- Email: [email protected]
Forecast of surgical risk in pancreatic cancer
Forecast of Surgical Risk in Pancreatic Cancer
Sergio Pedrazzoli, MD, Btwno Bcnadimani, MD, Cosimo Sperti, MD, Francesco Cappellauo, MD, Antonio Plccol~,MD, BS, and Carmello Militello, MD, Padua, Italy
In the surgical treatment of pancreatic cancer, pancreatic resections for cure and operations to bypass intestinal, blhary, or pancreatic obstructions can be performed m relation to tumor stage The poor postoperative quality and duration of life require careful evaluation of the surgical risk, since the incidence of mortality and morbidity after both operations IS high P21 Numerous studies on pancreatic cancer describe many parameters whose alteration 1s statlstlcally related to increased operative mortality and morbidity [1,3,4] However, the interpretation of these predictive parameters 1s difficult when related to the evaluation of the operative risk of a single patient To our knowledge, no reports exist m the literature on a method to evaluate the surgical risk for a single patient Such a method could lead to a rational approach toward multldlsclphnary treatment of patients with pancreatic cancer The aim of this study was to establish the relative importance of some parameters m predlctmg surglcal morbidity and mortality and to fmd a chmcally applicable algorithm which would allow the preoperative surgical risk of a single patient to be evaluated. Patients and Methods Three hundred twenty-one patients with pancreatic cancer were observed m our department between 1962 and 1983, and 305 underwent surgical treatment Those who underwent laparotomy only (71 patients) were excluded from the study because m these cases the surgical risk was mainly due to the illness In the remaining 234 patients the followmg preoperative parameters were considered age, weight loss (percentage of usual weight), total serum blhrubm level, duration of Jaundice, and total blood protein level A further 60 patients were excluded because at least one of these parameters were missing Of the remaining 174 patients, 126 underwent bypass operation (Group I) and 48, potentially curative resection From the Frrst Department of Surgery and the Department of Internal Medrcrne, Unwersrty of Padua, Padua, Italy Requests for reprrnts should be addressed to Sergro Pedrazzolt, Department of Surgery Clsnca Chrrurgrca I Unwerslty of Padua, Via Grusbnianr 2 35100 Padua Italy Supported rn part by a grant from the Italian Nabonal Research Councrl Special Project Oncology, ’ contract no 84 00725 44 Presented at the 14th European Federation Congress and 1985 Jubrlee of the International College of Surgeons, Geneva, Swrtzerland May 26-30, 1985
374
(Group II) (Table I) The median age of the Group I patients was 60 years and of the Group II patients, 59 years (ranges 29 to 85 years and 37 to 75 years, respectlvely) Group I had 88 men and 38 women and Group II, 28 men and 20 women Pancreatic cancer was staged according to Hermreck’s parameters (Table I) The operative mortality of the 174 patients eligible for our study was similar to that of the 234 patients who underwent bypass operation or pancreatic resections (19 5 percent versus 15 4 percent, chlsquare 2 10, p >O 05) In Group I, all patients underwent a blhary bypass that was associated with a gastroJeJunostomy m 36 patients In Group II, a pancreatoduodenectomy was performed in 34 patients, a total pancreatectomy in 7 patients (a mesenterlc resection was required m 5), and a distal pancreatectomy m 7 patients Patients m both groups were classlfled mto the followmg subgroups according to their postoperative course D, patients who died during the first postoperative month, C, those with local or general postoperative comphcatlons, and U, and those with an uneventful course (Table II) Dlscrlmmant analysis was performed on data from Groups I and II The variables evaluated (age, bdlrubmemla, duration of Jaundice, weight loss, and total protememla) were transformed into a three-score ordinal scale (Table III) Stepwlse dlscrlmmant analysis according to Rao’s criteria was performed using the dlscrlmmant program of the SPSS statlstlcal package [5] The stepwlse method allowed selection of the best set of dlscrlmmatmg variables within the five variables entered for analysis Thus, it was possible to identify which variables contnbuted to the three considered outcomes The stepwlse procedure began by selecting the single best-dlscnmmatmg variable, a second variable was chosen as the variable best able to improve the dlscrlmmatmg value m combmatlon with the first variable The third and followmg variables were selected m the same way If the remammg variables were no longer able to give further dlscnmmatlon, they were not taken into account Results
Group I: The variables were selected m the following order according to their dlscrlmmant value weight loss, duration of Jaundice, total blood protein level, and patient age The dlscrlmmant value of preoperative blhrubm plasma levels was_ not _ consldered significant
Usmg these four variables, a mean
The American
Journal of Surgery
Surgical Risk In Pancreatic Cancer
TABLE I
Characteristics of Patient Treatment Grow
1962-66
1967-71
I Bypass Stage I-11 Stage Ill Stage IV II Resectron Stage I-II Stage Ill Stage IV
36 (17) 0 5 31 2 (1) 0 2 0
12 (13) 0 6 6 8 (2) 7 1 0
Total
38 (18)
20 (15)
Number of Pabents’ 1972-76
1977-63
Total
25 (3) 0 8 17 15 (2) 9 6 0
53 (21) 0 12 41 23 (1) 15 7 1
126 (54) 0 31 95 48 (6) 31 16 1
40 (5)
76 (22)
176 (60)
Values In parentheses denote patrents who underwent bypass or resectron dunng the same period and were excluded from the study because of lack of one or more parameters l
TABLE II
TABLE III
Actual Outcome of the Patlents
Group
I Bypass
Actual Outcome U C
D
Total
II Resection
26 8
30 17
70 23
126 48
Total
34
47
93
174
C = compkcabons (local or general), D = dead, U = uneventful postoperative course
percentage of 88 percent of the patients who underwent a bypass operation could be correctly classified mto the three risk groups (Table IV) Group II: The variables were selected m the following sequence according to their dlscrlmmant value protememra, brhrubmemla, weight loss, and patient age The drscrlmmant value of the duration of Jaundice was not considered slgmficant A mean percentage of 83 percent of the patients who underwent potentially curative resection could be correctly classified mto the three risk groups (Table V) Similar percentages of correct classification into the three risk groups were observed when our patients were divided mto four groups accordmg to the year of operation (chl-square 0 35, p >O 05) (Figure 1) Operative mortality rates were 10 and 217 percent respectively m 20 patients resected by the same surgeon (SP) and m 28 patients treated by 7 dlfferent surgeons (chl-square 3 66, p >O 05) (Figure 2) The percentages of correct classification were 80 and 86 percent, respectively Comments A high serum blhrubm level, elevated prothrombm time, diabetes, age, obesity, pain, past heart trouble, and alcoholism were considered risk factors m bypass operations or pancreatic resections for pancreatic cancer [1,3,4,6,4 Nevertheless, it is not easy to foresee the chances of survival of an mdlvldual patient on the basis of these factors
Volume 153, April 1997
Scoring System of the Considered Parameters
Parameters
1
Score 2
3
Age (yr) Serum brlrrubrn (mg/dl) Duration of jaundice (d) Weight loss (% usual wt) Blood protein (g/100 ml)
<50 6
50-80 10-20 15-30 <20 5-6
>80 >20 >30 >20 <5
TABLE IV
Comparlsonof Actual Versus PredIcted Outcome In Patlents Who Underwent Bypass (Risk predlctlon based on dlscrlmlnant analysis)
Actual Rusk Group Patients Outcome (n) 26 30 70
U
Predicted Risk Group’ D C U (n = 22) (n = 29) (n = 75) 20 (76 9) 2 (6 7) 0 (0)
6 (23 1) 22 (73 3) 1 (1 4)
0 (0) 6 (20) 69 (98 6)
C = complication (local or general), D = dead, U = uneventful postoperative course Values in parentheses are percentages l
TABLE V
Comparisonof Actual Versur Predicted Outcome of Patients Who Underwent Resectlve Surgery (risk predlctlon based on dlscrlmlnant analysis)
Actual Rrsk Group Patients (n) D C U
8 17 23
Predicted Risk Group* D (n = 9)
(n 217)
(n i22)
7 (87 5) 2(11 8) 0 (0)
1(12 5) 13 (76 4) 3 (13)
0 (0) 2(11 8) 20 (87)
C = compkcation (local or general), D = dead, U = uneventful postoperative course * Values in parentheses are percentages
375
Pedrazzoh et al
39 cases
20 cases
100%
40 cases
75 ases
20
100%
::. ++ ++
I+’ ++ ++: :++ +
+I+ ++’
++’ ++’ ++’
1+++1 patlents
:+: +++ +: I+’
1967 1971 with
UmI
77 *** *** ***
50%* *
L
f ++++ +++ +++
1972 1976
correctly
\ \
predtcted
> \\
8
\
\
\
predlcted
Cl
actual
C
m
patlents
predicted
C
actual
D
I\‘\: ~~~ u
\y \
’
\ i \\
\
\’
:::::: :::::: :::::: ::::::
\
L :1::::: :I::::: :? s 1nglc surgeon
outcome
patfents
\
\
\
1977 1983
K \
A
c
0
m
patients
predicted
C
actual
U
m
patients
predlcted
U
actual
C
Figure 1 Actual and predicted outcome by time D = dead, C = complications (local or general), 0 = uneventful postoperative course
Buzby et al [S] developed a clmlcally applicable method to determine the magnitude of risk of a nutrltlonally dependent comphcatlon based on the profile of the nutritional status of the surgical patient Some patients affected by pancreatic cancer were included m then study Although their linear predictive model had classlfled as high or mtermedlate risk 89 percent of patients who ultimately had comphcatlons and 93 percent of the patients who later died, 54 percent of patients classlfled as high risk and 70 percent of those classlfled as mtermedlate risk had no comphcatlons Therefore, this model may be useful m estabhshmg whether a patient should receive preoperative nutrltlonal therapy On the other hand, it 1s not reliable m mdlcatmg the operative risk for a single patient and thus the choice between surgical or radlologlc management when both methods are available [9] In our retrospective study, using easily detectable parameters, operative deaths could be separated from postoperative comphcatlons and patients with a normal postoperative course The patients were assigned to these three groups with a total accuracy rate of 88 percent for bypass operation and 83 percent for pancreatic resections The patients who
376
‘::
3
I+’ ++ +*+
L 1966
1
++ +++ :+: ++ :++ +:+ :t+ ++
‘+I :++ +:+ +*+ +++ +++ +++ +*+ +++ +++ +++ f+’ ++-t +++ +++
1962
l.*:
i::
++’ ++’
50%
20
cases
cases
I:::::
I..... I.....
sz ieven 5
rgcons
m
patients
predlcted
D
actual
D
bmq
patients
predicted
C
actual
C
1-1
patients
predlcted
U
actual
U
0
patients
predicted
D
actual
C
actual
D
m
patients
predlcted
C
m
patients
predlcted
C
actual
U
m
patients
predicted
U
actual
C
Figure 2 Actual and predicted outcome In pat/en& operated on by a single surgeon or by seven different surgeons D = dead, C = complications (local or general), 0 = uneventful postoperative course
died after bypass operation were correctly classlfled m 77 percent of cases and were erroneously included m subgroup C m 23 percent of cases, whereas none of them were included m subgroup U Only 2 of the 100 patients who survived were found m subgroup D (Table IV) As for resectlve surgery, 7 of 8 patients (88 percent) who died were included m subgroup D, 1 (12 percent) m subgroup C by error, and none m subgroup U Only 2 of 40 patients who survived (5 percent) were placed m subgroup D The high dlscrlmmant power of our parameters was not affected by changes m medical or perloperatlve management over the years This further emphasizes the predlctlve value of our parameters even if the results of a prospective study are not yet available For exam-
The American Journal of Surgery
Surgical Risk in Pancreatic Cancer
ple, m 1982, Herter et al [7] found an essentially unchanged incidence of comphcatlons before and after 1970 In our survey, total mortality (16 7 percent) and morbidity (52 percent) rates after pancreatic resections as well as after blhary bypass with or without a gastroJeJunoanastomoas were similar to those m the literature [1,2,7,9,10,11] (Tables IV and V) A single surgeon (SP) performed 20 pancreatic resections with a 10 percent operative mortality Both patients who died and 1 of the 18 patients who survived (5 percent) were m subgroup D The remammg 28 pancreatic resections were performed by 7 different surgeons with a 217 percent mortality rate Of these, 5 patients who died were m subgroup D as well as, mistakenly, 1 of the 22 who survived In addition, one patient who died was classified m subgroup C (Figure 2) It should be stressed that all the surgeons who performed a pancreatic resection were experienced general surgeons and that the patients were assigned at random All the parameters considered, except weight loss, had been previously regarded as associated with a greater surgical risk of pancreatic cancer Nevertheless, they had never been analyzed m a multlvarlate fashion to evaluate their predictive power when dlfferently combined Although the patients described by Buzby et al [8] who had comphcatlons lost weight more rapidly, the mcluslon of this factor did not improve the predictive accuracy of the prognostic nutritional index The prognostic value already ascribed to weight loss by clmlcal experience was confirmed by dlscrlmmant analysis Weight loss was first among the predictive factors for bypass surgery, followed by duration of Jaundice, protememla, and age Total blhrubmemla and duration of Jaundice were excluded among the predictive parameters for bypass operations and pancreatic resection, respectively Jaundice 1s represented sufficiently by either of these parameters, even if different m the two groups Finally, the use of dlscrlmmant analysis allowed allocation of mdlvldual patients mto the classes of surgical risk [5] As this 1s a retrospective study, our results should be verified by a prospective study with the aim of estabhshmg the rehablhty of a correct classlflcatlon of patients other than those available for this analy91s
Summary Over a 22 year period, 305 patients underwent operations for pancreatic cancer Seventy-one patients who only underwent a laparotomy and 60
Volume 153, April 1987
patients with defective data were excluded from our study The remaining 174 patients were analyzed to fmd a clmlcally applicable algorithm to allow the preoperative surgical risk to be evaluated m a single patient One hundred twenty-six had undergone a bypass operation (Group I) and 48, a potentially curative resection (Group II) The two groups were divided mto the followmg three risk groups accordmg to the postoperative course D, patients who died during the first postoperative month, C, patients with postoperative comphcatlons, and U, patients with an uneventful course Preoperative weight loss, duration of Jaundice, serum blhrubm level, total protein level, and age of every patient were collected and transformed mto a three-score ordinal scale Stepwlse dlscrlmmant analysis of these data enabled us to correctly classify 88 percent of the patients who underwent palliative operations and 83 percent of those who had pancreatic resection This method 1s useful m identifying the risk group for a single patient using mdlvldual preoperative variables It may, moreover, suggest the most suitable treatment for each patient with pancreatic cancer
References 1 Braasch JW, Gray BN Consrderatrons that lower pancreatoduodenectomy mortakty Am J Surg 1977, 133 480-4 2 Morrow M, Hrlarrs B, Brennan MF Comparison of conventronal surgical resection, radroactrve rmplantatlon, and bypass procedures for exocrme carcrnoma of the pancreas, 19751980 Ann Surg 1984, 199 l-5 3 Chapurs Y, Catala M, Place S, Bonnette P Reflexrons sur le trartment des cancers du pancreas Chrrurgre 1983, 109 357-8 4 Andren-Sandberg A, lhse I Factors rnfluencmg survival after total pancreatectomy in patients with pancreatic cancer Ann Surg 1983, 198 605-10 5 Nre NH, Hull CH, Jenkins JG, Sternbrenner K, Bent DH SPSS statrsbcal package for the socral sciences 2nd ed New York McGraw-Hill. 1975 6 Grlsdorf RB, Spanos P Factors mfluencrng morbrdtty and mortality In pancreatrcoduodenectomy Ann Surg 1973, 177 332-7 7 Herter FP, Cooperaman AM, Ahlborn TN, Antrnorr C Surgical expertence with pancreatic and perrampullary cancer Ann Surg 1982, 195 274-81 8 Buzby GP, Mullen JL, Matthews DC, Hobbs CL, Rosato EF Prognostic nutrrbonal index In gastrointestinal surgery Am J Surg 1980, 139 160-7 9 Sarr MG, Cameron JL Surgical management of unresectable carcmoma of the pancreas Surgery 1982, 91 123-33 10 Nakase A, Matsumoto Y, Uchrda K, Honjo I Surgical treatment of cancer of the pancreas and the perrampullary region cumulative results in 57 rnstrtubons rn Japan Ann Surg 1977, 185 52-7 11 Moossa AR, Lewis J, Mackre CR Surgical treatment of pancreatic cancer Mayo Clan Proc 1979, 54 468-74
377
Sergio Pedrazzoli, MD, Btwno Bcnadimani, MD, Cosimo Sperti, MD, Francesco Cappellauo, MD, Antonio Plccol~,MD, BS, and Carmello Militello, MD, Padua, Italy
In the surgical treatment of pancreatic cancer, pancreatic resections for cure and operations to bypass intestinal, blhary, or pancreatic obstructions can be performed m relation to tumor stage The poor postoperative quality and duration of life require careful evaluation of the surgical risk, since the incidence of mortality and morbidity after both operations IS high P21 Numerous studies on pancreatic cancer describe many parameters whose alteration 1s statlstlcally related to increased operative mortality and morbidity [1,3,4] However, the interpretation of these predictive parameters 1s difficult when related to the evaluation of the operative risk of a single patient To our knowledge, no reports exist m the literature on a method to evaluate the surgical risk for a single patient Such a method could lead to a rational approach toward multldlsclphnary treatment of patients with pancreatic cancer The aim of this study was to establish the relative importance of some parameters m predlctmg surglcal morbidity and mortality and to fmd a chmcally applicable algorithm which would allow the preoperative surgical risk of a single patient to be evaluated. Patients and Methods Three hundred twenty-one patients with pancreatic cancer were observed m our department between 1962 and 1983, and 305 underwent surgical treatment Those who underwent laparotomy only (71 patients) were excluded from the study because m these cases the surgical risk was mainly due to the illness In the remaining 234 patients the followmg preoperative parameters were considered age, weight loss (percentage of usual weight), total serum blhrubm level, duration of Jaundice, and total blood protein level A further 60 patients were excluded because at least one of these parameters were missing Of the remaining 174 patients, 126 underwent bypass operation (Group I) and 48, potentially curative resection From the Frrst Department of Surgery and the Department of Internal Medrcrne, Unwersrty of Padua, Padua, Italy Requests for reprrnts should be addressed to Sergro Pedrazzolt, Department of Surgery Clsnca Chrrurgrca I Unwerslty of Padua, Via Grusbnianr 2 35100 Padua Italy Supported rn part by a grant from the Italian Nabonal Research Councrl Special Project Oncology, ’ contract no 84 00725 44 Presented at the 14th European Federation Congress and 1985 Jubrlee of the International College of Surgeons, Geneva, Swrtzerland May 26-30, 1985
374
(Group II) (Table I) The median age of the Group I patients was 60 years and of the Group II patients, 59 years (ranges 29 to 85 years and 37 to 75 years, respectlvely) Group I had 88 men and 38 women and Group II, 28 men and 20 women Pancreatic cancer was staged according to Hermreck’s parameters (Table I) The operative mortality of the 174 patients eligible for our study was similar to that of the 234 patients who underwent bypass operation or pancreatic resections (19 5 percent versus 15 4 percent, chlsquare 2 10, p >O 05) In Group I, all patients underwent a blhary bypass that was associated with a gastroJeJunostomy m 36 patients In Group II, a pancreatoduodenectomy was performed in 34 patients, a total pancreatectomy in 7 patients (a mesenterlc resection was required m 5), and a distal pancreatectomy m 7 patients Patients m both groups were classlfled mto the followmg subgroups according to their postoperative course D, patients who died during the first postoperative month, C, those with local or general postoperative comphcatlons, and U, and those with an uneventful course (Table II) Dlscrlmmant analysis was performed on data from Groups I and II The variables evaluated (age, bdlrubmemla, duration of Jaundice, weight loss, and total protememla) were transformed into a three-score ordinal scale (Table III) Stepwlse dlscrlmmant analysis according to Rao’s criteria was performed using the dlscrlmmant program of the SPSS statlstlcal package [5] The stepwlse method allowed selection of the best set of dlscrlmmatmg variables within the five variables entered for analysis Thus, it was possible to identify which variables contnbuted to the three considered outcomes The stepwlse procedure began by selecting the single best-dlscnmmatmg variable, a second variable was chosen as the variable best able to improve the dlscrlmmatmg value m combmatlon with the first variable The third and followmg variables were selected m the same way If the remammg variables were no longer able to give further dlscnmmatlon, they were not taken into account Results
Group I: The variables were selected m the following order according to their dlscrlmmant value weight loss, duration of Jaundice, total blood protein level, and patient age The dlscrlmmant value of preoperative blhrubm plasma levels was_ not _ consldered significant
Usmg these four variables, a mean
The American
Journal of Surgery
Surgical Risk In Pancreatic Cancer
TABLE I
Characteristics of Patient Treatment Grow
1962-66
1967-71
I Bypass Stage I-11 Stage Ill Stage IV II Resectron Stage I-II Stage Ill Stage IV
36 (17) 0 5 31 2 (1) 0 2 0
12 (13) 0 6 6 8 (2) 7 1 0
Total
38 (18)
20 (15)
Number of Pabents’ 1972-76
1977-63
Total
25 (3) 0 8 17 15 (2) 9 6 0
53 (21) 0 12 41 23 (1) 15 7 1
126 (54) 0 31 95 48 (6) 31 16 1
40 (5)
76 (22)
176 (60)
Values In parentheses denote patrents who underwent bypass or resectron dunng the same period and were excluded from the study because of lack of one or more parameters l
TABLE II
TABLE III
Actual Outcome of the Patlents
Group
I Bypass
Actual Outcome U C
D
Total
II Resection
26 8
30 17
70 23
126 48
Total
34
47
93
174
C = compkcabons (local or general), D = dead, U = uneventful postoperative course
percentage of 88 percent of the patients who underwent a bypass operation could be correctly classified mto the three risk groups (Table IV) Group II: The variables were selected m the following sequence according to their dlscrlmmant value protememra, brhrubmemla, weight loss, and patient age The drscrlmmant value of the duration of Jaundice was not considered slgmficant A mean percentage of 83 percent of the patients who underwent potentially curative resection could be correctly classified mto the three risk groups (Table V) Similar percentages of correct classification into the three risk groups were observed when our patients were divided mto four groups accordmg to the year of operation (chl-square 0 35, p >O 05) (Figure 1) Operative mortality rates were 10 and 217 percent respectively m 20 patients resected by the same surgeon (SP) and m 28 patients treated by 7 dlfferent surgeons (chl-square 3 66, p >O 05) (Figure 2) The percentages of correct classification were 80 and 86 percent, respectively Comments A high serum blhrubm level, elevated prothrombm time, diabetes, age, obesity, pain, past heart trouble, and alcoholism were considered risk factors m bypass operations or pancreatic resections for pancreatic cancer [1,3,4,6,4 Nevertheless, it is not easy to foresee the chances of survival of an mdlvldual patient on the basis of these factors
Volume 153, April 1997
Scoring System of the Considered Parameters
Parameters
1
Score 2
3
Age (yr) Serum brlrrubrn (mg/dl) Duration of jaundice (d) Weight loss (% usual wt) Blood protein (g/100 ml)
<50
50-80 10-20 15-30 <20 5-6
>80 >20 >30 >20 <5
TABLE IV
Comparlsonof Actual Versus PredIcted Outcome In Patlents Who Underwent Bypass (Risk predlctlon based on dlscrlmlnant analysis)
Actual Rusk Group Patients Outcome (n) 26 30 70
U
Predicted Risk Group’ D C U (n = 22) (n = 29) (n = 75) 20 (76 9) 2 (6 7) 0 (0)
6 (23 1) 22 (73 3) 1 (1 4)
0 (0) 6 (20) 69 (98 6)
C = complication (local or general), D = dead, U = uneventful postoperative course Values in parentheses are percentages l
TABLE V
Comparisonof Actual Versur Predicted Outcome of Patients Who Underwent Resectlve Surgery (risk predlctlon based on dlscrlmlnant analysis)
Actual Rrsk Group Patients (n) D C U
8 17 23
Predicted Risk Group* D (n = 9)
(n 217)
(n i22)
7 (87 5) 2(11 8) 0 (0)
1(12 5) 13 (76 4) 3 (13)
0 (0) 2(11 8) 20 (87)
C = compkcation (local or general), D = dead, U = uneventful postoperative course * Values in parentheses are percentages
375
Pedrazzoh et al
39 cases
20 cases
100%
40 cases
75 ases
20
100%
::. ++ ++
I+’ ++ ++: :++ +
+I+ ++’
++’ ++’ ++’
1+++1 patlents
:+: +++ +: I+’
1967 1971 with
UmI
77 *** *** ***
50%* *
L
f ++++ +++ +++
1972 1976
correctly
\ \
predtcted
> \\
8
\
\
\
predlcted
Cl
actual
C
m
patlents
predicted
C
actual
D
I\‘\: ~~~ u
\y \
’
\ i \\
\
\’
:::::: :::::: :::::: ::::::
\
L :1::::: :I::::: :? s 1nglc surgeon
outcome
patfents
\
\
\
1977 1983
K \
A
c
0
m
patients
predicted
C
actual
U
m
patients
predlcted
U
actual
C
Figure 1 Actual and predicted outcome by time D = dead, C = complications (local or general), 0 = uneventful postoperative course
Buzby et al [S] developed a clmlcally applicable method to determine the magnitude of risk of a nutrltlonally dependent comphcatlon based on the profile of the nutritional status of the surgical patient Some patients affected by pancreatic cancer were included m then study Although their linear predictive model had classlfled as high or mtermedlate risk 89 percent of patients who ultimately had comphcatlons and 93 percent of the patients who later died, 54 percent of patients classlfled as high risk and 70 percent of those classlfled as mtermedlate risk had no comphcatlons Therefore, this model may be useful m estabhshmg whether a patient should receive preoperative nutrltlonal therapy On the other hand, it 1s not reliable m mdlcatmg the operative risk for a single patient and thus the choice between surgical or radlologlc management when both methods are available [9] In our retrospective study, using easily detectable parameters, operative deaths could be separated from postoperative comphcatlons and patients with a normal postoperative course The patients were assigned to these three groups with a total accuracy rate of 88 percent for bypass operation and 83 percent for pancreatic resections The patients who
376
‘::
3
I+’ ++ +*+
L 1966
1
++ +++ :+: ++ :++ +:+ :t+ ++
‘+I :++ +:+ +*+ +++ +++ +++ +*+ +++ +++ +++ f+’ ++-t +++ +++
1962
l.*:
i::
++’ ++’
50%
20
cases
cases
I:::::
I..... I.....
sz ieven 5
rgcons
m
patients
predlcted
D
actual
D
bmq
patients
predicted
C
actual
C
1-1
patients
predlcted
U
actual
U
0
patients
predicted
D
actual
C
actual
D
m
patients
predlcted
C
m
patients
predlcted
C
actual
U
m
patients
predicted
U
actual
C
Figure 2 Actual and predicted outcome In pat/en& operated on by a single surgeon or by seven different surgeons D = dead, C = complications (local or general), 0 = uneventful postoperative course
died after bypass operation were correctly classlfled m 77 percent of cases and were erroneously included m subgroup C m 23 percent of cases, whereas none of them were included m subgroup U Only 2 of the 100 patients who survived were found m subgroup D (Table IV) As for resectlve surgery, 7 of 8 patients (88 percent) who died were included m subgroup D, 1 (12 percent) m subgroup C by error, and none m subgroup U Only 2 of 40 patients who survived (5 percent) were placed m subgroup D The high dlscrlmmant power of our parameters was not affected by changes m medical or perloperatlve management over the years This further emphasizes the predlctlve value of our parameters even if the results of a prospective study are not yet available For exam-
The American Journal of Surgery
Surgical Risk in Pancreatic Cancer
ple, m 1982, Herter et al [7] found an essentially unchanged incidence of comphcatlons before and after 1970 In our survey, total mortality (16 7 percent) and morbidity (52 percent) rates after pancreatic resections as well as after blhary bypass with or without a gastroJeJunoanastomoas were similar to those m the literature [1,2,7,9,10,11] (Tables IV and V) A single surgeon (SP) performed 20 pancreatic resections with a 10 percent operative mortality Both patients who died and 1 of the 18 patients who survived (5 percent) were m subgroup D The remammg 28 pancreatic resections were performed by 7 different surgeons with a 217 percent mortality rate Of these, 5 patients who died were m subgroup D as well as, mistakenly, 1 of the 22 who survived In addition, one patient who died was classified m subgroup C (Figure 2) It should be stressed that all the surgeons who performed a pancreatic resection were experienced general surgeons and that the patients were assigned at random All the parameters considered, except weight loss, had been previously regarded as associated with a greater surgical risk of pancreatic cancer Nevertheless, they had never been analyzed m a multlvarlate fashion to evaluate their predictive power when dlfferently combined Although the patients described by Buzby et al [8] who had comphcatlons lost weight more rapidly, the mcluslon of this factor did not improve the predictive accuracy of the prognostic nutritional index The prognostic value already ascribed to weight loss by clmlcal experience was confirmed by dlscrlmmant analysis Weight loss was first among the predictive factors for bypass surgery, followed by duration of Jaundice, protememla, and age Total blhrubmemla and duration of Jaundice were excluded among the predictive parameters for bypass operations and pancreatic resection, respectively Jaundice 1s represented sufficiently by either of these parameters, even if different m the two groups Finally, the use of dlscrlmmant analysis allowed allocation of mdlvldual patients mto the classes of surgical risk [5] As this 1s a retrospective study, our results should be verified by a prospective study with the aim of estabhshmg the rehablhty of a correct classlflcatlon of patients other than those available for this analy91s
Summary Over a 22 year period, 305 patients underwent operations for pancreatic cancer Seventy-one patients who only underwent a laparotomy and 60
Volume 153, April 1987
patients with defective data were excluded from our study The remaining 174 patients were analyzed to fmd a clmlcally applicable algorithm to allow the preoperative surgical risk to be evaluated m a single patient One hundred twenty-six had undergone a bypass operation (Group I) and 48, a potentially curative resection (Group II) The two groups were divided mto the followmg three risk groups accordmg to the postoperative course D, patients who died during the first postoperative month, C, patients with postoperative comphcatlons, and U, patients with an uneventful course Preoperative weight loss, duration of Jaundice, serum blhrubm level, total protein level, and age of every patient were collected and transformed mto a three-score ordinal scale Stepwlse dlscrlmmant analysis of these data enabled us to correctly classify 88 percent of the patients who underwent palliative operations and 83 percent of those who had pancreatic resection This method 1s useful m identifying the risk group for a single patient using mdlvldual preoperative variables It may, moreover, suggest the most suitable treatment for each patient with pancreatic cancer
References 1 Braasch JW, Gray BN Consrderatrons that lower pancreatoduodenectomy mortakty Am J Surg 1977, 133 480-4 2 Morrow M, Hrlarrs B, Brennan MF Comparison of conventronal surgical resection, radroactrve rmplantatlon, and bypass procedures for exocrme carcrnoma of the pancreas, 19751980 Ann Surg 1984, 199 l-5 3 Chapurs Y, Catala M, Place S, Bonnette P Reflexrons sur le trartment des cancers du pancreas Chrrurgre 1983, 109 357-8 4 Andren-Sandberg A, lhse I Factors rnfluencmg survival after total pancreatectomy in patients with pancreatic cancer Ann Surg 1983, 198 605-10 5 Nre NH, Hull CH, Jenkins JG, Sternbrenner K, Bent DH SPSS statrsbcal package for the socral sciences 2nd ed New York McGraw-Hill. 1975 6 Grlsdorf RB, Spanos P Factors mfluencrng morbrdtty and mortality In pancreatrcoduodenectomy Ann Surg 1973, 177 332-7 7 Herter FP, Cooperaman AM, Ahlborn TN, Antrnorr C Surgical expertence with pancreatic and perrampullary cancer Ann Surg 1982, 195 274-81 8 Buzby GP, Mullen JL, Matthews DC, Hobbs CL, Rosato EF Prognostic nutrrbonal index In gastrointestinal surgery Am J Surg 1980, 139 160-7 9 Sarr MG, Cameron JL Surgical management of unresectable carcmoma of the pancreas Surgery 1982, 91 123-33 10 Nakase A, Matsumoto Y, Uchrda K, Honjo I Surgical treatment of cancer of the pancreas and the perrampullary region cumulative results in 57 rnstrtubons rn Japan Ann Surg 1977, 185 52-7 11 Moossa AR, Lewis J, Mackre CR Surgical treatment of pancreatic cancer Mayo Clan Proc 1979, 54 468-74
377