Abstract
ABSTRACT. Anemia has long been known to be a complication of end-stage renal disease (ESRD), and it has been linked to cardiovascular morbidity and mortality. Although kidney transplant recipients (KTR) are prone to experiencing cardiovascular outcomes, little is known about the epidemiology of anemia in this population. With few exceptions, studies to date have not fully evaluated the associations between posttransplant anemia (PTA) and medications commonly used in KTR, particularly immunosuppressant drugs, angiotensin-converting enzyme inhibitors (ACEI) and angiotensin II receptor blockers (ARB). The authors aimed to specifically investigate possible associations between these drugs and PTA. Detailed medical information was retrospectively collected on 374 consecutive KTR from our transplant clinic. Univariate/multivariate linear regression models were used to test for associations between hematocrit (HCT) and other covariates, and logistic regression models were used to detect independent predictors of PTA, defined as HCT <33%. The mean time since transplantation was 7.7 yr, and mean creatinine was 2.2 mg/dl. The prevalence of PTA was 28.6%. Ten percent of all patients were on erythropoietin therapy, but only 41.6% of patients whose HCT was <30 received this treatment. From multivariate analyses, the authors found that female gender and lower renal function were associated with lower HCT (both P < 0.001). Patients on ACEI had significantly lower HCT (P = 0.005) compared with patients without such treatment. In addition, a significant curvilinear dose-response relationship was found between ACEI dose and HCT. Among the immunosuppressant drugs, mycophenolate mofetil (P = 0.05) and tacrolimus (P = 0.02) were associated with a lower HCT. The authors conclude that PTA is prevalent and undertreated in KTR. Several medications that are possibly modifiable correlates of PTR deserve further study.
Anemia, as a complication of end-stage renal disease (ESRD) is well studied (1). Reports indicate that anemia in this patient population is associated with numerous morbidities, including cardiovascular complications such as cardiac enlargement, left ventricular hypertrophy, congestive heart failure, and angina (2–9⇓⇓⇓⇓⇓⇓⇓). Furthermore, effective treatment of anemia has been demonstrated to decrease morbidity and improve survival (10–16⇓⇓⇓⇓⇓⇓). The known effect of anemia on cardiovascular disease in the ESRD population would suggest that anemia may be one of the major factors that explains the burden of cardiovascular disease in the kidney transplant recipient (KTR) population. Systematic investigation into the prevalence of posttransplant anemia (PTA) is therefore of critical importance.
Contrary to the ample data available regarding anemia in the ESRD population, much less is known about the epidemiology of PTA, and only few studies have systematically investigated this issue (17–21⇓⇓⇓⇓). Although these studies have contributed to our understanding of the prevalence of PTA, most of them did not systematically collect information on important prescription drug classes that are possibly associated with anemia, did not conduct rigorously controlled multivariate analyses, or were limited by small sample size for such analyses.
The present study aimed to establish the prevalence of PTA in a large US single-center transplant population and to describe the utilization patterns of recombinant human erythropoietin (rh-Epo) use in patients with PTA. As the centerpiece of this study, however, we sought to explore possible associations between PTA and frequently used prescription drugs in that population, specifically immunosuppressant agents, angiotensin-converting enzyme inhibitors (ACEI) and angiotensin II receptor blockers (ARB), using multivariate regression techniques that allowed us to rigorously control for confounding. In absence of such analyses in previous studies, we also sought to describe a possible dose-response relationship among ACEI, ARB, and PTA.
Materials and Methods
Study Population
For this study, we retrospectively reviewed the outpatient medical records of 374 patients with a functioning renal transplant who were transplanted at our hospital and received routine follow-up care at our transplant clinic between September 2000 and September 2001. We retrieved the following information on every patient from these charts: age (to be used as continuous variable), gender (dichotomous), time since transplantation (continuous), donor type (categorical: cadaveric, living), and the likely cause of native kidney disease (categorical: diabetes, hypertension, glomerulonephritis, reflux nephropathy, polycystic kidney disease, other). From laboratory studies, we obtained each patient’s hematocrit, mean corpuscular volume (MCV) (continuous), and serum-creatinine level (categorical: <1.5 mg/dl, 1.5 to 2 mg/dl, >2 to 3 mg/dl, >3 mg/dl). We extracted information on several medications, such as prescription, generic substance, and dose of ACEI, ARB, and prescription of several immunosuppressants: cyclosporine, azathioprine, rapamycine, mycophenolate mofetil (MMF), tacrolimus, and corticosteroids (all dichotomous). We also evaluated whether a patient received treatment with erythropoietin (rh-Epo).
Statistical Analyses
For data analyses, we used two separate approaches. The primary set of analyses used hematocrit values (HCT, in %) as the main outcome variable, which was found to be normally distributed in this population (skewness, −0.06; kurtosis, −0.08; Shapiro-Wilk test: P = 0.91). We used simple linear regression for univariate analyses and multiple linear regression to test for independent associations between HCT and all other variables. A priori, we decided to include all covariates in the multivariate models. The results reflect change in hematocrit, and 95% confidence intervals (CI) are given to describe the likely range of the true parameter. We tested the inherent assumptions of the linear regression approach by testing the residuals for normality, examining partial regression residual plots for all variables, and using standardized residuals analysis. The condition number was used to detect problems with multicollinearity in the full regression model. We also calculated the adjusted R-square as a measure for model fit.
In a second set of analyses, we modeled the likelihood of patients being anemic (defined as HCT < 33%) using univariate and multivariate logistic regression. Having model parsimony in mind, we forced age, gender, and level of kidney function into the multivariate model and then used an automated backwards elimination procedure to delete factors at P-value > 0.10 from the model one at a time. We then manually added each remaining variable into the model to ensure that other variables that were above the prespecified significance threshold did not confound the results. Goodness of fit was evaluated using the Hosmer and Lemeshaw test.
In an effort to describe a possible dose-response relationship between ACEI use or angiotensin II inhibitor use, we attempted to normalize the doses of the various compounds within each drug class to relative equipotence using the following conversion factors: we assumed that daily doses of 10 mg each of enalapril, fosinopril, lisinopril, and quinapril, 150 mg of captopril, and 5 mg of ramipril were equivalent in the ACEI class. Similarly, 50 mg of losartan, 150 mg of irbesartan, and 160 mg of valsartan were assumed equal in the ARB class. To explore dose-response relationships, we first tested categories of normalized dose for both ACEI and ARB (normalized dose < 1; normalized dose ≥ 1; reference group: no use). Normalized ACEI dose, and ARB dose were then each used as linear terms in the regression models. Additionally, squared terms were also calculated to help detect a nonlinear dose-response relationship.
We used SAS for Windows, release 8.2 (The SAS Corporation, Inc., Cary, NC) for all statistical analyses. The study was approved by the institutional review board of our hospital.
Results
The study population included 198 men (52.9%) and 176 women (47.1%) who had been transplanted for 7.7 yr on average. The mean age was 49 yr, and living versus cadaveric organs were represented equally among those whose donor type could be determined. Other relevant characteristics can be found in Table 1. The mean HCT among the 374 transplant carriers was 36 ± 6; 51.9% of patients had a HCT > 36, 19.5% between 33 and 36, 14.4% between 30 and 33, and 14.2% had a HCT < 30 (Figure 1). Erythropoietin was given in 10% of patients overall. However, even among those with a HCT < 30, only 41.5% of patients were on rh-Epo treatment (Figure 1).
Table 1. Patient Characteristics (n = 374)
Figure 1. Distribution of hematocrit and frequency of erythropoietin use (n = 374).
Two thirds of patients (63.4%) received an immunosuppressive regimen that contained both corticosteroids and cyclosporine A. The frequency of use of all individual immunosuppressant drugs can be gleaned from Table 1. ACEI were prescribed in 29.7% of patients, and 6.4% received ARB.
Several variables were significantly associated with HCT values in univariate analyses. On average, men had a HCT value that was 1.85 higher than women. Hematocrit was further associated with kidney function, compared with patients whose creatinine was < 1.5 mg/dl, patients with a serum creatinine of 1.5 to 2 mg/dl, 2 to 3 mg/dl, or >3 mg/dl were found to have a HCT value that was −1.24, −3.71, and −6.67 lower, respectively (Table 2). Among the medications studied and using patients with corticosteroids and cyclosporine A as the reference group, we found HCT values to be significantly lower in patients on ACEI (−1.9) and those on rapamycine (−4.9) compared with those patients without the respective drug. Patients whose regimen was corticosteroid-free had a HCT value that was lower by −2.4 than those on corticosteroid therapy.
Table 2. Determinants of hematocrit levels in 374 kidney transplant recipients
Next, we built a multivariate model including all variables to test for independent associations with HCT values (Table 2). Male gender remained a significant predictor, being associated with a HCT value that was 2.9 higher than in women (95% CI, 1.94 to 3.88; P < 0.001). Kidney function was also found to be independently associated with HCT. Compared with patients with creatinine < 1.5 mg/ml, those whose creatinine was 1.5 to 2 mg/dl, 2 to 3 mg/dl, and >3 mg/dl had a HCT value that was lower by −1.47 (95% CI, −2.75 to −0.19; P = 0.02), −3.95 (95% CI, −5.26 to −2.64; P < 0.001), and −7.25 (95% CI, −8.80 to −5.70; P < 0.001), respectively. Among the immunosuppressant drugs, we found that both MMF (−1.34; 95% CI, −2.65 to −0.02; P = 0.05) and tacrolimus (−2.32; 95% CI, −4.23 to −0.41; P = 0.02) were each associated with lower HCT values. ACEI therapy was independently associated with a −1.62 lower HCT value compared with no such therapy (95% CI, −2.71 to −0.54; P = 0.003). A similar effect was indicated for ARB therapy, but it did not reach statistical significance. All other covariates were not found to be independently associated with HCT values.
When introducing categories of normalized ACEI doses, we found indications for a dose-response relationship: compared with those patients who did not take an ACEI, patients who received an ACEI at a normalized dose < 1 had a −0.86 lower HCT value (P = 0.35), and patients who received a normalized dose of ≥ 1 had a −1.33 lower HCT value (P = 0.03). The findings for categories of ARB normalized dose were NS. We next introduced the normalized doses of ACEI and of ARB into the models as continuous variables, but we did not find a linear dose-response relationship of ACEI or ARB dose and HCT values. However, when adding a square term for dose into the model we found both the linear and the square ACEI dose term to be significant. Figure 2 illustrates the estimated dose-response curve between ACEI dose and HCT value. This finding is compatible with the concept of a diminishing marginal association between ACEI dose and HCT; our data suggest that each extra dose of ACEI dose is associated with a smaller decrement in HCT value. No violations of the assumptions inherent in the linear regression approach or problems with collinearity were detected. The adjusted R-square of the full model was 0.29.
Figure 2. Estimated dose-response curve between normalized angiotensin-converting enzyme inhibitor (ACEI) dose and hematocrit level. An ACEI dose equivalent corresponds to 10 mg each of enalapril, fosinopril, lisinopril, and quinapril, 150 mg of captopril, and 5 mg of ramipril. Dose-response estimation derived from a linear and a squared ACEI dose variable in the full multivariate linear regression model (both P < 0.001).
Finally, we analyzed whether the likelihood of being anemic (i.e., having a HCT < 33%) was associated with ACEI or ARB using logistic regression. From univariate analyses, we found that both drugs were associated with an increased likelihood of anemia; individuals on ACEI had 80% higher odds of being anemic compared with patients who did not receive this drug (odds ratio [OR], 1.80; 95% CI, 1.10 to 2.93). Similarly, there was some indication that patients on ARB had more than twice the odds of being anemic compared with patients who did not receive those medications (OR, 2.28; 95% CI, 0.96 to 5.41). From multivariate logistic regression models that adjusted for gender, level of kidney function, and receipt of tacrolimus (as yielded by the backwards selection process), we found that ACEI use was independently associated with a HCT < 33 (OR, 1.86; 95% CI, 1.04 to 3.35), whereas ARB were not (OR, 2.07; 95% CI, 0.89 to 3.81). These results remained robust in several sensitivity analyses; neither individual entry of the remaining variables into the model nor exclusion of those patients who received rh-Epo treatment changed the results substantially. The assumption of good model fit was not rejected (Hosmer and Lemeshaw test: P = 0.27).
Discussion
Although the issue of anemia after kidney transplantation has received increasing attention of late, the data on the exact prevalence of PTA are limited. We found that PTA (HCT < 33) is a common condition found in KTR, identifiable in 28.6% of this study population. The current study is the largest one evaluating PTA in US KTR. Other studies have found a similar prevalence of PTA, namely between 20 and 39.7% (17–21⇓⇓⇓⇓). The recent study by Mix et al. (21) found that the prevalence of anemia (HCT < 36) was 76% at transplantation and fell to 21% 1 yr posttransplant. However, 4 yr after transplantation the prevalence had risen to 36%. The proportion of patients in our study below this threshold (HCT < 36) was 48.1%. The different prevalences of anemia found in these studies is likely related to between-study differences in the definition of anemia. These studies have variably defined anemia as HCT < 33, HCT < 36, hemoglobin < 12/13 g/dl, and hemoglobin < 11.5 g/dl. To have a clinically relevant definition of anemia that could potentially affect patient care, we chose to define anemia as HCT < 33 in accordance with DOQI guidelines for the target HCT range of 33 to 36 (20).
Despite the high prevalence of anemia in KTR in our study, the use of rh-Epo was exceptionally low (10% of all patients). Particularly the observation that only 41.5% of patients with a HCT < 30 received rh-Epo was surprising. However, this finding is in accordance with previous reports by Yorgin et al. (19) as well as Vanrenterghem et al. (20), and it confirms their observations of rh-Epo therapy in only 8% and 5.2% of their study populations, respectively. Numerous studies in the CKD and ESRD populations have demonstrated improvement in quality of life and morbidity and mortality with treatment of anemia (10–16⇓⇓⇓⇓⇓⇓), which leads to the conclusion that management of anemia in KTR falls through the cracks of routine follow-up care in these patients.
With the notable exception of the European survey study (20), previous studies have only provided limited accounts of the possible associations between immunosuppressant drugs and PTA. In our study of 374 patients, we evaluated the role of six individual immunosuppressants on HCT and found MMF and tacrolimus to be independently associated with lower HCT. Both drugs, however, had been implicated in development of PTA in earlier studies (22,23⇓). Although more clearly defined for MMF, it is likely that the lower HCT seen in patients on these agents is related to bone marrow suppression caused by these drugs. Our analysis failed to reveal azathioprine as a risk factor for lower HCT, which is at odds with other previous reports that have reported such an association (18), but it confirms findings from the European survey study (20). Interestingly, the latter study did find an association between MMF and azathioprine and the presence of anemia. Unfortunately, the authors combined MMF and azathioprine in their analyses, which assumes an equal effect of those two drugs. Our findings suggest that this assumption may not be appropriate.
In addition to immunosuppressant agents, other medications that have been implicated in the development of PTA include ACEI. In fact, the erythrotoxic effect of ACEI has been used in KTR as treatment of choice for posttransplant erythrocytosis. In addition to posttransplant erythrocytosis, ACEI and ARB have been shown to have beneficial effects in a variety of disease states, including congestive heart failure, hypertensive nephrosclerosis, and diabetic nephropathy and proteinuric nondiabetic nephropathy (22,23⇓). A role for ACEI and/or ARB has also been suggested in the management of chronic allograft nephropathy; in fact, both ACEI and ARB were shown to be efficacious in prolonging graft survival in a rat kidney transplant model (24). However, concern over the side effect profile of these agents, including graft dysfunction, hyperkalemia, and anemia, has limited their use in KTR. We found a negative association between ACEI use and PTA even after multivariate adjustment for possible confounders. A similar trend was observed for ARB, but the statistical power to detect an independent effect was limited because only 6.4% of our study patients were prescribed an ARB. The literature on the association between ACEI or ARB and PTA is mixed. Vanrenterghem et al. (20) found an association between ACEI/ARB use and the likelihood of anemia, but it failed to detect an association between these drugs and hemoglobin levels. Unfortunately, ACEI and ARB were not analyzed separately in that study; rather, they were collapsed into a single categorical variable. Two other North American studies also failed to detect an independent association between ACEI and PTA (21,24⇓). Of note, our study had a markedly greater proportion of patients on ACEI or ARB than either of these studies. Other possible explanations for this discrepancy may include differences in mean time posttransplant or renal function. In the studies above, the mean renal function was better (S-Creat, 1.8 mg/dl) than in our population (S-Creat, 2.2 mg/dl), suggesting possible effect modification; the positive association between ACE/ARB use and lower HCT found in our study might have been due to a more pronounced effect of ACE/ARB on HCT in the setting of more advanced renal insufficiency. However, we failed to detect such effect modification between S-Creat and ACEI use when testing this hypothesis formally. Adding credence to the notion that ACEI may be associated with PTA is our finding of a curvilinear dose-effect relation between normalized ACEI dose and HCT. This type of dose-effect relationship is compatible with the concept of a diminishing marginal association between ACEI/ARB dose and HCT, suggesting that each extra dose of ACEI/ARB is associated with a smaller decrement in hematocrit.
There are certain limitations to our study, the most important one being its cross-sectional design. We are therefore unable to establish causality of the associations found. It is further possible that other confounding factors are operational, such as inflammation or impaired iron stores or utilization. Unfortunately, parameters on metabolic availability of iron, parathyroid hormone levels, or inflammation markers were unavailable in most patients. Furthermore, ethnicity was not recorded for many patients and was therefore unavailable for study. Lastly, the unanswered question whether anemia is of predictive value in KTR looms large. The strength of the study is its large sample size combined with a rigorous multivariate analytical approach, which allowed us to control better for confounding than was possible in the earlier studies to this topic.
In summary, this largest epidemiologic study of anemia in US KTR confirms previous observations that PTA is common and undertreated. Consistent with suggestions from previous studies, our data show that certain immunosuppressants, namely MMF and tacrolimus, are independently associated with PTA. Our novel findings of a dose-dependent association between ACEI use and PTA and a similar trend regarding ARB raise important questions for future study. Is lower HCT of prognostic value in KTR at all, and are therapeutic interventions to raise a low HCT (e.g., iron supplementation, rh-Epo therapy) efficacious in improving patient of graft survival? Are ACEI or ARB efficacious in improving patient or graft survival to outweigh the possible negative effects of a lower HCT? Do ACEI or ARB contribute to hyporesponsiveness to rh-Epo therapy? Considering the high prevalence of PTA in KTR and the present findings, further study in this area is clearly warranted.
- © 2004 American Society of Nephrology