Available online at
ScienceDirect www.sciencedirect.com Diabetes & Metabolism 41 (2015) 116–125
Original article
Frequency and predictors of confirmed hypoglycaemia in type 1 and insulin-treated type 2 diabetes mellitus patients in a real-life setting: Results from the DIALOG study B. Cariou a,∗ , P. Fontaine b , E. Eschwege c , M. Lièvre d , D. Gouet e , D. Huet f , S. Madani g , S. Lavigne h , B. Charbonnel a a
Department of Endocrinology, l’institut du Thorax, CHU de Nantes, hôpital Guillaume & René Laennec, boulevard Jacques-Monod, 44093 Nantes cedex 1, France b Department of Endocrinology and Diabetology, University Hospital of Lille, Lille, France c Inserm U-1018, Centre de Recherche en Epidémiologie et Santé des populations (CESP), Villejuif, France d Laennec Faculty of Medicine, Lyon, France e Hôpital Saint-Louis, centre hospitalier de La Rochelle, La Rochelle, France f Hôpital Saint-Joseph, Paris, France g Novo Nordisk, Paris, France h ITEC Services SAS, Cenon, France Received 13 June 2014; received in revised form 21 October 2014; accepted 21 October 2014 Available online 24 November 2014
Abstract Aim. – DIALOG assessed the prevalence and predictors of hypoglycaemia in patients with type 1 (T1DM) or insulin-treated type 2 diabetes mellitus (T2DM) in a real-life setting. Methods. – In this observational study, insulin-treated patients (n = 3048) completed prospective daily questionnaires reporting the frequency and consequences of severe/confirmed non-severe hypoglycaemia over 30 days. Patients (n = 3743) also retrospectively reported severe hypoglycaemia over the preceding year. Results. – In this prospective survey, 85.3% and 43.6% of patients with T1DM and T2DM, respectively, reported experiencing at least one confirmed hypoglycaemic event over 30 days, while 13.4% and 6.4%, respectively, reported at least one severe event. Hypoglycaemia frequency increased with longer duration of diabetes and insulin therapy. Strongly predictive factors for hypoglycaemia were previous hypoglycaemia, > 2 injections/day, BMI < 30 kg/m2 and duration of insulin therapy > 10 years. HbA1c level was not predictive of hypoglycaemia in either T1DM or T2DM. The confirmed hypoglycaemia rate was increased in the lowest compared with the highest tertile of HbA1c in T1DM, but not T2DM. At the time of enrolment, physicians reported severe hypoglycaemia in 23.6% and 11.9% of T1DM and T2DM patients, respectively, during the preceding year; the retrospective survey yielded frequencies of 31.5% and 21.7%, respectively. Also, severe hypoglycaemia led to medical complications in 10.7% and 7.8% of events in T1DM and T2DM patients, respectively, over 30 days. Conclusion. – Using a unique combined prospective and retrospective approach, the DIALOG study found a relatively high frequency of hypoglycaemia among insulin-treated patients. These findings emphasize the importance of a patient-centred approach for managing diabetes in which hypoglycaemia risk evaluation is critical. Trial registration. – ClinicalTrials.gov: NCT01628341. © 2014 Elsevier Masson SAS. All rights reserved. Keywords: Hypoglycaemia; Type 1 diabetes; Type 2 diabetes; Observational study Abbreviations: ACCORD, Action to Control Cardiovascular Risk in Diabetes; ADA, American Diabetes Association; BMI, Body Mass Index; CRF, Case Report Form; CVD, Cardiovascular Disease; DCCT, Diabetes Control and Complications Trial; GP, General Practitioner; HbA1c , Glycated Haemoglobin; HCP, Healthcare Professional; SMBG, Self-Measured Blood Glucose; T1DM, Type 1 Diabetes Mellitus; T2DM, Type 2 Diabetes Mellitus; UKPDS, United Kingdom Prospective Diabetes Study. ∗ Corresponding author. Tel.: +33 253 482 707; fax: +33 253 482 708. E-mail address: [email protected] (B. Cariou). http://dx.doi.org/10.1016/j.diabet.2014.10.007 1262-3636/© 2014 Elsevier Masson SAS. All rights reserved.
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
117
1. Introduction
2.2. Study setting and assessments
Iatrogenic hypoglycaemia is an acute complication of insulin treatment in type 1 (T1DM) and type 2 diabetes mellitus (T2DM) patients, but occurs more frequently in insulin-deficient states (T1DM) and advanced T2DM [1]. Most of the available reports on the incidence of hypoglycaemia are from clinical trials such as the diabetes control and complications trial (DCCT) for T1DM [2–4], and the United Kingdom Prospective diabetes study (UKPDS) [5] and action to control cardiovascular risk in diabetes (ACCORD) study for T2DM [6]. Both the DCCT and UKPDS demonstrated that the risk of hypoglycaemia increases as glycaemic control improves [3,5]. However, it is difficult to accurately assess the frequency of hypoglycaemia in real life: most of the estimates that have been published were relatively small prospective studies, or larger studies based on retrospective data from registries or databases [7–13]. Severe hypoglycaemia is associated with increased risks of morbidity and mortality, especially in patients with T2DM and high cardiovascular risk, and in the elderly [14,15]. Recently, accumulating evidence has indicated that hypoglycaemia is a marker of vulnerability especially in older patients with T2DM [6,16–18]. The link between severe hypoglycaemia and an increased risk of death was shown in a retrospective epidemiological analysis of the ACCORD study, although direct causality was not demonstrated [19]. Moreover, health-related quality of life is also decreased as the frequency of hypoglycaemia increases [14,20]. Consequently, hypoglycaemia is now one of the major considerations when adopting a patient-centred approach to the management of T2DM [21]. Nevertheless, there is a need for dedicated studies assessing the frequency of hypoglycaemia in a “real-life” setting. DIALOG was a large nationwide observational study using an original approach that combined both prospective and retrospective evaluations of the frequency of confirmed hypoglycaemia, as rigorously defined by American Diabetes Association (ADA) criteria [22,23]. The objectives of DIALOG were to evaluate the prevalence of severe and non-severe confirmed hypoglycaemia in a real-life setting among patients with T1DM and insulintreated T2DM in France, and to define predictive factors for hypoglycaemia. DIALOG is, to our knowledge, the first study conducted both prospectively and retrospectively in a very large (> 4000 patients) non-clinical-trial population.
All 2300 specialist physicians (1700 endocrinologists/diabetologists and 600 internists) who manage patients with diabetes in France were invited to take part in DIALOG, and 474 (20.6%) agreed to participate. Of a random representative sample of 8000 general practitioners (GPs) treating patients with diabetes, 725 (9.1%) agreed to participate (Appendix A; see supplementary material associated with this article online). Participating physicians were required to include consecutive patients who fulfilled the selection criteria and were seen within a predefined period (between 2 May and 28 June 2012). Investigators completed a case report form (CRF) for each patient, which included the physician’s record of the patient’s characteristics, diabetes history and number of severe hypoglycaemia events experienced within the last 12 months. Investigators also completed a non-inclusion registry detailing reasons for excluding any patients. Investigators provided patients with information on hypoglycaemia, and two self-administered questionnaires (Appendix A; see supplementary material associated with this article online) with instructions on how to complete them at home. Patients answered the prospective questionnaires every day for 30 days, recording whether they had experienced hypoglycaemia and responding to questions about each episode (including blood glucose level, if measured, and characteristics and consequences of the event). With the retrospective questionnaire, patients answered questions concerning severe hypoglycaemic episodes during the previous year.
2. Methods 2.1. Study design and conduct This was a multicentre, prospective and retrospective, observational study. The study protocol and patients’ information sheet were approved by the appropriate authorities as required by French regulations. The study was conducted in accordance with the principles of the declaration of Helsinki, guidelines for good pharmacoepidemiology practices [24] and regulatory requirements. Patients signed an information notice before being enrolled into the study.
2.3. Inclusion and exclusion criteria Patients aged ≥ 18 years with either T1DM or T2DM treated with insulin for at least 12 months who were able to self-monitor their blood glucose levels and fill in the questionnaires were eligible. Pregnant or breastfeeding women, or those with a pregnancy within the previous year, were excluded. 2.4. Study outcomes For the prospective survey, the primary endpoint was the percentage of insulin-treated patients who had at least one hypoglycaemic event (severe or confirmed non-severe) during the 30 days following enrolment. Secondary endpoints included hypoglycaemic events by type. Also, rates of hypoglycaemia were calculated as events/patient/month, and factors predictive of hypoglycaemia were identified. The endpoint for the retrospective survey was severe hypoglycaemic episodes during the preceding year. Medical consequences of severe hypoglycaemia were also recorded by both questionnaires. 2.5. Definition of hypoglycaemia During data entry and analysis, hypoglycaemic episodes were defined in accordance with ADA definitions for severe, documented symptomatic, asymptomatic, probable and
118
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
pseudohypoglycaemia [22,23] (Appendix A; see supplementary material associated with this article online). The main endpoint of the DIALOG study was the combination of confirmed hypoglycaemia, whether symptomatic or asymptomatic, and severe hypoglycaemia (Fig. S1; see supplementary material associated with this article online). Nocturnal hypoglycaemia was defined as any confirmed episode with time of onset between 00h00 and 06h00. 2.6. Statistical analysis It was calculated that to arrive at an estimate of the number of hypoglycaemic events in T1DM and T2DM patients with a precision of at least 3%, a total of 3811 patients were required (Appendix A; see supplementary material associated with this article online). Three analytical sets were defined (Fig. S2; see supplementary material associated with this article online). The overall set included all patients in the survey except for any who had a major deviation (such as a violation of one of the inclusion/exclusion criteria). The prospective set included all patients with a completed inclusion CRF with no major deviations and with an analysable prospective questionnaire. The main criteria for a questionnaire to be considered analysable were a record of at least one hypoglycaemic episode or completion of at least 25 days of the questionnaire (see Appendix A for the full definition of “analysable”; supplementary material associated with this article online). The retrospective set included all patients with a completed inclusion CRF with no major deviations and an analysable retrospective questionnaire. 2.7. Statistical techniques Quantitative variables with a normal distribution were analysed using Student’s t test. Qualitative variables were compared using a Chi2 test or Fisher’s exact test. Predictive factors for the presence of a hypoglycaemic event during the prospective 1-month follow-up were identified using a logistic regression model. First, univariate analysis tested the relationship between each baseline parameter and the relevant variable, using the Chi2 test (or Fisher’s exact test) for qualitative data, and Student’s t test [or analysis of variance (ANOVA)] for continuous variables. All variables significant at a 20% level were included in the second step, a multivariate analysis using a stepwise logistic regression model with a significance level set at 5%. 3. Results 3.1. Disposition of patients A total of 4444 patients were included (74% by 370 specialists and 26% by 320 GPs); 20 patients were excluded because of major deviations. Among the final 4424 included patients, 3048 questionnaires from the prospective set and 3743 questionnaires from the retrospective set were evaluable (Fig. S2; see supplementary material associated with this article online).
Table 1 Demographic data for patients in the Overall Set. T1DM (n = 1915) Male, n (%) 978 (51.2) Age, years 48.4 ± 16.2 111 (5.8) Patients ≥ 75 years old, n (%) 25.4 ± 4.4 Body mass index (BMI), kg/m2 Patients with BMI > 30 kg/m2 , 249 (13.0) n (%) HbA1c , % 7.9 ± 1.3 (63 ± 14) (calculated mmol/mol) Diabetes history 20.4 ± 12.9 Time since diagnosis, years Duration of insulin use, years, n (%) 10 1315 (68.8) Experienced severe 477 (25.0) hypoglycaemia during previous 12 months, n (%)a Number of severe episodes during previous 12 monthsa Mean ± SD 3.3 ± 8.3 2.0 (0, 135) Median (range) Complications of diabetes, n (%) Complications present 855 (44.6) 142 (7.4) Peripheral arterial disease Coronary disease 180 (9.3) 48 (2.5) Cerebrovascular disease Retinopathy 595 (31.1) 145 (7.6) eGFR (MDRD) < 60 mL/min Micro- or 307 (16.0) macroalbuminuria Diabetic foot ulcer 73 (3.8) Comorbidities 676 (35.3) Hypertension 687 (35.9) Dyslipidaemia Insulin regimen at enrollment Insulin pump 598 (31.2) NR 1 injection NR 2 injections 1 or 2 injections 214 (11.2) 1090 (56.9) > 2 injections Oral antidiabetic drugs at enrolmentb Metformin NA DPP-4 inhibitor NA NA Sulphonylurea and/or glinide Alpha-glucosidase NA inhibitor GLP-1 agonist NA
T2DM (n = 2509) 1352 (54.0) 66.3 ± 11.0 596 (23.8) 30.9 ± 5.9 1316 (52.5) 7.9 ± 1.2 (63 ± 13) 17.3 ± 9.3
986 (39.3) 915 (36.5) 606 (24.2) 329 (13.1)
2.1 ± 2.4 1.0 (1, 31) 1597 (64.9) 444 (17.7) 614 (24.5) 144 (5.7) 623 (24.8) 564 (22.5) 781 (31.1) 141 (5.6) 2025 (80.7) 1900 (75.7) 122 (4.9) 953 (38.0) 477 (19.0) NR 945 (37.7) 1387 (55.3) 382 (15.2) 823 (32.8) 60 (2.4) 187 (7.5)
Values are means ± standard deviation unless otherwise stated; T1DM: type 1 diabetes mellitus; T2DM: type 2 diabetes mellitus; eGFR (MDRD): estimated glomerular filtration rate (Modification of diet in renal disease); DPP-4: dipeptidyl peptidase-4; GLP-1: glucagon-like peptide-1; NR, not reported; NA: not applicable. a As reported by physician during examination. b As reported for T2DM patients only.
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
119
Table 2 Hypoglycaemia in the prospective set (n = 3048). Type 2 diabetes mellitus (n = 1731)
Type 1 diabetes mellitus (n = 1317) Frequency n (%)a
Allb Severee Overall non-severee Non-severe symptomatic Non-severe asymptomatic Nocturnal Severe nocturnal Diurnal Probable hypoglycaemiaf Pseudohypoglycaemiaf a b c d e f
1123 (85.3) 177 (13.4) 1112 (84.4) 1085 (82.4) 367 (27.9) 529 (40.2) 55 (4.2) 1089 (82.7) 521 (39.6) 289 (21.9)
Events/patient/month (mean ± SD) R1c R2d 7.4 0.2 7.1 6.1 1.1 0.8 0.1 6.5 1.2 0.6
± ± ± ± ± ± ± ± ± ±
7.48.6 ± 7.3 0.71.6 ± 1.3 7.38.5 ± 7.1 6.47.4 ± 6.3 3.03.8 ± 4.6 1.21.9 ± 1.3 0.31.2 ± 0.5 6.97.9 ± 6.9 2.53.0 ± 3.3 1.92.6 ± 3.2
Median (range)
5.0 (0, 51) 0.0 (0, 12) 5.0 (0, 48) 4.0 (0, 45) 0.0 (0, 39) 0.0 (0, 10) 0.0 (0, 3) 4.1 (0, 49) 0.0 (0, 30) 0.0 (0, 29)
Frequency n (%)a
755 (43.6) 110 (6.4) 722 (41.7) 682 (39.4) 134 (7.7) 190 (11.0) 21 (1.2) 704 (40.7) 472 (27.3) 373 (21.5)
Events/patient/month (mean ± SD)
1.7 0.1 1.6 1.5 0.2 0.2 0.0 1.5 1.1 0.7
± ± ± ± ± ± ± ± ± ±
3.34.0 ± 4.0 0.41.5 ± 0.9 3.23.9 ± 3.9 2.93.7 ± 3.6 0.92.4 ± 2.5 0.71.7 ± 1.4 0.11.1 ± 0.5 2.93.7 ± 3.6 3.64.2 ± 5.9 2.43.3 ± 4.4
Median (range)
0.0 (0, 29) 0.0 (0, 6) 0.0 (0, 28) 0.0 (0, 25) 0.0 (0, 19) 0.0 (0, 9) 0.0 (0, 3) 0.0 (0, 26) 0.0 (0, 30) 0.0 (0, 30)
Patients reporting at least one episode. Severe or confirmed non-severe hypoglycaemia for same patient. Rate of events including patients with no episodes. Rate of events in patients with one or more episodes. As individual patients could report both severe and non-severe events, the sum of these two categories is higher than the All value. Not included when identifying episodes of hypoglycaemia for main criterion.
3.2. Baseline characteristics Baseline data for the overall set are shown in Table 1, which are similar to characteristics of the prospective set (Table 1 in Appendix A; see supplementary material associated with this article online). To ensure that selection bias was avoided, investigators also recorded the characteristics of the non-included patients (n = 3066), who were comparable to the analysable patients in terms of age, gender and diabetes duration (data not shown), although the proportion of patients with T1DM was smaller among the non-included vs included patients (30.7% vs 43.3%, respectively), and a smaller proportion reported a history of hypoglycaemia (61.7% vs 88.3%, respectively). 3.3. Frequency of confirmed hypoglycaemia in the prospective survey The proportion of all patients reporting at least one hypoglycaemic episode during the 30-day follow-up was 61.6% for the prospective set (1878 of 3048 patients), with 85.3% for T1DM (event rate/month of 7.4 ± 7.4) and 43.6% for T2DM (event rate/month of 1.7 ± 3.3). Table 2 shows the results for the different categories of hypoglycaemia, presented as mean (SD) values for all patients and also calculated for only those patients with at least one episode of hypoglycaemia. Median (range) values are also shown. The distribution by number of hypoglycaemic episodes per patient per month are shown in supplementary figures (Appendix A; see supplementary material associated with this article online). Among T1DM patients, the largest group was those reporting > 5 episodes/month (48.8% of patients) whereas, among T2DM patients, it was those reporting 0 episodes (56.4%). Also, in both T1DM and T2DM, the frequency of non-severe confirmed hypoglycaemia was significantly higher
in patients seen by specialists rather than GPs (Table 2 in Appendix A; see supplementary material associated with this article online). In a second step, the frequency of confirmed and severe hypoglycaemia was investigated in different subgroups (Figs. S2 S3; see supplementary material associated with this article online). In T1DM, the frequency of hypoglycaemia increased with longer duration of both diabetes and insulin therapy, but was lower in patients aged ≥ 65 years vs younger patients and in patients with cardiovascular disease (CVD; CV + ) vs those without CVD (CV−). The pattern of severe hypoglycaemia was similar to that for overall confirmed hypoglycaemia except for a tendency towards a higher frequency in patients with CVD, although none of these differences was statistically significant (see Fig. 2 in Appendix A, see supplementary material associated with this article online). In T2DM, the pattern between subgroups for overall hypoglycaemia was similar to that in T1DM, except that the occurrence of hypoglycaemia did not differ significantly between patients aged < 65 years and ≥ 65 years. In addition, the occurrence of hypoglycaemia was significantly higher for patients with BMI < 30 kg/m2 vs ≥ 30 kg/m2 , and significantly higher in patients using insulin for > 10 years (59.0%) compared with the overall T2DM population (43.6%), although it was still lower than for the overall population with T1DM (85.3%). The frequency of hypoglycaemia was also significantly higher in patients with T2DM using a basal–bolus regimen (four injections/day; 56.0%) compared with those using basal insulin only (one injection per day; 30.8%), and in those using insulin secretagogues (35.2%) vs those not using them (23.5%). There were no statistically significant differences between subgroups for severe hypoglycaemia, possibly due to the small overall number of events (see Fig. 3 in Appendix A; see supplementary material associated with this article online).
120
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
3.4. Analysis of predictive factors for confirmed hypoglycaemia Of the 3048 patients in the prospective set, the multivariate analysis included the 1304 patients with T1DM and 1631 patients with T2DM who had complete records for all variables under consideration (Table 3). In both T1DM and T2DM, the strongest predictive factor was a history of previous hypoglycaemia. Other strongly predictive factors in T1DM were > 2 injections/day, BMI < 30 kg/m2 and duration of insulin therapy > 10 years. These factors were also predictive in T2DM, albeit with lower odds ratios. In T1DM, women are significantly more at risk than men. In T2DM, the use of sulphonylureas or glinides and absence of coronary heart disease were further predictive factors. Surprisingly, glycated haemoglobin (HbA1c ) levels were not predictive of hypoglycaemia in either T1DM or T2DM. To further assess the relationship between glycaemic control and hypoglycaemia risk, the rate of hypoglycaemia was analysed by tertiles of HbA1c (Table 4). In T1DM, the rate of confirmed hypoglycaemia (events/patient/month) was significantly increased in the lowest tertile [HbA1c < 7.0% (< 53 mmol/mol)] compared with the highest tertile [HbA1c > 8.4% (> 68 mmol/mol)]. In contrast, in T2DM the rate of hypoglycaemia was similar in both the highest and lowest tertiles of HbA1c . Table 3 Predictive factors in relation to the occurrence of hypoglycaemia. Factor Type 1 diabetes mellitus (n = 1304) History of severe or non-severe hypoglycaemia > 2 daily insulin injections vs 1 or 2 injectionsa BMI < 30 kg/m2 vs BMI > 30 kg/m2 Duration of insulin therapy > 10 years vs ≤ 10 years Age < 65 years vs age > 65 years Absence of peripheral arterial disease Female vs male Type 2 diabetes mellitus (n = 1631) History of severe or non-severe hypoglycaemia > 2 vs 1 daily insulin injectionsb Duration of insulin therapy > 10 years vs ≤ 10 years 2 vs 1 daily insulin injections Use of sulphonylurea or glinide Absence of coronary heart disease BMI < 30 kg/m2 vs BMI > 30 kg/m2
Odd ratio (95% CI) 8.17 (4.17–15.97) 2.75 (1.77–4.29) 2.14 (1.39–3.31)
The impact of the number of self-measured blood glucose (SMBG) tests per day on the frequency of confirmed hypoglycaemia was assessed in both T1DM and T2DM groups. The frequency of non-severe and severe hypoglycaemia increased in parallel with the number of SMBG tests in T1DM for all categories examined (Table 5) whereas, in T2DM, the frequency increased with the number of SMBG tests except for the severe and nocturnal severe hypoglycaemia categories. 3.5. Retrospective survey In the self-reported retrospective survey, a combined total of 969 patients (25.9%) reported experiencing a severe hypoglycaemic event during the previous year: 31.5% and 21.7% of patients with T1DM and T2DM, respectively (Table 6). More patients reported diurnal than nocturnal severe events in both T1DM and T2DM. In comparison, physicians reported on CRFs that 23.6% and 11.9% of patients with T1DM and T2DM had experienced severe hypoglycaemia over the same period. 3.6. Medical consequences of severe hypoglycaemia In the prospective set, 10.7% and 7.8% of severe events in T1DM and T2DM, respectively, entailed complications, including transport to an emergency unit (1.8% and 1.2%, respectively; Table 7). In addition, 7.1% and 10.2% of patients reported that their severe hypoglycaemia had prompted them to contact a healthcare professional (HCP) outside of their regular consultations. In the retrospective set, 18.6% and 9.2% of severe events in T1DM and T2DM, respectively, entailed complications, including loss of consciousness (15.5% and 6.7%, respectively) and transport to an emergency unit (6.0% and 3.7%, respectively). The percentage of patients contacting an HCP because of severe hypoglycaemia was 8.6% and 14.5%, respectively (Table 7).
2.00 (1.42–2.83)
4. Discussion 1.87 (1.27–2.74) 1.82 (1.08–3.09) 1.79 (1.26–2.53) 3.52 (2.50–4.95) 2.69 (2.03–3.57) 1.72 (1.34–2.21)
1.55 (1.15–2.09) 1.55 (1.19–2.01) 1.46 (1.15–1.86) 1.35 (1.09–1.66)
BMI: body mass index. a Patients using an insulin pump were considered as using > 2 injections. b Patients using an insulin pump were excluded.
DIALOG was a large study to assess the frequency of welldefined confirmed hypoglycaemia by a unique combination of complementary prospective and retrospective approaches. The results of DIALOG confirm that iatrogenic hypoglycaemia is a major concern in insulin-treated patients with diabetes in a reallife setting. In the main 1-month prospective survey, confirmed hypoglycaemic events occurred in 85.3% and 43.6% of patients with T1DM and T2DM, respectively. For T1DM, the percentage of patients reporting severe hypoglycaemia during the preceding 12 months was 23.6% by CRF and 31.5% by retrospective questionnaire (the latter with an event rate of 0.08 events/patient/month). By prospective questionnaire, 13.4% of patients reported experiencing severe hypoglycaemia (0.2 events/patient/month). For T2DM, a striking finding was the large proportion of patients who experienced severe hypoglycaemia over the preceding 12 months: 11.9% according to CRF and 21.7% according to retrospective questionnaire (0.05 events/month). By prospective questionnaire, 6.4% of patients
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
121
Table 4 Hypoglycaemia according to tertiles of HbA1c measured at selection visit: Prospective Set (n = 3048). Type 1 diabetes mellitus (n = 1303)
Allb n (%)a Events/patient/month Severe n (%)a Events/patient/month Non-severe n (%)a Events/patient/month Nocturnal n (%)a Events/patient/month Severe nocturnal n (%)a Events/patient/month Diurnal n (%)a Events/patient/month Type 2 diabetes mellitus (n = 1708)
Allb n (%)a Events/patient/month Severe n (%)a Events/patient/month Non-severe n (%)a Events/patient/month Nocturnal n (%)a Events/patient/month Severe nocturnal n (%)a Events/patient/month Diurnal n (%)a Events/patient/month
Tertile 1 HbA1c < 7.0% (< 53 mmol/mol) (n = 276)
Tertile 2 HbA1c 7.0–8.4% (53–68 mmol/mol) (n = 716)
Tertile 3 HbA1c > 8.4% (> 68 mmol/mol) (n = 311)
P-value (tertile 1 vs tertile 3)
230 (83.3) 8.0 ± 8.7
619 (86.5) 7.6 ± 7.2
262 (84.2) 6.3 ± 6.3
0.765 (Chi2 ) 0.0095 (T)
45 (16.3) 0.3 ± 0.8
85 (11.9) 0.2 ± 0.6
45 (14.5) 0.2 ± 0.9
0.538 (Chi2 ) 0.568 (T)
227 (82.2) 7.7 ± 8.5
618 (86.3) 7.4 ± 7.1
257 (82.6) 6.1 ± 6.1
0.901 (Chi2 ) 0.0095 (T)
102 (37.0) 0.7 ± 1.2
294 (41.1) 0.8 ± 1.3
127 (40.8) 0.7 ± 1.1
0.336 (Chi2 ) 0.592 (T)
13 (4.7) 0.1 ± 0.3
31 (4.3) 0.1 ± 0.3
11 (3.5) 0.0 ± 0.2
0.474 (Chi2 ) 0.529 (T)
220 (79.7) 7.2 ± 8.2
604 (84.4) 6.7 ± 6.7
253 (81.4) 5.5 ± 6.0
0.616 (Chi2 ) 0.005 (T)
Tertile 1 HbA1c < 7.0% (< 53 mmol/mol) (n = 397)
Tertile 2 HbA1c 7.0–8.4% (53–68 mmol/mol) (n = 899)
Tertile 3 HbA1c > 8.4% (> 68 mmol/mol) (n = 412)
P-value (tertile 1 vs tertile 3)
181 (45.6) 1.8 ± 3.4
401 (44.6) 1.7 ± 3.2
168 (40.8) 1.8 ± 3.4
0.167 (Chi2 ) 0.926 (T)
29 (7.3) 0.1 ± 0.4
45 (5.0) 0.1 ± 0.4
36 (8.7) 0.1 ± 0.6
0.454 (Chi2 ) 0.154 (T)
171 (43.1) 1.7 ± 3.3
388 (43.2) 1.6 ± 3.1
158 (38.3) 1.6 ± 3.3
0.172 (Chi2 ) 0.764 (T)
37 (9.3) 0.2 ± 0.7
98 (10.9) 0.2 ± 0.7
55 (13.3) 0.2 ± 0.7
0.071 (Chi2 ) 0.177 (T)
1 (0.3) 0.0 ± 0.1
11 (1.2) 0.0 ± 0.1
9 (2.2) 0.0 ± 0.2
0.013 (Chi2 ) 0.013 (T)
168 (42.3) 1.6 ± 3.0
377 (41.9) 1.4 ± 2.8
154 (37.4) 1.5 ± 3.0
0.151 (Chi2 ) 0.597 (T)
Event rates are shown as means ± SD; T: Student’s t test. a Patients reporting at least one episode. b Severe and confirmed non-severe.
with T2DM reported experiencing severe hypoglycaemia (0.1 events/month). For both T1DM and T2DM, prospectively reported event rates of severe hypoglycaemia were higher than those reported retrospectively (T1DM: 0.2 ± 0.7 and 0.08 ± 0.18 events/patient/month, respectively; T2DM: 0.1 ± 0.4 and 0.05 ± 0.15 events/patient/month, respectively). One possible reason is that patients were particularly conscientious about recording events during the prospective study and/or did not recall events very well when asked to think back over the previous year. The important implication is that event rates for severe
hypoglycaemia reported in retrospective observational studies may well underestimate the true frequency. Event rates have been reported here as mean values, as is the usual practice, with no adjustment to reflect the possibility that a few patients may have had many recurrent episodes, while others may have had none. Yet, it is apparent from the median values and ranges (Table 2) that some patients did indeed have very high numbers of events. These patients were not excluded from our analysis, as our aim was to report the actual frequency of hypoglycaemia as observed in real life.
122
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
Table 5 Hypoglycaemia according to number of self-measured blood glucose tests per day: prospective set (n = 3048). Type 1 diabetes mellitus
< 3 times/day (n = 99)
3– < 4 times/day (n = 221)
4–5 times/day (n = 548)
> 5 times/day (n = 449)
Total (n = 1317)
P-value (Chi2 ) for overall trend
46 (46.5) [36.6, 56.3]
160 (72.4) [66.5, 78.3]
497 (90.7) [88.3, 93.1]
420 (93.5) [91.3%, 95.8]
1123 (85.3) [83.4, 87.2]
< 0.001
Alla n (%)b 95% CI Severe n (%)b 95% CI Non-severe n (%)b 95% CI
10 (10.1) [4.2, 16.0]
22 (10.0) [6.0, 13.9]
69 (12.6) [9.8, 15.4]
6 (16.9) [13.5, 20.4]
177 (13.4) [11.6, 15.3]
0.040
44 (44.4) [34.7, 54.2]
156 (70.6) [64.6, 76.6]
494 (90.1) [87.7, 92.6]
418 (93.1) [90.8, 95.4]
1112 (84.4) [82.5, 86.4]
< 0.001
Type 2 diabetes mellitus
< 2 times/day (n = 22)
2– < 3 times/day (n = 361)
3–< 4 times/day (n = 594)
≥ 4 times/day (n = 549)
Total (n = 1729)
P-value (Chi2 ) for overall trend
56 (24.9) [19.2, 30.5]
111 (30.7) [26.0, 35.5]
266 (44.8) [40.8, 48.8]
321 (58.5) [54.3, 62.6]
754 (43.6) [41.3, 45.9]
< 0.001
6 (2.7) [0.6, 4.8]
25 (6.9) [4.3, 9.5]
38 (6.4) [4.4, 8.4]
41 (7.5) [5.3, 9.7]
110 (6.4) [5.2, 7.5]
0.091
53 (23.6) [18.0, 29.1]
100 (27.7) [23.1, 32.3]
256 (43.1) [39.1, 47.1]
312 (56.8) [52.7, 61.0]
721 (41.7) [39.4, 44.0]
< 0.001
Alla n (%)b 95% CI Severe n (%)b 95% CI Non-severe n (%)b 95% CI
Values are n (%) of patients reporting at least one episode. a Severe and confirmed non-severe. b Patients reporting at least one episode.
Table 6 One-year occurrence of severe hypoglycaemia: retrospective set. Type 1 diabetes mellitus (n = 1592) Frequency n (%)a Severe Nocturnal severe Diurnal severe a
Type 2 diabetes mellitus (n = 2151) Events/patient/month (mean ± SD)
Frequency n (%)a
0.08 ± 0.18 0.03 ± 0.08 0.05 ± 0.13
502 (31.5) 270 (17.0) 381 (23.9)
Events/patient/month (mean ± SD) 0.05 ± 0.15 0.01 ± 0.05 0.04 ± 0.12
467 (21.7) 128 (6.0) 385 (17.9)
Patients reporting at least one episode during the preceding 12 months; all events were self-reported.
Table 7 Medical consequences of severe hypoglycaemia. Prospective survey
Total of events with complicationsc Loss of consciousness Transport to emergency unit Hospitalization Fracture or trauma Contact with a healthcare professionald a
(30 days)
Retrospective survey
(12 months)
T1DM (n = 281)a n (%)b
T2DM (n = 166)a n (%)b
Total (n = 447)a n (%)b
T1DM (n = 1577)a n (%)b
T2DM (n = 1365)a n (%)b
Total (n = 2942)a n (%)b
30 (10.7) 26 (9.3) 5 (1.8) 2 (0.7) 2 (0.7) 20 (7.1)
13 (7.8) 7 (4.2) 2 (1.2) 5 (3.0) 0 (0.0) 17 (10.2)
43 (9.6) 33 (7.4) 7 (1.6) 7 (1.6) 2 (0.4) 37 (8.3)
293 (18.6) 244 (15.5) 95 (6.0) 61 (3.9) 31 (2.0) 136 (8.6)
125 (9.2) 92 (6.7) 51 (3.7) 54 (4.0) 5 (0.4) 198 (14.5)
418 (14.2) 336 (11.4) 146 (5.0) 115 (3.9) 36 (1.2) 334 (11.4)
Number of severe episodes. Events self-reported by patients completing the prospective questionnaire; percentages calculated from total number of events (presence of a characteristic/number of events). c Derived variable (at least loss of consciousness/transport to emergency room/hospitalisation/fracture or trauma). d Patients who responded “yes” to the question: Did you contact a healthcare professional as a result of this hypoglycaemia episode? b
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
The present results are broadly in line with those of previous studies [9,12,25,26]. In two small prospective studies (< 400 patients), for example, rates of severe hypoglycaemia (events/patient/year) in T1DM patients were reported as 1.15 [9] and 3.2 [12], and in T2DM patients as 0.35 [9] and 0.70 [12]. The results in the present prospective survey were 2.4 events/patient/year in T1DM and 1.2 events/patient/year in T2DM; thus, our results for T2DM are somewhat higher than those reported earlier. Previously reported event rates for severe hypoglycaemia have generally been lower for T2DM than for T1DM [9,12]. Surprisingly, in the present study, among those patients who experienced one or more episodes of hypoglycaemia during the 30-day follow-up, event rates (expressed as events/patient/month) for severe hypoglycaemia, including nocturnal hypoglycaemia, were closely similar in both T2DM and T1DM (R2 values, Table 2). In fact, large proportions of patients (40.2% and 11.0% in T1DM and T2DM, respectively) experienced nocturnal events. This is important because nocturnal hypoglycaemia is feared by patients and acts as a brake on intensification of insulin therapy. Another goal of DIALOG was to identify predictive factors for hypoglycaemia in real life. In our study, the most important predictive factors in both T1DM and T2DM were a history of severe or non-severe hypoglycaemia > 2 daily injections and duration of insulin therapy > 10 years. These are in line with previous reports [9]. In the UK Hypoglycaemia Study, event rates for T2DM treated with insulin for > 5 years were similar to those for T1DM treated for < 5 years [12]. However, in DIALOG, the frequency of hypoglycaemia in patients with T2DM with a duration of insulin therapy > 10 years remained lower than that of patients with T1DM, even when the duration of disease was shorter. Our analysis for predictive factors yielded some intriguing results. In T1DM, the risk of hypoglycaemia was higher in women than in men, an observation that has never been reported before. In patients with T2DM, the absence of coronary heart disease was a predictive factor for hypoglycaemia. It is possible that, in line with a patient-centred approach [21], physicians treating patients with coronary heart disease allow them to maintain fairly high glycaemic levels precisely to avoid hypoglycaemia. In addition, hypoglycaemia was less frequent in patients with T1DM aged ≥ 65 years. These results seem counterintuitive, as vulnerability to hypoglycaemia is normally greater with increasing age. Just as in patients with CVD, however, it is possible that older patients aim to achieve less stringent glycaemic targets. Moreover, there was no increased risk of hypoglycaemia in patients with an estimated glomerular filtration rate of ≤ 60 mL/min (data not shown; data were not available for patients with more advanced chronic kidney disease). An unexpected observation was that HbA1c was not predictive of hypoglycaemia in either T1DM or T2DM. It is conceivable that patients susceptible to hypoglycaemia aimed for less stringent glycaemic targets, and that the patients with lower HbA1c were those for whom hypoglycaemia was not a problem. Analysis of the rate of hypoglycaemia by tertiles of HbA1c led to discordant results between T1DM and T2DM.
123
Results from the DCCT showed that the rate of severe hypoglycaemia increased as HbA1c levels decreased [2] and, in our study, the risk of hypoglycaemia was highest in the lowest tertile in T1DM. In contrast, there was a trend towards a U-shaped association between tertiles of HbA1c and hypoglycaemia risk in T2DM, with the higher risk for extreme values. Interestingly, an association between higher HbA1c levels and severe hypoglycaemia was previously reported in T2DM in the Fremantle observational cohort study [13]. The number of SMBG tests was positively correlated with the reported occurrence of hypoglycaemia in both T1DM and T2DM. This was to be expected, as the definition of confirmed hypoglycaemia required a blood glucose measurement (except in cases of severe hypoglycaemia). Patients who were testing more often would therefore be more likely to detect asymptomatic events. This is a limitation of all studies of hypoglycaemia that rely on blood glucose measurements to confirm an event. It could also suggest that increasing the number of SMBG tests failed to prevent the risk of hypoglycaemia, and that high numbers of SMBG tests likely reflected high glycaemic variability in the patients concerned, a phenomenon not addressed in the present study. Our study also showed that severe hypoglycaemia incurred medical consequences such as hospitalisation. The economic implications for individual patients and the healthcare system in general should be assessed more thoroughly in health-economic analyses. The present study had a few limitations. Our definition of severe hypoglycaemia corresponded to the intentions of the ADA definition, but our patients did not specifically record whether, in the absence of a plasma glucose concentration, neurological recovery followed the return of plasma glucose to normal. However, it seems reasonable to assume this was the case if the patient was able to complete the questionnaire. Also, the data obtained were dependent on patient self-reporting and recall, and not all patients returned analysable questionnaires: 69% of all included patients qualified for the prospective set and 85% for the retrospective set. It is possible that the patients who failed to complete their questionnaires had different characteristics or event rates from those who did. However, the key demographic and diabetes-related characteristics were similar for both the overall and prospective sets. In fact, our study methodology incorporated a non-inclusion registry to ensure that the included population was representative of real-life patients as seen by physicians. Nevertheless, a recruitment bias cannot be excluded, as the included patients were probably more at risk than non-included ones, with a higher prevalence of both T1DM and previous hypoglycaemia. It should also be mentioned that most of our patients were recruited by specialists (endocrinologists) rather than by GPs; this imbalance might have led to the recruitment of patients with more severe hypoglycaemia, especially for patients with T2DM. The present study had a novel design and a number of strengths compared with previous studies. For the first time, severe hypoglycaemia was reported simultaneously in three ways in the same cohort: retrospectively through the physician’s CRF; retrospectively by questionnaire; and prospectively
124
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
by questionnaire. Hypoglycaemia was strictly defined to include only confirmed hypoglycaemia according to ADA definitions [23]. Also, the distribution of participating physicians was representative of real life. The number of patients was exceptionally large for a prospective survey of hypoglycaemia, and the frequency and rate of hypoglycaemia were assessed by both prospective and retrospective approaches. 5. Conclusion The DIALOG study provides a comprehensive, up-to-date record of the prevalence of hypoglycaemia, in particular severe hypoglycaemia, among insulin-treated patients with diabetes in real life. The relatively high frequency of hypoglycaemia reported underlines the importance of optimising patients’ insulin regimens to successfully control glycaemia while minimizing the occurrence of hypoglycaemia. The study also confirmed that previous episodes of hypoglycaemia increase the risk of future episodes, thus supporting the concept of a patient-centred approach for managing diabetes in which hypoglycaemia risk evaluation is critical. Contributors Bertrand Cariou was the lead investigator. All the authors except S. Lavigne contributed to designing the study, conducting it and collecting data. S. Lavigne collated the data and conducted the statistical analyses. B. Cariou specified the contents of the article and wrote the first draft in conjunction with a medical writer. All the authors reviewed and commented on drafts of the article. Disclosure of interest B. Cariou has taken part in advisory panels for, and received research support from, Novo Nordisk and Sanofi; attended advisory boards for Amgen, AstraZeneca, Genfit, Janssen, Novo Nordisk and Sanofi-Aventis; and acted as a consultant for Genfit. P. Fontaine has received fees for participation to advisory board for Novo Nordisk, Eli Lilly, Sanofi-Aventis, Merck Sharp & Dohme, Boehringer Ingelheim and GlaxoSmithKline. E. Eschwege has taken part in an advisory panel for the French National Authority for Health: Transparency Committee: Working group for Public Health Benefit Assessment of drugs; and attended advisory boards for Roche, Novartis and Novo Nordisk. M. Lièvre has attended advisory boards for Merck-Serono, Novartis, Medtronic, Terumo, Novo Nordisk, Abbott, Bard, Boston Scientific, Cordis, Takeda, AstraZeneca and Johnson & Johnson, and served on an advisory panel for Johnson & Johnson (no personal remuneration). D. Gouet has received research support from Novo Nordisk, Eli Lilly, Boehringer Ingelheim, Medtronic, Johnson & Johnson and Sanofi-Aventis. D. Huet has attended advisory boards for Sanofi-Aventis, Novo Nordisk and Novartis. S. Madani is an employee of Novo Nordisk. S. Lavigne is an employee of ITEC services SAS, a contract research organization paid by Novo Nordisk Pharmaceutique to collate and analyse the data. B. Charbonnel has received
fees for consultancy, speaking, travel and/or accommodation from AstraZeneca, Bristol-Myers Squibb, Boehringer Ingelheim, GlaxoSmithKline, Janssen, Eli Lilly, Merck Sharpe & Dohme, Novartis, Novo Nordisk, Roche, Sanofi-Aventis and Takeda. Acknowledgements The authors would like to thank Grace Townshend and Mark Nelson, Watermeadow Medical, Witney, UK, for their assistance in the writing, editing and submission of this manuscript. This assistance was funded by Novo Nordisk. The authors thank all of the physicians involved in the study for their participation. Role of the funding source: Novo Nordisk sponsored the study. The trial was designed by an independent scientific committee (Appendix A, see supplementary material associated with this article online) in collaboration with Novo Nordisk. Data were collected by a contract research organization (ITEC), which conducted the statistical analysis of the data and wrote the study report (S.L.). All authors had full access to the data and share final responsibility for the content of the manuscript and the decision to submit for publication.
Appendix A. Supplementary data Supplementary data (Figs. S1, S2, S3, S4 and Appendix) associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.diabet.2014.10.007. References [1] Frier BM, Heller SR. Epidemiology and impact of hypoglycaemia on patients with diabetes. Translational endocrinology & metabolism: hypoglycaemia in diabetes update, 3; 2012. p. 15–47. [2] Diabetes Control, Complications Trial Research Group. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl J Med 1993;329:977–86. [3] Diabetes Control and Complications Trial Research Group. Hypoglycemia in the Diabetes Control and Complications Trial. Diabetes 1997;46: 271–86. [4] Nathan DM, DCCT/EDIC Research Group. The diabetes control and complications trial/epidemiology of diabetes interventions and complications study at 30 years: overview. Diabetes Care 2014;37: 9–16. [5] UK Prospective Diabetes Study (UKPDS) Group. Intensive blood glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet 1998;352:837–53. [6] Seaquist ER, Miller ME, Bonds DE, Feinglos M, Goff Jr DC, Peterson K, et al. The impact of frequent and unrecognized hypoglycaemia on mortality in the ACCORD study. Diabetes Care 2012;35:409–14. [7] Bruderer SG, Bodmer M, Jick SS, Bader G, Schlienger RG, Meier CR. Incidence of and risk factors for severe hypoglycaemia in treated type 2 diabetes mellitus patients in the UK - a nested case-control analysis. Diabetes Obes Metab 2014;16:801–11.
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125 [8] Kulzer B, Seitz L, Kern W. Real-world patient-reported rates of non-severe hypoglycaemic events in Germany. Exp Clin Endocrinol Diabetes 2014;122:167–72. [9] Donnelly LA, Morris AD, Frier BM, Ellis JD, Donnan PT, Durrant R, et al. Frequency and predictors of hypoglycaemia in type 1 and insulin-treated type 2 diabetes: a population-based study. Diabet Med 2005;22:749–55. [10] Henderson JN, Allen KV, Deary IJ, Frier BM. Hypoglycaemia in insulin-treated type 2 diabetes: frequency, symptoms and impaired awareness. Diabet Med 2003;20:1016–21. [11] Pedersen-Bjergaard U, Pramming S, Heller SR, Wallace TM, Rasmussen AK, Jørgensen HV, et al. Severe hypoglycaemia in 1076 adult patients with type 1 diabetes: influence of risk markers and selection. Diabetes Metab Res Rev 2004;20:479–86. [12] UK Hypoglycaemia Study Group. Risk of hypoglycaemia in types 1 and 2 diabetes: effects of treatment modalities and their duration. Diabetologia 2007;50:1140–7. [13] Davis TM, Brown SG, Jacobs IG, Bulsara M, Bruce DG, Davis WA. Determinants of severe hypoglycaemia complicating type 2 diabetes: the Fremantle diabetes study. J Clin Endocrinol Metab 2010;95:2240–7. [14] Amiel SA, Dixon T, Mann R, Jameson K. Hypoglycaemia in type 2 diabetes. Diabet Med 2008;25:245–54. [15] McCoy RG, Van Houten HK, Ziegenfuss JY, Shah ND, Wermers RA, Smith SA. Increased mortality of patients with diabetes reporting severe hypoglycaemia. Diabetes Care 2012;35:1897–901. [16] ACCORD Study Group. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med 2008;358:2545–59. [17] ADVANCE Collaborative Group. Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes. N Engl J Med 2008;358:2560–72. [18] Duckworth W, Abraira C, Moritz T, Reda D, Emanuele N, Reaven PD, et al. Glucose control and vascular complications in veterans with type 2 diabetes. N Engl J Med 2009;360:129–39.
125
[19] Bonds DE, Miller ME, Bergenstal RM, Buse JB, Byington RP, Cutler JA, et al. The association between symptomatic, severe hypoglycaemia and mortality in type 2 diabetes: retrospective epidemiological analysis of the ACCORD study. BMJ 2010;340:b4909. [20] Barendse S, Singh H, Frier BM, Speight J. The impact of hypoglycaemia on quality of life and related patient-reported outcomes in type 2 diabetes: a narrative review. Diabet Med 2012;29: 293–302. [21] Inzucchi SE, Bergenstal RM, Buse JB, Diamant M, Ferrannini E, Nauck M, et al. Management of hyperglycaemia in type 2 diabetes: a patient-centered approach. Position statement of the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD). Diabetologia 2012;55:1577–96. [22] Workgroup on Hypoglycemia American Diabetes Association. Defining and reporting hypoglycaemia in diabetes: a report from the American Diabetes Association Workgroup on Hypoglycemia. Diabetes Care 2005;28:1245–9. [23] Seaquist ER, Anderson J, Childs B, Cryer P, Dagogo-Jack S, Fish L, et al. Hypoglycemia and diabetes: a report of a workgroup of the American Diabetes Association and the Endocrine Society. Diabetes Care 2013;36:1384–95. [24] International Society for Pharmacoepidemiology. Guidelines for good pharmacoepidemiology practices (GPP), 2007; 2013 [Available from: URL] http://www.pharmacoepi.org/resources/guidelines 08027.cfm [25] Verges B, Brun JM, Tawil C, Alexandre B, Kerlan V. Strategies for insulin initiation: insights from the French LIGHT observational study. Diabetes Metab Res Rev 2012;28:97–105. [26] Charbonnel B, Eschwege E, Charpentier G, Vialettes B, Chabrier G, Suissa S, et al. Insulin glargine in real use conditions in France [article in French]. Med Mal Metabol 2010;4:169–78.
ScienceDirect www.sciencedirect.com Diabetes & Metabolism 41 (2015) 116–125
Original article
Frequency and predictors of confirmed hypoglycaemia in type 1 and insulin-treated type 2 diabetes mellitus patients in a real-life setting: Results from the DIALOG study B. Cariou a,∗ , P. Fontaine b , E. Eschwege c , M. Lièvre d , D. Gouet e , D. Huet f , S. Madani g , S. Lavigne h , B. Charbonnel a a
Department of Endocrinology, l’institut du Thorax, CHU de Nantes, hôpital Guillaume & René Laennec, boulevard Jacques-Monod, 44093 Nantes cedex 1, France b Department of Endocrinology and Diabetology, University Hospital of Lille, Lille, France c Inserm U-1018, Centre de Recherche en Epidémiologie et Santé des populations (CESP), Villejuif, France d Laennec Faculty of Medicine, Lyon, France e Hôpital Saint-Louis, centre hospitalier de La Rochelle, La Rochelle, France f Hôpital Saint-Joseph, Paris, France g Novo Nordisk, Paris, France h ITEC Services SAS, Cenon, France Received 13 June 2014; received in revised form 21 October 2014; accepted 21 October 2014 Available online 24 November 2014
Abstract Aim. – DIALOG assessed the prevalence and predictors of hypoglycaemia in patients with type 1 (T1DM) or insulin-treated type 2 diabetes mellitus (T2DM) in a real-life setting. Methods. – In this observational study, insulin-treated patients (n = 3048) completed prospective daily questionnaires reporting the frequency and consequences of severe/confirmed non-severe hypoglycaemia over 30 days. Patients (n = 3743) also retrospectively reported severe hypoglycaemia over the preceding year. Results. – In this prospective survey, 85.3% and 43.6% of patients with T1DM and T2DM, respectively, reported experiencing at least one confirmed hypoglycaemic event over 30 days, while 13.4% and 6.4%, respectively, reported at least one severe event. Hypoglycaemia frequency increased with longer duration of diabetes and insulin therapy. Strongly predictive factors for hypoglycaemia were previous hypoglycaemia, > 2 injections/day, BMI < 30 kg/m2 and duration of insulin therapy > 10 years. HbA1c level was not predictive of hypoglycaemia in either T1DM or T2DM. The confirmed hypoglycaemia rate was increased in the lowest compared with the highest tertile of HbA1c in T1DM, but not T2DM. At the time of enrolment, physicians reported severe hypoglycaemia in 23.6% and 11.9% of T1DM and T2DM patients, respectively, during the preceding year; the retrospective survey yielded frequencies of 31.5% and 21.7%, respectively. Also, severe hypoglycaemia led to medical complications in 10.7% and 7.8% of events in T1DM and T2DM patients, respectively, over 30 days. Conclusion. – Using a unique combined prospective and retrospective approach, the DIALOG study found a relatively high frequency of hypoglycaemia among insulin-treated patients. These findings emphasize the importance of a patient-centred approach for managing diabetes in which hypoglycaemia risk evaluation is critical. Trial registration. – ClinicalTrials.gov: NCT01628341. © 2014 Elsevier Masson SAS. All rights reserved. Keywords: Hypoglycaemia; Type 1 diabetes; Type 2 diabetes; Observational study Abbreviations: ACCORD, Action to Control Cardiovascular Risk in Diabetes; ADA, American Diabetes Association; BMI, Body Mass Index; CRF, Case Report Form; CVD, Cardiovascular Disease; DCCT, Diabetes Control and Complications Trial; GP, General Practitioner; HbA1c , Glycated Haemoglobin; HCP, Healthcare Professional; SMBG, Self-Measured Blood Glucose; T1DM, Type 1 Diabetes Mellitus; T2DM, Type 2 Diabetes Mellitus; UKPDS, United Kingdom Prospective Diabetes Study. ∗ Corresponding author. Tel.: +33 253 482 707; fax: +33 253 482 708. E-mail address: [email protected] (B. Cariou). http://dx.doi.org/10.1016/j.diabet.2014.10.007 1262-3636/© 2014 Elsevier Masson SAS. All rights reserved.
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
117
1. Introduction
2.2. Study setting and assessments
Iatrogenic hypoglycaemia is an acute complication of insulin treatment in type 1 (T1DM) and type 2 diabetes mellitus (T2DM) patients, but occurs more frequently in insulin-deficient states (T1DM) and advanced T2DM [1]. Most of the available reports on the incidence of hypoglycaemia are from clinical trials such as the diabetes control and complications trial (DCCT) for T1DM [2–4], and the United Kingdom Prospective diabetes study (UKPDS) [5] and action to control cardiovascular risk in diabetes (ACCORD) study for T2DM [6]. Both the DCCT and UKPDS demonstrated that the risk of hypoglycaemia increases as glycaemic control improves [3,5]. However, it is difficult to accurately assess the frequency of hypoglycaemia in real life: most of the estimates that have been published were relatively small prospective studies, or larger studies based on retrospective data from registries or databases [7–13]. Severe hypoglycaemia is associated with increased risks of morbidity and mortality, especially in patients with T2DM and high cardiovascular risk, and in the elderly [14,15]. Recently, accumulating evidence has indicated that hypoglycaemia is a marker of vulnerability especially in older patients with T2DM [6,16–18]. The link between severe hypoglycaemia and an increased risk of death was shown in a retrospective epidemiological analysis of the ACCORD study, although direct causality was not demonstrated [19]. Moreover, health-related quality of life is also decreased as the frequency of hypoglycaemia increases [14,20]. Consequently, hypoglycaemia is now one of the major considerations when adopting a patient-centred approach to the management of T2DM [21]. Nevertheless, there is a need for dedicated studies assessing the frequency of hypoglycaemia in a “real-life” setting. DIALOG was a large nationwide observational study using an original approach that combined both prospective and retrospective evaluations of the frequency of confirmed hypoglycaemia, as rigorously defined by American Diabetes Association (ADA) criteria [22,23]. The objectives of DIALOG were to evaluate the prevalence of severe and non-severe confirmed hypoglycaemia in a real-life setting among patients with T1DM and insulintreated T2DM in France, and to define predictive factors for hypoglycaemia. DIALOG is, to our knowledge, the first study conducted both prospectively and retrospectively in a very large (> 4000 patients) non-clinical-trial population.
All 2300 specialist physicians (1700 endocrinologists/diabetologists and 600 internists) who manage patients with diabetes in France were invited to take part in DIALOG, and 474 (20.6%) agreed to participate. Of a random representative sample of 8000 general practitioners (GPs) treating patients with diabetes, 725 (9.1%) agreed to participate (Appendix A; see supplementary material associated with this article online). Participating physicians were required to include consecutive patients who fulfilled the selection criteria and were seen within a predefined period (between 2 May and 28 June 2012). Investigators completed a case report form (CRF) for each patient, which included the physician’s record of the patient’s characteristics, diabetes history and number of severe hypoglycaemia events experienced within the last 12 months. Investigators also completed a non-inclusion registry detailing reasons for excluding any patients. Investigators provided patients with information on hypoglycaemia, and two self-administered questionnaires (Appendix A; see supplementary material associated with this article online) with instructions on how to complete them at home. Patients answered the prospective questionnaires every day for 30 days, recording whether they had experienced hypoglycaemia and responding to questions about each episode (including blood glucose level, if measured, and characteristics and consequences of the event). With the retrospective questionnaire, patients answered questions concerning severe hypoglycaemic episodes during the previous year.
2. Methods 2.1. Study design and conduct This was a multicentre, prospective and retrospective, observational study. The study protocol and patients’ information sheet were approved by the appropriate authorities as required by French regulations. The study was conducted in accordance with the principles of the declaration of Helsinki, guidelines for good pharmacoepidemiology practices [24] and regulatory requirements. Patients signed an information notice before being enrolled into the study.
2.3. Inclusion and exclusion criteria Patients aged ≥ 18 years with either T1DM or T2DM treated with insulin for at least 12 months who were able to self-monitor their blood glucose levels and fill in the questionnaires were eligible. Pregnant or breastfeeding women, or those with a pregnancy within the previous year, were excluded. 2.4. Study outcomes For the prospective survey, the primary endpoint was the percentage of insulin-treated patients who had at least one hypoglycaemic event (severe or confirmed non-severe) during the 30 days following enrolment. Secondary endpoints included hypoglycaemic events by type. Also, rates of hypoglycaemia were calculated as events/patient/month, and factors predictive of hypoglycaemia were identified. The endpoint for the retrospective survey was severe hypoglycaemic episodes during the preceding year. Medical consequences of severe hypoglycaemia were also recorded by both questionnaires. 2.5. Definition of hypoglycaemia During data entry and analysis, hypoglycaemic episodes were defined in accordance with ADA definitions for severe, documented symptomatic, asymptomatic, probable and
118
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
pseudohypoglycaemia [22,23] (Appendix A; see supplementary material associated with this article online). The main endpoint of the DIALOG study was the combination of confirmed hypoglycaemia, whether symptomatic or asymptomatic, and severe hypoglycaemia (Fig. S1; see supplementary material associated with this article online). Nocturnal hypoglycaemia was defined as any confirmed episode with time of onset between 00h00 and 06h00. 2.6. Statistical analysis It was calculated that to arrive at an estimate of the number of hypoglycaemic events in T1DM and T2DM patients with a precision of at least 3%, a total of 3811 patients were required (Appendix A; see supplementary material associated with this article online). Three analytical sets were defined (Fig. S2; see supplementary material associated with this article online). The overall set included all patients in the survey except for any who had a major deviation (such as a violation of one of the inclusion/exclusion criteria). The prospective set included all patients with a completed inclusion CRF with no major deviations and with an analysable prospective questionnaire. The main criteria for a questionnaire to be considered analysable were a record of at least one hypoglycaemic episode or completion of at least 25 days of the questionnaire (see Appendix A for the full definition of “analysable”; supplementary material associated with this article online). The retrospective set included all patients with a completed inclusion CRF with no major deviations and an analysable retrospective questionnaire. 2.7. Statistical techniques Quantitative variables with a normal distribution were analysed using Student’s t test. Qualitative variables were compared using a Chi2 test or Fisher’s exact test. Predictive factors for the presence of a hypoglycaemic event during the prospective 1-month follow-up were identified using a logistic regression model. First, univariate analysis tested the relationship between each baseline parameter and the relevant variable, using the Chi2 test (or Fisher’s exact test) for qualitative data, and Student’s t test [or analysis of variance (ANOVA)] for continuous variables. All variables significant at a 20% level were included in the second step, a multivariate analysis using a stepwise logistic regression model with a significance level set at 5%. 3. Results 3.1. Disposition of patients A total of 4444 patients were included (74% by 370 specialists and 26% by 320 GPs); 20 patients were excluded because of major deviations. Among the final 4424 included patients, 3048 questionnaires from the prospective set and 3743 questionnaires from the retrospective set were evaluable (Fig. S2; see supplementary material associated with this article online).
Table 1 Demographic data for patients in the Overall Set. T1DM (n = 1915) Male, n (%) 978 (51.2) Age, years 48.4 ± 16.2 111 (5.8) Patients ≥ 75 years old, n (%) 25.4 ± 4.4 Body mass index (BMI), kg/m2 Patients with BMI > 30 kg/m2 , 249 (13.0) n (%) HbA1c , % 7.9 ± 1.3 (63 ± 14) (calculated mmol/mol) Diabetes history 20.4 ± 12.9 Time since diagnosis, years Duration of insulin use, years, n (%) 10 1315 (68.8) Experienced severe 477 (25.0) hypoglycaemia during previous 12 months, n (%)a Number of severe episodes during previous 12 monthsa Mean ± SD 3.3 ± 8.3 2.0 (0, 135) Median (range) Complications of diabetes, n (%) Complications present 855 (44.6) 142 (7.4) Peripheral arterial disease Coronary disease 180 (9.3) 48 (2.5) Cerebrovascular disease Retinopathy 595 (31.1) 145 (7.6) eGFR (MDRD) < 60 mL/min Micro- or 307 (16.0) macroalbuminuria Diabetic foot ulcer 73 (3.8) Comorbidities 676 (35.3) Hypertension 687 (35.9) Dyslipidaemia Insulin regimen at enrollment Insulin pump 598 (31.2) NR 1 injection NR 2 injections 1 or 2 injections 214 (11.2) 1090 (56.9) > 2 injections Oral antidiabetic drugs at enrolmentb Metformin NA DPP-4 inhibitor NA NA Sulphonylurea and/or glinide Alpha-glucosidase NA inhibitor GLP-1 agonist NA
T2DM (n = 2509) 1352 (54.0) 66.3 ± 11.0 596 (23.8) 30.9 ± 5.9 1316 (52.5) 7.9 ± 1.2 (63 ± 13) 17.3 ± 9.3
986 (39.3) 915 (36.5) 606 (24.2) 329 (13.1)
2.1 ± 2.4 1.0 (1, 31) 1597 (64.9) 444 (17.7) 614 (24.5) 144 (5.7) 623 (24.8) 564 (22.5) 781 (31.1) 141 (5.6) 2025 (80.7) 1900 (75.7) 122 (4.9) 953 (38.0) 477 (19.0) NR 945 (37.7) 1387 (55.3) 382 (15.2) 823 (32.8) 60 (2.4) 187 (7.5)
Values are means ± standard deviation unless otherwise stated; T1DM: type 1 diabetes mellitus; T2DM: type 2 diabetes mellitus; eGFR (MDRD): estimated glomerular filtration rate (Modification of diet in renal disease); DPP-4: dipeptidyl peptidase-4; GLP-1: glucagon-like peptide-1; NR, not reported; NA: not applicable. a As reported by physician during examination. b As reported for T2DM patients only.
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
119
Table 2 Hypoglycaemia in the prospective set (n = 3048). Type 2 diabetes mellitus (n = 1731)
Type 1 diabetes mellitus (n = 1317) Frequency n (%)a
Allb Severee Overall non-severee Non-severe symptomatic Non-severe asymptomatic Nocturnal Severe nocturnal Diurnal Probable hypoglycaemiaf Pseudohypoglycaemiaf a b c d e f
1123 (85.3) 177 (13.4) 1112 (84.4) 1085 (82.4) 367 (27.9) 529 (40.2) 55 (4.2) 1089 (82.7) 521 (39.6) 289 (21.9)
Events/patient/month (mean ± SD) R1c R2d 7.4 0.2 7.1 6.1 1.1 0.8 0.1 6.5 1.2 0.6
± ± ± ± ± ± ± ± ± ±
7.48.6 ± 7.3 0.71.6 ± 1.3 7.38.5 ± 7.1 6.47.4 ± 6.3 3.03.8 ± 4.6 1.21.9 ± 1.3 0.31.2 ± 0.5 6.97.9 ± 6.9 2.53.0 ± 3.3 1.92.6 ± 3.2
Median (range)
5.0 (0, 51) 0.0 (0, 12) 5.0 (0, 48) 4.0 (0, 45) 0.0 (0, 39) 0.0 (0, 10) 0.0 (0, 3) 4.1 (0, 49) 0.0 (0, 30) 0.0 (0, 29)
Frequency n (%)a
755 (43.6) 110 (6.4) 722 (41.7) 682 (39.4) 134 (7.7) 190 (11.0) 21 (1.2) 704 (40.7) 472 (27.3) 373 (21.5)
Events/patient/month (mean ± SD)
1.7 0.1 1.6 1.5 0.2 0.2 0.0 1.5 1.1 0.7
± ± ± ± ± ± ± ± ± ±
3.34.0 ± 4.0 0.41.5 ± 0.9 3.23.9 ± 3.9 2.93.7 ± 3.6 0.92.4 ± 2.5 0.71.7 ± 1.4 0.11.1 ± 0.5 2.93.7 ± 3.6 3.64.2 ± 5.9 2.43.3 ± 4.4
Median (range)
0.0 (0, 29) 0.0 (0, 6) 0.0 (0, 28) 0.0 (0, 25) 0.0 (0, 19) 0.0 (0, 9) 0.0 (0, 3) 0.0 (0, 26) 0.0 (0, 30) 0.0 (0, 30)
Patients reporting at least one episode. Severe or confirmed non-severe hypoglycaemia for same patient. Rate of events including patients with no episodes. Rate of events in patients with one or more episodes. As individual patients could report both severe and non-severe events, the sum of these two categories is higher than the All value. Not included when identifying episodes of hypoglycaemia for main criterion.
3.2. Baseline characteristics Baseline data for the overall set are shown in Table 1, which are similar to characteristics of the prospective set (Table 1 in Appendix A; see supplementary material associated with this article online). To ensure that selection bias was avoided, investigators also recorded the characteristics of the non-included patients (n = 3066), who were comparable to the analysable patients in terms of age, gender and diabetes duration (data not shown), although the proportion of patients with T1DM was smaller among the non-included vs included patients (30.7% vs 43.3%, respectively), and a smaller proportion reported a history of hypoglycaemia (61.7% vs 88.3%, respectively). 3.3. Frequency of confirmed hypoglycaemia in the prospective survey The proportion of all patients reporting at least one hypoglycaemic episode during the 30-day follow-up was 61.6% for the prospective set (1878 of 3048 patients), with 85.3% for T1DM (event rate/month of 7.4 ± 7.4) and 43.6% for T2DM (event rate/month of 1.7 ± 3.3). Table 2 shows the results for the different categories of hypoglycaemia, presented as mean (SD) values for all patients and also calculated for only those patients with at least one episode of hypoglycaemia. Median (range) values are also shown. The distribution by number of hypoglycaemic episodes per patient per month are shown in supplementary figures (Appendix A; see supplementary material associated with this article online). Among T1DM patients, the largest group was those reporting > 5 episodes/month (48.8% of patients) whereas, among T2DM patients, it was those reporting 0 episodes (56.4%). Also, in both T1DM and T2DM, the frequency of non-severe confirmed hypoglycaemia was significantly higher
in patients seen by specialists rather than GPs (Table 2 in Appendix A; see supplementary material associated with this article online). In a second step, the frequency of confirmed and severe hypoglycaemia was investigated in different subgroups (Figs. S2 S3; see supplementary material associated with this article online). In T1DM, the frequency of hypoglycaemia increased with longer duration of both diabetes and insulin therapy, but was lower in patients aged ≥ 65 years vs younger patients and in patients with cardiovascular disease (CVD; CV + ) vs those without CVD (CV−). The pattern of severe hypoglycaemia was similar to that for overall confirmed hypoglycaemia except for a tendency towards a higher frequency in patients with CVD, although none of these differences was statistically significant (see Fig. 2 in Appendix A, see supplementary material associated with this article online). In T2DM, the pattern between subgroups for overall hypoglycaemia was similar to that in T1DM, except that the occurrence of hypoglycaemia did not differ significantly between patients aged < 65 years and ≥ 65 years. In addition, the occurrence of hypoglycaemia was significantly higher for patients with BMI < 30 kg/m2 vs ≥ 30 kg/m2 , and significantly higher in patients using insulin for > 10 years (59.0%) compared with the overall T2DM population (43.6%), although it was still lower than for the overall population with T1DM (85.3%). The frequency of hypoglycaemia was also significantly higher in patients with T2DM using a basal–bolus regimen (four injections/day; 56.0%) compared with those using basal insulin only (one injection per day; 30.8%), and in those using insulin secretagogues (35.2%) vs those not using them (23.5%). There were no statistically significant differences between subgroups for severe hypoglycaemia, possibly due to the small overall number of events (see Fig. 3 in Appendix A; see supplementary material associated with this article online).
120
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
3.4. Analysis of predictive factors for confirmed hypoglycaemia Of the 3048 patients in the prospective set, the multivariate analysis included the 1304 patients with T1DM and 1631 patients with T2DM who had complete records for all variables under consideration (Table 3). In both T1DM and T2DM, the strongest predictive factor was a history of previous hypoglycaemia. Other strongly predictive factors in T1DM were > 2 injections/day, BMI < 30 kg/m2 and duration of insulin therapy > 10 years. These factors were also predictive in T2DM, albeit with lower odds ratios. In T1DM, women are significantly more at risk than men. In T2DM, the use of sulphonylureas or glinides and absence of coronary heart disease were further predictive factors. Surprisingly, glycated haemoglobin (HbA1c ) levels were not predictive of hypoglycaemia in either T1DM or T2DM. To further assess the relationship between glycaemic control and hypoglycaemia risk, the rate of hypoglycaemia was analysed by tertiles of HbA1c (Table 4). In T1DM, the rate of confirmed hypoglycaemia (events/patient/month) was significantly increased in the lowest tertile [HbA1c < 7.0% (< 53 mmol/mol)] compared with the highest tertile [HbA1c > 8.4% (> 68 mmol/mol)]. In contrast, in T2DM the rate of hypoglycaemia was similar in both the highest and lowest tertiles of HbA1c . Table 3 Predictive factors in relation to the occurrence of hypoglycaemia. Factor Type 1 diabetes mellitus (n = 1304) History of severe or non-severe hypoglycaemia > 2 daily insulin injections vs 1 or 2 injectionsa BMI < 30 kg/m2 vs BMI > 30 kg/m2 Duration of insulin therapy > 10 years vs ≤ 10 years Age < 65 years vs age > 65 years Absence of peripheral arterial disease Female vs male Type 2 diabetes mellitus (n = 1631) History of severe or non-severe hypoglycaemia > 2 vs 1 daily insulin injectionsb Duration of insulin therapy > 10 years vs ≤ 10 years 2 vs 1 daily insulin injections Use of sulphonylurea or glinide Absence of coronary heart disease BMI < 30 kg/m2 vs BMI > 30 kg/m2
Odd ratio (95% CI) 8.17 (4.17–15.97) 2.75 (1.77–4.29) 2.14 (1.39–3.31)
The impact of the number of self-measured blood glucose (SMBG) tests per day on the frequency of confirmed hypoglycaemia was assessed in both T1DM and T2DM groups. The frequency of non-severe and severe hypoglycaemia increased in parallel with the number of SMBG tests in T1DM for all categories examined (Table 5) whereas, in T2DM, the frequency increased with the number of SMBG tests except for the severe and nocturnal severe hypoglycaemia categories. 3.5. Retrospective survey In the self-reported retrospective survey, a combined total of 969 patients (25.9%) reported experiencing a severe hypoglycaemic event during the previous year: 31.5% and 21.7% of patients with T1DM and T2DM, respectively (Table 6). More patients reported diurnal than nocturnal severe events in both T1DM and T2DM. In comparison, physicians reported on CRFs that 23.6% and 11.9% of patients with T1DM and T2DM had experienced severe hypoglycaemia over the same period. 3.6. Medical consequences of severe hypoglycaemia In the prospective set, 10.7% and 7.8% of severe events in T1DM and T2DM, respectively, entailed complications, including transport to an emergency unit (1.8% and 1.2%, respectively; Table 7). In addition, 7.1% and 10.2% of patients reported that their severe hypoglycaemia had prompted them to contact a healthcare professional (HCP) outside of their regular consultations. In the retrospective set, 18.6% and 9.2% of severe events in T1DM and T2DM, respectively, entailed complications, including loss of consciousness (15.5% and 6.7%, respectively) and transport to an emergency unit (6.0% and 3.7%, respectively). The percentage of patients contacting an HCP because of severe hypoglycaemia was 8.6% and 14.5%, respectively (Table 7).
2.00 (1.42–2.83)
4. Discussion 1.87 (1.27–2.74) 1.82 (1.08–3.09) 1.79 (1.26–2.53) 3.52 (2.50–4.95) 2.69 (2.03–3.57) 1.72 (1.34–2.21)
1.55 (1.15–2.09) 1.55 (1.19–2.01) 1.46 (1.15–1.86) 1.35 (1.09–1.66)
BMI: body mass index. a Patients using an insulin pump were considered as using > 2 injections. b Patients using an insulin pump were excluded.
DIALOG was a large study to assess the frequency of welldefined confirmed hypoglycaemia by a unique combination of complementary prospective and retrospective approaches. The results of DIALOG confirm that iatrogenic hypoglycaemia is a major concern in insulin-treated patients with diabetes in a reallife setting. In the main 1-month prospective survey, confirmed hypoglycaemic events occurred in 85.3% and 43.6% of patients with T1DM and T2DM, respectively. For T1DM, the percentage of patients reporting severe hypoglycaemia during the preceding 12 months was 23.6% by CRF and 31.5% by retrospective questionnaire (the latter with an event rate of 0.08 events/patient/month). By prospective questionnaire, 13.4% of patients reported experiencing severe hypoglycaemia (0.2 events/patient/month). For T2DM, a striking finding was the large proportion of patients who experienced severe hypoglycaemia over the preceding 12 months: 11.9% according to CRF and 21.7% according to retrospective questionnaire (0.05 events/month). By prospective questionnaire, 6.4% of patients
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
121
Table 4 Hypoglycaemia according to tertiles of HbA1c measured at selection visit: Prospective Set (n = 3048). Type 1 diabetes mellitus (n = 1303)
Allb n (%)a Events/patient/month Severe n (%)a Events/patient/month Non-severe n (%)a Events/patient/month Nocturnal n (%)a Events/patient/month Severe nocturnal n (%)a Events/patient/month Diurnal n (%)a Events/patient/month Type 2 diabetes mellitus (n = 1708)
Allb n (%)a Events/patient/month Severe n (%)a Events/patient/month Non-severe n (%)a Events/patient/month Nocturnal n (%)a Events/patient/month Severe nocturnal n (%)a Events/patient/month Diurnal n (%)a Events/patient/month
Tertile 1 HbA1c < 7.0% (< 53 mmol/mol) (n = 276)
Tertile 2 HbA1c 7.0–8.4% (53–68 mmol/mol) (n = 716)
Tertile 3 HbA1c > 8.4% (> 68 mmol/mol) (n = 311)
P-value (tertile 1 vs tertile 3)
230 (83.3) 8.0 ± 8.7
619 (86.5) 7.6 ± 7.2
262 (84.2) 6.3 ± 6.3
0.765 (Chi2 ) 0.0095 (T)
45 (16.3) 0.3 ± 0.8
85 (11.9) 0.2 ± 0.6
45 (14.5) 0.2 ± 0.9
0.538 (Chi2 ) 0.568 (T)
227 (82.2) 7.7 ± 8.5
618 (86.3) 7.4 ± 7.1
257 (82.6) 6.1 ± 6.1
0.901 (Chi2 ) 0.0095 (T)
102 (37.0) 0.7 ± 1.2
294 (41.1) 0.8 ± 1.3
127 (40.8) 0.7 ± 1.1
0.336 (Chi2 ) 0.592 (T)
13 (4.7) 0.1 ± 0.3
31 (4.3) 0.1 ± 0.3
11 (3.5) 0.0 ± 0.2
0.474 (Chi2 ) 0.529 (T)
220 (79.7) 7.2 ± 8.2
604 (84.4) 6.7 ± 6.7
253 (81.4) 5.5 ± 6.0
0.616 (Chi2 ) 0.005 (T)
Tertile 1 HbA1c < 7.0% (< 53 mmol/mol) (n = 397)
Tertile 2 HbA1c 7.0–8.4% (53–68 mmol/mol) (n = 899)
Tertile 3 HbA1c > 8.4% (> 68 mmol/mol) (n = 412)
P-value (tertile 1 vs tertile 3)
181 (45.6) 1.8 ± 3.4
401 (44.6) 1.7 ± 3.2
168 (40.8) 1.8 ± 3.4
0.167 (Chi2 ) 0.926 (T)
29 (7.3) 0.1 ± 0.4
45 (5.0) 0.1 ± 0.4
36 (8.7) 0.1 ± 0.6
0.454 (Chi2 ) 0.154 (T)
171 (43.1) 1.7 ± 3.3
388 (43.2) 1.6 ± 3.1
158 (38.3) 1.6 ± 3.3
0.172 (Chi2 ) 0.764 (T)
37 (9.3) 0.2 ± 0.7
98 (10.9) 0.2 ± 0.7
55 (13.3) 0.2 ± 0.7
0.071 (Chi2 ) 0.177 (T)
1 (0.3) 0.0 ± 0.1
11 (1.2) 0.0 ± 0.1
9 (2.2) 0.0 ± 0.2
0.013 (Chi2 ) 0.013 (T)
168 (42.3) 1.6 ± 3.0
377 (41.9) 1.4 ± 2.8
154 (37.4) 1.5 ± 3.0
0.151 (Chi2 ) 0.597 (T)
Event rates are shown as means ± SD; T: Student’s t test. a Patients reporting at least one episode. b Severe and confirmed non-severe.
with T2DM reported experiencing severe hypoglycaemia (0.1 events/month). For both T1DM and T2DM, prospectively reported event rates of severe hypoglycaemia were higher than those reported retrospectively (T1DM: 0.2 ± 0.7 and 0.08 ± 0.18 events/patient/month, respectively; T2DM: 0.1 ± 0.4 and 0.05 ± 0.15 events/patient/month, respectively). One possible reason is that patients were particularly conscientious about recording events during the prospective study and/or did not recall events very well when asked to think back over the previous year. The important implication is that event rates for severe
hypoglycaemia reported in retrospective observational studies may well underestimate the true frequency. Event rates have been reported here as mean values, as is the usual practice, with no adjustment to reflect the possibility that a few patients may have had many recurrent episodes, while others may have had none. Yet, it is apparent from the median values and ranges (Table 2) that some patients did indeed have very high numbers of events. These patients were not excluded from our analysis, as our aim was to report the actual frequency of hypoglycaemia as observed in real life.
122
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
Table 5 Hypoglycaemia according to number of self-measured blood glucose tests per day: prospective set (n = 3048). Type 1 diabetes mellitus
< 3 times/day (n = 99)
3– < 4 times/day (n = 221)
4–5 times/day (n = 548)
> 5 times/day (n = 449)
Total (n = 1317)
P-value (Chi2 ) for overall trend
46 (46.5) [36.6, 56.3]
160 (72.4) [66.5, 78.3]
497 (90.7) [88.3, 93.1]
420 (93.5) [91.3%, 95.8]
1123 (85.3) [83.4, 87.2]
< 0.001
Alla n (%)b 95% CI Severe n (%)b 95% CI Non-severe n (%)b 95% CI
10 (10.1) [4.2, 16.0]
22 (10.0) [6.0, 13.9]
69 (12.6) [9.8, 15.4]
6 (16.9) [13.5, 20.4]
177 (13.4) [11.6, 15.3]
0.040
44 (44.4) [34.7, 54.2]
156 (70.6) [64.6, 76.6]
494 (90.1) [87.7, 92.6]
418 (93.1) [90.8, 95.4]
1112 (84.4) [82.5, 86.4]
< 0.001
Type 2 diabetes mellitus
< 2 times/day (n = 22)
2– < 3 times/day (n = 361)
3–< 4 times/day (n = 594)
≥ 4 times/day (n = 549)
Total (n = 1729)
P-value (Chi2 ) for overall trend
56 (24.9) [19.2, 30.5]
111 (30.7) [26.0, 35.5]
266 (44.8) [40.8, 48.8]
321 (58.5) [54.3, 62.6]
754 (43.6) [41.3, 45.9]
< 0.001
6 (2.7) [0.6, 4.8]
25 (6.9) [4.3, 9.5]
38 (6.4) [4.4, 8.4]
41 (7.5) [5.3, 9.7]
110 (6.4) [5.2, 7.5]
0.091
53 (23.6) [18.0, 29.1]
100 (27.7) [23.1, 32.3]
256 (43.1) [39.1, 47.1]
312 (56.8) [52.7, 61.0]
721 (41.7) [39.4, 44.0]
< 0.001
Alla n (%)b 95% CI Severe n (%)b 95% CI Non-severe n (%)b 95% CI
Values are n (%) of patients reporting at least one episode. a Severe and confirmed non-severe. b Patients reporting at least one episode.
Table 6 One-year occurrence of severe hypoglycaemia: retrospective set. Type 1 diabetes mellitus (n = 1592) Frequency n (%)a Severe Nocturnal severe Diurnal severe a
Type 2 diabetes mellitus (n = 2151) Events/patient/month (mean ± SD)
Frequency n (%)a
0.08 ± 0.18 0.03 ± 0.08 0.05 ± 0.13
502 (31.5) 270 (17.0) 381 (23.9)
Events/patient/month (mean ± SD) 0.05 ± 0.15 0.01 ± 0.05 0.04 ± 0.12
467 (21.7) 128 (6.0) 385 (17.9)
Patients reporting at least one episode during the preceding 12 months; all events were self-reported.
Table 7 Medical consequences of severe hypoglycaemia. Prospective survey
Total of events with complicationsc Loss of consciousness Transport to emergency unit Hospitalization Fracture or trauma Contact with a healthcare professionald a
(30 days)
Retrospective survey
(12 months)
T1DM (n = 281)a n (%)b
T2DM (n = 166)a n (%)b
Total (n = 447)a n (%)b
T1DM (n = 1577)a n (%)b
T2DM (n = 1365)a n (%)b
Total (n = 2942)a n (%)b
30 (10.7) 26 (9.3) 5 (1.8) 2 (0.7) 2 (0.7) 20 (7.1)
13 (7.8) 7 (4.2) 2 (1.2) 5 (3.0) 0 (0.0) 17 (10.2)
43 (9.6) 33 (7.4) 7 (1.6) 7 (1.6) 2 (0.4) 37 (8.3)
293 (18.6) 244 (15.5) 95 (6.0) 61 (3.9) 31 (2.0) 136 (8.6)
125 (9.2) 92 (6.7) 51 (3.7) 54 (4.0) 5 (0.4) 198 (14.5)
418 (14.2) 336 (11.4) 146 (5.0) 115 (3.9) 36 (1.2) 334 (11.4)
Number of severe episodes. Events self-reported by patients completing the prospective questionnaire; percentages calculated from total number of events (presence of a characteristic/number of events). c Derived variable (at least loss of consciousness/transport to emergency room/hospitalisation/fracture or trauma). d Patients who responded “yes” to the question: Did you contact a healthcare professional as a result of this hypoglycaemia episode? b
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
The present results are broadly in line with those of previous studies [9,12,25,26]. In two small prospective studies (< 400 patients), for example, rates of severe hypoglycaemia (events/patient/year) in T1DM patients were reported as 1.15 [9] and 3.2 [12], and in T2DM patients as 0.35 [9] and 0.70 [12]. The results in the present prospective survey were 2.4 events/patient/year in T1DM and 1.2 events/patient/year in T2DM; thus, our results for T2DM are somewhat higher than those reported earlier. Previously reported event rates for severe hypoglycaemia have generally been lower for T2DM than for T1DM [9,12]. Surprisingly, in the present study, among those patients who experienced one or more episodes of hypoglycaemia during the 30-day follow-up, event rates (expressed as events/patient/month) for severe hypoglycaemia, including nocturnal hypoglycaemia, were closely similar in both T2DM and T1DM (R2 values, Table 2). In fact, large proportions of patients (40.2% and 11.0% in T1DM and T2DM, respectively) experienced nocturnal events. This is important because nocturnal hypoglycaemia is feared by patients and acts as a brake on intensification of insulin therapy. Another goal of DIALOG was to identify predictive factors for hypoglycaemia in real life. In our study, the most important predictive factors in both T1DM and T2DM were a history of severe or non-severe hypoglycaemia > 2 daily injections and duration of insulin therapy > 10 years. These are in line with previous reports [9]. In the UK Hypoglycaemia Study, event rates for T2DM treated with insulin for > 5 years were similar to those for T1DM treated for < 5 years [12]. However, in DIALOG, the frequency of hypoglycaemia in patients with T2DM with a duration of insulin therapy > 10 years remained lower than that of patients with T1DM, even when the duration of disease was shorter. Our analysis for predictive factors yielded some intriguing results. In T1DM, the risk of hypoglycaemia was higher in women than in men, an observation that has never been reported before. In patients with T2DM, the absence of coronary heart disease was a predictive factor for hypoglycaemia. It is possible that, in line with a patient-centred approach [21], physicians treating patients with coronary heart disease allow them to maintain fairly high glycaemic levels precisely to avoid hypoglycaemia. In addition, hypoglycaemia was less frequent in patients with T1DM aged ≥ 65 years. These results seem counterintuitive, as vulnerability to hypoglycaemia is normally greater with increasing age. Just as in patients with CVD, however, it is possible that older patients aim to achieve less stringent glycaemic targets. Moreover, there was no increased risk of hypoglycaemia in patients with an estimated glomerular filtration rate of ≤ 60 mL/min (data not shown; data were not available for patients with more advanced chronic kidney disease). An unexpected observation was that HbA1c was not predictive of hypoglycaemia in either T1DM or T2DM. It is conceivable that patients susceptible to hypoglycaemia aimed for less stringent glycaemic targets, and that the patients with lower HbA1c were those for whom hypoglycaemia was not a problem. Analysis of the rate of hypoglycaemia by tertiles of HbA1c led to discordant results between T1DM and T2DM.
123
Results from the DCCT showed that the rate of severe hypoglycaemia increased as HbA1c levels decreased [2] and, in our study, the risk of hypoglycaemia was highest in the lowest tertile in T1DM. In contrast, there was a trend towards a U-shaped association between tertiles of HbA1c and hypoglycaemia risk in T2DM, with the higher risk for extreme values. Interestingly, an association between higher HbA1c levels and severe hypoglycaemia was previously reported in T2DM in the Fremantle observational cohort study [13]. The number of SMBG tests was positively correlated with the reported occurrence of hypoglycaemia in both T1DM and T2DM. This was to be expected, as the definition of confirmed hypoglycaemia required a blood glucose measurement (except in cases of severe hypoglycaemia). Patients who were testing more often would therefore be more likely to detect asymptomatic events. This is a limitation of all studies of hypoglycaemia that rely on blood glucose measurements to confirm an event. It could also suggest that increasing the number of SMBG tests failed to prevent the risk of hypoglycaemia, and that high numbers of SMBG tests likely reflected high glycaemic variability in the patients concerned, a phenomenon not addressed in the present study. Our study also showed that severe hypoglycaemia incurred medical consequences such as hospitalisation. The economic implications for individual patients and the healthcare system in general should be assessed more thoroughly in health-economic analyses. The present study had a few limitations. Our definition of severe hypoglycaemia corresponded to the intentions of the ADA definition, but our patients did not specifically record whether, in the absence of a plasma glucose concentration, neurological recovery followed the return of plasma glucose to normal. However, it seems reasonable to assume this was the case if the patient was able to complete the questionnaire. Also, the data obtained were dependent on patient self-reporting and recall, and not all patients returned analysable questionnaires: 69% of all included patients qualified for the prospective set and 85% for the retrospective set. It is possible that the patients who failed to complete their questionnaires had different characteristics or event rates from those who did. However, the key demographic and diabetes-related characteristics were similar for both the overall and prospective sets. In fact, our study methodology incorporated a non-inclusion registry to ensure that the included population was representative of real-life patients as seen by physicians. Nevertheless, a recruitment bias cannot be excluded, as the included patients were probably more at risk than non-included ones, with a higher prevalence of both T1DM and previous hypoglycaemia. It should also be mentioned that most of our patients were recruited by specialists (endocrinologists) rather than by GPs; this imbalance might have led to the recruitment of patients with more severe hypoglycaemia, especially for patients with T2DM. The present study had a novel design and a number of strengths compared with previous studies. For the first time, severe hypoglycaemia was reported simultaneously in three ways in the same cohort: retrospectively through the physician’s CRF; retrospectively by questionnaire; and prospectively
124
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125
by questionnaire. Hypoglycaemia was strictly defined to include only confirmed hypoglycaemia according to ADA definitions [23]. Also, the distribution of participating physicians was representative of real life. The number of patients was exceptionally large for a prospective survey of hypoglycaemia, and the frequency and rate of hypoglycaemia were assessed by both prospective and retrospective approaches. 5. Conclusion The DIALOG study provides a comprehensive, up-to-date record of the prevalence of hypoglycaemia, in particular severe hypoglycaemia, among insulin-treated patients with diabetes in real life. The relatively high frequency of hypoglycaemia reported underlines the importance of optimising patients’ insulin regimens to successfully control glycaemia while minimizing the occurrence of hypoglycaemia. The study also confirmed that previous episodes of hypoglycaemia increase the risk of future episodes, thus supporting the concept of a patient-centred approach for managing diabetes in which hypoglycaemia risk evaluation is critical. Contributors Bertrand Cariou was the lead investigator. All the authors except S. Lavigne contributed to designing the study, conducting it and collecting data. S. Lavigne collated the data and conducted the statistical analyses. B. Cariou specified the contents of the article and wrote the first draft in conjunction with a medical writer. All the authors reviewed and commented on drafts of the article. Disclosure of interest B. Cariou has taken part in advisory panels for, and received research support from, Novo Nordisk and Sanofi; attended advisory boards for Amgen, AstraZeneca, Genfit, Janssen, Novo Nordisk and Sanofi-Aventis; and acted as a consultant for Genfit. P. Fontaine has received fees for participation to advisory board for Novo Nordisk, Eli Lilly, Sanofi-Aventis, Merck Sharp & Dohme, Boehringer Ingelheim and GlaxoSmithKline. E. Eschwege has taken part in an advisory panel for the French National Authority for Health: Transparency Committee: Working group for Public Health Benefit Assessment of drugs; and attended advisory boards for Roche, Novartis and Novo Nordisk. M. Lièvre has attended advisory boards for Merck-Serono, Novartis, Medtronic, Terumo, Novo Nordisk, Abbott, Bard, Boston Scientific, Cordis, Takeda, AstraZeneca and Johnson & Johnson, and served on an advisory panel for Johnson & Johnson (no personal remuneration). D. Gouet has received research support from Novo Nordisk, Eli Lilly, Boehringer Ingelheim, Medtronic, Johnson & Johnson and Sanofi-Aventis. D. Huet has attended advisory boards for Sanofi-Aventis, Novo Nordisk and Novartis. S. Madani is an employee of Novo Nordisk. S. Lavigne is an employee of ITEC services SAS, a contract research organization paid by Novo Nordisk Pharmaceutique to collate and analyse the data. B. Charbonnel has received
fees for consultancy, speaking, travel and/or accommodation from AstraZeneca, Bristol-Myers Squibb, Boehringer Ingelheim, GlaxoSmithKline, Janssen, Eli Lilly, Merck Sharpe & Dohme, Novartis, Novo Nordisk, Roche, Sanofi-Aventis and Takeda. Acknowledgements The authors would like to thank Grace Townshend and Mark Nelson, Watermeadow Medical, Witney, UK, for their assistance in the writing, editing and submission of this manuscript. This assistance was funded by Novo Nordisk. The authors thank all of the physicians involved in the study for their participation. Role of the funding source: Novo Nordisk sponsored the study. The trial was designed by an independent scientific committee (Appendix A, see supplementary material associated with this article online) in collaboration with Novo Nordisk. Data were collected by a contract research organization (ITEC), which conducted the statistical analysis of the data and wrote the study report (S.L.). All authors had full access to the data and share final responsibility for the content of the manuscript and the decision to submit for publication.
Appendix A. Supplementary data Supplementary data (Figs. S1, S2, S3, S4 and Appendix) associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.diabet.2014.10.007. References [1] Frier BM, Heller SR. Epidemiology and impact of hypoglycaemia on patients with diabetes. Translational endocrinology & metabolism: hypoglycaemia in diabetes update, 3; 2012. p. 15–47. [2] Diabetes Control, Complications Trial Research Group. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl J Med 1993;329:977–86. [3] Diabetes Control and Complications Trial Research Group. Hypoglycemia in the Diabetes Control and Complications Trial. Diabetes 1997;46: 271–86. [4] Nathan DM, DCCT/EDIC Research Group. The diabetes control and complications trial/epidemiology of diabetes interventions and complications study at 30 years: overview. Diabetes Care 2014;37: 9–16. [5] UK Prospective Diabetes Study (UKPDS) Group. Intensive blood glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes (UKPDS 33). Lancet 1998;352:837–53. [6] Seaquist ER, Miller ME, Bonds DE, Feinglos M, Goff Jr DC, Peterson K, et al. The impact of frequent and unrecognized hypoglycaemia on mortality in the ACCORD study. Diabetes Care 2012;35:409–14. [7] Bruderer SG, Bodmer M, Jick SS, Bader G, Schlienger RG, Meier CR. Incidence of and risk factors for severe hypoglycaemia in treated type 2 diabetes mellitus patients in the UK - a nested case-control analysis. Diabetes Obes Metab 2014;16:801–11.
B. Cariou et al. / Diabetes & Metabolism 41 (2015) 116–125 [8] Kulzer B, Seitz L, Kern W. Real-world patient-reported rates of non-severe hypoglycaemic events in Germany. Exp Clin Endocrinol Diabetes 2014;122:167–72. [9] Donnelly LA, Morris AD, Frier BM, Ellis JD, Donnan PT, Durrant R, et al. Frequency and predictors of hypoglycaemia in type 1 and insulin-treated type 2 diabetes: a population-based study. Diabet Med 2005;22:749–55. [10] Henderson JN, Allen KV, Deary IJ, Frier BM. Hypoglycaemia in insulin-treated type 2 diabetes: frequency, symptoms and impaired awareness. Diabet Med 2003;20:1016–21. [11] Pedersen-Bjergaard U, Pramming S, Heller SR, Wallace TM, Rasmussen AK, Jørgensen HV, et al. Severe hypoglycaemia in 1076 adult patients with type 1 diabetes: influence of risk markers and selection. Diabetes Metab Res Rev 2004;20:479–86. [12] UK Hypoglycaemia Study Group. Risk of hypoglycaemia in types 1 and 2 diabetes: effects of treatment modalities and their duration. Diabetologia 2007;50:1140–7. [13] Davis TM, Brown SG, Jacobs IG, Bulsara M, Bruce DG, Davis WA. Determinants of severe hypoglycaemia complicating type 2 diabetes: the Fremantle diabetes study. J Clin Endocrinol Metab 2010;95:2240–7. [14] Amiel SA, Dixon T, Mann R, Jameson K. Hypoglycaemia in type 2 diabetes. Diabet Med 2008;25:245–54. [15] McCoy RG, Van Houten HK, Ziegenfuss JY, Shah ND, Wermers RA, Smith SA. Increased mortality of patients with diabetes reporting severe hypoglycaemia. Diabetes Care 2012;35:1897–901. [16] ACCORD Study Group. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med 2008;358:2545–59. [17] ADVANCE Collaborative Group. Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes. N Engl J Med 2008;358:2560–72. [18] Duckworth W, Abraira C, Moritz T, Reda D, Emanuele N, Reaven PD, et al. Glucose control and vascular complications in veterans with type 2 diabetes. N Engl J Med 2009;360:129–39.
125
[19] Bonds DE, Miller ME, Bergenstal RM, Buse JB, Byington RP, Cutler JA, et al. The association between symptomatic, severe hypoglycaemia and mortality in type 2 diabetes: retrospective epidemiological analysis of the ACCORD study. BMJ 2010;340:b4909. [20] Barendse S, Singh H, Frier BM, Speight J. The impact of hypoglycaemia on quality of life and related patient-reported outcomes in type 2 diabetes: a narrative review. Diabet Med 2012;29: 293–302. [21] Inzucchi SE, Bergenstal RM, Buse JB, Diamant M, Ferrannini E, Nauck M, et al. Management of hyperglycaemia in type 2 diabetes: a patient-centered approach. Position statement of the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD). Diabetologia 2012;55:1577–96. [22] Workgroup on Hypoglycemia American Diabetes Association. Defining and reporting hypoglycaemia in diabetes: a report from the American Diabetes Association Workgroup on Hypoglycemia. Diabetes Care 2005;28:1245–9. [23] Seaquist ER, Anderson J, Childs B, Cryer P, Dagogo-Jack S, Fish L, et al. Hypoglycemia and diabetes: a report of a workgroup of the American Diabetes Association and the Endocrine Society. Diabetes Care 2013;36:1384–95. [24] International Society for Pharmacoepidemiology. Guidelines for good pharmacoepidemiology practices (GPP), 2007; 2013 [Available from: URL] http://www.pharmacoepi.org/resources/guidelines 08027.cfm [25] Verges B, Brun JM, Tawil C, Alexandre B, Kerlan V. Strategies for insulin initiation: insights from the French LIGHT observational study. Diabetes Metab Res Rev 2012;28:97–105. [26] Charbonnel B, Eschwege E, Charpentier G, Vialettes B, Chabrier G, Suissa S, et al. Insulin glargine in real use conditions in France [article in French]. Med Mal Metabol 2010;4:169–78.
