E m e r g i n g T r e a t m e n t s a n d T e c h n o l o g i e s
Development and Validation of Stroke
Risk Equation for Hong Kong Chinese
Patients With Type 2 Diabetes
The Hong Kong Diabetes Registry
Stroke is among the most common
dence of stroke and related mortality than HRISTOPHER W.K. LAM, PHD
Health Organization MONICA project(2). Diabetic patients have a two- to five-fold increased risk of stroke, in part due to OBJECTIVE — We sought to develop stroke risk equations for Chinese patients with type 2
interactions between multiple risk factors (3). The Framingham Study (4) and U.K.
Prospective Diabetes Study (UKPDS) (5) RESEARCH DESIGN AND METHODS — A total of 7,209 Hong Kong Chinese type 2
diabetic patients without a history of stroke at baseline were analyzed. The data were randomlyand evenly divided into the training subsample and the test subsample. In the training sub- sample, stepwise Cox models were used to develop the risk equation. Validation of the U.K.
Prospective Diabetes Study (UKPDS) stroke risk engine and the current stroke equation was performed in the test dataset. The life-table method was used to check calibration, and the area under the receiver operating characteristic curve (aROC) was used to check discrimination.
cruited from a workforce (6), there is cur-rently no risk equation applicable to RESULTS — A total of 372 patients developed incident stroke during a median of 5.37 years
(interquartile range 2.88 –7.78) of follow-up. Age, A1C, spot urine albumin-to-creatinine ratio spite this number being projected to 42.3 (ACR), and history of coronary heart disease (CHD) were independent predictors. The perfor- million by 2030 (7). In this study, we val- mance of the UKPDS stroke engine was suboptimal in our cohort. The newly developed risk idate and develop stroke risk equations to equation defined by these four predictors had adequate performance in the test subsample. Thepredicted stroke-free probability by the current equation was within the 95% CI of the observed predict first stroke in Chinese type 2 dia- probability. The aROC was 0.77 for predicting stroke within 5 years. The risk score was com- puted as follows: 0.0634 ϫ age (years) ϩ 0.0897 ϫ A1C ϩ 0.5314 ϫ log (ACR) (mg/mmol) ϩ 0.5636 ϫ history of CHD (1 if yes). The 5-year stroke probability can be calculated by: 1 Ϫ0.9707EXP (Risk Score Ϫ 4.5674).
CONCLUSIONS — Although the risk equation performed reasonably well in Chinese type
METHODS — Since 1995, all newly
2 diabetic patients, external validation is required in other populations.
referred diabetic patients to the Prince ofWales Hospital in Hong Kong underwent Diabetes Care 30:65–70, 2007
comprehensive assessments of complica-tions and risk factors based on the Euro-pean DiabCare protocol (7a). Patientswith hospital admissions within 6 – 8 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● From the 1Department of Medicine and Therapeutics, The Chinese University of Hong Kong, Hong Kong, China; the 2Li Ka Shing Institute of Health Sciences, The Chinese University of Hong Kong, Hong Kong,China; the 3Department of Chemical Pathology, The Chinese University of Hong Kong, Hong Kong, China; lyzed patients had a history of stroke. Pa- the 4Oxford Centre for Diabetes, Endocrinology, and Metabolism, University of Oxford, Oxford, U.K.; the 5Nuffield Department of Clinical Medicine, University of Oxford, Oxford, U.K.; 6Worldwide Outcomes Research, Merck & Co., Inc., Whitehouse Station, New Jersey; and the 7Hospital Authority Head Office, Address correspondence and reprint requests to Professor Juliana C.N. Chan, Department of Medicine and Therapeutics, The Chinese University of Hong Kong, The Prince of Wales Hospital, Shatin, NT, Hong ment of insulin within 1 year of diagnosis Kong SAR, China. E-mail:
Received for publication 19 June 2006 and accepted in revised form 16 October 2006.
Abbreviations: ACR, albumin-to-creatinine ratio; ARB, angiotensin II receptor blocker; aROC, area
under the receiver operating characteristic curve; CHD, coronary heart disease; eGFR, estimated glomerularfiltration rate; SBP, systolic blood pressure; UKPDS, U.K. Prospective Diabetes Study.
tee, and written informed consent was ob- A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion 2007 by the American Diabetes Association.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. DIABETES CARE, VOLUME 30, NUMBER 1, JANUARY 2007 Risk equation for stroke in type 2 diabetes
glucose, A1C, lipid profile (total choles- terol, HDL cholesterol, and triglycerides analysis. Patients with transient cerebral nal and liver functions. A sterile, random, ischemia (code 435) were not included.
the albumin-to-creatinine ratio (ACR).
Details of assessment methods, laboratory count observation time and censoring.
RESULTS — Between 1995 and 2005,
the training data (n ϭ 3,652) and test data (n ϭ 3,559). Cox proportional hazard re- type 1 diabetes (n ϭ 332), uncertain type gression with the stepwise algorithm (P Ͻ 1 diabetes status (n ϭ 5), non-Chinese or ated Modification of Diet in Renal Disease 0.05 for entry and stay) was used to select unknown nationality (n ϭ 49), and past predictors at baseline for incident stroke.
history of stroke (n ϭ 325) were ex- (0.742 if female), where SCR is serum cre- diabetic patients were included in the fi- smoking status, history of coronary heart teristics of patients with and without in- of foot pulses, confirmed by an ankle-to- cident stroke. In this cohort, the median (46 – 67) and median disease duration 5 through dilated pupils were performed.
total-to-HDL cholesterol ratio, A1C, sys- changes due to diabetes, laser scars, or a 5.16% of patients (n ϭ 372) developed globin. In developing the current predict- stroke was 9.66 (95% CI 8.69 –10.64) per (stroke: 5.18% or 190) and 3,541 (stroke: history was defined as having a history of 5.14% or 182) patients, respectively.
During the follow-up period, 705 patients or stress test, myocardial infarction, an- gina coronary artery bypass graft surgery, stroke (among the 372 stroke events).
1 Ϫ S(j)EXP (Risk Score Ϫ Mean of the Risk Score), where X1, X2, . . . , Xp are baseline predic- tors and ␤1, ␤2, . . . , ␤p are, respectively, the estimated coefficients of baseline pre- stroke risk engine are listed in Table 2. Of dictors 1 to p, and S(j) is the survival func- tion over j years when the risk score takes sex, current smoking status, and total-to- inhibitors (statins) (10) and blockers of cant. In the stepwise algorithm, log10 ACR (11) reduced the risk of stroke by Ͼ20%.
lected as significant predictors of stroke, (codes 430 – 434 and 436) or deaths from was not selected by the new model. In the stroke (codes 430 – 434 and 436 – 438).
(ARB), other antihypertensive drugs, oral nese patients with type 2 diabetes, i.e., the while all others were classified as ischemic was performed in the test subsample. Cal- stroke. All diagnoses of stroke were con- ibration was checked using the same life- CI of the observed curve (Fig. 1). The pre- firmed by the attending physician on dis- dicted stroke probabilities (or stroke-free probability) by the new stroke risk equa- under the receiver operating characteris- guidelines of the hospital authority. Only tic curve (aROC) was utilized to indicate DIABETES CARE, VOLUME 30, NUMBER 1, JANUARY 2007 Yang and Associates
Table 1—Baseline clinical and biochemical characteristics of 7,209 Chinese type 2 diabetic
patients with no history of stroke divided according to the development of first stroke during a
in this cohort was 0.588 (95% CI 0.549 – median follow-up of 5.37 years
0.626). The unadjusted aROC for the newrisk equation was 0.749 (0.716 – 0.782).
Taking into consideration follow-up time RESEARCH DESIGN AND METHODS, the risk equa- tion for predicting the first stroke event can mates of models 1 and 2 listed in Table 2.
score ϭ 0.0634 ϫ age (years) ϩ 0.0897 ϫ ϩ 0.5636 ϫ history of CHD (1, if yes; 0, otherwise); the 5-year stroke probability ϭ 1 Ϫ 0.9707EXP (Risk Score Ϫ 4.5674). At the cut- off point of Ն5.3099 for the risk score, cor- over 5 years of follow-up, the sensitivity was 65.7% and specificity 74.9%. Sensitivities and specificities for other cutoff points are The predictive ability of the current risk equation for hemorrhagic stroke and isch- emic stroke was further estimated by using Use of oral antidiabetes drugs at baseline points in the test subsample. Using the risk Use of antihypertensive drugs at baseline rhagic stroke (n ϭ 32) and ischemic stroke (n ϭ 150) were 0.770 and 0.785, respectively, for 5 years of follow-up.
CONCLUSIONS — In this prospec-
Data are percent or median (interquartile range) unless otherwise indicated. eGFR was from the glomerular filtration rate from the modified Modification of Diet in Renal Disease formula. *Derived from ␹2 test.
†Derived from Wilcoxon two-sample test. ‡Data for 1,271 patients (stroke events ϭ 82) who enrolled before1 December 1996 were not available. ACEI, ACE inhibitor; DBP, diastolic blood pressure; LLD, lipid- in Chinese type 2 diabetic patients due to different risk profiles. Thus, there is aneed to develop a Chinese-relevant riskequation to predict incident stroke using Table 2—Parameter estimates of the risk equation for Hong Kong Chinese type 2 diabetic patients and reestimated hazard ratios of the
predictors used by the UKPDS stroke engine in the training subsample

Current smoking status (1 if yes; 0 otherwise) DIABETES CARE, VOLUME 30, NUMBER 1, JANUARY 2007 Risk equation for stroke in type 2 diabetes
study (6) consisting of a small cohort ofChinese male steel workers (n ϭ 4,400),the aROC was 0.78 and 0.82 for ischemicand hemorrhage stroke, respectively.
However, the 95% CIs were not reported.
Besides, the event rate was relatively lowin this community cohort with only 49ischemic strokes and 33 hemorrhagicstrokes in the training subsample and 21ischemic strokes and 15 hemorrhagicstrokes in the validation subsample (6).
This is compared with 372 strokes (190 inthe training subsample and 182 in the testsubsample) in our cohort.
tures with that developed from theUKPDS (5) and a Chinese cohort of steelworkers (6), all of which used commonly Figure 1—The predicted stroke-free probabilities by the UKPDS stroke engine and the Hong Kong (HK) Chinese stroke risk score, as well as the 95% CIs of the observed stroke-free probability over 8 years of observation in the test dataset. of seven predictors (age, sex, smoking sta-tus, total-to-HDL cholesterol ratio, SBP, for periodic assessments. These risk pre- ventricular hypertrophy. The inclusion of 22 additional nontraditional risk factors and markers of subclinical atherosclerotic status). Our equation consists of four pre- diseases, such as BMI, waist-to-hip ratio, compared with 0.61– 0.74 for other risk including smoking status, sex, and total- that have only minor contributions to the to-HDL cholesterol ratio, were not signif- icant in the analysis, whereas the effect of overfitting the risk equation (15). It may (14), a basic model of stroke equation was also increase the probability of inaccuracy diabetic population has a very high prev- Table 3—Sensitivity, specificity, and positive predictive values in the test data at selected risk scores and their corresponding 5-year stroke

*Calculated using the risk score and 5-year stroke probability equation. †The suggested cutoff point.
DIABETES CARE, VOLUME 30, NUMBER 1, JANUARY 2007 Yang and Associates
1996), use of aspirin was not included as 4. D’Agostino RB, Wolf PA, Belanger AJ, Kan- a predictor for stroke (P ϭ 0.6890). Sec- nel WB: Stroke risk profile: adjustment for ond, although atrial fibrillation is a risk antihypertensive medication: the Framing- factor for stroke (24), it has been included ham Study. Stroke 25:40 – 43, 1994 in some (25) but not all stroke risk equa- 5. Kothari V, Stevens RJ, Adler AI, Stratton UKPDS 60: risk of stroke in type 2 diabe- factors for cardiovascular and renal dis- et al. (25) also reported that inclusion of Diabetes Study risk engine. Stroke 33: cluding atrial fibrillation did not signifi- 6. Zhang XF, Attia J, D’Este C, Yu XH, Wu cantly improve the predicting accuracy of XG: A risk score predicted coronary heart tive effects of these risk factors. Although stroke risk equations. Besides, the preva- disease and stroke in a Chinese cohort.
lence of atrial fibrillation in the diabetic J Clin Epidemiol 58:951–958, 2005 7. Wild S, Roglic G, Green A, Sicree R, King has been reported (19,20), renal function H: Global prevalence of diabetes: esti-mates for the year 2000 and projections for 2030. Diabetes Care 27:1047–1053, study, only baseline measurements of risk factors were used to develop and validate 7a.Piwernetz K, Home PD, Snorgaard O, Ant- in the stroke equation. This may be due to buminuria as an expression of endothelial tee: Monitoring the targets of the St. Vin- equation to predict stroke in Chinese type 2 diabetic patients. It is noteworthy that the DiabCare initiative. Diabet Med10:371–377, 1993 general populations (4,14) but not in the may not adequately predict the event risk type of diabetes. Diabetes Care 8:114 – (5). The use of antihypertensive drugs, in- incidences of the event of interest. How- ever, the ranking may still be appropriate (26). Thus, further validation is required univariate Cox models in our patients but blood cell count is associated with macro- not selected in the stepwise algorithm.
Chinese patients with type 2 diabetes.
study, the effects of drugs on stroke were Acknowledgments — This study was par-
10. Colhoun HM, Betteridge DJ, Durrington ables. In this respect, the efficacy of ARB tially supported by a Merck Sharp & Dohme Foundation for Research and Development in Menys V, Fuller JH: Primary prevention of firmed in randomized clinical trials (16).
Diabetes, established under the auspices of the Chinese University of Hong Kong. R.J.S. was stratification and prediction of clinical of Health, for her assistance in data retrieval trolled trial. Lancet 364:685– 696, 2004 events (21). This periodic assessment may Julius S, Beevers G, de Faire U, Fyhrquist especially in clinical settings where close References
Pedersen O, Lindholm LH, Nieminen MS,Omvik P, Oparil S, Wedel H: Cardiovas- 1. Murray CJ, Lopez AD: Mortality by cause for eight regions of the world: Global Bur- den of Disease Study. Lancet 349:1269 – specialist physicians in carrying out these randomised trial against atenolol. Lancet 12. Hanley JA, McNeil BJ: The meaning and and determinants in cardiovascular diseases use of the area under a receiver operating in China. Part i. Morbidity and mortality characteristic (ROC) curve. Radiology patients did not have lipid testing over the monitoring. Circulation 103:462–468, 3. Almdal T, Scharling H, Jensen JS, Vester- gaard H: The independent effect of type 2 curve for long-term risk prediction. Stat First, use of aspirin might increase the risk 14. Chambless LE, Heiss G, Shahar E, Earp MJ, Toole J: Prediction of ischemic stroke with 20 years of follow-up. Arch Intern munities Study. Am J Epidemiol 160:259 – DIABETES CARE, VOLUME 30, NUMBER 1, JANUARY 2007 Risk equation for stroke in type 2 diabetes
patients: Hong Kong Diabetes Registry.
15. Harrell F: Regression Modelling Strategies Diabetologia 49:2299 –2308, 2006 with Applications to Linear Models, Logistic 19. Mann JF, Gerstein HC, Pogue J, Bosch J, Regression, and Survival Analysis. New Yusuf S: Renal insufficiency as a predictor radiologic features. Neurology 54:2298 – 24. Woo J, Lau EM: Risk factors predisposing trial. Ann Intern Med 134:629 – 636, 2001 tion: a longitudinal study. Neuroepidemi- 25. Moons KG, Bots ML, Salonen JT, Elwood Prevalence (MAP) Study. Diabetologia 48: PC, Freire de Concalves A, Nikitin Y, Siv- filtration rate in addition to albuminuria is enius J, Inzitari D, Benetou V, Tuomilehto 17. Gerstein HC, Mann JF, Yi Q, Zinman B, important in managing type II diabetes.
of stroke in the general population in Eu- 21. International Diabetes Federation: IDF global guideline for type 2 diabetes. Avail- fibrinogen and electrocardiography? J events, death, and heart failure in diabetic Epidemiol Community Health 56 (Suppl.
and nondiabetic individuals. JAMA 286: 26. Thomsen TF, McGee D, Davidsen M, Jor- 18. Yang XL, So WY, Kong AP, Clarke P, Ho 22. Saaddine JB, Engelgau MM, Beckles GL, gensen T: A cross-validation of risk-scores for coronary heart disease mortality based diabetes report card for the United States: JC: End-stage renal disease risk equations quality of care in the 1990s. Ann Intern Studies and Framingham Heart Study. Int DIABETES CARE, VOLUME 30, NUMBER 1, JANUARY 2007


Klicpera layout2

Klinische Studien_KORR1_Musterseiten 06.09.13 07:13 Seite 19 1.2 Signifikanz (p-Wert)Um welches Problem geht es in diesem Unterkapitel?Im Rahmen einer wissenschaftlichen Studie können in der Regel nicht alle Patienten untersucht werden. Das führt zu dem Problem, dass von einer Aus- wahl (Stichprobe) auf die Grundgesamtheit (Population) geschlossen werden muss. Wie aber ist es möglich, vo


CONSIDERAZIONI SULL’USO DEGLI IMMUNOSOPPRESSORI EQUIVALENTI DOPO TRAPIANTO D’ORGANO SOLIDO Introduzione I pazienti portatori di un trapianto d’organo solido (rene, fegato, cuore, polmone, pancreas, intestino) devono assumere quotidianamente farmaci immunosoppressori per la prevenzione del L’uso di questi farmaci ha rappresentato, e rappresenta tutt’ora, uno degli

Copyright © 2014 Articles Finder