Incidence and Predictors of Myocardial Infarction after Kidney Transplantation

PTMI.

PTMI was defined by identification of a posttransplantation Medicare claim (Part A or B) with an MI diagnosis (International Classification of Diseases, Ninth Revision [ICD-9] codes 410.x) or by death after transplantation with MI specified as a cause. For patients with multiple qualifying events, we selected the earliest event date as the date of PTMI. Observations were censored at the earliest of the following events: Loss to follow-up, loss of Medicare, 3 yr after transplant (the time when Medicare coverage ends after kidney transplantation in the absence of extenuating disability) to avoid censoring bias, death not caused by PTMI, or end of observation (December 31, 2000).

Recipient, Donor, and Transplant-Related Characteristics and Outcomes.

We collected the following recipient demographics and clinical characteristics for evaluation: Age, race, Hispanic ethnicity, gender, primary cause of ESRD, time from first dialysis until transplantation, college education, employment (full or part time), obesity (body mass index ≥30 kg/m²), and comorbid conditions. The date of first dialysis was defined as the earliest dialysis start date identified among records of UNOS, the Center for Medicare and Medicaid Studies (CMS Form 2728), and the USRDS treatment history file. Recipient characteristics and comorbidities were those reported by UNOS at the time of transplantation, supplemented with information on pretransplantation conditions from the CMS 2728 when available (i.e., conditions known at the time of ESRD reporting) and with pretransplantation claims data in cases of diabetes (ICD-9 250.x), MI, peripheral vascular disease (ICD9 440.2x, 443.9), hypertensive heart disease (ICD-9 402.x), dyslipidemia (ICD-9 272.0 to 272.5), cardiac arrhythmia (ICD-9 427.x), and tobacco use (ICD-9 305.1x). A diagnosis of diabetes was identified from claims data on the basis of one inpatient or two outpatients claims, with the date of onset set at the earliest claim, as previously validated for diabetes (Hebert method) (16). We distinguished diabetes related and unrelated to ESRD. Posttransplantation diabetes was defined as qualifying posttransplantation claims data in patients without an indication of diabetes before transplantation (related or unrelated to renal failure). After confirming with a sensitivity analysis that use of single claims produced similar diagnosis rates and risk associations with PTMI for dyslipidemia as the Hebert method (2% absolute difference in diagnosis rates and 3% absolute risk difference), we required single claims for other conditions. Donor and transplant procedure characteristics were derived from UNOS records, including donor age, race, gender, source (deceased versus living), the year of transplantation, recipient sensitization (panel of reactive antibodies ≥50%), the number of donor-recipient human leukocyte antigen mismatches, donor-recipient cytomegalovirus (CMV) seropairing, cold ischemia time, and delayed graft function defined as dialysis during the first 7 d posttransplantation. We collected immunosuppression data recorded on the UNOS recipient registration form for analysis on an intention-to-treat basis. We defined induction therapy as recorded administration of any of the following during the transplant hospitalization for the specified purpose of induction: Anti-thymocyte globulins, anti-CD3 monoclonal antibodies, and anti-CD25 antibody preparations. We ascertained maintenance immunosuppression on the basis of recorded use for maintenance therapy at transplantation discharge.

Incidence of MI on the Waiting List.

To provide context for the estimates of PTMI incidence, we estimated the incidence of first MI after wait-listing among transplant candidates who had Medicare as the primary payer and joined the waiting list in 1995 to 2000. We used Cox regression to estimate a probability distribution function adjusted to the average age, gender, race, and ethnic composition of the study sample that received a transplant. Participants for whom Medicare coverage began after the start of the waiting period were entered into the risk set by left truncation (17). Observations were censored at removal from the waiting list (including time of transplantation or death not due to MI), loss to follow-up, loss of Medicare, and end of study observation.

Recipient Factors.

Traditional risk calculators do not fully explain late ischemic heart disease risk in selected renal allograft recipients who are free of previous vascular events (22,23). In this study of MI risk from the time of transplantation, dominant risk factors aside from advanced age included pretransplantation MI and allograft failure. Nonetheless, traditional risk factors including male gender, diabetes, and dyslipidemia each were independently predictive of MI after transplantation, and smoking, obesity, and hypertensive heart disease [our surrogate for long-standing hypertension (24)] were unadjusted predictors. The time dependence and nested nature of some observed effects afford more detailed insight into the epidemiology of PTMI.

Although women faced a lower risk for PTMI than men, this relative protection diminished with time after transplantation for the study cohort as a whole. Taken with the absence of interactions between gender and age categories, this finding suggests that dissipation of the gender differential in general coronary heart disease risk may begin at a younger age after transplantation than in the general population (25,26). Diabetes without comorbid nephropathy was a relatively weak predictor of PTMI in this study, but risk increased to >50% above baseline in the substantial subset with diabetes as the cause of renal failure, and risk among all individuals with diabetes rose with time after transplantation. Although these effect sizes are lower than those of other recent reports of diabetes-related coronary syndrome risk after transplantation (5), this likely reflects our adjustment for a variety of independently predictive cardiovascular comorbidities that may be confounded with diabetes if not individually considered. Although we did not have information to adjust risk for pretransplantation diabetes duration, a diagnosis of diabetes as the cause of kidney failure is a reasonable surrogate for long-standing diabetes (27,28). Therefore, these patterns support increased risk for PTMI with longer duration of diabetes both before and after kidney transplantation. Whether more quantitative variable descriptions incorporating posttransplantation information would prevent the adjustment-related loss of predictive value of smoking, hypertension/hypertensive heart disease, and obesity was not discernible with the available data.

Race and ethnicity have been identified as important factors in the detection of cardiovascular events in the general population, but the direction of racial trends in diagnosis are inconsistent across studies (29,30). Our observation of reduced risk for PTMI among nonwhite individuals parallels the lower reported rates of coronary event hospitalizations among nonwhite dialysis patients (21). Further study is needed to determine whether these patterns reflect differential ascertainment, as from differential access to care, or true racial and ethnic risk variation.

The risk for PTMI among our study cohort was independently increased by detection of a variety of cardiovascular comorbidities before transplantation, including angina, peripheral vascular disease, and arrhythmias, and pretransplantation MI was a particularly strong predictor. In contrast, whereas Abbott et al. (5) found bivariate links between the risk for hospitalization for posttransplantation acute coronary syndromes and pretransplantation cardiovascular events, these associations attenuated to nonsignificance with covariate adjustment in their study. Possible explanations for such discrepancies include the interstudy differences in outcome measures and distinctions in the ascertainment of covariates themselves. In this work of Abbott et al. (5) and other studies (5,7,31), information on comorbidity is limited to that provided by CMS 2728, a data source with demonstrated specificity for cardiovascular diagnoses but low sensitivity as a result of underreporting and missing information (32). We combined CMS 2728 diagnoses with billing evidence of comorbid conditions, markedly increasing the frequencies of detected conditions and likely contributing to increased precision of condition-related effect estimates.

Donor- and Transplant-Related Factors.

We found that donor age, deceased source, delayed graft function, and allograft failure were independently predictive of PTMI. Donor hypertension may be a predisposing factor for cardiovascular disease after transplantation (10) and, given potential confounding of age and hypertension, may at least partially mediate the rise in PTMI risk with older donor age. The present study extends and confirms evidence of associations between renal failure and cardiovascular risk after transplantation (33), most specifically, corroborating the recent finding of Abbott et al. (5) that graft loss is a potent predictor of posttransplantation hospitalizations for coronary disease events. Although this association may indicate that factors that promote graft loss also promote PTMI, direct causality is plausible given consistent observation of azotemia as a precursor to increased cardiovascular risk in general (34) and MI risk specifically (2) in patients without transplants. Furthermore, because delayed graft function and graft failure are potentially modifiable complications, they are putative targets for PTMI risk reduction.

We also identified posttransplantation diabetes as a potentially modifiable risk factor for PTMI. Posttransplantation diabetes has been proposed to amplify cardiovascular risk after transplantation (13), but to our knowledge, ours is the first study to link this complication with cardiac events and with PTMI specifically. Framed in the context of recently reported associations of posttransplantation diabetes and all-cause death risk (14,18), our findings also suggest that PTMI may partially mediate the mortality excess observed after diagnosis of de novo diabetes in transplant recipients. Although posttransplantation diabetes maintained independent predictive value for death, this effect was weakened when considered with PTMI in our study. Of note, the lower cumulative incidence of posttransplantation diabetes in this study compared with that estimated by Kasiske et al. (18) from USRDS data sources using a similar method (15 versus 24% over 36 mo) is largely explained by sampling criteria. In contrast to Kasiske, we included patients who had pretransplantation diabetes and by definition are not at risk for the de novo complication and, in comprising nearly half of the sample, markedly dilute incidence percentages.

Several bivariate predictors of PTMI warrant mention. We observed higher PTMI risk among recipients who were CMV seropositive or received kidneys from seropositive donors compared with those with donor-negative/recipient-negative seropairings. Adjusted risk for the seropair with the highest risk for posttransplantation CMV disease (donor-positive/recipient- negative) approached but did not reach our chosen level of statistical significance (P = 0.03). In accordance with these unadjusted risk estimates, CMV infection has been linked with atherosclerotic risk among heart transplant recipients (35) and general patients without transplants (36). As approximately 15% of serologic pairing information was missing, failure to detect an independent association of CMV seropairing with PTMI may reflect misclassification error, which tends to reduce the estimated statistical significance of effects.

Increased time on dialysis has been associated with angiographic progression of coronary artery disease and noninvasive measures of coronary calcification in small cohorts (37,38). We found that although the unadjusted incidence of PTMI was lowest among patients who received a preemptive transplant and higher with dialysis duration beyond 1 yr compared with <1 yr, these associations were nonsignificant after covariate adjustment. In a study of late cardiac hospitalizations after transplantation that incorporated estimated renal function, Abbott et al. (7) found that pretransplantation dialysis duration independently predicted heart failure but not coronary syndrome events. These findings suggest that the impact of dialysis duration on PTMI risk may be accounted for, at least statistically, by the accumulation of comorbid conditions and/or overshadowed by the importance of allograft function.

The potential to modify posttransplantation cardiovascular risk with immunosuppressive regimens is actively debated. Whereas choice of immunosuppression has not been shown to reduce cardiovascular mortality, specific agents, including calcineurin inhibitor type, sirolimus, and corticosteroids, seem to have a variable impact on traditional cardiovascular risk mediators such as BP, serum lipids, and glycemic control (11,39). Our observation of a bivariate association between maintenance corticosteroid use and reduced risk for PTMI may seem surprising given the known adverse effects of steroids on cardiovascular risk factors, including hypertension, dyslipidemia, and glucose intolerance. Given the role of inflammation in cardiovascular risk (40), this association could reflect benefit via anti-inflammatory pathways. Alternatively, the association may be driven by selection factors, as patients who received a transplant without maintenance steroids between 1995 and 2000 represent a highly selected minority (<10% of the cohort) whose cardiovascular risk is unknown. As in the report of posttransplantation coronary syndrome hospitalizations (5), we did not detect independent associations between PTMI and maintenance or induction immunosuppressive agents analyzed on an intention-to-treat basis. Possible explanations for the lack of association include changes in immunosuppressive agents during follow-up, confounding by uncontrolled factors related to initial agent choice, or absence of differential net impact on PTMI risk perhaps as a result of an effective balance of distinct adverse effects.