Objective: We examined the association between glycated hemoglobin (HbA 1c ) and all-cause mortality in patients with type 1 diabetes mellitus.

Transcription

1 ORIGINAL Endocrine ARTICLE Care Glycemic Control and All-Cause Mortality Risk in Type 1 Diabetes Patients: The EURODIAB Prospective Complications Study Danielle A. J. M. Schoenaker, Dominique Simon, Nish Chaturvedi, John H. Fuller, Sabita S. Soedamah-Muthu, and the EURODIAB Prospective Complications Study Group School of Population Health (D.A.J.M.S.), University of Queensland, Brisbane, 4006 Queensland, Australia; Department of Diabetes, la Pitié Hospital and University Pierre et Marie Curie, Paris, France; INSERM CESP (D.S.), U-1018, Villejuif, France; National Heart and Lung Institute (N.C.), Imperial College London, London W2 1PG, United Kingdom; Department of Epidemiology and Public Health (J.H.F.), University College London, London WC1E 6BT, United Kingdom; and Division of Human Nutrition (S.S.S.-M.), Wageningen University, 6700 EV Wageningen, The Netherlands Context: Glycemic targets and the benefit of intensive glucose control are currently under debate because intensive glycemic control has been suggested to have negative effects on mortality risk in type 2 diabetes patients. Objective: We examined the association between glycated hemoglobin (HbA 1c ) and all-cause mortality in patients with type 1 diabetes mellitus. Design, Setting, and Patients: A clinic-based prospective cohort study was performed in 2764 European patients with type 1 diabetes aged years enrolled in the EURODIAB Prospective Complications Study. Outcome Measure: Possible nonlinearity of the association between HbA 1c and all-cause mortality was examined using multivariable restricted cubic spline regression using three (at HbA 1c 5.6%, 8.1%, and 11.8%) and five knots (additionally at HbA 1c 7.1% and 9.5%). Mortality data were collected approximately 7 years after baseline examination. Results: HbA 1c was related to all-cause mortality in a nonlinear manner after adjustment for age and sex. All-cause mortality risk was increased at both low (5.6%) and high (11.8%) HbA 1c compared with the reference (median HbA 1c : 8.1%) following a U-shaped association [P overall effect.008 and.04, P nonlinearity.03 and.11 (three and five knots, respectively)]. Conclusions: Results from our study in type 1 diabetes patients suggest that target HbA 1c below a certain threshold may not be appropriate in this population. We recognize that these low HbA 1c levels may be related to anemia, renal insufficiency, infection, or other factors not available in our database. If our data are confirmed, the potential mechanisms underlying this increased mortality risk among those with low HbA 1c will need further study. (J Clin Endocrinol Metab 99: , 2014) ISSN Print X ISSN Online Printed in U.S.A. Copyright 2014 by the Endocrine Society Received July 11, Accepted December 13, First Published Online January 2, 2014 Patients with type 1 diabetes mellitus are at increased mortality risk (1). Treatment of these patients is focused on prevention of micro- and macrovascular complications by aiming to keep glycated hemoglobin (HbA 1c ) levels within a normal range. A target HbA 1c level of less than 7% is currently recommended, with different goals Abbreviations: ACCORD, Action to Control Cardiovascular Risk in Diabetes; AER, albumin excretion rate; BMI, body mass index; BP, blood pressure; CI, confidence interval; CVD, cardiovascular disease; DCCT, Diabetes Control and Complications Trial; HbA 1c, glycated hemoglobin; HDL, high-density lipoprotein; HR, hazard ratio; IQR, interquartile range; PCS, prospective complications study. For editorial see page jcem.endojournals.org J Clin Endocrinol Metab, March 2014, 99(3): doi: /jc

2 doi: /jc jcem.endojournals.org 801 for selected individual patient groups (2). Results from intervention and observational studies in patients with type 2 diabetes mellitus have raised attention to safety of aiming for intensive glucose control. Intensive glycemic control has been suggested to have negative effects on mortality risk in type 2 diabetes patients from the Action to Control Cardiovascular Risk in Diabetes (ACCORD) study (3), but no statistically significant association was found between HbA 1c and all-cause mortality from metaanalyses of clinical trials (4, 5). An observational study by Currie et al (6) among patients with type 2 diabetes showed a U-shaped association between HbA 1c and allcause mortality with increased mortality risk at low and high levels, especially in patients treated with insulin. These findings may have important implications for treatment of patients with type 2 diabetes, but it remains unclear whether they may also apply to insulin-dependent, or type 1, diabetes patients. Risk factors for complications are different for type 1 and type 2 diabetes and earlier disease onset in type 1 diabetes increases the lifetime risk of complications more than in type 2 diabetes patients (7). Findings in type 2 diabetes patients can therefore not automatically be translated to type 1 diabetes patients and need further investigation. We aimed to explore the relationship between HbA 1c and all-cause mortality in a prospective cohort study of patients with type 1 diabetes. Materials and Methods Study design and population The EURODIAB Prospective Complications Study (PCS) is a clinic-based prospective cohort study including 3250 men and women aged between 15 and 60 years with type 1 diabetes. Participants were recruited from 31 centers in 16 European countries and examined between 1989 and Details on patient selection and the standardized methodology have previously been published elsewhere (8). Type 1 diabetes mellitus was defined as diabetes diagnosed before the age of 36 years with a continuous need for insulin within 1 year of diagnosis. Of those invited, 85% participated. Those with duration of diabetes less than 1 year and pregnant women were excluded. Ethics committee approval conformed to the Declaration of Helsinki was obtained at each center; all participants provided written informed consent. Seven years after the baseline examination, study participants were invited for re-examination. For the present study, participants were excluded due to the following reasons: participants could not be assessed or the center did not participate in follow-up examination (n 464) or no information on baseline HbA 1c (n 22), leaving 2764 participants for analysis. Outcome assessment Data on mortality were collected at follow-up. All events were captured by death certificates. In addition, questionnaires and hospital records were used to classify underlying cause of death. A comparison of allocation of cause of death was performed separately by two observers with 100% agreement. Exposure assessment Baseline HbA 1c was measured centrally with an enzyme immunoassay using a monoclonal antibody (The Royal London Hospital, London/Dako). HbA 1c values obtained were converted to Diabetes Control and Complications Trial (DCCT) (percentage) (9) and International Federation of Clinical Chemistry and Laboratory Medicine (millimoles per mole) values in line with the European Association for the Study of Diabetes recommendations (10). Assessment of baseline characteristics Blood pressure was recorded in a sitting position with a random zero sphygmomanometer (Hawskley) and taken as the mean of two measurements. Total and high-density lipoprotein (HDL) cholesterol and triacylglycerol were determined using fasting serum and standard enzymatic methods (Boehringer Mannheim). All analyses were performed centrally. Low-density lipoprotein cholesterol was calculated using the Friedewald equation if triglycerides were less than 4 mmol/l (11). Information on diabetes duration, daily insulin dose and frequency, antihypertensive medication use, physical activity, smoking status, severe hypoglycemic events, and cardiovascular disease (CVD) was collected by questionnaire. Severe hypoglycemic events were defined as the number of hypoglycemic attacks over the past year, serious enough to require the help of another person. Measurements were performed for the assessment of microvascular complications (retinopathy, neuropathy, and nephropathy) (12). Weight and height were measured with indoor clothing without shoes, and body mass index (BMI) was computed as weight (kilograms)/ height (square meters). Statistical analysis Baseline characteristics of the study population were presented by quintiles of baseline HbA 1c and mortality status at follow-up as mean (SD), median [interquartile range (IQR)] or n (percentage). Differences in characteristics were assessed by P for trend for quintiles of HbA 1c and by independent-samples t tests (continuous variables) or 2 tests (categorical variables) for mortality status at follow-up. In preliminary analyses, linear and quadratic models were fitted and compared using the log-likelihood ratio test. Restricted cubic spline regression (13) was used to examine the possible nonlinear relationship between baseline HbA 1c and all-cause mortality risk at follow-up. Splines are piecewise polynomial functions and hold a high degree of smoothness at points called knots. Restricted cubic spline regression models provide a more comprehensive insight in the dose-response relationship compared with an analysis using categorized data, particularly when the sample size does not allow the use of smaller categories. It allows the use of all data points to estimate risk at each level of exposure. Restricted cubic spline models were tested using three to five knots: the model fit, as assessed by Akaike s information criterion (14), did not substantially differ, and the shape of the association was comparable. Statistical measures are affected by the number of knots; we therefore present results using both three and five knots. Three knots were placed at the fifth, 50 th, and 95th percentiles (HbA 1c 5.6%, 8.1%, and 11.8%)

3 802 Schoenaker et al Glycemic Control and Mortality in Type 1 Diabetes J Clin Endocrinol Metab, March 2014, 99(3): Table 1. Baseline Characteristics of Type 1 Diabetes Patients in the EURODIAB PCS by Quintiles of Baseline HbA 1c (n 2764) Characteristic Quintiles of Baseline HbA 1c Q1 Q2 Q3 Q4 Q5 P Value a n HbA1c, %, mmol/mol (43.2) (56.3) (66.1) (77.0) (100.0).0001 All-cause deaths, n, % 21 (3.7) 13 (2.4) 19 (3.4) 21 (4.0) 27 (4.9).12 Mean follow-up, y Age, y Sex, n, %, male 287 (51.7) 275 (51.8) 295 (54.1) 257 (50.9) 245 (47.1).15 Physical activity, n, %.53 Physical inactivity 30 (5.4) 19 (3.5) 16 (2.9) 13 (2.5) 9 (1.7) Mild physical activity 176 (31.4) 156 (28.9) 182 (32.7) 177 (33.9) 204 (37.4) 1 time/wk Moderate physical 181 (32.3) 180 (33.3) 193 (34.7) 186 (35.6) 166 (30.5) activity 1 time/wk Vigorous physical 174 (31.0) 185 (34.3) 165 (29.7) 146 (28.0) 166 (30.5) activity 1 time/wk Ever-smoker, n, % 260 (45.9) 237 (43.2) 291 (52.1) 283 (53.7) 290 (52.4).0001 BMI, kg/m Diabetes duration, y Insulin dose, units/d 0.6 ( ) 0.7 ( ) 0.7 ( ) 0.7 ( ) 0.7 ( ).0003 Severe hypoglycemia 223 (39.3) 223 (40.4) 167 (29.9) 148 (28.0) 121 (21.8).0001 events, n, % Systolic blood pressure, mm Hg Antihypertensive 44 (7.8) 57 (10.4) 49 (8.8) 52 (9.9) 55 (10.0).15 medication use, n, % Triacylglycerol, mmol/l 0.8 ( ) 0.9 ( ) 0.9 ( ) 1.0 ( ) 1.2 ( ).0001 Total to HDL cholesterol 3.0 ( ) 3.0 ( ) 3.2 ( ) 3.3 ( ) 3.4 ( ).0001 ratio, mmol/l Cardiovascular disease, 52 (9.3) 43 (7.9) 44 (8.0) 46 (8.8) 52 (9.4).75 n, % Retinopathy, n, % 161 (33.8) 202 (46.6) 229 (50.3) 210 (51.5) 219 (49.1).0001 Neuropathy, n, % 108 (19.3) 168 (30.9) 178 (32.6) 201 (38.7) 260 (47.9).0001 Nephropathy, n, %.0001 Normoalbuminuria 437 (80.5) 395 (74.5) 388 (72.3) 330 (65.2) 299 (56.7) (AER 20 g/min) Microalbuminuria 81 (14.9) 101 (19.1) 109 (20.3) 120 (23.7) 159 (30.2) (AER g/ min) Macroalbuminuria (AER 200 g/ min) 25 (4.6) 34 (6.4) 40 (7.5) 56 (11.1) 69 (13.1) Values are mean SD, median (IQR), or n (%). a P for trend. and additionally two knots were placed at the 25th and 75th percentiles when using five knots (HbA 1c 7.1% and 9.5%). The median HbA 1c value (8.1%) was selected as the reference value (13). Spline models were adjusted for age (years) and sex. In additional analyses, adjustments were made for diabetes duration (years), physical activity (physical inactivity; mild physical activity 1 time/wk; moderate physical activity 1 time/ wk; vigorous physical activity 1 time/wk) (15), BMI (kilograms per square meter), systolic blood pressure (BP) (millimeters of mercury), total to HDL cholesterol ratio (millimoles per liter), severe hypoglycemic events (no, yes), and nephropathy [albumin excretion rate (AER) 20 g/min; AER g/min; AER 200 g/min]. Data analysis was performed using SAS software version 9.3 (SAS Institute Inc). The SAS Macro RCS based on Cox proportional hazards regression (16) was used to conduct the spline analysis. Two-sided values of P.05 were considered statistically significant. Results Excluded participants Participants excluded for the present study due to no participation in follow-up examination (n 464) or missing data on baseline HbA 1c (n 22) had higher baseline HbA 1c levels, diastolic BP, triacylglycerol, and total to HDL cholesterol ratio levels; were more often smokers; more often used antihypertensive medication; and had higher prevalence of CVD and neuropathy (Supplemental Table 1, published on The Endocrine Society s Journals Online web site at Participant characteristics During a mean follow-up period of 7.2 years (SD 1.1), 101 all-cause deaths occurred among 2764 participants

4 doi: /jc jcem.endojournals.org 803 Table 2. Baseline Characteristics of Type 1 Diabetes Patients in the EURODIAB PCS by Mortality Status at Follow-Up (n 2764) Mortality Status at Follow-Up Characteristic Survived Deceased P Value a n HbA 1c, %, mmol/mol (68.3) (71.6).02 Mean follow-up, y Age, y Sex, n, %, male 1307 (50.8) 52 (61.9).04 Physical activity, n, %.0001 Physical inactivity 78 (3.0) 9 (9.1) Mild physical activity 1 time/wk 853 (32.5) 42 (42.4) Moderate physical activity 1 time/wk 873 (33.3) 33 (33.3) Vigorous physical activity 1 time/wk 821 (31.3) 15 (15.2) Ever-smoker, n, % 1303 (49.1) 58 (57.4).03 BMI, kg/m Diabetes duration, y Insulin dose, units/d 0.7 ( ) 0.6 ( ).03 Severe hypoglycemia events, n, % 844 (31.7) 38 (37.6).21 Systolic blood pressure, mm Hg Antihypertensive medication use, n, % 222 (8.4) 35 (34.7).0001 Triacylglycerol, mmol/l 0.9 ( ) 1.1 ( ).009 Total to HDL cholesterol ratio, mmol/l 3.2 ( ) 3.9 ( ).002 Cardiovascular disease, n, % 210 (8.0) 27 (26.7).0001 Retinopathy, n, % 976 (45.3) 45 (71.4).0001 Neuropathy, n, % 844 (32.3) 71 (71.7).0001 Nephropathy, n, %.0001 Normoalbuminuria 1,809 (70.9) 40 (43.0) (AER 20 g/min) Microalbuminuria 546 (21.4) 24 (25.8) (AER g/min) Macroalbuminuria (AER 200 g/min) 195 (7.7) 29 (31.2) Values are mean SD, median (IQR), or n (%). a P from independent-samples t test or 2 test. (51% male). Among men and women at baseline, the mean age was 33 years (SD 10), the mean diabetes duration was 14.7 years (SD 9.3), and the mean baseline HbA 1c was 8.4% (SD 1.9) (68.3 mmol/mol). Baseline characteristics of the study population according to quintiles of baseline HbA 1c and mortality status at follow-up are presented in Tables 1 and 2, respectively. HbA 1c was positively associated with previous or current smoking, triacylglycerol, and total to HDL cholesterol ratio levels, and microvascular complications and negatively associated with severe hypoglycemia events (Table 1). The proportion of type 1 diabetes patients who died during follow-up was lowest in the second quintile [baseline HbA 1c 7.3% (56.3 mmol/mol)]. Participants who died during follow-up were older; had higher HbA 1c, systolic BP, triacylglycerol, and total to HDL cholesterol ratio; and had longer diabetes duration. They were also more likely to be male, physically inactive or ever-smoker, use antihypertensive medication, and had CVD or microvascular complications at baseline (Table 2). HbA 1c and all-cause mortality risk The log-likelihood ratio test comparing linear and quadratic models indicated a borderline significant better fit for the nonlinear quadratic model (P.05). To gain flexibility and describe and present this nonlinear association, we performed restricted cubic spline regression. Restricted cubic spline results predict hazard ratios for all-cause mortality between the fifth and 95th percentiles of the population HbA 1c distribution [HbA 1c between 5.6% and 11.8% (37.7 and mmol/mol)]. HbA 1c was associated with all-cause mortality risk in a nonlinear manner. Levels below and above the median HbA 1c (8.1%) (65.0 mmol/mol) were associated with an increased all-cause mortality risk, indicating a U-shaped association. Results from a univariate analysis show a P for overall effect of HbA 1c of.0005 (three knots) and.01 (five knots) and P for nonlinearity of.01 (three knots) and.04 (five knots). Complete data were available after adjustment for age and sex for 2655 participants, among whom 84 died during follow-up. Results from our multivariate

5 804 Schoenaker et al Glycemic Control and Mortality in Type 1 Diabetes J Clin Endocrinol Metab, March 2014, 99(3): Figure 1. Association between HbA 1c and all-cause mortality risk in type 1 diabetes patients in the EURODIAB PCS. Restricted cubic spline regression with three knots located at the fifth, 50th, and 95th percentiles (HbA 1c 5.6%, 8.1%, and 11.8%). The y-axis represents adjusted HRs for all-cause mortality for any value of HbA 1c compared with a reference HbA 1c of 8.1%. HRs are adjusted for age and sex (P for overall effect of HbA 1c.008; P for nonlinearity.03). Dashed lines indicate 95% CIs. spline model including three knots showed a similar U- shaped relationship (P for overall effect.008) (P for nonlinearity.03) (Figure 1). Both low HbA 1c ( 5.6%, 37.7 mmol/mol, hazard ratio (HR) 1.31, 95% confidence interval (CI) 0.67, 2.42) and high HbA 1c ( 11.8%, mmol/mol, HR 1.86, 95% CI 1.12, 3.05) were associated with an increased all-cause mortality risk compared with the reference (8.1%, 65.0 mmol/l). A similar relationship with all-cause mortality risk was found when using five knots (P for overall effect of HbA 1c.04) (P for nonlinearity.11) (Figure 2), even though statistical significance was affected by the increased number of degrees of freedom as a result of the increased number of knots. Allcause mortality risk was lowest between HbA 1c values of approximately 7 and approximately 8% (53.0 and 63.9 mmol/mol). The shape of this association also remained after additional adjustments for diabetes duration, physical activity, BMI, systolic BP, and total to HDL cholesterol ratio, even though nonlinearity did not remain statistically significant (P for overall effect of HbA 1c.04 and.05 for three and five knots, respectively) (P for nonlinearity.11 and.19 for three and five knots, respectively) due to the decreased number of events as a result of missing values (n 1818 participants; all-cause death events, n 60). The 95% CIs increased with the decreasing number of participants on which sections with very low or high HbA 1c values were based. Adjustment for baseline severe hypoglycemia events in addition to age and sex did not change the positive association between low or high HbA 1c and all-cause mortality risk (P for overall effect.008 and.04 for three and five knots, respectively) (P for nonlinearity.03 and.11 for three and five knots, respectively) (Supplemental Figure 1, A and B). When additionally adjusting for nephropathy, the magnitude of the association below the reference HbA 1c became stronger, whereas the association for HbA 1c greater than the reference was slightly attenuated (P for overall effect.03 and.11 for three and five knots, respectively) (P for nonlinearity.02 and.10 for three and five knots, respectively) (Supplemental Figure 2, A and B). We did not have enough fatal CVD cases to examine the relationship with HbA 1c ; however, the distribution of these cases across HbA 1c quintiles was comparable with the U-shaped distribution of all-cause mortality cases across quintiles of HbA 1c (data not shown). Discussion Results from our study in type 1 diabetes patients from the EURODIAB PCS point toward a nonlinear relationship between HbA 1c and all-cause mortality risk. HbA 1c levels below and above our selected reference HbA 1c (median value) were associated with increased all-cause mortality risk after adjustment for age and sex. This U-shaped relationship also remained after additional adjustments for diabetes duration, physical activity, BMI, systolic BP and total to HDL cholesterol ratio. To our knowledge, the relationship between HbA 1c and all-cause mortality in type 1 diabetes patients has not been described before. Our findings are in agreement with findings in type 2 diabetes patients from two retrospective cohort studies (6, 17) and the ACCORD trial (3). Currie et al (6) showed a U-shaped relationship between HbA 1c and all-cause mortality in the UK General Practice Research database. In patients with insulin treatment, compared with the HbA 1c reference decile [median HbA 1c (percentage): 7.5, IQR ], the HR of all-cause mortality in the lowest HbA 1c decile [median HbA 1c (percentage): 6.4, IQR ] was 1.79 (95% CI 1.45, 2.22) and

6 doi: /jc jcem.endojournals.org 805 Figure 2. Association between HbA 1c and all-cause mortality risk in type 1 diabetes patients in the EURODIAB PCS. Restricted cubic spline regression with five knots is located at the fifth, 25th, 50th, 75th, and 95th percentiles (HbA 1c 5.6%, 7.1%, 8.1%, 9.5%, and 11.8%). The y-axis represents adjusted HRs for all-cause mortality for any value of HbA 1c compared with a reference HbA 1c of 8.1%. HRs are adjusted for age and sex (P for overall effect of HbA 1c.04; P for nonlinearity.11). Dashed lines indicate 95% CIs (95% CI 1.49, 2.17) in the highest HbA 1c decile [median HbA 1c (percentage): 10.5, IQR ]. Huang et al (17) found a U-shaped relationship between mortality and HbA 1c in older patients in the Diabetes and Aging Study. Findings from the ACCORD trial (3) show that in type 2 diabetes patients with CVD or at least two risk factors for CVD or severe atherosclerosis, intensive glycemic control (target HbA 1c 6.0%) vs standard strategy (HbA 1c %) was associated with increased mortality (HR 1.22, 95% CI 1.01, 1.46). Furthermore, post hoc epidemiological analyses of the ACCORD trial show increased risk for 1% higher HbA1c for the intensive strategy (HR 1.66, 95% CI 1.46, 1.89, P.0001) and the standard strategy (HR 1.14, 95% CI 0.95, 1.38, P.17). The risk of death increased linearly in the intensive treatment group, with no suggestion of increased risk at low HbA 1c levels, whereas a shallow U-shaped relationship between HbA 1c and all-cause mortality was seen in the standard treatment group (P for nonlinearity.0184). This may indicate that the increased risk with intensive compared with standard treatment occurs in patients with average HbA 1c greater than 7% (18). In line with our findings, the lowest mortality risk in the standard treatment group was associated with average HbA 1c between 7% and 8% (19). In the Action in Diabetes and Vascular Disease: Preterax and Diamicron Modified Release Controlled Evaluation, Zoungas et al (20) showed a weak U-shaped relationship between HbA 1c and mortality with the increased risk at the low end only in those with a HbA 1c of 5.5%, an unusual low level for diabetes patients. They report a possible threshold HbA 1c level of 7%. No significant effect of intensive glycemic control on mortality risk was found in the Veterans Affairs Diabetes Trial (21) and meta-analyses of randomized controlled trials (4, 5). The Diabetes Control and Complications Trial (DCCT) of 6.5 years mean duration (22) reports on HbA 1c and cardiovascular factors among type 1 diabetes patients, showing a major reduction in microvascular disease in the intensive group (HbA 1c 7.0%) compared with the conventional group (HbA 1c 9.0%) (22). At the end of the observational postrandomization follow-up study Epidemiology of Diabetes Interventions and Complications (11 y mean duration), during which both treatment groups maintained an identical HbA 1c level, a positive association between mean HbA 1c over time and multiple measures of cardiac function was found, regardless of intensive vs conventional treatment during the DCCT (23). In addition, a reduced risk of 57% (95% CI 12% 79%, P.02) for nonfatal myocardial infarction, stroke, or CVD death was found in the intensive treatment group after a mean of 17 years of follow-up (24), indicating a legacy effect of early intensive glycemic control. Intervention trials intentionally aim for certain HbA 1c levels to achieve intensive glycemic control in participants. The relationship between HbA 1c and risk for all-cause mortality is, however, best assessed in a prospective observational study. The American Diabetes Association recommends, based on findings from the major clinical trials in type 2 diabetes patients, less stringent HbA 1c target goals for patients with a history of severe hypoglycemia, limited life expectancy, advanced microvascular or macrovascular complications, and extensive comorbid conditions and those with long diabetes duration (2). Type 1 diabetes patients generally have a longer glycemic exposure than those with type 2 diabetes because of its onset at a younger age and are therefore at increased risk for micro- and macrovascular complications. This risk factor profile could indicate that when considering optimum target HbA 1c for type 1 diabetes patients, the lower the better might not be appropriate.

7 806 Schoenaker et al Glycemic Control and Mortality in Type 1 Diabetes J Clin Endocrinol Metab, March 2014, 99(3): The EURODIAB PCS is a clinic-based study with a large sample size, in which the same standardized methods were used in each European center. However, our study has several limitations. First, 15% of the participants were excluded for the present study due to missing information, which may have caused selection bias. We found that excluded patients had a higher risk profile regarding lifestyle and biological risk factors, which may imply that our associations for HbA 1c and all-cause mortality could have been even stronger if these patients were included. However, even though baseline HbA 1c was higher among excluded participants, it remains unknown whether most patients were excluded below or above the reference and therefore if the positive association for low and/or high HbA 1c with mortality would have been stronger if these patients were included. Second, it remains unsure whether the observed association of an increased mortality risk at low HbA 1c represents an actual, even more a causal, effect or is a result of other factors associated with low HbA 1c and increased mortality that we did not measure. Residual confounding through unmeasured variables could still explain the results; specifically, confounding by comorbid conditions is a main concern (25). Additional medical problems might interfere with treatment and thereby increase mortality risk (19). For example, patients with more complications may have been treated differently because of fear of hypoglycemia; however, these factors have not been measured in our study. Low HbA 1c ( 6.5%) was associated with an increased mortality risk in a type 2 diabetes population in those with high levels of comorbidity, but this association was not shown in those with lower levels of comorbidity (26). Other unmeasured factors in our study that may influence the association between low HbA 1c and increased mortality include lack of adherence to medical advice, psychiatric conditions, cognitive function, social or financial crisis (19), hypertension, hypokalemia, QT interval (27), anemia, or other states of increased erythrocyte turnover. It is not known which factors exactly cause low HbA 1c levels and the exact mechanisms through which pregnancy, chronic liver disease, renal disease, and HIV affect HbA 1c are not completely understood (28). Third, we did not have enough power to stratify the association between HbA 1c and allcause mortality by cause of death. Finally, these observations must be interpreted with caution because of the limited statistical power for our analyses. Limited number of all-cause mortality events can be explained by the young age of our study population. However, the EURODIAB PCS as an observational study provides a unique opportunity to study the relationship between HbA 1c and allcause mortality prospectively in young type 1 diabetes patients. Our results are relevant to the discussion about the risk of intensive glycemic control in diabetes patients. Our findings confirm the reported increased risk of death associated with low HbA 1c shown in observational studies among type 2 diabetes patients. The potential mechanism underlying these findings remains unclear. Lower HbA 1c targets are mechanically associated with higher risk of hypoglycemia, a common complication in intensive glucose treatment. Results from a meta-analysis of randomized clinical trials indicates a 30% increased risk of severe hypoglycemia by intensive glycemic control (5). In the DCCT, a 2- to 3-fold increase in severe hypoglycemia events in the intensive group was seen compared with the conventional group (22). Recent findings from the Diabetes and Aging Study suggest a U- shaped association between hypoglycemia and HbA 1c in type 2 diabetes patients, with the highest risk for severe hypoglycemia reported among patients achieving nearnormal glycemia (HbA1c 6%) as well as poorly controlled patients ( 9%) (29). In our study, severe hypoglycemia events more often occurred in the lower compared with the higher HbA 1c quintiles (P for trend 0001); however, additional adjustment for severe hypoglycemia events did not attenuate the association between low or high HbA 1c with all-cause mortality. This adjustment was, however, limited to self-reported hypoglycemia events. Also, Gruden et al (30) did not find an association between severe hypoglycemia and cardiovascular events or an increase in the markers of inflammation and endothelial injury in the EURODIAB PCS. Additional adjustment for nephropathy in our study indicated a stronger magnitude of the association for HbA 1c below the reference, whereas the association for HbA 1c greater than the reference was slightly attenuated. This could indicate that the positive association between low HbA 1c and all-cause mortality risk might be stronger in those with nephropathy. In conclusion, results from our study suggest an elevated mortality risk at both low and high HbA 1c in type 1 diabetes patients. The higher risk of all-cause mortality at lower HbA 1c levels found in our study might indicate that target HbA 1c levels below a certain threshold may not be appropriate for all type 1 diabetes patients. We recognize that HbA 1c levels below 5.6% in individuals with type 1 diabetes may be related to anemia, renal insufficiency, infection, or other factors not available in our database. Findings on this relationship from observational and intervention studies among type 1 and 2 diabetes patients are inconsistent, and our data will require confirmation by other groups. If confirmed, the potential mechanisms underlying this increased mortality risk among those with low HbA 1c will need further study.

8 doi: /jc jcem.endojournals.org 807 Acknowledgments See Supplemental Data for a complete list of the members of the EURODIAB Prospective Complications Study Group. Address all correspondence and requests for reprints to: Danielle A. J. M. Schoenaker, MSc, The University of Queensland, School of Population Health, Herston, Queensland 4006, Australia. d.schoenaker@uq.edu.au. The EURODIAB Prospective Complications Study was supported by grants from the Wellcome Trust, the European Community, and Diabetes UK. Disclosure Summary: The authors have nothing to disclose. References 1. Soedamah-Muthu SS, Fuller JH, Mulnier HE, Raleigh VS, Lawrenson RA, Colhoun HM. All-cause mortality rates in patients with type 1 diabetes mellitus compared with a non-diabetic population from the UK general practice research database, Diabetologia. 2006;49(4): American Diabetes Association. Standards of medical care in diabetes Diabetes Care. 2012;34(suppl 1):S11 S The Action to Control Cardiovascular Risk in Diabetes Study Group. Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med. 2008;358(24): Buehler AM, Cavalcanti AB, Berwanger O, et al. Effect of tight blood glucose control versus conventional control in patients with type 2 diabetes mellitus: a systematic review with meta-analysis of randomized controlled trials. Cardiovasc Ther. 2011;31(3): Hemmingsen B, Lund SS, Gluud C, et al. Intensive glycaemic control for patients with type 2 diabetes: systematic review with meta-analysis and trial sequential analysis of randomised clinical trials. BMJ. 2011;343:d Currie CJ, Peters JR, Tynan A, et al. Survival as a function of HbA1c in people with type 2 diabetes: a retrospective cohort study. Lancet. 2010;375(9713): Dahl-Jørgensen K, Larsen J, Hanssen K. Atherosclerosis in childhood and adolescent type 1 diabetes: early disease, early treatment? Diabetologia. 2005;48(8): Stephenson J, Fuller JH, Eurodiab Complications Study Group. Microvascular and acute complications in IDDM patients: the EURODIAB IDDM Complications Study. Diabetologia. 1994; 37(3): Chaturvedi N, Bandinelli S, Mangili R, Penno G, Rottiers RE, Fuller JH. Microalbuminuria in type 1 diabetes: rates, risk factors and glycemic threshold. Kidney Int. 2001;60(1): Hanas R, John G Consensus Statement on the Worldwide Standardization of the Hemoglobin A1C Measurement. Diabetes Care. 2010;33(8): Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem. 1972;18(6): Soedamah-Muthu SS, Chaturvedi N, Witte DR, Stevens LK, Porta M, Fuller JH. Relationship between risk factors and mortality in type 1 diabetic patients in Europe: The EURODIAB Prospective Complications Study (PCS). Diabetes Care. 2008;31(7): Harrell FE. Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis. New York: Springer-Varlag New York, Inc; Akaike H. A new look at the statistical model identification. IEEE Trans Automat Contr. 1974;19(6): Tielemans SM, Soedamah-Muthu SS, De Neve M, et al. Association of physical activity with all-cause mortality and incident and prevalent cardiovascular disease among patients with type 1 diabetes: the EURODIAB Prospective Complications Study. Diabetologia. 2012; 56(1): Heinzl H, Kaider A. Manual for the SAS-Macro RCS. SAS Institute Inc; Available at: software/statistische-software/rcs/. 17. Huang ES, Liu JY, Moffet HH, John PM, Karter AJ. Glycemic control, complications, and death in older diabetic patients. Diabetes Care. 2011;34(6): Riddle MC, Karl DM. Individualizing targets and tactics for highrisk patients with type 2 diabetes: practical lessons from ACCORD and other cardiovascular trials. Diabetes Care. 2012;35(10): Riddle MC, Ambrosius WT, Brillon DJ, et al. Epidemiologic relationships between A1C and all-cause mortality during a median 3.4-year follow-up of glycemic treatment in the ACCORD Trial. Diabetes Care. 2010;33(5): Zoungas S, Chalmers J, Ninomiya T, et al. Association of HbA1c levels with vascular complications and death in patients with type 2 diabetes: evidence of glycaemic thresholds. Diabetologia. 2012; 55(3): Duckworth W, Abraira C, Moritz T, et al. Glucose control and vascular complications in veterans with type 2 diabetes. N Engl J Med. 2009;360(2): The Diabetes Control and 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(14): Genuth S, Backlund J-YC, Bayless M, et al. Effects of prior intensive versus conventional therapy and history of glycemia on cardiac function in type 1 diabetes in the DCCT/EDIC. Diabetes. 2013;62(10): Nathan D, Cleary P, Backlund J, et al. Intensive diabetes treatment and cardiovascular disease in patients with type 1 diabetes. N Engl J Med. 2005;353(25): Rutter M. Low HbA1c and mortality: causation and confounding. Diabetologia. 2012;55(9): Greenfield S, Billimek J, Pellegrini F, et al. Comorbidity affects the relationship between glycemic control and cardiovascular outcomes in diabetes: a cohort study. Ann Int Med. 2009;151(12): Tsujimoto T, Yamamoto-Honda R, Kajio H, et al. Vital signs, QT prolongation, and newly diagnosed cardiovascular disease during severe hypoglycemia in type 1 and type 2 diabetic patients. Diabetes Care. 2014;37(1): Gallagher EJ, Le Roith D, Bloomgarden Z. Review of hemoglobin A1c in the management of diabetes. J Diabetes. 2009;1(1): Lipska KJ, Warton EM, Huang ES, et al. HbA1c and risk of severe hypoglycemia in type 2 diabetes: the Diabetes and Aging Study. Diabetes Care. 2013;36(11): Gruden G, Barutta F, Chaturvedi N, et al. Severe hypoglycemia and cardiovascular disease incidence in type 1 diabetes: the EU- RODIAB Prospective Complications Study. Diabetes Care. 2012; 35(7):