SEARCH

SEARCH BY CITATION

Keywords:

  • Heart disease;
  • kidney transplantation

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix

Traditional risk factors do not adequately explain coronary heart disease (CHD) risk after kidney transplantation. We used a large, multicenter database to compare traditional and nontraditional CHD risk factors, and to develop risk-prediction equations for kidney transplant patients in standard clinical practice. We retrospectively assessed risk factors for CHD (acute myocardial infarction, coronary artery revascularization or sudden death) in 23 575 adult kidney transplant patients from 14 transplant centers worldwide. The CHD cumulative incidence was 3.1%, 5.2% and 7.6%, at 1, 3 and 5 years posttransplant, respectively. In separate Cox proportional hazards analyses of CHD in the first posttransplant year (predicted at time of transplant), and predicted within 3 years after a clinic visit occurring in posttransplant years 1–5, important risk factors included pretransplant diabetes, new onset posttransplant diabetes, prior pre- and posttransplant cardiovascular disease events, estimated glomerular filtration rate, delayed graft function, acute rejection, age, sex, race and duration of pretransplant end-stage kidney disease. The risk-prediction equations performed well, with the time-dependent c-statistic greater than 0.75. Traditional risk factors (e.g. hypertension, dyslipidemia and cigarette smoking) added little additional predictive value. Thus, transplant-related risk factors, particularly those linked to graft function, explain much of the variation in CHD after kidney transplantation.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix

Despite dramatic improvement in short-term outcomes, the rate of late kidney allograft failure has changed little in the past 25 years. Death with a functioning allograft is the leading cause of late allograft failure, and cardiovascular disease is the leading cause of death after kidney transplantation (1,2). Risk factors for coronary heart disease (CHD) in the general population are also risk factors for kidney transplant patients (3). However, risk-prediction equations defined and validated for the general population underestimate CHD risk for kidney transplant patients (4,5), and several nontraditional risk factors are reported to be associated with CHD after kidney transplantation (2,4,6–12). The relative ability of traditional and nontraditional, transplant-related, risk factors to predict CHD events remains unclear.

Most observational studies defining risk factors for CHD have used single-center data or large administrative data sets. Single-center studies often lack a large enough sample size or number of CHD events to achieve adequate statistical power and credible external validity. Studies using administrative data often leave uncertainty as to whether patients were followed closely and CHD events reported accurately. We collected data from 14 transplant centers for 23 575 kidney transplant patients from around the world for the Patient Outcomes in Renal Transplantation (PORT) study, and determined the best predictors of CHD both early and late after kidney transplant. These predictive risk factors were used to develop risk-prediction equations that can help clinicians risk-stratify their patients at clinically important time points, namely at the time of transplant, at 7 days posttransplant, and at any visit during 1–5 years posttransplant.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix

This study was approved by the Institutional Review Board of Hennepin County Medical Center, Minneapolis, MN, USA.

Transplant centers

We initially contacted 88 transplant centers in North America, South America, Europe, Asia and the Pacific Rim. Forty-three centers (49%) expressed interest in participating. We requested each of these centers to complete a brief survey about availability of data related to CHD and its risk factors. Twenty-four centers (27%) had the relevant data elements in existing databases and had databases available for research. In the end, 14 centers (16%) submitted usable data.

Patients, data elements and outcomes

The full PORT database included 37 076 kidney transplant patients. Only adult patients who underwent kidney-only transplants after January 1, 1990, were included in this analysis, because before that date transplant centers were less likely to collect detailed data regarding CHD events and potential CHD risk factors. Patients who underwent simultaneous pancreas–kidney transplants (n = 1599) and patients aged <18 years (n = 873) were also excluded. Our final study cohort thus consisted of 23 575 kidney transplant patients.

Participating centers submitted de-identified data to the PORT data-coordinating center at the Chronic Disease Research Group in Minneapolis, MN, USA. Recipient data elements included ethnicity; primary cause of kidney failure; body mass index (BMI); cytomegalovirus (CMV) and hepatitis serostatus; time from end-stage renal disease (ESRD) onset to transplant; history of comorbid conditions such as diabetes, hypertension, acute myocardial infarction (AMI), congestive heart failure, coronary revascularization, cerebrovascular accident, carotid endarterectomy, peripheral artery disease and cancer; smoking status at time of transplant and pretransplant blood pressure and lipid values. Donor data elements included age, sex, ethnicity and CMV serostatus. Transplant procedure data elements included panel reactive antigens (PRAs), HLA mismatches and cold ischemia time. Follow-up data included posttransplant adverse events such as cardiovascular disease, peripheral vascular disease, delayed allograft function, acute rejection, depression, CMV disease and cancer; and posttransplant serum creatinine, blood pressure, antihypertension medications and lipid values.

Estimated glomerular filtration rate (eGFR) was calculated using the four-variable Modification of Diet in Renal Disease Study equation (13). Delayed allograft function was defined as need for dialysis in the first week posttransplant. Acute rejection was defined as any rejection episode treated at the transplant center and not necessarily biopsy proven. Required patient data elements included unique patient identifier, transplant date, age at transplant, sex and donor type. Only patients with these data were included. Required center data elements included dates of death, graft failure and AMI. Only centers that could report data on these outcomes were included.

CHD was defined as fatal or nonfatal AMI, coronary revascularization (coronary artery bypass graft, angioplasty or stenting) or sudden death. All centers outside of North America (except centers in Tokyo, Japan, and Auckland, New Zealand) were visited by AKI or BLK to review data collection practices and learn about patterns of posttransplant follow-up care at each center.

Statistical analysis

The final study population was randomly divided into development and validation subsets, consisting of 16 509 (70%) and 7066 (30%) patients, respectively, stratified by transplant center. Risk-prediction models were developed using Cox proportional hazards regression analysis. Time to CHD was defined as time from transplant to the earliest date of fatal AMI, nonfatal AMI, coronary revascularization or sudden death. Patient follow-up was censored at the last known follow-up date. The first model predicted CHD risk during the first year posttransplant using risk factors available at the time of transplant. A second model, conditional on 7-day CHD-free graft survival, predicted CHD risk during the first year posttransplant using risk factors available at 7 days posttransplant. The final model predicted CHD risk during the subsequent 3 years using risk factors available at a randomly selected time point, representing a clinic visit, between 1 and 5 years posttransplant. This model included only patients who were alive with a functioning allograft at 1 year posttransplant. Details of the model development are specified in Appendix.

Performance of the final model was assessed using measures of model calibration and model discrimination as described in Appendix. Model calibration, or how closely predicted CHD outcomes agreed with actual outcomes, was measured as the slope of the prognostic index. A slope of 1.0 indicates good model calibration and a p-value >0.05 suggests that the slope is not statistically different from 1.0 (14). Model discrimination, or the ability to separate patients with higher and lower risk of CHD events, was assessed using a time-dependent c-statistic obtained from applying the model coefficients to the validation data set (15).

Once measures of calibration and discrimination for each model were assessed, the final covariates selected in the development subset were entered into a Cox model using the full study population to obtain final parameter estimates. The resulting parameter estimates (log hazard ratios) were rescaled to integers to obtain a clinically useful risk-prediction score. The risk score was calculated for each patient and mapped to a predicted probability of CHD using a Cox proportional hazards model with the risk score as the only covariate considered in the model. The predicted probability of CHD was estimated as one minus the resulting estimate of the survivor function. To assess regional variation in the predicted probability of CHD, a second probability mapping was conducted by stratifying the proportional hazards models by region (North America, Northern Europe, Southern Europe and Pacific Rim).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix

Median follow-up was 4.5 years (25th percentile, 2.0 years; 75th percentile, 7.8 years). Total cumulative incidence of CHD at 1, 3 and 5 years posttransplant was 3.1%, 5.2% and 7.6%, respectively.

Predicting CHD in the first year posttransplant

In the first-year posttransplant, 689 CHD events occurred, 49% nonfatal AMI, 38% coronary revascularization and 13% fatal AMI and sudden death. All factors listed in Table A.1 were considered in developing the CHD risk-prediction equation, after excluding variables with missing values for more than 50% of the population. For the predictive risk model derived from the Cox proportional hazards analysis of risk factors known at the time of transplant, the risk points associated with each clinical factor (Table 1) can be summed to determine an individual patient's CHD risk score. This risk score can be used to determine the absolute probability of CHD using Figure 1, panel A. For example, a patient with 14 points would have a 4.8% risk of CHD in the first year posttransplant.

Table A.1.  Recipient, donor and transplant characteristics considered in the Cox model predicting coronary heart disease risk at the time of transplant, and 1-year cumulative incidence of coronary heart disease in the model development sample (70% of the study population)
CharacteristicsPercent of population n = 23 575Model development sample (n = 16 509)
1-year cumulative incidence (%) (95% confidence interval)p
  1. AMI = acute myocardial infarction; CHF = congestive heart failure; CVA = cerebrovascular accident; ESRD = end-stage renal disease; HDL = high-density lipoprotein; LDL = low-density lipoprotein; NE = not estimable due to low prevalence; PRA = panel reactive antibodies.

  2. 1Baseline variables missing for more than 50% of the population and thus not eligible for the prediction model.

Recipient
Age (years)  <0.0001
 18–34220.7 (0.5–1.1) 
 35–49362.0 (1.7–2.4) 
 50–64344.7 (4.2–5.3) 
 ≥65 86.6 (5.4–8.1) 
Sex  0.0001
 Women402.4 (2.0–2.8) 
 Men603.4 (3.1–3.8) 
Body mass index (kg/m2)  <0.0001
 <18.5 92.4 (1.8–3.4) 
 18.5–24.9322.5 (2.1–3.0) 
 25.0–29.9213.5 (3.0–4.2) 
 30.0–34.9 94.2 (3.3–5.3) 
 ≥35.0 55.5 (4.0–7.5) 
 Unknown252.4 (2.0–3.0) 
Primary cause of ESRD  <0.0001
 Diabetes226.5 (5.7–7.4) 
 Hypertension113.5 (2.7–4.5) 
 Glomerulonephritis271.7 (1.4–2.2) 
 Cystic kidney disease102.1 (1.5–2.9) 
 Other241.8 (1.4–2.3) 
 Unknown 71.4 (0.8–2.3) 
Race  <0.0001
 White713.3 (3.0–3.7) 
 African American142.4 (1.9–3.2) 
 Asian 81.0 (0.6–1.7) 
 Other 2 8.0 (5.2–12.2) 
 Unknown 61.7 (1.1–2.9) 
Cytomegalovirus serostatus  0.4660
 Negative253.1 (2.6–3.7) 
 Positive413.1 (2.7–3.6) 
 Unknown342.8 (2.4–3.2) 
History of diabetes  <0.0001
 No741.9 (1.7–2.2) 
 Yes266.1 (5.4–6.9) 
History of hypertension  0.0012
 No322.2 (1.7–2.7) 
 Yes683.2 (2.9–3.6) 
Hepatitis C serostatus  0.9758
 Negative683.0 (2.7–3.4) 
 Positive 43.0 (1.9–4.8) 
 Unknown283.0 (2.5–3.5) 
PRA, most recent  0.1083
 <10%723.1 (2.8–3.4) 
 ≥10%123.5 (2.7–4.4) 
 Unknown172.4 (1.9–3.1) 
History of AMI  <0.0001
 No842.7 (2.4–3.0) 
 Yes 412.0 (9.7–14.8) 
 Unknown122.1 (1.5–2.9) 
History of CHF  <0.0001
 No733.1 (2.8–3.4) 
 Yes 2 9.4 (6.9–12.8) 
 Unknown252.1 (1.7–2.6) 
History of coronary revascularization  <0.0001
 No782.3 (2.0–2.6) 
 Yes 5 16.9 (14.5–19.5) 
 Unknown171.7 (1.3–2.3) 
History of CVA  <0.0001
 No853.0 (2.7–3.3) 
 Yes 35.9 (4.1–8.3) 
 Unknown122.1 (1.5–2.8) 
History of carotid endarterectomy  < 0.0001
 No663.5 (3.2–3.9) 
 Yes0.214.6 (6.3–31.6) 
 Unknown341.9 (1.6–2.3) 
History of peripheral arterial disease surgery  <0.0001
 No573.3 (2.9–3.6) 
 Yes 2 15.7 (12.3–19.9) 
 Unknown401.9 (1.6–2.3) 
History of cancer  <0.0001
 No763.1 (2.8–3.4) 
 Yes 46.3 (4.7–8.4) 
 Unknown201.9 (1.5–2.4) 
Time from ESRD onset to transplant (years)  0.0149
 0 (preemptive)122.2 (1.6–2.9) 
 >0 to ≤1273.0 (2.6–3.6) 
 >1 to ≤2152.8 (2.2–3.5) 
 >2 to ≤3104.3 (3.4–5.3) 
 ≥3253.0 (2.5–3.6) 
 Unknown103.1 (2.3–4.1) 
Pretransplant systolic blood pressure (mmHg)1  0.5153
 <120 33.5 (2.1–5.6) 
 120–129 43.0 (1.9–4.8) 
 130–139 72.4 (1.7–3.6) 
 140–159142.5 (1.9–3.3) 
 ≥1609.52.9 (2.2–3.9) 
 Unknown633.2 (2.8–3.5) 
Pretransplant diastolic blood pressure (mmHg)1  <0.0001
 <80134.5 (3.7–5.6) 
 80–84 71.9 (1.2–2.8) 
 85–89 51.7 (1.0–2.9) 
 90–99 82.2 (1.5–3.2) 
 ≥100 51.1 (0.6–2.3) 
 Unknown633.2 (2.8–3.5) 
Pretransplant total cholesterol (mg/dL)1  0.0002
 <160114.1 (3.3–5.2) 
 160–199123.4 (2.7–4.4) 
 200–239 84.2 (3.2–5.5) 
 240–279 33.1 (1.9–5.1) 
 ≥280 24.0 (2.1–7.6) 
 Unknown652.6 (2.3–2.9) 
Pretransplant HDL cholesterol (mg/dL)1  <0.0001
 <40104.6 (3.7–5.7) 
 40–49 64.3 (3.2–5.8) 
 50–59 43.9 (2.7–5.7) 
 ≥60 53.3 (2.3–4.8) 
 Unknown742.6 (2.3–2.9) 
Pretransplant LDL cholesterol (mg/dL)1  0.0001
 <100134.1 (3.3–5.0) 
 100–129 63.7 (2.7–5.1) 
 130–159 44.6 (3.2–6.7) 
 160–189 14.6 (2.4–8.6) 
 ≥190 1 4.8 (2.2–10.5) 
 Unknown752.6 (2.3–2.9) 
Pretransplant triglycerides (mg/dL)1  0.0021
 <150163.3 (2.6–4.0) 
 150–199 73.8 (2.8–5.1) 
 200–499124.1 (3.3–5.1) 
 ≥500 14.4 (2.1–9.1) 
 Unknown642.6 (2.3–3.0) 
History of smoking1  0.0001
 Yes113.2 (2.5–4.2) 
 No204.1 (3.4–4.8) 
 Unknown692.6 (2.4–3.0) 
Donor
Age (years)  0.1990
 0–17 62.7 (1.8–4.1) 
 18–34193.0 (2.4–3.7) 
 35–49242.5 (2.0–3.0) 
 50–64203.1 (2.5–3.7) 
 ≥65 52.8 (1.8–4.3) 
 Unknown283.5 (3.0–4.1) 
Sex  0.0002
 Women372.5 (2.2–3.0) 
 Men422.9 (2.5–3.3) 
 Unknown224.1 (3.4–4.8) 
Race  0.0043
 White423.5 (3.1–3.9) 
 African American51.9 (1.1–3.2) 
 Other/unknown532.7 (2.4–3.1) 
Cytomegalovirus serostatus  0.1876
 Negative253.3 (2.8–3.9) 
 Positive322.7 (2.3–3.2) 
 Unknown433.1 (2.7–3.5) 
Type  0.0099
 Deceased613.3 (2.9–3.7) 
 Living392.6 (2.2–3.0) 
Body mass index (kg/m2)1  <0.0001
 <18.5 21.5 (0.6–4.0) 
 18.5–24.9123.8 (3.0–4.8) 
 25.0–29.9 94.3 (3.3–5.5) 
 30.0–34.9 35.1 (3.5–7.5) 
 ≥35.0 23.0 (1.6–5.3) 
 Unknown712.6 (2.4–3.0) 
History of diabetes1  <0.0001
 Yes<1 9.4 (4.6–18.8) 
 No181.4 (1.0–1.9) 
 Unknown823.3 (3.0–3.6) 
History of hypertension1  <0.0001
 Yes 2 6.9 (4.6–10.4) 
 No 92.3 (1.6–3.3) 
 Unknown893.0 (2.7–3.3) 
Hepatitis C serostatus1  0.0001
 Positive 15.1 (2.8–9.3) 
 Negative323.7 (3.2–4.3) 
 Unknown672.6 (2.3–2.9) 
History of smoking1  0.2473
 Yes 34.2 (2.7–6.5) 
 No 43.6 (2.3–5.5) 
 Unknown932.9 (2.7–3.2) 
Transplant
Year  0.0051
 1990–1994232.3 (1.8–2.8) 
 1995–1999282.9 (2.5–3.5) 
 2000–2007493.4 (3.0–3.8) 
Number  0.2450
 First873.1 (2.8–3.4) 
 Subsequent132.5 (1.9–3.3) 
 Unknown0.2NE 
HLA mismatches  0.0008
 0–2322.7 (2.3–3.1) 
 3–6502.8 (2.5–3.2) 
 Unknown184.2 (3.5–5.0) 
Cold ischemia time (h)  0.0523
 0–12452.6 (2.3–3.0) 
 13–24203.3 (2.7–4.0) 
 25–35 53.8 (2.7–5.4) 
 ≥36 11.3 (0.4–4.0) 
 Unknown283.3 (2.8–3.9) 
Table 1.  Risk factors for coronary heart disease in the first year after kidney transplant1
CharacteristicPercent of population (n = 23 575)Adjusted hazard ratio (90% confidence interval)Risk score2
  1. ESRD = end-stage renal disease.

  2. 1Cox proportional hazards analysis for coronary heart disease events predicted by covariates known at the time of transplant.

  3. 2The risk score for an individual patient can be calculated by adding the individual points and then determining probability of coronary heart disease using probabilities in Figure 1 (panel A). The maximum number of points that can be assigned to a patient is 28.

  4. 3Includes acute myocardial infarction, congestive heart failure, coronary revascularization, cerebrovascular accident and peripheral arterial disease surgery (including all revascularization procedures for peripheral artery disease).

Recipient age (years)
 18–34221.000
 35–49362.07 (1.52–2.84)4
 50–64343.79 (2.80–5.12)7
 ≥65 84.99 (3.60–6.91)8
Recipient sex
 Women401.000
 Men601.22 (1.07–1.40)1
Recipient history of diabetes
 No741.000
 Yes262.00 (1.75–2.28)3
Recipient history of cancer
 No/none reported961.000
 Yes 41.38 (1.10–1.73)2
Number of cardiovascular comorbid conditions3
 0861.000
 1113.76 (3.24–4.36)7
 ≥2 35.89 (4.91–7.08)9
Donor type
 Living391.000
 Deceased611.24 (1.08–1.43)1
Body mass index (kg/m2)
 <35.0961.000
 ≥35.0 41.55 (1.24–1.94)2
Years from first ESRD treatment to transplant
 0 (preemptive)131.000
   >0 to ≤2421.27 (1.00–1.60)1
   >2451.41 (1.11–1.79)2
image

Figure 1. Predicted probability of coronary heart disease (dark line) with 95% confidence interval (lighter lines), and distribution of risk scores in the Patient Outcomes after Renal Transplant population (bars). (Panel A) Prediction of coronary heart disease in the first year posttransplant from variables available at the time of transplant (Table 1). (Panel B) Prediction of coronary heart disease in the first year posttransplant from variables available at the time of transplant and during the first week after transplant (Table 2). (Panel C) Prediction of coronary heart disease within 3 years after a visit occurring 1–5 years posttransplant (Table 3).

Download figure to PowerPoint

Predicting CHD in the first year posttransplant with data from week one

Using additional variables available at 1 week posttransplant (e.g. eGFR calculated at or near day 7 posttransplant, delayed graft function), a second CHD risk-prediction equation was developed (Table 2). This model was conditional on allograft survival and no CHD events in the first week posttransplant. Additional variables that were considered but not retained in the model are listed in Table A.2. In the first year posttransplant, 22 855 patients experienced 464 CHD events, 31% nonfatal AMI, 54% coronary revascularization and 16% fatal AMI and sudden death. An individual patient's score can be used to determine the absolute probability of CHD using Figure 1, panel B. For example, a patient with 14 points at 1 week posttransplant would have a 2.2% risk of CHD in the first year posttransplant.

Table 2.  Risk factors for coronary heart disease in the first year after kidney transplant, conditional on 7-day survival without a coronary heart disease event1
CharacteristicPercent of population (n = 22 855)Adjusted hazard ratio (90% confidence interval)Risk score2
  1. ESRD = end-stage renal disease.

  2. 1Cox proportional hazards analysis for coronary heart disease events predicted by covariates known within the first week after transplant.

  3. 2The risk score for an individual patient can be calculated by adding the individual points and then determining probability of coronary heart disease using probabilities in Figure 1 (panel B). The maximum number of points that can be assigned to a patient is 31.

  4. 3Includes acute myocardial infarction, congestive heart failure, coronary revascularization, cerebrovascular accident and peripheral arterial disease surgery (including all revascularization procedures for peripheral artery disease).

  5. 4Defined as need for dialysis in the first week after transplant.

Recipient age (years)
 18–34231.000
 35–49361.96 (1.37–2.81)4
 50–64343.08 (2.18–4.37)6
 ≥65 84.44 (3.05–6.46)8
Recipient sex
 Women401.000
 Men601.21 (1.02–1.42)1
Recipient history of diabetes
 No741.000
 Yes262.03 (1.73–2.40)4
Number of cardiovascular comorbid conditions3
 0871.000
 1104.09 (3.42–4.90)8
 ≥2 36.46 (5.18–8.05)10 
Body mass index (kg/m2)
 <30.0861.000
 ≥30.0141.46 (1.22–1.75)2
Years from first ESRD treatment to transplant
 0 (preemptive)131.000
 >0 to ≤2431.74 (1.25–2.42)3
 >2452.37 (1.71–3.29)5
Delayed graft function4
 No881.000
 Yes121.22 (0.99–1.50)1
Table A.2.  Recipient characteristics considered but not retained in the Cox model predicting coronary heart disease risk at any point within 3 years after a visit during years 1–5 posttransplant, and 3-year cumulative incidence of coronary heart disease in the model development sample (70% of the study population)
Posttransplant recipient characteristicsModel development sample (n = 13 711)
CharacteristicPercent of population (n = 19 578)3-year cumulative incidence (%) (95% confidence interval)p
  1. 1Excluding nonmelanoma skin cancers, including posttransplant lymphoproliferative disorder.

  2. 2Stroke/transient ischemic attack.

Years posttransplant  0.3625
 1 to <2425.4 (4.6–6.2) 
 2 to <3254.8 (4.0–5.8) 
 3 to <4195.4 (4.4–6.5) 
 4 to 5154.5 (3.6–5.7) 
Depression  0.0253
 No975.0 (4.6–5.5) 
 Yes 3 8.2 (5.4–12.1) 
Cytomegalovirus disease  0.1320
 No905.0 (4.5–5.5) 
 Yes105.9 (4.6–7.7) 
Nonmelanoma skin cancer  0.0075
 No985.0 (4.6–5.5) 
 Yes 2 9.5 (4.5–13.2) 
Cancer1  0.0844
 No985.1 (4.6–5.5) 
 Yes 2 7.7 (4.5–13.2) 
Congestive heart failure  <0.0001 
 No974.9 (4.4–5.3) 
 Yes 3 15.8 (11.2–22.1) 
Cerebrovascular accident2  <0.0001 
 No995.0 (4.6–5.5) 
 Yes 114.8 (8.8–24.5) 
Carotid endarterectomy  0.0032
 No99.9 5.1 (4.6–5.6) 
 Yes0.127.1 (7.5–72.4) 

Predicting CHD beyond the first year posttransplant

The model developed to predict CHD risk within 3 years of a clinic visit during years 1–5 posttransplant (Table 3) included the most recent eGFR calculated within 1 year before the risk assessment. The model was conditional on allograft survival at 1 year posttransplant, resulting in 19 578 patients with 669 CHD events in this 3-year observation period, 39% nonfatal AMI, 38% coronary revascularization and 23% fatal AMI and sudden death. An individual patient's score can be used to determine the absolute probability of CHD using Figure 1, panel C. For example, a patient with 14 points at 1 year posttransplant would have a 3.4% risk of CHD in the subsequent 3 years.

Table 3.  Risk factors for coronary heart disease within 3 years after a visit occurring 1–5 years after kidney transplant1
CharacteristicPercent of population (n = 19 578)Adjusted hazard ratio 90% confidence intervalRisk score2
  1. CVD = cardiovascular disease; eGFR = estimated glomerular filtration rate; ESRD = end-stage renal disease; PVD = peripheral vascular disease.

  2. 1Cox proportional hazards analysis for coronary heart disease events predicted by covariates known 1–5 years after transplant.

  3. 2The risk score for an individual patient can be calculated by adding the individual points and then determining probability of coronary heart disease using probabilities in Figure 1 (panel C). The maximum number of points that can be assigned to a patient is 64.

  4. 3Estimated glomerular filtration rate was unknown for 21% of patients. A missing indicator was included in the risk-prediction model but removed from the final risk score.

  5. 4Includes acute myocardial infarction, congestive heart failure, coronary revascularization, cerebrovascular accident and peripheral arterial disease surgery (including all revascularization procedures for peripheral artery disease).

  6. 5Defined as need for dialysis in the first week after transplant.

Recipient age (years)
 18–34231.000
 35–49362.37 (1.81–3.11)6
 50–64333.23 (2.47–4.22)8
 ≥65 73.58 (2.61–4.91)9
Recipient sex
 Women401.000
 Men601.30 (1.13–1.50)2
Recipient race
 White721.45 (1.13–1.86)3
 African American141.16 (0.85–1.58)1
 Other141.000
Most recent panel reactive antibodies at time of transplant
 <10%891.000
 ≥10%111.20 (0.99–1.46)1
Years from first ESRD treatment to transplant
 0 (preemptive)131.000
 >0 to ≤2441.22 (0.98–1.53)1
 >2431.41 (1.12–1.77)2
Acute rejection episode in prior year
 No951.000
 Yes 52.21 (1.71–2.86)5
Posttransplant lymphoproliferative disorder
 No99.71.000
 Yes0.34.01 (2.03–7.90)9
Recipient history of diabetes
 No661.000
 Pretransplant292.41 (2.08–2.79)6
 New onset posttransplant 51.86 (1.43–2.43)4
Most recent eGFR (mL/min/1.73 m2)3
 ≥50371.000
 40 to <50181.18 (0.98–1.42)1
 <40241.46 (1.23–1.72)3
Number of cardiovascular comorbid conditions4
 0861.000
 1101.80 (1.53–2.12)4
 ≥2 42.69 (2.17–3.33)7
Posttransplant CVD or PVD events
 Coronary revascularization 22.57 (2.05–3.22)6
 Acute myocardial infarction 22.37 (1.89–2.99)6
 PVD surgery 21.59 (1.26–2.01)3
Delayed graft function5
 No891.000
 Yes111.39 (1.17–1.66)2

We also examined CHD risk in a subset of patients from the 5 (of 14) study centers that routinely collected other traditional risk factors as defined by the Framingham Heart Study, such as total cholesterol, high-density lipoprotein cholesterol, smoking, blood pressure and use of antihypertensive medications (n = 3880 with 143 CHD events; Table A.3). We compared our third prediction model and a variation of it that included these Framingham risk factors, and found that the Framingham variables did not significantly improve the risk prediction (likelihood ratio test, p = 0.0937). Significance level and direction were similar for the coefficients for the prediction variables in this model, even after including these Framingham risk factors (data not shown). Of all these Framingham risk factors, total cholesterol, blood pressure and use of antihypertensive medications were correlated with CHD. However, the discriminatory ability of the prediction model was not significantly improved by adding these Framingham risk factors and the prediction model performed much better than the Framingham Heart Study equation (Figure 2). We also examined CHD risk in the subset of patients from the 7 (of 14) study centers that routinely collected information on smoking status at the time of transplant (n = 9785 with 361 CHD events; Table A.4). Smoking status did not independently predict CHD, significantly affect which variables predicted CHD events, or affect the discriminatory ability of the prediction model (data not shown). Smokers were found to be more often men and obese, and to have a history of cancer and longer dialysis duration than nonsmokers (data not shown). Thus, smoking was strongly correlated with other CHD risk factors.

Table A.3.  Data elements available for Framingham risk factors; patients (n = 3880, 143 coronary heart disease events) were included in a subanalysis to predict coronary heart disease at any point within 3 years after a visit during years 1–5 posttransplant
CenterData elements (% of patients)
Total cholesterolHDL cholesterolBlood pressureSmoking statusIncluded in subanalysis (%) (n)
  1. HDL = high-density lipoprotein.

19780955377 (446)
27973858869 (749)
39288968686 (963)
4626174 0 61 (1531)
5616178 053 (191)
1–574718339n = 3880
image

Figure 2. Comparison of discriminatory ability of the Patient Outcomes after Renal Transplant (PORT) prediction equation, the Framingham equation and the PORT equation modified to include traditional Framingham risk factors such as blood pressure, use of antihypertensive medications and total cholesterol. Smoking was not significant and therefore was not included in the PORT equation with the traditional Framingham risk factors. Time-dependent c-statistic for predicting coronary heart disease (CHD) events within 3 years after a clinic visit 1–5 years posttransplant.

Download figure to PowerPoint

Table A.4.  Smoking status, at time of transplant by center; smoking status of these patients (n = 9785, 361 coronary heart disease events) was included in a subanalysis to predict coronary heart disease at any point within 3 years after a visit during years 1–5 posttransplant
CenterSmoking status (% of patients)
CurrentFormerNeverMissingIncluded in subanalysis (%) (n)
114.918.919.247.05.9 (577)
218.527.442.012.211.2 (1093)
3 9.836.239.914.111.5 (2450)
420.232.245.4 2.025.0 (878) 
516.548.535.00 9.0 (1123)
6 7.630.359.2 2.92.4 (238)
721.2078.8035.0 (3426)
1–718.221.553.9 6.3n = 9785

Given the regional variation in incident rates of CHD (Figure A.1), the risk scores from each of the three models can be used to determine the absolute probability of CHD according to a patient's region (Figure A.2). For each of the three prediction models, the association of risk factors with CHD when stratified by region resulted in a similar hazard ratio in direction and magnitude of association (data not shown).

image

Figure A.1. Coronary heart disease-free posttransplant survival in different regions worldwide.

Download figure to PowerPoint

imageimageimage

Figure A.2. Predicted probability of coronary heart disease (CHD) for North America, southern Europe, northern Europe and Pacific Rim. (Panel A) In first-year posttransplant from variables available at the time of transplant (Table 1). (Panel B) In first-year posttransplant from variables available at 7 days posttransplant (Table 2). (Panel C) At any point within 3 years after a clinic visit occurring 1–5 years posttransplant (Table 3).

Model performance

The discriminatory ability of each of the three risk-prediction models was assessed using a time-dependent c-statistic, with bootstrapped 95% confidence intervals, created for each model (Figure A.3). Values of the c-statistic at each posttransplant time point can be interpreted as the discriminatory ability of the model, with values near 0.50 indicating no discriminatory ability and 1.0 indicating perfect discriminatory ability. c-Statistic values above 0.70 are considered acceptable, and values above 0.80 are interpreted as having excellent discriminatory ability. c-Statistic values were 0.80–0.85, 0.73–0.83 and 0.73–0.80 for the three models described, respectively (Figure A.3). Calibration was statistically acceptable for all three models, with parameter estimates of 1.07 (p = 0.25), 0.91 (p = 0.26) and 1.03 (p =−0.72), respectively.

image

Figure A.3. Discriminatory ability using time-dependent c-statistic (dark line) with 95% confidence interval (lighter lines). (Panel A) Predicting coronary heart disease (CHD) within the first year posttransplant from variables available at the time of transplant. (Panel B) Predicting CHD within the first-year posttransplant from variables available at 7 days posttransplant. (Panel C) Predicting CHD at any point within 3 years after a clinic visit occurring 1–5 years posttransplant.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix

In this retrospective observational study, we identified several important, nontraditional, transplant-related risk factors for CHD events that occurred both early and late after kidney transplant. Importantly, several of these risk factors are directly or indirectly related to kidney allograft function. In a subset analysis, some traditional CHD risk factors, identified in the Framingham Heart Study, correlated with CHD events, but added little to the predictive power of the transplant-related risk factors. Finally, we used this information to develop practical risk-prediction equations that allow clinicians to estimate CHD risk in individual patients.

The transplant-related risk factors identified in this analysis have been correlated to CHD in other studies. For example, poor kidney allograft function, a potentially modifiable risk factor, is associated with increased risk of CHD (6,7). Consistent with our findings, other observational studies have shown that deceased donors (8), acute rejection (2,4,6,8), obesity (9), longer duration of pretransplant dialysis (2,10,16), diabetes or hypertension as cause of ESRD (11,12) and delayed graft function (11) increase CHD risk. However, our large sample size and the large number of CHD events in our study allowed us to test the relative contribution of all of these risk factors together. For example, the risk of CHD associated with pretransplant diabetes is larger than the risk of new-onset posttransplant diabetes, presumably due to long-term effects of pretransplant diabetes compared with the shorter duration of new-onset diabetes (Table 3).

The statistical power of this study likely explains why we also detected some previously unreported associations with CHD, such as pretransplant history of cancer and posttransplant lymphoproliferative disorders. Prevalence of these risk factors was low (<5%), but because of the large sample size, they were significant in the risk-prediction models. These associations should be confirmed in future studies. Nevertheless, in the general population, CHD has been recognized as a long-term complication of surviving cancer treatment (17,18). Other general population studies have shown that cancer and CHD share similar risk factors, such as increased age and smoking (19).

In this study, we developed and validated a series of equations to estimate CHD risk. These equations give clinicians an opportunity to improve risk stratification to target preventive therapies appropriately. Modifiable, independent, risk factors that we identified include BMI, duration of pretransplant dialysis, acute rejection, posttransplant lymphoproliferative disorders, new-onset diabetes and eGFR posttransplant. The impact of risk factor modification on future CHD events should be determined by randomized controlled trials and cannot be determined by these risk-prediction equations. Our finding that some traditional Framingham risk factors added little to the prediction equations we developed should not be interpreted to mean that these traditional risk factors are not important or should not be treated. The lack of a stronger predictive ability in this analysis may be due to insufficient data on these risk factors being collected, especially late after transplant. Alternatively, traditional risk factors such as posttransplant hypertension are known to be related to kidney allograft function and merit intervention (20,21).

Our findings also complement the Framingham Heart Study CHD risk prediction, because the Framingham Heart Study did not include risk factors pertinent to kidney transplant patients. Pretransplant, especially while patients are on dialysis, some studies have noted a reverse epidemiology of Framingham risk factors (22). Almost all patients pretransplant tend to be hypertensive, and blood pressure is determined by the amount of fluid removed during dialysis. Thus, it was not surprising that transplant centers were not recording these risk factors pretransplant. After the first-year posttransplant, Framingham risk factors such as hypertension, cholesterol and smoking were available only for a subset of patients. However, these Framingham risk factors are modifiable and provide opportunity for risk reduction, particularly in high-risk patients identified in this study.

Given the large number of patients and CHD events across 14 transplant centers worldwide, external validity of these results is likely high. Nevertheless, our study has several important limitations. First, we included only centers that had systematically collected data on CHD and CHD risk; possibly, CHD management at these centers may differ from management at centers that do not collect these data. Second, despite selecting centers with the best data on CHD risk, only a subset of centers (n = 3880 patients) had routinely collected data on some traditional CHD risk factors, such as blood pressure, lipids and cigarette smoking. Although we excluded centers and data not collected routinely, irrespective of CHD, selection bias is possible in the collection of data from these patients. Third, we used a combined CHD endpoint of nonfatal AMI, coronary revascularization, fatal AMI and sudden death. However, this combined endpoint is an appropriate choice, given that it has typically been used in major clinical trials evaluating interventions for prevention of CHD. Fourth, we could not validate our prediction equation in an independent, equally large data set of kidney transplant patients, because such a resource is unavailable. Fifth, we could not include allograft loss as a risk factor because most transplant centers do not follow their patients after allograft loss. We also did not include immunosuppression as a risk factor because selection of an immunosuppressive regimen is confounded by other risk factors, and inclusion of such factors could lead to erroneous risk prediction. A randomized clinical trial of immunosuppression would provide a more appropriate data set for predicting CHD risk for different immunosuppressants. Sixth, we could not assess whether the impact of individual risk factors on CHD varies by region due to limited statistical power to fully assess interactions between regions. Lastly, novel CHD risk factors or future treatments may change the predictive capability of these risk-prediction equations.

In summary, transplant-related risk factors, particularly those linked to kidney graft function, explain much of the variation in CHD after kidney transplant. Kidney graft function is potentially modifiable, and thus provides an opportunity for reducing posttransplant CHD events. Application of the risk-prediction equations could help clinicians risk-stratify their patients.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix

This work was supported by a grant from Bristol-Myers Squibb Company, Princeton, NJ. The interpretation and reporting of these data are the responsibility of the authors. A.K.I. is a Robert Wood Johnson Foundation Faculty Scholar. The other authors report no conflicts of interest. The authors thank Chronic Disease Research Group colleagues Shane Nygaard, BA, for manuscript preparation and Nan Booth, MSW, MPH, for manuscript editing.

The PORT Investigators are Daniel Brennan, MD (Steering Committee), Washington University, Barnes Jewish Hospital, St. Louis, MO, USA; Jeffrey Connaire, MD, and Ajay Israni, MD, MS (Steering Committee), Hennepin County Medical Center, Minneapolis, MN, USA; Robert Gaston, MD, University of Alabama at Birmingham, Birmingham, AL, USA; John Gill, MD (Steering Committee), University of British Columbia St. Paul's Hospital, Vancouver, Canada; Christophe Legendre, MD, and Henri Kreis, MD, Hôpital Necker Adulte, Paris, France; Kai Lopau, MD, University of Würzburg, Würzburg, Germany; Arthur Matas, MD, and Bertram Kasiske, MD (Steering Committee), University of Minnesota, Minneapolis, MN, USA; Todd Pesavento, MD, Ohio State University, Columbus, OH, USA; Helen Pilmore, MD, Auckland City Hospital, Auckland, New Zealand; John Pirsch, MD, University of Wisconsin, Madison, WI, USA; Kazunari Tanabe, MD, and Kiyoshi Setoguchi, MD, Tokyo Women's Medical University, Tokyo, Japan; Armando Torres, MD, Domingo Hernandez, MD, and Esteban Porrini, MD, Hospital Universitario de Canarias, La Laguna, Tenerife, Spain; Yves Vanrenterghem, MD (Steering Committee), Universitaire Ziekenhuizen, Leuven, Belgium; Bruno Watschinger, MD, Universität Wien, Allgemeinen Krankenhaus (AKH) der Stadt Wien, Vienna, Austria.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix
  • 1
    United States Renal Data System. USRDS 2008 Annual Data Report: Atlas of End-Stage Renal Disease in the United States. Bethesda , MD : National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, 2008.
  • 2
    Kasiske BL, Guijarro C, Massy ZA, Wiederkehr MR, Ma JZ. Cardiovascular disease after renal transplantation. J Am Soc Nephrol 1996; 7: 158165.
  • 3
    Expert Panel on Detection EaToHBCiA. Executive Summary of the Third Report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III). JAMA 2001; 285: 24862497.
  • 4
    Kasiske BL, Chakkera HA, Roel J. Explained and unexplained ischemic heart disease risk after renal transplantation. J Am Soc Nephrol 2000; 11: 17351743.
  • 5
    Ducloux D, Kazory A, Chalopin JM. Predicting coronary heart disease in renal transplant recipients: a prospective study. Kidney Int 2004; 66: 441447.
  • 6
    Jardine AG, Fellstrom B, Logan JO et al. Cardiovascular risk and renal transplantation: Post hoc analyses of the Assessment of Lescol in Renal Transplantation (ALERT) Study. Am J Kidney Dis 2005; 46: 529536.
  • 7
    Meier-Kriesche HU, Baliga R, Kaplan B. Decreased renal function is a strong risk factor for cardiovascular death after renal transplantation. Transplantation 2003; 75: 12911295.
  • 8
    Rigatto C, Parfrey P, Foley R, Negrijn C, Tribula C, Jeffery J. Congestive heart failure in renal transplant recipients: Risk factors, outcomes, and relationship with ischemic heart disease. J Am Soc Nephrol 2002; 13: 10841090.
  • 9
    Aker S, Ivens K, Grabensee B, Heering P. Cardiovascular risk factors and diseases after renal transplantation. Int Urol Nephrol 1998; 30: 777788.
  • 10
    Abbott KC, Bucci JR, Cruess D, Taylor AJ, Agodoa LY. Graft loss and acute coronary syndromes after renal transplantation in the United States. J Am Soc Nephrol 2002; 13: 25602569.
  • 11
    Lentine KL, Brennan DC, Schnitzler MA. Incidence and predictors of myocardial infarction after kidney transplantation. J Am Soc Nephrol 2005; 16: 496506.
  • 12
    Lentine KL, Schnitzler MA, Abbott KC et al. De novo congestive heart failure after kidney transplantation: A common condition with poor prognostic implications. Am J Kidney Dis 2005; 46: 720733.
  • 13
    Levey AS, Coresh J, Greene T et al. Using standardized serum creatinine values in the modification of diet in renal disease study equation for estimating glomerular filtration rate. Ann Intern Med 2006; 145: 247254.
  • 14
    Steyerberg EW, Eijkemans MJ, Harrell FE Jr, Habbema JD. Prognostic modeling with logistic regression analysis: In search of a sensible strategy in small data sets. Med Decis Making 2001; 21: 4556.
  • 15
    Mandel M, Galai N, Simchen E. Evaluating survival model performance: A graphical approach. Stat Med 2005; 24: 19331945.
  • 16
    Vanrenterghem YF, Claes K, Montagnino G et al. Risk factors for cardiovascular events after successful renal transplantation. Transplantation 2008; 85: 209216.
  • 17
    Vaughn DJ, Palmer SC, Carver JR, Jacobs LA, Mohler ER. Cardiovascular risk in long-term survivors of testicular cancer. Cancer 2008; 112: 19491953.
  • 18
    Moser EC, Noordijk EM, Van Leeuwen FE et al. Long-term risk of cardiovascular disease after treatment for aggressive non-Hodgkin lymphoma. Blood 2006; 107: 29122919.
  • 19
    Chan AO, Jim MH, Lam KF et al. Prevalence of colorectal neoplasm among patients with newly diagnosed coronary artery disease. JAMA 2007; 298: 14121419.
  • 20
    Mange KC, Cizman B, Joffe M, Feldman HI. Arterial hypertension and renal allograft survival. JAMA 2000; 283: 633638.
  • 21
    Cosio FG, Pelletier RP, Sedmak DD, Pesavento TE, Henry ML, Ferguson RM. Renal allograft survival following acute rejection correlates with blood pressure levels and histopathology. Kidney Int 1999; 56: 19121919.
  • 22
    Kalantar-Zadeh K, Block G, Humphreys MH, Kopple JD. Reverse epidemiology of cardiovascular risk factors in maintenance dialysis patients. Kidney Int 2003; 63: 793808.

Appendix

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix
Model development

Variables with values missing for more than 50% of the population were not included in the models. All continuous variables were categorized before modeling according to clinically meaningful groups based on existing literature. Missing values were initially assigned to a separate category. The univariate relationship between each candidate variable and coronary heart disease incidence was then modeled using Kaplan–Meier methodology. The log-rank statistics obtained from Kaplan–Meier analyses were used to assess the univariate relationship between clinical factors and coronary heart disease occurrence. Variables with univariate associations and p < 0.05 were considered for inclusion in the final model. To obtain a parsimonious prediction model with clinical utility, a backward stepwise selection process was used for final variable selection. p < 0.10 was the primary criterion for variable retention. This more liberal criterion was used, not the traditional p < 0.05, to improve predictive ability (14). The resulting models were compared to models that used a still more liberal retention criterion of p < 0.50, and a more conservative criterion of p < 0.05 to assess model sensitivity to the p < 0.10 criterion. Missing value categories found not to be significantly different from a nonmissing category were combined with the category most closely associated with the missing value.

Method for model calibration and discrimination

For model calibration, the slope of the prognostic index was obtained from 200 bootstrap samples of the development data set. In each bootstrapped sample, covariate selection proceeded using a backward selection process with a retention criterion of p < 0.10. In each bootstrapped sample, the prognostic index was calculated as linear combination of the bootstrapped model coefficients with the values of the covariates for each observation in the bootstrapped sample. The prognostic index was then entered into a Cox model with the prognostic index being the only covariate in the model yielding an estimated slope of the prognostic index. This process was repeated for each of the 200 bootstrapped samples. The mean of the 200 slopes was calculated and used as a linear shrinkage factor. The prognostic index was then calculated for each observation in the validation set and a slope was estimated from a Cox model in which the only covariate was the prognostic index.

Discrimination is the ability of the model to separate patients with a higher risk of coronary heart disease events from those with a lower risk; this was assessed using a time-dependent c-statistic obtained from applying the model coefficients to the validation data set (15). We used the shrinkage factor to modify the parameter estimates prior to calculating the time-dependent c-statistic in the validation data set. Confidence intervals for the time-dependent c-statistic were obtained from 200 bootstrapped samples of the validation set.