Table 2.

Characteristics of studies included in systematic review of cardiovascular risk prediction models in renal transplant recipients.

Study	Objective	Outcome	Design	Sample size and inclusion criteria	Age/sex/events	Follow-up	Scoring system	Summary of results	Methodology quality
Kasiske et al. [25] 2000	To compare observed and expected incidence of IHD based on relationships of risk factors and IDH in FRS	IHD defined by MI or coronary revascularization or death due to IHD	Retrospective cohort study (United States)	n = 1124  (excluding IHD pretransplant (n = 107) and within the 1st year (n = 43), as well as RTR that did not survive >1 year with a functioning graft (n = 236))   n = 1500  (including IHD pre- or <1 year posttransplant)	40.1 ± 12.8 years (at time of transplant)     56.4% male 123 events excluding those with IHD within the 1st year posttransplant	At least 12 months Chart records reviewed from 1963 to 1997	Framingham [11]   (variables included age, gender, blood pressure, HDL cholesterol, diabetes, current smoking)	FRS predicted IHD but underpredicted risk in some populations, especially diabetes; overall actual to predicted risk of IHD was 1.28 (CI 1.20–1.40; P < 0.0001)	Sample population intentionally excluded angina pectoris and CHF, in contrast to the original Framingham cohort; IHD pretransplant and within the 1st year were also excluded; robust measures of performance were not included in this analysis.
Ducloux et al. [10] 2004	To determine incidence and risk factors for IHD and assess relevance of FRS in RTR	Coronary heart disease defined by document MI or coronary revascularization or typical angina	Prospective cohort study (France)	n = 344    RTR free of vascular disease, >12 months posttransplant, without acute rejection or Scr > 400 umol/L	51 ± 13.7 years   63% male (n = 217) 27 events in total	72 ± 14 months	Framingham [11]    (C-reactive protein and serum homocysteine also investigated)	Although the FRS predicted IHD in low-risk patients, it was underpredicted in high-risk patients: observed versus expected incidences were low-FRS = 0.6% versus 0.51%; high-FRS = 6.4% versus 2.8%	Sample population not identical to FRS; hypertension not significantly associated with CVD leading the authors to question sample size and follow up; c-statistics not used
Kiberd and Panek [28] 2008	To assess the ability of the FRS to predict CV events	MACE defined as fatal and nonfatal MI and invasive coronary artery therapy, cerebral vascular events, and other (CHF and PVD, rhythmic)	Prospective cohort study (United States)	n = 540    RTR >6 months posttransplant	48 ± 12 years   59% male (n = 322) events: n = 38 for cardiac n = 16 for stroke n = 92 for all combined	4.73 years	Framingham CV and stroke score [11]	FRS underestimated CVE across the entire cohort (observed to predicted risk 1.64 CI 1.19–2.94), but more so in patients aged 45–60 with CVD or diabetes (observed to predicted risk 2.74 CI 1.7–4.24)	Small number of events and sample limited ability to validate or develop new score; study used prevalent RTR > 6 months posttransplantation versus incident patients
Israni et al. [33] 2010	PORT dataset used to identify CHD predictive risk factors to develop risk-prediction equations at clinically important time points	CHD defined as fatal or nonfatal AMI, coronary revascularization, or sudden death.	Retrospective cohort study (14 centers worldwide)	n = 23 575  n = 3880 for FRS comparison*    kidney-only transplants; only centers that could report on required variables were included	18–34 years = 22% 35–49 years = 36% 50–64 years = 34% ≥65 = 8%   60% male 689 events in year 1; 669 events in years 1–5; 143 events for FRS comparison*	4.5 years   (data collected from Jan 1, 1990, to 2007)	2 models were developed to predict CV risk within 1 year and 1 model to predict risk within years 1–5 posttransplant (8, 7, and 12 variables, resp.)	The 3 models performed reasonably well with a time-dependent c-statistic of >0.75. Transplant related factors (e.g., DGR, AR, and eGFR) could predict CVE without FRS and FRS added little predicted value	Multicenter data with large sample size; models developed had many variables (8, 7, and 12), which predicted risk at clinically important time points. The model was not externally validated.
Silver et al. [11] 2011	To quantify predictive value of FRS and determine whether novel factors could improve	MACE defined by fatal or nonfatal MI or coronary revascularization or cardiac death	Retrospective cohort study (Canada—ethnically diverse cohort)	n = 956    Transplanted between 1998 and 2008, with >3 months of graft function	RTR with CVE: 58.3 ± 11 years RTR without CVE: 52.3 ± 12 years   RTR with CVE: 81% male RTR without CVE: 61% male 89 events in total	4.15 years	Framingham [11]   (C-reactive protein, uric acid, urine albumin-creatinine ratio also investigated)	FRS underpredicted events in all subgroups (actual to predicted event ratio was 1.2–8.4; P < 0.0001), but most notably in RTR with a history of diabetes or smoking. GFR was only non-FRS variable of predictive value after MVA (CRP, UA, and urine ACR did not).	Definition of MACE did not include angina or silent MI yet a much more inclusive definition was used for pretransplant CV history resulting in inconsistency; ethnicity was not accounted for as a confounder; small sample size
Soveri et al. [21] 2012	ALERT (a multicenter clinical trial) dataset used to develop and validate an equation for CV risk and mortality prediction in RTR	MACE defined by cardiac death, nonfatal MI, or coronary revascularization	Retrospective cohort study (multicenter-Northern Europe and Canada)	n = 1329    Renal and combined renal/pancreas, >6 months posttransplant; all were on CSA based IS, none were on statins: unstable angina within prior 6 months were excluded	50 ± 10.9 years   65.5% male (n = 917) 165 events in total	6.7 years	A 7-year MACE and mortality calculator for RTR   Variables included age, CHD, SCr, LDL, smoking, diabetes, time on dialysis, and number of transplants	A formula for a 7-year MACE and mortality prediction was developed using a 7-variable model; MACE model had a c-statistic of 0.738 and 0.740 in the assessment and test samples, and mortality model had a c-statistic of 0.734 and 0.720 in assessment and test samples, respectively	MACE prediction tool developed from population specific variables in a sufficiently sized dataset; model was internally validated and discrimination and calibration were both reported; generalizability limited to dataset inclusion criteria
Soveri et al. [35] 2013	To externally validate the 7-year MACE and mortality calculators for RTR using the PORT dataset	MACE defined by cardiac death, nonfatal MI, or coronary revascularization   All patient follow-up was censored at graft loss.	Retrospective cohort study (Europe and the United States)	n = 2967    Kidney-only transplants; only centers that could report on required variables were included	Age and sex not reported   (211 events in total)	Median follow-up with functioning graft was 4.7 years (33% had at least 7 years of follow-up)	7-year MACE and mortality calculator for RTR	MACE could be predicted with a discrimination of 0.740 but calibration indicated significant underestimation in risk in decile 5 and 9; mortality c-statistic was 0.721; underestimation of risk in decile 7 and overestimation in the highest risk decile	External validation of the 7-year MACE and mortality calculator for renal transplants in a large database. Some limitations with respect to model performance which can be attributed to differences in datasets.

FRS: Framingham risk score; RTR: renal transplant recipients; IHD: ischemic heart disease; MI: myocardial infarction; CHF: congestive heart failure; SCr: serum creatinine; CV: cardiovascular; CVD: cardiovascular disease; MACE: major adverse cardiovascular event; CHF: congestive heart failure; PVD: peripheral vascular disease; DGF: delayed graft function; AR: acute rejection; eGFR: estimated glomerular filtration rate; CRP: C-reactive protein; UA: uric acid; urine ACR: urine albumin-to-creatinine ratio; MVA: multivariate analysis; ALERT: Assessment of Lescol in Renal Transplantation; CSA: cyclosporine; IS: immunosuppression; CHD: coronary heart disease; PORT: patient outcomes in renal transplant; LDL: low density lipoprotein cholesterol.