J Nephrol (2017) 30:441–447

DOI 10.1007/s40620-016-0366-6

ORIGINAL ARTICLE

Survival and time-to-transplantation of peritoneal dialysis

versus hemodialysis for end-stage renal disease patients:
competing-risks regression model in a single Italian center
experience
Marta Rigoni1   · Emanuele Torri1,2 · Giandomenico Nollo1,3 · Diana Zarantonello4 ·
Alessandro Laudon4 · Laura Sottini4 · Giovanni Maria Guarrera5 · Giuliano Brunori4 

Received: 21 July 2016 / Accepted: 18 November 2016 / Published online: 29 November 2016
© The Author(s) 2016. This article is published with open access at Springerlink.com

Abstract  revealed a lower time-to-transplantation for PD versus HD

Aims  Despite several studies reporting similar outcomes patients (subhazard ratio: 2.34, p < 0.01, and 2.57, p < 0.01,
for peritoneal dialysis (PD) and hemodialysis (HD), the respectively). The waiting time for placement in the trans-
former is underused worldwide, with a PD prevalence of plant waiting list was longer in HD than PD patients (330
15% in Italy. In 2008, the Unit of Nephrology and Dialy- vs. 224 days, p < 0.01).
sis of the Healthcare Trust of the Autonomous Province of Conclusions  The extensive use of PD did not lead to any
Trento implemented a successful PD program which has statistically significant difference in mortality. Furthermore,
increased the proportion of PD incident patients from 7 to PD was associated with lower time to transplantation. PD
47%. We aimed to assess the effect of this extensive use may be a viable option for large-scale dialytic treatment in
of PD by comparing HD and PD in terms of survival and the advanced chronic kidney disease population.
time-to-transplantation.
Methods  A total of 334 HD and 153 PD incident patients Keywords  Dialysis survival · Hemodialysis · Advanced
were enrolled between January 2008 and December 2014. chronic kidney disease · Kidney transplantation · Peritoneal
After screening for exclusion criteria and propensity score dialysis · Competing-risks model
matching, 279 HD and 132 PD patients were analyzed.
Survival and time-to-transplantation were assessed by com-
peting-risks regression models, using death and transplan- Introduction
tation as primary and competing events.
Results  Crude and adjusted regression models for survival Hemodialysis (HD) and peritoneal dialysis (PD) are the
revealed the absence of significant differences between HD two common forms of dialysis therapy for end-stage renal
and PD cumulative incidence functions (subhazard ratio: disease (ESRD). Although PD is a well-established treat-
1.09, p = 0.62 and 1.34, p = 0.10, respectively). Differently, ment modality for advanced chronic kidney disease offer-
crude and adjusted regression models for transplantation ing several potential benefits, it is underused in Western
countries [1]. In 2008, only 11% of the dialysis population
was treated with PD worldwide [2]. PD prevalence varies
* Marta Rigoni
[email protected] significantly between regions. The proportion of dialysis
patients on PD reaches the 79% in Hong Kong [2]. Con-
1
Innovazione e Ricerca Clinica in Sanità – IRCS, Fondazione versely, PD initiation in the United States has traditionally
Bruno Kessler, Via Sommarive, 18, 38123 Trento, Italy
been low, never exceeding 15 to 16% of incident or preva-
2
Dipartimento Salute e Solidarietà Sociale, Provincia lent maintenance dialysis patients [1]. In Italy, recent data
Autonoma di Trento, Trento, Italy
from the Italian Study Group of Peritoneal Dialysis have
3
Biotech, Dipartimento di Ingegneria Industriale, Università di reported an incidence of PD modality of approximately
Trento, Trento, Italy
20% with a prevalence of 15% [3]. The underuse of PD
4
U.O. Nefrologia, APSS Trento, Trento, Italy does not seem justified in terms of safety and effectiveness;
5
Direzione Sanitaria APSS Trento, Trento, Italy it could be explained, instead, by a complex set of clinical,

13

442 J Nephrol (2017) 30:441–447

organizational, economic as well as patient-related fac- Methods

tors favoring HD use [4]. Indeed, data derived from both
national registries and large observational studies agree that Treatment center
the survival rate of patients treated with both modalities is
at least comparable [5, 6]. In 2008 the new clinical leadership of the Nephrology and
A randomized controlled clinical trial on dialysis modal- Dialysis Unit and the implementation of evidence-based
ity was attempted in the Netherlands, but only 38 of 773 policies led to the initiation of a large-scale PD program.
eligible subjects provided consent to be randomized to The program involved physician training in the use of PD,
receive the different treatment options [7]. Currently avail- systematic pre-dialysis information and education [14], and
able survival data are virtually all observational and thus clinical path standardization across the province. The man-
affected by confounding factors and limitations [8]. Indis- agement change was supported by promoting teamwork,
putable evidence for beneficial effects on survival can be collaboration, and performance feedback.
obtained only from randomized clinical trials, but it is
unrealistic to expect such an initiative in the near future [9]. Data collection and patients
Therefore, the acquisition of reliable and comprehensive
observational data is of the utmost importance to reliably This study was an observational, retrospective cohort study.
examine the clinical consequences of radical modifications Data were collected from 487 patients treated at the Unit
in dialysis therapy [9]. Over the last 10–15 years, as overall of Nephrology of the Santa Chiara Hospital in Trento
survival of dialysis patients has steadily improved and sta- from January 1, 2008, to December 31, 2014. Patients
tistical methods to analyze observational data have evolved, were observed until September 30, 2015, granting a mini-
a pattern of virtual equivalence in survival between patients mum follow-up period of 9 months. Patients decided to
on HD versus PD has emerged [8]. Previous studies showed undergo treatment on a voluntary basis, after a detailed
that age ≥65 years, frailty, cardiovascular disease, and informative discussion and clinical evaluation with physi-
diabetes mellitus could worsen survival outcomes in PD cians and nurses: 334 patients chose HD and 153 PD treat-
compared to HD patients [10–12]. However, focusing on ment. All patients gave informed consent for the collec-
transplantation outcomes, PD patients could benefited from tion and processing of data in anonymous form. Exclusion
a higher rate of kidney transplantation and shorter time-to- criteria were: previous transplantation, or death within 30
transplantation with respect to their HD counterparts [13]. days from the start of the treatment (i.e. in the adjustment
Further research is needed to improve PD uptake, and to to therapy period). Patients who chose to change dialysis
support appropriate and patient-centered decision-making treatment modality during the study period were censored.
in the real care context to improve patient survival and well- To correct the analysis for baseline covariates that could
being. The use of advanced statistical analysis techniques potentially affect the choice of dialysis modality, a propen-
may affect the quality of studies and has the potential to sity score model was implemented [16]. The estimation of
highlight the most effective strategies for improvement. the propensity score was performed with a logistic regres-
Since 2008, a concerted effort has been made in the sion model based on region of common support including
healthcare services of the Autonomous Province of Trento gender, age-group, cardiovascular disease, diabetes, chronic
to expand the use of PD through the implementation of a obstructive pulmonary disease (COPD), chronic liver dis-
capacity building program [14]. Following the introduc- ease, cancer, and hypertension as included variables. The
tion of this strategy, the incidence of PD patients progres- matching between HD and PD patients was done with the
sively increased to 47% in 2013. Significantly, the preva- Kernel Matching method. Distribution of baseline covari-
lence of PD patients grew to 20%, which is higher than the ates before matching was not uniform for diabetes mellitus,
Italian healthcare system average. Based on these results, while after propensity score matching it was uniform for
the present study was designed to analyze the outcomes all covariates. Demographic characteristics, survival, and
related to the increased use of PD over the 7-year program, comorbidities data were collected at the start of the dialytic
and compare PD and HD in terms of patient mortality and therapy and derived from the Provincial Register of Dialy-
time-to-transplantation. sis Patients.
To overcome limits due to the retrospective design of
the study, a propensity score model was implemented to Statistical analyses
assure matching of the two treatment groups. Furthermore,
survival data in ESRD patients were analyzed by compet- Survival analysis was performed using competing-risks
ing-risks regression [15], since in these patients death and regression models, according to the method of Fine
transplant act as competing events (i.e. the occurrence of and Gray [17]. The model is based on cumulative inci-
one event hinders or modifies the occurrence of the other). dence functions and subdistribution hazards risk (SHR)

13
J Nephrol (2017) 30:441–447 443

functions. For the survival study, the event of interest was Table 1  Characteristics of the study population after propensity
death and the competing event was transplantation. Crude score matching
and adjusted regression models were computed. Adjust- HD, n (%) PD, n (%) p
ment was performed for age-group (<65 vs. ≥65 years),
Total patients 279 132
cardiovascular disease, diabetes mellitus, and arterial
Males 194 (70) 90 (68) 0.78
hypertension.
Females 85 (30) 42 (32) 0.78
In order to evaluate the efficiency of the transplant
Cardiovascular disease 126 (45) 56 (42) 0.60
process (i.e. time to placement on the transplant waiting-
Diabetes mellitus 94 (34) 33 (25) 0.08
list, and time-to-transplantation), competing-risks regres-
COPD 44 (16) 20 (15) 0.87
sion models were calculated, considering transplant as
Chronic liver disease 29 (10) 15 (11) 0.77
the event of interest and death as the competing event.
Cancer 41 (15) 19 (14) 0.94
Crude and adjusted by age-group regression models were
Arterial hypertension 234 (84) 105 (80) 0.28
calculated. In the subgroup of patients on the transplant
Age at start of dialysis <65 years 107 (38) 56 (42) 0.43
waiting list (54 HD patients and 49 PD patients) differ-
Age at start of dialysis ≥65 years 172 (62) 76 (58) 0.43
ences in waiting time for placement in the list and in call
Median age at start of dialysis, years 69 [59–78] 69 [54–76] 0.20
time-to-transplantation after placement in the list were
[IQR]
assessed by Mann–Whitney U test. Data were expressed
as median and interquartile range (IQR) for not normally COPD chronic obstructive pulmonary disease, HD hemodialysis, IQR
distributed data, and number and percentage for categori- interquartile range, PD peritoneal dialysis
cal data. Categorical data were compared by the chi-
squared test and continuous data by the Mann–Whitney Survival analysis—death as event of interest
U test.
A p-value less than 0.05 indicated statistical signifi- Death occurred in 102 (37%) HD patients versus 46 (35%)
cance. All analyses were performed with Stata statistical PD patients. As shown in Fig. 1, the cumulative incidence
software, version 13.0 (StataCorp, Texas 77845 USA). of death was slightly but not significantly higher in PD
patients than in HD patients (SHR = 1.09, p = 0.62). The
risk of death in PD patients did not change significantly
(SHR = 1.34, p = 0.10) after adjustment for age-group, car-
Results diovascular disease, diabetes mellitus, and arterial hyper-
tension. Adjusted regression model values with covariate
Patient characteristics and clinical data contributions are reported in Table  2. Aging (≥65 years)
and cardiovascular disease increased the SHR of death,
Following screening for exclusion criteria and propensity while arterial hypertension decreased it. Diabetes mellitus
score matching, 279 HD patients and 132 PD patients had no statistically significant effect. However, the con-
were analyzed. The characteristics of the two groups of founders did not significantly modify the risk of death for
patients after the propensity score matching are shown PD patients in comparison to HD patients.
in Table 1. The two groups were uniform for each of the
considered features. The glomerular filtration rate at the Survival analysis—transplantation as event of interest
start of dialysis was between 4.5 and 6 ml/min (Chronic
Kidney Disease Epidemiology Collaboration formula). Transplantation occurred in 42 (32%) PD patients com-
The Kt/V ratio (urea clearance multiplied by treatment pared to 47 (17%) HD patients. As shown in Fig.  2, the
time/urea distribution volume) was maintained in the cumulative incidence of transplant was significantly higher
range of the Kidney Disease Outcomes Quality Initia- in PD patients than HD patients (SHR = 2.34, p < 0.01).
tive (KDOQI) guidelines (>1.2 per session in HD, 2.0 After adjustment for age-group, the SHR increased to 2.57
per week in PD). Regarding HD, 204 of 279 HD patients (Table  3). At multivariate analysis, age ≥65 years signifi-
(73%) were treated by fistulas or grafts, and 75 (27%) cantly decreased the likelihood of receiving a transplant
by tunneled central venous catheter. Regarding PD, 86 (Table 3).
of 132 PD patients (65%) adopted automated peritoneal
dialysis, while 46 (35%) practiced continuous ambulatory Time‑to‑transplantation
peritoneal dialysis. More than 90% of the patients were
put on dialysis after a period of pre-dialysis education A greater proportion of PD patients (49/132, 37%) were
and care at the hospital outpatient clinic. on the transplant waiting list than HD patients (54/279,
19%). The median time for placement on the transplant

13

444 J Nephrol (2017) 30:441–447

Fig. 1  Cumulative incidence function of hemodialysis (HD, continu- the competing event was transplantation. Adjustment was performed
ous line) and peritoneal dialysis (PD, dashed line) patients from 2008 for age-group, cardiovascular disease, diabetes mellitus, and arterial
to 2014, provided by crude (left) and adjusted (right) competing-risks hypertension
regression models. In the model the event of interest was death and

Table 2  Subdistribution hazard ratios of PD patients compared to waiting list was significantly longer (p < 0.01) in HD [330
HD patients according to the multivariate competing-risks regression days (IQR 222–663)] than PD patients [224 days (IQR
model 178–363)]. The median time to receive the call for trans-
Adjusted competing-risks regression—event of interest death plantation after placement in the waiting list was slightly
SHR (95% CI) p
but not significantly (p = 0.35) longer in HD [216 days (IQR
107–424)] than in PD patients [155 days (IQR 108–322)].
PD compared to HD 1.34 (0.95–1.91) 0.10
Age ≥65 years 4.22 (2.59–6.88) <0.01
Cardiovascular disease 2.08 (1.42–3.04) <0.01 Discussion
Diabetes mellitus 1.17 (0.82–1.65) 0.39
Arterial hypertension 0.46 (0.30–0.71) <0.01 Broadening the evidence using comparative studies of
CI confidence interval, HD hemodialysis, PD peritoneal dialysis,
dialysis outcomes and their impact on ESRD is essential to
SHR subdistribution hazard ratio drive changes in care patterns and to help the nephrology

Fig. 2  Cumulative incidence function of hemodialysis (HD, continu- regression models. In the model the event of interest was transplanta-
ous line) and peritoneal dialysis (PD, dashed line) patients from 2008 tion and the competing event was death. Adjustment was performed
to 2014, provided by crude (left) and adjusted (right) competing-risks for age-group

13
J Nephrol (2017) 30:441–447 445

Table 3  Subdistribution hazard ratios of PD patients compared to this finding. Indeed, PD patients showed a proportional and
HD patients according to the multivariate competing-risks regression slightly higher incidence of death than HD patients during
model
the whole treatment course, but without any statistically
Adjusted competing-risks regression—event of interest transplanta- meaningful difference.
tion Aging (≥65 years) and cardiovascular disease increased
SHR (95% CI) p the SHR of death. However, these confounders did not
significantly modify the risk of death in PD patients with
PD compared to HD 2.57 (1.68–3.92) <0.01
respect to HD patients. The issue of aging and cardiovas-
Age ≥65 years 0.06 (0.03–0.12) <0.01
cular disease in PD patients is still debated in the litera-
CI confidence interval, HD hemodialysis, PD peritoneal dialysis, ture [20]. Some studies conducted with different methods
SHR subdistribution hazard ratio reported a higher risk of death in PD patients than HD
patients, which increased with aging and cardiovascular
disease [10, 21, 22]. In contrast, Buemi et al. reported that
community reflect critically on its practices. The present in elderly and cardiac patients, PD was actually preferable,
study aimed to evaluate the effects of an increased use because in comparison with HD it reduced the hemody-
of PD in the Autonomous Province of Trento in terms of namic stress experienced by the patient and the incidence
survival and time to transplantation. This was performed of hypotension [23].
within a framework of outcome assessment to improve Although diabetes is recognized as a confounding vari-
the quality of care and quality of life of patients. In this able able to affect survival in PD versus HD patients, our
study we analyzed data from 279 HD patients and 132 PD study did not result in an increased risk of death for ESRD
patients over a period of 7 years with a minimum follow- patients. A recent systematic review analyzed mortality
up of 9 months. Since the study was an observational ret- outcomes in diabetic patients who underwent HD or PD.
rospective cohort study, the power of the results was poten- The analysis of 25 observational studies led to the conclu-
tially limited. sion that the available evidence was inconsistent, because
It is well known that the comparison of two different survival varied across study designs, follow-up periods, and
treatments under real life conditions (i.e. outside the con- patient subgroups [12].
text of a randomized clinical trial) is affected by the ‘con- Our results showed that hypertension decreased the
founding by indication’ problem, which needs to be prop- risk of mortality in ESRD patients, but it did not affect
erly addressed. Therefore, we constructed a propensity the survival differences between the two dialytic treat-
score model [16] to match PD and HD patients. The score ments. The explanation for this finding is still controver-
allowed us to analyze the baseline covariates that could sial. Some reports have indicated a paradoxical associa-
potentially affect the choice of dialysis modality and thus to tion between hypertension and mortality in hemodialysis
balance the risk of positive selection of the PD population. patients. According to this, a normal to low blood pressure
After propensity score matching we obtained two large seems associated with poor outcome, whereas high pres-
patient groups, which were homogenous regarding clinical sure potentially confers survival advantages, a phenomenon
characteristics, modality and length of treatment. termed ‘reverse epidemiology’ [24].
The comparative study was performed using competing- Hence, confounders analysis in a real clinical context
risks regression models. Accounting for competing risk showed that major clinical complications did not change
events permits the simultaneous analysis of outcomes and the cumulative incidence of death in the two patient groups
avoids overestimation of cumulative incidences [15]. In our in a meaningful way. Indeed, the two treatment modalities
study, the competing-risks model was applied twice: con- displayed a substantial equivalence, apart from a small non-
sidering death as the event of interest and transplantation significant negative trend in the PD group.
as the competing event, and vice versa. Survival analysis With respect to transplantation, the literature shows that
did not reveal any evidence of difference between PD and PD commonly registers a higher rate of kidney transplanta-
HD in terms of mortality. Consistently, crude and adjusted tion [25]. According to data from the Italian national reg-
regression models for survival revealed no significant dif- isters, this result seems to be due to the younger age of PD
ference in terms of cumulative incidence functions between patients and the higher prevalence of first dialysis experi-
patient groups. ence [26]. In our study, where the number of patients who
Although disputed by a recent analysis [18], the litera- could undergo PD was enlarged and patient age was com-
ture generally reports better survival for PD patients com- parable between the two groups, PD nonetheless had a
pared to HD patients during the first years of treatment reduced time-to-transplantation. Crude and adjusted regres-
[19]. Our study performed on two samples equivalent for sion models for transplantation revealed a significantly
frailty and complex disease characteristics did not confirm lower time-to-transplantation for PD compared to HD.

13

446 J Nephrol (2017) 30:441–447

According to our analysis, this finding seems mainly Acknowledgements  This work was partially funded by the Autono-
attributable to the reduction in the time to be placed on the mous Province of Trento (the Healthcare Research and Implementa-
tion Program at Fondazione Bruno Kessler).
waiting list for PD patients. The tendency to reduce time to
transplantation for PD patients was confirmed by the call Compliance with ethical standards 
time to transplant, although the difference in this case was
not significant. This finding could be related to a different Conflict of interest  We have read and understood the Journal of
profile of PD patients. Patients who undergo this method Nephrology’s policy on disclosing conflicts of interest and declare that
tend to be more empowered, and to pursue their care plan we have none.
by themselves. Having a strong social support network and Informed consent  Informed consent was obtained from all individ-
being functionally able is strongly associated with choosing ual participants included in the study.
PD [27]. This attitude can be translated into a more effi-
cient treatment and a better planning of the examinations Ethical approval  All procedures performed in studies involving
necessary for inclusion in the transplantation waiting list. human participants were in accordance with the ethical standards of
Moreover, these patients have to manage their own treat- the institutional and/or national research committee and with the 1964
Helsinki declaration and its later amendments or comparable ethical
ment daily, which can serve as a strong incentive to obtain standards. For this type of study formal consent is not required, since
quick inclusion in the list. Vice versa, the frequent hospi- it is retrospective.
tal admissions required for HD may negatively affect the
planning of the examinations necessary for inclusion in the Open Access  This article is distributed under the terms of the
transplantation waiting list. Creative Commons Attribution 4.0 International License (http://
Although not statistically significant, the HD slower creativecommons.org/licenses/by/4.0/), which permits unrestricted
time in call-to-transplantation could be partially explained use, distribution, and reproduction in any medium, provided you give
appropriate credit to the original author(s) and the source, provide a
by the fact that these patients had a greater tendency to ane- link to the Creative Commons license, and indicate if changes were
mia and more likely required blood transfusion with greater made.
use of erythropoiesis-stimulating agents due to an increase
in panel reactive antibody [28]. The presence of antibodies
generated by frequent transfusions may therefore hinder the
finding of a matching organ for transplantation, prolonging References
the waiting time in HD patients.
Overall our findings should be taken into account to 1. Khawar O, Kalantar-Zadeh K, Lo WK, Johnson D, Mehrotra
improve clinical practice and management of modality R (2007) Is the declining use of long-term peritoneal dialy-
sis justified by outcome data? Clin J Am Soc Nephrol CJASN
choice. This may be part of a general improvement strat- 2(6):1317–1328
egy aimed to maximize quality of life, patient-reported out- 2. François K, Bargman JM (2014) Evaluating the benefits of
comes, and cost-savings [29, 30]. home-based peritoneal dialysis. Int J Nephrol Renov Dis
7:447–455
3. Zaza G, Rugiu C, Trubian A, Granata S, Poli A, Lupo A (2015)
How has peritoneal dialysis changed over the last 30 years: expe-
rience of the Verona dialysis center. BMC Nephrol 16:53
Conclusions 4. Desai AA, Bolus R, Nissenson A, Bolus S, Solomon MD, Kha-
war O et  al (2008) Identifying best practices in dialysis care:
results of cognitive interviews and a national survey of dialysis
The aim of this study was to assess outcomes of survival providers. Clin J Am Soc Nephrol CJASN 3(4):1066–1076
and time-to-transplantation in the presence of an exten- 5. Yeates K, Zhu N, Vonesh E, Trpeski L, Blake P, Fenton S (2012)
sive use of PD in patients with ESRD in the Autonomous Hemodialysis and peritoneal dialysis are associated with similar
Province of Trento. Our data did not show differences outcomes for end-stage renal disease treatment in Canada. Neph-
rol Dial Transplant 27(9):3568–3575
in long-term survival between patients treated with PD 6. Kao T-W, Huang J-W, Hung K-Y, Chang Y-Y, Chen P-C, Yen C-J
and with HD. Differently, PD patients seemed to benefit et al (2010) Life expectancy, expected years of life lost and sur-
in terms of reduced time for placement on waiting list to vival of hemodialysis and peritoneal dialysis patients. J Nephrol
transplantation. 23(6):677–682
7. Korevaar JC, Feith G, Dekker FW, Van Manen JG, Boescho-
Our findings support the wider adoption of PD thanks ten EW, Bossuyt PM et al (2003) Effect of starting with hemo-
to its feasibility, and the evidence of positive outcomes and dialysis compared with peritoneal dialysis in patients new on
far-reaching benefits for patients. These reported results dialysis treatment: a randomized controlled trial. Kidney Int
and the analysis performed should help in devising future 64(6):2222–2228
8. Teixeira JP, Combs SA, Teitelbaum I (2015) Peritoneal dialysis:
studies with similar methods of competing risk survival update on patient survival. Clin Nephrol 83(1):1–10
analysis and management strategies to achieve an improve- 9. Ikizler TA (2012) Intensive hemodialysis: back to the beginning?
ment in the outcomes of patients undergoing dialysis care. J Am Soc Nephrol JASN 23(4):573–575

13
J Nephrol (2017) 30:441–447 447

10. Winkelmayer WC, Glynn RJ, Mittleman MA, Levin R, Pliskin of dialysis initiation. PLoS ONE. http://www.ncbi.nlm.nih.gov/
JS, Avorn J (2002) Comparing mortality of elderly patients on pmc/articles/PMC3948631/
hemodialysis versus peritoneal dialysis: a propensity score 21. Johnson DW, Dent H, Hawley CM, McDonald SP, Rosman JB,
approach. J Am Soc Nephrol JASN 13(9):2353–2362 Brown FG et al (2009) Association of dialysis modality and car-
11. Liu J-H, Chen J-Y, Lin S-Y, Lin H-H, Ting I-W, Liang C-C et al diovascular mortality in incident dialysis patients. Clin J Am Soc
(2013) Comparing Survival between peritoneal dialysis and Nephrol CJASN 4(10):1620–1628
hemodialysis patients with subclinical peripheral artery disease: 22. Han SS, Park JY, Kang S, Kim KH, Ryu D-R, Kim H et  al
a 6-year follow-up. Int J Med Sci 10(4):434–440 (2015) Dialysis modality and mortality in the elderly: a meta-
12. Couchoud C, Bolignano D, Nistor I, Jager KJ, Heaf J, Heim- analysis. Clin J Am Soc Nephrol CJASN 10(6):983–993
burger O et  al (2015) Dialysis modality choice in diabetic 23. Buemi M, Lacquaniti A, Bolignano D, Donato V, Fazio MR,
patients with end-stage kidney disease: a systematic review of Campo S et al (2008) Dialysis and the elderly: an underestimated
the available evidence. Nephrol Dial Transplant 30(2):310–320 problem. Kidney Blood Press Res 31(5):330–336
13. Fuquay R, Teitelbaum I (2012) Transplant outcomes and dialysis 24. Kalantar-Zadeh K, Kilpatrick RD, McAllister CJ, Greenland S,
modality. Contrib Nephrol 178:251–257 Kopple JD (2005) Reverse epidemiology of hypertension and
14. Zarantonello D, Laudon A, Dalprà A, Fratton F, Giovannini L, cardiovascular death in the hemodialysis population: the 58th
Zulian I et al (2015) Identikit of the patient who chooses the PD. annual fall conference and scientific sessions. Hypertension
Experience of the Autonomous Province of Trento. G Ital Nefrol 45(4):811–817
32(1):1–9 25. Snyder JJ, Kasiske BL, Gilbertson DT, Collins AJ (2002) A
15. Noordzij M, Leffondré K, van Stralen KJ, Zoccali C, Dekker comparison of transplant outcomes in peritoneal and hemodialy-
FW, Jager KJ (2013) When do we need competing risks meth- sis patients. Kidney Int 62(4):1423–1430
ods for survival analysis in nephrology? Nephrol Dial Transplant 26. Italian Register of Dialysis and Transplantation: http://ridt.sin-
28(11):2670–2677 italy.org
16. Austin PC (2011) An introduction to propensity score methods 27. Chanouzas D, Ng KP, Fallouh B, Baharani J (2012) What influ-
for reducing the effects of confounding in observational studies. ences patient choice of treatment modality at the pre-dialysis
Multivar Behav Res 46(3):399–424 stage? Nephrol Dial Transplant 27(4):1542–1547
17. Fine JP, Gray RJ (1999) A proportional hazards model for
28. Wetmore JB, Peng Y, Monda KL, Kats AM, Kim DH, Brad-
the subdistribution of a competing risk. J Am Stat Assoc bury BD et al (2015) Trends in anemia management practices in
94(446):496–509 patients receiving hemodialysis and peritoneal dialysis: a retro-
18. Merchant AA, Quinn RR, Perl J (2015) Dialysis modality and spective cohort analysis. Am J Nephrol 41(4–5):354–361
survival: does the controversy live on? Curr Opin Nephrol 29. Roggeri A, Roggeri DP, Zocchetti C, Bersani M, Conte F, ReNe
Hypertens 24(3):276–283 (Renal Lombardy Network) et  al (2016) Healthcare costs of
19. Kumar VA, Sidell MA, Jones JP, Vonesh EF (2014) Survival the progression of chronic kidney disease and different dialysis
of propensity matched incident peritoneal and hemodialy- techniques estimated through administrative database analysis. J
sis patients in a United States health care system. Kidney Int Nephrol. doi:10.1007/s40620-016-0291-8
86(5):1016–1022 30. Karopadi AN, Mason G, Rettore E, Ronco C (2013) Cost of peri-
20. Heaf JG, Wehberg S (2014) Relative survival of peritoneal
toneal dialysis and haemodialysis across the world. Nephrol Dial
dialysis and haemodialysis patients: effect of cohort and mode Transplant 28(10):2553–2569

13