Comparison of predicted with measured creatinine clearance in cardiac surgical patients

Comparison of Predicted With Measured Creatinine Clearance in Cardiac Surgical Patients G.K. Bloor, K.R. Welsh, S. Goodall, and M.V. Shah Objective: To determine whether creatinine clearance can be determined from a single plasma creatinine measurement in patients up to 5 days after cardiac surgery. Design: Observational longitudinal study. Setting: Cardiac intensive care unit in a tertiary referral center for cardiothoracic surgery. Patients: Seventy-five patients (54 men, 21 women) scheduled for elective coronary artery surgery (93 postoperative patient days). Interventions: Creatinine clearance measurement using a 4-hour urine collection and a single arterial blood sample. Measurements andMainResults: There was significant agreement (Deming analysis r = 0.63-0.84, correlation r = 0.76-0.95, p < 0.05) between the predicted creatinine clearance and the measured creatinine clearance on each of

the postoperative days. This was maintained even if the patients required inotrope or vasoconstrictor therapy, were receiving parenteral nutrition, or had changing renal function (Deming analysis r = 0.67-0.7; correlation r = 0.8-0.93, p < 0.001) but does not apply to patients with preexisting renal dysfunction (Deming analysis r = 0.36; correlation r = 0.57,p = 0.002). Conclusions: In cardiac surgical patients with normal preoperative renal function, predicted creatinine clearance is as reliable as measured creatinine clearance up to the fifth postoperative day.

ENAL FUNCTION may change rapidly in patients in the intensive care umt. There is a significant incidence of renal failure in these patients, which is associated with increased mortality. ~ Early detection of deteriorating renal function can facilitate therapeutic changes, which may avoid acute renal failure or limit the degree of subsequent damage. Creatmme clearance is commonly used to estimate glomerular filtration rate (GFR), but a 24-hour collection time often provides results too late to be of clinical use. A previous study 2 has validated the use of 4-hour urine collections in the measurement of creatinine clearance, but even these are cumbersome to perform. Methods to predict creatinine clearance from a single plasma creatinine concentration in patients with both normal and abnormal renal function have been vahdated previously. 2 However, it has been emphasized that the creatinine concentration has to be stable for at least 4 days to allow valid prediction) This is not the situataon in the postoperative period. This study was designed to investigate whether the prediction of creatanine clearance from a single plasma creatimne concentration is valid in the postoperative period after cardiac surgery.

(Hitachi 747 analyzer; Boehringer Mannheim, Germany). Creatinine clearance was measured using 4-hour urine collections as previously described. 2 It was determined using the following formula:

R

METHODS The study protocol was approved by the local Ethics Committee, and informed consent was obtained from each patient. Patients accepted for cardiac surgery were studied over a 2-month period. Intravenous cefuroxime, 1.5 g, was administered as prophylaxis before induction of anesthesia in all patients. Blood samples were obtained daily at 8 AM via an mdwelhng arterial catheter after removal of 5 mL of blood from the catheter and were stored in lithium heparin bottles until analysis within 4 hours. Urine samples were collected over a 4-hour period, arranged so that the blood sampling occurred within the urine collection time. The sex, age, and body weight of the patients were recorded. Plasma creatinme concentration was determined by a modified Jaff6 method using an alkaline picrate reaction

Copyright © 1996 by W.B. Saunders Company

KEY WORDS: postoperative period, plasma creatinine, creatinine clearance

Urine creatmme (mmol L) × urine volume (mL) x 1000 Plasma creatinine (p~mol/l) × time (minutes) Creatinine clearance was also predicted from the plasma creatinine concentration using the nomogram described by Slersbaek-Nielsen and colleagues, which requires the sex, age and body weight of the patient to be known 4 (Fig 1). Subgroups were established on the basis of preexisting renal dysfunction (defined as a predicted creatimne clearance less than 60 mL/min), number of postoperative days, inotrope, vasoconstrictor, and intravenous feeding therapies recorded. Inotrope therapy was used to maintain the cardiac index greater than 2.5 L/mln/m 2. Vasoconstrictor therapy was used to maintain the systolic blood pressure greater than 100 mmHg in the presence of a low measured systemic vascular resistance. Evidence of changing renal function, as defined by a greater than 30% reduction in measured creatinine clearance, was also noted. Between-batch coefficients of variation were 2.7% and 4.5% for plasma creatinine concentration at 50 and 500 ~mol/L, respectively. There was an mtra-lndividual coefficient of varianon of up to 13% for creatinine clearance measurement. Statistical analysis (C-stat; Cherwell Scientific, Oxford) by paired t test was applied for comparison of parametric data obtained by both methods. All results are expressed as mean (95% confidence limits). Differences or

From the Departments of Anaesthesia and Chemical Pathology, Leeds General Infirmary, Leeds, UK. Address repnnt requests to Dr G.K Bloor, FRCA, MRCP, Department of Anaesthesia, Eastbourne District General Hospttal, Kings Drtve, Eastbourne, East Sussex BN21 2UD Untted Kmgdom Copyright © 1996 by W B Saunders Company 1053-0770/96/1007-001353 00/0

Journal of Cardlothoracw and Vascular Anesthesia, Vo110, No 7 (December), 1996' pp 899-902

899

900

BLOOR ET AL

Clearance (mL/mm)

Welght (kg)

150 130 110 100 9O 80 70 60

120 110 100 90 80 70 60

50

50

40

40

30

30

Serum creatlnlne (mg/100 mL)

R

--50

Age (years)

2.40 --3o ~2o 35 45 55 65 75 85

.'---TT 13 - 1211

95

06

--1009 08 07 05 (I 4

20

10

--

Fig 1. Nomogram for rapid evaluation of endogenous-creatinine clearance. With a ruler, join weight to age. Keep ruler at crossing-point of line marked R Then move the right-hand side of the ruler to the approprmte serum creatinine value and read the patient's clearance from the left side of the nomogram. To convert the units for serum creatinine, multiply ~ m o l / L by 0.0113 to obtain milligrams per 100 mL. (Reprinted with permmslon,

correlations are considered to be significant at p < 0.05. Regression and Deming analysis indicating correlation coefficients, agreements, and slope equations were obtained by plotting measured creatinine clearance (x axis) against predicted creatinine clearance (y axis). Imprecision of plasma creatinine concentration and creatlnme clearance measurement were obtained as 5% and 7%, respectively.

4)

changed at least 30% (Demlng r = 0.68; correlation r = 0.8, n = 43, p < 0.001); during intravenous feeding (Deming r = 0.7; correlation r = 0.93, n = 19,p < 0.001); but not in patients with preexisting renal dysfunction (Deming r = 0.36; correlation r = 0.57, n = 27,p = 0.002). The mean predicted creatinlne clearance for all patients was 57 (53 to 62) mL/minute, whereas the mean measured Day 2

Day 1 RESULTS

Seventy-five patients (54 men, 21 women) were studied during a total of 93 patient days up to and including the fifth postoperative day. Preexisting renal dysfunction was evident in 16 patients. Renal support therapy, usually hemofiltration, was used for 2 patients. Twenty patients who were discharged from the intensive care unit within 24 hours of operation were followed up on the ward for 2 days. Because there was great variability in the plasma creatinine concentration, marked changes occurred for indwidual patients. The m e a n change in the plasma creatinine concentration was 9% (5% to 12%) for the first and second postoperatwe days. Comparison of measured creatmine clearance with predicted creatlnine clearance showed correlations (ranging between r = 0.84-0.95;p < 0.05) up to the fourth postoperative day. Deming analysis confirmed agreement (Deming r = 0.63-0.84; p < 0.001). On the fifth day the correlation was less (Deming r = 0.75; correlation r = 0.76, n = 7, p < 0.05)(Fig 2). Similar agreement was observed when inotropes were used ( D e m l n g r = 0.7; correlation r = 0.85, n = 71, p < 0.001); vasoconstrictors were used (Demlng r = 0.67; correlation r = 0.9, n = 48, p < 0.001); renal function

140

140

120

120

100

100

a0

80

60

60

40

40

20

20

20

40

60

8O

100

120

140

0

20

40

Day 3

60

80

100

120

140

Day 4

140

140

120

Z20

J

100 80 60

40 20 0

20

I0

60

80

100

120

lt0

~0

40

60

80

100

120

140

F.g 2 Correlations of measured (x axis) with predicted (y axis) creatinine clearance, days 1 to 4; "'r" is correlation coefficient, "'rd" is agreement by Deming analysis. Day 1 (n = 33, r = 0 92, rd = 0 63, p < 0 . 0 0 1 ) , day 2 ( n = 2 2 , r = 0 8 4 , r d = 0 8 0 , p <0.001), day 3 ( n = 2 0 , r = 0 . 9 0 , rd=O.68, p <0.001); day 4 ( n = 1 1 , • = 0 9 5 , rd = 0.84,p < 0.001}.

CREATININE CLEARANCE PREDICTION AFTER SURGERY

creatinine clearance was 66 (60 to 73) mL/minute. Although consistent correlation is confirmed, comparison by paired t test shows that there is significant difference between predicted and measured creatinine clearance in all groups, except patients on the fifth postoperative day (t = 0.8014,p = 0.45, n = 7), patients who showed a reduction in renal function (t = 1.1996, p = 0.24, n = 43), and patients with preexisting renal dysfunction (t = 1.2686, p = 0.22, n = 27). The mean prediction bias was -9.1 (5.8-12.4) mL/minute for all patients.

DISCUSSION

The liver produces creatine phosphate, which is converted to creatinine during muscle metabolism. The plasma creatinine concentration is determined by the balance between production and excretion of creatinine and can be a poor predictor of renal function in certain circumstances: (1) animal protein ingestion within the preceding 8 hours, (2) muscle damage; (3) starvation or catabolism leading to muscle breakdown; (4) drugs such as salacylates,5 cotrimoxazole, 6 and clmetidine, 7 which may inhibit tubular secretion of creatinine; (5) compounds that interfere with the laboratory analysis such as acetoacetate in ketosls 8 and bilirubm in jaundice9; (6) recent changes in renal function from which a new steady-state level in plasma creatlnine has not been reached1°; and (7) liver dysfunction. Muscle damage and catabolism are common In the period after major surgery. Liver dysfunction and renal damage are well recognized. Thus, there are good reasons to expect change in plasma creatlnine concentration associated either with nonrenal causes or with renal insult from which a new steady state has yet to be reached. Under these circumstances, the relationship between plasma creatinine concentration and creatinine clearance would be expected to break down. The use of creatlnine clearance measurement to assess renal function has been reviewed recently, I1 but postoperative patients were not included. Creatinine is secreted by the renal tubules, and this increases as the G F R decreases. Creatmine clearance is thus a poor measure of the GFR. When the G F R is at least 30 mL/min, measured creatinine clearance overestimates it by as much as 15%. Creatlnine clearance predictions are known to be 13% (range 6 to 20%) less than measured values and thus more closely reflect the GFR. 2 When the G F R is less than 30 mL/min, predicted creatinlne clearance is greater than measured creatinine clearance because of the increased relevance of nonrenal creatinine excretion. Measured creatinine clearance is less reproducible because of two extra problems. There is a collection error in the clinical setting because of the inaccuracies in timing and collection of the total volume of urine produced. Also, the "between-batch" coefficient of variation (CV) increases when a result is dependent upon two readings. The resulting CV is the square root of the sum of the square of each individual error. I2 This study has shown a 9% (5% to 12%) change in the

901

plasma creatxnine concentration associated with major surgery in patients who had an uneventful postoperative course and were discharged to the ward on the first postoperative day. A change of up to 4% could be attributable to repeat measurement variability. Thus, these figures represent true change in the plasma creatinine concentration. In a variety of circumstances, this study has shown significant correlation between predicted and measured creatmine clearance. These findings compare favorably with previously published results from healthy volunteers and patients with stable chronic renal failure of correlations (r) between 0.82 to 0.9, depending on the formula used. 13 Correlation of r = 0.76 (p < 0.05) was obtained for patients on the fifth postoperative day. However, only seven patients were studied, reflecting the small numbers of cardiac surgery patients who require intensive care for a prolonged period Correlation of r = 0.57 (p = 0.002) was obtained for patients with preexisting renal dysfunction. Prediction of creatinine clearance is unsuitable for these patients. Evidence of correlation IS not the same as showing agreement between the two methods. For all patients studied, there was a prediction bias of - 1 3 % (9% to 18%). Agreement is confirmed by Deming analysis. Considering the variability and overestimation of the creatinlne clearance measurement method, the authors confirm that for assessment of renal function, predicted creatinine clearance as at least as accurate as measured creatinine clearance. Many formulae have been described in the last 20 years to predict creatinine clearance. ~3,14The authors chose to use a nomogram to predict creatinine clearance and compare their correlations with previously published data using these well-known formulae. Prediction of creatinlne clearance from a single plasma creatinine result has previously been constdered too inaccurate in patients who have changing plasma creatinine concentrations.2.3 i1 The authors have shown that for assessment of renal function, the prediction methods based on a single plasma creatxnme concentration are comparable with creatinine clearance measurement in postoperative cardiac surgical patients. Additionally, they have shown that the plasma creatinine concentration changes in an unpredictable manner for individual patients. However, the effect is small and the correlation between predicted and measured creatlnine clearance is maintained. Doctors have previously been tested for their ability to determine creatmlne clearance when the plasma creatinine concentration, sex, age, and body weight are known. 3 The outcome was reported as little better than guesswork. It has therefore been suggested that routine biochemistry reports could include a computer-generated creatinlne clearance prediction per 70 kg of body weight, using sex and age data on the request card. It would then be a simple matter to proportionately increase or decrease the result for those over or under this weight, respectively.

902

BLOOR ET AL

REFERENCES 1 Cameron JS" Acute renal failure m the intensive care unit today Intens Care Med 12.64-70, 1986 2 Luke DR, Halstenson CE. Opsahl JA, et al. Validity of creatlmne clearance estimates in the assessment of renal funcUon Chn Pharmacol Therapeut 48'5,503-508, 1990 3. Smith SA' Est~matlon of glomerular filtration rate from the serum creatinlne concentration Postgrad Med J 64 204-208, 1988 4. Slersbaek-Nielsen K, Hansen JM, Kampmann J, et al. Rapid evaluation of creatlnme clearance Lancet 1.1133-1134, 1971 5 Burry HC, Dieppe PA Apparent reduction of endogenous creatinine clearance by salicylate treatment Br Med J 2.16-17, 1976 6 Shouval D, Ligumsky M, Ben-Ishay D' Effects of cotrlmoxazole on normal creatmlne clearance Lancet i'244-245, 1978 7 Larsson R, Bodemar G, Kagedal B, et al: The effects of clmetldine on renal function In patients with renal failure. Acta Med Scand 208.27-31, 1980

8. Cook JGH. Factors influencing the assay of creatlnlne Ann Clin Biochem 12'219-232, 1975 9. Osberg IM, Hammond 1<33" A solution to the problem of blhrubin interference with the kinetic Jaffe method for serum creatmme. Chn Chem 24 1196-1197, 1978 10 Payne RB, Morgan DB: Sodmm, water and acid-base balance' Teaching transient and steady states Med Educ 28.275280, 1977 11. Payne RB' Creatlmne clearance. A redundant clinical investigation Ann Clin Blochem 23:243-250, 1986 12. Morgan DB, Dillon S, Payne RB The assessment of glomerular function. Creatinine clearance or plasma creatlnine9 Postgrad Med J 54 302-310, 1978 13 Robertshaw M, Lai KN, Swamlnathan R: Prediction of creatlnine clearance from plasma creatimne' Comparison of 5 formulae Br J Chn Pharmacol 28.275-280, 1989 14. Cockcroft DW. Gault MH. Pre&ction of creatmine clearance from serum creatinme Nephron 16'31-41, 1976