giorda c b stroke 2007 giorda 1154 60

Incidence and Risk Factors for Stroke in Type 2Diabetic Patients

The DAI Study

Carlo Bruno Giorda, MD; Angelo Avogaro, MD; Marina Maggini, PhD; Flavia Lombardo, PhD;Edoardo Mannucci, MD; Salvatore Turco, MD; Stefania Spila Alegiani, PhD; Roberto Raschetti, PhD;

Mario Velussi, MD; Ele Ferrannini, MD; The DAI Study Group

Background and Purpose—Type 2 diabetes mellitus is a strong predictor of cerebrovascular disease, yet few studies haveassessed the incidence of stroke and the role of other risk factors in unselected type 2 diabetes mellitus populations.

Methods—We prospectively followed-up 14 432 type 2 diabetes mellitus patients, aged 40 to 97 years, with and withouta history of cardiovascular disease at enrollment, and we estimated the incidence of stroke and the hazards ratios withrespect to clinical variables.

Results—During a 4-year follow-up, 296 incident stroke events were recorded. In persons with no history of cardiovasculardisease, the age-standardized incidence of stroke (per 1000 person-years) was 5.5 (95% confidence interval, 4.2 to 6.8)in men and 6.3 (95% confidence interval, 4.5 to 8.2) in women. In persons with a history of cardiovascular disease, itwas 13.7 (95% confidence interval, 7.5 to 19.8) in men and 10.8 (95% confidence interval, 7.3 to 14.4) in women. Thehazards ratios of stroke incidence varied according to age, sex, and history of cardiovascular disease. Among men withno history, HbA1c and smoking were predictors of stroke. Among patients with a history, the risk factors were, in men,therapy with insulin plus oral agents, treated high total cholesterol and low HDL cholesterol, whereas in womenmicrovascular complications were a risk factor. Previous stroke was a strong predictor of stroke in both sexes.

Conclusions—Age and previous stroke are the main predictors of stroke in diabetes. The combined role of Hba1c,microvascular complications, low HDL cholesterol, and treatment with insulin plus oral agents highlights theimportance of diabetic history and clinical background in the development of stroke. (Stroke. 2007;38:1154-1160.)

Key Words: diabetes mellitus � epidemiology � risk factors

Comparisons of epidemiological data on diabetic andnondiabetic subjects in the general population have

clearly demonstrated that type 2 diabetes mellitus (T2DM) isan independent risk factor for ischemic stroke.1–5 For exam-ple, in the Framingham Study, the incidence of nonhemor-rhagic stroke has been found to be from 2.5- to 3.5-timeshigher among diabetic than among nondiabetic subjects.6

Furthermore, it has recently been proposed that, for women,diabetes be considered a cardiovascular disease risk equiv-alent for fatal stroke, in that the incidence of a first strokeis similar to the incidence of relapse among nondiabeticwomen.7 Although it can be hypothesized that there is aninteraction of various atherogenic factors (eg, hypertensionand coagulative, hemorheological, and lipidic alterations),which characterize T2DM and which could make the risk ofstroke in these patients different from that in nondiabeticsubjects, little is known about the incidence of stroke amongdiabetic patients.

The clinical history of diabetes and its systemic involve-ment could play an important role in the development ofcerebrovascular complications. In the UKPDS study, strokein diabetic subjects was prospectively evaluated with predic-tive formulas,8 yet the study population consisted of patientswith newly diagnosed diabetes selected with criteria typicalto clinical trials. Apart from analyses on middle-aged co-horts,5 relatively few prospective studies on cerebrovascularcomplications among diabetes patients older than 65 years ofage have been published.

Compared with other European countries, Italy is unique interms of diabetes care, in that there exists a network of �700public diabetes care units that provide care for up to 80% ofknown patients,9 performing such services as diagnosticconfirmation, therapy, prevention, and the early diagnosis ofacute and chronic complications through close follow-up of thepatient by a team of specialists and the scheduling of regular

Received July 25, 2006; accepted November 22, 2006.From Metabolism and Diabetes Unit (C.B.G.), ASL 8, Regione Piemonte, Chieri, Italy; Division of Metabolic Diseases (A.A.), University of Padova,

Italy; National Institute of Health (M.M., F.L., S.S.A., R.R.), Rome, Italy; University of Florence and Azienda Ospedaliera Careggi (E.M.), Florence,Italy; Federico II University (S.T.), Naples, Italy; Casa di Cura Pineta del Carso (M.V.), Aurisina, Trieste, Italy; University of Pisa School of Medicine(E.F.), Pisa, Italy.

Correspondence to Dr Carlo Bruno Giorda, Metabolism and Diabetes Unit, ASL 8, Via De Maria, 1 10023 Chieri (TO), Italy. E-mail [email protected]© 2007 American Heart Association, Inc.

Stroke is available at http://www.strokeaha.org DOI: 10.1161/01.STR.0000260100.71665.2f

1154 by guest on February 22, 2015http://stroke.ahajournals.org/Downloaded from

http://stroke.ahajournals.org/

check-ups. Most patients are referred to these care units bytheir general practitioner and care is free of charge.

The objective of the present study was to describe theincidence and the risk factors for stroke in T2DM patients,including elderly persons, who receive standard care on aregular basis at diabetes care units in Italy.

MethodsThe present analysis was conducted as part of the DAI Study, amulticenter cohort study on the prevalence and incidence of cardio-vascular events among T2DM patients (myocardial infarction, ische-mic heart disease, coronary artery bypass, coronary angioplasty,cerebral thromboembolism, and peripheral amputations). In thepresent work, we report the results of the incidence study. A detaileddescription of the DAI study methods has been reported else-where.10,11 A brief description is provided here.

The DAI study was begun in September 1998 by the IstitutoSuperiore di Sanita (the Italian National Institute of Health) togetherwith the Italian Association of Clinical Diabetologists (AMD) andthe Diabetes and Informatics Study Group (DIAINF). In 1998, 201diabetes care units volunteered to participate. To encourage partici-pation, data collection was limited to routine data. These care unitsare distributed throughout Italy and represent �29% of the existingunits in Italy, comprising 19 570 patients. The DAI study populationconsisted of all patients with T2DM diagnosed after 39 years of age,according to the 1985 criteria of the World Health Organization,12

who visited the participating care units in the 2 enrollment periods(between September and December 1998, and from March to June1999). In each unit, patients were chosen to create a samplerepresentative of the diabetic population seen at the units. Asystematic sampling technique, including in the cohort every fourthpatient, was used.

A total of 201 care units voluntarily participated in the prevalencestudy. In the incidence study described herein, 157 of the 201 unitsparticipated, for a total of 14 432 diabetic patients (7214 men and7218 women). At baseline, 2788 patients (1602 men and 1186women) had a history of cardiovascular disease complications.

Data Collection and DefinitionsThe data used in this analysis was collected in 4 waves of follow-upin the period 2000 to 2003. All of the diabetes units participated atthe first follow-up, 120 at the second, 97 at the third, and 77 at thefourth. During the baseline and follow-up visits, a standard ques-tionnaire was used to collect the following information: personal dataand life habits; anthropometric data; drug therapy; laboratory data(specified here); clinical history and data on complications (retinop-athy, blindness, and foot ulcer); and data on cardiovascular compli-cations. For patients who did not show up for the scheduled visits,information, including death, was obtained through telephone inter-views with the patient, a relative, or the treating physician.

The following tests were performed at baseline and at eachfollow-up visit. Plasma glucose, HbA1c, and lipid profile weredetermined in fasting state. HbA1c was measured by each unitlaboratory, and not by a centralized one. For this reason, for eachpatient the relative difference between the observed value and theupper limit of the normal range of the care unit was calculated.Urinary albumin excretion was obtained in a timed overnightcollection: microalbuminuria was defined as urinary albumin excre-tion of 30 to 300 mg/L in at least 3 successive measurements, in theabsence of other factors capable of causing proteinuria. A patientwas defined as “hypertensive” if systolic blood pressure was�140 mm Hg or diastolic blood pressure was �90 mm Hg, or if thepatient was undergoing antihypertensive treatment. Retinopathy wasassessed by a dilated and comprehensive eye examination and by theacquisition of high-quality stereoscopic photographs assessed by anophthalmologist. Family history of cardiovascular disease was de-fined when the patient had a first-degree relative (parent, sibling,child) who had had a major cardiovascular event before 55 years ofage. Alcohol consumption was calculated in equivalent mL of wine,

taking into account the daily consumption of wine, hard alcohol (mL�4), liqueurs (mL �2), and beer (mL �0.5).

Cardiovascular EventsStroke was defined according to WHO criteria for confirmed andpossible stroke (ie, a clinical syndrome consisting of a rapiddeveloping neurological deficit and persisting for �24 hours, orleading to death, in the absence of other diseases that could explainthe symptoms). A hospital discharge record or a specialist visit wasrequired to certify the event.

Other study events were myocardial infarction, coronary heartdisease, coronary artery bypass, coronary angioplasty, and peripheralamputations.

The presence of coronary heart disease was assessed by docu-mented clinical data and by ECG. Patients were classified as havingcoronary heart disease if they had one of the following: hospitaliza-tion for either fatal or nonfatal acute myocardial infarction or anepisode of angina; positive 12-lead ECG for previous acute myocar-dial infarction or angina coded using the Minnesota coding system(criteria I, 1 to 3; IV, 1 to 3; V, 1 to 2; and VII, 1), history of coronaryartery bypass graft, and percutaneous transluminal coronary angio-plasty. Ninety-eight percent of patients had at least one ECG in the12 months preceding enrollment to certify the absence of infarction.

Statistical AnalysisBased on the presence of cardiovascular disease at baseline, 2cohorts were identified. Cohort A consisted of persons with nohistory of cardiovascular disease. Cohort B consisted of persons withprevious cardiovascular disease. The analyses were performed sep-arately for the 2 cohorts. The outcomes considered in the analysiswere a first fatal or nonfatal stroke. All of the patients werefollowed-up from the baseline visit to the first of the followingevents: stroke (or any cardiovascular event in cohort A, because inthis cohort the objective was to estimate the incidence of first strokenot preceded by any cardiovascular event), death, or the end ofparticipation of the unit, or end of the study. Patients for whom noinformation was available at any of the follow-up waves (n�2062)were considered as lost to follow-up and were included in the cohortuntil the last available visit or proxy information.

Data are expressed as mean (�standard deviation) or median (andinterquartile range) for continuous variables and as proportions forcategorical variables. The incidence density of stroke was standard-ized based on the age distribution of the 1998 Italian population.Kaplan–Meier curves were used to construct the cumulative 4-yearincidence of stroke. Univariate and multivariate Cox proportionalhazard models were used to examine the risk factors for stroke.Preliminary data analysis was performed by univariate Cox propor-tional hazard models of all covariates, measured at the baseline visit(Table 1). Analyses were conducted separately for men and women,adjusting by age at baseline. The interaction between cholesterollevel and anti-hyperlipidemic therapy and between blood pressureand anti-hypertensive therapy was tested. All covariates for men andwomen with P�0.1 (P�0.05 in cohort B, given the smaller samplesize) were entered in the final multivariate models, performedseparately for men and women (Table 2). The proportional hazardsassumption was checked with Schoenfeld residuals; the results werechecked using a stepwise technique. All variables were entered intothe model without transformation, although the quadratic term forcontinuous covariates was first considered and assessed with thelikelihood ratio test. All analyses were performed using the Stata 8.0statistical package.

ResultsTable 1 summarizes the baseline characteristics for men andwomen, including a description of the cardiovascular diseasein cohort B. The study population consisted of a highproportion of elderly patients with visceral adiposity. Themost common cardiovascular risk factor was systolic hyper-tension, which was found in almost all patients. A high

Giorda et al The DAI Study 1155

by guest on February 22, 2015http://stroke.ahajournals.org/Downloaded from


percentage of patients were undergoing treatment, especiallythose with a previous cardiovascular event (cohort B). Thesecharacteristics match those reported in a recently publishedsurvey of 140 000 diabetes patients in Italy.14

Compared with men, women, especially those with cardio-vascular disease, had worse lipid profiles (ie, total cholester-ol, high-density lipoprotein, triglycerides) and were lesslikely to be treated with lipid-lowering drugs. As expected,

TABLE 1. Characteristics of 14 432 Persons With T2DM by Sex and Baseline Presence ofCVD: the DAI Study, Italy

CVD at Baseline

No (11 644) Yes (2788)

Characteristic Men Women Men Women

Mean�SD

Age at visit, y 64�9 66�9 67�8 70�8

Body mass index, kg�m�2 28�4 29�5 28�4 29�5

Waist circumference, cm 100�11 96�13 101�11 97�13

Cholesterol, mg/dL 207�47 220�43 209�45 221�45

HDL cholesterol, mg/dL 48�13 52�14 46�12 51�14

Median (IQ range)

Duration of diabetes, y 7 (3–13) 8 (3–14) 9 (4–15) 11 (5–18)

Systolic BP, mm Hg 140 (130–160) 150 (135–160) 140 (130–160) 150 (139–160)

Diastolic BP, mm Hg 80 (80–90) 80 (80–90) 80 (80–90) 80 (80–90)

Plasma glucose, mg/dL 153 (129–186) 158 (130–194) 158 (130–191) 165 (135–203)

Triglycerides, mg/dL 127 (91–180) 134 (99–183) 142 (102–198) 146 (108–201)

%

Hypertension† 79.5 86.3 89.6 93.7

HbA1c‡ 72.1 77.9 76.4 82.5

Alcohol intake 58.1 20.7 54.1 18.8

Smoking

Current 21.1 7.4 14.7 5.1

Former 30.9 5.1 45.9 7.3

Microvascular complication 35.0 33.6 46.5 45.4

Relatives with CHD 25.5 30.1 38.1 39.4

Glycemic control

Diet alone 18.6 14.0 12.8 9.5

Oral agents 67.4 67.1 65.2 60.1

Insulin�oral agents 5.7 9.1 8.0 14.2

Insulin alone 8.3 9.8 14.0 16.1

Antihypertensive treatment 44.2 59.4 71.0 80.6

Lipid-lowering treatment 10.1 17.3 25.1 26.1

Geographic area

Northern Italy 57.1 49.5 59.1 51.4

Central Italy 10.5 11.0 11.0 8.3

Southern Italy and the islands 32.4 39.6 29.9 40.3

Cardiovascular complication

AMI � � � � � � 45.4 22.4

Stroke � � � � � � 10.9 12.1

CHD � � � � � � 32.3 57.7

Other* � � � � � � 11.4 7.8

*Coronary revascularizations, limb amputations, and all combined events.†Defined as systolic BP �140 or diastolic BP �90, or antihypertensive treatment.‡Percentage of patients with values higher than the upper limit of the normal range.AMI indicates acute myocardial infarction; BP, blood pressure; CHD, coronary heart disease; CVD, cardiovascular

disease; HDL, high-density lipoprotein; IQ, interquartile; SD, standard deviation; T2DM, type 2 diabetes mellitus.

1156 Stroke April 2007



the average metabolic control was fair, confirming otherreports of good average HbA1c levels in Italian T2DMpatients receiving care at diabetes care units.14 The specificcardiovascular events were those typical to diabetes patients;

specifically, the most common events were acute myocardialinfarction and coronary heart disease (with acute myocardialinfarction more frequent among men and coronary heartdisease more frequent among women), followed by stroke,which was more frequent among women. The percentage ofpatients who smoked was �30% less than that for Italy’sgeneral population.15

During the follow-up period, a total of 296 strokes weredocumented (188 in cohort A and 108 in cohort B). The crudecumulative risk for the 2 cohorts is shown in Figure 1. Incohort A, the age-standardized incidence rate per 1000person-years was 5.5 (95% confidence interval, 4.2 to 6.8) inmen and 6.3 (95% confidence interval, 4.5 to 8.2) in women;in cohort B, it was 13.7 (95% confidence interval, 7.5 to 19.8)in men and 10.8 (95% confidence interval, 7.3 to 14.4) inwomen.

The multivariate analysis model for cohort A is shown inTable 3. In men, age, HbA1c, and smoking were independentpredictors of first stroke. In women, independent predictorswere age and microvascular complications.

The multivariate analysis model for cohort B is shown inTable 4. In men, the predictors of stroke were age, insulintherapy in combination with oral agents, previous stroke, andtreated high cholesterol levels. Because in this cohort it hadalready been demonstrated that individuals with high choles-terol had a greater likelihood of receiving aggressive lipid-lowering therapy as standard care,15 we determined thepercentage of statin-treated patients (categorized into tertilesby total cholesterol) in the years after the baseline evaluation:the incidence of treatment was 11.9% in the lower tertile(�189 mg/dL), 17.4% in the middle tertile (190 to 224mg/dL); and 33.2% in the upper tertile (�225 mg/dL;P�0.01). In women with previous cardiovascular disease, the

TABLE 2. Variables Considered for Inclusion in the Models

Continuous Variables

Age Age in years at enrolment (10 yearsincrements)

Duration Duration of diabetes in years (3 yearsincrements)

Waist circumference Waist circumference in centimeters(5-cm increments)

Cholesterol Total cholesterol at enrollment (40-mg/dlincrements)

HDL HDL at enrollment (5-mg/dl increments)

HbA1c Relative difference between the observedvalue and the upper limit of the normal

range of the unit (20% increments)

Categorical variables

Smoking Nonsmoker; current smoker; formersmoker at enrollment

Microvascular complication No; Yes (retinopathy, blindness,microalbuminuria, proteinuria, foot ulcer)

Lipid-lowering therapy No; Yes

Antihypertensive therapy No; Yes

Antihyperglycemic therapy Diet; oral agents; insulin�oral agents;insulin alone

Previous stroke atenrollment

No; Yes

Geographic area Northern Italy; Central Italy; SouthernItaly and Island

Kaplan–Meier curves for stroke in patients withtype 2 diabetes mellitus, with and without pre-vious cardiovascular disease (CVD), by sex.The DAI Study, Italy.




predictors of stroke were age, low high-density lipoproteincholesterol, microvascular complications, and previous stroke(hazard ratio, 3.1; 95% confidence interval, 1.7 to 5.5).

DiscussionThe age-adjusted incidence of stroke found for this largecohort of diabetic subjects confirms that stroke is an impor-tant complication among diabetic subjects. Compared withthe nondiabetic population, the incidence is from 2- to3-times greater.16 A particularly high incidence (approxi-mately double) was found for patients with previous cardio-vascular events, compared with patients with no history ofcardiovascular disease, which, to the best of our knowledge,is the first such finding.

In our study, age was found to have played an importantrole in the occurrence of both a first stroke and relapse.Although this is not a new finding for the general popula-tion,17 this is the first time that it has been shown in a clearmanner in a cohort of diabetic subjects not selected forage18,19 Among men with no history of cardiovascular dis-ease, blood glucose control at enrollment was predictive ofstroke. In other studies,4,5 the association was reported forhigh cut-off levels, exceeding 10% of Hba1c; in our study,each 20% increment above the normal glycohemoglobin

levels, corresponding to �1% point of Hba1c, was signifi-cant. This finding is consistent with a previous report,20 andwith regard to prevention it stresses the importance of strictmetabolic control in the daily routine of these patients.

Microvascular complications, retinal lesions, microalbu-minuria, and proteinuria, which were significantly associatedwith stroke among women, have been described as factorsthat are predictive of cardiovascular and cerebrovascularmorbidity and mortality among diabetic subjects.21 Given thatthese complications are correlated with a history of poorblood glucose control, they may also be an indication of theeffect of diabetes over time on the occurrence of stroke. Theharmful affects could be explained not only by endothelialdamage but also by factors such as hypertension, which isoften triggered by kidney damage.

The significant increase in the hazard ratio for therapy withinsulin plus oral agents in men with previous cardiovasculardisease should be carefully considered. It is not rare that inobservational studies, because of what can be defined as“indication bias,” drug therapy,22,23 particularly insulin ther-apy,24,25 has been found to be associated with a worsecardiovascular prognosis. However, in clinical trials, inten-sive treatment has been observed to have an anti-atherogeniceffect.26 In our study, it is more plausible that therapy with

TABLE 3. HRs and 95% CIs for Potential Predictors of Stroke Among 11 644 Patients With T2DM Without Previous CVD by Sex (CoxProportional Hazards Analysis), the DAI Study, Italy

Risk Factor Adjusted Only for Age Adjusted for Multiple Variables

Men Women Men Women

HR (95% CI) P HR (95% CI) P HR (95% CI) P HR (95% CI) P

Age at visit (10-y increments) 1.85 (1.47–2.34) 0.00 2.24 (1.73–2.90) 0.00 1.94 (1.46–2.57) 0.00 2.25 (1.69–3.00) 0.00

Duration (3-y increments) 1.09 (1.01–1.18) 0.03 1.04 (0.96–1.12) 0.35 1.03 (0.94–1.12) 0.54 1.01 (0.93–1.10) 0.81

Waist circumference (5-cm increments) 1.10 (1.00–1.20) 0.04 1.04 (0.96–1.13) 0.31 1.09 (0.99–1.19) 0.09 1.04 (0.96–1.13) 0.32

HbA1c (20% increments) 1.27 (1.11–1.46) 0.00 1.07 (0.92–1.26) 0.39 1.22 (1.04–1.43) 0.01 1.02 (0.85–1.21) 0.84

Smoker

No 1 1 1 1

Yes 2.26 (1.37–3.73) 0.00 1.21 (0.52–2.79) 0.66 2.29 (1.36–3.87) 0.00 1.18 (0.47–2.94) 0.73

Former 1.37 (0.85–2.21) 0.20 0.64 (0.20–2.04) 0.46 1.17 (0.70–1.96) 0.55 0.71 (0.22–2.26) 0.56

Microvascular complication

No 1 1 1 1

Yes 1.51 (1.01–2.26) 0.05 1.64 (1.09–2.47) 0.02 1.15 (0.74–1.80) 0.53 1.65 (1.06–2.56) 0.03

Antihypertensive therapy

No 1 1 1 1

Yes 1.48 (0.98–2.23) 0.06 0.91 (0.60–1.39) 0.66 1.41 (0.90–2.21) 0.13 0.79 (0.51–1.22) 0.28

Antihyperglycemic therapy

Diet 1 1 1 1

Oral agents 1.76 (0.88–3.54) 0.11 2.28 (0.92–5.68) 0.08 1.19 (0.58–2.46) 0.63 1.86 (0.74–4.71) 0.19

Insulin�oral agents 4.02 (1.69–9.55) 0.00 3.05 (1.07–8.69) 0.04 2.10 (0.80–5.51) 0.13 2.00 (0.65–6.20) 0.23

Insulin alone 2.33 (0.94–5.76) 0.07 2.65 (0.92–7.65) 0.07 1.61 (0.61–4.21) 0.34 2.00 (0.64–6.19) 0.23

Geographic area

Northern Italy 1 1 1 1

Central Italy 0.47 (0.21–1.10) 0.08 1.01 (0.49–2.07) 0.98 0.52 (0.22–1.21) 0.13 1.03 (0.50–2.14) 0.93

Southern Italy and the Islands 0.63 (0.40–1.00) 0.05 1.28 (0.83–1.96) 0.27 0.62 (0.37–1.01) 0.06 1.12 (0.71–1.77) 0.63

CI indicates confidence interval; HR, hazards ratio.




insulin plus oral agents was a proxy of severity and of theclinical course of the disease. In a recent national survey,patients undergoing this treatment showed the worst meta-bolic control.13

The significant hazard ratios for Hba1c in men, for mi-croangiopathic vascular complications in women, and fortherapy with insulin plus oral agents are difficult to explain,yet these combined factors as a whole highlight the role ofdiabetic history and clinical background in the developmentof stroke. Surprisingly, hypertension was not found to havebeen associated with stroke. Again, because the study popu-lation was not selected (with a very high proportion ofhypertensive elderly persons), age and the weight of diabetesitself may have attenuated the emergence of other risk factorsthat were found for younger cohorts.5,28 Furthermore, itcannot be excluded that the high percentage of patientsundergoing hypertensive treatment attenuated the role ofhypertension.

In patients with a history of cardiovascular disease, thehistory of a previous stroke was a strong risk factor forrelapse, confirming that arteriopathy, once established, tendsto worsen and that relapses occur independently of otherfactors. To our knowledge this is the first time this relation-ship has been reported for diabetic subjects. With regard tolipids, the role of reverse cholesterol transport was high-lighted by the consistently favorable effect of high-densitylipoprotein cholesterol in women. The protective role of totalcholesterol was unexpected, yet the interpretation of thisfinding warrants some caution. In the literature, the evidencedemonstrating that cholesterol is a risk factor for cerebral

vasculopathy is controversial.28 A possible and reasonableexplanation for this lies in the fact that persons with highcholesterol received aggressive treatment in the 4 years offollow-up (the use of statins in these individuals was 3-foldthat in persons with normal cholesterol). This explanation isconsistent with a previous report that showed that, in thiscohort, high cholesterol levels prompted aggressive treatmentas standard care11 By contrast, the deleterious effects of highcholesterol levels, however treated, can be interpreted as theinability of therapy to counter severe dyslipidemia in thisspan of time. It should be noted that in subjects with previouscardiovascular disease, men, but not women, living in thesouth enjoyed a substantially lower hazard for stroke ascompared with patients elsewhere in the country. This pro-tective effect, which resisted all adjustments, may be the firsttrace of a lifestyle effect to ever have been detected in patientswith diabetes. Adherence to a Mediterranean diet (which ishigher in southern Italy) has been convincingly associatedwith a reduction in cardiovascular risk.29,30 Postmenopausalwomen may be inherently more resistant to diet-related riskreduction, as suggested by a recent study.31

A limitation of this study is that the lack of centralizedlaboratory data undoubtedly reduced the reliability of theinformation. Given that the observation period was relativelybrief, there may not have been time for other potentiallysignificant factors to emerge. Furthermore, although diabetesis predominantly ischemic,27 the possibility that a quota ofevents could be hemorrhagic cannot be ruled out. Finally,studying cerebral vasculopathy only by evaluating its most

TABLE 4. HRs and 95% CIs for Potential Predictors of Stroke Among 2788 Patients With T2DM With Previous CVD by Sex (CoxProportional Hazards Analysis), the DAI Study, Italy

Risk Factor

Adjusted for Age Adjusted for Multiple Variables

Men Women Men Women

HR (95% CI) P HR (95% CI) P HR (95% CI) P HR (95% CI) P

Age at visit (10-y increments) 2.03 (1.42–2.91) 0.00 2.42 (1.67–3.51) 0.00 1.86 (1.28–2.70) 0.00 2.54 (1.67–3.88) 0.00

HDL (5-mg/dL increments) 0.97 (0.86–1.09) 0.56 0.89 (0.79–0.99) 0.03 0.99 (0.89–1.11) 0.89 0.88 (0.79–0.98) 0.02

Microvascular complication

No 1 1 1 1

Yes 1.06 (0.62–1.81) 0.85 2.11 (1.22–3.66) 0.01 0.74 (0.41–1.33) 0.32 1.89 (1.04–3.43) 0.04

Antihyperglycemic therapy

Diet 1 1 1 1

Oral agents 2.05 (0.63–6.69) 0.24 4.13 (0.56–30.29) 0.16 1.94 (0.59–6.40) 0.28 3.37 (0.46–24.82) 0.23

Insulin�oral agents 4.77 (1.33–17.13) 0.02 7.54 (0.99–57.65) 0.05 5.01 (1.35–18.50) 0.02 4.77 (0.61–37.32) 0.14

Insulin alone 2.32 (0.60–8.97) 0.22 5.64 (0.73–43.41) 0.10 1.79 (0.42–7.71) 0.43 3.95 (0.50–31.14) 0.19

Previous stroke at enrollment

No 1 1 1 1

Yes 2.78 (1.58–4.88) 0.00 2.74 (1.58–4.74) 0.00 2.83 (1.58–5.07) 0.00 3.12 (1.75–5.57) 0.00

Total cholesterol (40-mg/dLincrements)

0.69 (0.49–0.97) 0.03 1.08 (0.80–1.46) 0.60 0.68 (0.47–0.98) 0.04 1.09 (0.78–1.53) 0.60

Lipid-lowering therapy

No 1 1 1 1

Yes 1.02 (0.48–2.19) 0.96 1.72 (0.87–3.40) 0.12 1.07 (0.50–2.29) 0.87 1.80 (0.90–3.61) 0.10

Cholesterol�lipid-lowering therapy 2.03 (1.15–3.58) 0.02 0.81 (0.48–1.35) 0.42 1.92 (1.09–3.39) 0.02 0.75 (0.44–1.30) 0.31




serious manifestation does not allow earlier stages of arteri-opathy to be taken into consideration.

In addition to the large study population, the strengths ofthis study lie in the fact that it provides an accurate descrip-tion of the daily reality of diabetes care. In particular, unlikeother studies, we evaluated the incidence of stroke amongadvanced-age diabetic subjects, among whom the incidenceof cerebral vasculopathy is greater. In Italy, the percentage ofdiabetic subjects who seek care at the diabetes care units isvery high, reaching up to 80% of known cases;10 thus, ourresults most likely can be generalized to the entire T2DMpopulation in Italy. What emerges from our results in prac-tical terms is that a diabetes specialist with �2000 patientswill probably have to face �15 new cases of stroke each year,consisting mostly of older patients and those with a morecomplex clinical history of diabetes.

DisclosuresNone.

References1. Stegmayr B, Asplud K. Diabetes as a risk factor for stroke: a population

perspective. Diabetologia. 1995;38:1061–1068.2. Riddle MC, Hart J. Hyperglycemia recognised and unrecognised risk

factor stroke and transient ischemic attacks. Stroke. 1982;13:356–359.3. Cox NH, Lorians JW. The prognostic value of blood glucose and glyco-

sylated estimation in persons with stroke. Postgrad Med J. 1986;62:7–10.4. Burchfiel CM, Curb JD, Rodriguez BL, Abbott RD, Chiu D, Yano K.

Glucose intolerance and 22-year stroke incidence. The Honolulu HeartProgram. Stroke. 1994;25:951–957.

5. Letho S, Ronnemaa T, Pyorala K, Laakso M. Predictors of stroke inmiddle-aged patients with non-insulin-dependent diabetes. Stroke. 1996;27:63–68.

6. Kannel WB, McGee DL. Diabetes and cardiovascular disease: The Fra-mingham Study. JAMA. 1979;241:2035–2038.

7. Ho JE, Paultre F, Mosca L. Is diabetes mellitus a cardiovascular diseaserisk equivalent for fatal stroke in women? Data from the women’s poolingproject. Stroke. 2003;34:2812–2816.

8. Kothari V, Stevens RJ, Adler AI, Stratton IM, Manley SE, Neil HA,Holman RR. UKPDS 60: risk of stroke in type 2 diabetes estimated byUK Prospective Diabetes Study risk engine. Stroke. 2002;33:1776–1781.

9. http//www.epicentro.iss.it/quadri/pag/regioni/piemonte.htmpag.7.Accessed November 17, 2006.

10. The DAI Study Group. The DAI prospective study on macrovascularcomplications in patients with type 2 diabetes. Characteristics of the studypopulation. Available at: www.iss.it. Ann Ist Super Sanit. 2001;37:289–296.

11. The DAI Study Group. Prevalence of coronary heart disease in a cohortof type 2 diabetic patients in Italy: the DAI study. Diabetic Med. 2004;21:738–745.

12. Harris MI, Hadden WC, Knowler WC, Bennett PH. International criteriafor the diagnosis of diabetes and impaired glucose tolerance. DiabetesCare. 1985;8:562–567.

13. Italian Association of Clinical Diabetologist (AMD). Annals AMD 2006.Quality Indicators in Diabetes in Italy. Copyright AMD Rome; 2006.

14. The Italian Cardiovascular Epidemiological Observatory. Italian Heart J.2004;5(Suppl 3):49S–92S.

15. Giorda C, Maggini M, Alegiani SS, Turco S, Raschetti R. Secondaryprevention of coronary artery disease in high-risk diabetic patients. NutrMetab Cardiovasc Dis. 2003;13:238–243.

16. The Stroke Prevention and Educational Awareness Diffusion (SPREAD)Collaboration. The Italian Guidelines for stroke prevention. Neurol Sci.2000;21:5–12.

17. Lichtman JH, Krumholz HM, Wang Y, Radford MJ, Brass LM. Riskand predictors of stroke after myocardial infarction among the elderly:results from Cooperative Cardiovascular Project. Circulation. 2002;105:1082–1087.

18. Croxson SC, Burden AC, Bodington M, Botha JL. The prevalence ofdiabetes in elderly people. Diabet Med. 1991;8:28–31.

19. International Diabetes Federation. Diabetes Atlas 2003. Available fromhttp://www.idf.org/e-atlas/home/. Accessed July 17, 2005.

20. Selvin E, Coresh J, Shahar E, Zhang L, Steffes M, Sharrett AR. Gly-caemia (haemoglobin A1c) and incident ischaemic stroke: the Athero-sclerosis Risk in Communities (ARIC) Study. Lancet Neurol. 2005;4:821–826.

21. Miettinen H, Haffner SM, Lehto S, Ronnemaa T, Pyorala K, Laakso.Proteinuria predicts stroke and other atherosclerotic vascular diseaseevents in nondiabetic and non-insulin-dependent diabetic subjects. Stroke.1996;27:2033–2039.

22. Rodriguez MD, Llorca J. Bias. J Epidemiol Commun Health. 2004;58:635–641.

23. Mamdani M, Sykora K, Li P, Notmand TS-L, Steiner LD, Austin CP,Rochon A. Paula, Anderson MG. Reader’s guide to critical appraisal ofcohort studies: 2. Assessing potential for confounding. BMJ. 2005;330:960–962.

24. Murcia AM, Hennekens CH, Lamas GA, Jimenez-Navarro M, RouleauJL, Flaker GC, Goldman S, Skali H, Braunwald E, Pfeffer MA. Impact ofdiabetes on mortality in patients with myocardial infarction and leftventricular dysfunction. Arch Intern Med. 2004;164:2273–2279.

25. Muis MJ, Bots ML, Bilo HJ, Hoogma RP, Hoekstra JB, Grobbee DE,Stolk RP. High cumulative insulin exposure: a risk factor of atheroscle-rosis in type 1 diabetes? Atherosclerosis. 2005;181:185–192.

26. Nathan DM, Lachin J, Cleary P, Orchard T, Brillon DJ, Backlund JY,O’Leary DH, Genuth S; Diabetes Control and Complications Trial: Ep-idemiology of Diabetes Interventions and Complications Research Group.Intensive diabetes treatment and cardiovascular disease in patients withtype 1 diabetes. N Engl J Med. 2005;353:2643–2653.

27. Bells DS. Stroke in the diabetic patient. Diabetes Care. 1994;17:213–219.28. Prospective Studies Collaboration Cholesterol, diastolic blood pressure

and stroke: 13000 strokes in 450000 people in 45 prospective cohorts.Lancet 1995;346:1647–1653.

29. Trichopoulou A, Costacou T, Bamia C, Trichopoulos D. Adherence to aMediterranean diet and survival in a Greek population. N Engl J Med.2003;348:2599–2608.

30. Estruch R, Martinez-Gonzalez MA, Corella D, Salas-Salvado J, Ruiz-Gutierrez V, Covas MI; PREDIMED Study Investigators. Effects of aMediterranean-style diet on cardiovascular risk factors: a randomizedtrial. Ann Intern Med. 2006;145:1–11.

31. Howard BV, Van Horn L, Hsia J, Manson JE, Stefanick ML, Wassertheil-Smoller S, et al. Low-fat dietary pattern and risk of cardiovasculardisease: the Women’s Health Initiative Randomized Controlled DietaryModification Trial. JAMA. 2006;295:655–666.




The DAI Study GroupSalvatore Turco, Stefania Spila Alegiani, Roberto Raschetti, Mario Velussi, Ele Ferrannini and Carlo Bruno Giorda, Angelo Avogaro, Marina Maggini, Flavia Lombardo, Edoardo Mannucci,

Incidence and Risk Factors for Stroke in Type 2 Diabetic Patients: The DAI Study

Print ISSN: 0039-2499. Online ISSN: 1524-4628 Copyright © 2007 American Heart Association, Inc. All rights reserved.

is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231Stroke doi: 10.1161/01.STR.0000260100.71665.2f

2007;38:1154-1160; originally published online March 1, 2007;Stroke.

http://stroke.ahajournals.org/content/38/4/1154World Wide Web at:

The online version of this article, along with updated information and services, is located on the

http://stroke.ahajournals.org//subscriptions/

is online at: Stroke Information about subscribing to Subscriptions:

http://www.lww.com/reprints Information about reprints can be found online at: Reprints:

document. Permissions and Rights Question and Answer process is available in the

Request Permissions in the middle column of the Web page under Services. Further information about thisOnce the online version of the published article for which permission is being requested is located, click

can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial Office.Strokein Requests for permissions to reproduce figures, tables, or portions of articles originally publishedPermissions:


http://stroke.ahajournals.org/content/38/4/1154

http://www.ahajournals.org/site/rights/

http://www.lww.com/reprints

http://stroke.ahajournals.org//subscriptions/


giorda c b stroke 2007 giorda 1154 60

Documents