, 39), understanding whether earlier detection and treatment of CFRD and potentially CF prediabetes may slow pulmonary function decline and reduce mortality is critical. Frequently sampled OGTTs and CGM allow more precise characterization of early insulin deficits and dysglycemia, and the findings reported here provide insight into the meaning behind some commonly measured CGM variables and their relationships with different components of glucose metabolism in PwCF.

Lessons From the Frequently Sampled Oral Glucose Tolerance Test

The OGTT is most commonly used as a test to characterize glucose tolerance. However, measurements of glucose, insulin, and/or C peptide in response to ingestion of glucose can also be used to estimate insulin secretion and sensitivity (24). Ratios of change in insulin and glucose can be used as indices of insulin secretion, particularly when the gold-standard test of measuring insulin secretion, the hyperglycemic clamp, is unavailable or impractical, and these OGTT-derived estimates have been found to correlate reasonably with clamp measures in individuals without diabetes (24) as well as youth at risk for T2D (26). The oDI also predicts future risk of T2D (40). In the present study, there were declines in IGI and oDI by glucose tolerance category. Although the differences were not statistically significant among the CF groups, the original study was not powered to detect differences in these outcomes.

Individuals with CF are at high risk for CFRD given the impairments in insulin secretion present from a very young age. Animal models of CF implicate functional pancreatic β-cell defects early in life even in the absence of structural pancreatic abnormalities (41), and pancreatic autopsies in very young children with CF have demonstrated relative reductions in β-cell mass (3) independent of pancreatic exocrine damage. Hyperglycemia on CGM as well as abnormalities captured by fsOGTT in youth with CF reflect these early abnormalities in pancreatic β-cell function. These deficits place individuals with CF at high risk for progressive β-cell dysfunction over time with accumulating islet injury from pancreatic exocrine damage and inflammation. From the findings described in this report, we might extrapolate that changes in CGM variables, for example, increasing glycemic variability, may be a useful tool for monitoring declines in β-cell function over time. This hypothesis warrants further testing in larger longitudinal cohorts.

CGM measures in this analysis did not correlate with iAUCcpeptide:iAUCglucose, an fsOGTT-derived measure of insulin secretion that takes into account the entirety of the OGTT. One explanation for the lack of correlation between CGM and this fsOGTT measure may be that the patterns of dysglycemia on CGM present in this population reflect abnormalities in early insulin secretion, which are best captured by changes in insulin or C peptide:glucose within the first 30 minutes of the OGTT. There was a decrease in iAUCcpeptide:iAUCglucose with worsening glucose tolerance categories, suggesting overall declines in insulin secretion, but these changes did not reach statistical significance, likely a function of the smaller sample size in this study.

CGM also did not correlate with insulin sensitivity as calculated by the Matsuda index. This lack of correlation is consistent with the generally normal insulin sensitivity reported in CF individuals with early dysglycemia (1, 42). All participants in this study were also at baseline health when recruited and not studied in the context of a recent illness or pulmonary exacerbation, scenarios that could transiently decrease insulin sensitivity. Interestingly, insulin sensitivity as reflected by the Matsuda index was higher in CF NGT participants compared to HCs as well as to CF AGT and CFRD subgroups. Similarly, higher insulin sensitivity in CF NGT individuals compared to HCs has been described previously by Moran et al (43), as measured by hyperinsulinemic-euglycemic clamps well, and Battezzati (42), as measured by the OGTT modeling of insulin sensitivity. This reduction in insulin sensitivity has been postulated to represent an early compensatory response in the presence of reduced β-cell function, or could occur secondary to a catabolic state from malabsorption, undernutrition, and increased energy expenditure in CF. A recent paper investigating mechanism of hypoglycemia in adults with CF also reported increased insulin sensitivity, along with impaired counterregulatory hormone production, in individuals with hypoglycemia on OGTT (44). In our present study cohort, there was no correlation between free-living hypoglycemia captured by percentage of time less than 60 mg/dL on CGM and the Matsuda index. Additional studies exploring the mechanisms behind hypoglycemia in CF and potential associations with insulin sensitivity are needed.

Last, with the OGTT we explored insulin clearance across glucose tolerance categories. Reduced insulin clearance has been proposed as a risk factor for T2D, or as a compensatory response to insulin resistance, and differences in insulin clearance by race/ethnicity and age have been described (45). One study in PwCF described abnormalities in insulin clearance, with females displaying greater insulin clearance than males (46). In our sample, we found no differences in estimated insulin clearance among our CF participants and controls. In contrast to Battezzati and colleagues (42), female participants in our study displayed a trend toward lower insulin clearance compared to male participants with CF. Whether this plays a role in the higher rates of CFRD seen in females requires further study. Larger studies across a wider age group, with potentially more direct methods of measuring insulin clearance, are needed to further understand the role insulin clearance may play in the pathogenesis of CFRD.

Continuous Glucose Monitoring as a Screening Tool For Cystic Fibrosis–related Diabetes

Given the challenges of obtaining annual OGTTs for diabetes screening in the CF population (8), some centers have advocated use of CGM as a screening and diagnostic tool (10, 16, 47, 48). However, much remains unknown regarding the exact CGM variables and corresponding thresholds needed to guide diagnosis and timing of therapeutic intervention in the CF population. In populations with positive autoantibodies at high risk for type 1 diabetes, CGM detects early dysglycemia in the preclinical phase (49) and increased hyperglycemia and glycemic variability on CGM have also been described to predict progression to type 1 diabetes (50). In populations with CF, a number of small studies have associated hyperglycemia on CGM, before a diagnosis of CFRD, with impairments in lung function (11), increased P aeruginosa infections (33), and weight loss (13). Hyperglycemia on CGM may also identify individuals at risk for eventual development of CFRD (20, 22). These findings, however, have yet to be validated in larger, prospective studies.

In this report we explore the role of CGM as a screening tool for diabetes by comparing it to the gold-standard OGTT. The variables tested (average glucose, peak glucose percentage of time spent > 140 mg/dL [7.8 mmol/L], percentage of time > 180mg/dL [10.0 mmol/L], percentage of time > 200 mg/dL [11.1 mmol/L], SD, CV, and MAGE) performed poorly at distinguishing those with CFRD from those without diabetes. These CGM variables also had low sensitivity for identifying those with prediabetes. We then sought to determine the thresholds for these CGM variables that maximized sensitivity for detecting CFRD, with the goal of 100% sensitivity so that no individuals with CFRD would be missed. If such a threshold could save individuals from needing an OGTT, CGM would be a reasonable first-line screen for ruling out diabetes. Notably, there was a wide range in 2hG values among individuals exceeding this threshold and significant overlap in 2hG values between those who exceeded, compared to those who did not exceed, the identified threshold. Because this analysis was conducted with a convenience sample of study participants and the prevalence of CF AGT in this group is higher than reported in the general CF population, caution should be applied when extrapolating findings to the wider CF population.

As a screening tool, CGM has potential advantages and limitations when compared to the OGTT. The OGTT requires fasting and collection of multiple venous samples over a 2-hour period. In contrast, CGM provides more detailed characterization of free-living glucoses over 1 to 2 weeks, but testing conditions are not standardized and readings are subject to differences in diet and activity among individuals. Given the challenges of collecting an annual OGTT, our group and others have also examined the utility of alternate HbA_1c thresholds and alternate glycemic markers including fructosamine, glycated albumin, and 1,5-anhydroglucitol as screening tools for CFRD (51-53). Identified thresholds for HbA_1c have ranged from 5.5% to 5.8% with varying sensitivities of 68% to 94%. In our prior report, alternate markers were not found to outperform HbA_1c at identifying those with CFRD. As markers of average glycemia, these tests might reasonably be expected to reflect average glucose as captured by CGM. In the CF population with a high prevalence of early insulin insufficiency, CGM measures of hyperglycemia and glycemic variability primarily capture postprandial glucose excursions that align with early OGTT-identified abnormalities.

However, findings from this study also demonstrate that CGM cannot be used interchangeably with the OGTT for diagnosing CFRD as currently defined. Rather, prospective longitudinal studies are needed to determine which of these tests will better predict important nonglycemic clinical outcomes such as lung function and weight decline in PwCF. Identifying CGM thresholds predictive of clinical decline would inform the development of future glycemic-based intervention trials aimed at slowing disease progression in PwCF. Current OGTT-based criteria for diagnosing diabetes in the CF population have been extrapolated from populations at risk for T2D (7). Notably, CFRD as currently defined is associated with declines in pulmonary function and nutritional status and increased mortality from pulmonary failure (38, 39, 54), raising questions about the relevance of these glucose cutoffs for CF; increasingly studies are advocating a need to revisit current diagnostic cut points for CFRD and suggesting that lower thresholds for intervention may be more relevant for targeting CF outcomes (32, 34, 55-57).

Limitations

The data presented here represent a single center’s experience and are cross-sectional. We did not adjust for multiple comparisons when examining the relationships between CGM and fsOGTT because the analyses were exploratory. Given the inherent variability of CGM related to dietary intake and physical activity, future studies evaluating the utility of CGM as a diabetes screening tool should consider capturing glycemic data under standardized conditions, including a standardized meal or glucose load, for a set period of time during the duration of CGM wear. Although the OGTT has been considered the gold standard for diagnosing CFRD and prediabetes in CF, CGM outputs include many variables from which to choose for establishing potential screening and diagnostic thresholds, and alternate OGTT time points (such as the 30-minute or 60-minute glucose, and potentially measurement of insulin concentrations) may be equally sensitive markers, at the appropriate thresholds, for detecting early dysglycemia in the CF population. Furthermore, oral estimates of β-cell function have been derived and validated from comparisons to gold-standard intravenous clamp techniques in non-CF populations (26, 40, 58), and given differences in underlying diabetes pathophysiology, longitudinal studies are required to determine the validity of these measures for predicting diabetes and clinical decline in people with CF.

In conclusion, CGM measures of hyperglycemia, and particularly glycemic variability, correlate inversely with fsOGTT-derived estimates of β-cell function. Although CGM appears to have low sensitivity for detecting diabetes by current OGTT-based definitions, prospective, multicenter studies to determine the optimal glycemic targets and timing of glycemic-based interventions to optimize health in individuals with CF are needed. Looking to the future, the landscape of CF is changing, most notably with the widespread introduction of CF transmembrane conductance regulator modulators, therapies that have proved to be highly effective in slowing pulmonary function decline and improving BMI. However, limited quality of life and high mortality rates related to pulmonary disease remain, and future studies focusing on early insulin insufficiency, progression of β-cell failure, and the relationships between these measures and clinical decline in CF continue to be of key importance.

Glossary

Abbreviations

1hG

1-hour glucose

2hG

2-hour glucose

AGT

abnormal glucose tolerance

AUC

area under the curve

BMI

body mass index

CF

cystic fibrosis

CFRD

cystic fibrosis–related diabetes

CGM

continuous glucose monitoring

CV

coefficient of variation

FPG

fasting plasma glucose

fsOGTT

frequently sampled oral glucose tolerance test

HbA_1c

glycated hemoglobin A_1c

HC

healthy control

iAUC

incremental area under the curve

IGI

insulinogenic index

MAGE

mean amplitude of glycemic excursions

NGT

normal glucose tolerance

oDI

oral disposition index

OGTT

oral glucose tolerance test

PwCF

people with cystic fibrosis

T2D

type 2 diabetes

Additional Information

Disclosures: The authors have nothing to disclose.

Data Availability

Some or all data generated or analyzed during this study are included in this published article or in the data repositories listed in “References.”