Randomized Trial of the Insulin-Only iLet Bionic Pancreas for the Treatment of Cystic Fibrosis– Related Diabetes

Cystic fibrosis (CF)–related diabetes (CFRD) is one of the most common nonpulmonary complications of CF, occurring in 30–50% of adults (1). The development of CFRD is associated with a decline in lung function, compromised nutritional status, and higher mortality (2). Insulin therapy has been shown to improve these outcomes but adds a substantial treatment burden with frequent blood glucose monitoring, close attention to carbohydrate intake, and multiple daily injections (MDIs) or insulin pump therapy to meet glycemic goals (3–6). Recommended glucose targets are the same for CFRD as for type 1 and 2 diabetes (6,7); however, people with CFRD face unique challenges to achieving good glycemic control, including pancreatic exocrine insufficiency affecting gastrointestinal absorption and motility, high caloric dietary requirements to ensure appropriate nutritional status, and frequent pulmonary exacerbations and systemic glucocorticoid treatment that can lead to substantial changes in insulin requirement.

Recent advances in diabetes technology have included automated insulin delivery (AID) systems that use continuous glucose monitoring (CGM) and insulin pumps to automatically adjust subcutaneous insulin delivery based on insulin-determining control algorithms. AID systems have primarily been developed for and studied in people with type 1 diabetes and are associated with improved glycemic control in these patients (8–13). AID technology may offer benefits for people with CFRD, but there have been limited studies investigating these devices in this patient population (14,15).

The iLet bionic pancreas (BP) is an AID system shown to be effective in improving glycemic control in children and adults with type 1 diabetes (13,16–19). There are several unique features of the BP that may be particularly useful in the management of CFRD. Unlike currently available hybrid closed-loop (HCL) AID systems, the BP requires no information for initialization other than body mass, which is used only for dose scaling. The device is designed to continuously adapt to a wide range of insulin needs, making it well suited to address the fluctuations in insulin requirements occurring in those with CFRD. The BP has been studied across a wide range of insulin requirements as low as 11 units/day and as high as 145 units/day (13,20–23), which may be particularly useful in people with CFRD who have highly variable degrees of endogenous β-cell function and insulin needs. In addition, users of the BP do not count carbohydrates and, instead, enter qualitative meal announcements such as usual for me, more, or less. Finally, all insulin dosing for correction of hyperglycemia is automated, reducing the need for user engagement with the device.

We previously reported a pilot study investigating an earlier version of the BP in insulin-only and bihormonal (insulin plus glucagon) configurations in three adults with CFRD, showing beneficial trends in glucose control with both configurations (15). Here, we report the results of a clinical trial comparing the iLet BP in the insulin-only configuration in 20 adults with CFRD compared with their usual diabetes care over 2 weeks.

We conducted a single-center, open-label, random-order, crossover trial comparing the BP with usual care (UC). Each study arm was 2 weeks in duration without any run-in period. Study visits took place at the Massachusetts General Hospital Diabetes Research Center. There were no restrictions placed on diet or activity. The protocol was approved by the Massachusetts General Hospital institutional review board, and written informed consent was obtained from all participants.

Participants were all adults age ≥18 years with CF, had been diagnosed with diabetes, and managed their glucose with insulin. Participants were required to have a baseline HbA_1c ≥6% or a mean CGM glucose ≥125 mg/dL and to have a total daily dose of insulin >0.1 units/kg/day. Other eligibility criteria are available in Supplementary Table 1.

Baseline clinical data collection included questionnaires and chart review for demographics, anthropometric measures, and medical and diabetes history. Permuted block randomization with a block size of 2 was used to randomly assign participants to begin with either BP or UC. Participants had a scheduled phone call on days 2 and 5 of each study arm and a study visit on day 14 of each arm. After completion of the first study arm, participants started immediately on the next arm or waited up to 2 weeks in case of scheduling issues or interval illness.

Participants completed electronic surveys to capture patient-reported outcomes (PROs) at enrollment and at the end of each study arm. Surveys included the Diabetes Distress Scale; Hypoglycemic Confidence Scale; Diabetes Technology Attitudes; Insulin Dosing Systems: Perceptions, Ideas, Reflections, and Expectations (INSPIRE); Hypoglycemia Fear Survey; 5-level EQ-5D; World Health Organization Well-Being Index; and Bionic Pancreas User Opinion Survey (Supplementary Table 2). Participants completed daily e-mailed surveys to capture self-reported episodes of symptomatic hypoglycemia, treatment of hypoglycemia, exercise, and alcohol consumption occurring during both treatment arms.

The study device consisted of the iLet BP with embedded insulin dosing algorithms, an infusion set (Inset I; Unomedical), and the Dexcom G6 CGM. The control algorithms were the same as those used in previous type 1 diabetes studies, without modification (13,16,17). The default glucose target was set at 120 mg/dL and could be adjusted to a higher or lower target by ±10 mg/dL at the investigator’s discretion. Participants using MDIs for UC were typically started at a higher target to account for washout of long-acting insulin and then adjusted to a lower target on day 2. A different recurring target could also be set (e.g., during nighttime), as needed.

During the UC period, participants continued the same insulin delivery and glucose monitoring methods used at the time of screening. Participants on MDI therapy were given an insulin smart pen (InPen; Medtronic) to capture all rapid-acting insulin doses delivered during the UC arm. Participants were not allowed access to the InPen phone application so that they would have to continue their previous method of dose calculation. Participants not using the Dexcom G6 CGM as part of their UC were provided a blinded Dexcom G6 Pro to wear during the UC arm to collect the primary outcome data.

The prespecified primary outcome was the percentage of time the sensor-measured glucose level was in the target range of 70–180 mg/dL (time in range [TIR]) on days 3–14 of each study period. The first 2 days of the BP arm were excluded to allow for initial adaptation of the algorithms and to allow for washout of long-acting insulin for participants treated using MDIs. The first 2 days of the UC arm were similarly excluded. Key secondary outcomes included mean CGM glucose level, median time spent in sensor-measured hypoglycemic ranges (<54 and <70 mg/dL), median time spent in sensor-measured hyperglycemic ranges (>180 and >250 mg/dL), measures of glycemic variability (coefficient of variation [CV] and SD), and the number of episodes of symptomatic hypoglycemia. Safety outcomes were collected throughout the entire period of study participation and included episodes of severe hypoglycemia, diabetic ketoacidosis, and any study device–related issues.

Data were analyzed using SAS 9.4 software (SAS Institute Inc., Cary, NC). All analyses comparing the BP and UC arms followed the intention-to-treat principle, with the data from each participant analyzed according to the treatment assigned by randomization. A secondary per-protocol analysis that included participants adhering to the protocol, reflecting the time the BP was in use during the BP arm, was also performed. Normality was assessed using the Shapiro-Wilk test. The significance of differences between the BP and UC arms in primary and secondary outcomes was assessed using paired t tests for normally distributed data or Wilcoxon signed rank tests for nonnormally distributed data, adjusting secondary outcomes for multiple comparisons using the Bonferroni method. Wilcoxon signed rank tests were used to compare insulin daily dosing, self-reported symptomatic hypoglycemia episodes, and CGM use between arms. To confirm that the UC arm reflected typical glycemic control for this cohort, TIR collected at the screening visit was compared with TIR collected during the UC arm using paired t tests. PROs were compared between baseline data obtained at the time of enrollment and after the BP arm using Wilcoxon signed rank tests. Baseline questionnaires were used in this analysis instead of UC results to avoid potential bias from use of the BP during the course of the trial. A three-level random-effects model using quadratic B-spline functions was used to analyze intersubject and within-subject interday variations of glucose trends over a 24-h period and to compare glucose profiles in the UC and BP arms using previously described methods (24), as shown in Fig. 2. The glycemic outcomes of participants who failed to achieve the American Diabetes Association (ADA) recommendations of TIR ≥70% during UC were considered separately in a post hoc analysis. Carryover effects and period effects were tested using mixed-effects models with patient-level random effects, period fixed effects, and period × arm interactions. All P values are two-sided.

The sample size was selected to detect superiority of the BP relative to UC based on the primary outcome. According to the International Consensus on Time in Range, a change in TIR of at least 5% is considered to be clinically meaningful (25). Based on preliminary data in CFRD and type 1 diabetes, participants were predicted to have an average TIR of 50 ± 18% in the UC arm and 70 ± 12% in the BP arm. Nine participants were predicted to be required to detect a difference between arms using a paired two-sided t test with a power of 80% at an α of 5%, assuming a correlation pre- and posttreatment of 0.3. The sample size was increased to 20 participants to account for the possibility of increased variability in glycemic control in patients with CFRD, to gather data on safety, and to increase power for detecting differences in secondary outcomes.

The full trial protocol and data sets containing deidentified data generated and/or analyzed during the current study are available from the corresponding author on reasonable request. Complete participant-level CGM and BP insulin administration raw data for all participants during both arms of the trial are available in Supplementary Fig. 13.

Between April 2021 and June 2022, 30 adults age 18–65 years were consented and screened for study participation (Supplementary Fig. 3). Of these, three did not meet inclusion criteria, three declined to participate, and two were lost to follow-up. The remaining 22 participants were randomized. Two participants were allocated to a study arm but withdrew prior to beginning any study procedures and were excluded from all analyses. The remaining 20 participants completed the entire study and were included in the primary analysis. There were no unscheduled visits during either arm. Fourteen participants had a total of 47 unscheduled contacts with the study team during the trial, which included 31 occurring during the BP arm, 12 during the UC arm, and 4 during enrollment but not during either treatment arm. Reasons for contact included issues or questions about the BP (36%) or about other study devices or procedures (26%), reporting hyperglycemia lasting >2 h (9%), and scheduling or other administrative issues (30%).

Baseline characteristics of the participants who completed the trial (n = 20) are summarized in Table 1. The mean age was 40 ± 13 years, 50% were female, and average diabetes duration was 15 ± 10 years. At the time of screening, 18 participants (90%) used CGM, and 10 (50%) used an insulin pump, 8 of whom (40% of all participants) used a Food and Drug Administration–cleared HCL system (all t:slim X2 with Control-IQ). The mean HbA_1c level within the 6 months prior to screening was 7.4 ± 1.6% (57 ± 6 mmol/mol). Participants’ lung disease ranged from mild to severe, and all had pancreatic insufficiency. One participant was treated with glucocorticoids, beginning the first arm (BP) on a dose of prednisone 25 mg daily and gradually tapering to a dose of 10 mg at the end of the second arm (UC).

The primary outcome, TIR, was higher in the BP arm than the UC arm (75 ± 11% vs. 62 ± 22%, P = 0.001), with an estimated treatment difference of 12 percentage points (95% CI 5–19%) (Fig. 1). Secondary outcomes are presented in Table 2. Mean CGM glucose was lower in the BP arm than the UC arm (150 ± 19 vs. 171 ± 45 mg/dL, P = 0.007). The median percentage of time with sensor-measured glucose <54 mg/dL and <70 mg/dL was not significantly different between the two arms (P = 1.0 for both). The median percentage of time with sensor-measured glucose >180 mg/dL and >250 mg/dL was significantly lower in the BP arm than the UC arm (P = 0.014 for both). There was no significant difference in SD or CV between arms (P = 0.14 and 1.0, respectively). Participants in the BP arm self-reported a higher number of symptomatic hypoglycemia episodes on daily surveys compared with the UC arm (0.7 vs. 0.4 median episodes per day, P = 0.01).

Glycemic control with the BP was improved during both daytime and nighttime hours (Fig. 2 and Supplementary Table 4). Average TIR increased by 11 percentage points in the BP compared with the UC arm during the daytime hours of 6:00 a.m. to 11:59 p.m. (74 ± 10% vs. 63 ± 21%) and by 15 percentage points during the nighttime hours of 12:00 a.m. to 5:59 a.m. (77 ± 16% vs. 62 ± 26%).

The total daily dose of insulin was similar in both treatment arms at 0.45 units/kg/day (P = 0.15). Although basal insulin did not differ between arms (P = 0.14), participants received slightly greater bolus insulin delivery during the BP arm at 0.28 units/kg/day (interquartile range [IQR] 0.21–0.5) compared with 0.21 units/kg/day (IQR 0.16–0.27) during the UC arm (P = 0.004) (Supplementary Table 5). Sixteen participants (80%) ended the trial using the typical glucose target set at 120 mg/dL, two participants used a higher target of 130 mg/dL, and two used a lower target of 110 mg/dL. Participants used the meal announcement feature an average of 2.2 ± 1.0 times/day. During days 3–14, CGM was active 96.8% of the time in the UC arm and 94.4% in the BP arm, with no difference between arms (P = 0.16). Self-reported carbohydrate treatment for hypoglycemia, daily exercise, and alcohol consumption are shown in Supplementary Table 6. Exercise was reported for a median of 1.5 days (IQR 0–4.5) out of the 2 weeks during the BP arm and 0 days (IQR 0–5) during the UC arm.

In the subgroup of 11 participants who did not achieve the ADA-recommended goal of an average TIR ≥70% during the UC arm, the average TIR improved by 20 percentage points (from 47 ± 18% during UC to 67 ± 9% with the BP, P = 0.018), and the mean CGM glucose improved by 37 mg/dL (P = 0.053). In this subgroup, median time sensor-measured glucose was <54 and <70 mg/dL was not significantly different between the two arms, but the average time sensor-measured glucose was >180 and >250 mg/dL was significantly improved in the BP arm (P < 0.01 for both) (Supplementary Table 7).

Sensor-measured outcomes by participants’ UC modality are shown in Supplementary Table 8. Participants who were managed with HCL therapy at baseline had a similar average TIR during the BP arm (77 ± 10%) as in the UC arm (75 ± 10%). Participants managed with MDIs or nonautomated pumps during the UC arm, with or without CGM, had an average TIR of 53 ± 24% vs. 73 ± 12% during the BP arm.

CGM data for the first 48 h of BP use compared with the entire study period are shown in Supplementary Table 9. During the first 48 h of the BP arm, the average TIR was slightly lower (71 ± 15%), and the average CGM glucose and average time sensor-measured glucose was >180 and >250 mg/dL were higher than during days 1–14. The median time sensor-measured glucose was <54 mg/dL was very low at 0% (IQR 0–0.73) during the first 48 h.

Two episodes of pulmonary exacerbations managed with antibiotics on an outpatient basis occurred during the UC arm, none occurred during the BP arm, and five occurred after enrollment but outside of either study arm. Although acute illness may impact glycemic control, the TIR during UC for each of these two participants was similar to data collected at screening, and in the entire cohort, there was no significant difference between the TIR collected at screening and the TIR during UC (P = 0.68), suggesting that the UC arm accurately represents typical glycemic control in this cohort.

The per-protocol analysis showed similar findings as the intention-to-treat analysis (Supplementary Table 10). There were no significant period or carryover effects noted after adjustment for multiple comparisons (data not shown). Participant-level plots containing all CGM data throughout the UC arm and all CGM and insulin administration data throughout the BP arm for each participant are provided in Supplementary Fig. 13.

Survey results at the time of enrollment and after the BP arm are presented in Supplementary Table 11. Level of diabetes distress on the Diabetes Distress Scale and fear of hypoglycemia on the Hypoglycemia Fear Scale were low at baseline and did not change (P = 0.60 and P = 0.81, respectively). Hypoglycemic Confidence Scale, World Health Organization Well-Being Index, and Diabetes Technology Attitudes scores remained stable from enrollment to after the BP arm (P > 0.05 for all). On the INSPIRE survey, participants had a decrease in reported expectations after use of the BP (87–67, P = 0.003). On the Bionic Pancreas User Opinion Survey, 13 of the 19 participants who completed the survey (68%) agreed or strongly agreed that they felt less burdened in managing their diabetes while using the BP, and 12 of 19 (63%) felt that they had a greater peace of mind, spent less time thinking about diabetes, and would strongly recommend the device to others. Eleven of 19 (59%) felt freer with food choices while using the BP.

Adverse events are summarized in Supplementary Table 12. There were no severe and/or unexpected device-related adverse events in the BP arm. A total 19 adverse events were reported among 10 participants, with 7 events occurring during the BP arm and 5 during the UC arm. The remaining seven events occurred after enrollment but outside of either study arm. In the BP arm, six of the seven events consisted of hyperglycemia without ketosis, most related to a device issue. There were eight device issues reported in the BP arm (seven related to either the infusion set, tubing, insulin cartridge, or CGM sensor and one related to a temporarily nonresponsive touch screen interface). There were no episodes of severe hypoglycemia or diabetic ketoacidosis in either arm.

In this single-center, random-order, crossover trial, use of the insulin-only configuration of the iLet BP over 2 weeks significantly improved glycemic control in adults with CFRD. Compared with UC, use of the BP was associated with a clinically significant increase in average TIR along with decreases in mean CGM glucose and in sensor-measured hyperglycemia without an increase in sensor-measured hypoglycemia, which remained low throughout the study. While our previously published pilot study was the first to test a prototype BP in this population (15), this study represents the first adequately powered randomized clinical trial to investigate AID technology in this patient population. Results reported here suggest that these technologies may represent an important advancement in treatment options for people living with CFRD.

The randomized controlled pivotal trial of the iLet BP in 161 adult participants with type 1 diabetes demonstrated a statistically significant baseline-adjusted reduction in mean HbA_1c of 0.5%, along with an increase of 11 percentage points in the average TIR, reduction of 16 mg/dL in mean CGM glucose, reduction of 11 and 4 percentage points in the average percentage of time sensor-measured glucose was >180 mg/dL and >250 mg/dL, respectively, and no difference in sensor-measured hypoglycemia over a 3-month period (13,26). These results in adults with type 1 diabetes are notably similar in magnitude to our findings investigating the same device in adults with CFRD. Other studies investigating insulin pump and AID technology in this patient population are very limited. One nonrandomized prospective study in nine adults with CFRD reported that transition from MDIs to an insulin pump without automation was associated with improvement in HbA_1c over 6 months (27). Recently, we published a retrospective chart review study of 13 adolescents and adults with CFRD that noted improvements in TIR 3 months after patients transitioned from either MDIs or pump to an HCL system (t:slim X2 with Control-IQ) (14), further suggesting the potential benefits of AID systems for this patient population.

Although characterized by insulin deficiency similar to type 1 diabetes, CFRD is a unique form of diabetes with significant clinical differences that could theoretically impact the efficacy of AID technology. First, there is substantial variability in residual β-cell function in people with CFRD, ranging from mild insulin deficiency requiring only prandial rapid-acting insulin dosing to near complete β-cell dysfunction. In addition, people with CFRD are also predisposed to reactive hypoglycemia that can lead to challenges in prandial insulin dosing and timing, particularly in the setting of relative glucagon deficiency caused by islet destruction (28). Moreover, people with CF require a high-calorie diet to meet their nutritional needs, which often involves high carbohydrate intake and significant prandial hyperglycemia. Frequent CF exacerbations and glucocorticoid treatment often lead to insulin resistance and transient increases in insulin requirements. The finding that the BP performs similarly in this unique patient population as in people living with type 1 diabetes is reassuring and suggests that this technology may also be potentially useful for other forms of diabetes, including type 2 and pancreaticogenic diabetes.

In this study, the BP achieved better glycemic control while delivering a similar total daily insulin dose and slightly higher bolus amount as received during UC, even with only two meal announcements per day on average. Although glucose was slightly higher than the 2-week average during the first 48 h while the system was initially adapting, sensor-measured hypoglycemia rates remained low during this period. Those participants who 1) were not meeting the ADA therapy goal of an average TIR of at least 70% during the UC arm and/or 2) were using insulin pumps without automation or MDI therapy at baseline had the greatest improvements in glucose control with the BP.

Overall, the BP appeared to be safe in this patient population over this 2-week study period. The total number of adverse events reported was relatively high; however, only 12 of 19 occurred while participants were active in either study arm, and the majority were deemed unrelated to study procedures, which is not unexpected in this medically complex patient population. In addition, different approaches for participant reporting of hyperglycemia without ketones measures were used in each arm such that participants were instructed to contact the study team at any time they experienced persistent hyperglycemia while using the BP. However, they were not instructed to do so during UC unless serum ketones became elevated, leading to greater reporting of hyperglycemia without ketosis during the BP arm. Despite no difference in time spent in hypoglycemic ranges, the rate of reported symptomatic hypoglycemia and reported treatment with carbohydrates were higher in the BP arm, which may have been related to participants learning to trust the device and become more used to spending less time in hyperglycemic ranges. Symptomatic hypoglycemia remained low overall, occurring less than once per day in both arms.

The development of CFRD has been associated with a significant impact on quality of life related to the added treatment burden of diabetes care on top of many other substantial pulmonary and gastrointestinal treatments (29,30). This trial is the first to report diabetes-related PROs in the CFRD population. In this study, no significant improvement in PROs was observed with the BP. Of note, the adults with CFRD enrolled in this study had a very low baseline level of diabetes distress and fear of hypoglycemia, leaving little room for detectable improvement with BP use. In addition, the high uptake of diabetes technology as part of UC may also have affected both the baseline and follow-up results. On the INSPIRE survey, participants reported a high positive expectation for what a closed-loop system could do for them at baseline, but their actual experience after using the BP moderated their positive expectations perhaps because nearly one-half of the participants used an HCL system for their UC. Despite this, a majority of participants indicated that use of the BP led to reduced diabetes burden, increased peace of mind, and freer food choices, and most indicated that they would strongly recommend the device to others.

The strengths of this study included the random-order, crossover design in a real-world, free-living environment. A diverse group of participants with CF were included across wide ranges in age, diabetes duration, pulmonary disease severity, insulin requirement, and insulin delivery modalities. Limitations of the study also warrant consideration. The number of participants was small, and the study duration was short. Data on the performance of the BP during a pulmonary exacerbation or steroid treatment were limited. The high rate of diabetes technology use during the UC arm of this study, particularly CGM and HCL system use, may not be representative of the larger CFRD population. However, the baseline use of CGM and pumps in this study is not far off from that reported in a recently published survey (75% and 29%, respectively) (31), supporting the overall generalizability of the results. Although current guidelines suggest use of the same CGM TIR goals for CFRD as type 1 and type 2 diabetes (7), these glycemic targets have not been specifically shown to improve cystic fibrosis–specific outcomes, such as pulmonary function and nutrition, and may not be generalizable to this population. Finally, participants with very low insulin requirements and those with HbA_1c <6.0% or mean CGM glucose <125 mg/dL were excluded. Further studies will be needed to understand how this device performs over a longer duration, including during periods of illness, and in people with CFRD who have a greater degree of residual β-cell function at baseline, as well as in children and adolescents.

In conclusion, adults with CFRD achieved a clinically significant increase in TIR and improved average CGM glucose with low and similar rates of sensor-measured hypoglycemia while using the iLet BP compared with their usual diabetes care over a 2-week period. These results provide strong evidence supporting the potential utility of AID technology in people with CFRD and support the need for larger and longer clinical studies investigating the safety and efficacy of this device in this and other diabetes populations.

Acknowledgments. The authors thank the study participants for their participation as well as the clinical and research staff of the Massachusetts General Hospital and Boston Children’s Hospital and Brigham and Women’s Hospital cystic fibrosis centers.

Funding. The study was supported by National Institute of Diabetes and Digestive and Kidney Diseases grants 1R01DK119699 and T32DK007028. M.S.P. and J.S.S. received support from the Cystic Fibrosis Foundation (Harry Shwachman Clinical Investigator Award, EnVision CF: Emerging Leaders in CF Endocrinology II Program, and Clinical Research Scholars Program). Dexcom supplied discounted CGM devices and sensors for the study.

Duality of Interest. After completion of the study, S.J.R., M.A.H., and C.A.B. became employees of Beta Bionics, Inc. S.J.R., E.R.D., and F.H.E.-K. are inventors on patents and patents pending related to BP technology. S.J.R., E.R.D., F.H.E.-K., M.A.H., and C.B. are employees and equity holders in Beta Bionics, Inc. E.R.D. is on the board of directors of Beta Bionics, Inc. M.S.P. has received in-kind support and research grants from Dexcom and serves as a consultant for Anagram Therapeutics. After completion of the study, J.S.S. became an employee of and holds stock in Vertex Pharmaceuticals. G.S.S. has received personal fees from Vertex Pharmaceuticals outside the submitted work. M.S.P., G.S.S., and A.U. have received research funding from Vertex Pharmaceuticals unrelated to this work. A.U. served on an advisory board for Vertex Pharmaceuticals and as an unpaid board member of the Cystic Fibrosis Research Institute. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. J.S.S. and M.S.P. researched data, contributed to the discussion, and wrote the first draft of the manuscript. L.E.C, M.Y.O., C.A.B., M.A.H., S.G.G., R.B., E.G., A.S., P.M., and S.J.R. researched data, contributed to the discussion, and reviewed and edited the manuscript. H.Z. analyzed study data and reviewed and edited the manuscript. A.U., G.S.S., I.N., F.H.E.-K., and E.R.D. contributed to the discussion and reviewed and edited the manuscript. All authors approved the final version of the manuscript. M.S.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented as an abstract at the 2022 North American Cystic Fibrosis Conference, Philadelphia, PA, 3–5 November 2022.