Errata: Corrigendum Volume 16, Issue 3, 284–285, Article first published online: 21 May 2012
Professor Kunihiro Yamagata, Department of Nephrology, Faculty of Medicine, University of Tsukuba, 1-1-1 Ten-oudai, Tsukuba, Ibaraki 305-8575, Japan. Email: email@example.com
Early initiation of dialysis had been considered one of the most important methods for better prognosis of dialysis patients. One of the reasons for this was that long-term as well as short-term prognosis was poor with late initiation of dialysis. In this study, we analyzed the effects of residual renal function and comorbidity on both short- and long-term outcomes of ESRD patients. The subjects of this study were 20 854 patients who started renal replacement therapy (RRT) in 1989 and 1990, when we conducted national surveillance for new ESRD patients. The effects of glomerular filtration rate (GFR) at dialysis start and comorbidity conditions on survival were measured. Mortality hazard ratio (HR) was calculated using a Cox proportional hazard model. Multivariate analysis included pre-dialysis estimated GFR, age, sex, and underlying renal disease. The mean age of the subjects was 57.7 years old. Mean GFR at dialysis initiation was 5.00 mL/min per 1.73 m2 and GFR was significantly higher in patients with diabetes. The median survival time from the start of dialysis was 69 months, excluding subjects who died within 3 months; 1-year survival was 89.7%, while 2-year, 3-year, 5-year, 10-year, and 15-year cumulative survival rates were 79.3%, 71.1%, 57.8%, 37.3%, and 26.1%, respectively. For mortality risks, the higher the GFR at dialysis initiation, the worse the HR for mortality in both short-term and long-term prognoses by unadjusted analysis. However, after adjustments for age, gender, underlying renal diagnosis, and symptom at dialysis initiation, both late and early initiation of RRT did not affect long-term prognosis.
The number of patients who require renal replacement therapy (RRT) continues to increase, especially the number of elderly subjects with ESRD (1). To improve the outcome of these subjects, many believe that early initiation of dialysis is necessary. In 2006, Dialysis Outcomes Quality Initiative (DOQI) guidelines, based on expert opinion, recommended prompt initiation of dialysis at a Kt/V (urea) <2.0/week, equivalent to a glomerular filtration rate (GFR) of 10.5 mL/min per 1.73 m2(2). The values were based on observational evidence that starting dialysis “late”, defined by lower GFR, was associated with poorer nutritional state and appeared to increase morbidity and mortality (3)(4). Furthermore, late initiation of dialysis may result in not only short-term prognosis, but also 12-year long-term survival (5). However, given the complexity of the clinical conditions and the various factors that influence the decision to start dialysis (biochemical, psychological, financial, and social), there remains no specific set of circumstances that dictate when dialysis should begin. Furthermore, recent studies have suggested that early initiation of dialysis resulted in higher mortality and poorer prognosis (6). Those decreases of survival with higher GFRs at initiation could be partly explained by comorbidity.
In this study, we analyzed the effects of pre-dialysis renal function and comorbidity on short- and long-term outcomes from Japanese dialysis registry data. From this analysis, we can determine the effects of renal function and symptoms at initiation of dialysis on the short- and long-term outcomes of patients.
PATIENTS AND METHODS
The data of this study were obtained from the Japanese Society of Dialysis Therapy (JSDT) registry. The JSDT has been conducting an annual questionnaire survey of dialysis facilities throughout Japan since 1968, and several papers based on these surveys have been published (7)(1). The questionnaire comprises four pages, and the response rate for the first page is close to 100%. In addition to the regular questionnaire, incident patient survey (serum creatinine at the first pre-dialysis session, comorbid condition at the initiation of first dialysis) was conducted in 1988 and 1989. In this incident survey, 29 411 patients starting dialysis were identified from the JSDT registry. Of this cohort, 25 076 had serum creatinine recorded at the start of dialysis treatment. For this long-term study, we excluded the subjects under 18 or over 100 years of age, subjects with missing data regarding age, underlying renal diseases, and those lost to follow-up; finally, 20 854 patients were the subjects of this long-term outcome study.
The recorded variables at the initiation of dialysis were age, gender, underlying renal diagnosis, serum creatinine (sCr), and symptom or reason for dialysis initiation, including (i) nausea and appetite loss; (ii) congestive heart failure; (iii) intractable edema; (iv) oliguria; (v) neuropathy; (vi) unrecovered acute exacerbation of renal function; (vii) other uremic symptom; and (viii) no symptom. Renal function at the start of RRT was evaluated by estimated GFR (eGFR) using the Japanese equation factor based on Modification of Diet in Renal Disease Study (MDRD) formula (eGFR = 188Xserum creatinine−1.154 XAge−0.203X0.881 (X0.742, if female)) (8).
The primary outcome was mortality, with censoring at 31 December 2006, or at the date a patient received a transplant, withdrew, or died. The other objective was to evaluate eGFR at dialysis start, and also to determine if eGFR had an effect on survival.
Descriptive statistics are presented as mean with standard deviation or median with interquartile range, depending on the underlying distribution. Continuous and categorical variables were compared using the t-test (parametric variables) or the Wilcoxon rank sum test (non-parametric variables) and the χ2 test, respectively. Patient survival was estimated using the Kaplan–Meier method; median survival time and 95% confidence interval are presented. Survival curves by registry were compared using the log-rank test.
Cox proportional hazard model after confirming the proportionality was used to identify variables associated with mortality. We used five sets of three models: (i) 1-year mortality in all patients; (ii) 1-year mortality excluding subjects expired within 3 months; (iii) 5-year mortality in patients who survived on dialysis longer than 1 year; (iv) 10-year mortality in patients who survived on dialysis longer than 5 years; and (v) 15-year mortality in patients who survived on dialysis longer than 10 years. The first set of models included the GFR values before dialysis initiation. To explore the effect of GFR at dialysis initiation on survival, incident patients were classified into six groups according to GFR: 0–2 mL/min per 1.73 m2, 2–4 mL/min per 1.73 m2 (referent), 4–6 mL/min per 1.73 m2, 6–8 mL/min per 1.73 m2, 8–10 mL/min per 1.73 m2, and 10 mL/min per 1.73 m2 or greater. The second model, in addition to the above variable, included age, gender, and underlying renal disease. Underlying renal disease was classified into three categories. The third model adjusted for the symptom and comorbid condition at dialysis start in addition to the above variables.
P-values of <0.05 from two-sided tests were considered statistically significant. All statistical analyses were performed using SPSS, Version 17.0 (IBM Japan, Inc.).
Incident patient characteristics and outcome
The mean age of the subjects was 57.7 years old. Details of the subjects and underlying renal disease are shown in Table 1. Patients with diabetes were significantly older and showed a male predominance compared with other underlying renal disease patients. Up to 31 December 2006, 16 206 subjects died, among which 1606 subjects died within 3 months after starting dialysis, 253 subjects received renal transplantation, and 48 patients were withdrawn from dialysis treatment. Details of the number of subjects are shown in Figure 1. One-year survival of all patients was 82.7%, while excluding subjects who died within 3 months, it was 89.7%. The 2-year, 3-year, 5-year, 10-year, and 15-year Kaplan–Meier estimated cumulative survival rates were 73.2%, 65.6%, 53.3%, 34.1%, and 24.1%, respectively. Excluding subjects who died within 3 months, these rates were 79.3%, 71.1%, 57.8%, 37.3%, and 26.1%, respectively. Median survival time of the subjects was 69.0 months (95% CI: 67.2–70.8). Patients with chronic glomerulonephritis (CGN) showed significantly longer survival than those with other underlying renal diseases. Figure 2 shows Kaplan–Meier survival rate separated by underlying renal disease. Long-term prognosis of the patients with diabetic nephropathy (DMN) was also significantly worse than that of other underlying renal disease patients.
Indicates P < 0.05 between each. CGN, chronic glomerulonephritis; DMN, diabetic nephropathy; eGFR, estimated glomerular filtration rate.
Long-term survival by GFR, symptoms, and conditions at the initiation of dialysis
Table 2 shows GFR at dialysis initiation. GFR level was separated into six stages by 2-mL intervals. Subjects with less than 2 mL/min per 1.73 m2 were significantly younger than the subjects at the other stages. There were no significant age differences among GFR > 8 mL/min per 1.73 m2 groups. The mode of GFR at initiation of RRT was GFR 4–6 mL/min per 1.73 m2. Female predominance was observed in GFR < 4 mL/min per 1.73 m2 groups. The proportion of those with diabetes was also significantly lower in GFR less than 4 mL/min per 1.73 m2 groups. Figure 3 shows Kaplan–Meyer analysis of long-term survival separated by GFR at RRT initiation. The lower the initial GFR, the better the patients' prognosis. There were significant prognosis differences among GFR groups.
Table 2. Glomerular filtration rate (GFR) at dialysis initiation
GFR mL/min per 1.73 m2
Underlying renal disease (%DMN)
DMN, diabetic nephropathy.
Table 3 shows symptoms and comorbidity conditions at the initiation of dialysis. Patients who started without symptoms were significantly younger than the other patient group. The proportion of those with diabetes was higher in patients with congestive heart failure or intractable edema. Figure 4 shows Kaplan–Meier analysis of survival, separated by symptom at the start of RRT. The survival difference between congestive heart failure and unrecovered acute exacerbation of renal function was diminished after 5 years, while patients without symptoms at the start of RRT continued to show a higher long-term survival rate. We performed further analysis in subjects without symptoms at the dialysis initiation. Figure 5 shows the Kaplan–Meier analysis of survival separated by GFR at RRT initiation in subjects without symptoms. By univariate analysis, the GFR 6–8 mL/min per 1.73 m2 group and the GFR > 10 mL/min per 1.73 m2 group showed significantly higher hazard ratios (HR) than the GFR 4–6 mL/min per 1.73 m2 group (HR: 1.94, 95% CI: 1.10–3.39 in the GFR 6–8 mL/min per 1.73 m2 group; HR: 2.92, 95% CI: 1.69–5.07 in the GFR > 10 mL/min per 1.73 m2 group). After adjustment of patient age, gender, and underlying renal diseases, the GFR > 10 mL/min per 1.73 m2 group showed a significantly higher hazard ratio than the GFR 4–6 mL/min per 1.73 m2 group (HR: 2.05, 95% CI: 1.17–3.58).
Table 4 shows patient survival risks by Cox proportional hazard model analysis at 1 year after starting dialysis treatment. To compare the effect of renal function at dialysis initiation on survival, we selected the GFR 4–6 mL/min per 1.73 m2 group as a reference category. In model 1 (univariate analysis), the GFR < 2 and 2–4 mL/min per 1.73 m2 groups showed a significantly lower risk of death at 1 year, while the GFR > 6 mL/min per 1.73 m2 group showed 1.5 to 2.8 times more risk than the reference. After adjustment for underlying renal diseases, age, and gender (model 2), there was no HR difference between reference and GFR < 2 mL/min per 1.73 m2 groups. Furthermore, age and female gender had significant associations with risk of death. In model 3, patients with congestive heart failure, intractable edema, oliguria, and unrecovered acute exacerbation of renal function had significantly higher risks of death. After adjusting for these factors, the significantly higher mortality risk of females was diminished, and patients with GFR < 2 mL/min per 1.73 m2 did not show a survival difference from the reference group. Table 5 shows 1-year survival risks by Cox proportional hazard model analysis from 3 months to 1 year. In model 1, the HR of GFR < 4 mL/min/ per 1.73 m2 groups was slightly reduced. In model 2, there was no HR difference between those with diabetes and those with other underlying renal diseases. The significantly poorer prognosis of females after dialysis initiation was diminished by excluding the subjects who expired within 3 months. In model 3, there was no improved survival among GFR < 6 mL/min per 1.73 m2 groups and between eGFR 4–6 mL/min per 1.73 m2 and 8–12 mL/min per 1.73 m2 groups.
Table 4. One-year mortality risks by logistic regression analysis in all patients
Symptom at dialysis initiation (reference: no symptom)
Nausea, appetite loss
Congestive heart failure
Unrecovered acute exacerbation of renal function
Other uremic symptom
Table 6 shows the long-term effect of renal function at dialysis initiation on survival. After adjustment for age, gender, and primary renal diseases, late initiation of RRT (GFR < 4 mL/min per 1.73 m2 groups) did not affect patient survival, and also early initiation of RRT (GFR > 10 mL/min per 1.73 m2 groups) did not affect long-term survival.
Table 6. Logistic regression analysis for survival
5-year mortality risks by logistic regression analysis from 1 year to 5 years.
Number at risk
1- to 5-year survival rate
10-year mortality risks by logistic regression analysis from 5 years to 10 years.
Number at risk
5- to 10-year survival rate
15-year mortality risks by logistic regression analysis from 10 years to 15 years.
Despite the availability of several clinical guidelines for the timing of dialysis initiation (9–11), controversy remains about its optimal timing for patients without clinical symptoms or clinical indications, and the appropriate GFR level for initiation of dialysis is unknown. To improve the outcome of the ESRD population, many believe that early initiation of dialysis is one of the best methods. There are two main reasons for early dialysis initiation. The first is that, upon comparing 12-year survival rates of patients with early dialysis initiation (ED) (average creatinine clearance of the subjects: 12.9 mL/min) and those with late dialysis initiation (LD) (average creatinine clearance of the subjects: 2.1 mL/min), the survival rate of ED was 77%, whereas it was 51% in LD (5). Long-term exposure to uremic toxins may result in retention of uremia, and cause several uremic toxin-related complications (5). The second is that a residual renal creatinine clearance increase of 5 L/week per 1.73 m2 (6.94 mL/min per 1.73 m2) reduces relative 2-year mortality risk by 0.95, as determined by the CANUSA study (3). In this study, we compare the effects of the level of renal function at dialysis initiation on short- and long-term prognoses. We found that initial renal function significantly affects not only short-term prognosis but also long-term prognosis by unadjusted analysis. However, both short-term and long-term unfavorable effects were not observed in LD in this study. Furthermore, after adjustments for age, underlying renal diseases, gender, and symptom at dialysis initiation, the unfavorable effects of ED were diminished. The correlation between higher eGFR at dialysis start and increased mortality is consistent with recent observational analyses, which have failed to show a benefit from ED and hinted at a possible adverse association of high eGFR with mortality (6,12–15)(16). Lassalle et al. reported that age and patient condition strongly determine the decision to start dialysis and may explain most of the inverse association between eGFR and survival (17), and a randomized prospective trial for planned early vs. late dialysis initiation study (IDEAL study) suggested that planned early initiation of dialysis in patients with ESRD was not associated with increased mortality (18). Consequently, the appropriate timing of initiation of dialysis is just before the appearance of uremic encephalopathy, uremic pericarditis, congestive heart failure, intractable edema, oliguria, and malnutrition (6). The most important factor during consultation of predialysis chronic kidney disease patients is careful management to avoid the appearance of these symptoms when residual kidney function is reduced. Furthermore, a significantly higher survival rate was observed in subjects without symptom at dialysis initiation; those in the eGFR > 10 mL/min per 1.73 m2 group showed a significantly poorer prognosis after adjustment of covariate. Rosansky et al. reported that better kidney function at dialysis initiation was associated with higher mortality, and ED might be harmful from analysis of the healthiest group of patients undergoing dialysis (19). This was identical to the findings of our study. Further study is needed to elucidate the etiology of poorer prognosis in these subjects.
In this study, we also reported on the long-term prognosis of patients who started RRT in 1988 and 1989. Previous reports suggested that the annual mortality rate of Japanese dialysis patients was better than that of other countries (20). There were considered to be several reasons for this, such as racial differences, comorbidity differences, underlying renal disease differences, and renal transplantation acceptance rate differences. It is possible that younger and healthier patients often leave dialysis for renal transplantation; however, the renal transplantation acceptance rate is low in Japan, and younger and healthier patients often remain on dialysis and these subjects live longer on dialysis. This is one important reason for better overall dialysis outcomes in Japan. Actually, among 20 854 dialysis patients who initiated RRT in 1988 and 1989, only 1.2% of the subjects accepted renal transplantation. The average annual number of renal transplantations in Japan during the period was only 760 cases/year. During the same period, there were 12 325 cases/year in the USA (data source, Transplantation Communications of Japan). Furthermore, modality of peritoneal dialysis is very low, and approximately 97% of the subjects selected hemodialysis in Japan (1). Simple international comparison is difficult; however, the 10-year cumulative survival rate of 37.3% (from 90 days) was quite high compared with the USRDS 10-year survival probability of 20.4% (from 90 days) of incident ESRD patients in 1990 (21).
Our study has several limitations. First, a total of 14.7% of incident patients did not have a serum creatinine value recorded in the registry data. In addition, 8.6% of the patients were lost to follow-up. These patients were removed from the analysis. Second, serum creatinine was measured at different medical institutions, and in different laboratory settings, and we did not research the analytical method used for serum creatinine. Consequently, we could not perform calibration for serum creatinine assays. However, any errors introduced by the lack of calibration are likely to be small at the high serum creatinine values. The share for serum creatinine analysis by the Jaffe method in a major clinical laboratory test company was 76.3% in 1989 and 1990. Therefore, we used MDRD formula with Japanese coefficient factor, instead of Japanese eGFR equation (22), for estimation of GFR in this study. Third, we used serum creatinine-based estimation for GFR. There was a possibility that, for the subjects with higher eGFR, their true GFR was not reflected, but showed lowered endogenous creatinine production due to reduced muscle mass per BMI (sarcopenia or somatic malnutrition) (23). For further analysis, we should assess renal function by using insulin clearance, serum cystatin C-based formula, or means of urea and creatinine clearance from time urine collection.
In summary, we analyzed the effects of residual renal function and comorbidity on both short- and long-term outcomes in 20 854 patients who started renal replacement therapy in 1988 and 1989. The median survival duration from the start of dialysis was 69 months, excluding subjects who died within 3 months; the 1-year survival was 89.7%. The 2-year, 3-year, 5-year, 10-year, and 15-year cumulative survival rates were 79.3%, 71.1%, 57.8%, 37.3%, and 26.1%, respectively. For survival risks, the higher eGFR at dialysis initiation, the worse the hazard ratio of mortality in both short-term and long-term prognoses for unadjusted patients. These long-term unfavorable effects were diminished after adjustments for age, underlying renal diseases, gender, and symptom at dialysis initiation. Consequently, late dialysis initiation did not affect short-term or 10-year and more prognosis in Japan.
All the authors declare no competing interests. The authors wish to express the greatest appreciation to the Japanese Society for Dialysis Therapy, and the former chairmen of the committee (Drs M. Odaka, K. Sawanishi, K. Maeda, and T. Akiba), the principal investigators of all prefectures, and to all the personnel at the participating institutions involved in this survey. Without their tireless hard work, the compilation of data for this paper would not have been possible. Current members of the standing committee responsible for the Committee of Renal Data registry (Chairman, Y.T.) are: K.I., S.N., K.Y., T. Shinoda, O. Morita, I. Masakane, K. Suzuki, A. Wada, Y. Watanabe, Y. Itami, T. Shigematsu, T. Shoji, S. Marubayashi, T. Hamano, K. Tsuchida, K. Wakai, H. Nakamoto, N. Kimata, S. Ogata, M. Taniguchi, S. Hashimoto, T. Hasegawa, N. Hanafusa, N. Fujii, H. Nishi, J. Kazama, and Y.T.