Glucose tolerance in Canadian and French cystic fibrosis adult patients

Cystic fibrosis (CF)-related diabetes is associated with increased mortality. We analysed the clinical and glycemic profiles of two cohorts of patients treated according to the same guidelines in France and Canada. To investigate incidence differences in phenotypic and glucose abnormalities and to explore the evolution over a 4-year follow-up period, two cohorts of 224 Canadian and 147 French adult CF patients (≥18 years) without treated CF-related diabetes (CFRD) were followed over a 4 year period. In each of these groups, we investigated the longitudinal relationship between glucose tolerance and pulmonary function. An annual 2-hour oral glucose tolerance test was performed: fasting blood glucose (G0) and 2-h blood glucose (G2) were measured. Patients were classified at inclusion according to their glucose tolerance status: Normal glucose tolerant, abnormal glucose tolerant or de novo CFRD. Age, sex ratio and proportion of F508del homozygous patients were not statistically different between both cohorts. Canadian patients had better pulmonary function (median %FEV1 (IQR): 71.0 (55.0–82.0) vs. 64.0 (40.0–78.0), p < 0.001) and greater body mass index (BMI; median BMI in kg/m2) (IQR) 21.1 (19.5–22.8) vs. 19.9 (18.4–21.4), p < 0.001). Glucose values: G0 (5.4 (5.0–5.9) vs. 4.8 (4.5–5.1) mmol/L, p < 0.001) and G2 (7.6 (5.8–9.7) vs. 6.5 (5.2–8.5) mmol/L, p = 0.001) were higher in the Canadian cohort translating into a higher incidence of de novo CFRD diagnosis (19.2 vs. 9.8%, p = 0.003). Decline in FEV1 over time was not different between patients according to glucose tolerance groups. Despite higher glucose levels and incidence of de novo CFRD, Canadian CF patients have a better lung function and a higher BMI than French patients. In spite of these differences between the cohorts, the decline in FEV1 in patients with abnormal glucose tolerance is similar between these groups.

www.nature.com/scientificreports www.nature.com/scientificreports/ Studies have compared demographic data of European and North American CF children patients using CF registries 12,13 , suggesting differences in demographic data (height and weight) between populations. No data are available regarding difference in CFRD prevalence between Europe and North America except data from annual registry reports 14 . Furthermore, registry data concerning specific comorbidities like CFRD are not homogeneously collected, and most of the time registries do not collect data about other glucose metabolic abnormalitites. It is thus possible that both prevalence of CFRD and its impact on clinical CF condition differ between these two populations. The consequences of different glucose tolerance profiles on respiratory function have not been evaluated in large cohorts and over long periods. Similarly, the potential role of insulin deficiency and insulin resistance have not been described and compared in large cohorts [15][16][17] .
The objective of this study was to compare the following parameters between the Canadian and French cohorts: (1) clinical characteristics, (2) glucose and insulin values as well as CFRD incidence and (3) evolution of pulmonary function over a 4-year follow-up period.

Methods
Study population. Data were obtained from the Montreal Cystic fibrosis Cohort (MCFC) for Canada and from the Lyon Cystic Fibrosis Cohort (DIAMUCO) for France. Both research cohorts are investigating mechanisms of glucose intolerance as well as associations between glucose intolerance and clinical outcomes in adult patients with CF. The Montreal Cystic Fibrosis Cohort was established in 2004 and all available data at inclusion from patients included between 2004 to 2016 were considered for this analysis. The DIAMUCO Cohort was established in 2009 and all available data at inlusion from patients included between 2009 to 2012 were considered for this analysis. All available follow-up pulmonary function data were included in both cohorts over a 4-year period.
Informed consents have been obtained for all subjects included. The institutional review board of each participating hospital and research ethics board authorized the cohorts in accordance with the current ethical standards (Comité de Protection des Personnes in France, Comité d' éthique de la recherche in Canada), as well as the French data Protection Agency for DIAMUCO Cohort (Comission Nationale de l'Informatique et des Libertés CNIL).
For both cohorts, all adult patients (>18 years) with CF, pancreatic insufficiency and no previous history of treated CFRD were included. Patients who received a de novo CFRD diagnosis during recruitment were however included. Main exclusion criteria were pancreatic sufficiency and previous pulmonary transplantation. In patients with ongoing pregnancy, pulmonary exacerbation or current treatment with medication known to interfere with glucose metabolism (e.g. high dose oral steroids and enteral tube feeding), OGTT was delayed upon resolution of this temporary exclusion factor. OGTT results were accepted in patients taking stable, long-term (≥1 year) low-dose oral corticosteroids (maximal dose of 10 mg prednisone per day). At the time of the OGTT, all patients were clinically stable with no recent (<1 month) pulmonary exacerbation or symptoms of acute infection. Included patients represent around 80% of all patients with CF followed at our respective clinics.
Clinical and biological data. In both cohorts, a harmonized data collection process was used to extract required information from medical charts at inclusion. This included age, sex, CF related genotype, chronic bacteriological colonization and number of intravenous antibiotics courses in the previous year. Chronic bacteriological colonization was defined as follow: 50% or more of samples being positive for a specific bacteria in the preceding 12 months 18 . Other data such as BMI, FEV1, glycosylated hemoglobin (HbA1c) were also collected at inclusion and then between 12 and 18 months during a 4-year follow-up for FEV1. BMI was calculated using weight in kilograms divided by height in square meter (kg/m 2 ). Pulmonary function was measured by spirometry using the forced expiratory volume in 1 second in L (FEV1) using Hankinson 1999 formula for FEV1(%) 19 . HbA1c was measured using HPLC automate variant II (Biorad) in France, and immunotubidimeter (Bayer Health Care diagnosis) in Canada. Both HbA1c are aligned with international standards.
Oral glucose tolerance test (OGTT). OGTT was realized at the inclusion in both cohorts and was carried out after an overnight fast. Patients were given glucose (1.75 g per kg bodyweight, maximum 75 g) then plasma glucose was measured at start (G0), 1 hour (G1) and 2 hour (G2) 20 . Patients diagnosed with de novo CFRD underwent within 2 months a second OGTT to confirm the diagnosis. Plasma insulin was also measured at start (I0), 1 hour (I1) and 2 hour (I2) by immunoradiometric assay (BI-INS-IRMA, Cisbio Bioassays, France) and measures were centralized in Quebec, to obtain comparable values.

Statistical methods.
Values are expressed as median (interquartile range [IQR]) or percentage, as appropriate. Demographic and clinical data, insulin and glucose values from both cohorts were compared at inclusion of patients of each cohort, using non parametric tests (Chi 2 or Mann-Whitney tests, as appropriate). Subgroups analyses determined by glucose tolerance subgroup classification were then performed to compare clinical status, glucose and insulin values of both cohorts using non parametric tests (Chi 2 or Mann-Whitney tests, as appropriate). Correlation analysis between BMI and pulmonary function (FEV1), AUC glycemia and AUC insulin were made with Spearman's correlation. A linear mixed regression model with random intercept and random slope was fitted to assess the effect of glucose tolerance subgroup at entry in the cohort on the mean slope of www.nature.com/scientificreports www.nature.com/scientificreports/ FEV1 change over 4 years. Effect of glucose tolerance subgroup was controlled for the cohort and age. Interaction between covariates (cohort, NGT and AGT) and time were tested to characterize differences in longitudinal rates of change. The relationship between FEV1 decline and CFRD subgroup (56 patients) was not analyzed due to the small sample size and no available clinical data in the canadian group since confirmed de novo CFRD patients are excluded from the cohort after diagnosis.
Analyses were performed using SPSS software (version 24 by IBM, Chicago, USA) and SAS ® software (version 9.4, SAS ® Institute Inc., Cary, NC, USA). Area under the curve (AUC) for glucose and insulin was calculated using the software GraphPad Prism (GraphPad Software Inc; CA, USA). A probability value ≤ 0.05 was considered as statistically significant.

Results
Characteristics at baseline. Data of 224 Canadian and 147 French patients were included (See Table 1).
Demographic and clinical data are detailed in Table 1. No difference of sex ratio, proportion of F508del homozygous patients and age was observed between the 2 cohorts. The clinical status of Canadian group was better with higher BMI (median in kg/m 2   Glucose values according to glucose tolerance group. For each glucose tolerance category (see Table 2), Canadian patients displayed higher G0 median values and higher glucose median AUC as compared to French patients: p < 0.001 and p = 0.005 in NGT group, p < 0.001 and p < 0.001 in AGT group, and p = 0.005 and p = 0.006 in CFRD group.
Insulin values according to glucose tolerance group. NGT and AGT Canadian patients displayed higher insulin median AUC than French patients (p < 0.001 and p = 0.040), but this difference no longer exists for patients with de novo CFRD, p = 0.278 (see Table 2).
Insulin sensitivity (Stumvoll index) was higher for NGT French patients (p = 0.039) compared to Canadian NGT patients, but no difference was observed in the 2 other subgroups. Insulin resistance (HOMA-IR) was higher for Canadian NGT and AGT patients (p = 0.004 and p = 0.048) compared to French patients.
Despite higher level of insulin values for Canadian patients, the trends of the curve of insulin profile is similar for both Canadian and French NGT and AGT patients. Patients of both cohorts with NGT present a plasma insulin rise during the first hour of the OGTT (Fig. 1a), followed by moderated reduction at 2-h when glucose levels are trending downward. For all AGT patients, a similar insulin profile as for NGT-patients is observed for the 1 st hour of the OGTT (Fig. 1b), but then insulin levels remain high at the second hour. For CFRD patients (Fig. 1c), if insulin values are similar at the end of the test for both cohorts, it is slightly higher in Canadian patients at 1-h. However, the 1 h insulin peak observed in NGT and AGT patients is reduced by approximatively 30% for CFRD patients. Insulin values keep rising during the second hour of the test for Canadian and French CFRD patients. After controlling for cohort, we observed a significant positive correlation (Fig. 2) between BMI and pulmonary function (FEV1) for all glucose tolerance groups (respectively p < 0.001, p = 0.001, p < 0.001 for NGT, AGT and CFRD). No significant correlation was observed for all subgroups between BMI and AUC glycemia.

Insulin sensitivity and resistance and correlation analysis between BMI and pulmonary function, and between
www.nature.com/scientificreports www.nature.com/scientificreports/ Concerning BMI and AUC insulin, a significant correlation was observed in NGT subgroup (p = 0.004) but not in AGT and CFRD patients. No significant correlation between pulmonary function (FEV1) and AUC glycemia were observed for all glucose tolerance groups (data not shown).

Discussion
To our knowledge, this is the first study that compares the incidence of glucose abnormalities in adult patients with CF, in two large cohorts and their association with clinical status. Despite higher glucose levels and incidence of de novo CFRD, Canadian CF patients secrete more insulin and have a better pulmonary and nutritional status according to their FEV1 and BMI than French patients. To explore the mechanisms of these differences, we conducted correlation analyses but we observed no correlation between BMI and glycemia AUC. Canadian patients also have slower annual decrease of their FEV1 than French patients during follow-up. No significant difference was observed for pulmonary function change over time between AGT and NGT patients. These observations challenge the concept of a possible causal role of hyperglycemia and/or hypoinsulinemia favouring clinical status (BMI and/or FEV1) degradation.
In the context of a well-established limited insulin secretion in adult CF patients 21,22 , three key factors could contribute to the development of hyperglycemia: progression of insulin secretory deficiency, higher insulin resistance or higher insulin requirements 22,23 . In Canadian patients, we observe higher insulin secretory capacity, higher estimated insulin resistance but also higher CFRD incidence. Recently, insulin resistance variations has  www.nature.com/scientificreports www.nature.com/scientificreports/ emerged as a possible contributor to hyperglycemia for patients with CF 8,24 . Higher BMI could contribute to higher insulin resistance in Canadian patients 25,26 . As insulin is an anabolic hormone, as long as a certain degree of insulin secretion is preserved, this could allow a higher BMI 27 . Indeed, when insulin secretion further deteriorates leading to de novo CFRD, both cohorts do not present anymore differences for BMI and insulin secretion. The frequency of exacerbations which is a good marker of respiratory function stability does not seem to be implicated in observed insulin resistance differences as the number of antibiotic courses is similar in both cohorts. It is also possible that insulin secretion itself might explain clinical differences between cohorts rather than blood glucose. A higher insulin secretion level might allow to reach and/or maintain a higher weight 28 . In a context of a limited insulin secretion capacity, hyperglycemia could also play a role by favoring pulmonary exacerbations 29 and promoting oxidative stress 30 . Both higher insulin as well as lower glucose values can thus contribute to better lung function.
In contrast to previous reports, our results highlight that despite higher glucose values, Canadian CF patients have better pulmonary function and BMI. Despite these higher glucose values, Canadian patients also presented a slower annual FEV1 decline. When both cohorts are combined and the pulmonary evolution of patients with AGT is compared to patients with NGT, there is no difference in annual FEV1 decline. Various factors not related   www.nature.com/scientificreports www.nature.com/scientificreports/ to cystic fibrosis can also explain this observation. Among them, the role of genetic factors other than CFTR mutations may explain the observed differences between these two cohorts (modifier genes). In the present study, the distribution for F508del CFTR mutation proportions is similar between the two cohorts but it is now well established that other mutations are associated, for some of them, with lung disease severity and for others, with diabetes susceptibility 31,32 . This emerging important factor may also have a role in observed glucose and BMI differences. For example, some mutations in genes involved in a higher risk for type 2 diabetes (e.g. TCF7L2) are also associated with a higher and earlier CFRD prevalence 32 . Further investigations need to be done to compare modifier genes between the 2 cohorts. Secondly, despite comparable nutritional and clinical recommendations for CF care as well as health care systems (universal access) between the 2 cohorts, other factors can still influence glucose values as well as the nutritional status of CF-patients. For example, differences in qualitative and quantitative nutritional intake could impact both glucose tolerance and BMI. Unfortunately, nutritional intake was not assessed in the present study. If done in the future, such assessment could be limited by the precision of available tools (e.g. food journals, 24-hour dietary recall, etc.) as well as the fact that most tools are country specific thus

C F R D O G T T tim e p o in t (h o u rs )
C a n a d ia n p a tie n ts F re n c h P a tie n ts www.nature.com/scientificreports www.nature.com/scientificreports/ limiting the ability to compare different populations 33 . In addition, despite merging two large and well characterized cohorts, the sample size of some subgroups, such as de novo CFRD, remains small thus limiting our ability to explore some differences.
In order to interpret our results, the potential mechanisms of higher BMI in Canadian patients should be explored. Key factors involved in energy balance and nutritional intake, absorption and energy expenditure (physical activity, energy demand related to CF exacerbations, etc.) should be evaluated. Nutritional recommendations are similar in North America and Europe with a recommended energy intake range from 120 to 150% of energy needs for the healthy population of similar age, sex and size. Patients included in both countries are also exposed to similar pancreatic enzyme replacement therapy (PERT) protocol, starting at 500 U lipase/kg/meal to a maximal dose of 1000-2500 U lipase/kg/meal which should lead to similar nutrient absorption capacity. Thus the two cohorts should be exposed to similar quantitative nutritional intake and absorption. The frequency of exacerbations necessitating intravenous antibiotics does not seem to play a role on BMI values as the number of antibiotics course is similar in both cohorts. International guidelines for antibiotic use in CF are worldwide applied and this may contribute to the very close use of antibiotics in Canadian and French cohorts. However, chronic higher caloric intake and/or differences in physical activity may still be important factors in explaining the differences and may play a role in the higher BMI of Canadian patients. In addition, backgroud population differences in diabetes and obesity, which are both higher in Canada compared to France [34][35][36][37] , could also explain the disparities between French and Canadians independently of CF status. As previously reported herein, there is a positive correlation between BMI and FEV1 which could explain the higher FEV1 observed in Canadian patients as well as their lower mean annual FEV1 decline.
Observed association and differences do not imply causality and despite our careful assessement of two large and well characterized cohorts important underlying mechanistic factors were not measured in that study.
In conclusion, Canadian patients present a better clinical status (higher BMI, insulin secretion and FEV1) than French patients, but unexpectedly they also present a higher incidence of glucose abnormalities. In addition, patients in the abnormal glucose tolerance group do not have worse mean FEV1 decline over observed time than patients with normal glucose tolerance. To better understand the complex interplay between glucose tolerance and clincal status (BMI and/or FEV1) of adult patients with CF, further investigations should focus on potential underlying factors that may play a role in the observed differences.