Long-term blood pressure control in a cohort of peritoneal dialysis patients and its association with residual renal function

Transcription

1 Nephrol Dial Transplant (2001) 16: Original Article Long-term blood pressure control in a cohort of peritoneal dialysis patients and its association with residual renal function Murali K. Menon, David M. Naimark, Joanne M. Bargman, Stephen I. Vas and Dimitrios G. Oreopoulos Toronto Western Hospital, University Health Network, Toronto, Ontario, Canada Abstract Background. Hypertension is the prime contributor for cardiovascular mortality in the dialysis population. Peritoneal dialysis (PD) has been thought to improve blood pressure (BP) control in the short term, but the long-term benefits are not conclusively proven. We aimed to evaluate the degree of BP control in PD patients in the long term and analyse the factors associated with poor control. Methods. Data of all patients who were initiated on PD at one centre between July 1994 and July 1998 and completed at least 1 year of PD were analysed retrospectively at initiation of PD, at 6 months, and annually thereafter until 5 years or until discontinuation of therapy. Hypertension was defined as per WHOuISH criteria. A Blood Pressure Control Index was empirically defined to account for the effect of antihypertensives on measured BP. Factors associated with poor BP control were analysed. Results. Out of 207 patients (age 57.0"16.0 years, 103 male, 104 female) 91.3% were hypertensive at the start of PD. About 33.8% had diabetic nephropathy. Systolic and mean arterial pressure index improved in early phase reaching a nadir between 6 months and 1 year followed by steady progressive worsening through out the rest of follow up. On multiple linear regression analysis age (P-0.001), duration of hypertension prior to dialysis (P-0.001), and declining residual renal function, expressed as both average of urea and creatinine clearance (P s 0.002) and residual urine output (P-0.001) were independently associated with poor BP control. Diabetes (P s0.836), peritoneal transport (DuP 4 of creatinine at start) (Ps0.218), peripheral oedema (P s 0.479) and dose of erythropoetin (P s 0.488) were not associated. Conclusions. Initiation of PD results in early improvement of hypertension in end-stage renal disease (ESRD). Correspondence and offprint requests to: D. G. Oreopoulos, University Health Network, The Toronto Western Hospital, 399 Bathurst Street, Suite 6EW-539, Toronto, Ontario M5T 2S8, Canada. dgoreopoulos@msn.com BP control thereafter deteriorates steadily with time and this is associated with age, duration of hypertension, and declining residual renal function. This suggests that hypertension in ESRD patients is a progressive disease primarily related to falling glomerular filtration rate, the preservation of which might improve BP control and possibly modify cardiovascular risk. Keywords: blood pressure; hypertension; peritoneal dialysis; residual renal function Introduction Cardiovascular disease is the leading cause of mortality and morbidity in end-stage renal disease (ESRD) patients undergoing dialysis and renal transplantation [1], and this is most likely related to long-standing hypertension along with abnormal lipid profiles, ventricular hypertrophy and dysfunction as well as a high level of sympathetic activation [2]. Initiation of peritoneal dialysis (PD) often results in excellent blood pressure (BP) control [3], possibly better than with haemodialysis (HD). However the improvement is always not sustained [4] and many patients become progressively hypertensive as years go by. Better fluid volume control and greater clearance of vasoconstrictor factors compared with HD have been considered responsible for the initial improvement [3] whereas late decline of peritoneal ultrafiltration and fluid retention has been implicated as potential causes of long-term deterioration in BP control [3,5]. However, there is a specific subset of patients in whom the hypertension is volume independent [6]. Progression of pre-existing essential hypertension with ageing, erythropoietin administration, declining residual renal function, hyperparathyroidism, retention of digitalis like substances, etc., which are potential mechanisms of hypertension in the ESRD population [2] may also be possible causes for this late deterioration. # 2001 European Renal Association European Dialysis and Transplant Association

2 2208 M. K. Menon et al. In the present paper, we report on the temporal behaviour of BP in a cohort of PD patients over a 5-year period and analyse the relationship between various factors including residual renal function and BP in that cohort. Methods Data was retrospectively gathered on 207 patients who were started on PD between July 1994 and July 1998 at Toronto Western Hospital. Patients were included in the sample if they had completed at least 1 year of continuous follow up on PD. Subjects were excluded from the analysis if they died, were transplanted or were transferred to HD within 1 year of starting PD. Information was collected by reviewing patients clinic charts and missing laboratory data were retrieved from the computerized health data management system of the hospital. Information collected at the initiation of PD included age, gender, underlying renal disease, presence or absence of diabetes, presence or absence of hypertension (defined according to WHOuISH criteria: systolic blood pressure (SBP) P140 mmhg or diastolic blood pressure (DBP) P90 mmhg or current antihypertensive therapy) and its duration, weight, peripheral oedema score (0, none; 1, mild; 2, moderate; 3, severe), presence or absence of heart disease, left ventricular (LV) hypertrophy, LV mass and mass index. Residual renal function was assessed by urine output, residual creatinine clearance and residual urea clearance (from 24 h urine collections). The peritoneal transport characteristics of each patient were assessed within 2 months of the initiation of PD by the dialysate to plasma creatinine ratio at 4 h in a standard peritoneal equilibration test (PET). Total K t uv for urea and total creatinine clearance were computed from PD fluid, urine and plasma analysis using PD Adequest software (Baxter Healthcare Corporation, Chicago, IL, USA). Other clinical data collected at the initiation of PD included the number of antihypertensive medications, dose of erythropoietin, and number of lipid lowering drugs. The laboratory data collected at initiation included haemoglobin, haematocrit, parathyroid hormone, albumin, cholesterol, LDL-cholesterol, HDL-cholesterol and triglyceride concentrations. Similar clinical and laboratory information was collected serially at 6 months, 1 year and annually thereafter for up to 5 years from the date of initiating PD or until individuals either died, were transplanted or were transferred to HD or another centre. The same parameters of residual renal function as described above were collected. In the PD clinic, weight was recorded by the patient s primary nurse. BP was measured with a standard mercury sphygmomanometer in supine position. The degree of peripheral oedema was assessed by nurses and was assigned a score, which ranged from 0 (none) to 3 (severe) and validated by the attending physician. The list of antihypertensives (furosemide and metolazone were not categorized as antihypertensives) and lipid lowering drugs and other drugs as well as dose of erythropoietin were checked by the nurse during each clinic visit and documented. Twenty fourhour urine collections, for estimating urea and creatinine clearance, were done with each clinic visit and patients were instructed on the proper collection technique. The dates of cardiac events and outcomes and the dates of discontinuation of PD were noted in the clinic chart by the primary nurse. An important methodological problem in evaluating BP control related to dialysis modality per se in dialysis patients is the confounding effect of antihypertensive therapy. The patient who requires four antihypertensive agents in order to achieve a normal BP reading and another who has normal BP while on no drugs clearly do not have the same degree of BP control. In other words, BP as an outcome measure is obscured by the use of concurrent antihypertensive medications. In order to capture the effect of antihypertensives on prevailing BP we defined a new index, SBP index, which was calculated as: [SBPŠ 3½1q number of antihypertensivesš: A mean arterial pressure index ( MAP index ) was computed in a similar fashion as: [(0:3333 3SBPÞ qð0:6667 3DBPÞŠ 3½1q number of antihypertensivesš: The indices were defined empirically just to capture the effect of the two dependent variables namely BP and number of antihypertensives. Statistical analysis Means and standard deviations were computed for continuous variables and numbers and percentages for binary and categorical variables. Multiple linear regression analyses were undertaken in order to explore the independent contribution of various factors to BP. In these analyses, MAP index was the dependent variable and age at the initiation of PD, presence of diabetes, duration of hypertension prior to dialysis, residual renal function, the initial dialysate to plasma (DuP) creatinine ratio, presence or absence of peripheral oedema and dose of erythropoietin were the independent variables. Here, MAP index values and other repeated measures for clinic visits for the same patient were treated as independent observations. A priori, there were no biologically plausible reasons to suspect that these factors should interact and so only the main effects were considered in the model. Three separate multiple regression analyses were performed in which the residual renal function was expressed as the 24-h urine output, the residual creatinine clearance and the average of the residual creatinine clearance and urea clearance, respectively. The regression coefficients were estimated by the method of unweighted least squares. Wald tests were conducted in order to determine whether the regression coefficients were significantly different from zero. In all multivariate analyses, a P-value of -0.05um (Bonferroni correction, where m is the number of independent variables in the model) was considered to be significant. All statistical analyses were performed with SPSS v 8.0 (Chicago, IL, USA). Results Descriptive data on the 207 patients at the initiation of PD is shown in Table 1. The mean age was 57.0"16.0 years. There were 103 males and 104 females. There were 77 (37.2%) diabetics. In 70 (33.8%) patients diabetic nephropathy was the most probable cause of ESRD. Type I diabetes accounted for 28 and type II for 49 cases. Most patients (ns189 or 91.3%) were hypertensive at the time of initiation of PD. In

3 BP control in PD patients in association with residual renal function Table 1. Baseline characteristics of patients at initiation of PD (total number of patientss207) Variable Mean"SD or percentage Age (year) 57.0"16.0 Malesufemales (%) 49.8u50.2 Patients with diabetes (%) 37.2 Patients with hypertension (%) 91.2 Duration of hypertension 9.7"9.2 prior to PD (years) Patients with LV hypertrophy (%) 42.0 Patients with heart disease (%) 69 (33.3) SBP (mmhg) 151.1"23.6 DBP (mmhg) 82.5"12.7 SBP index 386"149 MAP index 268"101 Haemoglobin (gul) 94.2"16.6 Albumin (gul) 38.3"6.05 PTH (pmolul) 22.6"19.7 Total cholesterol (mmolul) 5.05"1.39 LDL cholesterol (mmolul) 3.09"1.18 Triglycerides (mmolul) 2.05"1.25 Initial 4 h DuP ratio 0.68"0.11 Initial weekly K t uv 2.7"0.6 Residual creatinine 5.69"4.13 clearance (mlumin) Residual urea clearance (mlumin) 2.66"2.16 Residual urine output (mluday) 634"496 Human recombinant erythropoietein 2787"4224 dose (Uuweek) Number of antihypertensives 1.54"0.9 Patients taking lipid-lowering drugs (%) 14.0 Table 3. Cardiac disease at start of PD 2209 Type of disease Number (%) LV hypertrophy* Any heart disease Coronary artery disease LV dysfunction Congestive heart failure Valvular heart disease Atrial fibrillation Others *Based on two-dimensional ECG in 113 patients and ECG in 94 patients. Table 2. Aetiology of ESRD Fig. 1. Mean arterial pressure over time from initiation of peritoneal dialysis. Number of patients whose data are available at a given time is shown just above the axis. Cause Number (%) Diabetes Hypertension Glomerulonephritis (unspecified) FSGS Interstitial nephritis Lupus PCKD Membranous nephropathy Miscellaneous Unknown Total (23.1%) patients hypertension was listed as the cause of ESRD. The break up of aetiology of ESRD is shown in Table 2. The prevalence of heart disease at initiation of PD was 33.3% (ns69). The type of heart disease is shown in Table 3. MAP, when plotted against time, showed an initial steady decline up to 1 year followed by rise and then fall towards the end of follow up (Figure 1). The number of antihypertensives showed an initial decline followed by a steady progressive increase (Figure 2) throughout follow up. When the composite of the above two parameters, namely the systolic and MAP indices, when averaged over all patients and plotted Fig. 2. Mean antihypertensive use over time from initiation of peritoneal dialysis. Number of patients whose data are available at a given time is shown just above the axis. over time, a definite U shaped pattern (Figure 3) was apparent. There was an initial improvement in these indices with a nadir occurring between 6 months and 1 year followed by a subsequent worsening. The initial decline was more apparent in SBP index while the subsequent upward slope was similar with both systolic and MAP index (Figure 3). The residual renal function whether defined as the 24-h urine output,

4 2210 M. K. Menon et al m 1y 2y 3y 4y 5y Fig. 3. Blood pressure control indices over time from initiation of peritoneal dialysis. Number of patients whose data are available at a given time is shown just above the axis. the residual creatinine clearance, or the average of the residual creatinine and urea clearances exhibited a progressive decline over the observation period (Figure 4) up to 4 years. After this there was an upward slope in the curve, which was primarily related to drop out of patients with poor renal function and subsequent decrease in sample size. This was evident by the widening standard deviation. To determine whether the upward trend in BP indices with time could also be related to patient drop out and not an actual deterioration we plotted both MAP index and residual renal function parameters of different batches of patients who had completed various periods of follow up against time (Figures 5 and 6). It showed that MAP index had a U shaped distribution with all patient groups whereas residual renal function clearly showed steady decline in all groups after first year. The results of the multiple regression analyses in which the MAP index was the dependent variable and age, diabetes, duration of hypertension prior to starting dialysis, residual renal function, presence or absence of peripheral oedema, DuP of creatinine at 4 h in the initial PET and erythropoietin dose were the independent variables are shown in Table 4. The numeric values of the regression coefficients and resulting Wald statistics were different depending on whether residual renal function was expressed as urine output, residual creatinine clearance or the average of the residual creatinine and urea clearances. However, the qualitative conclusions from these analyses were similar: age, duration of hypertension prior to the initiation of dialysis and the residual renal function were each independently associated with the MAP index (Table 4). Discussion Although there is abundant data in the field of hypertension in dialysis patients there is insufficient precise information available to recommend treatment strategies [2]. Fig. 4. Residual renal function over time from initiation of peritoneal dialysis. Number of patients whose data are available at a given time is shown just above the axis. MAP Index Mean y 1y 2y 3y 4y 5y Fig. 5. MAP index of groups of patients who completed various periods of follow-up time: 1 year (s), 2 years (h), 3 years (^), 4 years (m), 5 years (m); mean (...). This shows the U-shaped pattern of blood pressure control was evident in all groups of patients y 1y 2y 3y 4y 5y Fig. 6. Residual creatinine clearance of various groups of patients completing different periods of follow-up showed a steady decline after the initial year in all groups. 1 year (s), 2 years (h), 3 years (^), 4 years (m), 5 years (m); mean (...).

5 BP control in PD patients in association with residual renal function Table 4. (A) Multiple linear regression analysis of factors affecting MAP index. (B) Multiple linear regression analysis of factors affecting MAP index. (C) Multiple logistic regression analysis of factors affecting MAP index 2211 Variable b SE (b) t P value A Constant Age (per year) Diabetes (present vs absent) HTN duration (per year) Residual creatine clearance (per luday) Extra cellular fluid expansion (present vs absent) DuP ratio Erythropoietein (per U) B Constant Age (per year) Diabetes (present vs absent) HTN duration (per year) Average creatinine clearance and urea clearance (per luday) ECF expansion (present vs absent) DuP ratio Erythropoietein dose (per U) C Constant Age (per year) Diabetes (present vs absent) HTN duration (per year) Residual urine output (per mluday) Extra cellular fluid expansion (present vs absent) DuP ratio Erythropoietein dose (per U) This is especially distressing because cardiovascular disease remains the leading cause of death in patients undergoing dialysis and transplantation [1] and a significant proportion of this is most likely related to long-standing hypertension. Many studies have suggested that uncontrolled hypertension correlates with adverse outcome in ESRD patients [7] and normal BP is associated with better survival [8] but it is not clear whether attaining a normal BP with aggressive pharmacotherapy has the same benefit as achieving the same target without multiple medications. As almost 90% of ESRD patients are hypertensive by the time of initiation of dialysis [2], achieving normal BP without medications would actually mean attaining this with dialysis alone. Although some authors believe that hypertension can be controlled in the vast majority of dialysis patients by maintaining the dry weight alone [9,10] general experience points to the contrary. However, the benefits of long slow dialysis on BP control [8] are possibly not attributable to maintenance of dry weight alone. The 1996 core indicators project funded by the health care financing administration (HCFA) reported that 53% of haemodialysis patients and 29% of PD patients had hypertension when values )150 and 90 mmhg, systolic and diastolic respectively, were considered abnormal [11]. If more stringent criteria, as recommended by JNC VI [12], were used, a much higher proportion would have been hypertensive. A particularly troubling methodological issue in studies assessing BP control in dialysis patients is the confounding effect of concurrent use of antihypertensive agents. The degree of PB control of a patient who requires multiple antihypertensives in order to achieve a normal BP and another who has normal BP while on no drugs was clearly not identical. Although studies have used the number of antihypertensives as a surrogate for the severity of hypertension [3,5,13], no study has effectively captured these two dimensions of BP control, namely measured BP and number of antihypertensive agents simultaneously. In the current study, we introduce a new parameter, blood pressure control index which multiplies a BP measure either the systolic or MAP by the number of antihypertensive medications plus one. These indices, although empirical, have the advantage of capturing the dynamic aspects of BP control in groups of subjects. Early studies had recognized the favourable effect of PD in controlling hypertension [14]. However, it was soon realized that the beneficial effect was not sustained [3,4] and many patients had a resurgence of hypertension when followed over longer periods of time. The effect of late worsening of hypertension was difficult to judge from some of these studies because of lack of information on antihypertensive use [4]. In another study the measured BP was not actually rising but the antihypertensive use went up [3], still the upward trend was apparent. More recent studies on BP in PD have been cross sectional [13] and have reported

6 2212 M. K. Menon et al. a higher prevalence of hypertension in PD population although they did not examine changes in BP control over the long term. In the current study, we demonstrate a clear U shaped trend of hypertension control in patients on PD. The antihypertensive use steadily increased with time although measured BP tended to fluctuate after the initial decline. This is not surprising because physicians treat BP aggressively and a steady worsening hypertension may indicate ineffective medical therapy rather than deteriorating BP control! The disparity in the two parameters underscores the necessity for a composite parameter which represents BP control. The early improvement in BP is likely to be related to improved fluid volume control as the systolic pressure indices improved better than MAP. This has been recognized by many earlier investigators [3,4]. We found that subsequently hypertension progresses steadily and this was similar for both MAP and systolic pressure. The late decline has been at least partly attributed to fluid retention resulting from loss of peritoneal ultrafiltration [3,5]. Such ultrafiltration failure has a cumulative incidence of up to 30% at 2 years [5]. Thus, it can explain worsening hypertension only in a subset of patients. In the current study, we did not find clinically assessed extracellular fluid volume overload as an important factor accounting for the worsening of BP control in the multivariate analysis. However, clinical assessment of fluid volume status has limitations and alternate methods like bio-electrical impedance analysis (BIA), inferior vena caval diameter [15] and even chemical markers like cgmp and ANP [6,15] have been used as surrogates of fluid status though these methods also have their own limitations. High peritoneal transport characteristic poses a risk for poor ultrafiltration [16] but we did not find this to be an independently associated factor in BP control. We had examined the DuP 4 of creatinine at baseline in all patients and a proportion of patients could have changed their transport characteristics over time. This could not be demonstrated as periodic PETs were not done. Use of erythropoetin has been associated with high BP in PD patients. There may be a correlation with rising haematocrit [2] which implicates viscosity as a cause and worsening HTN has also been observed independent of changes in haematocrit or haemoglobin where the effect of loss of hypoxic vasodilatation [2,17] may be operative. We did not find dose of erythropoietin to be an independent risk factor for worsening BP control. Diabetes accounted for nearly one-third of patients in our series. It has been shown that diabetes as co-morbidity significantly increases the cardiovascular risk in dialysis patients [18]. In a multicentre Italian study comparing 301 diabetic vs 1689 non-diabetic patients undergoing CAPD cardiovascular disease was twice as frequent and relative risk of death 2.13 times higher (P-0.001) [19] among diabetics. However, it is not clear whether diabetes increases the risk of hypertension as even non-diabetic ESRD patients already have an extremely high prevalence of hypertension. Our results show that diabetes was not an independent risk factor for worsening hypertension. The factors that were independently associated with worsening BP control were age, duration of hypertension prior to initiation of PD, and residual renal function. We evaluated three parameters of residual renal function, namely creatinine clearance, average of urea and creatinine clearance and urine volume separately along with other factors in a multivariate analysis. Markers of solute clearance and urine volume were evaluated in separate analyses because these parameters are interdependent. Largescale studies have demonstrated inverse relation between BP and glomerular filtration rate (GFR). The Modification of Diet in Renal Disease study group demonstrated that the prevalence of hypertension varied inversely with GFR (from 66% at a GFR of 83 mluminu1.73 m 2 to 95% at a GFR of 12 mluminu 1.73 m 2 ) in a cohort of 1795 patients with chronic renal insufficiency [20]. Our study demonstrates that this inverse relation is valid even at lower levels of GFR. In other words, hypertension in ESRD patients is a relentlessly progressive disease. Initiation of PD leads to short-term control of the situation and accounts for the initial downward trend in the U shaped curve. As time advances with loss of residual renal function dialysis therapy fails to hold back the progressing disease resulting in the upward trend in the U shaped curve. As the other factors like age and duration of hypertension are totally unmodifiable every effort should be made to maintain the residual renal function in ESRD patients on PD by avoidance of nephrotoxic drugs and other insults. In conclusion, BP in ESRD patients treated with PD improves initially reaching a nadir by 6 months to 1 year followed by steady worsening over the years. Age, duration of hypertension prior to initiation of dialysis and residual renal function (average of urea and creatinine clearance as well as urine volume) are independently associated with such late decline. Every effort should be made to preserve the residual renal function as this may improve BP control and favourably impact cardiovascular morbidity and mortality. References 1. United States Renal Data System. USRDS 2000 Annual Data Report. US Department of Health and Human Services. The National Institute of Health, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD, June Malliox LU, Haley WE. Hypertension in ESRD patients: pathophysiology, therapy, outcomes and future directions. Am J Kid Dis 1998; 32: Saldhana L, Weiler EWJ, Gonick HC. Effect of continuos ambulatory peritoneal dialysis on blood pressure control. Am J Kidney Dis 1993; 21: Lamiere N. Cardiovascular risk factors and blood pressure control in continuous ambulatory peritoneal dialysis. Perit Dial Int 1993; 13: S394 S395

7 BP control in PD patients in association with residual renal function 5. Slingeneyer A, Canaud B, Mion C. Permanent loss of ultrafiltration capacity of peritoneum in long term peritoneal dialysis. An epidemiological study. Nephron 1983; 33: Lang SM, Wolfram G, Gerzer R et al. Characterization of subtypes of hypertension in CAPD patients by cyclic guanosine monophosphate. Perit Dial Int 1999; 19: Foley RN, Parfrey PS, Harnett JD et al. Impact of hypertension on cardiomyopathy, morbidity and mortality in end-stage renal disease. Kidney Int 1996; 49: Chara B, Calemard E, Laurent G. The importance of treatment time and blood pressure control in achieving long term survival on dialysis. Am J Nephrol 1996; 16: Zucchelli P, Santoro A, Zuccala A. Genesis and control of hypertension in hemodialysis patients. Seminar Nephrol 1988; 8: Charra B, Calemard E, Laurent G. Control of hypertension and prolonged survival on maintenance hemodialysis. Nephron 1983; 33: Rocco MV, Flanagan MJ, Beaver S et al. Reports from the 1995 core indicators for peritoneal dialysis study group. Am J Kidney Dis 1997; 39: Joint National Committee on Detection, Evaluation and Treatment of High Blood Pressure. The Sixth Report of the Joint National Committee on Detection, Evaluation and Treatment of High Blood Pressure. NIHuNHLBI. Arch Int Med 1997; 157: Cocchi R, Espoti EG, Fabbri A et al. Prevalence of hypertension in patients on peritoneal dialysis: results of an Italian multicentre study. Nephrol Dial Transplant 1999; 14: Flynn CT, Hibbard J, Dohrman B. Advantages of continuous ambulatory peritoneal dialysis to the diabetic with renal failure. Proc Europ Dial Transplant Assoc 1979; 16: Leunissen KM, Kouw P, Kooman JP et al. New techniques to determine fluid status in hemodialyzed patients. Kidney Int 1993; (Suppl 41): S50 S Tsamaloukas AH, Saddler MC, Murata GH et al. Symptomatic fluid retention inpatients on continuous peritoneal dialysis. J Am Soc Nephrol 1995; 6: Heidenreich S, Rhan KH, Zidek W. Direct vasopressor effect of recombinant erythropoetin on renal resistance vessels. Kidney Int 1991; 39: Maiorca R, Cancarini GC, Brunori G et al. Morbidity and mortality of CAPD and hemodialysis. Kidney Int 1993; (Suppl 40): S4 S Viglino G, Cancarini GC, Catizone L et al. Ten years experience of CAPD in diabetics: comparison of results with non-diabetics. Italian Cooperative Peritoneal Dialysis Study Group. Nephrol Dial Transplant 1994; 9: Buckalew VM Jr, Berg RL, Wang SR et al. Prevalence of hypertension in 1,795 subjects with chronic renal disease: the modification of diet in renal disease study baseline cohort. Modification of Diet in Renal Disease Study Group. Am J Kidney Dis 1996; 28: Received for publication: Accepted in revised form: