Risk factors for increased variability in dialysis delivery in haemodialysis patients

Nephrol Dial Transplant (2003) 18: 2112 2117 DOI: 10.1093/ndt/gfg297 Original Article Risk factors for increased variability in dialysis delivery in haemodialysis patients K. Scott Brimble, Darin J. Treleaven, Joye St Onge and Euan J. Carlisle McMaster University, Department of Medicine, Hamilton, ON, Canada Abstract Background. Numerous events may occur during a haemodialysis session, leading to variation in the quantity of dialysis received. The purpose of this study was to identify risk factors for variability in haemodialysis delivery. Methods. Variability in dialysis delivery was expressed by the coefficient of variation (CV%) and calculated for the volume of blood processed (VBP) for all treatments and the monthly urea reduction ratio (URR) in each patient over an 8 month period. The univariate and multivariate relationships between various predictor variables and the URR and VBP CV% were determined. Results. Eighty-nine patients were identified who met study criteria. The mean VBP and URR CV% were 10.3 ± 4.7 and 5.4 ± 3.8%, respectively. Patients with tunnelled catheters and total nursing-care patients had higher VBP and URR CV%, as evaluated by multivariate analysis. Patients with inadequate dialysis (mean URR <65%) had a higher VBP CV% than those patients with mean URR values 65% (14.8 ± 5.4 vs 9.7 ± 4.5%; P ¼ 0.01). An accurate determination of the URR in 90% of patients required 14 measurements in patients with catheters vs three and two measurements in arteriovenous fistulae and polytetrafluoroethylene grafts, respectively. Conclusions. This study demonstrated that the use of a venous tunnelled catheter and dialysis in a total nursing-care unit were the only factors independently associated with greater variability in both VBP and URR. Attention to individual dialysis sessions in patients with tunnelled catheters or in a total nursingcare unit is prudent, particularly when identifying reasons for under-dialysis. Correspondence and offprint requests to: Scott Brimble, MD, Suite 708, 25 Charlton Avenue E, Hamilton, ON, L8N 1Y2, Canada. Email: brimbles@mcmaster.ca Keywords: access; adequacy; catheters; haemodialysis; urea reduction ratio; variability Introduction The importance of dialysis dose, as measured by either the Kt/V or urea reduction ratio (URR), in relation to patient morbidity and mortality has been clearly demonstrated [1 5]. Clinical practice guidelines [6,7] recommend monthly monitoring of these adequacy measures to ensure optimal patient care. More frequent measurements have been suggested for patients who are non-compliant with the dialysis prescription or who have frequent problems with dialysis delivery, such as poor access blood flow [6]. Numerous unanticipated events may occur during a patient s haemodialysis session, which can lead to variation in the quantity of dialysis received from one session to the next. Access recirculation or dialyser clotting are examples that may result in decreased efficiency of solute clearance. Alternatively, failure to account for treatment interruptions or premature discontinuation may lead to a reduction in treatment duration [6]. All of these events will reduce the overall quantity of dialysis delivered and yet may not be readily apparent from periodic review of adequacy measures without careful attention to all treatment sessions. A previous study on a cohort of prevalent haemodialysis patients had identified that patients with arteriovenous fistulae (AVF) or polytetrafluoroethylene (PTFE) grafts received more dialysis on adequacy measurement days than on non-measurement days, whereas patients with catheters did not [8]. A hypothesis arising from that study and based on clinical observation was that venous tunnelled catheters may be subject to greater variability in dialysis delivery between all treatments as compared with AVF or PTFE grafts. The purpose of this study, therefore, was to identify potential risk factors, including access type, ß 2003 European Renal Association European Dialysis and Transplant Association

Variability in haemodialysis delivery 2113 for session-to-session variability in haemodialysis delivery. Subjects and methods Patients were identified from a haemodialysis census database at St Joseph s Hospital in Hamilton, ON, Canada. Only patients who had been on haemodialysis over the last 8 months were considered and patients were excluded if any of the following had been changed over the study period: access type, prescribed blood or dialysate flow, prescribed treatment duration, type of haemodialysis machine or level of nursing care required. Level of nursing care was defined as follows. Total nursing-care patients require a dialysis nurse for the entire set up and delivery of the dialysis treatment while selfcare patients monitor and adjust the dialysis treatment themselves. Demographic and comorbidity data were obtained, as well as information on the dialysis prescription. All dialysis patients at this centre have a detailed clinical summary of their medical history on the dialysis chart, which has been written by the primary nephrologist and is updated regularly. The presence of diabetes mellitus (DM), coronary artery disease (CAD), left ventricular (LV) dysfunction and peripheral vascular disease (PVD) was ascertained by reviewing this summary. In addition, patients were also considered to have DM if they were on insulin or oral hypoglycaemic agents. Patients were also considered to have CAD if they had a known positive stress test or coronary angiogram or previous coronary artery angioplasty or bypass grafting. All dialysis treatments over the preceding 8 months were reviewed for volume of blood processed (VBP), treatment duration, dialysis shift, level of nursing care, target weight, ultrafiltration volume, hypotension (defined as any documented systolic blood pressure <100 mmhg), dialyser clotting and URR. All patients were dialysed on high-flux dialysers. VBP on dialysis was recorded after each session directly from the haemodialysis machine. VBP represents the product of the blood flow calculated from the calibrated blood pump, integrated over the total time of the dialysis session. Average blood flow per session was calculated by dividing the VBP by the treatment duration recorded for each session. URR was measured mid-week using the blood re-infusion sampling technique, as described previously [6]. The pre-dialysis urea level was assessed on blood that was drawn immediately prior to haemodialysis initiation whereas the post-dialysis urea was drawn 5 min after the patient s blood had been reinfused. This procedure does not have a direct effect on the VBP for that session. Four types of catheters were used in this study; 64% of patients who dialysed with a venous tunnelled catheter utilized an Optiflow (C.R. Bard Inc.) catheter. The remaining catheters used were Highflow (CardioMed), Uldall (Cook Canada) and Permcath (Quinton Instruments). The vascular access creation strategy at this centre at the time of the study was to create AVF in any patient felt to have suitable vessels. PTFE grafts or catheters were placed in patients who did not have suitable vessels or in whom a primary AVF had failed. In the event a patient required initiation of dialysis prior to fistula maturation, a tunnelled catheter was placed. Continuous variables are expressed as means ± SD. Variability in dialysis delivery was expressed by the coefficient of variation (CV% ¼ SD/mean) and calculated for the VBP, blood flow, treatment duration and the monthly URR in each patient. The number of averaged repeated measurements required to determine with 95% probability the patient s true URR and VBP was calculated according to the formula [9]: N ¼ (1.96 CV%/D o ) 2 where N is the number of measurements required per person and D o is the level of precision specified (as a percentage of the true variable value). A D o value of 10% was used in this study. The univariate relationships between various predictor variables and the URR and VBP CV% were determined using either a two-tailed Student s t-test or analysis of variance (ANOVA) for categorical variables and linear regression for continuous variables. Multivariate correlates were determined using multiple linear regression on significant (P < 0.10) univariate variables. Data analysis was carried out using the SPSS v. 9.0 (SPSS Inc.) software package. Results Two-hundred and ninety-five patients were identified who had been on haemodialysis over the preceding 8 months. The dialysis charts of 125 of these patients were sampled at convenience by two of the authors (J.S. and S.B.) for further review and 89 patients were identified who met the additional specified study criteria. Eighty-six patients (96.6%) were dialysed three times a week and 8650 haemodialysis treatment sheets were reviewed overall in the final analysis (95% complete). The baseline characteristics of the patients are shown in Table 1 and have been described previously [8]. Univariate predictors of the VBP and URR CV% are shown in Table 2. The mean VBP and URR CV% were 10.3 ± 4.7 and 5.4 ± 3.8%, respectively, and were significantly correlated (r ¼ 0.62, P < 0.0001). Patients with tunnelled catheters and total nursing-care patients had both a higher VBP and URR CV%. There was no difference in the VBP or URR CV% amongst the Table 1. Baseline patient characteristics Characteristic Value Mean age (years) 64.7 Females (%) 43.8 CAD (%) 50.6 DM (%) 38.2 LV dysfunction (%) 19.1 PVD (%) 20.2 Dialysis vintage (months) 41.2 Access type (%) Fistula 43.8 Graft 43.8 Tunnelled catheter 12.4 Dialysis location (%) Total care 64.0 Self-care 36.0

2114 K. S. Brimble et al. Table 2. Univariate predictors of VBP and URR CV% Mean VBP CV% P-value Mean URR CV% P-value Test statistic (95% CI) (95% CI) Overall (89) 10.3 (9.2 11.3) 5.4 (4.5 6.2) Access Catheter (11) 17.0 (12.3 21.7) 11.6 (7.4 15.9) Fistula (39) 8.8 (7.5 10.2) <0.001 a 4.7 (3.7 5.7) <0.001 a Graft (39) 9.9 (9.0 10.9) <0.001 a 4.3 (3.6 5.0) <0.001 a ANOVA b Shift c Morning (30) 11.3 (8.8 13.7) 5.6 (3.8 7.5) Afternoon (32) 10.3 (9.4 11.8) NS 5.2 (3.7 6.8) NS Evening (21) 8.4 (6.7 10.1) NS 4.9 (4.2 5.6) NS ANOVA b Gender Male (50) 9.3 (8.3 10.4) 5.2 (4.1 6.1) Female (39) 11.6 (9.7 13.5) 0.03 5.7 (4.2 7.2) NS t-test CAD Absent (44) 9.3 (7.7 10.9) 5.1 (4.0 6.2) Present (45) 11.3 (10.1 12.6) 0.05 5.7 (4.3 7.0) NS t-test LV dysfunction Absent (72) 9.9 (8.7 11.1) 5.0 (3.7 6.3) Present (17) 12.1 (10.3 13.9) NS 5.5 (4.5 6.5) NS t-test DM Absent (55) 9.5 (8.1 10.9) 5.0 (3.9 6.1) Present (34) 11.7 (10.3 13.0) 0.04 6.0 (4.7 7.4) NS t-test PVD Absent (71) 10.1 (8.9 11.3) 5.3 (4.4 6.3) Present (18) 11.2 (9.1 13.4) NS 5.6 (3.6 7.6) NS t-test Nursing care Total (57) 12.1 (10.9 13.4) 6.3 (5.1 7.5) Self (32) 7.2 (6.0 8.3) <0.001 3.7 (3.2 4.3) 0.003 t-test Age NS NS Pearson correlation Weight NS 0.05 Pearson correlation Hypotensive episode frequency NS NS Pearson correlation Dialyser clotting frequency NS NS Pearson correlation a Comparison group is catheters. b Post-hoc analysis utilizing Tukey s honestly significant difference. c Patients were excluded from the analysis if their shift changed during the study. Table 3. Multivariate predictors of VBP and URR CV% Variable VBP CV% coefficient P-value URR CV% coefficient P-value (95% CI) (95% CI) Access a 6.1 (3.7 8.5) <0.001 6.6 (4.6 8.7) <0.001 Nursing care b 3.9 (2.2 5.1) <0.001 1.6 (0.2 3.1) 0.025 Female gender 1.5 (0 3.1) 0.05 CAD 1.9 (0 3.2) 0.039 Total r 2 0.48 0.39 a Catheter vs AVF/graft. b Total nursing vs self-care. different types of catheters used (data not shown). Patients with CAD, DM and females had a higher VBP CV% while larger patients had a higher URR CV%. Age, dialysis vintage, the presence of LV dysfunction, dialysis shift, ultrafiltration volume and frequency of hypotensive or dialyser clotting episodes were not associated with either URR or VBP CV%. Multivariate linear regression analysis (Table 3) indicates that only use of tunnelled catheters and requirement for total nursing-care were factors independently associated with both a higher VBP and URR CV%. Female gender and CAD were also independently associated with a higher VBP CV%. There was no evidence of interaction amongst any of these significant variables. The CV% was also determined for blood flow and treatment duration, the two components of VBP. The blood flow CV% was significantly higher (17.1 ± 7.1%) in patients with catheters than in patients with AVF (8.8 ± 4.1%) or grafts (9.0 ± 2.7%) (P < 0.0001), whereas there was no difference observed in treatment duration (data not shown). To ensure that the greater variability observed in catheters was not due simply to deterioration in catheter performance over the length of

Variability in haemodialysis delivery 2115 Table 4. Comparison of mean URR values and frequency of intradialytic events in patients according to access type and level of nursing care Mean URR Hypotensive episode Dialyser clotting (SD) frequency frequency Access Catheter (11) 67.1 (5.4) 14.9 (15.3) 1.00 (1.3) Fistula (39) 71.3 (5.6) a 14.4 (19.7) 0.42 (1.1) Graft (39) 74.1 (6.0) b 15.1 (25.3) 0.76 (1.0) Nursing care Total (57) 71.2 (6.2) 17.0 (23.9) 0.61 (1.1) Self (32) 73.4 (5.8) c 10.7 (16.5) 0.69 (1.1) a P ¼ 0.1 and b P ¼ 0.002 vs catheter group. c P ¼ 0.1 vs total nursing-care group. the study, the mean URR for each month was calculated for patients who utilized catheters. There was no significant difference between the beginning and end months of the study in the mean URR (67.1 ± 6.8 vs 66.3 ± 6.8%; P ¼ 0.39). Table 4 shows the average URR values and frequencies of intradialytic events according to access type and level of nursing care. Patients with catheters had lower URR values than patients with PTFE grafts (P ¼ 0.002) and a trend towards lower URR values compared to patients with AVF (P ¼ 0.1). Patients requiring total nursing-care had a tendency to lower URR values than patients in self-care (P ¼ 0.1). A trend towards an increased frequency of dialyser clotting was also observed in patients with catheters, although this did not achieve statistical significance (P ¼ 0.21). Patients requiring total nursing-care had a tendency to an increased number of hypotensive episodes, although this too did not reach significance (P ¼ 0.19). The effect of access type on the number of measurements required to estimate the true URR and VBP with a 95% probability for each patient was determined (Table 5). To accurately determine the URR in 90% of the patients within 10% of the true value required 14 measurements in patients with catheters as opposed to three and two measurements in patients with AVF and grafts, respectively. Similar results were seen when the VBP was measured (Table 5). To determine whether there was a relationship between variability in session-to-session dialysis delivery (VBP CV%) and dialysis adequacy (URR), patients were divided into two groups based on a mean URR cut point of 65%. The mean VBP CV% in the 11 patients with a mean URR <65% was 14.8 ± 5.4% compared with 9.7 ± 4.5% in the remaining patients with a URR of 65% (P ¼ 0.01). Discussion The use of a venous tunnelled catheter and dialysis in a total nursing-care unit were the only factors independently associated with greater variability in both VBP and URR, as measured by the CV%, a commonly used and accepted measure of variability [10,11]. In addition, it was demonstrated that greater variability in the VBP Table 5. Number of measurements required to accurately (±10%) assess URR and VBP Access type 50% of patients (median) 90% of patients URR VBP URR VBP Overall 1 4 4 9 Catheters 5 8 14 20 AVF 1 3 3 8 Grafts 1 4 2 7 was a major contributor to inadequate dialysis delivery, as measured by the URR. This has important implications for the management of patients who are receiving inadequate dialysis, in particular those who utilize a tunnelled catheter for vascular access or dialyse in a total nursing-care unit. There are several reasons why variability in the amount of dialysis delivered may be important. The majority of events contributing to variability in dialysis delivery, including dialyser clotting, access dysfunction, treatment interruptions and premature discontinuation of haemodialysis for staff/unit or patient convenience, will lead to a decrease in the quantity of dialysis received on a given day. Perhaps more importantly, management of a low URR (or Kt/V) may be ill-fated if the aetiology of these low adequacy measures has more to do with factors leading to greater variability in dialysis delivery and less to do with the dialysis prescription. For example, if a patient has an average URR of 61% (objective 65%), increasing the blood flow or treatment duration may not be particularly effective if the main problem is recurrent dialyser clotting or patient-initiated shortening of treatment sessions. Several explanations are possible for the greater variability observed in patients requiring higher levels of nursing care. A self-care patient may be more compliant with the dialysis prescription than a patient in a total nursing-care unit and may have greater incentive than dialysis nurses to ensure the full treatment is being received, leading to more consistent treatment sessions. In addition, a self-care patient at this centre is selected based on their overall favourable clinical status. Even after adjustment for comorbidity

2116 K. S. Brimble et al. and access type, such a patient is likely healthier with a better functioning access, resulting in more consistent haemodialysis treatment than a patient who requires assistance with dialysis. The use of a venous tunnelled catheter for dialysis access was the strongest predictor of both the URR and VBP CV%. We and others [12 14] have demonstrated that this form of access has been associated with lower adequacy measures. To our knowledge, this is the first study to demonstrate that lower adequacy measures are associated with higher variability in dialysis delivery. This was supported by the observation that patients with mean URR values below the desired level of 65% had significantly greater variability in VBP than did patients whose URR was >65%. This variability in VBP appeared to be due almost exclusively to variability in blood flow. Factors which contribute to shortened treatment times, including treatment interruptions or premature discontinuation by staff or patients, did not appear to be as important. Nevertheless, factors which could contribute to variability in access blood flow, including recurrent hypotension and dialyser clotting, did not appear to fully explain the greater variability in patients with catheters. However, the lack of association with these factors may have resulted from low statistical power in the group; there was a tendency to increased dialyser clotting in patients with catheters and frequency of hypotension in patients requiring total nursing-care that did not reach statistical significance. Other potentially important factors, including elevating venous pressures, access recirculation and use of intraluminal thrombolytic therapy [15], were not evaluated but may have contributed to the differences between the patient groups observed. A recent study has shown that use of catheters compared with AVF for vascular access was associated with a relative mortality risk of 1.54 and 1.70 in diabetics and non-diabetics, respectively [16]. This effect was independent of adequacy measures and various comorbid conditions and appeared to be due to both infectious and cardiac complications. Although the increase in mortality risk from cardiac causes in patients with catheters was not related to dialysis adequacy per se, the possibility that variability in dialysis dose on a day-to-day basis contributed to the increased risk cannot be excluded. As an example, patients with greater variability in VBP or URR may also be less likely to reach their fluid removal target on a given day and hence be at higher risk of cardiac complications from recurrent volume overload. A recent study published in abstract form [17] has shown that variability in Kt/V (as measured by the intrapatient Kt/V SD) was strongly associated with patient mortality, independent of age, diabetic status, ethnicity or gender. However, the influence of access type was not specifically evaluated. This study has demonstrated that, due to the increased variability in URR observed in patients with catheters, a greater number of URR and VBP measurements must be reviewed in order to be reasonably certain of the overall adequacy of a patient s dialysis regimen. A previous study [9] has shown in a cohort of 50 patients based on only three measurements that only one to two URR or Kt/V readings are required to accurately determine adequacy of dialysis in prevalent haemodialysis patients. However, the majority of these patients dialysed via an AVF and patients with diabetes, systemic illness or recent hospitalization were specifically excluded. Thus, when reviewing the most recent URR in a patient with a catheter, one must view the result with healthy scepticism. It may be advisable, based on this study s findings, to recommend more frequent URR (or Kt/V) measurements in patients with catheters as part of a routine strategy, particularly if the URR values are borderline or below target levels. A previous study on the same cohort of patients [8] had shown that patients with AVF and PTFE grafts received more dialysis on adequacy measurement days than on non-measurement days, whereas patients with catheters did not. It was hypothesized that greater variability (or unpredictability) in catheters could lead to a failure to detect a significant and systematic bias favouring more dialysis delivery on adequacy measurement days in patients with catheters. The greater variability overall between dialysis sessions in patients with catheters shown in this analysis would appear to support this explanation. There are several limitations to the study. Although data collection was thorough, it is a retrospective study and correlation of the observed variability to important clinical outcomes was beyond the scope of this study. In addition, VBP has not been shown to correlate with clinical outcomes and is clearly not a substitute for more appropriate adequacy measures, such as URR or Kt/V. Nonetheless, similar findings were observed with both the URR and VBP, providing support for the use of VBP as a method to evaluate all treatment sessions. To summarize, variability in URR and VBP, as measured by the CV%, was highest in patients with tunnelled catheters and who dialysed in a total nursingcare unit. Particular attention to individual dialysis sessions in patients with tunnelled catheters or in a total nursing-care unit is prudent, particularly when attempting to identify reasons for under-dialysis and planning an intervention strategy. Acknowledgements. We are indebted to Dr Geoff Norman for fruitful discussions on the measurement of variability in haemodialysis delivery and to Dr David Churchill who provided several insightful comments. Conflict of interest statement. None declared. References 1. Held PJ, Port FK, Wolfe RA et al. The dose of haemodialysis and patient mortality. Kidney Int 1996; 50: 550 556 2. Parker T, Husni L, Huang W, Lew N, Lowrie EG, Dallas Nephrology Associates. Survival of haemodialysis patients in the United States is improved with a greater quantity of dialysis. Am J Kidney Dis 1994; 23: 670 680

Variability in haemodialysis delivery 2117 3. McClellan WM, Soucie JM, Flanders D. Mortality in end-stage renal disease is associated with facility-to-facility differences in adequacy of haemodialysis. J Am Soc Nephrol 1998; 9: 1940 1947 4. Hakim RM, Breyer J, Ismail N, Schulman G. Effects of dose of dialysis on morbidity and mortality. Am J Kidney Dis 1994; 23: 663 669 5. Owen WF, Lew NL, Liu Y, Lowrie EG, Lazarus JM. The urea reduction ratio and serum albumin concentration as predictors of mortality in patients undergoing haemodialysis. N Engl J Med 1993; 329: 1001 1006 6. NKF-K/DOQI clinical practice guidelines for haemodialysis adequacy, update 2000. Am J Kidney Dis 2001; 37 [Suppl 1]: S7 S64 7. Deziel C, Hirsch DJ, Hoult P et al. Clinical practice guidelines for the delivery of haemodialysis. Canadian Society of Nephrology. J Am Soc Nephrol 1999; 10 [Suppl 13]: S306 S310 8. Brimble KS, Onge JS, Treleaven DJ, Carlisle EJ. Comparison of volume of blood processed on haemodialysis adequacy measurement sessions vs regular non-adequacy sessions. Nephrol Dial Transplant 2002; 17: 1470 1474 9. Kloppenberg WD, Stegeman CA, Hooyschuur M, van der Ven J, de Jong PE, Huisman RM. Assessing dialysis adequacy and dietary intake in the individual haemodialysis patient. Kidney Int 1999; 55: 1961 1969 10. Taylor JR. An Introduction to Error Analysis. University Science Books, Mill Valley, CA, 1982; 28 30 11. Virga G, Amici G, Mastrsimone S et al. Day-to-day variability of adequacy indexes in peritoneal dialysis. Nephrol Dial Transplant 1999; 14: 2932 2936 12. Palevsky PM, Washington MS, Stevenson JA et al. Barriers to the delivery of adequate haemodialysis in ESRD network 4. Adv Renal Repl Ther 2000; 7 [Suppl 1]: S11 S20 13. Palevsky PM, Washington MS, Stevenson JA et al. Improving compliance with the dialysis prescription as a strategy to increase the delivered dose of haemodialysis. An ESRD network 4 quality improvement project. Adv Renal Repl Ther 2000; 7 [Suppl 1]: S21 S30 14. Seghal AR, Snow RJ, Singer ME et al. Barriers to adequate delivery of haemodialysis. Am J Kidney Dis 1998; 31: 593 601 15. Clase CM, Crowther MA, Ingram AJ, Cina CS. Thrombolysis for restoration of patency to haemodialysis central venous catheters: a systematic review. J Thromb Thrombolysis 2001; 11: 127 136 16. Dhingra RK, Young EW, Hulbert-Shearon TE, Leavey SF, Port FK. Type of vascular access and mortality in U.S. haemodialysis patients. Kidney Int 2001; 60: 1443 1451 17. Bosch JP, Nabut J, Alquist M, Hegbrant J, Walters BA. Intrapatient variation in dialysis dose predicts mortality for dialysis patients. JASN 2002; 13: 21A Received for publication: 10.12.02 Accepted in revised form: 11.4.03