Nephrology Dialysis Transplantation

Size: px

Start display at page:

Download "Nephrology Dialysis Transplantation"

Denis McCormick
6 years ago
Views:

1 Nephrol Dial Transplant (1998) 13: Original Article Nephrology Dialysis Transplantation Urea kinetic modelling are any of the bedside Kt/V formulae reliable enough? Adrian Covic1, David J. A. Goldsmith2, Ken Hill3, Michael C. Venning3 and Peter Ackrill3 1Renal Department, Parhon Hospital, Iasi, Romania, 2Renal Unit, Guy s Hospital, London, and 3Withington Hospital AKU, Withington Hospital, West Didsbury, Manchester, UK Abstract as a whole the biggest difference from UKM mean Background. Longevity on dialysis is determined by values was obtained using Barth s and Calzavara s many factors. One of these has increasingly been seen formulae (delta of and respectively to be dialysis dose. There are several methods for (P<0.05)). calculating dialysis dose. We wanted prospectively to Conclusions. The best correlations were seen with the test gold-standard UKM-Kt/V with various shortcut Daugirdas 2 formula (r2=0.953). Also, comparing bedside formulae, to see whether reliance on the grouped formulae containing ln(co/ct) terms with latter approach was likely to lead to errors in over- or those incorporating the (Co Ct)/Co ratio (i.e. the urea underprescribing dialysis regimens. reduction) there was a better correlation for all formulae Methods. Ten bedside formulae for the calculation of employing the logarithmic transformation (r2= Kt/V (urea) were compared with UKM Kt/V values, cf. r2= ). Nevertheless no bedside in a month-long study involving 507 dialysis sessions formula had the accuracy of UKM-Kt/V. in 50 patients in a single-centre in-patient haemodialysis unit. Key words: urea kinetic modelling, haemodialysis; Results. For patients with UKM Kt/V<0.8 (median dialysis adequacy 0.69, n=140), simplified formulae had a difference (delta) of from the calculated UKM resulting in an inter-method variability ranging from 13 to 57%. The least difference was seen with the Introduction Calzavara formula (P=NS), maximum difference with the Barth formulae (P<0.05). No statistically significis the correct assessment of dialysis adequacy. Gotch s An issue of increasing importance to all nephrologists ant differences were seen when comparing Daugirdas 1 and 2 and Keshaviah formulae with UKM, for analysis of the NCDS data, proposed the Kt/V (urea) patients with UKM Kt/V<0.8. For patients with as an important correlate of clinical outcome [1]. UKM Kt/V in the range (median 1.06, n= Recent studies have shown how efficient the use of 285) the extreme recorded values from simplified forand monitoring of dialysis [2], and also in predicting urea kinetic modelling (UKM ) is in the quantification mulae were ( least different) and (most different) from the UKM mean, with an inter-method patient mortality [3]. On this basis, there is increasing variability ranging between 1.1% (Basile method) to recognition of the need for, and acceptance of, Kt/V 23.1% (Calzavara). No statistically significant differmany renal units [4,5]. However, the need for some (urea) as a surrogate for dialysis dose delivered in ence were seen when comparing Daugirdas 1 and 2, Keshaviah, and Lowrie formulae with UKM, for time to be spent using a computer programme has led patients with UKM Kt/V For patients with to attempts to use various short-cuts, including bedthe highest UKM Kt/V values (>1.4; median 1.58, side Kt/V formulae, which can be used by nephrolog- n=72), all simplified formulae gave Kt/V values lower ists as they evaluate patients clinically. These formulae than UKM Kt/V: the minimum difference was [6 15] all make use of the relationship between pre- using Jindal (P=NS, intermethod variability of 4.4%), and post-dialysis blood urea nitrogen (BUN). while the maximum was seen when using Calzavara Unfortunately, as has recently been observed [16], (P<0.05; difference=0.69; intermethod variability of these simplifications rely on various assumptions, 43.7%). There was also no statistically significant which, depending on the extent to which they are valid, difference for Basile and Kerr methods. For the group can lead to very marked, and highly clinically relevant, differences in calculated Kt/V. For example, for a Correspondence and offprint requests to: D. J. A. Goldsmith, Renal hypothetical patient of 70 kg undergoing a dialysis Unit, Guy s Hospital, London SE1 9RT, UK. with a single pair of BUN data, Kt/V values from European Renal Association European Dialysis and Transplant Association

2 Kt/V bedside formulae 3139 (borderline/underdialysed) to 1.43 (well dialysed) can The statistical analysis was performed with a CSTAT be obtained. In part the reason for concern with these (Oxford Statistics) programme, using Student s t test, correla- bedside techniques also rests on their proneness to tion analysis and regressions to the mean. Significance was greater sampling error [17], one reason for which is taken as two-tailed P<0.05. For each patient s dialysis session the absolute difference between UKM and simplified the well-known phenomenon of post-dialytic urea Kt/V was derived ( delta method ), and taking all dialysis rebound [18 20]. sessions for all patients, a mean delta method was derived The purpose of this study was to evaluate, using for each method compared with UKM. In order to assess UKM Kt/V as the gold standard the best-known agreement between UKM and the various simplified bedside simplified formulae, in an attempt to rank methods, a series of Bland and Altman analyses [23] was these in order of precision and accuracy. constructed. Agreement between methods was therefore evaluated by calculating the bias as the mean difference (mean delta method) and the standard deviation for that Subjects and methods mean difference; 95% distribution of these difference was depicted schematically. The study included all 50 in-centre patients from the Withington AKU haemodialysis programme (mean age 48.5±17.7 years, male 36 : female 14); 32 patients were on Results double-needle access, 14 on single-needle access to arteriovenous fistulae, and four on permanent (tunnelled) single- Five hundred and seven dialysis sessions, in 50 patients, lumen dialysis subclavian catheters. The number of modelling over 30 days were included in this analysis. The descripsessions for each patient ranged from minimum eight to tion of the Kt/V results obtained from urea kinetic maximum 12. For the study period of 30 days, the following modelling and from the simplified formulae and shown dialysis parameters were unchanged : dialysis session time in Table min (machine start to machine stop); dialysate flow For patients with UKM Kt/V<0.8 (median 0.69, 500 ml/min; blood flow 200 ml/min; same dialysis membrane for each patient (Gambro-Lundia 3N and 5N, Fresenius F6 n=140), simplified formulae had a difference (delta) and F60); same dialysis machine (Fresenius 2008E); same of to from the calculated UKM value patient dry weight (JVP, oedema, postural BP fall, cramps); resulting in an inter-method variability ranging from same total ultrafiltration; and same dialysate sodium/calcium/conductivity 13 to 57%. The least difference was seen with the profile and dialysate temperature. Patient Calzavara formula (P=NS), while maximum differ- medication, including antihypertensives and erythropoietin, ence was seen with the Barth formulae (P<0.05). No were unaltered. No patient included in this analysis had any statistically significant difference were seen when comsignificant intercurrent illness during the study period. Special paring Daugirdas 1 and 2 and Keshaviah formulae attention was paid to the real dialysis time, so that timewith UKM, for patients with UKM Kt/V<0.8. counters were fitted to all machines for all sessions, to record For patients with UKM Kt/V in the range 0.8 to effective dialysis duration (excluding any unwanted interruptions, e.g. due to dialysis hypotensive episodes). For each 1.4 (median 1.06, n=285) the extreme recorded values dialysis session for each patient the following were recorded: from simplified formulae were ( least different) BUN at beginning (C0), and at the end (Ct) of the session and (most different) from the UKM mean, with (latter obtained 3 min after slowing the pump speed to an inter-method variability ranging between 1.1% 50 ml/min, which value correlated extremely well with that (Basile method) to 23.1% (Calzavara). No statistically of a sample taken 30 min after the cessation of dialysis significant difference were seen when comparing [18 20]), true dialysis time (Ct), the intradialytic weight loss Daugirdas 1 and 2, Keshaviah, and Lowrie formulae (Uf ), patients dry weight ( Wt); and weekly a full biochemical with UKM, for patients with UKM Kt/V and haematological profile was obtained. For patients with the highest UKM Kt/V values For this study the reference Kt/V was obtained from a (>1.4; median 1.58, n=72), all simplified formulae computer model ( Fresenius Pack-H [21,22]). The following were used in these calculations: manufacturers filter clearance gave Kt/V values lower than UKM Kt/V: the min- ( K ), the effective blood flow and dialysis duration ( T= imum difference was using Jindal (P=NS, inter- 240 min ), patients weekly haematocrit. Table 1 summarizes the simplified formulae used for comparison. Table 2. Kt/V results obtained with UKM and simplified formulae Table 1. Bedside formulae for simplified calculation of Kt/V values Formulae Mean SD Median Minimum Maximum Jindal Kt/V=0.04((C0 Ct)/C0 100) 1.2 UKM Keshaviah Kt/V=1.162ln(C0/Ct) Jindal Lowrie Kt/V=ln(C0/Ct) Keshaviah Barth Kt/V=0.031((C0 Ct)/C0 100) 0.66 Lowrie Daugirdas 1 Kt/V= ln(ct/c t Uf/Wt) Barth Daugirdas 2 Kt/V= Daugirdas ln(ct/c t)+(4 3.5 Ct/C0) Uf/Wt Daugirdas Calzavara Kt/V=(C0 Ct)/((C0+Ct)/2) Calzavara Ijely Kt/V=0.018((C0 Ct)/C0) 100 Ijely Basile Kt/V=0.023((C0 Ct)/C0 100)0.284 Basile Kerr Kt/V=0.042((C0 Ct)/C0 100) 1.48 Kerr

3 3140 Table 3. Formulae determining extreme differences from UKM Kt/V values A. Covic et al. Kt/V<0.8 Kt/V Kt/V>1.4 SN Pts DN Pts Minimum D Calzavara Basile Jindal Keshaviah Kerr Maximum D Barth Calzavara Calzavara Barth Barth, Calzavara method variability of 4.4%), while the maximum was hensively independently validated by measures of outcome, seen when using Calzavara (P<0.05; difference=0.69; seduce with promises of accuracy, speed, and inter-method variability of 43.7%). There was also no usefulness. But is this algebraic promiscuity justified statistically significant difference for Basile and Kerr [16,24,25]? What can we make of the different values methods. that each of these competing formulae delivers, and For the group as a whole the biggest difference from which most closely approximates to the gold-standard UKM mean values was obtained using Barth s and of UKM Kt/V? It is these questions that this report Calzavara s formulae (delta of and respectively attempts to answer, using data from a very large (P<0.05)). The best correlations were seen with number of dialysis sessions over a short period of time the Daugirdas 2 formula (r2=0.953). Also, grouping in a stable group of dialysis patients dialysed in a formulae containing ln(co/ct) terms Kesheviah, consistent way (medium-length dialysis sessions, slow Lowrie, Daugirdas and those incorporating the blood flow). In reality, a prospective cohort study (Co Ct)/Co ratio (i.e. the urea reduction) Jindal, using hard end-points such as mortality and morbid Barth, Calzavara, Ijely, Basile, and Kerr there was events can only fully complete the picture, but that is a better correlation for all formulae employing the beyond the scope of this paper, and no such trial is logarithmic transformation (r2= cf. r2= currently being pursued. To allow a sensible interpreta ). tion of our data, great care was employed in the design In reality, as which UKM Kt/V category a patient and execution of this study to eliminate potential should be assigned to is not known, nor can safely be confounding factors, such as changes in haematocrit, assumed, we also analysed the formulae with respect dry weight, ultrafiltration, time on dialysis and urea to the type of dialysis access, single-vs double-needle. rebound [17 20]. It would be of interest to repeat this Here, for single-needle access the minimum difference exercise with high-flux dialysis patients. was seen using the Kesheviah formula (0.033) whilst Our data show a wide and clinically significant range for double-needle patients this was seen using the Kerr of Kt/V values, depending upon which simplified for- formula (0.026). A summary is shown in Table 3. mula is used, with an overall inter-method variability For an overview of the 10 methods comparisons, a (mean differences for each method) of 24%. Of particu- Bland and Altman matrix was constructed [23]. This lar concern is the fact that these differences are at their is shown in Table 4 and in Figures 1 10 most apparent and important at the extremes of dia- lysis dose ( Kt/V<0.8 and>1.4) ranging from 44 to 57%). Discussion In an attempt to answer the question of which is the best simplified formula, we first tried to catalogue the Developed to simplify urea kinetic modelling and to methods yielding the greatest differences. This is not permit very rapid evaluation of dialysis-adequacy, straightforward, taking into account the excellent val- numbers of bespoke formulae have proliferated over idation studies extant in the literature [6 15] and the recent years. These simpler methods, none compremethod excellent coefficient of correlation for each single with UKM values obtained in our study Table 4. (r2>0.939). Formulae more naturally modelling the exponential decline in BUN across a dialysis session were better correlated than those formulae that did Method Mean of the 95% confidence Standard differences between interval of the deviation of not logarithmically transform BUN. But as we have UKM and method mean the mean observed before [26], a good correlation does not values (necessarily) a good biological agreement make. So, for the first time this question is addressed using the Daugirdas correct statistical approach [23]. From Table 4 it can Daugirdas be seen that non-logarithmic formulae produce results Keshaviah that are unacceptably different from UKM Kt/V and Barth Lowrie also that these results are skewed towards extreme Basile values: Kerr and Jindal methods overestimating small Jindal Kt/V values (<1.0; Figures 3, 10), while Basile, Ijely, Calzavara and Calzavara methods overestimate higher Kt/V Kerr values (>1.0; Figures 1,2,4). Ijely So, which formula can we recommend? From

4 Kt/V bedside formulae 3141 Fig. 1. Blands & Altman analysis: UKM vs Basile formulae. Fig. 2. Blands & Altman analysis: UKM vs Ijely formulae.

5 3142 A. Covic et al. Fig. 3. Blands & Altman analysis: UKM vs Kerr formulae. Fig. 4. Blands & Altman analysis: UKM vs Calzavara.

6 Kt/V bedside formulae 3143 Fig. 5. Blands & Altman analysis: UKM vs Daugirdas 1. Fig. 6. Blands & Altman analysis: UKM vs Daugirdas 2.

7 3144 A. Covic et al. Fig. 7. Blands & Altman analysis: UKM vs Barth. Fig. 8. Blands & Altman analysis: UKM vs Keshaviah.

8 Kt/V bedside formulae 3145 Fig. 9. Blands & Altman analysis: UKM vs Lowrie. Fig. 10. Blands & Altman analysis: UKM vs Jindal.

9 3146 A. Covic et al. Table 2, the only conclusion is that there is no uniquely 9. Barth RH. Direct calculation of Kt/V. A simplified approach applicable formula for all categories of patient, but it to monitoring of haemodialysis. Nephron 1988; 50: Daugirdas JT. The post: pre dialysis plasma urea nitrogen ratio is reasonable to say that the Daugirdas and Kesheviah to estimate Kt/V and NPCR: Validation. Int J Artif Organs logarithmic equations consistently more closely 1989; 12: approached the UKM Kt/V results, for low medium 11. Daugirdas JT. Second generation logarithmic estimates of single- and high dialysis dose ranges, and for single- or double- pool variable volume Kt/V: an analysis of error. J Am Soc Nephrol 1993; 4: needle access. The Bland and Altman agreement ana- 12. Calzavara P, Vianello A, Da Porto A et al. Comparison between lysis for Daugirdas formulae (and also for Kesheviah) three mathematical models of Kt/V. Int J Artif Organs 1988; shows that one can expect a minimum difference from 11: UKM Kt/V values of , with the lowest 13. Ijely GK, Raja RM. Simplified calculation of PCR and Kt/V. limit of agreement possible being to Abstract 24th Annual JASN Meeting 1991; Basile C, Casino F, Lopez T. Percent reduction in blood urea and to Although of concern, these differ- concentration during dialysis estimates Kt/V in a simple and ences might be regarded as acceptable. Moreover, the accurate way. Am J Kidney Dis 1990; 15: scatter of difference is uniform, without any skew, 15. Kerr PG, Argiles A, Canaud B, Flavier JL, Mion CM. Accuracy confirming its universal applicability. Finally, the of Kt/V estimations in high-flux haemodiafiltration using percent reduction of urea: incorporation of urea rebound. Nephrol Dial intermethod variability between mean UKM and mean Transplant 1993; 8: Kesheviah/Daugirdas values was only %. 16. Movilli E. Simplified approaches to calculate Kt/V. It s time for In conclusion, no bedside formula is perfect, but an agreement. Nephrol Dial Transplant 1996; 11: some are less imperfect than others. Logarithmic 17. Lai Y-H, Guh J-Y, Chen H-C, Tsai J-H. Effects of different methods are to be preferred. Common sense dictates sampling methods for measurement of the post-dialysis blood urea nitrogen on urea kinetic modeling parameters in patients that bedside analyses should never supplant the rouundergoing long-term haemodialysis. ASAIO J 1995; 41 2: tine application of more comprehensive analyses, unless prospective studies supply evidence of compar- 18. Tattersall JE, DeTakats D, Chamney P, Greenwood RN, able reliability. Farrington K. The post-hemodialysis rebound: predicting and quantifying its effect on Kt/V. Kidney Int 1996; 50: Pflederer BR, Torrey C, Preister-Coary A et al. Estimating References equilibrated Kt/V from an intradialytic sample. Effect of cardiopulmonary and access recirculations. Kidney Int 1995; 48: 1. Gotch FA, Sargent JA. A mechanistic analysis of the National Co-operative Dialysis Study (NCDS). Kidney Int 1985; 28: 20. Pflederer BR, Torrey C, Lau AH, Daugirdas JT. Post-dialysis urea rebound after long session length (3.5 to 4.5 hours) 2. Depner TA. Assessing adequacy of hemodialysis urea modeling. haemodialysis. J Am Soc Nephrol 1993; 4: 377 Kidney Int 1994; 45: von Ahrenholz P, Holtz M, Falkenhagen D, Klinkmann H. 3. Held PJ, Port FK, Wolfe RA et al. The dose of hemodialysis Clearance-bestimmungen fur die harnstoffgefuhrte Hamodialyse. and patient mortality. Kidney Int 1996; 50: Z Urol Nephrol 1988; 81: Barth RH. Urea modelling and Kt/V: a critical appraisal. Kidney 22. Ahrenholz P, Falkenhagen D, Klinkmann H. A simplified Int 1993; 43 [Suppl 41]: S procedure to compute dialysis time and frequency by means of 5. Hakim RM, Depner TA, Parker TF. Adequacy of haemodialysis. urea kinetics. Int J Artif Organs 1988; 115: Am J Kidney Dis 1992; 20: Bland MJ, Altman DG. Statistical methods for assessing agree- 6. Jindal KK, Manuel A, Goldstein MB. Percent reduction of the ment between two methods of clinical measurement. Lancet blood urea concentration during dialysis (PRU ), a simple and 1986; 1: accurate method to estimate Kt/V urea. ASAIO Trans 1987; 24. Bosticardo GM, Alloatti S, Avale U, Paternoster G, Cestonaro 33: G, Giacchino F. Single pool urea kinetic model (UKM) and 7. Kesheviah PR, Hanson GI, Berkseth RO, Collins AJ. A simpli- simplified formulae: comparison of results. J Am Soc Nephrol fied approach to monitoring in vivo therapy prescription. Trans 1993; 12: 355 Am Soc Artif Organs 1988; 34: Daugirdas JT. Rapid methods of estimating Kt/V: three formu- 8. Lowrie EG, Teehan BP. Principles of prescribing dialysis las compared. ASAIO Trans 1990; 36: M therapy: implementing recommendations form the National 26. Harty JC, Goldsmith DJA, Boulton H et al. Limitations of the Co-operative Dialysis Study. Kidney Int 1983; 23 [Suppl 13]: peritoneal equilibration test in prescribing and monitoring dia- S lysis therapy. Nephrol Dial Transplant 1995; 10: Received for publication: Accepted in revised form:

Dialysis Adequacy (HD) Guidelines

Dialysis Adequacy (HD) Guidelines Peter Kerr, Convenor (Monash, Victoria) Vlado Perkovic (Camperdown, New South Wales) Jim Petrie (Woolloongabba, Queensland) John Agar (Geelong, Victoria) Alex Disney (Woodville,