Glucose tolerance in patients with cystic fibrosis: five year prospective studyBMJ 1995; 311 doi: https://doi.org/10.1136/bmj.311.7006.655 (Published 09 September 1995) Cite this as: BMJ 1995;311:655
- Susanne Lanng, research fellowa,
- Annelise Hansen, nursea,
- Birger Thorsteinsson, consultantb,
- J(empty set)rn Nerupa,
- Christian Koch, consultanta
- a CF Center Copenhagen, Department of Paediatrics, Rigshospitalet, Copenhagen, Denmark
- b Department of Medicine F, Hiller(empty set)d Hospital, Hiller(empty set)d, Denmark
- Steno Diabetes Center, Gentofte, Denmark J(empty set)rn Nerup, professor. Correspondence to: Dr Lanng.
- Accepted 26 June 1995
Objectives: To study prevalence and incidence of diabetes mellitus in patients with cystic fibrosis.
Design: Five year prospective study with annual oral glucose tolerance tests.
Setting: CF Center Copenhagen, Denmark.
Subjects: 191 patients with cystic fibrosis aged above 2 years.
Main outcome measures: Glucose tolerance, plasma glucose concentrations after fasting and after glucose loading, and haemoglobin A1c levels.
Results: Prevalence of diabetes increased from 11% (n=21) to 24% (n=46) during study, with annual age dependent incidence of 4-9%. Diabetes was diagnosed at median age of 21 (range 3-40
Conclusions: Prevalence and incidence of diabetes in cystic fibrosis patients was high and increased with age. Since hyperglycaemic symptoms, fasting hyperglycaemia, and increased levels of glycated haemoglobin did not reliably identify diabetes mellitus, we recommend annual oral glucose tolerance tests in all cystic fibrosis patients aged over 10 years.
In this five year study of patients with cystic fibrosis we performed annual glucose tolerance tests and found cumulative incidence of diabetes of 24% in patients aged 20 and 76% in those aged 30
Since diabetes in cystic fibrosis is associated with deterioration in overall clinical status and since late diabetic complications may develop, these patients should be identified
Presence of hyperglycaemic symptoms, fasting hyperglycaemia, and increased levels of glycated haemoglobin did not reliably identify diabetes mellitus in patients with cystic fibrosis
Annual screening with oral glucose tolerance test is therefore recommended for all cystic fibrosis patients aged over 10
In the past three decades survival of patients with cystic fibrosis has steadily improved to a median of about 30 years in several centres.1 2 3 4 5 Following the increased life expectancy, glucose intolerance has been observed with increasing frequency in these patients.6 7 8 We conducted a five year prospective study of glucose tolerance in a large unselected group of patients with cystic fibrosis in order to define the incidence of diabetes mellitus in cystic fibrosis.
Measurement of glycated haemoglobin has been suggested as an alternative to the oral glucose tolerance test for assessing glucose metabolism, and serial testing of parameters of glycaemic control has been suggested for diagnosing diabetes and predicting progression to diabetes in high risk groups (such as subjects with impaired glucose tolerance,9 Pima Indians,10 pregnant women,11 patients with chronic pancreatitis,12 and patients with cystic fibrosis13 14). Wetherefore annually measured glycated haemoglobin concentrations as well as performing oral glucose tolerance tests to test whether screening with this parameter could identify cystic fibrosis patients with undiagnosed diabetes. We also studied whether impaired glucose tolerance was a risk marker for the development of diabetes in such patients. Finally, we studied whether those patients who had normal glucose tolerance at all five oral glucose tolerance tests had stable plasma glucose concentrations after fasting and glucose loading and stable glycated haemoglobin levels.
Subjects and methods
During 1989, 245 patients attended our cystic fibrosis centre. The diagnosis of cystic fibrosis was based on the presence of abnormal electrolyte concentrations in sweat and a typical clinical picture.15 16 We included all 226 patients aged over 2 years in our five year prospective study (1989-93) with at least one annual oral glucose tolerance test. The study was conducted in accordance with the Declaration of Helsinki and was approved by the municipal medical ethics committee of Copenhagen and Frederiksberg. We obtained informed consent from all participants or their parents.
The oral glucose tolerance test was carried out in all non-diabetic patients. After an overnight fast patients were given 1.75 g glucose monohydrate/kg body weight (maximum 75 g), dissolved in 200-300 ml lemon flavoured water, to be drunk within 3-4 minutes. During the test patients rested, fasted, and did not smoke. Capillary plasma glucose concentrations were measured before, and 60 and 120 minutes after, the glucose load with a glucose dehydrogenase method (Merck).17 According to World Health Organisation criteria, a capillary plasma glucose concentration </=8.8 mmol/l at 2 hours after the glucose load is normal, 8.9-12.1 mmol/l indicates impaired glucose tolerance, and >/=12.2 mmol/l indicates diabetes.18 When the test indicated diabetes we accepted the diagnosis only in patients with symptoms of hyperglycaemia
(polyuria, polydipsia, and loss in weight). In patientswith a diabetic test but without hyperglycaemic symptoms we repeated the test within one month in order to confirm the diagnosis. If diabetes was confirmed no further oral glucose tolerance tests were performed.
After the oral glucose tolerance test the patients' blood was sampled for measurements of glycated haemoglobin (haemoglobin A1c), liver function, and precipitins against Pseudomonas aeruginosa. Finally, we asked the patients (or their relatives) about the daily number of pancreatic enzyme capsules that they took (as a measure of the exocrine pancreatic function) and the use of any drugs with a potential risk of influencing glucose tolerance. Measurements of haemoglobin A1c,19 precipitins against P aeruginosa,20 and presence of the (delta)F508 mutation21 were as described elsewhere. Pulmonary function was assessed monthly in all patients aged over 6 years by forced vital capacity and forced expiratory volume in one second and recorded on an electronic spirometer (Spiroton, Draeger). The results are expressed as percentages of the reference values for sex and height.22
Statistical evaluation (with MEDSTAT statistical software23) included non-parametric tests (Kruskal-Wallis, Mann-Whitney, Fisher's, and the one-tailed Page test) and life table analysis. A probability of <0.05 (two-tailed) was accepted as significant.
A total of 191 of the 226 patients completed the study (table I). The other 35 patients were excluded because of incomplete data: 14 had moved to other centres, 17 had fewer than five oral glucose tolerance tests, one died (from acute myelocytic leukemia), two became pregnant, and one had a heart-lung transplant. Of these 35 patients, 26 had normal glucose tolerance, two had impaired glucose tolerance, four were diabetic at their last oral glucose tolerance test, and three had unknown glucose tolerance. All but one (with impaired glucose tolerance) of the 19 patients excluded at entry because they were under 2 years old hadnormal glucose tolerance at the latest oral glucose tolerance test during the study period.
During the study the prevalence of diabetes increased from 11% (n=21) to 24% (n=46) in the 191 cystic fibrosis patients (fig 1), with an average annual incidence of 3.8%. In the 131 patients aged 10 or more at entry the prevalence increased from 16% (n=21) to 34% (n=44) (average annual incidence 5.0%), while in the 47 patients aged 20 or more it increased from 25% (n=12) to 53% (n=25) (average annual incidence 9.3%) (fig 1). The prevalences of normal glucose tolerance decreased from 73% (n=139) to 58% (n=110) in the total study population, from 63% (n=82) to 48% (n=63) in patients aged >/=10 years, and from 43% (n=20) to 38% (n=18) in patients aged >/=20 years (fig 1). The prevalences of impaired glucose tolerance in the total population and in patients aged >/=10 years were relatively stable over the five years, 15 (13-18)% and 18 (15-21)%, respectively, whereas the prevalence of impaired glucose tolerance in patients aged >/=20 years decreased steadily from 32% (n=15) to 9% (n=4).
The incidence of diabetes increased with age. No patients aged less than 10 years developed diabetes during the study. Nine of the 75 patients who did not have diabetes at the start of the study and who were aged 10-19 developed diabetes, compared with 16 of the 35 patients aged 20 years or more (P<0.0005). The cumulative incidence of diabetes in the patients was 3%, 24%, and 76% at ages 10, 20, and 30 respectively (fig 2).
Of the 46 patients (25 males) who were diabetic at the end of the study, 25 (13 males) had had diabetes diagnosed during the five years. The median age at diagnosis of diabetes was 21 (range 3-40): 22 (3-40) in males and 17 (10-28) in females (P=0.089) (fig 3). In 15 patients diabetes was initially suspected on clinical grounds (polyuria, polydipsia, loss in weight, or exacerbation of lung infections) and was confirmed by an oral glucose tolerance test. Two male patients, who were diabetic at the start of the study, had developed diabetes before the age of 10 (at 3 and 8 years old). They had hyperglycaemic symptoms, including ketonuria, at the time of diagnosis, and their HLA types were DR3/4 and DR4. In 31 patients without symptoms of hyperglycaemia, diabetes was indicated by the annual oral glucose tolerance test, and the diagnosis was confirmed by a second test. Another 18 asymptomatic patients' annual glucose tolerance tests indicated diabetes, but tests repeated within a month indicated normal glucose tolerance in nine patients and impaired tolerance in nine.
Seven patients developed diabetes after the start of prednisone treatment for bronchopulmonary aspergillosis, after one month to four years of prednisone. Forty diabetic patients received insulin treatment during the study (with a median (range) dose of 0.19 (0.09-2.1 3)IU/kg/day). At the end of the study only 34 were treated with insulin (0.33 (0.11-1.62) IU/kg/ day). Four patients stopped insulin treatment because of withdrawal of prednisone, while the other two were unwilling to continue insulin treatment.
GLUCOSE TOLERANCE AND GLYCATED HAEMOGLOBIN LEVELS
At diagnosis of diabetes, the median (range) fasting plasma glucose concentration was 6.7 (4.1-9.0) mmol/l and the median haemoglobin A1c level was 5.8% (4.8-9.3%). Of the 25 patients who had diabetes diagnosed during the study, four had fasting plasma glucose concentrations above the upper normal limit of 7.8 mmol/l18 and four had haemoglobin A1c levels above the upper normal limit of 6.4%. Eight patients with normal glucose tolerance and two with impaired glucose tolerance had fasting plasma glucose concentrations >/=7.8 mmol/l; two of these patients with normal glucose tolerance and one with impaired glucose tolerance developed diabetes during the study. Two patients with normal glucose tolerance and two with impaired glucose tolerance had haemoglobin A1c levels >6.4%; both of the patients with impaired glucose tolerance developed diabetes during the study. Table II shows predictive values and odds ratios for impaired glucose tolerance, fasting hyperglycaemia, and haemoglobin A1c levels above the upper normal limit.
A total of 640 glucose tolerance tests were classified as normal, and 143 indicated impaired glucose tolerance. The corresponding figures for patients aged 10 or more were 376 and 115, and those for patients aged 20 or more were 96 and 40. No patient had impaired glucose tolerance at all tests, but 95 patients had impaired glucose tolerance at least once during the study, including 21 of the 25 patients who developed diabetes during the study. Of the 108 cases of impaired glucose tolerance found during the first four years of the study, the following oral glucose tolerance test indicated normal tolerance in 63 cases, impaired tolerance in 30 cases, and diabetes in 15 cases.
Only 71 patients had normal glucose tolerance at all five tests during the study. In these patients fasting plasma glucose concentrations increased insidiously during the study from 5.2 mmol/l to 5.6 mmol/l, as did plasma glucose concentrations after glucose loading, from 5.9 mmol/l to 6.3 mmol/l, and haemoglobin A1c levels, from 5.1% to 5.4% (table III). Throughout the study haemoglobin A1c levels were lower in the 71 patients with normal glucose tolerance at all tests (overall median level 5.2% (range 3.7-6.5%)) than in the 74 patients with variable glucose tolerance (normal or impaired, but not indicating diabetes) (overall level 5.3% (3.2-6.5%)), in the 25 patients who developed diabetes during the study (5.8% (4.1-11.0%)), and in the 21 patients who had diabetes at entry (6.5% (4.8-13.9%)) (P<0.00001).
In our study the patients had a median age of 13.6 years, and during thestudy the prevalence of glucose intolerance (impaired tolerance and tolerance indicating diabetes) increased from 27% to 42% and the prevalence of diabetes had increased from 11% to 24%. The annual incidence of diabetes was relatively constant, averaging 3.8% a year in the total population, and increased with age: 5.0% a year in patients aged 10 or more and 9.3% a year in patients aged 20 or more. Arrigo et al studied 32 patients with cystic fibrosis (mean age 12.8 years) who underwent oral glucose tolerance tests on two occasions two years apart.24 In their study the prevalence of glucose intolerance (impaired tolerance and tolerance indicating diabetes, WHO criteria18) increased from 37.5% to 50%. Two patients developed diabetes, corresponding to an incidence of 3% a year (confidence interval 1% to 21%). To our knowledge no other longitudinal studies of the incidence of diabetes in cystic fibrosis exist.
It is unknown whether the presence of diabetes affects survival of patients with cystic fibrosis.1 2 25 There is evidence, however, that the overall clinical status deteriorates gradually in diabetic1 6 and prediabetic26 patients with cystic fibrosis compared with non-diabetic patients and that insulin treatment of diabetic patients improves lung function and reduces the number of lung infections with Haemophilus influenzae and Streptococcus pneumoniae.27 Moreover, diabetic patients with cystic fibrosis are probably no less prone to developing late diabetic complications than are patients with other types of diabetes of similar duration and glycaemic control.8 It is therefore important to diagnose and treat diabetes in patients with cystic fibrosis and to identify those patients at risk of developing diabetes.
Since diabetes in cystic fibrosis is often asymptomatic (67% of newly diagnosed patients in our study), it is often underdiagnosed. Thus, the reported prevalences of diabetes requiring insulin treatment of 4.1% and 5.2% in large groups of American (median age 12.5 years)5 andCanadian (mean age 17.3)2 cystic fibrosis patients were lower than would have been expected from our findings: prevalences of diabetes of 11% at entry (median age 13.5) and 24% at exit (median age 17.3). Only 16% of our diabetic patients had abnormally high fasting plasma glucose concentrations (>/=7.8 mmol/l)18 and haemoglobin A1c levels (>6.4%) at the time of diagnosis of diabetes. Regular oral glucose tolerance tests are therefore mandatory to diagnose diabetes. We suggest that annual oral glucose tolerance tests are started as part of clinical practice in cystic fibrosis patients over the age of 10, as our two patients who developed diabetes below this age both had frank symptoms of diabetes at onset.
Identifying cystic fibrosis patients who are at risk of diabetes is desirable. As in non-insulin dependent diabetes,9 18 impaired glucose tolerance seems to be a risk marker for the development of diabetes in cystic fibrosis, with an age dependent odds ratio of 3-6 compared with normal glucose tolerance, and 21 of the 25 patients who developed diabetes during the study had impaired glucose tolerance at least once. However, since 58% of the tests that showed impaired glucose tolerance were followed the next year by tests showing normal tolerance (in accordance with the rate of reversion of impaired to normal glucose tolerance of 28-67% reported elsewhere9), the presence of impaired glucose tolerance is not a reliable predictor of diabetes in cystic fibrosis. From this and previous studies we can therefore conclude that cystic fibrosis patients at risk of developing diabetes cannot be reliably identified by the presence of impaired glucose tolerance, fasting hyperglycaemia, haemoglobin A1c levels above the upper normal limit, residual β cell function,28 cystic fibrosis genotype29 (though exocrine pancreatic insufficiency may have to be present for diabetes to develop in cystic fibrosis30 31 32), HLAtype,33 presence of islet cell cytoplasmic antibodies,24 33 or family history of diabetes.24 33 34
Only 37% of our patients were classified as normal at all five oral glucose tolerance tests. Within this group of patients plasma glucose concentrations, after both fasting and glucose loading, and haemoglobin A1c levels increased insidiously, supporting the theory that β cell function deteriorates progressively in cystic fibrosis patients.
The reason for the development of diabetes in cystic fibrosis is not fully understood. In cystic fibrosis patients who are not diabetic but have exocrine pancreatic insufficiency the number of islets of Langerhans is normal or slightly reduced,35 but in cystic fibrosis patients with diabetes the number of islets and β cells are reduced by 30-50%.35 36 37 The configuration of the islets is altered so that they exist in disorganised clusters separated by broad bands of fibrous tissue and fatty replacement and degeneration,*RF 6, 35, 36, 37
*which, together with impairment of blood supply, may contribute to the abnormal endocrine function. Since the discovery and cloning of the cystic fibrosis transmembrane regulator gene it has been possible to search for its expression at the mRNA and gene product levels in various tissues. Its mRNA is present in the centroacinar cells of the intercalated duct of the pancreas but not in the surrounding serous acini or in the islets of Langerhans.38 Correspondingly, the major mutation genotype in cystic fibrosis ((delta)F508) affects the severity of the exocrine pancreatic insufficiency, whereas the endocrine pancreatic function is unrelated to this genotype.29
Our longitudinal study of unselected cystic fibrosis patients has shown that glucose tolerance deteriorates gradually, with an age dependent annual incidence of diabetes of 4-9%, which resulted in a doubling of the prevalence of diabetes to 24% within five years. Cystic fibrosis patients at risk of diabetes cannot be identified with certainty, although patients with impaired glucose tolerance are at higher risk than patients with normal glucose tolerance. Since all cystic fibrosis patients with diabetes cannot be identified by the presence of hyperglycaemic symptoms, increased fasting plasma glucose concentrations, or increased haemoglobin A1c levels, we recommend that annual oral glucose tolerance tests be carried out in all cystic fibrosis patients over the age of 10.
We thank the staff at the CF Department 5063, Rigshospitalet, for their skilled assistance.
Conflict of interest None.