Content uploaded by Pieter ter Wee
Author content
All content in this area was uploaded by Pieter ter Wee on Oct 23, 2014
Content may be subject to copyright.
Original Article
Economic impact of extended time on peritoneal dialysis as a result of
using polyglucose: the application of a Markov chain model to forecast
changes in the development of the ESRD programme over time
Tom J. G. Weijnen
1
, Henk W. van Hamersvelt
2
, Paul M. Just
3
, Dick G. Struijk
4
, Yuvan I. Tjandra
5
,
Piet M. ter Wee
6
and Frank Th. de Charro
1
1
Centre for Health Policy and Law, Erasmus University Rotterdam, Rotterdam, The Netherlands,
2
Department of
Nephrology, University Hospital St Radboud, Nijmegen, The Netherlands,
3
Renal Division, Baxter Healthcare Inc.,
Deerfield, IL, USA,
4
Renal Unit, AMC, University of Amsterdam, Amsterdam, The Netherlands,
5
Department of
Nephrology, Red Cross Hospital, The Hague, The Netherlands and
6
Department of Nephrology, Vrije Universiteit
Medical Center, Amsterdam, The Netherlands
Abstract
Background. The use of polyglucose as a peritoneal
dialysis (PD) fluid extends time on PD treatment. It
is anticipated, therefore, that the share of patients
treated with PD will be positively influenced. The rela-
tionship between extension of PD treatment time and
an increase of the PD treatment share, however, is
complex and needs further investigation. In this paper,
a Markov chain model was applied to investigate the
impact of extended time on PD treatment for the PD
share in all dialysis patients in The Netherlands.
Furthermore, the economic impac t of the extended
time on treatment (ETOT) was explored.
Methods. Scenarios were forecast over a 10 year
period using aggregate data from the End-Stage
Renal Registry in The Netherlands (Renine). Three
scenarios were simulated in which the median PD
technique survival was extended by 8, 10 and 12
months. Two other scenarios explored the impact of
the combined effect of ETOT of 10 months together
with a 10% and 20% increase of PD inflow shares.
Reductions of costs to society due to ETOT were
estimated using Dutch cost data on renal replacement
therapies.
Results. PD share increa ses from 30.0% in the null
scenario to 34.5% in the scenario with an ETOT of
10 months and an increased PD inflow share of 20%.
The reduction in total costs to society of the renal
replacement therapies is 0.96%. The average societal
costs per discounted patient year for haemodialysis
(HD) are J84 100. For PD, these costs are J60 300.
A shift from HD to PD results in average cost savings
of 28% per patient year.
Conclusions. In view of high dialysis costs to society,
a reduction of 0.96% can be considered to be relevant
for healthcare policy makers.
Keywords: economic evaluation ; extended time on
treatment; Markov chain modelling; peritoneal
dialysis; polyglucose
Introduction
Although glucose is the principal osmotic agent used
in peritoneal dialysis (PD), it is sometimes associated
with a relatively short duration of effective ultrafiltra-
tion. Research has sho wn that the use of polyglucose
as a PD fluid extends time on PD treatment [1]. It can
therefore be expected that the PD share in dialysis
patients will be positively influenced. This implies that
the annual costs of an end-stage renal disease (ESRD)
programme will decrease, because costs to society of
PD are lower than costs of haemodialysis (HD) [2].
In this paper a Markov chain model was applied,
to investigate the impact of extended time on PD
treatment (ETOT) in The Netherlands. Anonymous
and aggregated data from the Dutch Renal Disease
Registry (Renine) was used to forecast patient numbers
over a 10 year period in six different scenarios. A null
scenario describes a situation where there is no impact
resulting from ETOT. To model the effects of ETOT,
three scenarios were defined where the transition rates
in the Markov chain model were adapted. In addition,
two scenarios described the effects of ETOT together
with increased PD inflow shares. The effect measures
Correspondence and offprint requests to: T. J. G. Weijnen, MSc,
PO Box 1738, NL-3000 DR Rotterdam, The Netherlands. Email:
weijnen@frg.eur.nl
Nephrol Dial Transplant (2003) 18: 390–396
#
2003 European Renal Association–European Dialysis and Transplant Association
by guest on June 13, 2013http://ndt.oxfordjournals.org/Downloaded from
were: the PD share in total ESRD patients at the end
of the 10 year model period; the discounted patient
years shifted from HD to PD over the model period;
and the impact of this shift in patient years on the costs
of the ESRD programme.
Subjects and methods
In general, patients in an ESRD programme are vulnerable
to competing risks. A patient has a certain probability of
staying on the same treatment, being transferred to other
treatments, recovering hisuher renal functioning or dying. If
technique survival on PD is extended through the use of
polyglucose in all patients, ETOT could be directly trans-
lated into a higher PD share. However, polyglucose is only
effective in reducing technique failure in patients suffering
from ultrafiltration failure. Other patients leave PD treat-
ment for different reasons, such as peritonitis. Moreover,
patients can receive a kidney transplant or die during the
additional time they could have stayed on PD treatment as
a result of using polyglucose. Consequently, the benefits of
ETOT are not transparent.
The total effect of the use of polyglucose in ultrafiltration
patients can be divided into direct and indirect effects. The
direct effect results from the increase in PD treatment time.
The indirect effect results from the increase in the number of
patients starting on PD instead of HD. It is expected that
more patients will start on PD because the chances of staying
for a longer period on stable PD treatment are higher.
Markov chain model
A Markov chain model simulates the dynamics in a patient
population, distributed over a number of treatment and age
groups. Continuous time is divided into a sequence of
discrete periods. The model estimates the number of patients
in the different treatment and age groups at the end of each
period. In this application of the Markov chain model, a
period equals 1 month. Consequently, a scenario covering
10 years consists of 120 periods. In the first month of the
model period, the calculations begin with a known distribu-
tion of patients over treatment and age groups. The number
of patients per treatment and age groups at the end of the
month are derived from: (a) the number of patients entering
the ESRD treatment—this is the inflow; (b) the distribu-
tion of inflow over the treatment and age groups; (c) the
application of the transition rates that indicate the number
of patients moving to another treatment or age group
within the period; and (d) the transition rates of patients
ending the treatment—this is the outflow from the model.
The transition rates and the relative distribution of the
new patients over the groups are exogenously determined.
The transition rates were calculated from the actual transi-
tions that occurred in The Netherlands in the period 1997–
1999. The relative distribution of the new patients over the
groups equals the average relative distribution over the years
1997–1999.
Figure 1 shows a simplified representation of the Markov
chain for three treatment groups and two calculation periods.
The number of patients in each group at the end of the
month is used as a starting point for the calculations in
the next period. Again, the patients entering the different
treatment and age groups are added to the patient numbers
at the start and the transition rates are applied.
Since transitions differ over age, four age groups were
distinguished in the model: 0–44, 45–64, 65–74 and 75q
years. Patients can move from a younger age group to an
older group. Five modes of treatment were distinguished:
full care-centre haemodialysis (FCCHD); limited care-centre
haemodialysis (LCCHD); home haemodialysis (HHD); PD;
and living with a functioning donor kidney (KTX).
Within treatment groups, two sub-treatments were dis-
tinguished for transitions occurring in the first year of treat-
ment and transitions occurring in later years. The reason for
this is that in some treatments, transitions in the first year
are higher than in later years. The model has a total of
40 different groups (4 age groups
3 5 modes of treatment3 2
sub-treatments) to describe developments in the patient
population. In addition to the 40 treatment groups, the
model has two ‘absorbing states’. These states are related to
death and to recovery of patients’ own kidney function. Both
death and recovery constitute the ‘outflow’ from the model.
Data
Anonymous and aggregated data were provided by the End-
Stage Renal Registry in The Netherlands (Renine). Renine
Fig. 1. The Markov chain.
391Economic impact of using polyglucose
by guest on June 13, 2013http://ndt.oxfordjournals.org/Downloaded from
has data available for transitions between modes of treat-
ment, incidence in the ESRD programme and mortality [3].
These data were used to calculate the transition rates and the
distribution of the new patients over the treatment and age
groups
On the basis of the number of new patients per age group
in the period 1990–1999, a linear trend was estimated
describing the incidence in the four age groups per million
of the population. This trend, combined with predictions of
the age-specific growth of the Dutch population, was used to
calculate the absolute number of new patients per age group
in the 10 year model period [4,5].
The null scenario
The null scenario explores the developments without ETOT.
The incidence per million of the age-specific population in
the 10 year model period is assessed according to the calcu-
lated linear trend. The distribution of these new patients over
treatment groups is equal to the average distribution in the
period 1997–1999. In the null scenario, the transition rates
are the average transition rates from the 3 year period 1997–
1999. Monthly transition rates were calculated as the ratio of
the transitions actually observed in a month and the number
of patients at the beginning of that month.
Scenarios with extended time on PD
Three scenarios were defined to explore the direct effects of
ETOT. These scenarios differ from the null scenario in the
transition rates from PD to HD. For the period 1990–1999,
technique survival on PD was calculated. Figure 2 shows
the technique survival on PD, for patients starting their PD
treatment in the period 1990–1999. Using the Renine data,
it was found that for The Netherlands, median time on PD
until a transition to HD was 3.9 years for all PD patients.
In PD patients with ultrafiltration failure, Wilkie et al. [1]
found a median extension of technique survival on PD of
22 months as a result of the use of polyglucose as a dialysis
fluid. Peers et al. [6] found an extended continuous ambu-
latory peritoneal dialysis (CAPD) technique survival of at
least 1 year. It was estimated that incidence of ultrafiltra-
tion failure in PD patients is 5% in the first year of treatment,
30% after 3 years of PD treatment and 50% after 4 years [7,8].
To represent these estimates in the Markov model, the transi-
tion rates from PD to HD in the first year of treatment
were kept constant. The assumption is that in the first year
of treatment there is no effect of the use of polyglucose. In
one scenario it was assumed that the median time until
technique failure on PD increases by 10 months for all PD
patients. This corresponds with ;22 months of ETOT for
the 50% of patients with ultrafiltration failure. Two other
scenarios were defined, a scenario with a lower median
ETOT of 8 months and a scenario with a higher extension of
median technique survival of 12 months.
The monthly transition rates for later years on treatment
in the Markov model were adjusted to express the increased
time on PD treatment in such a way that the median time
on PD of 3.9 years increased by 8, 10 and 12 months.
Figure 2 shows the survival curve for the scenario with the
ETOT of 10 months, compared with the observed survival
curve in the period 1990–1999.
The adjusted transition rates were obtained by correcting
for the adjusted failure rate as an impact of ETOT. The
theoretical survival (s
t
) can be defined as:
s
t
~(1{r)
t
ð1Þ
where r is the failure rate and t is the number of periods. If
T is the median time on PD in months, the failure rate r
can be described as:
r~1{e
(ln(0:5)=T)
ð2Þ
The extended median time is T
e
sTqextension. The
adjusted failure for each ETOT can be described as:
r
e
~1{e
(ln(0:5)=T
e
)
ð3Þ
where r
e
is the adjusted failure rate for ETOT.
The adjusted transition rates were calculated by multi-
plying all transition rates from PD to HD with a factor r
e
ur.
With a median time on PD of 3.9 years, the monthly failure
rate (r) is equal to 0.0147 and the failure rate for an extended
Fig. 2. Technique survival on PD with adjusted failure-rate.
392 T. J. G. Weijnen et al.
by guest on June 13, 2013http://ndt.oxfordjournals.org/Downloaded from
time on PD (r
e
) of 8 months is equal to 0.0126. As a
consequence, the correction factor for all transition rates
from PD to HD is 0.8649. For the extended times of 10 and
12 months, the correction factors for all transition rates
are 0.8366 and 0.8101, respectively.
Scenarios with increased PD incidence and extended
time of 10 months
Because of ETOT, PD treatment will be more attractive to
patients and doctors. It can therefore be anticipated that the
share of patients starting their renal replacement therapy
with PD will increase. The relationship between ETOT and
increasing incidence in PD is unknown. Two scenarios have
been defined in which the PD incidence increases by 10% or
by 20% compared with the inflow share in the null scenario.
The relative distribution of PD incidence over the age groups
has been kept equal to this distribution in the null scenario.
Compared with the null scenario, total ESRD incidence does
not change. Also, the share of transplants in the new patient
does not differ. In these two scenarios with increased PD
incidence, the extended time on PD treatment is 10 months.
Costs
To estimate the cost impact of extended time on PD treat-
ment, total costs of the renal replacement programme in five
scenarios were compared with total costs in the null scenario.
The cost estimates were based on a cost study in The
Netherlands [5]. In this study, total costs to society were
calculated per patient per year for different treatment and
age groups. Treatment groups were: FCCHD, LCCHD,
HHD, CAPD, continuous cycling peritoneal dialysis (CCPD)
and KTX; age groups were: 0–44, 45–64, 65–74 and 75q
years. The costs of PD used in the estimates consisted of
a weighted average of CAPD and CCPD costs. The weights
were the age-related shares of CAPD and CCPD patients in
total PD patients. Costs to society included costs of dialysis,
other dialysis-related healthcare costs and patient costs.
Costs of home adaptation for PD patients and travel costs
for HD and PD patients were also included.
Costs in the first year on treatment and costs in the later
years on treatment were distinguished for each treatment and
age group. Total costs in the first year of treatment were
higher than in the later years. In the first year, costs of
hospitalization, vascular access operations and of training
of patients are higher than in later years. In the dialysis
treatment groups, costs in the two older age groups of
65–74 and 75q are higher than costs in the two younger age
groups. This is due to the higher costs of hospitalization in
the older age groups.
In the original cost study, additional costs of polyglucose
were not taken into account. In The Netherlands, these
costs amount to J3313 per patient per year. The extra costs
of polyglucose were added to the costs of PD for the
additional patient years as a result of ETOT. In general,
resource use was valued at real cost to society, not at
reimbursement rates. Originally costs were calculated at a
1996 price level. Using a price index for healthcare services,
costs have been updated to 1999 prices. To calculate total
costs of dialysis over a 10 year period in the six scenarios,
yearly costs were discounted using a rate of 4%.
Results
Figure 3 presents total estimated inflow per million
of the population per age group of ESRD patients in
the period 1990–2009. Figures from 1990 to 1999 are
actual inflow figures. The figures from 2000 to 2009 are
estimates using linear trend extrapolations for the
10 year period. This estimated trend has been used to
calculate the total number of new patients in the
estimation period.
The estimated trend indicates a stronger growth
of incidence in ESRD patients per million of the
population in the older age groups of 65–74 and 75q
compared with the age groups 0–44 and 45–64.
Table 1 summarizes the distribution of total number
of new patients over HD and PD treatment in shares
of total new patients per age group. The shares in the
Fig. 3. Inflow of ESRD patients per age group per million of the population. Realized inflow in the period 1991–1999 and estimated trend.
393Economic impact of using polyglucose
by guest on June 13, 2013http://ndt.oxfordjournals.org/Downloaded from
null scenario are derived from the average relative
distribution of new patients over treatment and age
groups in the period 1997–1999 in The Netherlands.
Table 1 also presents the relative distribution in the
two scenarios with a 10% or 20% increased PD inflow
share.
Figure 4 presents the development of the PD shares
in the six scenarios in the 10 year model period from
2000 to 2009. The realized PD shares for 1999 are also
presented. The predicted PD share in the null scenario
initially increases from 30.5% at the beginning of 2000
to 31.0%, but decreases to 30.0% at the end of 2009
as a result of changes in the age distribution of the
dialysis population. The scenarios in which the median
time on PD treatment was increased by 8, 10 and
12 months, result in PD shares at the end of 2009
increasing to 30.8%, 31.0% and 31.1%, respectively.
If the PD infl ow share increases by 10% or 20%, the
share of PD in total dialysis patients increases to
32.7% or 34.5% at the end of 2009.
Extended time on PD treatment results in a shift
of life years from HD to PD. The increase of PD
inflow shares enhances this shift of life years from HD
to PD. In Table 2, the total number of discounted
life-years on HD for the different scenarios over the
10 year modelling period are presented. The decrease
of patient years in HD equals the shift to PD. In the
scenario with an ETOT of 10 months, 1.0% of the
discounted life-years in the null scenario shifts from
HD to PD. In the scenarios with an ETOT of 8 or 12
months, this shift is 0.8% and 1.1%, respectively. In the
scenarios with an ETOT of 10 months and increased
inflow shares of 10% or 20%, the proportion of shifted
discounted life-years is 2.9% and 4.9%, respectively.
Table 3 shows the total discounted costs to society
of HD and PD treatment forecasted in different
scenarios.
In the ‘median survivalq10 months’ scenario, the
reduction in total co sts is 0.19% compared with the
null scenario. In the ‘median survivalq8 months’ and
in the ‘median survivalq12 months’ scenarios, the cost
reduction is 0.15% and 0.22%, respectively. In the
scenarios where the PD inflow share increases by 10%
and 20% together with an ETOT of 10 months, the
cost reduction is 0.58% and 0.96%, respectively. In the
null scenario the total costs per discounted patient
Table 1. Incidence per age group in the null scenario and the two scenarios with increased inflow shares, in shares of total new patients per age
group
Age group Scenarios HD PD KTX
0–44 Null scenario 52.3 41.5 6.2
10% increased PD inflow shareqETOT of 10 months 48.2 45.7 6.2
20% increased PD inflow shareqETOT of 20 months 44.0 49.8 6.2
45–64 Null scenario 60.9 35.5 3.7
10% increased PD inflow shareqETOT of 10 months 57.3 39.1 3.7
20% increased PD inflow shareqETOT of 20 months 53.7 42.6 3.7
65–74 Null scenario 76.6 23.0 0.4
10% increased PD inflow shareqETOT of 10 months 74.3 25.3 0.4
20% increased PD inflow shareqETOT of 20 months 72.0 27.6 0.4
75q Null scenario 84.6 15.4 0.0
10% increased PD inflow shareqETOT of 10 months 83.1 16.9 0.0
20% increased PD inflow shareqETOT of 20 months 81.5 18.5 0.0
Fig. 4. The PD share in the six scenarios over a 10 year model period (2000–2009) and the start value for the analysis at the end of 1999.
394 T. J. G. Weijnen et al.
by guest on June 13, 2013http://ndt.oxfordjournals.org/Downloaded from
year for HD are J84 100. The total costs per
discounted patient year for PD are J60 300. These
costs constitute average HD and PD costs over the
model period, weighed by the total patient years per
treatment and age group. In the model period, a shift
from HD to PD results in cost savings of on average
28% per patient year.
Discussion
A Markov chain model was used to estimate the effect
of the use of polyglucose as a dialysis fluid in patients
with ultrafiltration failure over a 10 year period. The
total effect was distinguished between a direct effect
and an indirect effect. The direct effect results from
the extended time on PD treatment and the indir ect
effect occurs as the impact of increased PD inflow
shares. These increasing PD inflow shares are expected
to result from the improved perspect ive for patients
who stay longer on a stable PD treatment. Thr ee
scenarios were defined in which the direct effect of
ETOT was modelled by increased median technique
survival of 8, 10 and 12 months for all PD patients.
Two other scenarios combined the extended time of
treatment of 10 months with increased PD inflow
shares of 10% and 20%. The null scenario was based
on technique success for HD, PD and KTX, as it was
observed in the Dutch ESRD regis try in the period
1997–1999. The effect measures were: the increased
PD share in the total ESRD programme at the end of
the 10 year period; the shift of discounted patient years
from HD to PD in the model period; and the cost
impact as a result of ETOT.
In the scenarios in which the median time on PD
treatment increases by 8, 10 and 12 months, the pre-
dicted PD share at the end of 2009 increases to 30.8%,
31.0% and 31.1%, respectively. The scenario with the
10 mon ths extended time on PD treatment and a 10%
increase of the PD inflow shares shows a PD share of
32.7% at the end of 2009. When the PD inflow share
increases by 20%, the share of PD in total dialysis
patients increases to 34.5% at the end of the model
period.
Compared with the null scenario, the scenarios in
which median technique success increases by 10 months
combined with increased PD inflow shares of 10%
and 20%, show a shift in discounted patient years
from HD to PD of 2.9% and 4.9%, respectively.
Although this seems to be low, the absolute number of
discounted patient years shifted from HD to PD is 988
and 1638 patient years respectively. The scenarios
with an ETOT of 8, 10 and 12 months show a shift
in patient years from HD to PD of 0.8%, 1.0% and
1.1%, respectively. In the model period, average
societal costs per patient on PD are estimated to be
28% lower than average societal costs per HD patie nt
(PD: J60 300 per year; HD J84 100 per year). Total
societal costs of dialysis can decrease by 0.6% to 1.0%
as a result of both the direct and indirect effects. This
implies for The Netherlands that societal costs
decrease by J35.4 million in the scenario with an
ETOT of 10 months and an increased PD inflow
share by 20%.
The analysis indicates that the impact resulting from
increased inflow shares is larger than the direct effect
of the ETOT. An additional analysis was performed
to investiga te the sensitivity of different assumptions
for our conclusion. In this sensitivity analysis, a
number of scenarios was used to investigate the
influence of increasing PD inflow sh ares and increased
time on treatment on the overall PD share. In one
group of scenarios, the PD-inflow share was increased
by 50%, in 10 steps of 5%. In another group of
scenarios the time on treatment was increased in
36 steps of 1 month. The model results from these
scenarios were used to estimate the influence of
increased PD-inflow shares and ETOT on overall
PD shares. Ordinary least square regressions were
applied to estimate these relationships. The regres sions
Table 2. Discounted patient years shifted from HD to PD over the
10 year model period
Scenarios Total number
of discounted
patient years in
HD patients
Discounted
patient years
shifted from
HD to PD
Share of
discounted patient
years shifted from
HD to PD
Null scenario 33641.7
ETOT of
8 months
33374.9 266.8 0.8%
ETOT of
10 months
33316.7 325.0 1.0%
ETOT of
12 months
33261.4 380.3 1.1%
10% increased
PD inflow
shareqETOT
of 10 months
32654.2 987.5 2.9%
20% increased
PD inflow
shareqETOT
of 10 months
32004.2 1637.5 4.9%
Table 3. Discounted societal costs of dialysis over the 10 year model
period (millions of Euros)
Scenarios HD PD Total
costs
Cost savings
compared
with the
null scenario
Savings
as % of
total costs
Null scenario 2828.9 898.6 3727.5
ETOT of
8 months
2806.5 915.3 3721.8 5.7 0.15%
ETOT of
10 months
2801.6 918.9 3720.5 7.0 0.19%
ETOT of
12 months
2797.0 922.4 3719.4 8.1 0.22%
10% increased
PD inflow
shareqETOT
of 10 months
2746.2 959.6 3705.8 21.7 0.58%
20% increased
PD inflow
shareqETOT
of 10 months
2691.9 1000.2 3692.1 35.4 0.96%
395Economic impact of using polyglucose
by guest on June 13, 2013http://ndt.oxfordjournals.org/Downloaded from
indicated that an increase of the PD inflow by 1%
causes the PD shares to increase by 0.57%. An increase
of the technique survival on PD by 1% causes the PD
share to increase by 0.11%. The sensitivity analysis
confirmed our conclusion that the influence of
increasing PD shares is larger than the influence of
ETOT.
It is possible that PD patients with ultrafiltration
failure switch to CCPD when using polyglucose to
extend time on PD treatment. This could have an
impact on costs because CCPD is more expensive
than CAPD. It is unknown whether ETOT would
result in a substantial increase in CCPD. The possible
cost impact has already been partially incorporated
into our cost estimates. The PD costs per patient per
year used in the cost estimates are an average of
CAPD and CCPD costs weighted by the age-specific
shares of CAPD and CCPD in total PD patients. The
question is whether ETOT changes these age-specific
shares. The relationship between ETOT and a switch
from CAPD to CCPD needs further investigation
before a possible impact on costs can be estimated.
This study concentrated on the effect the use of
polyglucose has on PD shares in total renal replace-
ment therapy. The results from the Markov modelling
indicated higher PD shares. Consequently, direct and
indirect effects of extended time on PD treatment
result in cost savings that can amount to 1.0% of the
total cost of dialysis treatment in The Netherlands.
Although the relative cost savings are low, the absolute
cost savings are substantial. These cost savings make
it possible to offer dialysis treatment to more patients
with equal budgets. In The Netherlands, the cost
savings would allow for 452 more HD patient years or
635 more PD patient years in the 10 year model period.
On a worl dwide scale, the dialysis budget is estimated
to be $100 billion [9]. In view of these high dialysis
costs to society, a saving of 1.0% can be considered to
be relevant for healthcare policy makers.
Acknowledgements. The authors would like to thank Martin
Nieuwenhuizen of Renine for his help with the data gathering and
analysis. Preliminary results from this study were presented at the
ERA-ETDA Conference, 17–20 September 2000, Nice; at the ASN
Conference, 13–16 October 2000, Toronto; and at the ISPD
Conference, 26–29 June 2001, Montreal. This work was supported
by an unrestricted grant from Baxter USA.
References
1. Wilkie ME, Plant MJ, Edwards L, Brown CB. Icodextrin 7.5%
dialysate solution (glucose polymer) in patients with ultrafiltra-
tion failure: extension of CAPD technique survival. Perit Dial Int
1997; 17: 84–87
2. Peeters P, Rublee D, Just PM, Joseph A. Analysis and
interpretation of cost data in dialysis: review of Western
European literature. Health Policy 2000; 54: 209–227
3. de Charro F, Ramsteijn P. Renine, a relational registry. Nephrol
Dial Transplant 1995; 10: 436–441
4. Ramsteijn PG, de Charro FTh, Geerlings W, Krediet RT,
Tjandra YI. Statistical Report 1998. The Development of the
Renal Replacement Program in The Netherlands. (In Dutch)
Renine, Rotterdam: 1998; 37–44
5. de Wit GA, Ramsteijn PG, de Charro FTh. Economic
evaluation of end stage renal disease treatment. Health Policy
1998; 44: 215–232
6. Peers EM, Scrimgeour AC, Haycox AR. Cost-containment in
CAPD patients with ultrafiltration failure. Clin Drug Invest 1995;
10: 53–58
7. Heimburger O, Wang T, Lindblom B. Alternations in water and
solute transport with time on peritoneal dialysis. Perit Dial Int
1999; 19 [Suppl 2]: S83–S90
8. Kawaguchi Y. National comparisons: optimal peritoneal
dialysis outcomes among Japanese patients. Perit Dial Int 1999;
19 [Suppl 3]: S9–S16
9. Klefter R, Nielsen B. A hundred billion dollars a year on
dialysis. How to calculate and what to compare. Int J Artif Org
1998; 21: 319–320
Received for publication: 1.2.02
Accepted in revised form: 19.9.02
396 T. J. G. Weijnen et al.
by guest on June 13, 2013http://ndt.oxfordjournals.org/Downloaded from
