Predictive factors for acute renal failure in thoracic and thoracoabdominal aortic aneurysm surgery

Predictive factors for acute renal failure in thoracic and thoracoabdominal aortic aneurysm surgery H a z i m J. Sail, M D , S t u a r t A. H a r l i n , M D , Charles C. Miller, P h D , D i m i t r i o s C. I l i o p o u l o s , M D , A n t i Joshi, M D , M o h a s c i T a b o r , MS, R o l a n d Zippel, M D , and G e o r g e V. Letsou, M D , Houston, Tex.

Purpose: The purpose of this study was to analyze the factors associated with acute renal failure in total descending thoracic and thoracoabdominal aortic aneurysm surgery. Methods: A total of 234 patients underwent thoracoabdominal aortic aneurysm or total descending thoracic aneurysm repair between January 1991 and January 1994. Eighty-five women and 149 men were evaluated. The median age was 67 years (range 8 to 88 years). Seventy-seven patients had type I thoracoabdominal aortic aneurysm, 99 had type II, 51 had type I I I or IV, and 7 had total descending thoracic aneurysm. Factors such as age, sex, aneurysm type, and visceral and distal aortic perfusion were examined with univariate fourfold table and multivariate logistic regression analysis. Results: Acute renal failure, defined as an increase in serum creatinine by I m g / d l per day for two consecutive days after surgery, occurred in 41 (17.5%) of 234 patients. Thirty-six (15%) of 234 patients required dialysis. Twenty (49%) of 41 patients with acute renal failure died. O f the 21 survivors with renal failure, renal failure resolved in 18 (86%) within 30 days of surgery. The univariate odds ratio of death, given acute renal failure, was 6.7 (95% confidence interval [CI] 3.2 to 14.2, p < 0.0001). No significant association was found between the probability of acute renal failure and age, sex, hypertension, right renal artery reattachment, or renal bypass. Factors associated with increased risk of acute renal failure in multivariate analysis were visceral perfusion (odds ratio [OR] = 3.6 95%, CI 1.2 to 11.0, p < 0.02), left renal artery reattachment ( O R = 4.4 95%, CI 1.6 to 11.9, p < 0.004), preoperative creatinine >_2.8 m g / d l (OR = 10.3, 95% CI 12.0 to 411.8, p < 0.0001), and simple clamp technique (OR = 3.4 95%, CI 1.07 to 10.76, p < 0.04). Direct univariate correlation was seen between preoperative creatinine and acute renal failure ( O R = 3.2 per m g / d l increase, 95% CI 2.7 to 10.1, p < 0.0001). Conclusion: Postoperative acute renal failure after thoracoabdominal and total descending thoracic aortic aneurysm surgery is associated with preoperative creatinine level, visceral perfusion, left renal artery reattachment, and simple cross-clamp technique. (J Vase Surg 1996;24:338-45.)

Surgical repair o f thoracoabdominal aortic aneurysms (TAAA) and descending thoracic aortic aneurysms (DTAA) is becoming commonplace. 1-7 This is coupled with a decrease in the incidence ofneurologic complications and an increase in the survival rate. However, despite these achievements the incidence o f From Baylor College of Medicine, The Methodist Hospital. Presented at the Twentieth Annual Meeting of the Southern Association for Vascular Surgery, Naples, Fla., Jan. 24-27, 1996. Reprint requests: Hazim J. Sail, MD, 6550 Farmin, Suite 1603, Houston, TX 77030. Copyright © 1996 by The Society for Vascular Surgery and International Society for Cardiovascular Surgery, North American Chapter. 0741-5214/96/$5.00 + 0 2 4 / 6 / 7 4 5 2 1

338

renal failure that occurs with this surgery remains high and is associated with high early and late mortality rates. Surgical adjuncts to decrease renal failure include perfusion o f the renal arteries with cotd Ringer's lactate solution and various methods o f blood perfusion to the kidneys, s'9 Perfusion methods include distal aortic per fusion and direct perfusion o f the renal ostia (visceral perfusion)fi Treatment during surgery with agents such as mannitol and dopamine is common. Newer txials with calcium channel blockers, free radical scavengers, and angiotensin antagonists have also been r e p o r t e d ) °-12 The aim of this study was to evaluate both

JOURNAL OF VASCULAR SURGERY Volume 24, Number 3

Sail et al.

339

LL at~al appendage

~___._J Biomedicus pump

Heat exchanger

Fig. 2. Visceralperfusion technique. Fig. 1. Distal aortic perfusion technique.

preoperative and intraoperative factors that lead to acute renal failure in 234 patients who underwent TAAA and DTAA repair. MATER]LAL AND M E T H O D S Two hundred thirty-four consecutive patients underwent surgery by the senior author (H. J. S.) for TAAA or DTAA between January 1991 and January 1994. Their charts were reviewed, and the information was entered into a database that formed the basis for this report. Onc hundred forty-nine patients were male; 85 were female. The median age was 67 years; ages ranged from 8 to 88 years. One hundred seventysix patients had hypertension, 73 had chronic obstructive pulmonary disease, and 21 had diabetes mellitus. One hundred two patients were smokers. The extent of the aneurysms were as follows: 77 patients had TAAA type I, 99 type II, 25 type III, 26 type IV, and 7 total DTAA. Acute renal failure was defined by an increase in serum creatinine of 1 m g / d l each day above the preoperative value for two con-

secutive days or by requirement of hemodialysis.13 The median preoperative creatinine level was 1.1 mg/dl; preoperative creatinine levels ranged from 0.9 to 13.7 mg/dl. The operative technique for repair of these aneurysms has been reported previously in detail. 1 All patients received general anesthesia, Swan-Ganz catheters, large central venous access for fluid resuscitation, and bilateral arterial radial lines for monitoring blood pressure. A double-lumen endotracheal tube was used. The patient was placed in the right lateral oblique decubitus position. The TAAA incision was made, and the chest was entered through the sixth intercostal space. The left lung was collapsed during the surgical repair. The abdominal viscera were rotated medially, and the ieft renal artery, a landmark for TAAA classification, was exposed. The celiac axis and the superior mesenteric, right, and left renal arteries were reattached together to the graft in a single island of tissue; occasionally the left renal artery was used as a separate button. Beginning in September 1992 cerebral spinal fluid drainage was used to keep cerebral spinal fluid pres-

340

JOURNALOF VASCULARSURGERY September 1996

Sail et al.

T a b l e I. Univariate analysis o f p a t i e n t characteristics a n d acute renal failure

Variable

No. of patients (%)

No. of renal failures (%)

All patients Preoperative creatinine (mg/dl) 0.4-0.9 0.9-1.0 1.1-1.3 1.4-13.7 Preoperative renal insufficiency Otherwise TAAAtype I Otherwise TAAAtype II Otherwise TAAAtype III Otherwise TAAAtype IV Otherwise Descending Otherwise Right renal artery Otherwise Left renal artery Otherwise Distal perfusion (all) Otherwise Visceral perfusion (all) Otherwise Distal perfusion (only) Otherwise Visceral perfusion (only) Otherwise Both visceral and distal perfusion Otherwise Simple clamp Otherwise Paraplegia No paraplegia

234 (100.0)

41 (17.5)

47 57 59 71 24 210 77 157 99 135 25 209 26 208 7 227 136 98 100 134 146 88 58 176 90 144 2 232 56 178 86 148 36 197

3 8 3 27 10 31 4 37 26 15 4 37 7 34 1 40 27 14 26 15 22 19 17 24 6 35 1 40 16 25 18 23 13 28

(20.1) (24.4) (25.2) (30.3) (10.3) (89.7) (32.9) (67.1) (42.3) (57.7) (10.7) (89.3) (11.1) (88.9) (3.0) (97.0) (58.1) (41.9) (42.7) (57.3) (62.4) (37.6) (24.8) (75.2) (38.5) (61.5) (0.9) (99.1) (23.9) (76.1) (36.8) (63.2) (15.5) (84.5)

(6.4) (14.1) (5.1) (38.0) (41.6) (14.8) (5.2) (23.6) (26.3) (11.1) (16.0) (17.7) (26.9) (16.4) (14.3) (17.6) (19.9) (14.3) (26.0) (11.2) (15.1) (21.6) (29.3) (13.6) (6.7) (24.3) (50.0) (17.2) (28.6) (14.0) (20.9) (15.5) (36.1) (14.2)

OR

95% CI*

Pt

3.23

1.93-5.39

0.001 (0.001)

4.13

1.68-10.11

0.001

0.18 1 2.85 1 0.89 1 1.89 1 0.78 1 1.49 1 2.79 1 0.64 1 2.63 1 0.22 1 4.80 1 2.45 1 1.44 1 3.41 1

0.06-0.52

0.001

1.42-5.73

0.003

0.29-2.73

0.83

0.77-4.83

0.18

0.09-6.65

0.82

0.73-3.01

0.27

1.39-5.61

0.003

0.33-1.27

0.20

1.29-3.35

0.006

0.09-0.55

0.001

0.29-78.36

0.23

1.19-5.02

0.01

0.73-2.85

0.30

1.55-7.51

0.002

Simple clamp, Cross damp-and-go technique without perfusion adjuncts. Fifty-six patients had both distal and visceralperfusion. Perfusion types labeled (all) = total number of perfusions of the specified type regardless of whether the other type was present. Perfusion types labeled (only) = specifiedperfusion type alone. Paraplegiarepresents both early and late paraplegia. For dichotomous variablesOR represents a test against a reference category whose referent OR equals 1. For continuous data OR refers to increase in odds associatedwith a one-unit increasein variablevalue. This value can be converted to an OR for an Nunit change in the variable by taking the natural logarithm of the OR, multiplying this value by the number of variable units desired, and taking the exponent of the product. Although continous data are presented in quarriles, ORs are against the continuous variable. *95% CI reflects the units against which its companion OR is computed. CIs are test-based. }p Value = probabilityof type I statisticalerror (commonpvalue).Valueswithout parentheses are Pearson Z2probabilities.Probabilityvalues in parentheses are univariate logistic regression likelihood ratio p values. sure less t h a n 10 m m H g . This p r o c e d u r e was used for all patients except those with r u p t u r e , previous back operation, or mycotic aneurysm. 134 Perfusion o f the distal aorta or visceral arteries was used as an a d j u n c t in m o s t cases. Distal aortic perfusion alone, as illustrated in Fig. 1, was used w h e n the a n e u r y s m did n o t extend below the renal arteries ( D T A A and TAAA type I). Visceral perfusion, as illustrated in Fig. 2, was used in c o n j u n c t i o n with distal perfusion for repair o f a n e u r y s m s that e x t e n d e d below the renal arteries (TAAA type II). F o r distal aortic per fusion the left atrium was c a n n u l a t e d with a no. 2 6 Bard aortic c a n n u l a (C.R.Bard, Inc., Haverhill,

Mass.) c o n n e c t e d to a Biomedicus p u m p (Biomedicus, Minneapolis). T h e left femoral artery was exposed t h r o u g h a l o n g i t u d i n a l g r o i n incision a n d c a n n u l a t e d with an 18F to 2 2 F c a n n u l a also conn e c t e d to the p u m p . T h e p a t i e n t received heparin, a n d the p u m p was initiated. T h e flow was d e p e n d e n t o n the proximal aortic pressure, which was kept at m o r e than 100 m m H g . I n cases o f femoral artery occlusion or previous aortofemoral bypass, the distal thoracic aorta or the u p p e r a b d o m i n a l aorta was used as inflow. Visceral pcrfusion a c c o m p a n i e d the latter part o f the clamping sequence used in TAAA type I I repair. We used n o further adjunctive renal protective modalities

JOURNAL OF VASCULAR SURGERY Volume 24, Number 3

such as dopamine or diuretic therapy. After the proximal aortic anastomosis and the same manner of aortic cross-clamping and distal aortic perfusion shown in Fig. I were completed, the clamp was moved down on the aorta below the renal arteries, and the celiac, superior mesenteric, and renal arteries underwent perfusion with no. 9 Pruitt catheters. The flow rates collectively ranged from 150 to 450 ml/min; median flow was approximately 250 ml/min. Successful flow was indicated by patient urine output of 1/2 cc/min. Intraoperative data. The median aortic clamp time was 45 minutes; times ranged from 13 to 131 minutes. The median renal perfusion time was 32 minutes; times ranged from 14 to 52 minutes. The median visceral perfusion time was 33 minutes; times ranged from 14 to 55 minutes. Visceral ischemia was measured for patients with TAAA type II. The median visceral ischemic time for these patients was 3 minutes; the times ranged from 2 to 6 minutes. Statistical analysis. Patient data were abstracted from medical charts and maintained in a computer database. Statistical analyses were performed with SAS software (SAS Institute, Inc. SAS Software Release 6.10, Cary, N. C.). Descriptive statistics for continuous variables are given by median and range. Dichotomous variable descriptive statistics are presented as frequencies. Univariate odds ratios for dichotomous risk factors were computed with standard fourfold table methods with test-based 95% confidence intervals (CI). Univariate odds ratios (OR) for continuous variables were computed with logistic regression analysis. Multivariate risk factor sets were: evaluated initially with stepwise logistic regression analysis followed by best subsets selection. For the final model adjusted O R with 95% CI, Hosmer-Lemeshow goodness-of-fit, and a Cox/Snell generalized r-square were computed (SAS Institute, Inc. SAS/STAT Software, Changes and Enhancements, Release 6.10. The LOGISTIC Procedure). The series of multivariate risk factors identified in the final logistic model was further evaluated by MantelHaenszel stratification to examine the appropriateness of the logistic model's exponential form given the data. RESULTS Univariate analysis of patient characteristics and resulting acute renal failure are presented in Table I. The overall incidence of acute renal failure was 41 (17.5%) of 234 patients. Thirty-six (15%) of 234 patients required hemodialysis. The mortality rate for those patients with acute renal failure was 20 (49%) of

Sail et aL

341

Figure 3 ROC Curve for Multiple Logistic Model

1.°1

0.9 0.8 0.7

0.6 0.5

0.4 0.3 0.2 0.1 0.0 . . . . . . . . . . .

0.0

,

. . . .

0.1

,

. . . .

,

. . . .

j

. . . .

,

. . . .

,

. . . .

,

. . . .

j

. . . .

,

. . . .

0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 - Specificity

,

1.0

Fig. 3. Receiver-operating characteristic curve. 41 patients. Of the 21 survivors with renal failure, renal failure resolved in 18 (86%) within 30 days of surgery. Univariate analysis revealed that the following factors were associated with the development of acute renal failure: left renal artery reattachment, TAAA type II, elevated preoperative creatinine, preoperative renal insufficiency, visceral perfusion, simple aortic clamp technique, and distal aortic perfusion. Factors not associated with statistical significance after univariate analysis included age, sex, hypertension, right renal artery reattachment, TAAA type I, type III, and type IV, DTAA, smoking, chronic obstructive pulmonary disease, renal perfusion time, and the amount and type of blood products given. The significant factors in the univariate model were subjected to multivariate analysis and are shown in Table II. Preoperative renal insufficiency was a univariate risk factor for renal failure, but its effect did not persist in multivariate analysis. TAAA type II became insignificant when studied in this manner (p < 0.23) but was retained in the model to account for extent of aneurysm for which visceral perfusion was used. Left renal artery reattachment (p < 0.004) and visceral perfusion (p < 0.02) continued to be significant, as did simple aortic clamp technique. In the multivariate logistic model distal aortic perfusion was used as the reference category for perfusion type.

342

j'OURNAL OF VASCULARSURGERY September 1996

Sail et al.

Table II. Multiple logistic regression model Variable

Parameter estimate

Adjusted OR

95% CI

p

Intercept Creatinine >2.8 Left renal artery Visceral perfusion Simple clamp TAAA type II

-3.9355 4.2531 1.4784 1.2801 1.2202 0.5569

70.33 4.39 3.60 3.39 1.75

12.01-411.75 1.62-11.91 1.18-10.97 1.07-10.76 0.67-4.52

0.0001 0.0001 0.004 0.02 0.04 0.25

Creatinine >2.8 = Preoperative creatinine >2.8 mg/dl. Model developed by stepwise multiple logistic regression analysis.

Both visceral perfusion and simple aortic clamping had threefold, significant increases in odds over distal aortic perfusion. Elevated preoperative creafinine (>2.8 mg/dl) proved to be the most significant variable (p < 0.001). Preoperative creafinine levels taken as a continuous variable did not remain significant in multivariate analysis. The cutpoint value of 2.8 mg/dl used in the multivariate model had the highest classification accuracy (86%) of all cutpoint values evaluated. Highest overall classification accuracy for the final model (shown in Table II) was 87%. A receiver-operating characteristic curve is shown in Fig. 3. This curve displays the sensitivity versus 1-specificity for the range of model predicted probabilities. The closer the curve goes to the upper left-hand corner of the graph, the better the predictive accuracy of the model is. This semiquantitative method of evaluating model accuracy provides a visual description of the tradeoff between sensitivity to detect and specificity to classify observations over the range of the model's predictive domain. The greater the area under the curve is, the greater the predictive accuracy of the model is. Area can range from 0 to 1. For this model the area under the curve is 0.823. is R-square for the final logistic model was 0.37, indicating that 37% of the variance in postoperative acute renal failure could be explained by the risk factors listed in Table II. Model fit was good (Hosmer-Lemeshow statistic 4.45 with 6 dr, p = 0.6157, with 69% of expected values being greater than 5), and no important departures from the exponential form of the logistic model were detected in the stratified analysis. DISCUSSION

The incidence of acute renal failure for TAAA or DTAA repair varies between 4% and 29% depending on the series. 3'5'6'8 Previously the incidence has been associated with age, male sex, renal occlusive disease, preoperative renal failure, stroke, elevated preoperative creatinine, and visceral ischemia, s,8 In this series acute renal failure developed in 41 (17.5%) of 234

patients undergoing DTAA or TAAA repair. Of the patients who had acute renal failure, 36 (88%) of 41 went on to require hemodialysis, and 20 (49%) of 41 patients died. Acute renal failure after repair of TAAA and extensive DTAA is associated with both high early and high late death rates. Age, sex, hypertension, right renal artery reattachment, DTAA, TAAA types I, III, IV, renal perfusion time, and chronic obstructive pulmonary disease were not associated with an increased risk of renal failure. In univariate analysis the risk factors revealed were TAAA type II, elevated preoperative creatinine, preoperative renal insufficiency, visceral perfusion, simple aortic clamp technique, and left renal artery reattachment. Paraplegia was also associated with acute renal failure (OR, 3.4; 95% CI, 1.6 to 7.5; p < 0.04) but not independently with increased mortalky. The association between acute renal failure and paraplegia is unlikely to be causal but probably represents a common underlying relationship with endorgan ischemic injury. Distal aortic perfusion was protective against renal failure. A working hypothesis during the exploration of these data was that the increased acute renal failure associated with visceral perfusion was due at least in part to confounding between aneurysm extent (TAAA type II in particular) and the perfusion technique required to support the kidneys during this type of repair. Multivariate statistical methods did not separate these effects mathematically, indicating that the risk associated with visceral perfusion is not accounted for entirely by the extent of the aneurysm. We also considered the possibility that duration of artificial circulation could account for the apparent difference between the effects of visceral and distal perfusion. Because repair of aneurysms requiring visceral perfusion tends to take longer than operations that can be accomplished with distal perfusion, we examined the effect of renal perfusion time on probability of renal failure. We were unable to detect any significant effect of perfusion time, so duration of artificial circulation does not appear to have been a

JOURNAL OF VASCULARSURGERY Volume 24, Number 3

factor in the increased risk associated with visceral perfusion. Distal aortic perfusion demonstrated a protective effect against acute renal failure in patients undergoing DTAA and TAAA. Although the exact mechanism is unclear, distal aortic perfusion provides indirect nonpulsatile flow to the viscera, and there is usually no ischemic time to the kidneys with this method unless the patient has intraoperative or postoperative hypotension. The :mechanisms behind the differences in renal failure rates seen with visceral and distal perfusion are currently unexplained. A number of clinical trials have recently been published that have studied the adjunctive effects of different regimens on renal failure. Loop diuretics and free radical scavengers did not prove effective, although an increased return of glomerular filtration rates occurred after renal failure with hemodilution. 16 Inhibition of endothelial derived relaxing factor aggravated ischemic renal failure. 17 3mother study suggested that endothelin antagonists may provide a protective effect on the development of postischemic renal failure. ~s A profound increase in angiotensin II was found in dogs undergoing nonpulsatilc cardiopulmonary bypass, with pulsatilc perfusion attenuating this response. The authors noted that angiotensin II causes profound splanchnic vasoconstriction that may aggravate renal ischemia and lead to renal failure. 19 Finally, a normalization of renal cortical blood flow was demonstrated after a converting enzyme inhibitor was infused in dogs undergoing thoracic aortic crossclamping .20 We conclude that several factors contribute to the risk of acute postoperative renal failure. Patients with elevated preoperative creatinine are at particular risk for the development of renal failure; in our series a threefoM increase in univariate odds of acute renal failure was seen with each milligram per deciliter increase. Distal aortic per fusion demonstrated a protective effect against acute renal failure in patients undergoing DTAA and TAAA repair, although direct perfusion of the renal arteries with nonpulsatile blood flow has; demonstrated a deleterious effect. Use of a simple cross-clamp technique is also associated with a worse renal outcome. The finding that direct visceral perfusion is associated with increased risk is unexpected and difficult to explain. It is apparently not due to aneurysm extent nor to duration of artificial circulation. Further experimental studies are warranted to elucidate the mechanisms of renal failure with direct per fusion. Such studies include randomized trials of pulsatile versus nonpulsatile per fusion

Sat et al.

343

adjuncts, use versus nonuse of direct perfusion, and studies of pharmacologic agents with specific renal action that may be useful in further reducing the serious problem of renal failure in TAAA and DTAA surgery. Grateful acknowledgment to Amy Wirtz Newland for editorial assistance. REFERENCES 1. Sail HJ, Bartoli S, Hess KR, et al. Neurologic deficit in patients at high risk with thoracoabdominal aortic aneurysms: the role of cerebral spinal fluid drainage and distal aortic perfusion. J Vase Surg 1994;20:434-43. 2. Svensson LG, Crawford ES, Hess KR, Coselli JS, Sail HJ. Experience with 1509 patients undergoing thoracoabdominal aortic operations. J Vase Surg 1993;17:357-70. 3. Frank SM, Parker SD, Rock P, et al. Moderate hypothermia with partial bypass and segmental sequential repair for thoracoabdominal aortic aneurysm. J Vase Surg 1994;19:68797. 4. Archer CW, Wynn MM, Archibald J. Naloxone and spinal fluid drainage as adjuncts in the surgical treatment of thoracoabdominal and thoracic aneurysms. Surgery I990;108:755-62. 5. Schepens/VIA, Defauw JJ, Hamerlinjnck MP, Vermeulen FE. Risk assessment of acute renal failure after thoracoabdominal aortic aneurysm surgery. Ann Surg 1994;219:400-7. 6. Golden MA, Donaldson MC, Whittemore AD, Mannick JA. Evolving experience with thoracoabdominal aortic aneurysm repair at a single institution. J Vase Surg 1986;i3:792-7. 7. Cox GS, O'Hara PJ, Hertzer NR, Piedmonte MR, Krajewski LP, Beven EG. Thoracoabdominal aneurysm repair: a representative experience. J Vase Surg 1992;15:780-8. 8. Svensson LG, Coselli IS, Sail HJ, Hess KR, Crawford ES. Appraisal of adjuncts to prevent acute renal failure after surgery on the thoracic or thoracoabdominal aorta. J Vase Surg 1989;10:230-9. 9. Sail HJ, Cox GS. A new technique for left renal cryopreservation. J Am Coll Surg 1994;178:629-31. 10. Waite RB, White G, Davis JH. Beneficial effects ofverapamil on postischemic renal failure. Surgery 1983;94:276-82. 11. Pedraza-Chaverri J, Tapia E, Bobadilla N. Ischemia-reperfusion induced acute renal failure in the rat is ameliorated by the spin-trapping agent a-phenyl-N-tert-butyl Nitrone (PBN). Ren Fail 1992;14:467-71. 12. Abbott WM, Abel RM, Beck CH. The reversal of renal cortical ischemia during aortic occlusion by mannitol. J Surg Res i974;16:482-7. 13. Hou SH, Bushinsky DA, Wish JB, Cohen JJ, Harrington JT. Hospital acquired renal insufficiency: a prospective study. Am J Med 1983;74:243-8. 14. Hollier LH, Money SR, Naslund TC, et al. Risk of spinal cord dysfunction in patients undergoing thoracoabdominal aortic replacement. Am J Surg 1992;164:210-4. 15. Bamber D. The area above the ordinal dominance graph and the area below the receiver operating characteristic graph. J Mathem Psychol 1975;12:387-415. I6. Wolgast M, Bayati, Hellberg O, Kallskog, Nygren K. Osmotic diuretics and hemodilution in postischemic renal failure. Ren Fail 1992;14:297-302. 17. Chintala MS, Chin PJS, Vemapulli S, Watldns RW, Sybertz EJ. Inhibition of endothelial derived relaxing factor (EDRF)

JOURNAL OF VASCULARSURGERY

344

Sail et al.

aggravates ischemic acute renal failure in anesthetized rats. Naunyn-Arch Pharm 1993;348:305-10. 18. Kusumoto K, Kubo K, Kandori H, et al. Effects of a new endothelin antagonist, TAK-044, on post-ischemic acute renal failure in rats. Life Sci 1994; 55:301-10. 19. Watkins L, Lucas SK, Gardner TJ, Potrer A, WalkerWG, Gott VL. Angiotensin II levels during cardiopulmonary bypass: a comparison of pulsatile and nonpulsatile flow. Surg Forum 1979;30:229-30. 20. Joob AW,Dunn C, Miller E, Freedlender A, Kron IL. Effect of left atrial to left femoral artery bypass and renin-angiotensin system blockade on renal blood flow and function during and after thoracic aortic occlusion. J Vase Surg 1987;5:329-35. Submitted Jan. 31, 1996; accepted Apr. 29, 1996. APPENDIX Receiver-operating characteristic ( R O C ) curves are plots o f logistic regression m o d e l sensitivity versus 1-specificity. 1-specificity is also the n u m b e r o f false-positive results divided by the n u m b e r o f nonevents, which can be interpreted as the falsepositive rate. Because the desirable relationship between sensitivity and 1-specificity is negative (high sensitivity with low false-positive rate), an optimal R O C curve will rise into the upper left hand corner o f the graph (Fig. 3). The faster the rise is, the better the predictive accuracy o f the model is. Sensitivity and specificity o f the model vary according to predicted probability cutpoints over the model's predictive range that are used to predict w h e t h e r observations bearing certain m o d e l characteristics will be classified

September 1996

as events or nonevents. For example, with the model developed for this study, a patient with a preoperative creatinine greater than 2.8, a left renal artery reattachment, and a clamp-and-go operation w o u l d have a high predicted probability (in this case [exp (42531 + 1,4784 + 1.2202- 3.9355) / l q- exp (4.2531 + 1.4784+ 1.2202- 3.9355)] :

95%; s e e

Table

II

for

coefficients) o f developing postoperative renal failure. This model instance w o u l d be very specific for renal failure (on the order o f 100%), because the person w o u l d be almost ccrtain to sustain it, but sensitivity w o u l d be low (approximately 5%; for both sensitivity and specificity model probability relationships are n o t shown on the R O C curve). That is, if we were to require that patients had to have creatinine >2.8, left renal artery reattachment, and simple clamp operation before we w o u l d consider t h e m to be at risk for renal failure, then nearly everyone we expected to have it w o u l d have it, but so w o u l d m a n y others we did n o t expect to have it. This is high sensitivity with low specificity. The point o f the R O C curve is to show that m o s t o f the area in the graph is contained underneath the curve, and this is so for the entire potential probability cutpoint range. The X and Y axes range from 0 to 1, so the m a x i m u m total area o f the graph is equal to 1. The "area u n d e r the curve is a semiquantitative measure o f the predictive accuracy o f the model. The greater the area is, the better the predictive accuracy is. In this study area u n d e r the curve is 0.823. A l t h o u g h no specific guidelines exist for comparing R O C areas, 0.823 is considered to represent g o o d accuracy.

DISCUSSION Dr. G. Melville Williams (Baltimore, Md.). Dr. Sail and the Baylor group have presented us with a rich experience in the repair of extensive ancurysms focused on the knotty and vexing problem of postoperative renal failure. Their 234-patient experience permits analysis of important potential etiologic factors by multivariate analysis. Some factors such as preoperative renal failure are logical and important. Patients having a preoperative serum creatinine in excess of 2.8 were 70 times more likely to have postoperative trouble. Similarly, those requiring left renal artery reconstruction separately had a fourfold increased risk. Baffling to me and to the authors is why perfusion of the viscera and renal arteries during the construction of the large patch anastomosis significantly increased the chances of renal dysfunction. This is counterintuitive, and the

authors provide no reasonable explanation from their data. Neither aneurysm extent nor the duration of renal perfusion correlated with postoperative renal dysfunction in this group of patients. I am left concluding that the less than ideal flow provided (which in this case was a mean of 250 ml/min to the visceral renal segment) caused harm. Low flow was worse than no flow. This is very valuable information. I have two questions. First, the authors used a heat exchanger in their per fusion circuit. In the article they did not say what this was used for, or what the body temperature was reduced to. Did this help in any way, or was this considered in their analysis? We believe that reducing the temperature to 32 ° throughout the body is helpful, although this is difficult to prove with our much smaller numbers.

JOURNALOYVASCULARSURGERY Volume 24, Number 3

My second question is, why were the flows to the kidney and the viscera so low? A flow of 250 m l / m i n is at least one-fifth lower than normal. Because outflow was split between tJhe legs/pelvis and renal/visceral segments, should the outflow to the legs be simply sacrificed and all the flow provided to the kidneys and viscera? Dr. Sail. We could not really accurately decide why patients undergoing direct visceral perfusion had an increased risk:. As you commented, we would expect them to have done better. A couple of things came to mind when we tried to sort it out. The first is the issue of pulsatile versus nonpulsatile direct perfusion to the kidneys. Nonpulsatile flow has been demonstrated to activate the renin angiotensin system and possibly increase renal vasoconstriction, which may have a deleterious effect. The second possible explanation we came up with was the actual trauma of putting the: catheter within the orifice of the artery. It may cause some sort of cholesterol embolization down in the artery, which may lead to renal failure. In answer to your other questions, we also use passive cooling to :approximately 32 °, and then we actively rewarm to 37 ° before coming offthe clamp. With passive cooling we have mainly observed that patients are less coagulopathic and come off the pump better. In regard to your question of why we maintain the flows we do to the individual viscera; we were unable to measure the flows to each cannula specifically. Rather, this was an aggregate flow through the entire system that ranged from 150 to 450 cc/min with an average of approximately 250 cc. That is really about all the pump would allow us in this specific circumstance.

Sail et al.

345

Dr. Cambria (Boston, Mass.). At Massachusetts General Hospital we have used a little different technical approach in the repair of these aneurysms. We continue to use a clamp-and-sew technique and adjunctive use of continuous cold per fusion to the renal arteries during the period of aortic cross-clamping. I have a couple of questions for you. In the patients who required dialysis, how many of these had substantial preoperative elevations of serum creatinine? I am aware of the previous work from your group that showed no benefit of cold per fusion, but given the current results with warm pump blood perfusion, are yon prepared to try it again? Last, with respect to the technical details of reconstruction of the left renal artery, isn't the positive predictive value of the requirement for a separate left renal artery reconstruction more related to total ischemic rime to the entire functioning renal mass than it is to any particular technique of surgical reconstruction? Dr. Sail. In regard to how many patients requiring hemodialysis had preoperative renal insufficiency, we looked at creatinine as a continuous variable, and as the creatinine went up, the incidence of renal failure went up. For the patients with preoperative renal insufficiency (or creatinine greater than 2), the incidence was approximately 40%. We have by and large abandoned direct cold crystalloid per fusion to the kidneys in favor of our distal aortic perfusion adjuncts. We studied ischemic time to the kidneys as a separate variable, but this did not come out as a predictive indicator in our data.

Pacific Vascular Research Foundation 1997 Wylie Scholar Award in Academic Vascular Surgery The Wylie Scholar Award provides up to 3 years of career development support for a promising young vascular surgeon. The award consists of a grant of $50,000 per year for 3 years. Funding for the second and third years is subject to review of acceptable progress reports. The award is nonrenewable and may be used for research support, essential expenses, and other academic purposes at the discretion of the Scholar and the medical institution. The award may not be used for any indirect costs. The candidate must be a vascular surgeon who has completed an accredited residency in general vascular surgery within the past 5 years and who holds a full-time appointment at a medical school accredited by the Liaison Committee on Medical Educators in the United States or the Committee for the Accreditation of Canadian Medical Schools in Canada. The applicant must be recommended by the administration (the Dean or fiscal officer) of the medical school and the head of the applicant's department or division. Only one candidate is eligible per institution. Applications are due by February 1, 1997, for awards granted July i , 1997. Applications may be obtained by writing to: Pacific Vascular Research Foundation, Wylie Scholar Award, 601 Montgomery Street, #900, San Francisco, CA 94111 (415) 291-7201.