Low Intraindividual Variability of Cyclosporin A Exposure

Reduces Chronic Rejection Incidence and Health Care Costs

BARRY D. KAHAN,* MARIA WELSH,* DIANA L. URBAUER,

MELINDA B. MOSHEIM,* KATHLEEN M. BEUSTERIEN,

MARTHA R. WOOD,

LINDA P. SCHOENBERG,* JOSEPH DICESARE,

STEPHEN M. KATZ,* and

CHARLES T. VAN BUREN*
*Division of Immunology and Organ Transplantation, Department of Surgery, University of Texas Houston
Health Science Center - Medical School and

Biometrics Consulting, Houston, Texas;

Covance Health
Economics and Outcomes Services, Inc., Washington, DC; and

Novartis Pharmaceuticals Corporation, East
Hanover, New Jersey.
Abstract. The present study applied a receiver operating char-
acteristic (ROC) analysis to assess the role of intraindividual
variability of cyclosporin A (CsA) drug exposure in predispos-
ing renal transplant recipients to the occurrence of chronic
rejection, as well as to increased health care costs using a
resource-based economic analysis. Two hundred and four adult
renal transplant recipients were treated with tapering doses of
prednisone (Pred) and with a concentration-controlled strategy
that selected doses of the olive oil-based formulations of CsA
(Sandimmune

) that achieved target concentrations based on

serial pharmacokinetic profiles. The ROC analysis revealed an
inflection point of plots of the coefficient of variation (%CV)
of CsA exposure versus the risk of chronic rejection at 28.4%
for the average concentration (C
av
), i.e., the dosing interval-
corrected area under the concentrationtime curves, and 36%
for the trough concentration (C
0
). The incidence of chronic
rejection over a period of 5 yr was 24% among the less variable
(LV) versus 40% among the variable (V) cohort. The economic
analysis revealed that the total mean facility and physician
costs per patient were $48,789 versus $60,998, respectively
(P 0.01). The degree of variability displayed by any indi-
vidual could only be predicted by serial measurements of CsA
concentrations, and not by demographic features, laboratory
determinations, clinical characteristics, individual or mean val-
ues of any observed CsA concentration, or other pharmacoki-
netic parameters calculated following a single drug exposure.
Thus, strategies that reduce intrapatient variability of CsA
exposure over time may lead to reductions in chronic allograft
loss and in treatment costs.
Oil-based liquid and gel capsule formulations of cyclosporin A
(CsA), the cornerstone for most immunosuppressive regimens
for the past 20 yr (1,2), display marked inter- and intraindi-
vidual variations in drug absorption, distribution, metabolism,
and elimination (3). These characteristics, combined with the
dangers inherent in either under- or overimmunosuppression,
result in the narrow therapeutic range that characterizes CsA as
a critical dose drug. Because the outcomes of CsA therapy do
not correlate predictably with patient characteristics, such as
body size or organ function, and because there are no known
intermediate end points to titrate drug dosage to patient re-
sponses, treatment with this agent has been routinely moni-
tored by serial estimates of trough concentrations (C
0
), i.e., the
amount of drug in whole blood samples collected immediately
before the next dose (4). This method, however, has only
partially compensated for the interindividual pharmacokinetic
differences, because C
0
concentrations show a weak correla-
tion with total drug exposure (5).
Similar to findings with other critical dose drugs, a pharma-
cokinetic approach using estimates of the area under the con-
centrationtime curve (AUC) or the corresponding dosing in-
terval-corrected value, C
av
(5), offers a more reliable indicator
of an individual patients proclivity toward inadequate immu-
nosuppression (6,7) versus nephrotoxicity (8). Thus, we
adopted a concentration-control strategy that individualizes
long-term CsA doses to maintain target C
av
values (9). Anal-
ysis of the utility of this approach to reduce the likelihood of
chronic rejection in 204 CsA-prednisone-treated renal trans-
plant recipients revealed a significant impact of the degree of
intraindividual variability of drug exposure over time post-
transplant (10). Although this analysis identified variability, it
did not determine the percent coefficient of variation that
reflected the greatest sensitivity with the least probability of a
false diagnosis of chronic rejection. Therefore, the purposes of
the present analysis were: (1) to extend the original database
for an additional 12 mo of clinical and pharmacokinetic follow-
up; (2) to use a receiver operating characteristic (ROC) anal-
ysis (11) to establish the inflection point on a plot of intrain-
Received March 5, 1999. Accepted October 2, 1999.
Correspondence to Dr. Barry D. Kahan, Division of Immunology and Organ
Transplantation, Department of Surgery, University of Texas Houston Health
Science Center - Medical School, 6431 Fannin, Suite 6.240, Houston, TX
77030. Phone: 713-500-7400; Fax: 713-500-0785; E-mail bkahan@
orgtx71.med.uth.tmc.edu
1046-6673/1106-1122
Journal of the American Society of Nephrology
Copyright 2000 by the American Society of Nephrology
J Am Soc Nephrol 11: 11221131, 2000
dividual coefficient of variation (%CV) of CsA exposure,
estimated based on either C
av
or C
0
values, versus occurrence
of chronic rejection; and (3) to determine whether the patient
cohorts with variable versus less variable behaviors show dif-
ferent health care costs.
Materials and Methods
Clinical Material
The cohort of 204 adult patients spanning 20 to 72 yr of age
received renal transplants under a dual-drug immunosuppressive reg-
imen of prednisone (Pred) administered in tapering doses, and the
olive oil-based liquid or corn oil-based gel capsule formulation of
CsA (Sandimmune

; Novartis, Basel, Switzerland) administered at

concentration-controlled doses based on C
av
values during the study
period from February 25, 1988 to July 27, 1994. Pred was tapered
from 120 mg on the day of surgery to 30 mg by day 7, 15 mg by day
90, and 10 mg by day 180 (12). None of the patients in this study
received either the microemulsion formulation of CsA (Neoral

;
Novartis) or triple-drug therapy with the nucleoside synthesis inhib-
itors azathioprine or mycophenolate mofetil. The median follow-up
time for patients who did not experience chronic rejection was 45.6
mo; the longest follow-up was 98.6 mo. Patients were withdrawn from
the study upon the occurrence of the end point of biopsy-proven
chronic rejection. The demographic features of each transplant recip-
ient were collated: age, gender, ethnicity, organ donor source (living
related/unrelated versus cadaveric), body mass index, panel-reactive
antibody, HLA-mismatch, diagnosis of diabetes, pretransplant trans-
fusions, donor gender and age, as well as first versus retransplant.
CsA Concentration Measurements
Whole blood samples for pharmacokinetic profiles were obtained
before (C
0
; trough) as well as 2, 4, 6, 10, 14, and 24 h after dosing for
patients treated with a once-daily CsA regimen, and before as well as
2, 4, 6, 8, 10, and 12 h after dosing for patients treated with a
twice-daily CsA regimen. Whole blood CsA concentrations were
estimated using a monoclonal selective antibody in the fluorescence
polarization immunoassay (TDx

; Abbott Laboratories, North Chi-

cago, IL) (13).
CsA Concentration-Control Regimen
Pretransplant pharmacokinetic studies were used to select starting
CsA doses, as described previously (14,15). In the initial posttrans-
plant period, CsA was delivered by continuous intravenous infusion
for 48 to 72 h at a dose calculated to produce a steady-state concen-
tration (C
ss
) of 400 50 ng/ml (16). Thereafter, the continuous
intravenous infusion CsA dose was tailored by linear extrapolation
based on the ratio of the observed-to-target C
ss
values. The infusion
was discontinued upon satisfactory absorption of a concomitant,
orally administered CsA dose, i.e., documentation that maximum
concentration (C
max
) minus C
ss
was greater than 700 ng/ml CsA.
Patients were initially assigned to a once- versus a twice-daily oral
dose regimen according to their CsA clearance rate, i.e., the ratio of
the intravenous dose (mg/kg per d) to the C
ss
(ng/ml) (17): Values
325 ml/min indicated a once-daily regimen (325), and those above
this rate a twice-daily regimen (325) (18). Thereafter, the dosing
interval was selected to maintain C
0
250 50 ng/ml, and the actual
CsA dose (in mg) was adjusted to maintain C
av
550 50 ng/ml for
the first posttransplant month. During the subsequent 2 mo, dosing
intervals were selected to maintain C
0
200 50 ng/ml, and twice
weekly pharmacokinetic profiles guided dose adjustments to achieve
C
av
500 50 ng/ml. From 3 to 6 mo posttransplant, monthly
pharmacokinetic profiles guided dose adjustments to achieve C
av

450 50 ng/ml and C
0
175 50 ng/ml. From 6 to 12 mo
posttransplant, alternate month pharmacokinetic profiles guided dose
adjustments to achieve C
av
400 50 ng/ml and C
0
150 50
ng/ml. Thereafter, pharmacokinetic profiles were performed every 3
to 6 mo to guide dose adjustments to maintain C
av
350 50 ng/ml
and C
0
100 50 ng/ml. Comparison of the target with the (mean
observed) and [one quartile range] of C
av
values at each posttransplant
interval showed reasonable application of the concentration-control
strategy: namely, 550 ng/ml (555.62) [128] during the first month;
500 ng/ml (504.70) [95] from months 1 to 3; 450 ng/ml (432.14) [85]
for months 4 to 6; 400 ng/ml (393.22) [72] for months 7 to 12; and 350
ng/ml (351.29) [65] for months 13 to 90, respectively (9). If an
adverse event occurred or if the CsA dose had to be adjusted, a
pharmacokinetic profile was performed after at least three (and usu-
ally seven) dosing intervals. C
0
independent of the pharmacokinetic
profiles were not measured in this protocol.
Pharmacokinetic Parameter Calculations
Whole blood steady-state CsA concentrationtime data were ana-
lyzed by standard noncompartmental methods (19). The data set
included a total of 4678 pharmacokinetic profiles from 204 pa-
tientsan increase of 793 profiles over our previous report (10). The
highest measured whole blood CsA concentration and the correspond-
ing sampling time were defined as C
max
and t
max
, respectively. The
drug concentrations at the beginning and at the end of the dosing
period were designated as C
0
, and C
12
or C
24
, respectively. The linear
trapezoidal rule was used to calculate the AUC from concentration
values within the dosing interval, and corrected to the C
av
by dosing
interval adjustment (AUC/, in hours). The initial absolute bioavail-
ability (F) was estimated by the dose-corrected AUC after oral versus
intravenous infusion (14). In addition to the mean (SD) and the
median, absolute, and dose-corrected values of the pharmacokinetic
parameters, the intrapatient %CV, defined as ([SD/mean] 100), was
calculated for each pharmacokinetic parameter. The mean numbers of
profiles were similar for patients stratified by the demographic char-
acteristics of age and race (data not shown).
Clinical Management and Diagnosis of Rejection
After the first 6 mo, patients were examined every 90 to 180 d in
the Transplant Center, depending on the exigency of other medical
complications. Each visit included a physicians assessment by history
and physical examination, as well as a complete blood count and
Sequential Multiple Analysis of 20 chemical constituents laboratory
test panel. Alternate visits also included a 24-h urinary protein deter-
mination. Histopathologic evidence of chronic rejection by renal
transplant biopsy was mandated in patients experiencing deterioration
of renal function as evidenced by an elevation of serum creatinine
30% above baseline, by proteinuria, and/or by progressive/persis-
tent hypertension refractory to two-agent antihypertensive therapy.
The diagnosis was always confirmed by the presence of histopatho-
logic features of obliterative vascular disease, including arterial and/or
arteriolar endothelial and smooth muscle changes, which were fre-
quently accompanied by glomerulopathy. Tubular atrophy and/or
interstitial fibrosis alone were not deemed sufficient conditions to
establish the diagnosis of chronic rejection.
Treatment Cost Calculations
Nurse reviewers examined the medical charts for 195 of the 204
patients in a blinded manner. The reviewers identified medical ser-
J Am Soc Nephrol 11: 11221131, 2000 Low Intraindividual Variability of CsA Exposure 1123
vices provided over the 5-yr period after the transplant procedure or
until the patient reached one of the following end points: switched to
another immunosuppressive regimen, retransplant, or death. The spe-
cific medical services abstracted included all hospitalizations, inpa-
tient procedures, outpatient visits, and procedures related to renal
transplant care. The average follow-up period for the economic anal-
ysis was 54 mo, and was comparable between groups.
Facility and physician services were assigned costs based on Medi-
care reimbursement rates in 1997. Inpatient facility costs were based
on abstracted Diagnosis Related Group (DRG) codes and correspond-
ing Medicare payment rates. Inpatient physician services were deter-
mined based on the abstracted International Classification of Diseases,
9th Revision, Clinical Modification (ICD-9-CM) procedure codes.
For each ICD-9-CM code, the appropriate procedure-based Current
Procedural Terminology (CPT) codes for both surgeons and anesthe-
siologists were selected (20). If a patient was admitted to the hospital
without undergoing major procedures, physician costs were assigned
according to the length of hospital stay. Hospital outpatient profes-
sional services also were identified by CPT codes. CPT codes were
matched to the corresponding relative value unit from the Resource-
Based Relative Value Scale (21), which was then used to estimate
physician costs based on the 1997 Medicare Fee Schedule for Hous-
ton, Texas (20).
Outpatient facility costs were estimated using the Ambulatory
Patient Group (APG) payment system and the corresponding Medi-
care reimbursement levels (22). This system is similar to the Ambu-
latory Payment Classification (APC) hospital outpatient department
payment system to be adopted by Medicare in the near future. The
costs of outpatient laboratory tests were estimated using the 1997
Clinical Laboratory Fee Schedule for Texas (23).
Statistical Analyses
To compare the occurrence of chronic rejection to the distribution
of demographic factors and clinical features among the 204 patients,
we used t tests for continuous variables (such as recipient age, donor
age, and dry weight) and Fisher exact tests for categorical variables
(such as race, gender, and donor source). For multiple laboratory
determinations, clinical parameters, and pharmacokinetic values (such
as hemoglobin, total protein, and number of antihypertensive medi-
cations), we used t tests to compare the mean values of the clinical
parameters of patients who did not versus those who did experience
chronic rejection during the observation period. The entire follow-up
period was subdivided into total time before the diagnosis of chronic
rejection, as well as subsets of mean values within the time intervals
of 3 to 6, 6 to 12, 12 to 24, and 24 to 36 mo after transplantation.
Logistic regression models were used to assess whether an indi-
vidual clinical parameter was associated with the occurrence of
chronic rejection, while controlling for the influence of demographic
factors or laboratory values that we have already demonstrated to be
related to the occurrence of chronic rejection in this patient cohort.
Backward elimination was then used to determine which factors/
values would remain in the model. After the pharmacokinetic vari-
ables that influenced the occurrence of chronic rejection were identi-
fied, ROC curves were constructed to depict the ability of individual
variables to predict the occurrence of chronic rejection (11,24). Each
ROC curve expressed the capacity for a clinical parameter to predict
rejection, taking into account both the accurate predictions (sensitiv-
ity, or true-positive rate) and inaccurate predictions of chronic rejec-
tion (false-positive rate). The time to chronic rejection was compared
between cohorts using a KaplanMeier analysis. A general linear
models procedure was used to estimate the variability of clinical
parameters over time using a repeated-measures ANOVA
(MANOVA), as well as univariate and multivariate analyses, to test
the hypothesis that variability neither decreased nor increased over
time (25). All analyses were performed using SAS version 6.12 on a
personal computer (26).
Costs were compared between groups using the MannWhitney
test, a nonparametric alternative to the t test. All statistical tests were
two-tailed and performed with P 0.05 as the upper limit of signif-
icance.
Results
Association of Demographic Factors and Chronic
Rejection
Table 1 shows the demographic characteristics that were
significantly different between the 71 (34%) patients who
experienced chronic rejection and the 133 (66%) who remained
free of this complication. During the additional 12 mo of
follow-up since the termination of the previous study (10), 14
of the 147 patients still at risk experienced the onset of chronic
rejection. The new data set confirmed the significant adverse
impact of three demographic factors shown in the previous
study (10) to increase the occurrence of chronic rejection:
African-American race (P 0.01), an acute rejection episode
(P 0.0001), and delayed graft function (P 0.006). In
addition, the present study revealed the adverse impact of
another factor on chronic rejection: drug-induced nephrotoxic-
ity (P 0.0001), which was defined as an increase of at least
25% over the baseline value of serum creatinine, without
evidence of rejection upon transplant biopsy but with reversal
upon CsA dose reduction. Although an initial univariate anal-
ysis suggested that chronic rejection was directly associated
Table 1. Demographic variables that significantly correlated
with the occurrence of chronic rejection
a
Characteristic
Chronic Rejection
P Value
b
No
(n 133)
Yes
(n 71)
Race 0.01
Caucasian 65 (49%) 31 (44%)
African-American 28 (21%) 29 (41%)
Hispanic 37 (28%) 11 (15%)
Asian 3 (2%) 0 (0%)
Acute rejection episode 0.0001
no 100 (75%) 33 (46%)
yes 33 (25%) 38 (54%)
Delayed graft function 0.006
no 118 (89%) 52 (73%)
yes 15 (11%) 19 (27%)
CsA-induced nephrotoxicity 17 (13%) 33 (46%) 0.0001
a
CsA, cyclosporin A.
b
Comparisons between values on each line in patients without
and with chronic rejection used the Fisher exact test for
homogeneity, comparing the proportion of patients for each
characteristic afflicted with chronic rejection versus the proportion
not affected (two-sided P value).
1124 Journal of the American Society of Nephrology J Am Soc Nephrol 11: 11221131, 2000
with dry weight (P 0.0135) and inversely with CsA dose per
day (P 0.0113), multivariate logistic regression models
showed that these factors were not significant. Clinical out-
come was not associated with the other demographic factors,
including donor or recipient age or gender, donor source,
repeat transplant, mode of previous dialysis treatment versus
preemptive transplant, diagnosis of diabetes mellitus, HLA
mismatch, or incidence of infection (data not shown).
Pharmacokinetic Values Associated with an Increased
Risk of Chronic Rejection
This study revealed an inverse association between the frac-
tion of patients free of the occurrence of chronic rejection and
the %CV values for C
av
or C
0
(Figure 1). There were strong
associations between the occurrence of chronic rejection and
the %CV both of observed and dose-corrected values for C
av
(P 0.002 and P 0.001, respectively) to a greater extent
than C
0
(P 0.004 and P 0.052), but only for the observed
%CV of C
max
(P 0.033) (Table 2). There was no statistically
significant difference among the overall mean (or immediately
precedent) values of the observed (or dose-corrected) trough
concentrations (C
0
or C
12/24
), C
max
, or C
av
between subjects
free of versus those afflicted with chronic rejection (data not
shown).
To determine the inflection point at which the %CV pro-
vided the most sensitive prediction of chronic rejection, we
performed ROC analyses. For a given %CV, the ordinate value
shows the percentage of patients with the diagnosis of chronic
rejection (true-positive results), and the abscissa value shows
the percentage of patients without evidence of chronic rejection
(false-positive results). Figure 2 shows that the ROC plot of
%CV C
av
has a lower inflection point, i.e., 28.4%, and includes
a larger area of predictive significance, i.e., 6301 units, than the
%CV C
0
, with an inflection point at 36% and an area of 5898
units. (An ROC analysis failed to show a significant predictive
effect of C
max
values [data not shown].) Thus, the analysis
Figure 1. Association of freedom from chronic rejection with lower values of the percent coefficient of variation (%CV) of cyclosporin A (CsA)
concentrations. The numbers of patients in each 5% value of CV for average concentration (C
av
) (Panel A) or trough concentration (C
0
)
(Panel B) are shown inside each bar. Although cohorts including one or two patients have been excluded from the figure, the actual number
of chronic rejectors per total number of patients in each of these cohorts for C
av
is 3 to 8% (1 of 1), 8 to 13% (1 of 2), 54 to 59% (0 of 1);
and for C
0
, 3 to 8% (1 of 1), 69 to 74% (1 of 2), and 117 to 122% (0 of 1).
Table 2. Association between %CV of pharmacokinetic
variables and the occurrence of chronic rejection
a
Parameter
Chronic Rejection
P Value
b
No Yes
C
0
mean 40.79 46.91 0.004
median 38.73 44.56
SD 14.89 12.89
range 14.79 to 118.92 25.00 to 81.99
C
max
mean 37.29 40.43 0.033
median 36.98 38.20
SD 9.62 10.66
range 12.47 to 62.34 22.38 to 77.35
C
av
mean 30.33 34.42 0.002
median 30.27 34.14
SD 8.38 9.22
range 2.98 to 57.85 11.03 to 61.12
C
0
/mg
mean 46.34 51.82 0.052
median 42.99 49.54
SD 19.23 18.90
range 9.45 to 143.00 22.07 to 126.99
C
av
/mg
mean 31.36 36.56 0.001
median 30.27 33.83
SD 10.19 11.59
range 11.55 to 66.41 16.32 to 73.94
a
%CV, percent coefficient of variation; C
0
, trough
concentration; C
max
, maximum concentration; C
av
, average
concentration.
b
P values determined by two-sided t test.
J Am Soc Nephrol 11: 11221131, 2000 Low Intraindividual Variability of CsA Exposure 1125
defined the variable (V) cohort as the group of patients with
%CV of C
av
28.4% and C
0
36%, and the less variable
(LV) group as those patients with values below the inflection
point (Figure 3).
The inflection points of %CV C
av
28.4% or %CV C
0

36% yielded similar values of sensitivity (77.5%) and speci-
ficity (39%) (Table 3). Furthermore, both C
0
and C
av
param-
eters showed higher negative predictive values (76.5 and
76.8%, respectively) than positive predictive values (40.4 and
40.7%, respectively). These findings suggest that a low coef-
ficient of variation is a better predictor of patients who will not
experience chronic rejection than a high coefficient is of those
who will develop this complication.
Time-to-Event Analysis
Although patients did not undergo protocol biopsies at stip-
ulated intervals posttransplant, the LV cohorts showed a longer
time than the V cohorts to the diagnosis of chronic rejection
(Figure 4). For variability of both C
av
and C
0
, the Kaplan
Meier analysis revealed significant differences between the
cohorts, i.e., P 0.001 and P 0.006, respectively.
Lack of Association between the Degree of
Pharmacokinetic Variability and Demographic Factors,
Clinical Characteristics, or Laboratory Values
The demographic, clinical, and laboratory values (data not
shown) were similar between members of the V and LV
cohorts, suggesting that it is unlikely that membership in each
cohort reflected comorbid conditions. The mean follow-up
periods in the LV and V cohorts were 3.49 and 3.58 yr,
respectively (P 0.73, NS). Furthermore, the mean number of
pharmacokinetic profiles per patient was only slightly greater
for the V group (23.86 8.24) than the LV group (21.1
7.64; P 0.018 by two-sample t test).
In addition, there was no correlation between the mean con-
Figure 2. Receiver operating characteristic (ROC) analysis of the correlation between %CV and the occurrence of chronic rejection. For a given
%CV (as indicated by the number juxtaposed to the broken line) of the C
av
(Panel A) or C
0
(Panel B), the ordinate values show the
corresponding true-positive rate (fraction of patients with that %CV who suffered from chronic rejection), and the abscissa values show the
corresponding false-positive rate (fraction of patients with that CV who did not suffer from chronic rejection). The inflection point (indicated
by the dot) was chosen as the optimal diagnostic value. The area between the ROC curve and the diagonal line is shown as a numerical value,
reflecting the degree to which a parameter shows a predictive benefit.
Figure 3. Percentage of patients experiencing chronic rejection strat-
ified by %CV. The inflection points of the ROC curves (Figure 2)
partitioned patients based on the %CV of C
av
or C
0
.
Table 3. Sensitivity/specificity analysis of %CV C
av
and
%CV C
0
values to predict the occurrence of
chronic rejection
a
Parameter (%)
%CV Values
C
av
C
0
Strict CV threshold 28.4 36.0
Sensitivity 77.5 77.5
Specificity 39.1 39.9
Positive predictive value 40.4 40.7
Negative predictive value 76.5 76.8
a
Abbreviations as in Table 2.
1126 Journal of the American Society of Nephrology J Am Soc Nephrol 11: 11221131, 2000
centrations or pharmacokinetic parameter values (C
0
, C
12/24
,
C
max
, or C
av
) among patients in the V versus LV cohorts for C
av
(Table 4) or C
0
(data not shown). Furthermore, there was no
relationship between the degree of posttransplant variability for
C
av
or C
0
and the absolute oral bioavailability or initial drug
clearance rate, as determined using paired intravenous and oral
administration of CsA in the early postoperative period (data not
shown).
Distribution of Variability in the Population
Frequency plots of the fraction of patients with individual
%CV values revealed that the LV cohort, defined as C
av

28.4% and C
0
36%, comprised only 33% of patients. Serial
comparisons revealed that the patients in the LV cohort showed
relatively constant %CV values over time (data not shown).
Similarly, examination of the %CV at various intervals post-
transplant confirmed that patients in the V cohort within the
first posttransplant year did not show a decrease in intrapatient
variability over time using repeated-measures ANOVA and
univariate (GreenhouseGeisser or HuynhFeldt) and multi-
variate (Wilks statistic) tests.
Health Economics
Table 5 shows the medical resource utilization over the
5-yr posttransplant period among the V and LV CsA bio-
availability groups. Fewer patients in the LV group were
rehospitalized compared to those in the V group (62%
versus 83%, P 0.05). The LV group had a mean of 2.5
rehospitalizations per patient, compared with 4.0 rehospi-
talizations per patient among the V group (P 0.05).
Furthermore, the mean length of stay for the transplant
hospitalization was shorter in the LV group compared to the
V group (8.3 versus 10.6 d, respectively; P 0.05; data not
shown). Compared to the V group, patients in the LV group
also received significantly fewer administrations of Solu-
Medrol antirejection therapy (30% versus 64%, P 0.05).
Eliminating from the analysis all rehospitalizations for non-
renal-related conditions (cardiovascular disorder, lung dis-
ease, fracture, etc.) did not diminish the differences between
the groups. Differences were also observed in outpatient
resource utilization. The LV group had a lower mean num-
ber of outpatient renal care visits compared to the V group
(17 versus 21, P 0.05) (Table 5). In addition, the mean
number of outpatient renal scan procedures per patient was
lower among members of the LV group compared to the V
group: Approximately 33% of the LV patients underwent at
least one renal scan, compared with 52% of the patients in
the V group (P 0.05) (Table 5). A total of 25 (36%)
patients in the LV group and 32 (25%) patients in the V
group reached an end point of switching immunosuppres-
sive regimens, retransplant, or death. In addition, seven
(10%) of the LV patients and six (5%) of the V patients were
lost to follow-up.
The differences observed in medial resource utilization be-
tween the V and the LV groups were reflected in renal care
costs. Over the entire period, mean per-patient facility costs for
all renal-related rehospitalizations for the LV and V groups
were $11,788 and $23,391, respectively (P 0.01; data not
shown). Mean per-patient outpatient facility costs also were
lower for the LV group ($2,823) than for the V group ($3,541;
P 0.05). Overall, the total inpatient and outpatient health
care costs, including both facility and physician costs and
including the initial hospitalization, for the LV and V groups
were $48,789 (median $37,322) and $60,998 (median
$49,646), respectively (P 0.01) (Figure 5).
Figure 4. Time-to-chronic rejection for patient cohorts stratified as variable (V) versus less variable (LV). Time-to-event plots based on %CV
of C
av
(Panel A) and C
0
(Panel B) for the LV () versus the V (O) cohorts. KaplanMeier analysis revealed for Panel A, P 0.001, and
for Panel B, P 0.006.
Table 4. Mean concentration values of all pharmacokinetic
profiles among the 204 patients stratified by %CV
of C
av
a
Mean Concentration
(ng/ml)
Chronic Rejection
No Yes
28.4 28.4 28.4 28.4
C
0
236.8 235.1 237.4 227.3
C
12/24
221.1 214.4 225.5 208.4
C
max
1035.9 975.0 962.5 929.6
C
av
456.1 444.7 449.6 423.7
a
None of the differences was statistically significant by t test
(two-sided P values, all 0.2).
J Am Soc Nephrol 11: 11221131, 2000 Low Intraindividual Variability of CsA Exposure 1127
Discussion
The present study suggests an adverse impact of intraindi-
vidual variability of CsA exposure upon both the incidence of
chronic rejection and the costs of health care among 204 renal
transplant recipients followed for 5 yr. An ROC analysis was
used to assign patients to LV versus V cohorts, based on their
%CV of C
av
or C
0
CsA concentrations, measured by pharma-
cokinetic profiles performed every 6 mo. Patients selected to
be in the LV cohorts of %CV C
av
or %CV C
0
displayed a
significantly longer time to the occurrence of chronic rejection
and lower overall health care costs. The 33% of renal trans-
plant recipients who were members of the LV cohort could not,
unfortunately, be discriminated from members of the V cohort
based on any demographic, clinical, or laboratory characteris-
tic, but only by serial pharmacokinetic profiling.
Low pharmacokinetic variability may confer relatively con-
stant CsA immunosuppressive exposure. Indeed, the high de-
grees of intraindividual variability documented with many
widely used drugs, such as dihydropyridine calcium channel
blockers, have also been associated with adverse impacts,
particularly among individuals afflicted with illnesses in which
a lack of efficacy has dire clinical consequences, i.e., the
chemotherapeutic treatment for human immunodeficiency dis-
ease.
In addition to finding a role of variability, the present study
detected an association between the diagnosis of nephrotoxic-
ity and the occurrence of chronic rejection. This association,
which had been suspected but not shown previously, may be
explained in at least two ways. This phenomenon may repre-
sent an error in diagnosis, wherein the nephrotoxic event in
fact indicates a progressive albeit subclinical rejection or a
nonimmunologic nephropathic process. In those cases, CsA
dose reduction may improve renal function, presumably by
ameliorating the drug-induced nephrotoxic component of the
overall injury. Alternatively, the drug-induced nephrotoxic in-
jury may lower the intrinsic resistance of the allograft to a
subsequent chronic rejection process. Although drug-induced
renal dysfunction represents a major indication for CsA dose
reduction, it appears likely that this factor alone would predis-
pose to chronic rejection, unless the CsA dose is drastically
lowered beyond a certain amount (27). Indeed, during any
Table 5. Medical resource utilization during 5-yr posttransplant period
Type of Medical Care Less Variable (n 69) Variable (n 126)
Rehospitalizations
no. of patients rehospitalized, n (%) 43 (62) 105 (83)
a
no. of rehospitalizations per patient (mean SD) 2.5 2.5 4.0 3.5
a
Rehospitalizations, excluding non-renal-related
b
no. of patients rehospitalized, n (%) 26 (37) 84 (67)
a
no. of rehospitalizations per patient (mean SD) 1.8 1.7 2.5 2.4
a
Antirejection therapy administration, n (%)
Solu-Medrol 21 (30) 81 (64)
a
OKT3 6 (9) 17 (13)
ATG 6 (9) 15 (12)
Outpatient renal care visits (mean SD)
c
17 10 21 12
a
Outpatient procedures (mean SD)
renal biopsy 0.09 0.3 0.05 0.2
ultrasound 0.43 0.5 0.56 0.5
renal scan 0.33 0.5 0.52 0.5
a
chest x-ray 0.39 0.5 0.47 0.5
a
P 0.05 for difference between groups. Other differences were not significantly different.
b
Rehospitalizations attributable to nonrenal conditions (e.g., cardiovascular disorders, lung disease, etc.).
c
Any visit that included CsA monitoring and/or renal procedures.
Figure 5. Comparison of the total mean 5-yr costs for renal care
among patients with variable and less variable pharmacokinetic pat-
terns of CsA exposure. The hatched areas indicate facility costs, and
the black-shaded areas indicate physician costs, in U.S. dollars. *P
0.05 for the difference in outpatient costs between groups; **P 0.01
for the difference in inpatient costs between groups.
1128 Journal of the American Society of Nephrology J Am Soc Nephrol 11: 11221131, 2000
given time interval, there was no significant difference be-
tween the mean observed or dose-corrected C
av
values, or
between the median CsA doses administered to the groups of
putatively nephrotoxic patients afflicted with (381.45 mg/d;
n 71) versus those free of (337.16 mg/d; n 133) chronic
rejection. The only significant association other than intraindi-
vidual variability with the diagnosis of nephrotoxicity was the
dose-corrected C
max
(P 0.039; data not shown), a finding
that confirms our previous observation (7).
The association between the inflection points reflecting low
intraindividual variability of drug exposure (%CV C
av

28.4% and %CV C
0
36%) was more robust for its negative
than for its positive predictive value for the diagnosis of
chronic rejection. The limited capacity of a high level of
variability to predict patients who would experience chronic
graft failure may be the result of preeminent and inconsistent
risk factors that overpower the biopharmaceutic effect of vari-
ability, including CsA-resistant induction of B cell antibody
production, preexistent donor kidney injury and/or limited re-
nal mass (28), or independent pharmacodynamic variabilities
of the efficacy of drug effect. For example, Batiuk et al. (29)
observed that CsA produces incomplete degrees of and inter-
individual differences in inhibition of calcineurin activity, the
putative target of drug action. We plan to study the association
between estimates of kinetic variability and calcineurin activity
in the same manner that Vozeh et al. (30) documented a linear
relation between theophylline concentrations and lung func-
tions in asthmatic patients. The planned studies may help
clarify the association between pharmacokinetic parameters
and dynamic slopes or maximal effects of CsA at its cal-
cineurin target.
The finding that %CV C
0
offers a more useful (albeit less
sensitive) measure of variability than %CV C
av
extends earlier
findings of an association between a high degree of trough
concentration variability in the early posttransplant period and
acute rejection episodes in renal (31) and heart and lung (32)
transplant recipients. Furthermore, Savoldi et al. (33) found
that among a cohort of 157 renal transplant recipients, patients
with a %CV of C
0
below the median value of 31% showed a
significantly greater incidence of functioning allografts than
did patients with higher variability (mean period of 7 2.3 yr).
The present analysis extends these findings by identifying
36% as the inflection point for C
0
.
It seems more likely that variable oral absorption of CsA,
rather than drug clearance rates, explains the present findings,
since our previous studies failed to document significant
changes in CsA clearance rates over time posttransplant in the
absence of concomitant drug therapy altering cytochrome P450
3A4 activity (7). This hypothesis is consistent with the obser-
vations of Sanathanan and Peck (34): Variability of absorption
of a variety of drugs enhances the effects of pharmacokinetic
variation. However, it will be important to combine pharma-
cokinetic profiling with erythromycin breath tests (35) to ex-
clude the influence of changes in hepatic disposition on drug
variability.
All patients were concentration-controlled based on adjust-
ment of CsA doses to achieve similar levels of exposure, i.e.,
C
av
350 50 ng/ml, based on pharmacokinetic profiles. It
is impossible to ascertain whether a different target C
av
value
would reduce the risk of patients experiencing chronic rejec-
tion versus nephrotoxicity, and indeed whether patients main-
tained at this C
av
would show less impact of variability on
outcome. Furthermore, the inflection points of ROC curves
may vary, depending on the concentration targets, the patterns
of patient care, and the precision of the concentration moni-
toring programs either to obtain precisely timed C
0
samples or
to perform pharmacokinetic profiles for C
av
values. Therefore,
transplant centers should perform their own ROC analyses to
examine the impact of %CV; the findings may vary according
to not only patient characteristics but also the immunosuppres-
sive regimen.
One might propose several explanations for the occurrence
and biologic implications of intraindividual pharmacokinetic
variability. First, because it persists over a long time period and
is not associated with age, variability is unlikely to reflect
maturational effects akin to those observed in pediatric and
adolescent populations (36). Second, it seems unlikely that low
variability is merely a reflection of better patient compliance to
the immunosuppressive regimen. Patients in the V cohort nei-
ther admitted nor seemed to display evidence of noncompli-
ance to a greater degree than did their LV counterparts (data
not shown). Indeed, patients in the V cohort underwent outpa-
tient follow-up a significantly greater number of times within
each time interval than patients in the LV cohort (Table 5).
Third, the long-term persistence of the pharmacokinetic vari-
ability suggests that it is unrelated to recovery of the impaired
gastrointestinal function associated with chronic renal failure.
Thus, one can only speculate that variability is due to episodic
absorptive variations caused by coadministered over-the-
counter medications and/or a variety of foods in the diet.
Unfortunately, there are no reliable, quantitative, and clinically
relevant surrogate techniques to evaluate intestinal factors that
might predispose patients to variable drug absorption.
At least two strategies may be envisioned to overcome the
adverse impact of high degrees of CsA variability described
herein. On one hand, combination therapy of CsA with addi-
tional agents (mycophenolate mofetil [(37)] and/or sirolimus
[(38)]) may potentiate the immunosuppressive effects and pos-
sibly exert direct actions to mitigate smooth muscle cell pro-
liferationa pathognomonic feature of chronic rejection. A
3-yr study is under way to compare the effects of addition of
sirolimus to reduce the incidence of chronic rejection among
CsA-treated patients. On the other hand, an improved biophar-
maceutical formulation of CsA might increase the proportion
of patients in the LV group. The new microemulsion CsA
formulation Neoral

(39) seems to afford more consistent

absorption, at least during the first 12 mo, i.e., a mean %CV of
18% compared to 36% for the oil-based formulation (B. Ka-
han, unpublished data). Indeed, potential cost benefits associ-
ated with the microemulsion formulation have been shown in
a retrospective medical chart review of patients at Canadian
centers participating in a multinational, randomized clinical
trial comparing Neoral

to Sandimmune

(40).
The economic evaluation only sought to examine the corre-
J Am Soc Nephrol 11: 11221131, 2000 Low Intraindividual Variability of CsA Exposure 1129
lation between pharmacokinetic variability and costs derived
from the perspective of the transplant service. If a broader
perspective was adopted and costs for other renal-related ser-
vices, such as dialysis, were incorporated, it is likely that costs
for these services would be disproportionately higher in the V
group versus the LV group because a higher proportion of
patients in the former group experienced chronic rejection
(39% versus 22%, respectively) and likely required institution
of dialysis. Thus, the differences in costs incurred within the
two groups may have been higher if a broader perspective were
evaluated. Future studies should use a decision analysis that
determines the costs associated with true-positive, false-posi-
tive, true-negative, and false-negative results of the estimates
of interindividual variability to assess the value of serial phar-
macokinetic profiling to effectively predict the occurrence of
chronic rejection and to test whether the present values for
variability are optimal.
This study provides a longitudinal view of the impact of
pharmacokinetic variability on chronic rejection incidence, as
well as on inpatient and outpatient medical resource utilization,
over the first 5 yr after renal transplantation. Members of the
LV cohort displayed a reduced risk of chronic rejection and
incurred significantly lower treatment costs. The findings sug-
gest that more consistent drug absorption among the renal
transplant recipient population may improve long-term out-
comes and result in substantial cost advantages.
Acknowledgments
This work was supported by Grant NIDDK 38106-11 from the
National Institute of Diabetes and Digestive and Kidney Diseases. The
health economic analysis was supported by an independent contract to
Covance from Novartis Pharmaceuticals Corp.
References
1. Borel JF: Pharmacology of cyclosporin. IV. Pharmacological
properties in vivo. Pharmacol Rev 41: 259371, 1990
2. Kahan BD. Cyclosporine. N Engl J Med 321: 17251738, 1989
3. Kahan BD: Overview: Individualization of cyclosporine therapy
using pharmacokinetic and pharmacodynamic parameters.
Transplantation 40: 457476, 1985
4. Bowers LD: Therapeutic monitoring for cyclosporine: Difficul-
ties in establishing a therapeutic window. Clin Biochem 24:
8187, 1991
5. Kahan BD, Grevel J: Overview: Optimization of cyclosporine
therapy in renal transplantation by a pharmacokinetic strategy.
Transplantation 46: 631644, 1988
6. Kasiske BL, Heim-Duthoy K, Rao KV, Awni WM: The relation
between cyclosporine pharmacokinetic parameters and subse-
quent acute rejection in renal transplant recipients. Transplanta-
tion 46: 716722, 1988
7. Lindholm A, Kahan BD: Influence of cyclosporine pharmacoki-
netics, trough concentrations, and AUC monitoring on outcome
after kidney transplantation. Clin Pharmacol Ther 54: 205218,
1993
8. Meyer MM, Munar M, Udeaja J, Bennett W: Efficacy of area
under the curve monitoring in renal transplantation. J Am Soc
Nephrol 4: 13061315, 1993
9. Kahan BD, Welsh M, Rutzky LP: Challenges in cyclosporine
therapy: The role of therapeutic drug monitoring by area under
the curve monitoring. Ther Drug Monit 17: 621624, 1995
10. Kahan BD, Welsh M, Schoenberg L, Rutzky LP, Katz SM,
Urbauer DL: Variable oral absorption of cyclosporine: A biop-
harmaceutical risk factor for chronic renal allograft rejection.
Transplantation 62: 599606, 1996
11. Beck JR, Schultz EK: The use of relative operating characteristic
(ROC) curves in test performance evaluation. Arch Pathol Lab
Med 110: 1320, 1986
12. Browne B, Kahan BD: Renal transplantation. In: Surgical Clinics
of North America: Horizons in Transplantation, edited by Kahan
BD, Philadelphia, Saunders, 1994, pp 10971116
13. Lindholm A, Napoli K, Rutzky L, Kahan BD: Specific mono-
clonal radioimmunoassay and fluorescence polarization immu-
noassay for trough concentration and area-under-the-curve mon-
itoring of cyclosporine in renal transplantation. Ther Drug Monit
14: 292300, 1992
14. Kahan BD, Welsh M, Rutzky L, Lewis R, Knight R, Katz S: The
ability of pretransplant test-dose pharmacokinetic profiles to
reduce early adverse events after renal transplantation. Trans-
plantation 53: 345351, 1992
15. Chueh SC, Kahan BD: Pretransplant test-dose pharmacokinetic
profiles: Cyclosporine microemulsion versus corn oil-based soft
gel capsule formulation. J Am Soc Nephrol 9: 297304, 1998
16. Dunn J, Grevel J, Napoli K, Lewis RM, Van Buren CT, Kahan
BD: The impact of steady-state cyclosporine concentrations on
renal allograft outcome. Transplantation 49: 3034, 1990
17. Gibaldi M: Biopharmaceutics and Clinical Pharmacokinetics.
Philadelphia, Lea and Febiger, 1991
18. Lindholm A, Welsh M, Rutzky L, Kahan BD: The adverse
impact of high cyclosporine clearance rates on the incidences of
acute rejection and graft loss. Transplantation 55: 985993, 1993
19. Benet LZ, Galeazzi RL: Noncompartmental determination of the
steady-state volume of distribution. J Pharmacol Sci 68: 1071
1074, 1979
20. Health Care Financing Administration: Federal Register 61:
5955659716, November 1996
21. IRP Systems: IRP ICD-9-CM to CPT-4 Cross-Walk of Database
as of October 1, 1997, IRP Systems, Inc., Wilmington, MA,
19841997
22. Ambulatory Patient Groups Definitions Manual, version 2.0,
Wallingford, CT, 3M Health Information Systems, 1997
23. Health Care Financing Administration: Clinical Diagnostic Lab-
oratory Fee Schedule, National/Carrier, 1996
24. Metz CE: Basic principles of ROC analysis. Semin Nucl Med 8:
283298, 1978
25. Agresti A: Categorical Data Analysis, New York, John Wiley
and Sons, 1990
26. Elandt-Johnson RC, Johnson NL: Survival Models and Data
Analysis, New York, John Wiley and Sons, 1990
27. Burke JF Jr, Pirsch JD, Ramos EL, Salomon DR, Stablein DM,
Van Buren DH: Long-term efficacy and safety of cyclosporine in
renal transplant recipients. N Engl J Med 331: 358363, 1994
28. Tillius SG, Tilney NL: Both alloantigen-dependent and -inde-
pendent factors influence chronic allograft rejection. Transplan-
tation 59: 313318, 1995
29. Batiuk TD, Feldzgeritta P, Halloran PF: Calcineurin activity is
only partially inhibited in leukocytes of cyclosporine-treated
patients. Transplantation 59: 14001404, 1995
30. Vozeh S, Kewitz G, Perruchoud A, Tschan M, Kopp C, Heitz M:
Theophylline serum concentration and therapeutic effect in se-
vere acute bronchial obstruction: The optimal use of intrave-
1130 Journal of the American Society of Nephrology J Am Soc Nephrol 11: 11221131, 2000
nously administered aminophylline. Am Rev Respir Dis 125:
181184, 1982
31. Holt DW, Marsden JT, Johnston A, Bewick M, Taube DH:
Blood cyclosporine concentrations and renal allograft dysfunc-
tion. Br Med J 293: 10571059, 1986
32. Best NG, Trull AK, Tan KK, Hue KL: Blood cyclosporin concen-
trations and the short-term risk of lung rejection following heart-
lung transplantation. Br J Clin Pharmacol 34: 513520, 1992
33. Savoldi S, Maiorca R, Maderna M, Chiappini R, Corbetta G:
Low intrapatient variability of blood cyclosporine levels is cor-
related with excellent graft survival. Transplant Proc 29: 288
289, 1997
34. Sanathanan LP, Peck CC: The randomized concentration-con-
trolled trial: An evaluation of its sample size efficiency. Con-
trolled Clin Trials 12: 780794, 1991
35. Watkins PB, Murray SA, Winkelman LG, Heuman DM, Wrigh-
ton SA, Guzelian PS: Erythromycin breath test as an assay of
glucocorticoid-inducible liver cytochromes P-450: Studies in rats
and patients. J Clin Invest 83: 688697, 1989
36. Rodman JH: Pharmacokinetic variability in the adolescent: Im-
plications of body size and organ function for dosage regimen
design. J Adolesc Health 15: 654662, 1994
37. Sollinger HW: Mycophenolate mofetil for the prevention of
acute rejection in primary cadaveric renal allograft recipients.
U.S. Renal Transplant Mycophenolate Mofetil Study Group.
Transplantation 60: 225232, 1995
38. Kahan BD, Podbielski J, Napoli KL, Katz SM, Meier-Kriesche
HU, Van Buren CT: Immunosuppressive effects and safety of a
sirolimus/cyclosporine combination regimen for renal transplan-
tation. Transplantation 66: 10401046, 1998
39. Kahan BD, Dunn J, Fitts C, Van Buren D, Wombolt D, Pollak R:
Reduced inter- and intrasubject variability in cyclosporine phar-
macokinetics in renal transplant recipients treated with a micro-
emulsion formulation in conjunction with fasting, low-fat meals,
or high-fat meals. Transplantation 59: 505511, 1995
40. Kingma I, Ludwin D, Dandavino R: Economic analysis of Ne-
oral in de novo renal transplant patients in Canada. Clin Trans-
plant 11: 4248, 1997
J Am Soc Nephrol 11: 11221131, 2000 Low Intraindividual Variability of CsA Exposure 1131