ORIGINAL ARTICLE

Glycemic Variability Is Higher in Type 1Diabetes Patients with MicrovascularComplications Irrespective of Glycemic Control

Jan Soupal, MD, Jan Skrha, Jr., MD, Martin Fajmon, MSc, Eva Horova, MD, PhD,

Milos Mraz, MD, PhD, Jan Skrha, MD, DrSc, MBA, and Martin Prazny, MD, PhD

Abstract

Background: Increased glycemic variability (GV) may be associated with diabetes complications. Our study assessed therelationship between microvascular complications (MVCs) and GV calculated from continuous glucose monitoring (CGM)data in type 1 diabetes patients.Subjects and Methods: Thirty-two patients with type 1 diabetes (16 with and 16 without MVC) participated in this cross-sectional study. Vibration perception threshold (VPT), microalbuminuria, and fundoscopy were used to detect MVC. CGMdata were recorded for 2 weeks and analyzed using proprietary software. Total SD (SDT), coefficient of variation (CV), andmean amplitude of glycemic excursions (MAGE) were compared.Results: Patients with any MVC had significantly higher GV, calculated from CGM, than patients without MVC (SDT,4.1 – 0.6 vs. 3.4 – 0.8 mmol/L [P = 0.010]; CV, 0.43 – 0.06 vs. 0.38 – 0.08 [P = 0.032]; MAGE, 6.9 – 1.2 vs. 5.9 – 1.2 mmol/L[P = 0.014]) but comparable glycated hemoglobin (HbA1c) (70 – 9 vs. 69 – 10 mmol/mol [8.6 – 0.8% vs. 8.5 – 0.9%], differencenot significant). No significant difference in GV was found between the two groups when using only self-monitored bloodglucose (SMBG) data. A positive association was found between VPT and SDT in all patients (r = 0.51, P = 0.0026).Conclusions: Patients with type 1 diabetes and any MVC had significantly higher GV calculated from CGM, but not fromSMBG, than patients with comparable glycemic control but without complications. This supports the hypothesis that in-creased GV might be associated with MVC in type 1 diabetes and that HbA1c may not describe diabetes control completely.

Introduction

Diabetes mellitus is characterized by hyperglycemiaand increased risk of diabetes complications in patients

with both type 1 and type 2 diabetes.1,2 It is proven that tightglycemic control reduces the risk of development and pro-gression of both microvascular (MVCs) and macrovascularcomplications.3,4 Glycated hemoglobin (HbA1c) is used asthe gold standard for the estimation of glycemic control.5,6

However, glycemic disorders have two components: chronicsustained hyperglycemia and acute glycemic fluctuations.7

Moreover, patients with similar HbA1c and mean plasmaglucose levels can have markedly different daily glucose ex-cursions.8 These short-term glucose fluctuations have also beensuggested as contributing to the risk of long-term complica-tions of diabetes.9 It has been shown that hyperglycemia acti-vates oxidative stress and plays a major role in overproductionof superperoxide by the mitochondrial electron chain,10,11

which in turn triggers glucose-mediated tissue damage.

The role of glycemic variability (GV) in similar patho-physiological pathways is the subject of debate. At themolecular level, there is experimental evidence that hyper-glycemic variability is associated with persistent expression ofthe nuclear factor jB p65 gene in aortic endothelial cells bothin vitro and in mice without diabetes. This expression persistsfor at least 6 days of subsequent normoglycemia and leads toan increase in levels of nuclear factor jB–dependent proteins,such as monocyte chemoattractant protein 1 and vascularcell adhesion molecule 1, which are implicated in diabetes-associated vascular injury.12 There are other studies showinga link between GV and oxidative stress13,14; however theseresults have not been consistently reproduced in humans.15,16

It is important that, from the clinical point of view, thequestion of whether there is an association between GV andcomplications of diabetes, and, if so, what is its clinical sig-nificance, remains unanswered. In the subanalysis of theCoronary Artery Calcification in Type 1 diabetes (CACTI)study, it was reported that the presence of coronary artery

3rd Department of Internal Medicine, First Faculty of Medicine, Charles University, Prague, Czech Republic.

DIABETES TECHNOLOGY & THERAPEUTICSVolume 16, Number 4, 2014ª Mary Ann Liebert, Inc.DOI: 10.1089/dia.2013.0205

198

calcium as a possible marker of subclinical atherosclerosis isassociated with GV in men with type 1 diabetes.17

In the prospective observational study performed by Bragdet al.,18 the authors found that GV was a predictor of theprevalence of peripheral neuropathy and showed a borderlinesignificance in the prediction of its incidence. On the otherhand, two retrospective analyses of the Diabetes Control andComplications Trial (DCCT) datasets concluded that GVhas no, or only a minor, contribution to MVCs in type 1 dia-betes.19,20 GV has been calculated from the standard self-monitored blood glucose (SMBG) data in many studies.However, even if the full seven-point glucose profile wasobtained, some potentially important peaks or nadirs werealways missed because they occurred between two mea-surements. In contrast, continuous glucose monitoring (CGM)gives the concentration of glucose in subcutaneous tissueapproximately every 5 min and therefore provides muchmore glucose data. It should now be preferred for the as-sessment of GV.8,21

The aim of our study was to contribute to the discussionabout GV and complications of diabetes. We have calculatedGV using CGM and analyzed its association with MVCs inpatients with type 1 diabetes.

Research Design and Methods

Study population

We studied outpatients with type 1 diabetes who wereregistered at the 3rd Department of Internal Medicine, 1st

Faculty of Medicine, Charles University, Prague, Czech Re-public. The study was approved by the Ethics Committee ofthe 1st Faculty of Medicine, Charles University. Thirty-twopatients with type 1 diabetes participated in this cross-sectional study. Sixteen patients did not have any MVCs, and16 had one or more complications. In the group of patients withMVC, four patients had one type of MVC, nine patients had acombination of two types of complications, and three subjectshad all three. All patients were on multiple daily injections(n = 17) or continuous subcutaneous insulin infusion (n = 15).

Other characteristics of the patients are shown in Table 1. Onlypatients with type 1 diabetes mellitus with a duration of 10–30years were allowed to participate. Diagnosis of type 1 diabeteswas confirmed by low C-peptide levels and the patient’s clin-ical appearance. These patients had to have been stable whilereceiving therapy for at least 3 months. Subjects with macro-vascular complications were not allowed to enter the study.Patients with acute diabetes complications, such as ketoaci-dosis, within the past 6 months and/or severe noncomplianceand/or any concomitant therapy influencing glucose metabo-lism were not allowed to participate either. The patients’therapy was maintained as usual during CGM.

All patients were screened for the presence of MVCs. Thepresence of diabetic retinopathy was assessed by indirectophthalmoscopy in artificial mydriasis by an experiencedophthalmologist. Both nonproliferative and proliferativeforms of diabetic retinopathy were included.

Microalbuminuria was measured as the urinary albumin-to-creatinine ratio in a random spot collection. Albumin-to-creatinine ratio values of ‡ 2.5 g/mol for males and‡ 3.5 g/mol for females were considered positive for micro-albuminuria.22

The vibration perception threshold (VPT) was assessedwith a biothesiometer (Bio-Medical Instrument Co, Newbury,OH) and used to evaluate the presence and severity of theimpairment of distal sensory nerves. We proceeded in accor-dance with a methodology approved by the American Dia-betes Association and the American Association of ClinicalEndocrinologists.23 For this assessment, the patient was lyingsupine and had been previously made familiar with themethod. The probe of the instrument was placed vertically overthe dorsal hallux, and the vibration amplitude (given in volts)was increased until the patient could feel the vibration. A meanof three readings was used to derive the value for each foot. AVPT of > 15 V was regarded as impaired VPT, VPT < 15 V isregarded as normal, whereas a value over 25 V is classified asabnormal and means the presence of sensory neuropathy. Wedecided to use a cutoff level of 15 V simply to distinguishbetween normal and impaired vibration perception—this

Table 1. Characteristics of Patients With and Without Microvascular Complications

Characteristic Group with MVC (n = 16) Group without MVC (n = 16) P value

Male sex [n (%)] 8 (50) 7 (44) 0.81Age (years) 44 – 10 39 – 13 0.27Duration of diabetes (years) 21 – 5 18 – 6 0.14Mean glucose CGM (mmol/L) 9.2 – 1.2 9.2 – 1.6 0.94HbA1c [% (mmol/mol)] 8.6 – 0.8 (70 – 9) 8.5 – 0.9 (69 – 10) 0.68Fasting plasma glucose (mmol/L) 7.5 – 3.6 8.9 – 3.3 0.26BMI (kg/m2) 25.9 – 2.4 25.5 – 2.7 0.64CGM readings per patient (n) 3,786 – 287 3,672 – 198 0.57VPT (V) 17.3 – 7.9 9.1 – 2.4 0.001a

ACR (g/mol) 21.5 – 32.0 0.98 – 0.74 0.018a

CSII/MDI (n) 8/8 7/9 0.99iPro2/Dexcom SEVEN (n) 7/9 7/9 1.00ACCU-CHEK/OneTouch (n) 12/4 13/3 0.99Total daily dose of insulin (IU) 52 – 17 51 – 20 0.93

Data are mean – SD values, unless stated otherwise.aP values < 0.05 are statistically significant.ACR, albumin-to-creatinine ratio; BMI, body mass index; CGM, continuous glucose monitoring; CSII, continuous subcutaneous insulin

infusion; HbA1c, glycated hemoglobin; MDI, multiple daily injections; MVC, microvascular complication; VPT, vibration perceptionthreshold.

GV AND MICROVASCULAR COMPLICATIONS 199

allowed us not to miss any incipient neuropathy and clearlyidentify patients with normal vibration perception.

CGM, GV, and biochemical methods

Patients were monitored for 12–14 consecutive days witheither an iPro2� (Medtronic, Northridge, CA) or Dexcom�

(San Diego, CA) SEVEN� system. Two sensors were succes-sively applied to every participant, obtaining a total of ap-proximately 3,700 measurements per patient. For a particularpatient, we always used the same type of CGM devicethroughout the study. CGM was used in the masked mode inorder to avoid any possible patient intervention that couldaffect the CGM results (the masked mode of the Dexcom de-vice has to be activated through Dexcom’s Technical Sup-port). Sensors were inserted into abdominal subcutaneoustissue and calibrated according to the standard recommen-dations of the manufacturers. The patients were instructed tomeasure their blood glucose with a personal blood glucosemeter (ACCU-CHEK� [Roche, Basel, Switzerland] or One-Touch� [LifeScan, Milpitas, CA]) at least four times per day.

Patients were invited to a total of three visits. CGM wasinserted during the first and the second visit and downloadedto the computer for the analysis during the second and thethird visit, respectively. Subsequently, for more detailedanalysis, the data were exported from the original softwareinto our proprietary software, which then calculated differentparameters of GV. There are large numbers of measures ofGV. However, no consensus exists at this moment about thestandard method.24,25 We have used a selection of previouslypublished and frequently used formulas: total standard de-viation (SDT), coefficient of variation (CV),8 and mean am-plitude of glucose excursions (MAGE).26

Blood samples were taken at the third visit between 7:00and 8:00 a.m. after an overnight fast. The fasting plasmaglucose level was measured by a routine enzymatic method.HbA1c was analyzed by a high-performance liquid chroma-tography method on a Variant II� analyzer (Bio-Rad, Her-cules, CA). The normal reference range of HbA1c in ourlaboratory is 20–42 mmol/mol (4–6%).

Statistical analysis

Statistical evaluation was performed by Statistics for Win-dows version 10 software (SPSS, Inc., Chicago, IL). Basic de-scriptive statistics were calculated for the relevant parameters.Analysis of variance, t test, or Wilcoxon’s, Mann–Whitney,and Kolmogorov–Smirnov tests were used to compare data.Pearson’s and Spearman’s correlations were used for analysisof associations between parameters. Multiple linear regres-sion analysis was carried out to disclose the statistically in-dependent associations of GV with other variables that couldpossibly influence the incidence of complications. Data areexpressed as mean – SD values. A value of P < 0.05 was con-sidered statistically significant.

Results

GV was significantly higher in the group of patients withany MVC (Fig. 1) compared with patients without MVC (SDT,4.1 – 0.6 vs. 3.4 – 0.8 mmol/L [P = 0.010]; MAGE, 6.9 – 1.2 vs.5.9 – 1.2 mmol/L [P = 0.014]; CV, 0.43 – 0.06 vs. 0.38 – 0.08[P = 0.032]). Patients with microalbuminuria (n = 7) had higherGV compared with subjects without any MVC (SDT, 4.3 – 0.5vs. 3.6 – 0.8 mmol/L [P = 0.044]; CV, 0.46 – 0.07 vs. 0.39 – 0.06[P = 0.023]; MAGE, 7.5 – 0.9 vs. 6.1 – 1.2 mmol/L [P = 0.008]).Relative to the patients without any MVC, the subgroup ofpatients with diabetic retinopathy (n = 14) had significantlyelevated SDT (4.1 – 0.7 vs. 3.5 – 0.8 mmol/L [P = 0.031]), and asimilar relationship was observed in the subgroup of patientswith impaired VPT (n = 8; SDT, 4.2 – 0.7 vs. 3.6 – 0.8 mmol/L[P = 0.041]). Moreover, a positive association was found be-tween VPT and SDT in all patients (r = 0.51, P = 0.0026) (Fig. 2).

Then we performed a multivariate linear regression anal-ysis of the association between GV (measured by SDT andMAGE) and complications adjusted for confounding factors(HbA1c, mean glucose, duration of diabetes, sex, and age).We observed significant effects only for parameters of GV,thus excluding any significant influence by the confounders(Table 2).

Significantly, when we calculated SD and MAGE from theSMBG data used for calibration of CGM devices (four values

FIG. 1. Glycemic variability and any microvascular complication: (A) total SD (SDT), (B) mean amplitude of glucoseexcursions (MAGE), and (C) coefficient of variation (CV). CGM, continuous glucose monitoring. Open columns, no micro-vascular complication; solid columns, any microvascular complication.

200 SOUPAL ET AL.

per day), the differences in GV between the patient groupswith and without complications disappeared. Using theseSMBG data reduces the differences below the level of statis-tical significance (SD, 3.7 – 0.9 vs. 4.3 – 0.9 mmol/L [P = 0.070,difference not significant]).

Finally, we did not observe any difference in the treatmentof patients with and without MVC (total daily dose of insulin,52 – 17 vs. 51 – 20 units [difference not significant]; daily basalinsulin dose, 26 – 8 vs. 25 – 9 units [difference not significant];total daily bolus count, 4.3 – 1.2 vs. 3.9 – 1.2 [difference notsignificant]). Patients with and without MVC did not differ inglycemic control (Table 1). No significant gender differenceswere observed in this study.

Discussion

Although there is increasing evidence of an associationbetween GV, especially in the postprandial period, and therisk of developing both MVCs and macrovascular complica-tions in type 2 diabetes,27 this has not been found consistentlyin patients with type 1 diabetes. This is quite interesting be-cause patients with type 1 diabetes are usually consideredmore homogeneous than patients with type 2 diabetes. Webelieve that the homogeneity of the patient group is veryimportant. We intentionally decided to perform the study inpatients with type 1 diabetes. These patients usually lack theother components of metabolic disorder so typical for patientswith type 2 diabetes. This should increase the chance ofproving an association between GV and complications in type1 diabetes. The homogeneity of patients with type 1 diabetesin our study was further increased by the exclusion of patientswith macroangiopathic complications.

In contrast to our findings, analyses of the DCCT/Epidemiology of Diabetes Interventions and Complications(EDIC) found no significant relationship between GV and thedevelopment or progression of retinopathy, nephropathy,

FIG. 2. Positive association between vibration perception threshold (VPT) and total SD (SDT) for all patients (r = 0.51,P = 0.0026).

Table 2. Multivariate Analysis of the Association

Between Parameters of Glycemic Variability

and Microvascular Complications Adjusted

for Glycated Hemoglobin, Mean Glucose,

Duration of Diabetes, Sex, and Age

Oddsratio

95% confidenceinterval P value

SDT

HbA1c 2.41 0.51–16.73 0.275MG (CGM) 0.42 0.09–1.18 0.106Duration of diabetes (years) 1.12 0.93–1.39 0.234Age (years) 1.01 0.92–1.08 0.969Sex 0.38 0.03–3.02 0.375SDT 7.5 1.83–52.08 0.004a

MAGEHbA1c 2.11 0.48–11.47 0.326MG (CGM) 0.53 0.15–1.35 0.201Duration of diabetes (years) 1.07 0.89–1.3 0.482Age (years) 1.02 0.95–1.11 0.536Sex 0.48 0.04–3.65 0.487MAGE 2.83 1.3–8.17 0.007a

aP values < 0.05 are statistically significant.HbA1c, glycated hemoglobin; MAGE, mean amplitude of glycemic

excursions; MG (CGM), mean glucose calculated from continuousglucose monitoring; SDT, total SD.

GV AND MICROVASCULAR COMPLICATIONS 201

and diabetic neuropathy.28,29 Nevertheless, regarding GV,there are limitations to the DCCT/EDIC.28,29 First, the DCCTwas not primarily designed to assess GV. Second, and mostimportantly, for the estimation of GV the authors used five- toseven-point blood glucose profiles taken quarterly. However,even when five- to seven-point glucose profiles are obtaineddaily, certain potentially important peaks or nadirs will bemissed. The results of our study suggest that four-point glu-cose profiles are not sufficient to match GV to complications,whereas CGM data allow us to reveal the association clearly.In that context, although several studies have been publishedabout the clinical impact of GV, relatively little attention hasbeen paid to the datasets from which GV has been calculated.The range of datasets varies from five to seven-point glucosemeasurements taken at 3-month intervals used in theDCCT19,28,29 to 70 measurements over 4 weeks reported byBragd et al.18 and CGM at 5-min intervals describedby Monnier and colleagues.14 There is a study published byBaghurst et al.21 dealing with the minimum frequency ofglucose measurements from which GV can be consistentlyassessed. These data showed that GV estimated by SDT be-comes unreliable if observations are more than 2–4 h apartand that MAGE becomes unreliable if glucose measurementsare more than 1 h apart. CGM therefore seems to be preferableto SMBG as a source of data for the calculation of GV becauseit can capture all its complexity.8,21

Last but not least, many different parameters are used todescribe GV. Some parameters have a very simple formula(SD and CV); some are more complex (MAGE). We pro-grammed the proprietary software to read and merge CGMdata and then calculated chosen parameters of GV. Our re-sults suggest that the more complex parameters provide noadditional information over SDT. However, it is crucial to notethe basic fact that parameters of GV tell us different thingsabout different aspects of the complex phenomenon of bloodglucose behavior. Based on our study we cannot, of course,conclude that SDT and CV are the best parameters of GV forall situations. However, thanks to their simplicity they areeasy to calculate routinely as a component of diabetes care.

There are some limitations of our study that should benoted. First is the relatively small number of patients in ourcohort. Despite this, we strictly followed inclusion and ex-clusion criteria, and so the patient sample is homogeneous.Another limiting factor is the fact that the study is cross-sectional, not prospective. It can inform us only about asso-ciation between GV and MVC, not causation. However, theassessment of causality was not the aim of our study. Thecurrent study should be interpreted as an exploratory conceptfor future studies.

In our analysis, patients with MVCs had significantlyhigher GV calculated from CGM data than the patientswithout complications, although they did not differ in gly-cemic control. This result did not hold for GV calculated fromSMBG data. The association between GV and complicationswas independent of sex, age, diabetes duration, and HbA1c inour patients.

The results of our study support the hypothesis that higherGV may be associated with MVC in patients with type 1 di-abetes and that HbA1c alone may not describe diabetes controland the risk of complications completely. If this hypothesis isproved correct, it will have a significant impact on the clinicalpractice and care of diabetes patients. Our study, however,

cannot answer the emerging questions as to whether causalityexists between GV and complications in diabetes and whetherthe reduction of GV decreases the risk of complications. Thesequestions can only be answered by a prospective observa-tional study on a large group of diabetes patients using CGM.We believe that the results of our study may be used as a‘‘bridge’’ to such a future study and facilitate its design.

Acknowledgments

This study was initiated, designed, and performed by theinvestigators and supported by Research Project P25/LF1/2of Charles University in Prague. The authors thank JanaPacnerova for technical assistance.

Author Disclosure Statement

No competing financial interests exist.

References

1. The effect of intensive treatment of diabetes on the devel-opment and progression of long-term complications in in-sulin dependent diabetes mellitus. The Diabetes Control andComplications Trial (DCCT) Research Group. N Engl J Med1993;329:977–986.

2. Intensive blood-glucose control with sulphonylureas or in-sulin compared with conventional treatment and risk ofcomplications in patients with type 2 diabetes (UKPDS 33).UK Prospective Diabetes Study (UKPDS) Group. Lancet1998;352:837–853.

3. Stratton IM, Adler AI, Neil HAW, Matthews DR, Manley SE,Cull CA, Hadden D, Turner RC, Holman RR: Association ofglycaemia with macrovascular and microvascular compli-cations of type 2 diabetes (UKPDS 35): prospective obser-vational study. BMJ 2000;321:405–412.

4. Sustained effect of intensive treatment of type 1 diabetesmellitus on development and progression of diabetic ne-phropathy: the Epidemiology of Diabetes Interventions andComplications (EDIC) study. DCCT/EDIC Research Group.JAMA 2003;290:2159–2167.

5. Hirsch IB, Brownlee M: Should minimal blood glucose var-iability become the gold standard of glycemic control?J Diabetes Complications 2005;19:178–181.

6. Hanefeld M: Postprandial hyperglycaemia: noxious effectson the vessel wall. Int J Clin Pract Suppl 2002;129:45–50.

7. Monnier L, Colette C: Glycemic variability: should we andcan we prevent it? Diabetes Care 2008;31(Suppl 2):S150–S154.

8. Siegelaar SE, Holleman F, Hoekstra JBL, DeVries JH: Glu-cose variability; does it matter? Endocr Rev 2010;31:171–182.

9. Brownlee M, Hirsch IB: Glycemic variability: a hemoglobinA1c-independent risk factor for diabetic complications.JAMA 2006;295:1707–1708.

10. Brownlee M: Biochemistry and molecular cell biology ofdiabetic complications. Nature 2001;414:813–821.

11. Brownlee M: The pathobiology of diabetic complications: aunifying mechanism. Diabetes 2005;54:1615–1625.

12. El-Osta A, Brasacchio D, Yao D, Pocai A, Jones PL, RoederRG, Cooper ME, Brownlee M: Transient high glucose causespersistent epigenetic changes and altered gene expressionduring subsequent normoglycemia. J Exp Med 2008;205:2409–2417.

13. Ceriello A, Esposito K, Piconi L, Ihnat M, Thorpe J, Testa R,Bonfigli AR, Giugliano D: Glucose ‘‘peak’’ and glucose

202 SOUPAL ET AL.

‘‘spike’’: impact on endothelial function and oxidative stress.Diabetes Res Clin Pract 2008;82:262–267.

14. Monnier L, Mas E, Ginet C, Michel F, Villon L, Cristol JP,Colette C: Activation of oxidative stress by acute glucosefluctuations compared with sustained chronic hyperglyce-mia in patients with type 2 diabetes. JAMA 2006;295:1681–1687.

15. Siegelaar SE, Barwari T, Kulik W, Hoekstra JB, DeVries JH:No relevant relationship between glucose variability andoxidative stress in well-regulated type 2 diabetes patients.J Diabetes Sci Technol 2011;5:86–92.

16. Wentholt IM, Kulik W, Michels RP, Hoekstra JB, DeVries JH:Glucose fluctuations and activation of oxidative stress intype 1 diabetes patients. Diabetologia 2008;51:183–190.

17. Snell-Bergeon JK, Roman R, Rodbard D, Garg S, Maahs DM,Schauer IE, Bergman BC, Kinney GL, Rewers M: Glycaemicvariability and coronary artery calcium: The Coronary Ar-tery Calcification in Type 1 Diabetes Study. Diabet Med2010;27:1436–1442.

18. Bragd J, Adamson U, Backlund LB, Lins PE, Moberg E,Oskarsson P: Can glycaemic variability, as calculated fromblood glucose self-monitoring, predict the development ofcomplications in type 1 diabetes over a decade? DiabetesMetab 2008;34:612–616.

19. Kilpatrick ES, Rigby AS, Atkin SL: Effect of glucose vari-ability on the long-term risk of microvascular complicationsin type 1 diabetes. Diabetes Care 2009;32:1901–1903.

20. Lachin JM, Genuth S, Nathan DM, Zinman B, Rutledge BN;DCCT/EDIC Research Group: Effect of glycemic exposureon the risk of microvascular complications in the DiabetesControl and Complications Trial—revisited. Diabetes 2008;57:995–1001.

21. Baghurst PA, Rodbard D, Cameron FJ: The minimum fre-quency of glucose measurements from which glycemicvariation can be consistently assessed. J Diabetes Sci Technol2010;4:1382–1385.

22. Bakker AJ: Detection of microalbuminuria. Receiveroperating characteristic curve analysis favors albumin-to-creatinine ratio over albumin concentration. Diabetes Care1999;22:307–313.

23. Boulton AJ, Armstrong DG, Albert SF, Frykberg RG, Hell-man R, Kirkman MS, Lavery LA, Lemaster JW, Mills JL Sr,

Mueller MJ, Sheehan P, Wukich DK; AmericanDiabetesAssociation; American Association of Clinical Endo-crinologists: Comprehensive foot examination and riskassessment: a report of the Task Force of the Foot CareInterest Group of the American Diabetes Association, withendorsement by the American Association of Clinical En-docrinologists. Diabetes Care 2008;31:1679–1685.

24. Rodbard D: Interpretation of continuous glucose monitoringdata: glycemic variability and quality of glycemic control.Diabetes Technol Ther 2009;11(Suppl 1):S-55–S-67.

25. Greven WL, Beulens JW, Biesma DH, Faiz S, de Valk HW:Glycemic variability in inadequately controlled type 1 dia-betes and type 2 diabetes on intensive insulin therapy: across-sectional, observational study. Diabetes Technol Ther2010;12:695–699.

26. Service FJ, Molnar GD, Rosevear JW, Ackerman E, Gate-wood LC, Taylor WF: Mean amplitude of glycemicexcursions, a measure of diabetic instability. Diabetes 1970;19:644–655.

27. Nalysnyk L, Hernandez-Medina M, Krishnarajah G: Gly-caemic variability and complications in patients with dia-betes mellitus: evidence from a systematic review of theliterature. Diabetes Obes Metab 2010;12:288–298.

28. Kilpatrick ES, Rigby AS, Atkin SL: The effect of glucosevariability on the risk of microvascular complications in type1 diabetes. Diabetes Care 2006;29:1486–1490.

29. Siegelaar SE, Kilpatrick ES, Rigby AS, Atkin SL, Hoekstra JB,Devries JH: Glucose variability does not contribute to thedevelopment of peripheral and autonomic neuropathy intype 1 diabetes: data from the DCCT. Diabetologia 2009;52:2229–2232.

Address correspondence to:Jan Soupal, MD

3rd Department of Internal MedicineFirst Faculty of Medicine

Charles UniversityU Nemocnice 1

128 08 Prague 2, Czech Republic

E-mail: jan.soupal@seznam.cz

GV AND MICROVASCULAR COMPLICATIONS 203