Adequate Early Cyclosporin Exposure is Critical to Prevent Renal Allograft Rejection: Patients Monitored by Absorption Profiling



This study used receiver operating characteristic analysis to investigate the properties of area under the concentration-time curve during the first 4 h after cyclosporin-microemulsion dosing (AUC0−4) and cyclosporin (CyA) levels immediately before and at 2 and 3 h after dosing (C0, C2 and C3) to predict the risk of biopsy-proven acute rejection (AR) at 6 months. Ninety-eight kidney transplant recipients treated with CyA-microemulsion-based triple therapy immunosuppression were studied on post-transplant days 3, 5, and 7, and at increasing intervals thereafter.

The most sensitive and specific predictor of AR was AUC0−4. Of the single time-point measurements, the measurement properties of C2 were closest to those of AUC0−4, and superior to those of C3. The relationship between C0 and subsequent AR was weak and did not reach statistical significance. On day 3, CyA AUC0−4≥ 4400 ng . h/mL and C2≥ 1700 ng/mL were each associated with a 92% negative predictive value for rejection in the first 6 months. Pharmacokinetic measurements on or after day 5, and measurements on day 3 in patients with delayed graft function, were not predictive of AR.

Adequate exposure within the first 3 days post transplantation may be critically important in preventing subsequent rejection.


The importance of therapeutic drug monitoring for cyclosporin (CyA) has long been recognised, as the therapeutic window in which adequate T-cell immunosuppression can be achieved with a minimum risk of adverse effects is narrow. The importance of the relationship between pharmacokinetic monitoring and outcome came to prominence in the early 1990s, when studies by Kahan and colleagues (1) and subsequently Schroeder et al. (2) found a strong correlation between clinical end-points in renal transplantation and a number of pharmacokinetic measures. In particular, the area under the concentration-time curve (AUC) for CyA was shown to be a sensitive predictor for acute rejection (AR) and graft survival at 1 year. Further studies revealed that both exposure and the degree of day-to-day variability in exposure were predictive of the incidence of acute (3) and chronic (4) rejection in renal transplant recipients.

Although AUC measurements have long been established as clinically predictive means of assessing CyA exposure, they are not considered practicable by most transplant centers. Indeed, most centers continue to routinely assess exposure by measuring trough levels of CyA, i.e. at the end of the 12-h dosing period, immediately before the next dose is taken (C0). As the C0 measurement has been shown to be a poor predictor of clinical outcome (1,2), efforts to establish simpler approaches to CyA monitoring involving fewer measurements than a full AUC have led to the development of sparse sampling algorithms and abbreviated AUCs, which are in clinical use in some centers (5–8).

This research was facilitated by the introduction of the CyA microemulsion formulation of CyA which has improved absorption characteristics, provides more consistent and predictable pharmacokinetics than the original Sandimmune formulation, and has been associated with some improvement in clinical outcomes (9–11). A series of studies (12–16) investigating the link between CyA exposure with CyA microemulsion and clinical outcomes has focused attention on the absorption phase of drug exposure: i.e. the first 4 h following dosing (AUC0−4). This parameter has been demonstrated in renal allograft recipients to be predictive of both full AUC and clinical outcomes, including the incidence of both AR and nephrotoxicity (15,17). The significance of the absorption phase to therapeutic drug monitoring has led to the concept of CyA-microemulsion absorption profiling (CyA-microemulsion AP) monitoring (16,18), which aims to provide clinically applicable methods for the assessment of drug exposure during this pharmacokinetic phase.

The simplest expression of the CyA-microemulsion-AP paradigm would be the identification of a single time point at which a measurement of CyA blood concentration is associated with overall drug exposure and clinical outcome. Within the absorption phase, peak blood levels of CyA are reached at approximately 2 h, and hence C2 measurements have been a major focus of interest. In heart and liver transplant patients, C2 has been found to correlate closely with AUC0−4, and to be associated with clinical outcome (19–21). In renal transplantation, C3 has also demonstrated a closer relationship with AUC0−4 and clinical outcomes than C0 (22). Whether C2 or C3 is more closely correlated with AUC0−4, and the relationship between each of these parameters, C0 and clinical outcome, needs to be further addressed (15).

In this study we investigated the utility of AUC0−4, C0, C2 and C3 measurements, using receiver operating characteristic (ROC) curves to assess the sensitivity and specificity of the four pharmacokinetic parameters in the prediction of clinical outcome. These findings suggested target levels of CyA exposure for routine therapeutic drug monitoring.


The study population comprised 98 consecutive primary kidney transplant recipients transplanted between August 1998 and December 1999 at the Queen Elizabeth II Health Sciences Centre in Halifax. All patients were followed for a minimum of 6 months.

Collaboration with in-patient and out-patient nursing staff and with the drug monitoring laboratory facilitated timely blood sampling and same-day AUC reporting. Patients gave informed consent for renal transplantation and for immunosuppression. The method of adjusting CyA dosing based on AUC0−4 had been adopted in our center as a clinical practice standard on the basis of earlier evidence and experience with this strategy.


The diagnosis of AR was confirmed by percutaneous core needle biopsy, and classified according to the Banff criteria (16,17).

Delayed graft function (DGF) was defined as absence of > 10% spontaneous drop in serum creatinine by the third postoperative day, regardless of the need for dialysis.

CyA was assessed by whole blood concentration, determined by a monoclonal, parent-compound-specific radioimmunoassay (RIA; DiaSorin CYCLO-Trac™ whole blood RIA kit, Stillwater, Minnesota, MN; cross-reactivity with CyA metabolites < 2.0%, 4.1–10.9% coefficient of variation in our laboratory, linear from 25 to 1200 ng/mL). Sirolimus was measured by high performance liquid chromatography (HPLC) with ultraviolet detection by a reference laboratory (Covance Laboratories, Indianapolis, Indiana, IN; coefficient of variation 2.6–13.0%, linear from 2.5 to 150.0 ng/mL).

Immunosuppression was as follows:

  • 1Prednisone, CyA microemulsion, mycophenolate mofetil (MMF): 54 patients, 55%
  • 2Prednisone, CyA microemulsion, and either everolimus (40-O-[2-hydroxyethyl]-rapamycin) or MMF (as part of a randomized double-blinded study): 17 patients, 17%
  • 3Prednisone and CyA microemulsion alone: 11 patients, 11%
  • 4Prednisone, CyA microemulsion, sirolimus (as part of an open-labeled multicenter study): eight patients, 8%
  • 5Prednisone, CyA microemulsion, MMF and either basiliximab, humanized anti-interleukin 2 receptor antibody or placebo (as part of a randomized double-blinded study): six patients, 6%
  • 6Prednisone, CyA microemulsion, MMF and antileukocyte function antibody (ANTILFA) or placebo (as part of a randomized double-blinded study): two patients, 2%

CyA microemulsion was started at 9–14 mg/kg/day in two separate doses. During the study period and based on accumulating data, the starting dose of CyA microemulsion was progressively increased so as to enlarge the proportion of patients exceeding the 4400 ng . h/mL AUC0−4 target threshold (previously defined [15]) within the first 5 days post-transplant. AUC0−4 was performed on days 3, 5, 7, 10, and 14, and at weeks 3, 4, 6, 8 and 12. Blood for CyA levels was drawn before (C0) and at 1, 2, 3, and 4 h after the morning dose of CyA microemulsion (C1, C2, C3, and C4, respectively). AUC0−4 was determined by the following regression formula derived in our retrospective study database (15), using only C1, C2 and C3: AUC0−4 = 256 + C1 + 0.9 0 × C2 + 1.4 × C3. The cyclosporin microemulsion dose was adjusted to achieve an AUC0−4 target range of 4400–5500 ng . h/mL (15), regardless of C0, throughout the first 6 months post transplant. Assuming a linear relationship between dose and systemic exposure for CyA microemulsion within the same patient, dose adjustments were made by multiplying the ratio of mid-range target AUC0−4 (5000 ng . h/mL) to each patient's current AUC0−4 by their previous CyA microemulsion dose to obtain the new dose (new dose = previous dose × 5000/current AUC0−4).

Methylprednisolone 500 mg i.v. was given intraoperatively followed by prednisone 1 mg/kg p.o. daily, and tapered by 5 mg every other day to 20 mg daily for the first month, 15 mg daily for the 2nd month and 10 mg daily for the 3rd month. In MMF-treated patients, MMF 1 g p.o. bid was started as soon as oral medication could be tolerated and was adjusted to maintain WBC > 4.0 × 109/L. Twelve patients were part of a randomized double-blind study and were assigned to receive either MMF 1 g p.o., everolimus 0.75 mg p.o., or everolimus 1.5 mg post-operative p.o. bid. In sirolimus-treated patients, sirolimus was given as a 6-mg loading dose on the first postoperative day followed by sirolimus 2 mg p.o. daily, and then adjusted to keep sirolimus trough blood levels above 5 ng/mL.


The primary outcome was biopsy-proven AR at six months.


Student's t-test was used to compare means of the continuous variables. Fisher's exact test was used to compare the incidence of AR between the different groups. p < 0.05 (two-tailed) was considered significant throughout. Statistical calculations were performed with SPSS 9.0 (SPSS Inc, Chicago, IL).

Receiver operating characteristic curve analysis

Receiver operating characteristic curves depict the relationship between sensitivity and specificity for a continuous test variable (with multiple possible cut-points), such as a therapeutic drug monitoring parameter, in the prediction of a dichotomous outcome variable, such as AR. In this study, ROC curves were used to examine the relationship between different indices of CyA pharmacokinetics and the clinical end-point of AR at 6 months.

The interpretation of ROC curves is as follows: if a test were to predict an outcome perfectly (i.e. with 100% sensitivity and 100% specificity) the curve would follow the y-axis vertically and then proceed horizontally along the top border of the plot. In contrast, for a test in which the sensitivity and specificity proved no better than chance, the curve would follow a 45° line, reflecting the fact that as the cut-point changes in the direction of increased sensitivity, specificity decreases proportionately. Quantitatively, a test's usefulness is given by the area under the ROC curve; the area under a perfect ROC curve is 1.0, whereas a ROC curve for a test that is no better than chance has an area of 0.5. Standard measures of statistical significance and confidence intervals can be deployed to determine if the area under any given ROC curve significantly exceeds 0.5.

For any cut-point in a continuous variable, a two-by-two table can be constructed showing the number of cases above and below the cut-point, with and without the outcome. These data can be used to determine the positive and negative predictive values of the test. As a prespecified analysis, we sought to determine the C2 and C3 thresholds that were associated with the same negative predictive value (i.e. the probability of remaining rejection free if above the threshold) as achieved by AUC0−4 ≥ 4400 ng/h/mL in this dataset.

For the ROC analyses examining the ability of each pharmacokinetic parameter to predict subsequent AR, the following subgroup analyses were prespecified: by presence or absence of DGF; by use of diltiazem; and by cadaveric vs. live donor transplantation.


Baseline characteristics

The demographic and clinical characteristics of the 98 patients are summarised in Table 1. The mean CyA microemulsion dose at day 3 was 11.7 ± 2.0 mg/kg/day.

Table 1. Demographic and clinical characteristics at baseline
 Mean ± SD (%)
Age (y)44 ± 11
Diabetes mellitus21
On diltiazem43

Clinical outcomes

Delayed graft function occurred in 12% of patients, and 15% of patients had at least one AR episode in the first 6 months after transplantation.

Relationship of CyA concentration at individual time points to AUC0−4

Table 2 summarises the standard pharmacokinetic parameters C0 and AUC0−4 on days 3, 5 and 7 post-transplant. While mean C0 levels remained stable, there was a trend of increasing mean AUC0−4 levels over the first week as a result of a combination of dose adjustment and absorption improvement.

Table 2. Cyclosporin trough levels and AUC0−4 on days 3, 5, and 7 after transplantation
Mean ± SD
Mean ± SD
Mean ± SD
  1. AUC0−4= area under the concentration-time curve during the first 4 h after cyclosporin-microemulsion dosing

Day 3457 ± 1975593 ± 177811.7 ± 2.06.8–21.5
Day 5432 ± 1405567 ± 164411.2 ± 4.06.3–26.1
Day 7447 ± 1456008 ± 146410.8 ± 4.15.0–26.1

Pearson correlations between AUC0−4, C0, C2, and C3 using data from day 3 are shown in Table 3. C0 correlated poorly with AUC0−4, whereas of the other two single time-point measurements, C2 provided a more accurate prediction of AUC0−4 than C3. On day 3, the correlation between C2 and AUC0−4 was 0.89 (r2 = 0.79, p < 0.0001), and that of C3 with AUC0−4 was 0.83 (r2 = 0.69, p < 0.0001). Similar results were obtained using data from post-transplant days 5–14. However, the relationship between C2 and AUC0−4 over time was more consistent (r2 ≥ 0.71) than that between C3 and AUC0−4 (r2 ≥ 0.46).

Table 3. Relationship between AUC0−4, C0, C2, and C3 on day 3
  1. AUC0−4= area under the concentration-time curve during the first 4 h after cyclosporin-microemulsion dosing

AUC0−4Pearson correlation1.0000.3190.8900.829
p (2-tailed) 0.0020.0000.000
C0Pearson correlation 1.0000.0720.282
p (2-tailed)  0.5060.008
C2Pearson correlation  1.0000.659
p (2-tailed)   0.000
C3Pearson correlation   1.000
p (2-tailed)    

Relationship of AUC0−4 and single time-point measurements to clinical outcome

The ROC analysis for AUC0−4, C0, C2, and C3 in the prediction of AR is presented in Figure 1 and summarised in Table 4. Of the four assessment methods, AUC0−4 was the most highly predictive of subsequent rejection, with an area under the ROC curve of 0.758 (Figure 1). With the dosing regime described, 72 patients (73%) had an AUC0−4 greater than 4400 ng . h/mL by day 3. Of these patients, only six (8.3%) subsequently experienced AR, whereas nine of the 26 patients below this threshold (34.6%) had an AR episode. This yields a relative risk for AR of 4.2 (95% confidence interval [CI] 1.6–10.5) for patients below this threshold level of AUC0−4 (p = 0.0031). The negative predictive value of AUC0−4 greater than 4400 ng . h/mL at day 3 is 92% (95% CI 81–97%) (Table 5).

Figure 1: ROC analysis: prediction of subsequent acute rejection: sensitivity and specificity of AUC0–4, C0, C2 and C3 at day 3.

Figure 1: ROC analysis: prediction of subsequent acute rejection: sensitivity and specificity of AUC0–4, C0, C2 and C3 at day 3.

Table 4. Receiver operating characteristic (ROC) analysis: area under ROC curve examining the properties of AUC0−4, C0, C2, and C3 at day 3 in the prediction of subsequent acute rejection
  1. AUC0−4= area under the concentration-time curve during the first 4 h after cyclosporin-microemulsion dosing

Table 5. Two-by-two table showing relationship between reaching AUC0−4 target on day 3 and subsequent 6-month acute rejection
AUC0−4 day 3Acute rejectionNoTotal
YesPredictive value
  • AUC0−4= area under the concentration-time curve during the first 4 h after cyclosporin-microemulsion dosing

  • 1

    The probability of experiencing rejection if below threshold

  • 2

    The probability of remaining rejection free if above threshold

< 4400 ngh/mL 91726Positive1 = 9/26 × 100 = 34%
≥ 4400 ngh/mL 66672Negative2 = 66/72 × 100 = 92%
Sensitivity 9/15 × 100 = 60%   
Specificity66/83 × 100 = 80%   

Of the single-measurement time points at day 3 (Figure 1), the greatest sensitivity and specificity in predicting AR was achieved by C2 measurement, with an area under the ROC curve of 0.744. By contrast, C0 measurement was the least predictive with an area of 0.657, not conventionally statistically significant, whereas C3 was marginally less predictive than C2 with an area of 0.700. A threshold C2 CyA level of 1700 ng/mL on day 3 yields a 92% negative predictive value for AR (Table 6), and was achieved in 63% of patients (p = 0.15 for comparison with 73% of patients above target by AUC0−4).

Table 6. Thresholds for 92% negative predictive values, and proportion of patients in whom the threshold was achieved or exceeded
 Threshold for 92%
negative predictive
Proportion of patients
above threshold on day 3
(%, 95%CI)
  1. AUC0−4= area under the concentration-time curve during the first 4 h after cyclosporin-microemulsion dosing

  2. 1 Not calculated for C 0 because no statistically significant relationship between variables

AUC0−44400 ng/h/mL73 (64–82)
C21700 ng/mL63 (52–73)
C31550 ng/mL61 (50–71)

Although measurements of AUC0−4 and C2 at day 3 were predictive of subsequent AR, measurements at day 5 showed only a weak trend towards prediction, and measures at day 7 were not predictive (Table 7).

Table 7. Receiver operating characteristic (ROC) analysis: area under ROC curve examining the properties of AUC0−4 on days 3, 5, and 7 in the prediction of subsequent acute rejection
  1. AUC0−4= area under the concentration-time curve during the first 4 h after cyclosporin-microemulsion dosing

Day 30.7580.613–0.9030.002
Day 50.6040.386–0.8220.274
Day 70.5500.373–0.7260.568

Subgroup analysis

A total of 12 patients in the population of 98 patients had DGF. In this group there was no relationship between any pharmacokinetic measurement and subsequent AR. In the 86 patients without DGF, predictive properties for AUC0−4 were improved (ROC area 0.872, 95% confidence intervals 0.791–0.953) compared with the population as a whole (ROC area 0.758, 95% confidence intervals 0.613–0.903), but this difference was not statistically significant. Comparable results in subgroup analysis were observed for C2 measurements (data not shown). For both AUC0−4 and C2, there was a trend towards stronger predictive properties in live donor than cadaveric transplants (including patients with DGF), but this did not reach statistical significance. Use of diltiazem did not affect the predictive value of AUC0−4 and C2 measurements.

Serum creatinine

At 6-month follow up, serum creatinine in patients with day-3 AUC0−4 ≥ 4400 ng . h/mL was 154 ± 63 µmol/L, not significantly different from 176 ± 45 µmol/L in those below the AUC target (p = 0.19). A similar trend, which also did not reach statistical significance, was observed for serum creatinine in patients with day-3 C2 ≥ 1700 ng/mL compared with those below this threshold (154 ± 62 µmol/L vs. 168 ± 55 µmol/L, respectively, p = 0.36).


Variability in the absorption profile of CyA has been reduced by the introduction of CyA microemulsion, but appropriate monitoring remains an important issue, given the drug's narrow therapeutic window (23). C0 monitoring is still widely used in clinical practice despite the poor relationship between this measure and AUC, AR and CyA nephrotoxicity (24). Other indices of CyA exposure (formal AUC0−4, and AUC0-12) have been previously shown to predict the risk of AR, and these data have created awareness of the potential importance of pharmacokinetic studies in the management of patients treated with CyA (16,18).

In retrospective (15) and more recently in prospective studies (17) we have previously shown that formal AUC0−4 accurately predicts the incidence of AR in renal transplant patients treated with CyA microemulsion. In this study we have documented the predictive utility of an abbreviated method of estimating AUC0−4 (based on C1, C2 and C3 only), which should be more feasible in clinical practice than the formal 5-point AUC0−4. We have further shown that a single measurement at C2 also provides useful prognostic information in the prediction of AR. Of the three single time-point measurements assessed in this study (C0, C2 and C3), C2 was confirmed as having the strongest predictive ability, at least during the early post-transplant period. This is in keeping with the characteristic Tmax of CyA microemulsion at approximately 2 h, and is consistent with theoretical considerations which suggest that the maximum concentration of CyA achieved is likely to be predictive of immunosuppressive potency (25).

On the basis of our results, we have set specific day 3 target thresholds for AUC0−4 and C2, which if achieved are associated with a 6-month risk of AR of less than 10%. The target thresholds are ≥ 4400 ng/h/mL for AUC and ≥ 1700 ng/mL for CyA levels at C2. The negative predictive value of an AUC0−4 that exceeds this threshold at day 3 is approximately 92%. A threshold C2 CyA level of 1700 ng/mL on day 3 also yields a 92% negative predictive value for AR. In other words, for patients above these thresholds, the risk of AR in the first 6 months after transplantation was 8% in this cohort in which CyA exposure monitoring was maintained.

Cyclosporin exposure, measured by AUC0−4 or C2, at days 5 or 7 was not predictive of the AR risk. Though the power to detect a difference at these later time points was lower than for day 3 data because of missing data (e.g. at day 5 the power to detect a 15% difference in rejection rate was only 40%), the trend towards predictive abilities at these time points was very weak, in contrast with the highly statistically significant result at day 3 (Table 7). This suggests that achieving an adequate level of CyA exposure by day 3 after transplantation may be critical in the prevention of AR. If both the critical importance of early adequate immunosuppression in predicting the longer-term immunologic outcome and the prognostic importance of monitoring during the period we observed in this heterogeneous population are confirmed in other studies, it will be important to also examine the effect of different immunosuppressive regimes on the predictive properties of the thresholds. Because of the limited sample size, and in particular because of the low number of rejection events in this series, multivariate analysis was not possible in the current study. We therefore were unable to examine the interaction between the immunosuppressive regime and CyA levels in these data. Whether CyA thresholds differ for different immunosuppressive regimes or whether adequate early exposure to any one drug is sufficient to reduce the risk of subsequent rejection remains an important outstanding question.

These threshold levels were not predictive in patients with DGF. Though this may be a reflection of the small numbers of patients in this category, it is also possible that this reflects the significantly greater risk of AR these patients face independently of their exposure to CyA, and corresponds with findings in previous studies (15). The trend to improved predictive properties of CyA exposure measurements in live donor compared with cadaveric transplants is likely accounted for by the absence of DGF in this live donor group. Use of diltiazem did not impact upon the predictive value of our day 3 threshold levels for AUC0−4 and C2. That these observations are in keeping with theoretical constructs validates the robustness of the cut-off points defined by this study.

In the absence of evidence from randomized trials of different strategies, these and other observational studies have emphasised the importance of the absorption phase of CyA microemulsion's pharmacokinetic profile in predicting overall CyA exposure and clinical outcomes. Early achievement of adequate exposure levels is associated with reduced risk of subsequent AR. In previous work we had defined an upper limit for the AUC0−4 range of 5500 ng . h/mL in the prevention of nephrotoxicity (15). An upper limit for C2 monitoring based on nephrotoxicity data is not currently available, but mean 6-month serum creatinine levels in those with early exposure above target was 154 ± 62 µmol/L, compared with 168 ± 55 µmol/L (p = 0.36) in those with levels below target, suggesting that adverse effects in the form of nephrotoxicity are unlikely to be marked. In this cohort of patients, AUC0−4 targets ≥ 4400 ng . mL were maintained throughout the first 6 months. Progressive reduction of target AUC0−4 over this time period, analogous to the reduction in C0 targets in conventional monitoring protocols, would be expected to result in lower 6-month serum creatinine values than those observed in this cohort. Given the critical importance of early exposure in the prevention of rejection in our data, it is likely that this could be achieved without an adverse impact in terms of rejection.

In renal, cardiac and hepatic transplantation, the superiority of absorption-phase profiling compared with standard trough level CyA monitoring has been shown (18). This study confirms the validity of the approach in renal transplantation, and identifies C2 measurement as an effective method of assessing rejection risk. The single-point C2 measurement is clinically applicable and places no more strain on the resources of the transplant unit than does conventional C0 measurement, though some reorganisation of services to facilitate blood collection at this time point would be required. As the C2 (or if practicable, AUC0−4) measurement offers clinical advantages in the management of CyA-microemulsion-treated patients, we hope that the transplantation community will give consideration to a more widespread adoption of this practice.


The findings of this study verify that AUC0−4 measurements at day 3 are a critical indicator of AR risk, and also demonstrated that C2 measurements have valuable predictive potential. As C2 is a more practical form of monitoring AP than AUC0−4, we recommend that, on the basis of the specific observations made in this study, at day 3 in CyA-microemulsion-treated renal transplant recipients, C2 levels should be at least 1700 ng/mL. The adoption of this simple, practicable CyA-microemulsion-AP strategy has the potential to reduce the risk of AR and early graft loss resulting from inadvertent under-immunosuppression.