Journal Logo

Clinical Transplantation

Exposure-response relationships for everolimus in de novo kidney transplantation: defining a therapeutic range

Kovarik, John M.1 7; Kaplan, Bruce2; Tedesco Silva, Helio3; Kahan, Barry D.4; Dantal, Jacque5; Vitko, Stefan6; Boger, Robert1; Rordorf, Christiane1

Author Information



Everolimus (Certican, Novartis Pharmaceuticals) is a proliferation signal inhibitor currently undergoing evaluation for synergistic activity with cyclosporine to prevent rejection after solid organ transplantation. Calcineurin inhibitors such as cyclosporine and tacrolimus inhibit proinflammatory cytokine synthesis, thus interfering with the progression of T cells from a quiescent (G0) to a competent (G1) state. Macrolide immunosuppressants such as everolimus and sirolimus, however, inhibit the T cell response to these cytokines leading to arrest of the cell cycle at the G1-S interface (1,2). Clinical trials have demonstrated the potential of combining macrolide immunosuppressants with calcineurin inhibitors in the management of de novo kidney allograft recipients (3–5). Exposure-response evaluations from clinical pharmacokinetic studies have indicated that both immunosuppression and certain adverse events are dose- and concentration-related for this class of drugs (6,7). Such evaluations can serve as a helpful adjunct in defining a therapeutic concentration range for the safe and effective use of these immunosuppressants in transplantation (7).

Two international, randomized, double-blind phase 3 trials have recently completed the 1-year follow-up period comparing two everolimus dose regimens versus mycophenolate mofetil (MMF) as part of a triple immunosuppressive regimen with cyclosporine and corticosteroids. Serial everolimus trough concentrations were obtained at each protocol-scheduled visit over the first 6 months after transplantation in both trials. These data were prospectively collected to explore for exposure-efficacy and exposure-safety relationships as a basis for defining a therapeutic concentration range for the use of everolimus in kidney transplantation.

Conventionally, the lower limit of the therapeutic range is defined as that concentration that provides a minimal but clinically relevant level of immunosuppression. In this analysis, the efficacy response was the incidence of freedom from biopsy-confirmed acute rejection in the first 6 months posttransplant and was judged, in part, relative to the incidence in the MMF control arm of the studies. The upper limit of a therapeutic range is conventionally associated with a dose-limiting drug-related adverse event. In this regard, we explored exposure-response relationships for several safety measures identified in everolimus and sirolimus development studies as possibly drug- and concentration-related for this class of macrolide immunosuppressants (3–8). The safety parameters in this evaluation were the incidences of hypercholesterolemia, hypertriglyceridemia, leukocytopenia, and thrombocytopenia.

Although it is not yet clear whether routine therapeutic drug monitoring will be necessary in the general kidney transplant population receiving everolimus, there will likely be patient subpopulations for which such an adjunct could be helpful to better individualize treatment or ascertain compliance with the prescribed regimen. Such subpopulations may include pediatric patients, patients with hepatic impairment (biotransformation is the primary route of everolimus elimination), or patients receiving comedications that are potent inducers or inhibitors of hepatic metabolism that could influence the disposition of everolimus. The analyses reported herein sought to define a therapeutic everolimus concentration range in de novo kidney transplantation.


Study design and immunosuppressive regimens.

Two randomized, double-blind, international efficacy trials were performed to evaluate everolimus as immunoprophylaxis of acute rejection episodes in kidney transplantation. The studies were identical in terms of treatments and clinical assessments with the following exceptions: in one study baseline assessments could be made either pre- or posttransplant but only posttransplant in the other study; in one study everolimus was begun when the graft was functional and in the other study it was begun within 24 hr posttransplant; and in one study additional lipid analyses were performed (only the lipid analyses common to both studies are included here). These differences did not preclude pooling the pharmacokinetic data for the present evaluation. The protocols were reviewed by investigational review boards at the individual study centers and patients gave written informed consent to participate. A total of 1,171 de novo renal allograft recipients (583 and 588 from the two separate studies) were enrolled from 98 clinical centers. Patients were randomized to receive either everolimus 0.75 mg bid (n=387), everolimus 1.5 mg bid (n=393), or MMF 1 g bid (n=391) in addition to cyclosporine and corticosteroids. Everolimus was begun posttransplant administered as a tablet formulation (Certican, Novartis Pharmaceuticals) given simultaneously with cyclosporine (Neoral, Novartis) according to a twice-daily schedule. Cyclosporine was begun at the time of transplantation with subsequent dose titration to bring trough concentrations in the range 150–400 ng/ml in the first month posttransplant and then 100–300 ng/ml thereafter. Corticosteroids were dosed based on a protocol-specified taper.


Clinical visits took place at baseline before transplantation, and then on day 1, weeks 1 and 2, and months 1, 2, 3, and 6 after transplantation. At each visit blood samples were obtained for laboratory biochemistry and hematology evaluations and (with the exception of baseline and day 1) for the determination of the everolimus and cyclosporine morning trough concentration. Patients with presumed acute rejection episodes underwent renal biopsy before or within 48 hr of initiation of antirejection treatment.


Everolimus concentrations were determined in blood by validated liquid chromatography, mass spectrometric methods at two central laboratories, one for each study. Both methods followed a similar approach. Briefly, after addition of 40-O-(3,-hydroxy)propyl-rapamycin as internal standard and ammonium hydroxide, samples were extracted with methyl-tert-butyl-ether. The organic extract was evaporated to dryness and reconstituted in mobile phase. Samples were analyzed using liquid chromatography with mass spectrometry under isocratic chromatographic conditions. Assay performance was assessed on the basis of calibration concentrations ranging from 0.2 to 50.0 ng/ml and quality control concentrations ranging from 0.3 to 30.0 ng/ml determined in duplicate with patient samples. The lower limit of quantification was 0.2 and 0.4 ng/ml at the two laboratories. Variation in accuracy and precision was <10% over the range of calibration and quality control concentrations.

Temporal patterns in laboratory parameters.

Laboratory parameters were compared between treatment groups in a mixed effects model with study, treatment, subject-nested-within-treatment, visit, and treatment-by-visit interaction terms. If a significant Treatment-effect was observed, subsequent pairwise comparisons at each visit were made. Data are presented as mean±SD.

Exposure-response relationships.

The exposure parameter was the everolimus trough concentration (Cmin). The efficacy parameter was the incidence of biopsy-confirmed acute rejection episodes. Safety measures were the incidence of hypercholesterolemia defined as >6.5 mmol/liter (250 mg/dl); hypertriglyceridemia defined as >2.9 mmol/liter (250 mg/dl); leukocytopenia defined as <4×109/liter; and thrombocytopenia defined as <100×109/liter. These laboratory cut-off values were prespecified in the study protocols as clinically notable values.

Each patient’s average Cmin, rejection status, maximum cholesterol and triglyceride levels, and minimum leukocyte and platelet counts, over the first 6 months after transplantation were derived from these data in a total of 695 evaluable patients. The Cmins were divided into five equal distribution groups with 139 patients per quintile: 1.0–3.4, 3.5–4.5, 4.6–5.7, 5.8–7.7, and 7.8–15 ng/ml.

The fractions of patients whose laboratory parameters were above and below the laboratory cut-off values and the fraction of patients free of acute rejection were determined in each quintile and subjected to the median-effect analysis (9). This model relates the fraction of the population affected (fa) and unaffected (fu=1–fa) with respect to a given response, on the one hand, to the drug exposure (D) and the exposure at which half the population are affected (Dm, or “median effect”), on the other hand. This is parameterized as: fa/fu=(D/Dm)m. The relationship is linearized on the logarithmic scale as follows: log(fa/fu)=m·log(D)–m·log(Dm). In this relationship, m is a Hill-type coefficient describing the sigmoidicity in the exposure-response relationship. The resulting log(fa/fu) versus log(Cmin) relationship was assessed by linear regression analysis. Goodness-of-fit of the model to the data was indicated by a regression coefficient (r value) >0.8; a regression P ≤0.05 indicated a significant relationship between exposure and a given response.


Study population and drug exposure.

Pharmacokinetic blood samples for everolimus and/or cyclosporine were received from 1052 patients: 695 were in the everolimus treatment arms and 357 were in the MMF control arm. There were a total of 3355 everolimus trough concentrations (Cmins) with 4.6±1.6 samples per patient (range 1–6). As summarized in Table 1, everolimus Cmins remained stable over the 6-month observation period. There were a total of 5232 cyclosporine trough concentrations. Under the blinded study conditions, cyclosporine trough levels were similar in everolimus- and MMF-treated patients over the observation period (P =0.80).

Table 1
Table 1:
Drug concentrations and laboratory parameters

Freedom from rejection.

Over the first 6 months posttransplant, 129 everolimus-treated patients had one or more biopsy-confirmed acute rejection episodes yielding an incidence of 81.4% freedom from rejection. By way of reference, the incidence of freedom from rejection in the control arm with MMF was 77.0%. Among everolimus-treated patients, those experiencing an acute rejection episode had significantly lower average Cmins compared with their rejection-free peers: 4.6±3.2 vs. 6.0±2.4 ng/ml (P =0.0001).

The incidences of freedom from rejection are summarized in Table 2 in each exposure group. The upper panel of Figure 1 shows the median-effect diagnostic plots with associated regression lines for the fraction of patients with a given response versus everolimus Cmin quintile. Freedom from rejection was well described by the median effect model (r=0.91; slope=1.10;P =0.03). Similar outcomes were observed when each of the component studies was assessed separately (data not shown). At everolimus troughs <3.4 ng/ml (quintile 1), the percent of patients rejection-free was 67.6% indicating that low everolimus exposure added minimally to the baseline dual therapy of cyclosporine and corticosteroids. Everolimus troughs between 3.5 and 7.7 ng/ml (quintiles 2 to 4) yielded 81.3 to 85.6% freedom from rejection. Everolimus troughs 7.8 to 15 ng/ml (quintile 5) were associated with 91.4% freedom from rejection. Based on these distributions, patients were divided into three Cmin exposure groups: <3, 3–7, and >7 ng/ml. Figure 2 shows the Kaplan-Meier survival curves for each group. Cox proportional hazard analysis indicated that the relative risk of experiencing a biopsy-confirmed acute rejection episode was 2.80 if Cmin was <3 ng/ml compared with a Cmin of 3–7 ng/ml (P <0.0001). A further, but statistically nonsignificant, reduction in risk was associated with Cmins >7 ng/ml compared with the middle exposure group: relative risk, 0.85 (P =0.52).

Table 2
Table 2:
Percent of patients in exposure groups with efficacy and safety responses
Figure 1
Figure 1:
Upper panel, Median-effect plot showing the fraction of patients affected/unaffected (fa/fu) with respect to a given response versus the everolimus trough level (Cmin) and the associated regression line. The concentration points are plotted at the median value in each exposure quintile. Responses are freedom from rejection (triangles), hypertriglyceridemia (squares), and thrombocytopenia (circles). Lower panel, Sigmoid exposure-response plot with symbols as in upper panel. The portion of the curves above 15 ng/ml are extrapolations based on the median-effect model. The solid vertical line at 3 ng/ml represents the putative lower therapeutic level for everolimus when combined with conventionally dosed cyclosporine and corticosteroids. The dashed vertical line at 15 ng/ml represents the upper limit of exposure in the study population.
Figure 2
Figure 2:
Kaplan-Meier estimates of the percentage of patients free of biopsy-confirmed acute rejection in various everolimus trough level (Cmin) ranges.


Mean lipid levels are listed in Table 1 and the mean trajectories are shown in Figure 3. The percentage of patients receiving lipid-lowering agents (statins, cholestyramine, gemfibrozil) was similar in the first month in those on everolimus compared with those on MMF being less than 11% in both groups. Between months 1 and 6 the percentage increased in both study groups. Specifically, for everolimus-treated patients, this was 23.9, 34.3, 45.4, and 50.8% at months 1, 2, 3, and 6, respectively. For MMF-treated patients, this was 18.6, 24.0, 31.6, and 36.7%, respectively.

Figure 3
Figure 3:
Mean serum levels of total cholesterol (squares) and triglycerides (diamonds) over the first 6 months posttransplant in patients receiving MMF (open symbols) and everolimus (filled symbols) in a cyclosporine-corticosteroid regimen. Bars represent 1 SDM.

Cholesterol levels increased after transplantation entering a plateau between month 1 and 2. The plateau resulted from counter-measure therapies including dietary restrictions, corticosteroid dose reductions, and lipid-lowering agents. At all visits from week 1 to month 6, values were significantly higher in everolimus-treated patients. The overall incidence of hypercholesterolemia (>6.5 mmol/liter) in the MMF arm was 59.9% and in the everolimus group, 81.3%. As listed in Table 2, the incidence of hypercholesterolemia rose in the first four sequential everolimus Cmin quintiles from 75.5 to 87.1% but due to a decline in the fifth quintile, a significant median-effect model fit was not achieved (r=0.52; slope=0.336;P =0.37).

As shown in Figure 3, triglycerides followed a similar trajectory as cholesterol with the plateau entered between months 2 and 3. The general incidence of hypertriglyceridemia (>2.9 mmol/liter) in the MMF arm was 47.1% and in the everolimus treatment group, 68.5%. In everolimus-treated patients, the incidence rose across the full Cmin exposure range from 59.0 to 77.0% and was well described by the model (r=0.94; slope=0.739 P =0.02) as illustrated in the upper panel of Figure 1.


Leukocyte counts decreased over the first 2 months and then stabilized in both treatment arms as listed in Table 1. Leukocyte counts were significantly lower in everolimus-treated patients from week 1 to month 1. The general incidence of leukocytopenia (<4×109/liter) in the MMF treatment group was 17.8% and in the everolimus group was 14.7%. Across the everolimus exposure range, median-effect analysis demonstrated a flat exposure-response relationship (r=0.19; slope=0.113;P =0.76) with no indication of an increase in leukocytopenia with rising exposure to everolimus as detailed in Table 2.

Postoperative platelet counts recovered in the first week and then stabilized in both treatment groups; they were higher in MMF-treated patients at most visits as listed in Table 1. The general incidence of thrombocytopenia (<100×109/liter) was 7.3% in the MMF study arm and 11.5% in the everolimus group. In everolimus-treated patients the incidence was relatively constant in the three lowest Cmin quintiles (7.2–10.1%) and subsequently rose in the fourth (13.7%) and fifth (17.3%) quintiles as detailed in Table 2. Although median-effect analysis demonstrated a moderate correlation between everolimus Cmin and the incidence of thrombocytopenia (r=0.68), the initial flatness in the relationship in the first three exposure quintiles yielded a nonsignificant fit of the model (slope=0.536;P =0.21) as shown in the upper panel of Figure 1. The incidence of clinically meaningful thrombocytopenia, defined in the protocol as <75×109/liter, was 3.4 and 4.2% in the MMF and everolimus groups, respectively.

Concurrent cyclosporine exposure.

Figure 4 shows the distribution of cyclosporine trough levels for patients in the everolimus quintiles. Each patient’s cyclosporine level is their average over 6 months posttransplant. Although cyclosporine levels were significantly different among everolimus quintiles (P <0.001), inspection of the figure indicates that patients in the three middle everolimus quintiles had similar cyclosporine exposure (medians 209, 210, and 227 ng/ml), whereas patients in the lowest and highest quintiles contributed most to the between-quintile differences. Patients in the lowest everolimus quintile had median cyclosporine levels of 184 ng/ml that were 26 ng/ml less than those in the middle quintile and patients in the highest everolimus quintile had median cyclosporine levels of 237 ng/ml that were 27 ng/ml above those in the middle quintile. This is likely due to the fact that both everolimus and cyclosporine are substrates of the isozyme CYP3A and the counter-transporter P-glycoprotein. Consequently, patients with low bioavailability and/or fast clearance of one drug may handle the other drug in a similar manner and congregate in the lower exposure quintile. Similarly, patients who have good bioavailability and/or slow clearance of these drugs will congregate in the highest quintile. Such patterning regarding clearance has been reported for sirolimus and cyclosporine in renal transplant patients (10). Given the relatively small median differences among cyclosporine levels in the everolimus quintiles and the large region of overlap in cyclosporine levels as shown in Figure 4, a confounding influence on the everolimus exposure-response relationships would be minimal.

Figure 4
Figure 4:
Cyclosporine trough concentrations (Cmin) averaged over 6 months from patients in each everolimus exposure quintile. Shown are the individual patient values (circles, 139 patients per quintile), the group median (bar), and the central 50% of the data (box).


Pharmacokinetic blood sampling was an integral component of the phase 3 safety-efficacy trials with everolimus. It was prospectively included in the study protocols to provide a large database from which to explore potential exposure-response relationships to define a therapeutic concentration range for this agent in kidney transplantation. A clear relationship between everolimus troughs and freedom from acute rejection was observed. Everolimus troughs less than 3.4 ng/ml were associated with a freedom from rejection of 67.6% which was a marginal improvement relative to dual therapy (cyclosporine-prednisone) with a reported incidence of about 49 to 56% (11,12). Everolimus troughs in the range 3.4 to 7.7 ng/ml were associated with a freedom from rejection of 81.3 to 85.6%, similar to that of 77.0% with triple therapy including MMF in the control arm. Although a further increase in everolimus troughs in the range of 7.8 to 15.0 ng/ml yielded a numerically higher incidence of freedom from rejection of 91.4%, the Kaplan-Meier curve did not suggest a pronounced improvement in this trough range. The importance of achieving everolimus troughs of more than 3 ng/ml was underscored by the Cox proportional hazards analysis that indicated a significantly higher risk of acute rejection of nearly 3-fold when levels were less than compared with levels above this cut-off value. On this basis, 3 ng/ml appears to be a lower therapeutic concentration for the use of everolimus in a regimen with conventionally dosed cyclosporine.

The occurrence of hyperlipidemia posttransplant is clearly multifactorial and has been reported in patients receiving only cyclosporine and corticosteroids (13). In our study, cholesterol and triglyceride levels rose posttransplant in both everolimus-treated and MMF-treated patients. From week 1 to month 6 levels of both lipids were significantly higher in the everolimus treatment group. Even at relatively low exposure to everolimus with troughs in the range 1.0 to 3.4 ng/ml, the incidences of hypercholesterolemia and hypertriglyceridemia were numerically higher compared with the respective incidences in the MMF treatment group: 75.5 vs. 59.9% for cholesterol and 59.0 vs. 47.1% for triglycerides, respectively. As everolimus exposure rose, so also did the incidences of hyperlipidemia. Hypertriglyceridemia was significantly related to concentration over all five everolimus exposure quintiles. This was generally the case for hypercholesterolemia up to troughs of 7.7 ng/ml; however, in the highest exposure quintile a decrease in incidence precluded a good model fit with the median-effect paradigm. The plateau in the cholesterol and triglyceride trajectories as shown in Figure 3 indicate that hyperlipidemias are controllable with counter-measure therapy; hence, hypercholesterolemia and hypertriglyceridemia do not appear to be dose-limiting for the use of everolimus in this transplant population.

Macrolide immunosuppressants such as everolimus and sirolimus are protein translation inhibitors that block growth factor—driven transduction signals in the T cell response to alloantigen. Their inhibitory effects, however, are not restricted to T cells. They also inhibit the signals provided by some hematopoietic and nonhematopoietic cell growth factors (1,2). Although inhibition of smooth muscle cell proliferation may constitute a beneficial effect, inhibition of other factors may contribute to the spectrum of drug-related and concentration-related adverse events observed with this class of agents. The myelosuppressive effects of everolimus are presumably related to drug-induced inhibition of the cytokine transduction signals necessary for the proliferation and differentiation of bone marrow elements. Platelets appeared to be more sensitive to these effects than leukocytes. Although leukocyte counts were on average lower in the first posttransplant month in everolimus-treated patients compared with MMF-treated patients, the overall incidence of clinically notable counts <4×109/liter was similar in both groups (14.7and 17.8%, respectively) and did not show any apparent concentration relationship in the everolimus group. Furthermore, the incidences of clinically meaningful decreases in leukocyte counts to <3×109/liter were low in both groups: 4.6 and 8.2%, respectively.

Thrombocytopenia may be a dose-limiting adverse event for macrolide immunosuppressants. At low exposure to everolimus corresponding to troughs up to 5.7 ng/ml, the incidence of thrombocytopenia (<100×109/liter) was similar to that in the MMF-treated group. At troughs of more 5.7 ng/ml, the incidence rose as shown in Figure 1 and in Table 2. Even in the highest quintile of everolimus troughs (7.8–15.0 ng/ml), the incidence did not exceed 17.3%. It is noteworthy that this evaluation was based on a clinically notable platelet decrease. The study protocols also defined a cut-off value for clinically meaningful thrombocytopenia of <75×109/liter. Platelet decreases of this magnitude had a low incidence of 4.2% over the full everolimus exposure range. Consequently, no specific upper limit for an everolimus therapeutic range was apparent using the incidence of thrombocytopenia as the safety criteria. Overall, everolimus was used with manageable safety up to Cmins of 15 ng/ml; hence, an upper therapeutic limit likely lies above this level. An overview of the putative therapeutic range is depicted in the lower panel of Figure 1.

In the present phase 3 trials protocol-mandated cyclosporine troughs were 150–400 ng/ml in month 1 and 100–300 ng/ml thereafter in order to fulfill the immunosuppressive needs of patients randomized to the MMF control arm. Currently there is interest in using macrolide immunosuppressants with lower doses and concentrations of cyclosporine to take advantage of potential immunosuppressive synergy between the two classes of agents and to lessen cyclosporine-related nephrotoxicity. When used in such regimens, the relationship between everolimus troughs and freedom from rejection as characterized in our evaluation may change. Methods such as those applied here may be useful to determine the exposure-response relationships under these conditions and to define a regimen-specific therapeutic range if necessary.


The authors thank Stephane Berthier, Louis McMahon, and Martine Attinger for performing the everolimus and cyclosporine analytical assays, and Chyi-Hung Hsu and Amy Racine for assistance with the statistical evaluation.


1. Schuler W, Sedrani R, Cottens S, et al. SDZ RAD, a new rapamycin derivative: pharmacological properties in vitro and in vivo. Transplantation 1997; 64: 36.
2. Sehgal SN. Rapamune (sirolimus, rapamycin): an overview and mechanism of action. Ther Drug Monit 1995; 17: 660.
3. Kahan BD, Wong RL, Carter C, Katz SH, Von Fellenberg J, Van Buren C, Appel-Dingemanse S. Phase I study of a 4-week course of RAD in quiescent cyclosporine-prednisone-treated renal transplant recipients. Transplantation 1999; 68: 1100.
4. Kahan BD, Kaplan B, Lorber MI, Winkler M, Cambon N, Boger RS, for the RADB157 Study Group. RAD in de novo renal transplantation: comparison of three doses on the incidence and severity of acute rejection. Transplantation 2001; 71: 1400.
5. Kahan BD, Rapamune US Study Group. Efficacy of sirolimus compared with azathioprine for reduction of acute renal allograft rejection: a randomised multicentre study. Lancet 2000; 356: 194–202.
6. Kovarik JM, Kahan BD, Kaplan B, Lorber M, Winkler M, Rouilly M, Gerbeau C, Cambon N, Boger R, Rordorf C. Longitudinal assessment of everolimus in de novo renal transplant recipients over the first post-transplant year: pharmacokinetics, exposure-response relationships, and influence on cyclosporine. Clin Pharmacol Ther 2001; 69: 48.
7. Kahan BD, Napoli KL, Kelly PA, et al. Therapeutic drug monitoring of sirolimus: correlations with efficacy and toxicity. Clin Transplant 2000; 14: 97.
8. Murgia MG, Jordan S, Kahan BD. The side effect profile of sirolimus: a phase I study in quiescent cyclosporine-prednisone-treated renal transplant patients. Kidney Int 1996; 49: 209.
9. Chou T, Talalay P. Quantitative analysis of dose-effect relationships: the combined effects of multiple drugs or enzyme inhibitors. Adv Enzyme Regul 1984; 22: 27.
10. Kaplan B, Meier-Kriesche HU, Napoli KL, Kahan BD. Correlation between pretransplantation test dose cyclosporine pharmacokinetic profiles and posttransplantation sirolimus blood levels in renal transplant recipients. Ther Drug Monit 1999; 21: 44.
11. Nashan B, Moore R, Amlot P, et al. Randomised trial of basiliximab versus placebo for control of acute cellular rejection in renal allograft recipients. Lancet 1997; 1: 1193.
12. Kahan BD, Rajagopalan PR, Hall M, United States Simulect Study Group. Reduction of the occurrence of acute cellular rejection among renal allograft recipients treated with basiliximab, a chimeric anti-interleukin-2-receptor monoclonal antibody. Transplantation 1999; 67: 276.
13. Quaschning T, Mainka T, Nauck M, Rump LC, Wanner C, Kramer-Gurth A. Immunosuppression enhances atherogenicity of lipid profile after transplantation. Kidney Int 1999; 71 (suppl): S235.
© 2002 Lippincott Williams & Wilkins, Inc.