Clarification of Risk Factors for Abdominal Operations

in Patients with Hepatic Cirrhosis

R. NEAL GARRISON, M.D., F.A.C.S., HENRY M. CRYER, M.D., DANIEL A. HOWARD, M.D.,
HIRAM C. POLK, JR., M.D., F.A.C.S.

Celiotomy in cirrhotic patients is reported to bear a high risk From the Department of Surgery, the University of Louisville
of operative morbidity and mortality. We reviewed 100 consec- School of Medicine, and the Veterans Administration
utive, cirrhotic patients who underwent nonshunt celiotomy. Medical Center, Louisville, Kentucky
Thirty patients died and major complications occurred in another
30 patients. Hospital mortality rate was 21% in 39 biliary op-
erations, 35% in 26 procedures for peptic ulcer disease, and
55% in nine colectomies. Fifty-two variables were compared
between survivors without complication, survivors with com- dergoing nonshunt celiotomies. Stone'5 has suggested that
plications, and nonsurvivors. A computer-generated, multivariant the Child's classification is useful in predicting the out-
discriminant analysis yielded an equation predictive of survival. come of any major surgical procedure although no data
Utilizing coagulation parameters, presence of active infection, were presented in his paper to support this conclusion.
and serum albumin, the equation predicted survival with 89%
accuracy. In a similar fashion, amount of operative transfusions, Pugh et al.'6 proposed a modification of the Child's clas-
absence of postoperative ascites, pulmonary failure, gastroin- sification, adding prothrombin time to replace the nu-
testinal bleeding, and culture-positive urine predicted survival tritional assessment in a series of patients undergoing
with 100% accuracy. We conclude that celiotomy in the cirrhotic transection of the esophagus with bleeding esophageal
patient is truly associated with very high morbidity and mortality, varices. In addition, several surgeonS2,3,5 " have reported
and preoperative assessment can predict survival with 89% ac- various preoperative variables that correlate with the high
curacy.
operative mortality in the cirrhotic patient. However,
Jackson' concluded that the prognosis for survival in cir-
DATIENTS WITH HEPATIC CIRRHOSIS frequently require rhotics cannot be predicted with current methods of eval-
an abdominal procedure. It has been estimated' that uation and that the ultimate outcome in an individual
ten per cent of all patients with liver disease undergo patient is not easily determined prior to the procedure.
operative procedures during the final 2 years oftheir lives, The purpose ofthis study was to further document the
when they represent the least satisfactory surgical risks. operative morbidity and mortality associated with an ab-
Furthermore, many investigators2'- have documented a dominal procedure in cirrhotic patients. An additional
high risk of morbidity and mortality associated with ab- objective was to utilize a computer-generated multivariant
dominal procedures in this group of patients. It is im- discriminant analysis to devise an equation for the ac-
portant to have a tool to assess the risk of operation in curate, preoperative assessment of operative risk.
cirrhotic patients before surgery. Many studies have at-
tempted to predict operative mortality by assessing vari- Patients and Methods
ables in patients undergoing portasystemic shunt pro-
cedures. Several of these authors'2-'4 have proposed mul- One hundred consecutive patients with liver cirrhosis
tifactorial indices to more accurately estimate surgical who underwent a celiotomy at University Hospital
risks. Child's classification remains the most successful. (Louisville General Hospital) and the Veterans Admin-
On the other hand, there is a lack of data concerning istration Medical Center in Louisville, Kentucky, between
the estimation of surgical risk in cirrhotic patients un- 1975 and 1982 were retrospectively reviewed. No patient
who underwent a portasystemic shunt or who had pre-
viously undergone that procedure was included in this
Presented at the Ninety-Fifth Annual Meeting ofthe Southern Surgical
review. Fifty-four preoperative, operative, and postop-
Association, December 5-7, 1983, Hot Springs, Virginia. erative variables were listed for each patient and cata-
Reprint requests: R. Neal Garrison, M.D., University of Louisville logued for three patient groups: survivors without post-
School of Medicine, Department of Surgery, Louisville, KY 40292. operative complications, survivors with postoperative
Submitted for publication: January 4, 1984. complications, and nonsurvivors.

648
Vol. 199 * No. 6 CELIOTOMY IN CIRRHOTIC PATIENTS 649
Preoperative data listed included age, sex, coexisting survived although they suffered at least one major com-
medical disease, mode of presentation for the underlying plication. Forty patients survived without complications.
operative disease, and the presence of encephalopathy, Sepsis with multiple organ system failure was the cause
malnutrition, ascites, and concomitant infection or peri- of death in 26 patients (87%) while bleeding accounted
toneal contamination. Malnutrition was listed as present for two deaths (7%). One death was attributed to meta-
only if so stated in the medical record. Ascites was con- static carcinoma and the last to methanol poisoning.
sidered present if noted before surgery or documented A variety of abdominal procedures were performed.
during the surgical procedure. Preoperative serum en- Thirty-nine biliary tract procedures constituted the largest
zymes, albumin and bilirubin levels, coagulation param- single group and resulted in mortality in eight patients.
eters, and hematologic values were similarly recorded. A cholecystectomy alone was performed on 29 occasions
Operative data listed included the operative procedure, with a mortality rate of 17% while an additional ten pa-
its duration, transfusion requirements, and the use of all tients underwent a choledochotomy with three deaths
drains or tubes that exited the peritoneal cavity. Post- resulting. The second most common indication for op-
operative variables noted were mortality and morbidity, eration was a complication of gastroduodenal ulcer dis-
length of hospital stay, the development of positive bac- ease. Perforation of the ulcer was the listed cause for
terial cultures or subsequent organ system failure, and exploration in 11 patients and five of these died. Four
antibiotic treatments. deaths were noted in 12 patients with bleeding as the
All patients had proven cirrhosis from either a previous indication for operative intervention. Two patients re-
biopsy or one taken at the time of the operation. Each quired correction of gastric outlet obstruction and one
patient was assigned a Child's classification based on the patient was explored for gastric lymphoma. A total of 26
most recent laboratory data and progress notes prior to patients underwent an operation on the gastroduodenal
the procedure. Pulmonary failure was defined as ventilator complex with a mortality rate of 35%. Nineteen of these
dependence for 5 days or more. If the patient was overtly procedures were performed as emergencies and all nine
oliguric or the serum creatinine exceeded 2 mg/dl, renal deaths occurred in this subset.
failure was considered present. Liver failure was defined Nine patients underwent colon resection resulting in
as a total serum bilirubin level greater than 2 mg/dl with five deaths. Perforated diverticuli required emergency op-
a concomitant doubling of the liver enzyme levels over eration in four patients and accounted for three deaths.
normal values. Gastrointestinal failure was defined as Three patients underwent an elective colectomy for car-
gastrointestinal bleeding that required the transfusion of cinoma and one death resulted. The last death in this
two or more units of blood products within a 24-hour group occurred in a trauma patient who underwent ex-
period, or endoscopic evidence of bleeding. A decrease teriorization of a colon injury.
in cardiac output requiring pharmacologic support when Five patients underwent open liver biopsy for the di-
the vascular volume was normal constituted cardiac agnosis of metastatic carcinoma with one death attributed
failure. to the metastatic disease. Five additional patients un-
The occurrence of each variable within patient groups derwent a negative exploratory laparotomy with four re-
was subjected to statistical analysis by. chi square for cat- sultant deaths. An exploration had been initiated to search
egorical variables and analysis of variance for continuous for a correctable source of infection in a septic patient
variables. All 54 variables except Child's classification with multiple organ system failure in each of these deaths.
were then included in a stepwise multivariant discriminant The lone survivor in this group was a patient with a
function analysis. An equation was derived to predict preoperative diagnosis of a pancreatic pseudocyst who
survival. A total of 66 patients with complete data was was found to have only a loculated collection of ascites.
included in this analysis. The discriminating equation Five patients had small bowel procedures. Small bowel
that this analysis developed was tested on all patients in obstruction required lysis of adhesions in two patients,
the study and the accuracy of the prediction evaluated. while two others needed small bowel resections for stran-
One equation for prediction of survival was derived for gulated inguinal hernias; these latter two patients died.
all preoperative variables and a second equation for op- An additional patient had a jejunal resection for a bleeding
erative and postoperative variables. arteriovenous malformation and later died. Three patients
required exploration for trauma and two deaths occurred
Results in patients with liver and splenic bleeding. Various pan-
creatic, splenic, and vascular cases accounted for the re-
A total of 100 consecutive patients, 30 women and 70 maining procedures with one death noted. Although op-
men, with hepatic cirrhosis underwent a nonshunt celi- erative indications and procedures showed wide differ-
otomy during the period of review. Thirty patients died ences in mortality, these were not statistically different
during the postoperative period. An additional 30 patients between large groups.
 Ann. SUr8. * June 1984
650 GARRISON AND OTHERS
TABLE 1. Preoperative Variables Differ Between Groups
Survivors
Preoperative Variable Without Complications Complications Nonsurvivors
(n = 40) (n = 30) (n = 30)
Ascites present* 12.5% 33% 70%
Malnutrition present* 7.5% 16.7% 43%
Emergency procedure* 15.0% 40% 80%
Active infection present* 7.5% 16.7% 47%
Bilirubin mg/dl meant 1.2 ± 0.2 2.0 ± 0.5 4.1 ± 1.0
Albumin g/dl meant 3.7 ± 0.6 3.3 ± 0.2 2.5 ± 0.1
WBC 103 meant 8.3 ± 0.6 9.5 ± 1.3 14.5 ± 1.8
PT > Control seconds meant +0.3 ± 0.3 +1.5 ± 0.6 +4.5 ± 1.0
PTT > Control seconds meant -3.0 ± 0.8 -0.8 ± 2.2 +7.9 ± 2.4
Child's class meant 1.25 ± 0.1 1.6 ± 0.1 2.4 ± 0.14
*
p < 0.01 by chi square. t p < 0.01 by analysis of variance.

All 54 preoperative, operative, and postoperative vari- ination can be obtained by the addition of further vari-
ables were subjected to statistical analysis. Preoperative ables, including those not statistically significant between
variables that were found to be different (p < 0.01) be- groups, a coefficient is derived for each variable. The
tween survivors without complications, survivors with coefficients give appropriate weight to the variable in the
complications, and nonsurvivors are listed in Table 1. equation. Utilizing the variables and coefficients found
All preoperative variables except Child's classification to be of value in predicting survival and nonsurvival, the
were then subjected to multivariant stepwise discriminant equation was solved for each of the patients in the study,
analysis. Those variables found to be significant in dis- and group means were computed for the survival group
criminating between survivors and nonsurvivors are listed and the nonsurvival group.
in Table 2 in decreasing order of importance. An equation The equation and variables found to be of value along
was then generated to predict survival using all preop- with their coefficients are illustrated in Table 3. An equa-
erative variables. In formulating this equation, the com- tion was derived utilizing the three preoperative variables
puter program first picks the one variable that is most of albumin (preoperative value), infection/contamination
discriminatory between survivors and nonsurvivors. Each (yes = 1, no = 2), and the difference in seconds between
successive variable for the predictive equation is chosen the actual value and the control value for the partial
in a stepwise fashion. The variable picked adds the most thromboplastin time. Infection/contamination refers to
information in discriminating between the two groups active intra-abdominal infection or contamination from
without overlapping information supplied by the previous perforated viscus. The equation for all three preoperative
variables used in the equation. When no further discrim- variables is: -4.90019 + [(albumin X 0.86204) + (in-
fection X 1.23263) + (PT-PTTc X -0.05189)].
The group mean closest to the value for each patient
TABLE 2. Preoperative Variables Differ Between
Survivors and Nonsurvivors determined the group in which the patient belonged. It
was found that, of these patients with values more negative
Mortality if Mortality if than the group mean for nonsurvivors, 93% died and, of
Variable Variable Present Variable Absent the patients with values more positive than the group
Child's classification
A 10% 50% TABLE 3. Preoperative Variables: Multivariant
B 31% 30% Discriminant Analysis*
C 76% 18%
Ascites 58% 11% Variable Coefficient
Infection/contamination 64% 21%
Emergency procedure 57% 10% Albumin (preoperative value) 0.86204
Poor nutrition 62% 22% Infection/contamination (yes = 1, no = 2) 1.23263
Bilirubin 2 3 mg/dl 62% 17% PT-PTT control (preoperative values) -0.05189
Albumin < 3 mg/dl 58% 12% Constant = -4.90019
PT > Control 47% 7%
PT > 1.5 sec/control 63% 18% * Equation: Patient score = constant + Z (variable coefficient X vari-
PTT > Control 54% 18% able value).
WBC > 10,000 54% 19%
Group Means
p < 0.01 for all variables listed. Survivors 0.62527
Listed in order of decreasing importance in predicting survival. Nonsurvivors -1.33986
Vol. 199 No. 6 CELIOTOMY IN CIRRHOTIC PATIENTS
-

651
S SURVIVORS
= TABLE 5. Operative and Postoperative Variables Differ
N = NONSURVIVORS Between Survivors and Nonsurvivors
8 s sS S Mortality if Mortality if
N 5SS S
N N N S SSSS SSS SSSS SS SS
S Variable Variable Present Variable Absent
NNNN NNS N N NS NN NNNSNNSNS SSSS SS SS S
-3.0 -2.0 -1.0 0 1.0 2.0 Pulmonary failure 100% 0%
Cardiac failure 92% 8%
-1.34 .63 Antibiotics > 2 82% 10%
MEAN MEAN
NONSURVIVOR SURVIVOR Renal failure 73% 9%
GROUP GROUP Liver failure 66% 8%
FIG. 1. Preoperative variables: accuracy ofpredictive equation. All patients Infected ascites 85% 16%
with values greater than the group mean for survivors lived. Ninety- Ascitic leak 82% 18%
three per cent of patients with values less than the group mean for GI bleed 86% 21%
nonsurvivors died. Patients with values between the means were classified Positive sputum culture 58% 18%
correctly 66% of the time. Required secondary
operation 81% 20%
>1 course antibiotics 73% 21%
mean for survivors, 100% lived. This is illustrated in Figure Postoperative bleeding 100% 26%
Positive urine cultures 65% 22%
1. Patients with values between the two means were pre- Positive blood cultures 61% 21%
dicted accurately 66% of the time. Overall, the equation Wound infection 61% 22%
predicted survival with 90% accuracy and nonsurvival Blood > 2 units 69% 22%
Fresh frozen plasma 61% 19%
with 71% accuracy. Accumulative accuracy of the entire
equation in predicting survival or nonsurvival was 83% p < 0.01 for all variables listed.
(Table 4). Listed in order of decreasing importance in predicting survival.
All operative and postoperative variables were then
analyzed in a similar fashion. Those variables found to overall operative mortality of 30% and an additional
be different between survivors and nonsurvivors are listed morbidity rate of 30%. Sepsis with multiple system organ
in Table 5 in order of decreasing significance with respect failure was by far the most common cause of death and
to predicting outcome. Transfusion requirements were seen in 87% of our failures. In an attempt to explain this
the only significant operative variable found to contribute extraordinarily high mortality rate, we analyzed multiple
to this prediction. All operative and postoperative vari- preoperative, operative, and postoperative variables.
ables were then subjected to multivariant stepwise dis- Stepwise multivariant discriminant analysis was utilized
criminant function analysis in a fashion similar to the to determine which variables were most predictive of out-
previous analysis. Pulmonary failure, operative blood come.
transfusion requirements, stress gastrointestinal bleeding, Neither the organ system involved in the surgical pro-
urinary tract infection, and the development of ascites cedure nor the operative procedure performed was help-
in the postoperative period were selected as most pre- ful in predicting survival, although the presence of intra-
dictive of outcome. These variables with their coefficient abdominal infection or contamination from a perforated
weights and the means for each group are listed in Table viscus contributed significantly to a poor outcome. In-
6. Overall, when tested against patients in our study, fection/contamination was associated with a 64% mor-
this equation predicted outcome with a 98% accuracy
(Table 7). TABLE 6. Operative and Postoperative Variables:
Multivariant Discriminant Analysis*
Discussion
Variable Coefficient
Celiotomy in the cirrhotic patient has been reported 5.23786
Pulmonary failure (yes = 1, no = 2)
to carry a high risk mortality.2 3'5'8'9 In reviewing 100 Blood requirements (number of units) -0.14948
patients with histologically proven cirrhosis undergoing UGI bleed
nonshunt celiotomy, we confirmed this finding with an (yes = 1, no = 2) 1.46929
Urinary infection
(yes = 1, no = 2) -0.53948
Ascites development
TABLE 4. Preoperative Variables: Accuracy of Predictive Equation (yes = 1, no = 2) -1.39626
Constant = -8.74144
Correct No. of No. of
Group (%) Survivors Nonsurvivors * Equation: patient score = constant + Z (variable coefficient x vari-
able volume).
Survivors 89 40 5
Nonsurvivors 71 6 15 Group Means
46 20 Survivors 1.62694
Total 83 Nonsurvivors -3.87098
 GARRISON AND OTHERS Ann. Surg. * June 1984
652
TABLE 7. Postoperative Variables: Accuracy ofPredictive Equation different between survivors and nonsurvivors in our pa-
Correct No. of No. of tients.
Group (%) Survivors Nonsurvivors The only operative variable related to mortality was
the transfusion requirement. When operative transfusions
Survivors 100 70 0 of more than 2 units ofblood were required, the mortality
Nonsurvivors 93 2 28
rate of 69% was observed compared to only 22% when
Total 98 72 28 2 units or less were required. Similar findings were seen
with fresh frozen plasma replacements; operative mor-
tality rate compared to 21% when it was not present. tality of 61% occurred when these transfusions were
Operative urgency was similarly related to mortality. needed compared to only 19% when none was required.
Death occurred in 57% of the patients who underwent These findings were expected and were simply an exten-
surgical treatment on an urgent basis compared to only sion of the preoperative coagulation parameters.
10% mortality on an elective basis. These findings are We think it is interesting that drain use was not sta-
consistent with those of Dobemeck et al.5 Their study tistically different between survivors and nonsurvivors.
found a mortality rate of 45% for emergency procedures, In patients with ascites present, the usage of drains com-
while only 11% of elective operations had a similar out- pared to their nonuse led to mortality rates of 53% and
come in nonbleeding cirrhotic patients. Child's classifi- 65%, respectively. However, drains did play an important
cation was related to the operative mortality. Mortality role leading to mortality in patients who developed an
rate in 50 Class A patients was 10 per cent, increased to ascitic leak. In 10 of 26 patients dying from sepsis, the
31% in 29 Class B patients, and was found to be 76% in source of the infection was judged to be secondary to
21 Class C patients. Furthermore, each ofthe components infection ofascites following the development of an ascitic
of the Child's classification was found to be statistically leak. In seven of these ten patients the leak developed
different between these groups with the exception of en- from a drain site, and in six patients the organisms cul-
cephalopathy, which in our series was poorly documented tured, Staphylococcus and Streptococcus, suggested con-
in the hospital records reviewed. Stone'5 has reported that tamination from outside the peritoneal cavity.
although the Child's criteria were initially written to pre- Mortality rates for various operative procedures were
dict the operative risk of patients undergoing portasys- high and not statistically different between larger groups.
temic shunt procedures, they are equally useful in pre- Procedures performed on the biliary tract had a 21% mor-
dicting the outcome of any major surgical procedure. tality rate in our series, which compares favorably to
Furthermore, elevated bilirubin, hypoalbuminemia, and several recent studies. Schwartz9 reported a mortality rate
the presence of ascites have been found to be different of 27%. Aranha2 confirmed this finding with a rate of
between survivors and nonsurvivors in studies by other 26% of cirrhotic patients undergoing cholecystectomy.
authors' 6 reporting on nonshunt celiotomy in cirrhotics. Patients undergoing operation for the complications of
Coagulation parameters had a similar bearing on sur- peptic ulcer disease were found to have an operative mor-
vival in our study. When the prothrombin time was greater tality rate of 35%. This percentage again compares fa-
than that of control, 47% of all patients died compared vorably with the 57% mortality rate reported by Wirthlin"
to only seven per cent when the prothrombin time was in cirrhotic patients undergoing procedures for nonvar-
less than or equal to the control value. Ifthe prothrombin iceal upper gastrointestinal bleeding. The rate from our
time of the patient was greater than 1.5 seconds over study also correlates with the 23% mortality rate reported
control time, the mortality rate increased to 63%. These by Doberneck5 for nonbleeding cirrhotic patients under-
findings are consistent with reports by other authors2'9 going resection or repair of the gastrointestinal tract. Fifty-
who found that prolonged prothrombin time was related three per cent of the patients undergoing a colon resection
to mortality in a significant way. Aranha2 found pro- in our series died. To our knowledge, there are no other
thrombin time to be the variable most predictive of sur- reports of colectomy in this patient population. The re-
vival. mainder of our patients underwent a variety of abdominal
Our analysis listed white blood cell counts of greater surgical procedures. Although the number of patients un-
than 10,000 as associated with a high mortality rate of dergoing each individual procedure was small, the overall
54% compared to only 19% when this count was less than mortality rate in this group was 31%, which is consistent
10,000. This finding is consistent with the fact that intra- with the overall mortality percentage seen in our study.
abdominal infection or contamination is the leading con- Morbidity was also high in our study. Eighty-five post-
tributor to mortality in our patients. All other preoperative operative complications occurred in 30 surviving patients.
factors that we reviewed, including the number of pre- The majority of these complications were related to in-
existing diseased organ systems, were not significantly fection. By far, the most common cause of death in our
Vol. 199 - No. 6 CELIOTOMY IN CIRRHOTIC PATIENTS 653
series was sepsis with multiple organ system failure ac- (p < 0.01) between survivors and nonsurvivors in our
counting for 87% of the deaths. This finding is consistent study. These included Child's classification, elevated
with separate reports by Schwartz9 and Aranha,2 where serum bilirubin, hypoalbuminemia, the presence of as-
sepsis accounted for 33% of mortalities in each of those cites, malnutrition, active infection/contamination, an
studies. The subsequent development of multiple organ elevation of the white blood cell count, prolongation of
system failure was linearly related to mortality. The mor- the prothrombin time, prolongation of the partial throm-
tality rate was only two per cent if no organ systems boplastin time, and the need for an emergency procedure.
failed, nine per cent if one organ system failed, 50% if However, when all preoperative variables were subjected
two organ systems failed, 75% if three organ systems failed, to stepwise multivariant discriminant functional analysis,
and 100% if four or more organ systems failed. Further- only three variables were utilized to predict the outcome
more, ifpulmonary failure complicated the postoperative of patients in our series. We excluded Child's classification
course, a mortality rate of 100% was noted. as a predictor of survival as it represents a multivariant
Several investigators have developed formulas for pre- formula in itself.
dicting operative risk in cirrhotic patients undergoing The final equation, thus generated by our analysis,
procedures other than portasystemic shunts. In 1955, utilized only three preoperative variables in predicting
Cayer3 found that the mortality rate in cirrhotic patients the outcome. These variables were the absolute serum
with ascites undergoing a variety of operations was 50% albumin concentration, the presence of infection/con-
and that ascites, hypoalbuminemia, prolonged prothrom- tamination, and the number of seconds partial throm-
bin times, and anemia were different between survivors boplastin time deviated from its control value. The equa-
and nonsurvivors. Wirthlin" found that the preoperative tion predicted survival with 89% accuracy and predicted
serum bilirubin level was the most reliable indicator for nonsurvival approximately 71% of the time. An overall
poor surgical risk in cirrhotic patients undergoing pro- accuracy rate of 83% was noted. A series of variables can
cedures for nonvariceal upper gastrointestinal bleeding. be selected utilizing this technique, each ofwhich supplies
The presence of varices with past history of ascites, ele- information to the predictive equation but does not over-
vated serum ammonia levels, prolonged prothrombin lap with the information supplied by the previous variables
time, and hypoalbuminemia were all significant between in the equation. Similar techniques have been utilized by
survivors and nonsurvivors. He developed a five-point Irvin and Zeppa'7 for peptic ulcer disease and by Simert'0
scoring system to assess surgical risk. Stone'5 found that and Cello4 for portasystemic shunt procedures. This
the Child's classification criteria were useful in predicting equation is expected to be accurate since it was derived
the outcome of any major procedure in the cirrhotic pa- from the patients upon whom it was tested. We are cur-
tient. Pugh"6 utilized a modification of the Child's clas- rently undertaking a prospective analysis of its predictive
sification by substituting prothrombin time for nutrition. value as well as a retrospective study upon patients cared
He developed a point system that predicted operative for in other university-affiliated hospitals.
mortality in a series of patients subjected to transection In a similar manner, the operative and postoperative
of the esophagus for bleeding esophageal varices. Dob- variables were subjected to the stepwise discriminant
erneck5 reported 102 nonbleeding cirrhotic patients un- function analysis. The variables chosen by the computer
dergoing an operative procedure. He found that an in- in the development of the equation were pulmonary fail-
creased mortality rate was associated with the following: ure, operative blood transfusion requirements, gastroin-
a serum albumin level greater than 3.5 mg/dl, an elevation testinal bleeding in the postoperative period, the devel-
of alkaline phosphatase values, prothrombin time greater opment of a positive urine culture, and finally, the de-
than 2 seconds/control, an emergency procedure, an al- velopment of ascites. When applied to the patients in our
imentary tract operation, the presence ofascites, operative study, this equation predicted survival with 100% accuracy
blood loss of more than 1000 cc, and postoperative com- and nonsurvival with 93% accuracy. This predictive
plications. Dobemeck also found that the number of sig- equation using postoperative data is not useful as a tool
nificant risk factors present was predictive of operative in itself. However, it does demonstrate a combination of
mortality ranging from five per cent for zero to one factor factors that lead to mortality in our patients. The equa-
up to 67% when more than six factors were present. On tions suggest that, if these factors can be prevented or
the other hand, Jackson' concluded that the prognosis altered, the chance for survival for the patient may be
for survival in cirrhotics could not presently be predicted improved.
in clinical, laboratory, or physiologic studies alone. He
found that the ultimate outcome in the individual patient Conclusions
is not easily determined prior to operation. Operative risk for procedures other than portasystemic
We found ten preoperative variables that were different shunts in cirrhotic patients is high. In our series, an overall
654 GARRISON AND OTHERS Ann. Sur * June 1984
mortality rate of 30% was noted with an additional 30% survival after therapeutic shunts. Results of a discriminant func-
morbidity rate in surviving patients. This unacceptable tion and linear logistic regressions analysis. Am J Surg 1981;
141:257-265.
rate of death is consistent for all groups of operative pro- 5. Doberneck RC, Sterling WA, Allison DC. Morbidity and mortality
cedures. after operation in nonbleeding cirrhotic patients. Am J Surg
1983; 146:306-309.
Analysis of operative and postoperative factors leading 6. Henrikson EC. Cirrhosis of the liver with special reference to the
to a mortality identifies multiple problems that must be surgical aspects. Arch Surg 1936; 32:413-451.
solved to decrease this operative mortality rate. These 7. Hughson W. Portal cirrhosis with ascites and its surgical treatment.
Arch Surg 1927; 15:418-440.
obstacles include normalization of coagulation parame- 8. Lindenmuth WW, Eisenberg MM. The surgical risk in cirrhosis of
ters, improvement in nutritional status, and control of the liver. Arch Surg 1963; 86:77-84.
preoperative ascites. 9. Schwartz SI. Biliary tract surgery and cirrhosis: a critical combination.
Preoperative assessment can predict survival with ap- Surgery 1981; 90:577-583.
10. Simert G, Persson T, Vang J. Factors predicting survival after por-
proximately 90% accuracy in cirrhotic patients requiring tacaval shunt: a multiple linear regression analysis. Ann Surg
abdominal procedures. Multivariant stepwise discrimi- 1978; 187:174-178.
11. Wirthlin LS, Urk HV, Malt RB, et al. Predictors of surgical mortality
nant analysis may be a useful tool in developing equations in patients with cirrhosis and nonvariceal gastroduodenal bleed-
to predict operative risk from variables assessed before ing. Surg Gynecol Obstet 1974; 139:65-68.
surgery for a variety of high risk procedures. 12. Child CG, Turcotte JF. The Liver and Portal Hypertension. Phil-
adelphia: WB Saunders, 1964.
13. Linton RR. The selection of patients for portacaval shunts. Ann
Surg 1951; 134:433-443.
References 14. McDermott W. The double portacaval shunt in the treatment of
chronic ascites. Surg Gynecol Obstet 1960; 110:457-469.
1. Jackson FC, Christophersen EB, Petemel WW, et al. Preoperative 15. Stone HH. Preoperative and postoperative care. Surg Clin North
management of patients with liver disease. Surg Clin North Am Am 1977; 57:409-419.
1968; 48:907-930. 16. Pugh RNH, Murray-Lyon IM, Dawson JL, et al. Transection of
2. Aranha GB, Sontag SJ, Greenlee HB. Cholecystectomy in cirrhotic the oesophagus for bleeding oesophageal varices. Br J Surg 1973;
patients: a formidable operation. Am J Surg 1982; 143:55-60. 60:646-649.
3. Cayer D, Sohmer MF. Surgery in patients with cirrhosis. Arch Surg 17. Irvin GL, Zeppa R. Predicted survival in peptic ulcer patients based
1955; 71:828-838. on computer analysis of preoperative variables. Ann Surg 1976;
4. Cello JP, Deveney KE, Trunkey DD, et al. Factors influencing 183:594-598.

DISCUSSION DR. WARD 0. GRIFFEN, JR. (Lexington, Kentucky): It is clear that

the authors have done a magnificent job of identifying preoperative,
DR. ROBERT ZEPPA (Miami, Florida): We are indebted to Dr. Polk operative, and postoperative risks in these patients and showing the
and his colleagues for an elegant study that has sought to quantify those ones that are important.
variables that predict survival or the lack thereof in a group of high- The reason I asked to discuss the paper was on account of a patient
risk patients. The importance of this contribution can not be overes- that one of our senior residents, Dan Procter, and I took care of last
timated. It is only when, as the authors have done, you obtain quantitative year, doing a straightforward left hemicolectomy for cancer in a patient
information concerning the predictive value of a finite and manageable with cirrhosis, who died 42 days later of sepsis. I asked him to look up
group of patients-can you then do prospective studies of logical pre- our cases of colectomy in patients who were cirrhotic.
operative and postoperative therapeutic interventions that perhaps will In a 5-year period of time, he found six such patients, and three of
lower the horrendous mortality that has been demonstrated here. the six patients died, all of them of sepsis. They had the same features
The analysis of the postoperative mortality clearly demonstrates that that Dr. Polk has indicated already. The three nonsurvivors all had
this high-risk cohort of patients will not tolerate a complication. Further, ascites at the time of the operation; one had contamination from a
if these are anticipated, or can be anticipated, as demonstrated by Dr. perforation of her diverticular disease. They all had intra-abdominal
Polk's data, then perhaps a therapeutic game plan can be directed toward anastomoses made.
the prevention of these sorts of complications and/or earlier intervention I wonder, and I ask the authors: Would it be possible to reduce the
when the clues that have been accumulated by the data collection suggest risk, at least in the patients undergoing colon resection who are cirrhotics,
that they are about to occur. by avoiding an intraperitoneal anastomosis?
The data presented suggest that, at least in elective patients, as much I do not know the answer to that, but I wonder if that is possibly an
preoperative physiological data should be obtained, and the deficits added risk. In other words, if you do colostomy and forget about trying
demonstrated as a function of the collection of these data corrected, to put the bowel back together, it might provide you with more survivors.
before surgery.
It is most intriguing to me to compare these data with those we have DR. WILLIAM J. MILLIKAN (Atlanta, Georgia): The Emory surgical
observed in our nonalcoholic cirrhotic population who have been sub- liver group appreciates the opportunity to review this manuscript, and
jected to shunts. In our analysis of 91 patients with nonalcoholic cirrhosis we feel that this will come to be recognized as a major contribution to
who were operated on between July 1971 and July 1981, 22% of the the field of liver disease.
group were on steroid therapy for chronic, aggressive hepatitis; this, of Second, we appreciated the opportunity to review the manuscript
course, is associated with hyperbilirubinemia and disturbances in the with our biometrist back in Atlanta, because this paper has so much
hepatic circulating enzymes. Yet the operative mortality for this group data that unless you have the opportunity to review the manuscript, a
was only 4%, and these are major operations. lot of it gets lost in the slides.
Two questions: Were your cirrhotic patients all alcoholic in this par- We would like to make two points that we relearned from the paper,
ticular series? And in the utilization of the Child categorization, how and ask two questions.
many of the variables needed to be present before you moved the patient (Slide) In a paper that was presented here in this room 2 years ago,
from Group A to Group B to Group C? Dr. Warren presented the Emory 10-year experience with patients op-