Glycemia and Insulin Secretion in Cystic Fibrosis 2 Years After Elexacaftor/Tezacaftor/Ivacaftor: PROMISE-ENDO

Abstract

Background:

Elexacaftor/tezacaftor/ivacaftor (ETI) is a highly effective therapy that improves lung disease in people with cystic fibrosis (pwCF), but its effect on glucose tolerance and insulin secretion is unclear.

Methods:

PROMISE is a multicenter prospective, observational study of ETI in pwCF ≥12 years and at least one F508del allele. The PROMISE Endocrine substudy (PROMISE-ENDO) enrolled participants at 10 CF Centers where hemoglobin A1c (HbA1c) was collected and 3-hour oral glucose tolerance tests (OGTT) conducted to examine glucose tolerance, glucose excursions, and insulin secretory rates (deconvolution of C-peptide) and sensitivity (oral minimal model) prior to ETI and 12 to 18 months and 24–30 months following ETI initiation. Longitudinal mixed effects models were used to test within-subject ETI effects.

Results:

At baseline, 79 participants completed OGTTs (39 [49%] male, median [IQR] age 19.6 [14.7, 27.3] years, BMI z-score 0.12 [−0.51, 0.65]). At 12–18 months n = 68 and at 24–30 months n = 58 completed OGTTs. At 24–30 months, fasting glucose (mg/dL) decreased (94 [92, 96] to 90 [88, 93], P = .02) in the subset not on insulin therapy (n = 61), but no differences in 1-hour or 2-hour glucose were found. HbA1c decreased from 5.8% (5.6%, 5.9%) to 5.5% (5.4%, 5.6%), P < .001 by 24–30 months. Although insulin sensitivity (mU/L⁻¹ min⁻¹) decreased (8.4 [7.2, 9.5] vs 6.8 [5.8, 7.9], P = .03), no changes in oral disposition index were found, P = .14.

Conclusion:

After 2 years of ETI, fasting glucose and HbA1c showed modest decreases. Glucose tolerance varied, and overall measures of insulin secretion did not deteriorate.

Cystic fibrosis (CF) is a multi-system condition arising from recessive mutations in the gene encoding the CF transmembrane conductance regulator (CFTR) and is associated with shortened life expectancy largely due to chronic lung damage and, ultimately, pulmonary failure. With improved life expectancy due to advances in care, CF-related comorbidities are becoming more prevalent and impose additional medical and psychological burden on patients and families. CF-related diabetes (CFRD) is one such comorbidity: robustly collected data suggest up to 20% of individuals aged 10 to 20 years and 40% to 50% of adults with CF have CFRD (1). CFRD arises largely from progressive insulin secretion defects, is more likely to occur in the setting of pancreatic exocrine insufficiency (PI), and has traditionally been associated with worse clinical outcomes, including worse pulmonary function, increased pulmonary exacerbation rates, greater prevalence of important sputum pathogens, poorer nutritional status (2), and up to a 6-fold greater mortality (3). In fact, suggesting a role for the prediabetic state, declines in body mass index (BMI) and lung function may begin up to 4 years prior to a confirmed diagnosis of CFRD (4).

Among the latest advances in care are the development of CFTR modulator therapies, including ivacaftor for individuals with defects limited to CFTR function (5) and lumacaftor/ivacaftor, tezacaftor/ivacaftor, and elexacaftor/tezacaftor/ivacaftor (ETI) for the treatment of the most common CFTR mutation (F508del) which impacts CFTR production, transport to plasma membrane, and function. ETI was approved by the Food and Drug Administration (FDA) in the fall of 2019 for individuals aged ≥ 12 years and for those ≥ 2 years in April 2023; the treatment has been considered disease-modifying for many, given the improvements in lung function, hospitalization rates, and quality of life observed in the vast majority of F508del-affected patients. However, data on glycemic outcomes in CFTR-treated individuals has been limited. Following the introduction of ivacaftor: (i) improvements in insulin secretion during oral glucose tolerance testing (OGTT) and acute glucose infusion were found in 5 patients (6); (ii) observational data emerged that ivacaftor treatment was associated with CFRD resolution in some patients (7); and (iii) US and UK CF registry data identified lower rates of CFRD diagnosis in the 5 to 6 years following initiation of ivacaftor therapy (8). Thus, hope emerged that CFTR modulator therapy would delay or even reverse CFRD. Indeed, glucose tolerance improved following 1 year of lumacaftor/ivacaftor in a French multicenter study of 40 F508del homozygous individuals aged ≥ 12 years with impaired glucose tolerance (IGT) or CFRD (9) but not in the US-based PROSPECT study of lumacaftor/ivacaftor (10), or in a small Italian study of 6 individuals with dysglycemia or CFRD (11). However, study teams acknowledged that lumacaftor/ivacaftor is only modestly effective with respect to pulmonary outcomes, unlike ETI and ivacaftor, which are considered highly effective CFTR modulator therapies (HEMT).

The robust effects demonstrated in phase 3 trials of ETI (12) upon multiple CF disease manifestations re-invigorated interest in the potential for ETI to also alter the landscape of progressively worsening glucose tolerance and insulin secretion defects in CF, while also acknowledging that some damage to the islet and the islet neighborhood may be irreversible. The PROMISE Endocrine substudy (PROMISE-ENDO) aimed to examine changes in glucose tolerance, glucose excursion, insulin secretion and sensitivity, and hemoglobin A1c (HbA1c) at 12 to 18 and 24 to 30 months after initiation of clinically prescribed ETI.

Methods

Target Population

Participants were age ≥12 years at enrollment, had at least one F508del mutation, were ETI-naïve prior to being clinically prescribed ETI by their US Cystic Fibrosis Foundation–accredited CF center team, and were recruited from one of 10 CF Foundation Therapeutic Development Network centers designated for PROMISE-ENDO. The larger PROMISE study is a prospective, observational study of clinically prescribed ETI being conducted at 56 US Cystic Fibrosis Foundation–accredited CF centers and includes pulmonary, nutritional, microbiology, inflammatory, gastrointestinal, and patient-reported outcomes (NCT04038047). More complete details have been previously published (13, 14). Unique exclusion criteria for enrolling into PROMISE-ENDO included established CFRD with fasting hyperglycemia, confirmed autoimmune type 1 diabetes, use of long-acting oral hypoglycemic agents within 4 weeks of baseline visit, or use of systemic glucocorticoids within 14 days prior to the baseline visit. Use of insulin and other diabetes medications were recorded. Measurements and study procedures were performed before and following 12 to 18 months and 24 to 30 months of ETI therapy. The protocol was approved by the appropriate Institutional Review Board for all sites, and consent or assent (as appropriate) was obtained from study participants.

Study Procedures

Anthropometric measures

Weight (kg) and height (cm) were measured, and BMI calculated as weight in kilograms divided by height in meters squared. BMI z-scores were calculated based on CDC reference growth charts using Stata’s zanthro function for all participants, with data in individuals > 20 years normalized to the maximum CDC reference range age of 20 years (15). This approach allowed for initial analyses to include the entire cohort and to prevent “losing” participants who transitioned from “pediatric” to “adult” during the study period.

Oral glucose tolerance test

An intravenous catheter was placed for blood sampling for a 3-hour, 10-sample OGTT after the participant fasted overnight. Baseline blood samples were collected at t = −10 and −1 minute before consumption of dextrose (1.75 g/kg; max 75 g) over 2 minutes starting at t = 0. Additional blood samples were collected at t = 10, 20, 30, 60, 90, 120, 150, and 180 minutes. For participants with weight < 40.9 kg, a modified 5-sample OGTT was conducted (samples collected at −1 minute prior to consumption of dextrose, and t = 30, 60, 90, 120 minutes post). For the subset for whom an intravenous line could not be placed or maintained, samples were drawn at baseline and at 60 and 120 minutes following dextrose ingestion. Basal insulin (insulin glargine or detemir or rapid-acting insulin for individuals on an insulin pump) was continued in the subset treated with basal insulin, but rapid-acting bolus insulin dosing and any rapid-acting oral diabetes therapies (eg, repaglinide) were to be withheld for at least 6 hours prior to OGTT.

Biochemical analysis

Blood samples were collected in P-800 tubes (BD Biosciences; San Jose, CA, USA) and placed on ice. Samples were immediately centrifuged at 4 °C, separated, and frozen at −80 °C for subsequent analysis in the Diabetes Research Center Radioimmunoassay and Biomarkers Core at University of Pennsylvania. Plasma glucose was measured in duplicate by the glucose oxidase method using an automated glucose analyzer (YSI 2300; Yellow Springs Instruments, Yellow Springs, OH, USA). Plasma insulin and C-peptide were assayed in duplicate by double-antibody radioimmunoassay (Millipore; Billerica, MA, USA). HbA1c was measured at local sites in all PROMISE participants at baseline, and at 6, 12 to 18, and 24 to 30 months post-ETI visits.

Glucose tolerance

Glucose tolerance was defined by OGTT as CFRD without fasting hyperglycemia (CFRD without FH; fasting glucose < 126 mg/dL and 2-hour glucose ≥ 200 mg/dL), CFRD with fasting hyperglycemia (CFRD with FH; fasting glucose ≥ 126 mg/dL and 2-hour glucose ≥ 200 mg/dL), impaired glucose tolerance (IGT; 2-hour glucose ≥ 140 and < 200 mg/dL), indeterminate glucose tolerance (IND; 1-hour glucose ≥ 200 mg/dL and 2-hour glucose < 140 mg/dL), and normal glucose tolerance (NGT₂₀₀; 1-hour glucose < 200 mg/dL and 2-hour glucose < 140 mg/dL). Because insulin secretion defects are evident with 1-hour OGTT glucose as low as 155 mg/dL (16), more stringent criteria were set for normal glucose tolerance (NGT₁₅₅; 1-hour glucose ≤ 155 mg/dL and 2-hour glucose < 140 mg/dL) and early glucose intolerance (EGI; 1-hour glucose ≥ 155 mg/dL and 2-hour glucose < 140 mg/dL). An additional categorization included development of hypoglycemia, defined as glucose < 65 mg/dL during the OGTT.

Calculations

Insulin secretory rate (ISR) was estimated from C-peptide values and derived by parametric deconvolution of C-peptide kinetics using a 2-compartment model which includes plasma glucose (17) with WinSAAM software 3.0.8 (University of Pennsylvania, New Bolton Center, Kennett Square, PA). Incremental area under the curve (iAUC) was calculated for glucose, insulin, and ISR for 0–30 minutes (iAUC₃₀), 0–120 minutes (iAUC₁₂₀), and 0–180 minutes (iAUC₁₈₀) with baseline values subtracted by the trapezoidal method using Stata 17 software (StataCorp, LLC, College Station, TX, USA). Insulin sensitivity (SI) was estimated using oral minimal model-based calculations (17, 18). Beta-cell (β-cell) function estimated by the oral disposition index (DI) was calculated as the product of ISR (iAUC₃₀) and SI.

Statistical analyses

Graphs were generated to permit visual inspection of individual participant data at baseline and follow-up. Normality was assessed by visual inspection and the Shapiro-Wilk test, with extreme/outlying values Winsorized to yield near-normal residuals (19). Due to skew on several measures, baseline continuous data were reported as median (interquartile range [IQR]). Categorical data were reported as count (percentage). To first assess for differences in glucose tolerance from baseline to 12–18 and 24–30 months, the number of participants in each category at each time point was determined and visualized using Sankey plots. Sankey bar charts are an extension of longitudinal bar charts that add Sankey-style overlays between the bars at adjacent time points. These overlays illustrate transitions between states (here, glucose tolerance status) over time. The thickness of the path is proportional to the conditional probability of transitioning from the original state (glucose tolerance category) at an earlier time point to the current state/category. Compared to a simple vertical bar chart, Sankey plots provide deeper insight into the data by visualizing which groups are driving changes in the categories over time (20). Additionally, rates of hypoglycemia were compared across visits with McNemar’s Chi-squared test for paired nominal data.

To assess for within-subject ETI effects, defined as the change in glucose excursion and insulin secretion over 12 to 18 months and 24 to 30 months, we used longitudinal mixed effects models with the slope as the outcome and time (12–18 months and 24–30 months post ETI vs baseline) as the main effect (21). This method (i) allows for robust variance estimates that account for possible non-normality of slope measures; (ii) takes advantage of improved efficiency in a statistical model that adjusts for subject-level covariates; (iii) benefits from the strong correlation of slope within a subject over time; and (iv) models variance-covariance of the outcome with subject-specific slope terms using an unstructured correlation structure through the specification of random effect terms. Model convergence and goodness-of-fit were assessed for all models.

Graphical inspection of OGTT-related ISR reflected 2 potential phases of change: 0 to 40 minutes (early phase/ascent) and 75 to 180 minutes (descent). To test for changes in the ISR trajectory during these 2 phases, slopes were compared between visits using mixed effects models (visit*time) adjusted for ISR at the start of the phase. These analyses were stratified by baseline glucose tolerance, as normal insulin secretion patterns are altered with worsening glucose tolerance.

We examined changes in ISR iAUC₃₀ and DI between baseline and 24 to 30 months using box-and-whisker plots with individuals grouped by baseline glucose tolerance category. We also investigated potential differences in sex, age, height, weight, BMI, and ppFEV₁ according to glucose tolerance status and its change across visits, and whether glucose and insulin parameters differed according to pre-ETI modulator use.

Individuals with CFRD and a peak OGTT C-peptide <3.5 ng/mL at the baseline visit were also examined separately, with the hypothesis that recovery of insulin secretion in the setting of such severe insulin deficiency would be unlikely to respond to ETI and that inclusion could potentially obscure results.

All statistical analyses were performed using Stata 18 (StataCorp LLC, College Station, TX, USA). Statistical tests were 2-tailed and considered significant at P < .05.

Results

Baseline OGTTs were available in 79 individuals (39 male/40 female) with median (IQR) age: 19.6 years (14.7, 27.3); percent predicted FEV₁ (ppFEV₁): 90 (66.5, 101.8); BMI: 20.1 kg/m² (19, 23.2) for participants aged ≤ 20 years and 22.9 (20.7, 24.2) for those aged > 20 years; BMI Z-score: 0.09 (−0.51, 0.91) for participants aged ≤ 20 years and 0.20 (−0.46, 0.49) for those aged > 20 years, and HbA1c of 5.6% (5.3, 5.9) (Table 1). One individual was pancreatic sufficient based on fecal elastase but treated chronically with pancreatic enzyme replacement. Ten individuals had 3 to 5 OGTT timepoints sampled that included 0, 60, 120 minutes and allowed for determination of glucose tolerance category only. At baseline, 12 individuals were treated with insulin: insulin pump (n = 1); 70/30 insulin prior to overnight continuous gastrostomy feeds (n = 1); multiple daily injections with both long-acting and short-acting insulin (n = 5); long-acting insulin only (n = 4); and short-acting insulin only (n = 1). Forty-two (53%) were treated with another CFTR modulator therapy prior to initiation of ETI: 6 ivacaftor, 14 lumacaftor/ivacaftor, 22 tezacaftor/ivacaftor.

Table 1.

Participant characteristics at baseline (n=79)

Age, years		19.6 (14.7, 27.3)
Male, n (%)		39 (49.4)
Weight, kg		59 (52.3, 69)
BMI, kg/m2
Aged ≤20 years (n=41)		20.1 (19, 23.2)
Aged >20 years (n=38)		22.9 (20.7, 24.2)
BMI Z-score†
Aged ≤20 years		0.09 (−0.51, 0.91)
Aged >20 years		0.20 (−0.46, 0.49)
FEV1, %-predicted		90 (66.5, 101.8)
Modulator therapy^, n (%)
No		37 (46.8)
Ivacaftor		6 (7.6)
Lumacaftor/Ivacaftor		14 (17.7)
Tezacaftor/Ivacaftor		22 (27.8)
Glucose Tolerance, n
NGT200	NGT 155	25	11
IND	EGI	14	28
IGT	IGT	26	26
CFRD	CFRD	14	14
OGTT profile
Fasting glucose, mg/dL		94 (89, 103)
Fasting glucose, mg/dL ++ (n=70)		93 (89, 100)
1-hour glucose, mg/dL		195 (165, 226)
2-hour glucose, mg/dL		141 (107, 183)
HbA1c, % (n=77)		5.6 (5.3, 5.9)

Data are median (IQR) for continuous variables.

Abbreviations: BMI, body mass index; FEV₁, forced expiratory volume in 1-second; NGT₂₀₀, normal glucose tolerance (with 1-hour glucose <200 mg/dL & 2-hour glucose <140 mg/dL); NGT₁₅₅, normal glucose tolerance (with 1-hour glucose <155 mg/dL & 2-hour glucose <140 mg/dL); EGI, early glucose intolerance (with 1-hour glucose ≥155 mg/dL & 2-hour glucose <140 mg/dL); IGT, impaired glucose tolerance; CFRD, cystic fibrosis-related diabetes; OGTT, oral glucose tolerance test.

^†Participants with age >19.9 years were converted to age 19.9 in order to generate BMI Z-scores based upon CDC reference data for youth

^{^}within 90 days of baseline data

⁺⁺not receiving basal insulin

Figure 1 presents the flow diagram of data from participants who underwent OGTT at baseline, 12 to 18 months, and 24 to 30 months. For the 58 participants who completed OGTT at 24 to 30 months, median (IQR) weight was 66.4 kg (57.9, 71.5) and ppFEV₁ was 100 (89, 110). At 24 to 30 months, median (IQR) BMI and BMI Z-score were 22.2 kg/m² (19.8, 24.6) and 0.04 (−0.41, 0.99), respectively, for participants with baseline age ≤ 20 years (n = 35), and 23.9 kg/m² (21.5, 25.7) and 0.49 (−0.16, 0.91) for those with baseline age > 20 years (n = 23).

Figure 1.

Flow diagram of OGTT data collected at PROMISE Endocrine substudy baseline and follow-up visits.

Abbreviations: mos, months; OGTT, oral glucose tolerance test.

At both 12 to 18 months and 24 to 30 months, fasting glucose decreased in participants not on basal insulin (Table 2). While fasting C-peptide was higher compared to baseline at 12 to 18 months but not at 24 to 30 months, fasting ISR, derived using a 2-compartment model that includes C-peptide and glucose, was higher at both 12 to 18 months and 24 to 30 months in the overall group. No differences were found in 1-hour or 2-hour OGTT glucose concentrations between baseline and following 12 to 18 months or 24 to 30 months of clinically prescribed ETI. Also shown in Table 2, and visualized using Sankey plots in Fig. 2, are changes in glucose tolerance categories over time using categorizations that include either NGT₂₀₀ and IND, or NGT₁₅₅ and EGI, as defined above. When comparing baseline glucose tolerance in the n = 57 with results at both baseline and 24 to 30 months, glucose tolerance improved in 14 individuals, worsened in 12 individuals, and did not change in 31 individuals. The use of ordinal random effects models to formally test changes in glucose tolerance category over time were considered, but did not yield convergence, likely due to insufficient sample size. Rates of hypoglycemia (defined as glucose ≤ 65 mg/dL) during the OGTT were similar across visits (Table 2).

Table 2.

Characteristics of the analytic sample (n=70)^a at baseline and following 12–18 months and 24–30 months of clinically-prescribed elexacaftor/tezacaftor/ivacaftor

		n	Baseline		12–18 months		P b	24–30 months		P c
HbA1c, %		68	5.8 (5.6, 5.9)		5.5 (5.4, 5.7)		<0.001	5.5 (5.4, 5.6)		<0.001
HbA1c, %d		58	5.6 (5.5, 5.7)		5.4 (5.3, 5.5)		<0.001	5.3 (5.2, 5.4)		<0.001
HbA1c, % treated with insulin		10	7.1 (6.6, 7.7)		6.5 (6.2, 6.8)		0.001	6.8 (6.1, 7.5)		0.13
Fasting insulin, uIU/mLe		61	8.1 (7.1, 9.1)		9.7 (8.9, 10.5)		0.005	8.8 (7.8, 9.9)		0.22
Fasting C-peptide, ng/mLe		61	1.4 (1.3, 1.6)		1.6 (1.5, 1.7)		0.04	1.5 (1.3, 1.7)		0.33
Fasting ISR, pmol/L/min		60	28.7 (25.6, 31.7)		32.5 (29.6, 35.4)		0.02	33.7 (29.2, 38.2)		0.04
Fasting ISR, pmol/L/mine		53	30.0 (26.9, 33.1)		33.1 (30.2, 36.0)		0.08	34.4 (29.1, 39.7)		0.13
Glucose Tolerance, n
NGT200	NGT 155		21	9	26	8		21	9
IND	EGI		11	23	10	28		6	18
IGT	IGT		24	24	17	17		15	15
CFRD	CFRD		14	14	11	11		15	15
OGTT profile
Fasting glucose, mg/dL		70	97 (94, 100)		94 (91, 97)		0.05	94 (91, 98)		0.17
Fasting glucose, mg/dLe		61	94 (92, 96)		91 (89, 93)		0.01	90 (88, 93)		0.02
1-hour glucose, mg/dL		68	199 (189, 210)		201 (188, 214)		0.71	198 (187, 208)		0.71
1-hour glucose, mg/dLe		60	192 (182, 203)		193 (180, 206)		0.86	191 (180, 202)		0.79
2-hour glucose, mg/dL		70	158 (142, 175)		151 (134, 167)		0.18	156 (138, 174)		0.71
2-hour glucose, mg/dLe		61	140 (128, 153)		132 (119, 145)		0.20	138 (122, 153)		0.75
Hypoglycemia +, n (%)		56	16 (29)		23 (41)		0.09	17 (35)f		0.35
Peak glucose, mg/dL		52	216 (201, 232)		217 (201, 232)		0.98	215 (201, 230)		0.85
Time to peak glucose, min		52	78 (70, 87)		64 (58, 70)		0.001	73 (64, 82)		0.28
Peak ISR, pmol/L/min		60	190 (171, 209)		192 (173, 211)		0.85	187 (166, 207)		0.75
Time to peak ISR, min		60	87 (78, 95)		76 (68, 84)		0.02	83 (74, 92)		0.40
Insulin sensitivity, mU/L−1.min−1		59	8.4 (7.2, 9.5)		7.6 (6.4, 8.8)		0.28	6.8 (5.8, 7.9)		0.03
Disposition indexg		57	11480 (8897, 14063)		10956 (8466, 13447)		0.74	9410 (7403, 11418)		0.14

Data are mean (95% confidence interval) for continuous variables.

Abbreviations: ISR, insulin secretory rate; NGT₂₀₀, normal glucose tolerance (with 1-hour glucose <200 mg/dL & 2-hour glucose <140 mg/dL); NGT₁₅₅, normal glucose tolerance (with 1-hour glucose <155 mg/dL & 2-hour glucose <140 mg/dL); EGI, early glucose intolerance (with 1-hour glucose ≥155 mg/dL & 2-hour glucose <140 mg/dL); IGT, impaired glucose tolerance; CFRD, cystic fibrosis-related diabetes; OGTT, oral glucose tolerance test.

^aAnalytic sample comprised of n=55 with data at all three visits, n=12 with data at baseline and 12–18 months only, and n=3 with data at baseline and 24–30 months only.

^bComparison between baseline and 12–18 months using mixed effects reduced maximum likelihood regression at significance level p <0.05.

^cComparison between baseline and 24–30 months using mixed effects reduced maximum likelihood regression at significance level p <0.05.

^dResults in this row exclude participants who received insulin therapy (any type) at all visits, and 1 participant who initiated insulin therapy after baseline; participants who discontinued insulin therapy between visits not excluded (n=3).

^eResults in this row exclude participants who received basal insulin therapy at all visits, and 1 participant who initiated basal insulin therapy after baseline; participants who discontinued basal insulin therapy between visits not excluded (n=2).

^fResult is based on n=48 with full 3-hour OGTTs at baseline and 24–30 months.

^gDisposition index calculated as insulin sensitivity*iAUC ISR 30min

Figure 2.

Sankey plots showing the percentage of participants in each glucose tolerance category at baseline and follow-up visits and the magnitude of change in category across visits, as illustrated by the size of the flow paths.

Abbreviations: CFRD, cystic fibrosis–related diabetes; EGI, early glucose intolerance (with 1-hour glucose ≥ 155 mg/dL & 2-hour glucose < 140 mg/dL); IGT, impaired glucose tolerance; mos, months; NGT₁₅₅, normal glucose tolerance (with 1-hour glucose < 155 mg/dL and 2-hour glucose < 140 mg/dL); NGT₂₀₀, normal glucose tolerance (with 1-hour glucose < 200 mg/dL and 2-hour glucose < 140 mg/dL).

Mean HbA1c (95% CI) was lower at 24 to 30 months (5.5% [5.4, 5.6]) compared to baseline (5.8% [5.6, 5.9], P < .001, [n = 68]). This difference persisted when comparisons were restricted to the subset not treated with insulin (24–30 months 5.3% [5.2, 5.4] vs baseline 5.6% [5.5, 5.7], P < .001 [n = 58]), Fig. 3. For additional comparison, HbA1c from the larger PROMISE cohort (n = 397) was examined and similarly found to be lower over time (5.8% [5.7, 5.9] at 24–30 months vs 6.0% [5.9, 6.1] at baseline, P = .007). In the small subset of n = 25 individuals treated with insulin in the larger PROMISE study, HbA1c was not different at 24 to 30 months vs baseline (7.0% [6.3, 7.7] vs 7.0% [6.3, 7.6], P = .97).

Figure 3.

Hemoglobin A1c (HbA1c) at baseline and after 12 to 18 and 24 to 30 months (mos) of ETI therapy, according to whether an individual was on insulin therapy (orange) or not (blue). Dashed lines at 5.7% and 6.5% represent prediabetes and diabetes thresholds, respectively. Triangles represent the mean HbA1c at baseline (5.8%), at 12 to 18 mos (5.5%), and at 24 to 30 mos (5.5%).

Figure 4 presents glucose and ISR curves by baseline glucose tolerance category. Glucose concentrations between 120 and 180 minutes decreased at 12 to 18 months in individuals with IGT and CFRD but were not different by 24 to 30 months. When the 5 individuals with CFRD and a baseline peak C-peptide < 3.5 ng/mL were examined separately (Supplementary Fig. S1 (22)), this subgroup appeared to have lower OGTT glucose concentrations at 150 and 180 minutes and an increase in ISR at 12 to 18 months and 24 to 30 months, although sample sizes were too small for statistical comparison. Next, changes in the slope of the ISR curves during the first 40 minutes of the OGTT were examined by glucose tolerance category. Steeper ascending slopes, representing faster rises in ISR, were demonstrated in NGT, IGT, and CFRD participants at 24 to 30 months relative to baseline (Fig. 4, Supplementary Table S1 (22)). Steeper descending slopes in the later stages of the OGTT (between 75–180 minutes) were also demonstrated in NGT and IGT participants at 24 to 30 months relative to baseline and represent faster turn off of insulin with declines in glucose. Although time to peak glucose and time to peak ISR were faster by 12 to 18 months, these measures were not different from baseline by 24 to 30 months (Table 2). In contrast, insulin sensitivity (SI) showed no appreciable change at 12–18 months but was lower at 24 to 30 months, from 8.4 (7.2, 9.5) at baseline to 6.8 (5.8, 7.9), P = .03. DI, a composite of insulin secretion and sensitivity, did not change significantly across baseline and follow-up visits (Table 2).

Figure 4.

Plasma glucose (left panel) and insulin secretory rates (right panel) according to baseline glucose tolerance status (normal glucose tolerance [NGT]; early glucose intolerance [EGI]; impaired glucose tolerance [IGT]; CF-related diabetes [CFRD]) at baseline visit (orange), 12 to 18 months (mos) post-ETI (teal), and 24 to 30 months post-ETI (dark blue). An asterisk denotes a statistically significant difference in OGTT glucose between baseline and follow-up (Wilcoxon matched-pairs signed-rank test).

Glucose, insulin, and C-peptide OGTT responses at baseline vs 12 to 18 months and 24 to 30 months after ETI are presented in Supplementary Table S2 (22). No differences in glucose iAUC₃₀, iAUC₁₂₀, and iAUC₁₈₀ across visits were found. Insulin iAUC, both iAUC₃₀ and iAUC₁₂₀, demonstrated modest increases at 12 to 18 months, but this increase was not maintained at 24 to 30 months. No differences were observed for insulin iAUC₁₈₀ at 12 to 18 months or 24 to 30 months. C-peptide iAUC and modeled ISR similarly did not change across visits.

Our examination of changes in ISR iAUC₃₀ and DI between baseline and 24 to 30 months according to baseline glucose tolerance category (Supplementary Figs. S2a and S2b (22)) revealed heterogeneity in DI over time, with both increases and decreases in DI. Within the NGT group, for example, declines in DI and glucose tolerance appeared to be driven by SI, as median ISR iAUC₃₀ was stable, although sample sizes were too small for formal statistical comparison.

Glucose and insulin parameters were not different between baseline and 24 to 30 months after adjusting for pre-ETI modulator use (n = 37 naïve vs n = 42 on a modulator, including 6/42 on ivacaftor at baseline). When participants with improved glucose tolerance status or stable NGT status (n = 16) were compared to those who had worsening glucose tolerance status or stable CFRD status (n = 22), no differences in baseline characteristics between groups were found (data not shown). No differences in sex, age, height, or weight were observed when participants were stratified according to change in glucose tolerance (improved, unchanged, or worsened). Changes in ppFEV₁ and BMI z-score from baseline to 24–30 months did not demonstrate a clear relationship with change in glucose tolerance status (Fig. 5). Baseline OGTT was consistent with CFRD in the one participant with pancreatic sufficiency and improved at 24 to 30 months (NGT). However, the impact of ETI on insulin secretion could not be gleaned as this participant only had 3-sample OGTTs.

Figure 5.

a. Percent predicted FEV₁ (ppFEV₁) at baseline and after 24-30 months (mos) of ETI therapy, according to whether an individual’s glucose tolerance status was categorized at baseline as normal (NGT, light blue), early intolerant (EGI, purple), impaired (IGT, orange), or diabetic (CFRD, navy blue), and whether their glucose tolerance status worsened, did not change, or improved from baseline to 24-30 mos.

b. BMI z-score at baseline and after 24-30 months (mos) of ETI therapy, according to whether an individual’s glucose tolerance status was categorized at baseline as normal (NGT, light blue), early intolerant (EGI, purple), impaired (IGT, orange), or diabetic (CFRD, navy blue), and whether their glucose tolerance status worsened, did not change, or improved from baseline to 24-30 mos

Discussion

This relatively large prospective, observational study examined glucose regulation and insulin secretion using frequently sampled OGTTs in individuals with CF ≥ 12 years of age at baseline clinically treated with ETI. After 24 to 30 months of ETI, fasting plasma glucose and HbA1c showed modest improvements. However, OGTT glucose excursions and insulin secretion on average did not change. Although insulin sensitivity decreased by 24 to 30 months, no net change in β-cell function as represented by the oral disposition index was observed; these data suggest compensatory increases in insulin secretion were operative. These study findings also demonstrate on average persistence of CFRD and β-cell function impairments in the ETI era. Not unexpectedly, over the 24 to 30 months of observation, a subset of participants displayed declines in β-cell function. Importantly, others exhibited stable or improved β-cell function over a time period in which declines in β-cell function might be expected. However, the extent to which ETI is responsible for improvement or stabilization cannot be determined by this study. Additionally, no clear relationships with baseline insulin secretion, pulmonary function, and nutritional status, or changes in pulmonary function and nutritional status were evident. As ETI has now been FDA approved in youth as young as age 2 years, the potential of early ETI initiation to disrupt progressive defects in glucose regulation and insulin secretion as well as progression to CFRD will be an important clinical consideration.

Following the introduction of HEMT, a small but growing number of publications have hinted at improvements in glucose tolerance and glycemia (9, 23), although other studies have been conflicting (10, 24). Early case series described increased insulin secretion post-ivacaftor (6); and analyses of the UK and US CF Patient registries found lower incidence of CFRD 5 years post-ivacaftor compared to the rest of the CF population (8). Following the widespread introduction of the highly effective modulator ETI, a growing number of

studies have reported on glycemia post-ETI with measures including HbA1c, continuous glucose monitoring (CGM), and OGTT, most focused on follow-up outcomes within the first year of treatment (25–29).

Our findings confirm decreases in HbA1c reported in studies within the first year post-ETI (27, 29–31) and add to these findings that a lower HbA1c is maintained out to 2 years post-ETI in individuals without established CFRD. Although glycemic excursions and OGTT-derived disposition index did not change, as has been reported in smaller studies (11, 27), fasting glucose decreased, albeit modestly. Possible mechanisms to explain the decrease in fasting glucose after ETI include lowering of overnight glucose via decreased hepatic glucose output and/or modulation of the glucagon hypersecretion observed in some people with CF (32). The findings of higher fasting insulin secretory rates could contribute to both of these mechanisms. Lower overnight glucose may translate to lower average glucose and therefore lower HbA1c. Studies that have examined glycemia post-ETI with direct, free-living measures such as CGM have been mixed, however. One prospective, observational study of CGM in 23 adults with and without CFRD identified improvements in hyperglycemia and glycemic variability in the 3 to 12 months following the introduction of ETI (33). In contrast, other studies comparing CGM pre-ETI to data 3 to 12 months post-ETI (30, 31, 34) have not confirmed these findings. Non-glycemic factors such as higher hemoglobin (35) have been postulated to influence HbA1c post-ETI (27). However, a large, recent study examining HbA1c post-ETI confirmed decreases in HbA1c even after adjusting for hemoglobin concentrations, although other mechanisms affecting hemoglobin glycation rates may still be operative (31) and require further study.

We also detected changes in the slope of the ISR curve, demonstrating faster insulin secretion after ETI in the earlier phases of the OGTT in NGT, IGT, and CFRD participants. Early phase insulin secretion defects have been well-described in individuals with pancreatic insufficient CF, even in young children and individuals with NGT (16, 36–38). Similar partial restoration of normal insulin secretion dynamics in the absence of frank improvements in glucose tolerance has been found with 6 months of treatment with the type 2 diabetes medication sitagliptin (39), and in secondary analyses of mixed meal tolerance testing performed in individuals clinically prescribed ETI as part of a larger longitudinal study of insulin secretion in CF (40). The pathophysiology underlying impaired insulin secretion is likely multifactorial, with contributions from decreased β-cell mass, abnormal islet vascularization and innervation, abnormalities in pancreatic paracrine signaling, and inflammation (41–43). Notably, several case series have reported reduced insulin doses and discontinuation of insulin requirements in individuals with CFRD following ETI (44–46). Despite no net changes in C-peptide iAUC, our findings suggest that ETI may modulate some of the insulin secretory impairments observed in CF, driving improvements in glycemia, fasting C-peptide, and ISR within the first year, followed by plateauing of effects in the second year, although the exact mechanisms require further investigation.

Variability in glucose tolerance categories over time has been described in people with CF (47), and the extent to which ETI contributed to changes in glucose tolerance observed here are uncertain. However, a retrospective study of OGTTs in 46 adults with CF and a minimum of 4 years follow-up, conducted prior to the widespread introduction of ETI, described decreases in insulin sensitivity without compensatory increases in insulin secretion resulting in increased rates of abnormal glucose tolerance over time (48). A recent case-control, longitudinal study including 9 adults treated with ETI compared to 8 controls measured insulin secretion with mixed meal tolerance tests and the gold-standard glucose-potentiated arginine (GPA) clamp at baseline and 5 years (in cases) vs 3 years (in controls). Glucose potentiated insulin and C-peptide responses to GPA declined over time in ETI-naïve “disease controls” but not in those treated with ETI, while early insulin secretion during a mixed meal test was quicker (suggesting some restoration of normal insulin secretion pattern) in the ETI group. Taken together, these studies suggest ETI may interrupt the natural deterioration in β-cell function otherwise observed over time in people with CF (40).

Our findings are notable for modest improvements in glucose and insulin response 2 years following ETI, with the most significant changes from baseline occurring within the first year. However, the specific mechanisms underlying the observed effects of CFTR modulators on glucose homeostasis are incompletely understood. Given CFTR protein expression has been identified in a small subset of pancreatic β-cells (49, 50) as well as α-cells (32), CFTR modulators have been posited to have a direct effect on β-cell and α-cell function, but data are conflicting (49, 51–53). Alternatively, indirect effects on β-cell function could occur through reduction in inflammation of the islet milieu and improvements in insulin sensitivity through decreased systemic inflammation (41, 42). Increases in weight and BMI have also been described following ETI (29, 54). These changes have largely been attributed to increases in fat mass rather than lean mass (30, 54), and this increase in fat mass could confer an increase in systemic insulin resistance and offset improvements in insulin secretion. Whether or not ETI may influence tissue-specific changes in insulin sensitivity or hepatic insulin clearance are unclear and require further study. Finally, ETI is not expected to undo reductions in β-cell mass and scarring that disrupt islet vasculature and innervation.

Limitations of our study include the lack of CGM data and mixed meal tolerance tests to characterize free-living glycemia and lack of hyperglycemic clamps to assess insulin sensitivity. However, OGTTs are the recommended test for diabetes screening and diagnosis, and conducting additional measures including hyperglycemic clamps, particularly across multiple CF centers, poses challenges. HbA1c was measured in local clinical labs; this lack of centralization is a potential source of between-site heterogeneity but is not expected to increase within-subject heterogeneity or impact findings of lower HbA1c. Additionally, while we recognize the potential for type I error as a limitation given multiple statistical tests, the OGTT measures tested are highly correlated. We therefore deemed an adjustment for multiple comparisons inappropriate given the untenability of the assumption that random variation underlies all associations in these data (55). Although a subset of individuals with CF in the United States are not treated with ETI, the vast majority of these individuals have severe CFTR genotypes that are not amenable to current CFTR modulator therapies, often have worse disease severity, and therefore would not be appropriate controls. The investigators pursued somewhat “agnostic” expectations with respect to changes in glucose tolerance and insulin secretion and sensitivity: (i) glucose tolerance status is assigned categorically and may not reflect subtle improvements in insulin secretion; (ii) in the event ETI does impact insulin secretion, islet function may be so compromised in some individuals due to scarring and fatty infiltration that insulin secretion is irrecoverable; and (iii) improvements with ETI might be expected in the subset of individuals in whom systemic inflammation drives the unmasking of insulin secretion defects. Unfortunately, no clear relationships between changes in glucose excursion and insulin secretion with changes in clinical parameters were evident. Finally, the rapid clinical initiation of ETI in eligible individuals prevented the prospective collection of comparator data, and the research community is limited by the lack of longitudinal, pre-ETI data for historical context in which to place these findings. Nonetheless, universal deteriorations in glucose tolerance and insulin secretion were notably absent, and follow-up of participants out to 54-months will provide additional insight into progression.

This study represents the longest and largest study to date to capture longitudinal OGTT metrics including both estimates of insulin secretion and sensitivity up to 2.5 years, prospectively, post-ETI. Ongoing studies will be important to follow the natural history of insulin sensitivity and insulin secretory changes to inform screening and management strategies in the HEMT era. Furthermore, studies assessing glycemia in children introduced to ETI early in life will help determine the role of earlier introduction of ETI in slowing or preventing progression to CFRD.

Abbreviations:

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

EGI

early glucose intolerance

ETI

elexacaftor/tezacaftor/ivacaftor

FH

fasting hyperglycemia

HbA1c

glycated hemoglobin

HEMT

highly effective CFTR modulator therapy

iAUC

incremental area under the curve

IGT

impaired glucose tolerance

IND

indeterminate glucose tolerance

IQR

interquartile range

ISR

insulin secretory rate

NGT

normal glucose tolerance

OGTT

oral glucose tolerance test

SI

insulin sensitivity

Data Availability

Some or all datasets generated during and/or analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.