Introduction

Transplantation is the preferred therapy for end-stagerenal failure, as it improves a patient’s quality of life,encourages occupational rehabilitation and is cost-ef-fective compared to the alternative of dialysis [6]. Graftsurvival at 1 year following transplantation has steadilyimproved over the past two decades and is now around90%–95% for low-risk patients. However, for trans-plant patients who survive beyond the 1st year, im-provements in the rate of graft failure have been moremodest [5].

Chronic rejection, a slow progressive process, is re-sponsible for 25%–35% of graft losses after 1 year. Oneof the major risk factors for the development of chronicrejection is the occurrence of acute rejection and,especially, episodes of acute rejection which are severe,recurrent or late [9, 12]. Thus, the development of

therapies to reduce the occurrence of acute rejection ofrenal transplants is of primary importance.

Cyclosporine, azathioprine and prednisolone areconventionally used formaintenance therapy in a numberof different regimens; even so, acute rejection will stilloccur in 30%–50% of patients and particularly in thosewho are at high immunological risk.Over 90%of episodesoccurring in the 1st yearwill occur in the 1st 6months aftertransplantation, and approximately 90% of all episodesof acute rejection can be treated successfully.

Patients and methods

Study design

Basiliximab (Simulect) is a high-affinity IL-2 receptor monoclonalantibody designed to prevent acute rejection episodes in renalallograft recipients when administered as immunoprophylaxis in

ORIGINAL ARTICLETranspl Int (2002) 15: 486–493DOI 10.1007/s00147-002-0458-1

James B. Chilcott

Michael W. Holmes

Stephen Walters

Ronald L. Akehurst

Bjorn Nashan

The economics of basiliximab (Simulect)in preventing acute rejectionin renal transplantation

Received: 7 November 2001Revised: 14 May 2002Accepted: 10 June 2002Published online: 17 September 2002� Springer-Verlag 2002

J.B. Chilcott (&) Æ M.W. HolmesS. Walters Æ R.L. AkehurstSchool of Health and Health RelatedResearch, The University of Sheffield,Regent Court, 30 Regent Street, Sheffield,S1 4DA, UKE-mail: j.b.chilcott@sheffield.ac.ukTel.: +44-114-2220689Fax: +44-114-2220785

B. NashanClinic for Visceral and TransplantationSurgery, Medizinische HochschuleHannover, Hannover, Germany

Abstract An economic evaluationwas undertaken alongside a multi-centre international trial of basili-ximab. Resource usage within thetrial was assessed, and the cost im-plications of using basiliximab eval-uated. Recipients of a primarycadaveric kidney transplant wererecruited into a double-blind trialand received either placebo (n=186)or basiliximab (n=190). Clinicaloutcomes and resource usage weremonitored in the 12 months follow-ing transplantation. Local unit costswere obtained, and global analysiswas undertaken using health sectorpurchasing-power parity rates. Nostatistically significant differences

were found in the mean cost oftreatment per patient. The mean costof treatment was $47,940 for basil-iximab patients and $46,280 forplacebo patients, a mean differenceof $1,660 (95% confidence interval(CI): –$4,150, $7,360; P=0.58).Basiliximab produces clinical benefitin terms of preventing episodes ofacute rejection, whilst the differencein the total resource usage and costof treatment is not statisticallysignificant.

Keywords Costs and cost analysis ÆKidney transplantation Æ Immuno-suppressive agents

addition to baseline immunosuppressive therapy with microemul-sion formulation cyclosporine and steroids. An international,multicentre trial (CHIB 201) designed to test the safety and efficacyof this drug, is reported elsewhere [10].

In this double-blind trial, 376 adults who underwent primarycadaveric kidney transplantation were randomly allocated to re-ceive a 20-mg infusion of basiliximab on day 0 (day of surgery) andon day 4 (n=190) or to receive placebo (n=186). The primaryclinical outcome measure was incidence of acute-rejection episodesduring the 6 months after transplantation. Among the secondaryend points was the effect on the use of health system resourcesduring the trial period. The primary outcome for the economicevaluation was the aggregate mean total cost of resources used perpatient over the 12-month post-transplantation period.

This study was performed in accordance with operating pro-cedures designed to ensure adherence to GCP as required by thefollowing:

• Directive 91 507 EEC: The Rules Governing Medicinal Prod-ucts in the European Community

• US Code of Federal Regulations dealing with clinical studies(21 CFR, including parts 50 and 56 concerning informed con-sent and IRB regulations)

• Declaration of Helsinki, concerning medical research in hu-mans (Recommendations Guiding Physicians in BiomedicalResearch Involving Human Subjects, Helsinki 1964, amendedTokyo 1975, Venice 1983 and Hong Kong 1989).

Informed consent was obtained from all patients, and approvalwas obtained from the Institutional Review Board at each partic-ipating institution.

The trial involved centres in Germany, France, the UK,Canada, Switzerland, Norway and Belgium. This paper presents ananalysis of resource use and costs across the whole trial, usingpooled data from all countries. Results by individual country aresummarised, and costs associated with key clinical events are esti-mated.

Statistical analysis

Differences in resource used and statistical confidence limits arereported for every category of resource collected. Parametric con-fidence intervals (CIs) were calculated by standard methods [4]. CIswere calculated only if the number of subjects in each of the groups(basiliximab or placebo) experiencing the event and not experi-encing the event were greater than 5 [1].

The non-parametric Mann-Whitney U-test was used to com-pare differences between the groups in the number of suspectedrejection episodes. Because the distribution of costs for both ba-siliximab and placebo are highly skewed, traditional statisticalmethods are inappropriate to conclude on a difference betweenmean costs. The bootstrap method [3], however, assumes that thetrue distribution of data is their empirical distributions. Desgagneet al. [2] report that ‘‘…the most reliable result is the P-value cal-culated with the bootstrap.’’

For the cost analysis, the 95% CIs for the observed mean dif-ference were calculated by a bias-corrected and accelerated non-parametric bootstrap percentile method. A two-tailed P value wasobtained by computing the bootstrap Z statistic for testing equalityof means. A P value of less than or equal to 0.05 was considered tobe significant.

The sample size of the study was designed to capture the ex-pected clinical difference in acute rejection rate, not the expecteddifference in costs. Using the standard power calculations [1] on alog-transformed cost indicates that the study would be expected tohave an approximately 50% power to identify a treatment costdecrease equal to the cost of basiliximab, that is, to demonstratecomplete cost recovery.

Resource use

The main purpose of the paper is to describe and analyse the dif-ferences in health system resources used between the groups ofpatients who received basiliximab and those who received placeboin the trial. The study did not attempt to capture all the costs of atransplantation. Costs included were only those that might varywith immunosuppression therapies.

The time period considered was for the trial only. There are, ofcourse, likely to be longer-term cost implications if the clinicalbenefits associated with basiliximab reduce long-term graft lossesor if differences in dialysis usage are sustained over a longer periodof time, but these are not considered in this paper.

It was expected that overall resource use (e.g. drug usage, di-agnostic tests, laboratory tests, consultations) would be positivelyassociated with overall length of stay in hospital. Statistical tests forinteraction were carried out on the total and follow-up length ofstay in hospital.

Pooling of data

An analysis of variance (ANOVA) model was applied to the data,with the total number of days in hospital (length of stay (LoS)including initial and follow-up hospitalisations) as the dependentvariable and the drug and country as explanatory variables. Therewas no statistical evidence of an interaction between the drug andcountry (F=0.76 on 6, 362 df, P=0.60). Therefore, the effect of thedrug on LoS did not differ across the countries.

There was no evidence of a treatment effect (F=0.19 on 1, 368df, P=0.67), that is, LoS did not vary by treatment (basiliximab orplacebo). However, there was a country effect (F=3.52 on 6, 368df, P=0.002) with evidence of differing LoS across the countries.

After the initial hospitalisation, 63.7% of the basiliximab groupcompared with 67.2% of the placebo group had a follow-up hos-pitalisation. The mean LoS (all patients) for these follow-up hos-pitalisations were 15.7 days (SD 23.4) and 16.6 days (SD 26.4) forthe basiliximab and placebo groups, respectively.

The ANOVA model above was reapplied but with the totalnumber of days for follow-up hospitalisations as the dependentvariable and the drug and country as explanatory variables. Therewas no statistical evidence of an interaction between the drug andcountry (F=0.70 on 6, 362 df, P=0.65). Therefore, the effect of thedrug on the number of days of follow-up hospitalisations did notdiffer across the countries.

There was no evidence of a treatment effect (F=0.12 on 1, 368df, P=0.73) or a country effect (F=0.87 on 6, 368 df, P=0.52).That is, the LoS for follow-up hospitalisations did not vary bytreatment (basiliximab or placebo) or country.

Statistically there is no evidence of an interaction, so, therefore,no strong justification to report resource use separately by country.It is reasonable to pool the resource use data, particularly as smallnumbers of patients were recruited in certain countries (e.g. Bel-gium and Norway).

A similar analysis was conducted using clinical outcomes [10],that is, death, graft loss or first rejection episode, as the dependentvariable and drug and country as the independent variables, in-cluding an interaction term. There was no statistically significantevidence of a treatment–country interaction. The hypothesis thatthe effect of basiliximab on both the main resource and clinicaloutcomes was independent of country could not be rejected.Therefore, it was legitimate to pool the data across all countries.

Economic evaluation and analysis

The economic analysis was undertaken from a hospital perspective.Total estimated costs per patient were calculated by estimating thecost of treatment (hospitalisations, dialysis, pharmaceuticals,

487

laboratory tests, diagnostic tests and outpatient consultations) forthe 12-month post-transplantation period.

All resources were valued at 1997 prices. Unit costs were taken,where possible, from the local service providers or national datasources. Discounting was not employed, since costs and benefits areestimated for a 12-month period only.

The primary analysis compared the aggregate mean cost perpatient in each group at 12 months post-transplantation.

The trial provides data on the resources consumed by everypatient in the following categories: drugs used, hospitalisations,laboratory tests, diagnostic tests, outpatient visits, dialysis. Unitcosts have been obtained or estimated for each item of resourceused for each country involved in the trial.

In the drugs used category, the unit cost is the cost per mil-ligramme for the drug. The total cost for each drug for eachpatient is calculated from the number of days on therapy multi-plied by the average daily dose, multiplied by the cost per milli-gramme. The one exception to this is in the ‘‘ATG/ALG’’ item.The total resource usage for anti-thymocyte globulin (ATG) andanti-lymphocyte globulin (ALG) are grouped together in the sameresource item ‘‘ATG/ALG’’ and have not been separately iden-tified. This is problematic in the costing exercise as these drugshave markedly different costs per milligramme, for example, inthe UK the cost per milligramme of ATG is approximately fivetimes the cost per milligramme of ALG. Thus, for this category,in some countries a cost per infusion of either ATG or ALG hasbeen obtained, and this has been multiplied by the number ofdays that the patient received therapy. The costs of basiliximabare taken from actual and targeted costs for the specific countries;costs are per vial and this cost was multiplied by the number ofdoses, either one or two. Drug costs are grouped into steroidtherapy, cyclosporine therapy and other immunosuppressants in-cluding acute rejection rescue therapy.

To calculate dialysis costs, the number of days spent on dialysiswas estimated as the days between the graft loss and either the dayof death or the end of the trial (365 days). The proportion of theyear the patient spent on dialysis was then found and multiplied bythe annual unit dialysis costs [8, 13].

In order to facilitate a cost analysis for the trial as a whole, thelocal-currency unit costs are converted to US$ costs by means of apurchasing-power parity (PPP) rate. The most recently availablehealth care PPP listing is published by the Organisation for Eco-nomic Co-operation and Development (OECD) [11] and was cal-culated for 1996. The PPP rates used for the countries involved inthe trial are given in Table 1. The use of PPPs was preferred toexchange rates, as PPPs reflect more closely the true purchasingdifferences in the health sectors of the countries concerned and nottheir overall strength in internationally traded goods and services.A PPP US$ unit cost vector (available from the authors) is thusobtained for each country in the trial.

No unit costs were available for Norway and Belgium. Sincethese countries accounted for approximately 10% of the total pa-tients in the trial, excluding them from the cost analysis was notjustified. For unit costs that are missing in particular countries, a

simple average of PPP US$ unit costs was taken of the availablecosts.

The country-specific vectors of unit costs have been applied tothe quantities of resources used by each patient to calculate thecosts of treatment, in PPP US$. The costs of each individual patienthave been summed for each arm of the trial so that an average costper basiliximab-arm patient and an average cost per placebo-armpatient could be calculated, together with associated statisticalparameters.

A multivariate linear regression model was used to investigatehow the average cost of treatment for a patient was related to thedifferent types of clinical events experienced. The model which bestexplained the variation in total treatment cost was established usingthe F test and with respect to the R2 statistic. The dependentvariable was the cost per patient (excluding the cost of basiliximab)and the independent variables were: steroid-responsive rejection(Y/N), steroid-resistant rejection (Y/N), hospitalisation for infec-tion (Y/N), hospitalisation for an adverse event (Y/N), and graftloss (Y/N).

Results

Population demographics

Table 2 shows the countries included in the study andassociated numbers of patients enrolled, by treatmentgroup. The German transplant centres contributed 45%of the study patients and the French 21%. The treatmentallocation was well balanced across all countries.

The demographic details of the trial population aregiven in Table 3. The population was well balanced, interms of gender, race and age, across the treatment arms.Approximately 66% of patients receiving basiliximabwere male (126 of 190) compared with 63% of the

Table 1. 1996 PPP rates

Country PPP Unit

Belgium 25.88 BEF per $Canada 0.78 CAD per $Switzerland 1.63 CHF per $Germany 1.27 DM per $France 4.29 FF per $UK 0.4 £ per $Norway 6.89 NOK per $USA 1 $ per $

Table 2. Country by group

Country Basiliximab Placebon (%) n (%)

Belgium 8 (4.2) 7 (3.8)Canada 15 (7.9) 13 (7.0)Switzerland 12 (6.3) 13 (7.0)Germany 85 (44.7) 84 (45.2)France 41 (21.6) 41 (22.0)UK 19 (10.0) 18 (9.7)Norway 10 (5.3) 10 (5.4)Totals 190 186

Table 3. Gender, race and age by group

Characteristic Basiliximab Placebo

n 190 186Male gender 126 118% Male 66.3 63.4Race (Caucasian) 179 179% Caucasian 94.2 96.2Mean age 47.4 47.0SD 12.3 13.3

488

placebo group (118 of 186). The mean age of bothtreatment groups was approximately 47 years.

Summary of clinical outcomes

At 12 months there were 14 (4%) deaths, 48 (13%)grafts were lost, 37 (10%) nephrectomies occurred andthere were no retransplantations (Table 4).

The proportion of patients experiencing an episode ofacute rejection in the 12 months following transplanta-tion was statistically significantly lower in the basilix-imab group than in the placebo groups (v2=10.18 on 1df, P=0.001); this is shown in Table 5. No other sig-nificant differences were found between clinical end-points for the two groups.

Resource use results

Patients on basiliximab had very similar resource use topatients on placebo in terms of cyclosporine therapy,steroid therapy, immunosuppressive therapy for acuterejection, concomitant medications, hospitalisations,laboratory tests, diagnostic tests and outpatient medicalconsultations. No statistically significant differences inany dimension of resource use were found, except forpatient consultations (other).

The differences in the proportion of patients usingeach type of immunosuppressive therapy are shown in

Fig. 1. Figure 2 shows that the mean number of daysof hospitalisation for initial, follow-up and for allhospitalisations for basiliximab patients, is smallerthan for patients on placebo, though these differencesare not statistically significant.

Cost analysis results

The following analysis describes the costs for the treat-ment groups in the trial; the group means, observedmean difference, 95% CIs and P values are presented inTable 6.

• Steroid therapy: all patients received steroid therapy.A reduction of $133 ($240, $10) (P=0.054) was as-sociated with the use of basiliximab.

• Cyclosporine therapy: all patients, with the exceptionof one basiliximab patient in Germany, received cy-closporine orally; a subgroup of patients received cy-closporine intravenously. There was no statisticallysignificant difference in costs.

• Immunosuppressive therapy for acute rejection: notall patients received immunosuppression therapy. Nosignificant difference was found.

• Hospitalisations: all patients, with the exception ofthree (one basiliximab, two placebo) had diagnostictests recorded (one missing, two uncertain missing orvalid, treated as valid).

Table 4. Summary of graft losses and deaths

Outcomen

Basiliximab Placebo Difference

190 186 95% CI for the difference

n (% yes) n (% yes) (Basiliximab–placebo)

Lower Upper

Death 9 4.7% 5 2.7% 2.0% –1.8% 5.9%Graft losses 23 12.1% 25 13.8% –1.3% –8.1% 5.4%Nephrectomy 17 8.9% 20 10.8% –1.8% –7.8% 4.2%

Table 5. Suspected rejectionepisodes Number of

suspectedrejection episodes

Basiliximab Placebo

n 190 186

n (% Yes) n (% Yes)

0 118 62.1 84 45.21 49 25.8 77 41.42 17 8.9 19 10.23 5 2.6 5 2.74 1 0.5 1 0.5Any rejectionepisode

72 37.9 102 54.8a

aP value from Mann-Whitney test: P=0.004

489

• Laboratory tests: all patients, with the exception ofone patient in the placebo group, had laboratory testsrecorded (missing data).

• Diagnostic tests: all patients, with the exception ofthree (one basiliximab, two placebo) had diagnostictests recorded (one missing, two uncertain missing orvalid, treated as valid).

• Approximately 90% of patients in each treatmentgroup were recorded as receiving outpatient consul-tations.

• Outpatient visits: approximately 90% of patients ineach treatment group were recorded as receivingoutpatient consultations.

• Dialysis: 46 patients received dialysis following graftloss (22 basiliximab, 24 placebo). Overall, there was anon-significant reduction of $1,378 (95% CI: $3,920, –$2,080).

Average cost of treatment by therapy

The average cost of treatment (PPP US$) for patientsreceiving basiliximab was $47,900 (95% CI: $44,000,$52,000) including the cost of basiliximab and $44,400excluding the cost of basiliximab. For patients receivingplacebo the average cost of treatment was $46,300 (95%CI: $42,300, $50,300). The standard deviation of costswas approximately $28,000 for both arms of the trial.These results show no statistically significant differencein the cost of treatment between the two groups(P=0.577). The mean cost difference is $1,660 (95% CI:–$4,150, $7,360). The bootstrap analysis indicates thatthere is a 24% probability that basiliximab is cost savingwhen compared with conventional therapy.

Cost effectiveness

The number of suspected rejection episodes during the12-month post-transplantation period was consideredthe most appropriate measure available for the cost-ef-fectiveness analysis. Table 5 shows that the number ofpatients suffering any suspected rejection episodes dur-ing the 12-month post-transplantation period was sig-nificantly lower (P=0.001) in the basiliximab groupthan in the placebo group (37.9% vs 54.8%, differenceof –16.9%; 95% CI: –29% to –4.0%). Thus, the incre-mental cost-effectiveness ratio per suspected rejectionepisode avoided for treatment on basiliximab is $1,660/0.169=$9,823.

The number needed to treat for benefit (i.e. no sus-pected rejection episode in the 12-month post-trans-plantation period) is 5.9 (95% CI: 3.8 to14.6). That is,six patients needed to be treated with basiliximab ratherthan placebo in order for one additional patient tobenefit (i.e. have no suspected rejection episode in the12-month post-transplantation period).

Country-specific results

Figure 3 presents the initial and follow-up LoS bycountry. Cost results are summarised by country inFig. 4. In no individual country was there a significantdifference between cost of treatment using basiliximaband using placebo.

Numbers of patients in several countries were quitelow, thus not too much should be read into the differ-ences in costs between countries. Those differences mayreflect chance variations in patient characteristics. Nev-ertheless, the cost differences do appear to be verymarked and themselves worthy of further investigation.In particular, the question of whether extra resources areassociated with better outcomes is worthy of investiga-tion.

Fig. 1. Differences in proportions of patients with immunosup-pressive therapy

Fig. 2. Differences in LoS in hospital

490

Costs of clinical events

The F statistics for the model and each of the indepen-dent variables are highly statistically significant(P<0.001) and are given in Table 7. The R2 statisticindicates that the model explains 77% (59% adjusted) ofthe variability in treatment cost. The coefficients of thecost model give the average costs associated with each ofthe key clinical events (Table 7). For example, the modelindicates that the cost associated with experiencing acuterejection is approximately $7,660 when the patient re-sponds to steroid therapy, and around $22,000 when thepatient is resistant to steroid therapy.

Discussion

The basiliximab and placebo groups had very similarresource use in terms of cyclosporine therapy, steroidtherapy, hospitalisations, diagnostic tests and outpatientmedical consultations. No statistically significant differ-ences in any dimension of resource use were foundexcept in the case of ‘‘other patient consultations’’.Similarly, there were no statistically significant differ-ences in the costs of treatment in the above categories,nor were the overall average costs of treatment perpatient over the 1st 12 months following transplantationsignificantly different between the two groups.

The cost model estimates the costs associated with thekey clinical events post-transplantation. The potential forcost recovery in the 1st year, from the reduction in theprobability of acute rejection, would be in the order of$1,300 ($7,660·0.17) per transplant recipient; this is inaccordance with the average cost saving of $1,800 ex-cluding the cost of basiliximab. This indicates that there isstrong potential for partial cost recovery, whilst thebootstrap analysis demonstrates a 24% chance of com-plete cost recovery in the 1st year. The strong link betweenacute rejection and graft failure also indicates a potentialfor long-term improvements in graft survival from the useof basiliximab. If this link were validated, then therewould be potential for massive economic benefits fromreductions in costs associated with graft failure and returnto dialysis. Further research is required on the long-termeconomic impact of basiliximab.

A study of similar design undertaken in the USA [7]demonstrated a reduction in follow-up hospitalisations.

Table 6. Cost analysis results for the difference of means (basiliximab–placebo)

Parameter Observed mean cost Difference 95% CI P

Basiliximabn=190

Placebon=186

Basiliximab–placebo

Lower Upper

$ $ $ $ $

Steroid therapy 417 550 –133 –240 –10 0.054Cyclosporine therapy 5,764 5,820 –56 –960 890 0.907

Immunosuppression therapyAll patients 1,985 2,694 –709 –1,960 840 0.33Patients with therapy:

basiliximab n=66,placebo n=74

5,716 6,772 –1056 –4,400 3,420 0.617

Hospitalisation 20,210 19,879 331 –2,740 3,710 0.853Laboratory tests 7,415 7,715 –300 –1,680 1,150 0.662Diagnostic tests 1,621 1,570 51 –300 410 0.777Outpatient visits 1745 1449 296 –190 820 0.237Costs of dialysis 5,223 6,601 –1,378 –3,920 2,080 0.396

Total cost of treatmentExcluding cost ofbasiliximab

44,380 46,276 –1,896 –8,110 4,480 0.555

Including cost ofbasiliximab

47,936 46,276 1,660 –4,150 7,360 0.577

Fig. 3. Initial and follow-up LoS in hospital by country

491

In this multinational study, however, the reduction inhospitalisations was not statistically significant. Thisdilution of the link between treatment effect and re-source use may reflect the tendency for patients in thecountries included within this multinational study to bekept in hospital for much longer for their initial period

of care. Thus, whereas prevention of early episodes ofacute rejection may save a readmission in the USA, thiswould not necessarily lead to an earlier hospital dis-charge following transplantation in some of the coun-tries involved in this study.

Figure 3 shows the LoS by country. The wide varia-tion between countries in LoS and overall resource usedoes not appear to be associated with differencesin patient outcomes, and further investigation is war-ranted. It might be possible for costs of hospitalisationto be reduced in many countries with no detriment topatients. The overall conclusion from the data presentedhere is that basiliximab produces clinical benefit in termsof preventing episodes of acute rejection, whilst thedifference in the cost of treatment over the 1st 12 monthsis not statistically significant.

Acknowledgements The research was funded by Novartis, Basel,Switzerland.

Table 7. Cost of key clinical events

Event Costcoefficient

SE t Significance

$ $

C0 constant 25,072 1,655 15.15 0.000C1 steroid-responsiverejection

7,660 2,061 3.72 0.000

C2 steroid-resistanttherapy

22,561 2,940 7.67 0.000

C3 infection 11,739 2,245 5.23 0.000C4 adverse events 12,148 1,887 6.44 0.000C5 graft loss 53,329 2,883 18.50 0.000

Fig. 4. Mean total cost perpatient by country including thecost of basiliximab in PPP US$units

References

1. Altman DG (1991) Practical statisticsfor medical research. Chapman & Hall,London

2. Desgagne A, Castilloux A, Angers J,LeLorier J (1998) The use of thebootstrap statistical method for thepharmacoeconomic cost analysis ofskewed data. Pharmacoeconomics13:487–497

3. Efron B, Tibshirani R (1993) An in-troduction to the bootstrap. Chapman& Hall, New York

4. Gardner MJ, Altman DG (1989)Statistics with confidence: confidenceintervals and statistical guidelines.BMJ, London

5. Hariharan S, Johnson CP, BresnahanBA, Taranto SE, McIntosh MJ,Stablein D (2000) Improved graft sur-vival after renal transplantation in theUnited States, 1988 to 1996. N Engl JMed 342:605–612

6. Karlberg I, Nyberg G (1995) Cost-effectiveness studies of renal trans-plantation. Int J Technol Assess HealthCare 11:611–622

7. Lorber MI, Fastenau J, Wilson D,DiCesare J, Hall ML (2000) A pro-spective economic evaluation of basil-iximab (Simulect) therapy followingrenal transplantation. Clin Transpl14:479–485

492

8. Mallik NP (1997) The cost of renalservices in Britain. Nephrol DialTransplant 12 [Suppl 1]:25–28

9. Massy ZA, Guijarro C, Kasiske BL(1995) Clinical predictors of chronicrenal allograft rejection. Kidney IntSuppl 52:S85–S88

10. Nashan B, Moore R, Amlot P, SchmidtAG, Abeywickrama K, Soulillou JP(1997) Randomised trial of basiliximabversus placebo for control of acutecellular rejection in renal allograftrecipients. Lancet 350:1193–1199

11. Organisation for Economic Co-opera-tion and Development (1998) OECDHealth Data 1998 (on CD-ROM). Or-ganisation for Economic Co-operationand Development, OECD HealthPolicy Unit, Paris

12. Pirsch JD, Ploeg RJ, Gange S,D’Alessandro AM, Knechtle SJ,Sollinger HW, Kalayoglu M, BelzerFO (1996) Determinants of graftsurvival after renal transplantation.Transplantation 61:1581–1586

13. Pritchard SS (1997) The cost of dialysistreatment for patients with end-stagerenal disease in France. Nephrol DialTransplant 12 [Suppl 1]:29–32

493