Background
Methods
Literature search
Eligibility criteria
Screening process
Data extraction
Critical appraisal
Reliability and clinical usability of available models
Statistical analysis
Results
Study selection and characteristics
Risk of bias
Development of prediction models
Reference | Derivation model | Recruitment years | Median FU time / Prediction horizon | Study Settings | Derivation cohort size | Internal Validation cohort size/method | Age range | Predictors | Outcomes | Modeling Method | Model accessibility | C statistic (95% CI) | Calibration | external validation |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Wang 2003 [17] a | 10-year Risk model of CVD | 1992 ~ 2002 | 6.1/10y | CMCS | 31,728 | No | 35–64 | Age, Gender, TC, SBP, HDL-C, smoking, FG | Fatal or nonfatal CVD | COX regression | Yes/Risk equations | Male 0.78 (0.76–0.81) Female 0.76 (0.72–0.80) | NR | NO |
Liu2004 [15] | CHD risk model | 1992–1993 1996–1999 | 10/10y | CMCS | 30,121 | No | 35–64 | Age, Gender, TC, BP, HDL-C, smoking, DM | Fatal CHD | COX regression | Yes/Risk equations | Male 0.76 (0.70–0.82) Female 0.74 (0.70–0.78) | Hosmer-Lemeshow test | NO |
Zhang2005 [27] | 10-year CVD risk prediction score | 1974–1980 | 13.5/10y | Beijing | 3000 | 1400/ random split-sample | 18–74 | Age, SBP, DBP, TC, BMI, smoking | Fatal or nonfatal CVD | COX regression | Yes/Risk equations | CHD events: training dataset 0.76/validation dataset 0.76; IS events: training dataset 0.72/validation dataset 0.78 | Hosmer– Lemeshow test | No |
Wu 2006 [16] | 10-year Risk prediction model of ICVD | 1983–1984 | 15.1/10y | USA-PRC cohort | 9903 | No | 35–59 | Age, Gender, SBP,TC,BMI,smoking, DM | Fatal or nonfatal CVD | COX regression | Yes/Risk Sheet/ online calculator | Optimal model: male 0.80 (0.76–0.83)/female 0.79 (0.76–0.83) simplified model: male 0.79 (0.76–0.83)/female 0.78 (0.75–0.82) | Hosmer– Lemeshow test | Yes |
Yang 2016 [13] | China-PAR | 1998 2000–2001 | 12.3/10y | InterASIA MUCA (1998) | 21,320 | 21,320/10*10 cross-validation | 35–74 | Age, Gender, SBP/Rx, TC, HDL-C, smoking, DM, WC, GR, FHAC, Urbanization | Fatal or nonfatal CVD | COX regression | Yes/Online calculator | Male 0.79 (0.78–0.81) Female 0.81 (0.79–0.82) | Hosmer– Lemeshow test and slope | Yes |
Hu 2017 [29] | Cardiovascular death prediction model | 1994 | 8.8/10y | Taiwan | 381,963 | No | 20+ | Age, Gender, BMI, smoking, physical activity, anemia, SBP, FG, TC, HDL, LDL, proteinrria, uric acid, CKD, CRP, heart rate, hypertension treatment | CVD death | COX regression | No | 0.91 (0.90–0.92) | NR | No |
Li 2017 [18] a | Risk prediction model of CVD | 2004 | 3.09/5y | Shandong | 50,990 | 21,853/10*10 cross-validation | 20+ | Age, Gender, BMI, DM, CKD, abnormal electrocardiogram, smoking, hypertension, dyslipidemia | Fatal or nonfatal CVD | COX regression | NO | Training dataset: male 0.84 (0.82–0.85)/Female 0.90 (0.88–0.91) Validation dataset: male 0.84 (0.81–0.86)/female 0.89 (0.87–0.91) | NR | No |
Pylypchuk 2018 [30] | PREDICT equations | 2002 | 4.2/5y | New Zealand | 401,752 | 166,611/geographical split-sample | 30–74 | Age, Gender, NZDep, smoking history, diabetes, SBP, TC/HDL, OBPLM | Fatal or nonfatal CVD | COX regression | Yes/Risk equations | Male 0.73 (0.72–0.73) Female 0.73 (0.72–0.73) | Calibration slope | No |
Li 2020 [31] | Risk prediction model of CVD | 2004 | 10/10y | Taiwan | 1481 | 740/bootstrap resampling | 40+ | Age, Gender, Marital status, BMI, smoking, physical activity, eGFR, ACR, history of heart disease, history of stroke, ABI | Fatal or nonfatal CVD | COX regression | NO | 0.88 (0.83–0.93) | Hosmer– Lemeshow test | No |
Yang 2020 [32] | CVD prediction model for high-risk CVD population | 2014 | 3/3y | Zhejiang | 19,953 | 9977/random split-sample | 35+ | Age, Gender, Family income, smoking, drinking, obesity, WC, TC, TG, LDL, FG, action capability, Self-care ability, Daily activity ability, pain, anxiety, History of hypertension/diabetes/dyslipidemia; Family history of hypertension/ischemic stroke and cerebral infarction; Hypoglycemic drugs use | CVD events | Random forest/CART/ multivariate regression/ NaïveBayes/ Bagged trees /Ada Boost | No | optimal model (random forest) from 6 models: Male 0.82/female 0.68 | Hosmer– Lemeshow test | NO |
Huang2021 [33] | GBCS prediction model | 2003–2008 | 12/10y | China/Guangzhou | 15,000 | 12,721/10*10 cross-validation | 50+ | Age, Gender, SBP, antihypertensive medication use, ever smoking, and diabetes status | Fatal or nonfatal CVD | COX regression | Yes/Risk equations | Training dataset: male 0.69 (0.67–0.71)/female 0.73 (0.71–0.74) Validation dataset: male 0.67 (0.65–0.70)/female 0.72 (0.70–0.73) | NR | No |
Wang 2015 [28] | CVD lifetime risk model | 1992 | 18/lifetime | CMCS | 21,953 | No | 35–84 | SBP/DBP, non-HDL-C, HDL-C, BMI, Diabetes, Smoking | Fatal or nonfatal CVD | Kaplan-Meier method | Yes/Risk sheet | NR | NR | No |
Predictors in the development papers
External validation of prediction models
Framingham | Framingham | PCE | WHO charts for east Asia | Asian equation | China-PAR b | Risk model (Optimal) b | Risk model (Simplied) b | |
---|---|---|---|---|---|---|---|---|
Wilson1998 | D’Agostino2008 | Stone2013 | WHO2019 | Asia2007 | Yang2016 | Wu2006 | Wu2006 | |
n = 2 | n = 5 | n = 11 | n = 2 a | n = 1 | n = 6 | n = 1 | n = 1 | |
Location of the validation cohorts | ||||||||
Single-province in mainland China | 0 | 3 | 2 | 0 | 0 | 4 | 0 | 0 |
Multi-province in mainland China | 2 | 0 | 6 | 2 | 1 | 2 | 1 | 1 |
China HongKong | 0 | 1 | 1 | 0 | 0 | 0 | 0 | 0 |
Ethic Chinese | 0 | 1 | 2 | 0 | 0 | 0 | 0 | 0 |
Participant age in the validation cohorts | ||||||||
Min, Median | 30 | 30 | 35 | 40 | 30 | 35 | 35 | 35 |
Max, Median | 75 | 74 | 79 | 80 | 75 | 74 | 59 | 59 |
Size of the Validation cohorts | ||||||||
Sample size, median[range] | 27,901 [25,682–30,121] | 7157 [438–27,721] | 20,886 [425–70,838] | 23,329 [27,321–29,337] | 25,682 | 21,631 [3347–70,838] | 15,100 | 15,100 |
Events,median[range] | 366 [191–542] | 880 [45–3732] | 622 [21–1493] | 1070 [1045–1091] | 542 | 1209 [190–3732] | 347 | 347 |
Recruitment years of the Validation cohorts | ||||||||
< 2000 year | 2 | 2 | 6 | 2 | 1 | 1 | 1 | 1 |
2001 ~ 2010 year | 0 | 2 | 4 | 0 | 0 | 5 | 0 | 0 |
> 2010 year | 0 | 1 | 1 | 0 | 0 | 0 | 0 | 0 |