| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Clinical |
Department of Nephrology,1 Hospital Universitario Fundación Alcorcón; Department of Nephrology,2 Hospital Universitario La Paz, Madrid; Department of Nephrology,3 Segovia General Hospital, Segovia; Hospital Universitario Fundación Alcorcón,4 Madrid, Spain,b
Correspondence to: J. Portolés, Department of Nephrology, Fundación Hospital Alcorcón, Avda Villaviciosa 1, Alcorcón 28922 Madrid, Spain. jmportoles{at}fhalcorcon.es
 |
ABSTRACT |
|---|
 TOP ABSTRACT PATIENTS AND METHODSRESULTSDISCUSSIONDISCLOSUREREFERENCES |
|---|
Objective: To study the prognostic factors for
mortality and hospital admission for patients on peritoneal dialysis
(PD). Method: Biannual data on individual characteristics,
clinical and analytical progress, treatment, and events were studied for a
cohort of incident patients undergoing PD (2003–2006) in a reference
area of 8.8 million people. Results: 489 patients (age 53.58 years, 61.6% male)
with 3-year follow-up were included. They presented at inclusion with Charlson
Comorbidity Index (CCI) of 5.25; previous cardiovascular (CV) event, 23.7%;
diabetes mellitus (DM), 19.1%; and hypertension (HT), 89.9%. Annual
hospitalization rate per patient-year at risk was 0.6. The variables that
predicted admission were CCI [odds ratio (OR) 1.14 per point], DM (OR 1.66),
and previous CV event (OR 1.90). Anemia maintained significance when corrected
for CCI: hemoglobin, 0.79 per 1 g/dL Hb; CCI, 1.15 per point. Annual mortality
rate was 5.4%. Those that died were older (67.47 vs 52.78 years) and had a
higher CCI (8.35 vs 5.0), a lower initial Hb (11.5 vs 12.2 g/dL), a higher
hospital admission rate, a higher annual rate of peritonitis, more previous CV
events (50.0% vs 22.1%), and higher prevalence of DM (38.5% vs 17.9%).
Survival analysis identified the following prognostic factors: CCI [hazard
ratio (HR) 1.51 per point], CV event (HR 2.85), DM (HR 2.52), age (HR 1.06 per
year), and mandatory referral to PD (HR 6.54). The effect of CV events and DM
persisted after correction for age, and that of choice of technique after
correcting for CCI and/or age. Conclusions: The CCI is useful for risk estimation in
PD patients. Previous CV event, DM, and age are the most relevant risk
factors. Control of anemia has prognostic value for hospital admissions.
Mandatory referral to PD is associated with higher mortality. The prognosis in
PD depends on predialysis patient management.
KEY WORDS: Cardiovascular event; dialysis choice; epidemiology; mortality.
Recent studies suggest that, during the first 2 years of follow-up, the survival rate of patients with chronic kidney disease (CKD) who begin peritoneal dialysis (PD) is the same as or better than that for those that begin hemodialysis (HD). However, the majority of these studies showed higher mortality rates in PD during the second year. Due to the difficulty in randomly assigning dialysis modality (1), these studies are mostly observational but provide valuable information on the prognostic factors for both techniques (2). In general, age and the presence of diabetes at the beginning of treatment are the main factors associated with mortality in PD.
The primary objective of the present study was to establish the prognostic factors for mortality and hospital admission in incident PD patients. The secondary objective was to describe the therapeutic management, progress, and characteristics of admissions, and the causes of death in this cohort.
|
PATIENTS AND METHODS |
|---|
|
TOPABSTRACTPATIENTS AND METHODSRESULTSDISCUSSIONDISCLOSUREREFERENCES |
|---|
The PD Center Group (GCDP) consists of 18 public hospitals treating a healthcare area of 8.8 million people in the central region of Spain. For 3 years (January 2003 to January 2006), all patients that began PD in this area were included and followed up until treatment discharge or death. Demographic parameters, etiology, comorbidity, origin, and elective or mandatory PD data were collected on admission. The Charlson Comorbidity Index (CCI) (3), a score that includes 16 comorbidity factors together with the patient's age, was used for comorbidity. Diagnoses of diabetes mellitus (DM) and previous cardiovascular (CV) events were based on clinical criteria: stroke, peripheral arterial ischemia, coronary artery disease, and NYHA grade-II or higher heart failure. Data for the following were collected at the beginning of the study and every 6 months: efficacy, residual renal function (RRF), peritoneal transport, objectives, and treatment for anemia and blood pressure control. Cases of peritonitis, hospital admissions, or discharges from the program were collected as they happened. Compliance with the efficacy objectives, anemia control, and blood pressure control according to the guidelines was recommended (4–7).
The design, management, and analysis of the database were done by the scientific committee without the participation of supporting companies. A data manager audited and edited the data by ranges and logical routines. Patients gave their informed consent at treatment inclusion. The statistical management and analyses were performed with SPSS software, version 11.0 (SPSS Inc., Chicago, IL, USA). The group discussed intermediate analysis annually.
The numeric variable data are shown as mean and standard deviation. Comparisons were performed by ANOVA or chi-square according to the nature of the variables. Certain numerical variables were categorized by tertiles (i.e., age, RRF, and CCI) for risk analysis comparing the upper and lower tertiles. Logistical regression was used to establish prognostic and risk odds for hospital admission. The Kaplan–Meier test for survival was applied, considering death as an event and exit from the program for any other reason (change in dialysis modality, recovery of renal function, transplant, or transfer) as censored. The Cox proportional hazards model was used to establish risk values, including the significant variables and those that significantly modified the coefficients of the model. Survival data are shown as mean survival probability and 95% confidence interval (CI). All rates obtained (for mortality, hospital admissions, and peritonitis) are reported in real time in each patient's treatment.
DESCRIPTION OF THE COHORT
We included in this analysis 489 patients treated between 2003 and 2006;
average follow-up was 13.36 (range 1 – 36) months. The most relevant
characteristics at inclusion were as follows: mean age 53.58 ± 16.1
years, 61.6% men, comorbidity measured by CCI 5.25 ± 2.54, 19.1%
diabetics, and 23.7% with a previous CV event
(Table 1). Before starting PD,
9.15% had suffered an acute myocardial infarction, 12.68% had peripheral
artery disease, 4.78% had suffered a stroke, and 6.44% had suffered episodes
of heart failure. During the study, 89.9% were hypertensive (82% on
treatment), 19.4% had a hemoglobin < 11 g/dL; 89.9% were on PD by choice
and the rest by order of the physicians; and 26.48% of the patients were on
the transplant waiting list during the first 6 months.
|
The most prevalent etiologies for CKD were glomerular 25.5%, diabetic nephropathy 16%, vascular ischemia 12.4%, interstitial 13.3%, and adult polycystic kidney disease 10.6%. The initial technique used was continuous ambulatory PD (CAPD) in 65.5% of patients and automated PD (APD) in the rest. Table 1 shows other cohort descriptive data.
By the end of follow-up, 21.4% of the patients had been transplanted, 5.5% had died, 7.4% had switched to HD, 2.1% had recovered enough RRF to discontinue PD, and the rest remained on PD. Only 4 patients (0.8%) were lost to follow-up.
|
RESULTS |
|---|
|
TOPABSTRACTPATIENTS AND METHODSRESULTSDISCUSSIONDISCLOSUREREFERENCES |
|---|
Comorbidity at initiation of treatment predicted hospital admission. The OR and 95% CI obtained were, for CCI 1.14 (1.06 – 1.23) per point; for DM 1.66 (1.05 – 2.63); and for previous CV event 1.90 (1.24 – 2.91). Neither gender nor etiology was associated with differences in hospitalization rate. Age did not predict admission considering it as a continuous variable, but it was at odds with the first to third tertiles (<46 and >62 years): 1.58 (1.02 – 2.46). Each gram per deciliter point above baseline hemoglobin was associated with a lower risk of admission 0.80 (0.69–0.90), which persisted when corrected for CCI (OR Hb 0.79 and OR CCI 1.15). The other parameters did not have prognostic value.
MORTALITY
Twenty-eight patients died during follow-up: calculated mortality rate was
5.2% per year at risk and estimated annual mortality rate was 5.4% with a 95%
CI of 3.7% – 7.8%. The probability of survival estimated by
Kaplan–Meier at first year was 95.5% and 86.6% at 2 years. The most
relevant causes of death were CV (42.8%), infectious (21.4%), withdrawal from
PD (7.1%), respiratory (3.6%), and gastro-hepatic (3.6%). The patients that
died had an admission rate 2.15 times higher (95% CI 1.54 – 3.0). The
patients that died were older and had more comorbidity, higher rates of DM and
previous CV events, and lower Hb at the beginning of treatment, and had
suffered more episodes of peritonitis with a higher number of admissions
(Table 1). We did not find
differences in other factors analyzed (gender, efficacy, treatment technique,
peritoneal transport, RRF, blood pressure control, obesity).
Patients that had changed from HD had a higher mortality rate than those that initiated renal replacement therapy with PD (11.5% vs 4.6%, p = 0.009). Also, those receiving PD through choice had a better prognosis than those forced to accept PD due to concomitant pathology (3.5% vs 20.4% mortality, p < 0.001). The main reasons for being obliged to accept PD were unavailable vascular access (57.1%), DM (10.2%), ischemic cardiopathy (8.2%), and poor HD tolerance (4.1%). Patients that chose the technique had a peritonitis rate of 0.46 episodes per year at risk, while those that were forced to accept the treatment had a rate 1.78 times higher (95% CI 1.26 – 2.46). Comparisons between those that chose and those that were forced to accept the treatment are detailed in Table 2.
|
The survival analysis identified patient-dependent prognostic factors such as comorbidity and being forced to accept PD. The most relevant comorbidity elements were age, diabetes, and previous CV events. The remaining elements included in the CCI were not individually significant (Table 3). The negative effect of DM or of previous CV events at the initiation of PD persisted after correction for age. The effect of dialysis treatment choice persisted after correcting for CCI and/or age (Figures 1 and 2).
|
|
|
For the multivariate analysis, several models were tested using Cox proportional hazards analysis. The best predictive model for mortality by Cox proportional hazards model included age, existence of previous CV events, and forced inclusion on PD. The HR and CI for each variable were, for age, 1.06 (1.02 – 1.09) per year; previous CV event, 2.4 (1.05 – 5.48); and forced inclusion in PD 6.05 (2.65 – 13.89).
The analysis by RRF tertiles at initiation (corrected for CCI) showed a tendency that did not reach statistical significance. The HR of the third tertile for RRF was 0.35 CI (0.11 – 1.10) and the CCI was 1.61 CI (1.36 – 1.90). The remaining elements evaluated at the beginning of treatment, such as gender, blood pressure control, dialysis dose, and peritoneal transport, were not statistically significant in univariate or corrected models.
|
DISCUSSION |
|---|
|
TOPABSTRACTPATIENTS AND METHODSRESULTSDISCUSSIONDISCLOSUREREFERENCES |
|---|
The inclusion of incident patients from a broad cohort that covered a reference population of more than 8 million people, and the prospective design with a follow-up time of up to 3 years, ensured the representative nature of the cohort. If we draw a comparison with a study carried out in the same area with patients on HD, our patients on PD were 11 years younger but presented a similar CCI (if we correct points for age in the index) (8).
The demographic characteristics of the population studied do not differ much from those described in the European literature (9–11). The average age and percentage of diabetic patients is similar to other European studies such as NECOSAD (20%) (9) and Heaf's Danish Registry study (19%) (12), and somewhat below Schaubel's Canadian study (25% diabetics) (13). These percentages are very different from the American registries, in which the proportion of diabetic patients is clearly higher (45% in the Vonesh et al. study) (14).
The rate of hospitalization in our series was lower than that of the USRDS registry (15), perhaps due to a lower comorbidity rate in our series. Cardiovascular disease was the main reason for admission in several published series and in the USRDS registry (15,16). Some studies have indicated that the rate of hospitalization increases with age and comorbidity, and is higher in diabetic patients and in women (17). In our study, the variables that were closely associated with a higher hospitalization rate were the CCI, previous CV pathology, and lower baseline Hb levels. This last factor persisted after correction for associated comorbidity, thus showing the importance of adequately treating anemia before beginning dialysis treatment. Other HD studies also found that baseline Hb has independent prognostic value (8). In fact, this would be the only risk factor found in our cohort that could be easily modified.
The majority of studies found similar or lower mortality rates for the first 2 years in patients undergoing PD and HD (fundamentally in nondiabetic patients or young diabetics) (12,14,18–20), with an increase in risk of mortality beginning in the second year (9,14,21), especially in patients over 65 years of age. The mortality rate found in our cohort was lower than that described by others, perhaps due to a short follow-up and to somewhat lower age and DM prevalence. In the NECOSAD study (3), which included 480 PD patients, patient survival during the first 2 years was found to be 80%. As in other studies (2,14,18,22–30), age and diabetes were factors independently associated with higher mortality. In the CHOICE study (20), however, with a follow-up time of up to 7 years, no differences between diabetic and nondiabetic patients were found. Some studies have suggested that patients with CV pathology have lower survival rates on PD than on HD (31,32). Cardiovascular pathology is frequent in patients that begin PD in our area and predicts not only mortality but also a higher rate of hospitalization. In other studies, lower risk of death due to stroke in nondiabetic patients on PD has been reported, but that risk was higher in diabetics (33).
None of the factors mentioned (age, diabetes, baseline CV comorbidity) are modifiable with medical intervention once PD is initiated. Early systematic prevention of CV pathology in the early phases of CKD would be the most effective method for reducing the high CV mortality in PD (34). As in other studies, we did not find any association between patient survival and the other factors studied (35). Residual renal function is another factor for better prognosis in the literature (36) that did not reach statistical significance in our cohort.
A recent meta-analysis (37) showed that baseline high transport status will identify patients with a worse prognosis. However, our study, as with other studies performed in our geographical area (38), did not confirm this. One must keep in mind that, in our country, we have all the technical resources available such as free use of APD and glucose polymers to avoid volume overload in patients with underlying ultrafiltration deficits.
Mujais and Story (22) found a lower mortality rate at 6 months in patients treated with APD compared to patients treated with CAPD. In our study, the type of dialysis technique did not influence patient survival; however, our percentage of patients on APD (33.5%) was much lower than in the other study (around 60%). In either case, nonrandomized treatment assignment makes it difficult to interpret its role in increased risk.
Our results showed a higher mortality rate in patients switching from HD compared to those whose first renal replacement therapy was PD, confirming the findings reported by Mujais and Story (22).
Adequate and individualized information about dialysis options is of great importance (39–41). The majority of our patients voluntarily chose PD treatment, but a small percentage of cases were referred to PD for medical reasons. An interesting note in this study was the finding that obligatory inclusion in PD, compared to patient choice, was the principal factor determining mortality rates. Patients that were forced to start PD had more comorbid conditions, but the effect persisted when adjusted for baseline comorbidity. The rate of peritonitis was also higher in the forced-PD group, suggesting the importance of patient involvement. Therefore, forced initiation of PD is a very high mortality risk factor in PD, whether or not preexisting factors (comorbidity, time spent on HD) or other modifiable factors, such as lack of patient cooperation, influence the choice of modality. Poor self-care in a daily home-treatment that has not been chosen may facilitate complications that would affect overall prognosis, and this should be specifically evaluated in patient training and follow-up.
The main limitations of the present study are the observational design and the low number of fatal events but, given the large number of patients analyzed, its statistical strength is high. Analysis of other factors that have also been associated with mortality, such as nutritional status and inflammation, were not included because the data were not collected uniformly at all sites.
In summary, the Charlson Comorbidity Index is a good predictive tool but it does not significantly improve the prognostic weight of age, the presence of DM, and previous CV disease combined. Prognosis on PD depends on patient management prior to initiation of the treatment. Systematic prevention of CV risk in CKD consultation before undergoing PD may reduce CV events and improve the results once in treatment. Anemia is the only risk factor that is independently modifiable and deserves special attention in the initial phases of PD. Patients that have no choice but to start treatment with this technique have a worse prognosis and it is independent of associated comorbidity. These patients may require special training programs and more attention.
|
DISCLOSURE |
|---|
|
TOPABSTRACTPATIENTS AND METHODSRESULTSDISCUSSIONDISCLOSUREREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
b These centers are included in the public research renal network REDinREN
(Instituto Carlos III de Investigación, Red 6/0016) and Instituto Reina
Sofía de Investigación Nefrológica.
Received 2 April 2008; accepted 4 July 2008.
|
REFERENCES |
|---|
|
TOPABSTRACTPATIENTS AND METHODSRESULTSDISCUSSIONDISCLOSUREREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  W. K. Lo ABSOLUTE FREE CHOICE FOR DIALYSIS MODALITY SELECTION -- IS IT POSSIBLE? Perit. Dial. Int., March 1, 2009; 29(2): 142 - 143. [Full Text] [PDF]
|
|
|
|
|
|
D. C. Mendelssohn INCREASING PD UTILIZATION: SHOULD SUITABLE PATIENTS BE FORCED? Perit. Dial. Int., March 1, 2009; 29(2): 144 - 146. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
