Impact of obesity on the evolution of outcomes in peritoneal dialysis patients.

BACKGROUND: Some studies reveal that obesity is associated with a decrease in mortality in haemodialysis (HD) patients. However, few studies have addressed the association between body mass index (BMI) and peritoneal dialysis (PD) patients.
METHODS: We performed this longitudinal, retrospective study to evaluate the impact of obesity on PD patients, using data from the Catalan Registry of Renal Patients from 2002 to 2015 (n = 1573). Obesity was defined as BMI ≥30; low weight: BMI <18.5; normal range: BMI = 18.5-24.99; and pre-obesity: BMI = 25-29.99 kg/m2. Variations in BMI were calculated during follow-up. The main outcomes evaluated were the technique and patient survival.
RESULTS: Obesity was observed in 20% of patients starting PD. We did not find differences in sex or PD modality, with the obesity group being older (65.9% are ≥55 years versus 59% non-obese, P = 0.003) and presenting more diabetes mellitus and cardiovascular disease (CVD) (47.9% obese versus 25.1% non-obese and 41.7% versus 31.5%, respectively). We did not observe differences in haemoglobin, albumin and Kt/V in obese patients. Regarding peritonitis rate, we did not find any difference between groups, presenting more peritonitis patients on continuous ambulatory peritoneal dialysis and aged ≥65 years [sub-hazard ratio (SHR) = 1.75, P = 0.000 and SHR = 1.56, P = 0.009]. In relation to technique survival, we found higher transfer to HD in the obese group of patients in the univariate analysis, which was not confirmed in the multivariate analysis (SHR = 1.12, P = 0.4), and we did not find differences in mortality rate. In relation to being transplanted, the underweight group, elderly and patients with CVD or diabetic nephropathy presented less probability to undergo kidney transplantation (SHR = 0.65, 0.24, 0.5 and 0.54, P < 0.05). Obese patients did not present differences in survival with weight changes but in normal-weight patients, a gain of 7% of the basal weight during the first year had a protective effect on death risk (hazard ratio 0.6, P = 0.034).
CONCLUSIONS: Obese and non-obese patients starting on PD had similar outcomes.

INTRODUCTION

The prevalence of obesity [defined as body mass index (BMI) ≥30 kg/m2] has increased in the last decades. In fact, the members of the Global Burden of Disease Obesity Collaborators reported that in 2015, 5% of children and 12% of adult population were obese [1]. In 2015, elevated BMI was related to 7.5% of deaths, with cardiovascular disease (CVD) being reported as the main cause of mortality in this population, followed by diabetes mellitus (DM), chronic kidney disease (CKD) and cancer [1]. Obesity is one of the principal modifiable cardiovascular risk factors (CVRF), at the same time being a risk factor for some chronic diseases such as DM, hypertension and CVD in the general population [2, 3]. However, obesity has been described as a protective factor for death in some groups of patients like elderly individuals in nursing homes, patients with some malignancies and hospitalized patients [4].

In the last decades, in the general population, there has been an increase of obesity in CKD patients and also in those who require renal replacement therapy (RRT), rising from 14.9% to 19.0% in haemodialysis (HD) patients, from 17.4% to 24.6% in peritoneal dialysis (PD) patients and from 11.2% to 16.1% in kidney transplantation (KT) patients from 2005 to 2016 in Catalonia [5]. Multiple epidemiological studies have demonstrated an inverse association between classic CVRF for CVD and mortality in HD patients [4, 6]. In addition, paradox obesity in HD patients has been described as a universal phenomenon that does not differ by sex, age, smoking, diabetic status, race/ethnicity, geographic regions or dialysis dose [7-10].

On the other hand, results among patients undergoing PD have been inconsistent. For a long period of time, obesity has been considered as a relative contraindication for the initiation of PD due to greater possibility of mechanical complications such as abdominal hernia [11], rapid decline in residual kidney function [12] and peritonitis [13]. In terms of technique and patient survival, Ahmadi et al. [14] performed a systematic review and meta-analysis because of the discrepancies observed between different groups and concluded that obese patients present a major risk to transfer to HD and lower mortality during the first year, but these differences in relation to mortality are not maintained over time.

In the present study, we have analysed the influence of obesity and the effect of variation of BMI on PD patients, in terms of technique and patient survival, with data obtained from the Catalan Renal Registry.

MATERIALS AND METHODS

After gaining the approval of the Institutional Review Board, we used data from the Registry of Renal Patients of Catalonia (RMRC). This is a mandatory population-based registry covering 7.5 million people that collects information on all patients with end-stage renal disease (ESRD) requiring RRT in Catalonia. At the time of starting RRT and at every change of treatment throughout RRT, a registration form is filled in. Every year an update has to be carried out and sent to the RMRC up to the finalization of RRT, death of patient or loss of follow-up.

A retrospective observational study has been carried out with the collection of data from patients starting PD in Catalonia from 2002 to 2015. Because the follow-up period is long, we have differentiated the data into two periods, from 2002 to 2006 and from 2007 to 2015. Patients who started PD during the first 90 days after beginning RRT who were resident in Catalonia, and aged >18 years and <85 years were included and followed-up from the start of RRT until December 2016.

Patients were classified in the following four groups depending on BMI at the moment of initiation of RRT: underweight (BMI <18.5); normal weight (BMI ≥18.5 and <25); pre-obesity (BMI ≥25 and <30) and obesity (BMI ≥30 kg/m2). A sub-analysis of obese patients in obesity Type I (30–34.9) and obesity Type II (≥35) was also carried out. To evaluate the change in weight, we calculated the percentage of change between basal weight and weight during the follow-up for each year and then the mean for the different periods was obtained: [(Weight at follow-up − Basal weight)/(Basal weight)]  × 100.

The collected variables were PD adequacy (Kt/V), nutrition status, which was assessed by serum albumin at the beginning of PD, accumulated probability of developing the first peritonitis based on the technique and initial weight, and also technique and patient survival. Related to technique survival, we considered the following events: death, transfer to HD or end of study period [31 December 2016 (censorship)]. Time for technique survival was evaluated from the start of the RRT to the end of the technique and with competitive risks (the different events compete with each other). To analyse patient survival for intention-to-treat, the time was evaluated from the beginning of the RRT until the death of the patient or time until the end of observation period (censorship). If a patient died within the first 90 days after transfer to HD, then the death was attributed to PD because we considered that this reflected the health status of patients during PD therapy. Transfer to HD was considered when patients were in HD for >90 days.

For the multivariate analysis, the following explanatory variables were taken into account: cause of ESRD, gender, CVRF (DM, age, hypertension and hypercholesterolaemia) and any cardiovascular event (ischaemic heart disease, heart failure, peripheral vascular disease and cerebrovascular disease).

Comparisons between groups by BMI at the time of initiation PD were performed by Chi-square test for categorical data and analysis of variance for continuous data (P < 0.05 was considered significant). Baseline characteristics of the study cohort were expressed as a number and a proportion or mean ± standard deviation (SD). The statistical approach to calculating the adjusted model was done using a generalized estimating equation, which is used to estimate the parameters of a generalized linear model with a possible unknown correlation between outcomes.

Cumulative incidence competing risk functions and competing risk regression were used to calculate the incidence of technique and patient mortality during PD. We also performed a survival analysis for intention-to-treat to evaluate the survival of all patients included in the study from the beginning of RRT, using the Kaplan–Meier in the univariate and the Cox regression in the multivariate analysis. All statistical tests were considered significant if P < 0.05 for two-tailed tests. Analyses were performed using STATA software version 13.

RESULTS

Baseline demographic and clinical characteristics

Among the 1573 patients included, weight and/or height information were not reported in 41 patients (2.6%); hence, the study population was 1532 incident patients on PD.

Baseline characteristics are reported in Table 1. Patients were divided in four groups depending on their BMI at the start of PD. There were 307 (20%) obese patients [232 patients (15.1%) obese Type I (BMI 30–34.9) and 75 (4.9%) obese Type II (BMI ≥35 kg/m2)]. The profile of an incident obese patient on PD is a man with diabetic nephropathy, aged >55 years with some cardiovascular morbidity who initiated PD in the recent period.

Table 1.

Baseline demographic and clinical  characteristics

Variable	Basal BMI					P-value
	Underweight (n = 35)	Normal (n = 631)	Overweight (n = 559)	Obesity (n = 307)	Total (n = 1532)
	n (%)	n (%)	n (%)	n (%)	n (%)
Period
2002–06	11 (31.4)	195 (30.9)	132 (23.6)	53 (17.3)	391 (25.5)	0.000*a
2007–15	24 (68.6)	436 (69.1)	427 (76.4)	254 (82.7)	1141 (74.5)
PD treatment
DPCC	19 (54.3)	354 (56.1)	294 (52.6)	163 (53.1)	830 (54.2)	0.648a
DPAC	16 (45.7)	277 (43.9)	265 (47.4)	144 (46.9)	702 (45.8)
Age, yearsb
<45	15 (44.1)	170 (28.1)	71 (13.3)	37 (12.8)	293 (20.0)	0.000*a
45–54	5 (14.7)	111 (18.3)	100 (18.8)	62 (21.4)	278 (19.0)
55–64	4 (11.8)	114 (18.8)	115 (21.6)	73 (25.2)	306 (20.9)
≥65	10 (29.4)	211 (34.8)	246 (46.2)	118 (40.7)	585 (40.0)
Sex
Men	13 (37.1)	418 (66.2)	406 (72.6)	200 (65.1)	1037 (67.7)	0.000*a
Women	22 (62.9)	213 (33.8)	153 (27.4)	107 (34.9)	495 (32.3)
Cause of CKD
Standard	14 (40.0)	284 (45.0)	194 (34.7)	89 (29.0)	581 (37.9)	0.000*a
DM	3 (8.6)	92 (14.6)	130 (23.3)	107 (34.9)	332 (21.7)
Others	18 (51.4)	255 (40.4)	235 (42.0)	111 (36.2)	619 (40.4)
Malignanciesc
No	30 (85.7)	575 (93.6)	502 (91.8)	284 (92.5)	1391 (92.5)	0.270a
Yes	5 (14.3)	39 (6.4)	45 (8.2)	23 (7.5)	112 (7.5)
Cirrhosis and liver diseased
No	29 (82.9)	602 (96.0)	534 (95.7)	297 (96.7)	1462 (95.7)	0.002*a
Yes	6 (17.1)	25 (4.0)	24 (4.3)	10 (3.3)	65 (4.3)
CVDe
No	21 (60.0)	465 (74.0)	350 (62.7)	179 (58.3)	1015 (66.4)	0.000*a
Yes	14 (40.0)	163 (26.0)	208 (37.3)	128 (41.7)	513 (33.6)
DM
No	31 (88.6)	472 (74.8)	327 (58.5)	133 (43.3)	963 (62.9)	0.000*a
Yes	4 (11.4)	159 (25.2)	232 (41.5)	174 (56.7)	569 (37.1)
BMI variation
Loss >4	3 (10.7)	94 (16.8)	149 (29.7)	82 (30.0)	328 (24.1)	0.000*
Loss 1–4	2 (7.1)	63 (11.3)	82 (16.3)	54 (19.8)	201 (14.8)
Remain	2 (7.1)	41 (7.3)	33 (6.6)	24 (8.8)	100 (7.3)
Gain 1–7	8 (28.6)	169 (30.2)	137 (27.3)	79 (28.9)	393 (28.9)
Gain >7	13 (46.4)	192 (34.3)	101 (20.1)	34 (12.5)	340 (25.0)
Dyslipidaemiaf
No	21 (63.6)	411 (70.7)	380 (74.2)	213 (76.9)	1025 (73.1)	0.139a
Yes	12 (36.4)	170 (29.3)	132 (25.8)	64 (23.1)	378 (26.9)
Laboratory data	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)
Haemoglobin	11.54 (1.88)	12.01 (1.72)	11.87 (1.46)	11.7 (1.35)	11.88 (1.57)	0.039*
Albumin	3.61 (0.72)	3.67 (0.51)	3.73 (0.47)	3.7 (0.44)	3.7 (0.48)	0.469
RCP	13.54 (29.21)	6.63 (18.01)	7.6 (15.5)	8.98 (17.01)	7.63 (17.31)	0.141

Chi-squared test.

Total = 1462.

Total = 1503.

Total = 1527.

Total = 1528.

Total = 1403.

DPAC, continuous ambulatory PD; DPCC, continuous cycling PD. Asterisks denote statistical significance.

Regarding weight variation from baseline (Table 1), 49.8 and 41.4% of obese patients had some weight loss and gain, respectively. Similar findings were observed in patients who were overweight, whereas weight gain was more frequent in patients who were underweight and normal weight at baseline (75 and 64.5%, respectively). In relation to laboratory data, all groups presented correct control of anaemia (>11 g/dL) and albumin according to the recommendations of clinical guidelines.

Adequacy parameters

Concerning Kt/V, all groups presented values >1.7 during the first 4 years of follow-up (Table 2).

Table 2.

Total Kt/V and evolution during 4 years of follow-up

1-year follow-up	Basal BMI					P-value
	Underweight	Normal	Overweight	Obesity	Total
	n = 29	n = 475	n = 437	n = 225	n = 1166
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)
ktv_dp_total	2.59 (0.74)	2.59 (0.78)	2.58 (0.90)	2.60 (0.73)	2.59 (0.82)	0.9851
ktv_dp_renal	0.84 (0.57)	0.85 (0.48)	0.86 (0.44)	0.89 (0.40)	0.86 (0.45)	0.6781
ktv_dp_peritoneal	1.75 (0.34)	1.74 (0.59)	1.72 (0.73)	1.71 (0.61)	1.73 (0.64)	0.9161
Second year follow-up	n = 21	n = 264	n = 267	n = 122	n = 674
ktv_dp_total	2.98 (0.68)	2.42 (0.78)	2.45 (0.74)	2.54 (0.80)	2.47 (0.77)	0.0091
ktv_dp_renal	0.94 (0.56)	0.69 (0.47)	0.79 (0.42)	0.84 (0.40)	0.77 (0.44)	0.0011
ktv_dp_peritoneal	2.03 (0.55)	1.73 (0.60)	1.66 (0.60)	1.70 (0.71)	1.70 (0.63)	0.0471
Third year follow-up	n = 14	n = 153	n = 151	n = 58	n = 376
ktv_dp_total	4.10 (6.10)	2.30 (0.69)	2.40 (0.88)	2.35 (0.63)	2.41 (1.35)	0.0001
ktv_dp_renal	0.65 (0.52)	0.64 (0.48)	0.74 (0.50)	0.77 (0.38)	0.70 (0.48)	0.1211
ktv_dp_peritoneal	3.45 (5.87)	1.66 (0.50)	1.65 (0.63)	1.58 (0.50)	1.70 (1.21)	0.0001
Fourth year follow-up	n = 6	n = 81	n = 74	n = 34	n = 195
ktv_dp_total	2.33 (0.99)	2.24 (0.69)	2.33 (0.60)	2.28 (0.52)	2.28 (0.64)	0.8391
ktv_dp_renal	0.52 (0.48)	0.59 (0.52)	0.75 (0.47)	0.74 (0.41)	0.68 (0.49)	0.0911
ktv_dp_peritoneal	1.81 (0.77)	1.65 (0.49)	1.57 (0.40)	1.53 (0.43)	1.60 (0.46)	0.3141

Clinical outcomes

Peritonitis episodes were reported at a rate of 0.25 during first 3 years of PD treatment. We did not find any difference related to BMI in patients who developed peritonitis during the first 3 years of follow-up (Table 3).

Table 3.

Summary of outcomes related to BMI group

Outcomes	Normal weight	Underweight	Overweight	Obesity
Risk of peritonitisa
Adjusted HR	1	0.77	1.03	1.06
95% CI	−	0.35–1.72	0.81–1.30	0.80–1.40
P-value	−	0.529	0.834	0.687
Undergoing KTb
Adjusted HR	1	0.65	1.14	1.13
95% CI	−	0.44–0.97	0.95–1.36	0.90–1.42
P-value	−	0.034	0.165	0.287
Transfer to HDc
Adjusted HR	1	1.12	0.98	1.12
95% CI	−	0.57–2.22	0.78–1.23	0.86–1.45
P-value	−	0.741	0.864	0.408
Mortality on PDd
Adjusted HR	1	1.42	0.76	0.8
95% CI	−	0.62–3.25	0.57–1.02	0.55–1.16
P-value	−	0.405	0.071	0.241
Patient survivale
Adjusted HR	1	1.76	0.78	0.81
95% CI	−	0.97–3.19	0.62–0.98	0.61–1.07
P-value	−	0.061	0.033	0.141

Adjusted by type of PD and age.

Adjusted by age, some CVD and primary kidney disease.

Adjusted by age and primary renal disease.

Adjusted by period, age, some CVD and primary kidney disease.

Patient survival until death, or time until the end of observation period. Adjusted by period, type of PD, age, some CVD and primary kidney disease.

CI, confidence interval; HR, hazard ratio.

KT was not reduced in obese incident patients on PD. The probability of undergoing KT was significantly lower in underweight patients, diabetic nephropathy, patients with history of CVD and the elderly population (Table 3 and Figure 1).

FIGURE 1

Probability of undergoing KT.

Concerning the probability of being transferred to HD, we found a higher incidence of transfers to HD in obese patients (23.73%) compared with patients with normal weight (19.65%, P < 0.05), but these differences disappeared when we adjusted the analysis by age (Table 3). Thus, in the multivariate analysis, the risk factors associated with transfer to HD were age [≥65 years sub-hazard ratio (SHR) = 1.86, P = 0.000] and diabetic nephropathy (SHR = 1.54, P = 0.002; Figure 2). To further analyse the survival of the PD technique, we performed an analysis including transfer to HD and death of patient and considering KT a competitive risk. In the univariate analysis, we found a reduced technique survival in the obese population (P = 0.015), but these differences disappeared in the multivariate analysis (Figure 3).

FIGURE 2

Multivariate analysis risk factors to transfer to HD.

FIGURE 3

Multivariate analysis risk factors to technique survival (transfer to HD and death of patient).

In terms of mortality in PD patients, we did not find differences in probability of death between the different BMI groups (Table 3). Patients who started PD in the period from 2007 to 2015 had a lower risk of death than patients starting PD from 2002 to 2006. Other risk factors for death were age, history of CVD and diabetic nephropathy (Figure 4). In the sub-analysis, which included grade of obesity, we found lower risk of death in obesity Grades II and III and a tendency to lower risk of death in obese Grade I compared with non-obese patients. It is important to note that only 75 patients were included in this group (Figure 5). In the intention-to-treat analysis, we evaluated how the initial BMI group affected PD incident patient survival even if they changed to TR or HD. Underweight patients presented worse patient survival (Figure 6).

FIGURE 4

Multivariate analysis of mortality.

FIGURE 5

Multivariate analysis of risk of mortality. Sub-analysis stratifying by types of obesity.

FIGURE 6

Intention-to-treat analysis of patient survival depending of BMI. DPAC, continuous ambulatory PD; DPCC, continuous cycling PD.

Finally, we analysed the probability of death in relation to BMI variations during the follow-up (Table 4). Interestingly, variations of BMI did not modify patient survival in the obesity group (Figure 7). However, in normal-weight patients, an increase of ≥7% in respect to the basal weight was found to be protective (Figure 8 and Table 4).

Table 4.

Summary of multivariate models of mortality and patient survival in relation to change in BMI

Changes in BMI	Same BMI	Loss 1–4	Loss >4	Gain 1–7	Gain >7
Probability of dying in obesea
Adjusted HR	1	1.12	0.45	1.16	0.78
95% CI	–	0.38–3.32	0.14–1.49	0.41–3.24	0.2–3.0
P-value	–	0.833	0.192	0.779	0.715
Probability of dying in non-obese (normal + overweight + underweight)b
Adjusted HR	1	0.59	0.57	0.59	0.64
95% CI	–	0.30–1.14	0.31–1.04	0.33–1.05	0.36–1.14
P-value	–	0.118	0.065	0.074	0.133
Survival in PD for intention-to-treat in obesec
Adjusted HR	1	1.35	1.00	1.27	0.40
95% CI	–	0.54–3.35	0.41–2.45	0.54–3.01	0.11–1.40
P-value	–	0.524	0.997	0.583	0.152
Survival in PD for intention-to-treat in non-obese (normal + overweight + underweight)d
Adjusted HR	1	1.00	0.97	0.76	0.59
95% CI	–	0.61–1.62	0.62–1.54	0.48–1.20	0.37–0.94
P-value	–	0.976	0.911	0.240	0.027

Adjusted by age.

Adjusted by period, age, CVD and CKD.

Adjusted for period, age, CKD and change of PD to HD or KT.

Adjusted for period, type of initial PD, age, CVD and change from PD to HD or KT.

CI, confidence interval; HR, hazard ratio.

FIGURE 7

Intention-to-treat analysis of patient survival depending of BMI variations in obese patients.

Intention-to-treat analysis of patient survival depending of BMI variations in normal weight patients. DPAC, continuous ambulatory PD; DPCC, continuous cycling PD.

DISCUSSION

Obesity is one of the principal modifiable CVRF in the general population [2, 3] and its protective effect in HD patients is well described [4, 6]. Nevertheless, the effect of obesity in PD population is unclear, and in Catalonia, there is not a standardized exclusion criterion with regards to obesity and PD inclusion. The main objective of our study was to analyse the relation between obesity and BMI variation with technique and patient survival in PD patients in Catalonia.

In the general population, there has been an increase of obesity in the CKD community during the last decades, affecting nearly 30% of incident patients on dialysis in the year 2002 in the USA [15]. Pliakogiannis et al. [16] published dates from the Canadian Registry highlighting a prevalence of 13.5% of obesity in PD from 1994 to 1998; Gilbertson et al. [17] observed that 22% of the PD population in the USA were obese from 1995 to 2000, McDonald et al. [18] described 17% of PD patients were obese in Australia and New Zealand from 1991 to 2002, and Qureshi et al. [19] illustrated a prevalence of 12% in Brazil from 2004 to 2007. In our cohort, obesity affected 20% of incident PD patients (307) from 2002 to 2015, and it is remarkable that the majority of cases started PD after 2007 compared with the period between 2002 and 2006 (82.7% of obese cases started PD after the year 2007) (P = 0.000).

As per the general population and as per data published by McDonald et al. [18] from the ANZDATA Registry, in our cohort of PD obese patients, we observed a higher prevalence of DM II and CVD (47.9% obese compared with 25.1% non-obese and 41.7% in obese compared with 31.5%, respectively) than in the non-obese population. On the other hand, despite finding more DM II in obese patients, Obi et al. [20] did not find more hypertension or risk of myocardial infarction, or other cardiac diseases, in the obese group compared with the normal weight one.

There are very few studies analysing PD adequacy in the obese PD population, and the results are conflicting. In 2002, Aslam et al. [21] did not find differences in the initial Kt/V of 104 patients with a high BMI (>27 kg/m2) compared with the control group of 104 patients with normal BMI [20-27] who were matched for age, gender, presence of DM and Charlson Comorbidity Index.

More recently, another group in the USA [22] published a single-centre experience in a small group of obese patients, including obesity Class III, and did not find any differences. In our study, we have analysed a large cohort of >1500 patients, 20% of them obese, and we did not observe differences between groups during the first and fourth year of follow-up, highlighting only a better Kt/V during the second and third year in the underweight group. It is important to analyse these data with caution, because this group was very small in our population, composed of only 28 and 23 patients during the second and third years of follow-up, respectively. One possible explanation for better Kt/V in underweight patients could be that this group of patients has a low body volume, and body volume is in the denominator in the Watson formula. As Akula et al. [22] published, one possible explanation for this is the fact that obese patients potentially present a larger abdominal surface participating in solid and fluid exchange, and that the fat tissue is not producing many uraemic toxins or participating in urea distribution volume.

Concerning peritonitis, McDonald et al. [23] analysed data from the ANZDATA Registry, including a large cohort of >10 000 patients who received PD, and recorded time to first develop peritonitis and episodes of peritonitis per patient-year over a 12-year period. In our study, we did not find differences between BMI groups but, in contrast, they found that higher BMI was associated with a shorter time to develop a first peritonitis episode, independent of other risk factors.

Obi et al. [20] studied peritonitis-related and non-peritonitis-related hospitalization and they detected higher incidence of peritonitis-related hospitalization across higher BMI categories in all adjustment models. In contrast , we did not find differences in time to develop a first peritonitis episode, even after stratifying patients into obesity Grades I–III.

Regarding KT, in contrast to the data published by Obi et al. [20], Lievense et al. [24] described that obese incident PD patients had the same likelihood of undergoing KT as the entire PD cohort. In that line, we did not observe differences in the probability of receiving a KT in obesity group.

In relation to technique survival, the obese group did not show more incidence of transfer to HD in the multivariate analysis as described by some authors previously in prospective observational [25] and multicentre studies [21]. In contrast to our results, data published in a meta-analysis [14] and some cohort studies [17, 18, 20, 26] showed more technique failure in obese patients. One possible explanation for our results showing better technique survival in the obese population could be our peritonitis rate, as peritonitis is one of the most important causes of technique failure. Most studies described obesity as a risk factor for peritonitis episodes and in our population, we did not find differences between BMI groups.

Ahmadi et al. [14] published a meta-analysis to analyse the association of BMI and mortality in PD patients. After excluding overlap data, they only included four papers and concluded that underweight patients were associated with higher 1-year mortality and being overweight with lower 1-year mortality. They also explained that although the association of obese patients with first-year mortality was not significant, both meta-analysed studies [17, 19] showed that being obese at baseline was associated with lower 1-year mortality. These differences in relation to mortality are not maintained over time. In our cohort, we did not find differences in probability of death between the four groups of BMI, but when we performed the sub-analysis comparing non-obese, obese Grade I and obese Grades II and III, we found less risk of death in obese Grades II and III (SHR = 0.30, P = 0.025) and a tendency of less risk of death in obese Grade I (SHR = 0.74, P = 0.052). It is important to note that there were only 75 patients included in this last group.

We also analysed the probability of death in relation to BMI variations during the follow-up and we did not find any difference. Qureshi et al. [19] also evaluated changes of BMI over time, and they observed a significantly higher mortality in normalized weight with a decrease ≥3.1%. In that line, in the intention-to-treat analysis, we found that underweight patients presented worse survival (SHR = 2.05, P = 0.014). Regarding variations in BMI, we did not observe any difference in the obese population; however, an increase of ≥7% respect to the basal weight represented a protective factor in non-obese patients on PD (adjusted hazard ratio = 0.59, P = 0.027).

It is important to note that obese patients had more DM and CVDs, and both entities related to higher mortality.

Our study has several strengths and limitations. The main limitation is that it was a retrospective study based on registry data; we do not have data about the type of dialysis fluid used (use of icodextrin and biocompatible or bioincompatible fluids), and as in most epidemiological studies, we only used BMI as an indicator of obesity. Although BMI has been accepted as one of the most reliable anthropometric indices for obesity, it has a limited ability to differ between muscle mass, adiposity and water. Many PD patients could be overhydrated, especially at the beginning of the treatment, and changes in water component could be related to changes in body weight. Despite the fact that most PD units use clinical criteria and bioimpedance in clinical practice to analyse the state of hydration, this article used retrospective cohort data, and these data are not available.

The main strength is that a large population of incident PD subjects was studied, and our results are in concordance with other published epidemiological studies; therefore, these results could be extrapolated.

In conclusion, we observed that the prevalence of obese PD patients is growing, and this entity is not related to worst outcomes. The obese PD population presented more prevalence of DM and CVD, but they did not have differences in adequacy parameters, risk of peritonitis, technique failure and probability of undergoing KT in terms of either mortality or patient survival. In addition, variations of BMI are not related to changes in mortality rate or patient survival in the obese population, but an increase of >7% in BMI in non-obese patients is supposed to be a protective factor.