Portal Vein Thrombosis Associated With Cirrhosis Clinical
Importance
I. James
Sarfeh, MD are found to have preexisting portal vein throm¬ bosis by angiography or at operation, presumably because of severe portal outflow block. These patients may repre¬ sent a different subpopulation of cirrhotics because their hepatic hemodynamics have already made long-term adjustments to total deprivation of portal inflow. Thus their clinical presentation as well as outcome after the standard portal decompressive procedures might be differ¬ ent from those cirrhotics without portal vein thrombosis. To test these theoretical considerations, cirrhotic patients with and without portal vein thrombosis who had under¬ gone portal decompression for variceal hemorrhage.
\s=b\ A group of 86 cirrhotics undergoing therapeutic variceal decompressive procedures were studied. Patients with portal vein thrombosis (PVT) comprised 21% of the group and more frequently had uncontrollable hemorrhage at an earlier stage of liver disease. Mortality in the 18 cirrhotics with PVT was higher (56%), mostly as a result of rebleeding. Despite adjustments for stage of liver disease and type of operation, in no cirrhotic with PVT did postshunt encephalopathy develop, compared with 32% incidence in patients without PVT (P < .05). It is concluded that patients with cirrhosis and PVT represent a different subpopulation of cirrhotics. Once adequate variceal decompression has been achieved, their prognosis should be superior to cirrhotics without PVT because their hepatic hemodynamics are unaffected by total shunting, hence precluding further impairment of liver function as a result of acute reduction of hepatic blood
however,
flow.
Hospital records of all patients with biopsy-proved cirrhosis who underwent operation for active or previous variceal hemorrhage were reviewed from 1972 through April 1978. The patients were all treated at Albany (NY) Medical Center and Albany Veterans Administration Hospitals. Follow-up information was obtained from records of hospital, clinic, and office visits. Included for study were patients who could be followed up for at least six months after operation. Thus, of the 97 cases reviewed, 86 patients with a minimum six months follow-up period were accepted for study. Of the 11 patients not accepted, one had portal vein thrombosis. The mean length of follow-up in patients without portal vein thrombo¬ sis was 16 months, ranging from a minimum of seven through a maximum of 40 months, compared with a mean of 19 months and a range of 12 through 43 months in patients with portal vein thrombosis. Portal vein thrombosis was demonstrated by selective angiogra¬ phy or splenoportography or at operation. Clinical stages of liver disease were established using Child's criteria." Operative circum¬ stances were classified as follows: (1) Elective: All bleeding had ceased within seven days of operation without use of splanchnic vasoconstrictors or balloon tamponade. (2) Urgent: Bleeding was
(Arch Surg 114:902-905, 1979) diversion and after with varices in cirrhotic Postshunt and accelerated are hepatic failure are the major causes of the poor results.1 Many investigators believe that these complications arise from acute deprivation of portal blood flow from the liver after the standard portacaval or mesocaval shunts."' Thus, newer variceal decompressive procedures that maintain hepatic portal inflow have been investigated in attempts to improve the surgical results."'7·" Some cirrhotic patients,
mortality portal Surgihigh.1cal morbidity bleeding esophageal patients encephalopathy
'
publication May 8, 1979. Department of Surgery, Albany (NY) Medical College. Reprint requests to Department of Surgery, Albany Medical College, Albany, NY 12208 (Dr Sarfeh). Accepted
From the
for
SUBJECTS AND METHODS
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Arizona Health Sciences Library User on 06/09/2015
Table 1.—Clinical Data of Patients With Cirrhosis and Variceal Hemorrhage
68 55 51 14-80
patients
Mean age, yr Median age, yr
Age range, yr
Portal Vein 18
Total No. of patients No. of deaths
53 51 41-68
Mortality, %_
Thrombosed Portal Vein
68 26 38
56
10
Causes of Death, No. (%) Liver failure/coagulopathy_19(73)_2 (20)
14
4(15) 3(12)
Other
Preoperative Hepatic
2(11) 4(22) 12(67)
16(24) 30(44)
and Causes of Death
Portal Vein
Rebleeding
Cause of cirrhosis, No. 53 Laennec's Other 15 Operative circumstances, No. (%) Elective 22(32)
Urgent Emergency
Mortality
Patent
Thrombosed
Patent Portal Vein No. of
Table 3—Assessment of
8(80)
Function
not significantly different in either but a group (Table 2), greater percentage of patients without portal vein thrombosis had albumin levels less than 3.0 g/dL ( 4.3; < .05, Yates' correction). The inci¬ of moderate or severe preoperative encephalopathy dence and ascites was less in patients with portal vein thrombo¬ sis, but the differences were not statistically significant. A higher proportion of patients with thrombosed portal veins were Child's class A, while fewer were class C when compared with patients who had patent portal veins ( - 4.07; < .05, Yates' correction).
Bilirubin levels
were
=
Table
2.—Preoperative
Assessment of Liver Function Patent Portal Vein No. of Patients 36
3.0
20
% 53 18 29
7
12
Albumin, g/dL
Thrombosed Portal Vein No. of Patients 7 7
39
4
39 22
10
3
17
29 32
43 47
12 3
66 17
34 16 18
50 24 26
11
61
5
28
39 21 8
57 31 12
14 3
78
25 24 19
37
10 6 2
56 33
=
Mortality
"
>3.5 3.0-3.5 .05, Yates' correction). =
Postoperative Encephalopathy in Survivors Of the 42 surviving patients without portal vein throm¬ bosis, 14 had substantial postoperative encephalopathy,
requiring continued therapy
after hospital discharge or readmission for treatment, whereas none of the eight surviving patients who had portal vein thrombosis had such severe encephalopathy. To make more valid comparisons,
31
10 32
!'MCS indicates mesocaval shunt; PCS, portacaval shunt.
all
surviving
class C patients with shunts other than mesocaval anastomoses have been excluded in the following analysis. Thus, when these patients are excluded, 32% of survivors without portal vein thrombosis had significant encephalopathy compared with none of the survivors with portal vein thrombosis ( 2 4.95; < .05, Yates' correction). Of the five class A and survivors with patent portal veins included in the above comparison, severe encephalopathy developed in two after mesocaval
portacaval
or
=
shunting.
COMMENT
This study has confirmed that portal vein thrombosis is frequently present in patients with bleeding eosphageal varices and cirrhosis. Several findings in the study would suggest that clinically these patients represent a different subpopulation of cirrhotics. The most significant of these findings was the absence of postshunt encephalopathy in the surviving cirrhotics with portal vein thrombosis when
compared with the 32% incidence in those with patent portal veins. This difference was present despite omission
of Child's class C survivors (mostly in the group without portal vein thrombosis) as well as those who had other than portacaval or mesocaval shunts, thus allowing for more valid comparisons. One interpretation of this finding might be that a higher proportion of surviving patients with portal vein thrombosis underwent mesocaval interpo¬ sition operations, so that lack of encephalopathy might be attributable to this operation. A functioning and effective mesocaval shunt, however, is hemodynamically similar to a side-to-side portacaval shunt in that both splanchnic as well as hepatic portal decompression are achieved.-" A more likely explanation for the absence of postshunt encephalopathy in the group of cirrhotics with portal vein thrombosis is that in response to loss of portal flow, long-term compensatory hepatic arterial changes had taken place. Thus, the subsequent portal decompressive operations did not alter the preexisting hepatic hemody¬ namics, sparing further deterioration of hepatic function.
This observation confirms that of Warren and others "" that patients whose portal flows are acutely reduced or reversed after total or "nonselective" portal decompressive
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Arizona Health Sciences Library User on 06/09/2015
operations have a high incidence of postshunt encephalopa¬ thy and liver failure. When compared with cirrhotics with patent portal veins, a significantly greater proportion of those with portal vein thrombosis had variceal hemorrhage of such magnitude that emergency operation was necessitated. This would suggest that compensatory collateral flow in these patients is less likely to prevent variceal hemorrhage or that bleeding occurs before adequate collaterals can develop. Along similar lines a greater proportion of patients with portal vein thrombosis had less advanced stages of liver disease, possibly because of the total portal obstruction resulting in earlier hemorrhage. In this series, variceal rebleeding after portal decom¬ pression in patients without preexisting portal vein throm¬ bosis occurred in 10% of patients. In this group, shunt failure occurred in a higher percentage (33%) of patients who had mesocaval interposition grafts as their primary procedure. Since in this procedure the length of prosthetic graft may be substantial, and kinking of the graft by adjacent structures (such as the third portion of the duodenum) can occur readily, shunt thrombosis is a distinct possibility as observed. Furthermore, in one such patient in the series, variceal bleeding recurred despite a patent mesocaval interposition graft. This phenomenon has been observed by others" and has been ascribed to altered flow patterns as a result of "jet streaming," resulting in diversion of gastrosplenic venous blood aw-ay from the graft. In the group of cirrhotics w-ith preexisting portal vein thrombosis, the prohibitively high (50%) variceal rebleeding rate attests to the difficulties in the surgical management of such patients. In this group, contrary to the group without portal vein thrombosis, the mesocaval interposition shunt was slightly, although not significant¬ ly, superior to the other procedures employed. Portal vein thrombectomy followed by portacaval shunt would appear unsatisfactory, as observed in this series, because of the chronicity of the thrombotic process, as well as the frequent extension of thrombus to areas in which extrac¬ tion of the thrombotic material is difficult. An alternative procedure in such patients would be proximal splenorenal shunt following splenectomy but the technical difficulties in the performance of this operation compared with the mesocaval interposition shunt would suggest that for many surgeons the latter should be the procedure of choice in cirrhotic patients with portal vein thrombosis. Variceal
rebleeding may occur in a few of these patients after the procedure, and angiography should then be performed since mesocaval graft thrombosis may not be the course. As demonstrated in one such patient in the present study, thrombus may have extended to the confluence of the splenic and superior mesenteric veins, resulting in segmentalization of the left and right sides of the splanchnic '-·''' venous system.1 The interposition graft would thus be prevented from providing adequate variceal decompres¬ sion. Under such circumstances splenectomy should be curative. In conclusion, this retrospective study would suggest that cirrhotics with portal vein thrombosis may well repre¬ sent a subpopulation of patients different from those without portal thrombosis. Variceal hemorrhage may occur at an earlier stage of progression of liver disease, and the likelihood of nonoperative control is less. If after operation adequate variceal decompression is achieved in cirrhotics with portal vein thrombosis, it may be anticipated that the incidence of postshunt encephalopathy will be far less than in cirrhotics with patent portal veins who undergo total
portal decompressive operations.
This study AM-18765.
was
supported
in part bv Public Health Service grant
References 1. Malt RA: Portasystemic shunts: I. N Engl Med 295:24-29, 1976. 2. Malt RA: Portasystemic shunts: II. N Engl J Med 295:80-86, 1976. 3. Resnick RH, Iber FL, Ishihara A, et al: A controlled study of the therapeutic portacaval shunt. Gastroenterology 67:843-857, 1974. 4. Jackson FC, Perrin EB, Felix WR, et al: A clinical investigation of the portacaval shunt. Ann Surg 174:672-701, 1971. 5. Britton RC: The clinical effectiveness of selective portal shunts. Am J Surg 133:506-511, 1977. 6. Salam AA, Warren WD, LePage JR, et al: Hemodynamic contrasts between selective and total portal-systemic decompression. Ann Surg 173:827-844, 1971. 7. Adamson RJ, Butt K, Iyer S, et al: Portacaval shunt with arterialization of the portal vein by means of a low flow arteriovenous fistula. Surg Gynecol Obstet 146:869-876, 1978. 8. Warren WD, Salam AA, Hutson D, et al: Selective distal splenorenal shunt. Arch Surg 108:306-314, 1974. 9. Child CG III: The Liver and Portal Hypertension. Philadelphia, WB Saunders Co, 1964, vol 1, p 50. 10. Brownlee AK: Statistical Theory and Methodology: Science in Engineering. New York, John Wiley & Sons Inc, 1965. 11. Rikkers LF, Rudman D, Galambos JT, et al: A randomized, controlled trial of the distal splenorenal shunt. Ann Surg 188:271-282, 1978. 12. Stone R, Wilson S, Passro E: Gastric portal hypertension. Am J Surg 136:73-79, 1978. 13. Witte CL, Ovitt TW, Witte MH, et al: Left-sided segmental portal hypertension following interposition mesocaval shunt. Surg Gynecol Obstet 145:169-174, 1977.
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Arizona Health Sciences Library User on 06/09/2015
J
Importance
I. James
Sarfeh, MD are found to have preexisting portal vein throm¬ bosis by angiography or at operation, presumably because of severe portal outflow block. These patients may repre¬ sent a different subpopulation of cirrhotics because their hepatic hemodynamics have already made long-term adjustments to total deprivation of portal inflow. Thus their clinical presentation as well as outcome after the standard portal decompressive procedures might be differ¬ ent from those cirrhotics without portal vein thrombosis. To test these theoretical considerations, cirrhotic patients with and without portal vein thrombosis who had under¬ gone portal decompression for variceal hemorrhage.
\s=b\ A group of 86 cirrhotics undergoing therapeutic variceal decompressive procedures were studied. Patients with portal vein thrombosis (PVT) comprised 21% of the group and more frequently had uncontrollable hemorrhage at an earlier stage of liver disease. Mortality in the 18 cirrhotics with PVT was higher (56%), mostly as a result of rebleeding. Despite adjustments for stage of liver disease and type of operation, in no cirrhotic with PVT did postshunt encephalopathy develop, compared with 32% incidence in patients without PVT (P < .05). It is concluded that patients with cirrhosis and PVT represent a different subpopulation of cirrhotics. Once adequate variceal decompression has been achieved, their prognosis should be superior to cirrhotics without PVT because their hepatic hemodynamics are unaffected by total shunting, hence precluding further impairment of liver function as a result of acute reduction of hepatic blood
however,
flow.
Hospital records of all patients with biopsy-proved cirrhosis who underwent operation for active or previous variceal hemorrhage were reviewed from 1972 through April 1978. The patients were all treated at Albany (NY) Medical Center and Albany Veterans Administration Hospitals. Follow-up information was obtained from records of hospital, clinic, and office visits. Included for study were patients who could be followed up for at least six months after operation. Thus, of the 97 cases reviewed, 86 patients with a minimum six months follow-up period were accepted for study. Of the 11 patients not accepted, one had portal vein thrombosis. The mean length of follow-up in patients without portal vein thrombo¬ sis was 16 months, ranging from a minimum of seven through a maximum of 40 months, compared with a mean of 19 months and a range of 12 through 43 months in patients with portal vein thrombosis. Portal vein thrombosis was demonstrated by selective angiogra¬ phy or splenoportography or at operation. Clinical stages of liver disease were established using Child's criteria." Operative circum¬ stances were classified as follows: (1) Elective: All bleeding had ceased within seven days of operation without use of splanchnic vasoconstrictors or balloon tamponade. (2) Urgent: Bleeding was
(Arch Surg 114:902-905, 1979) diversion and after with varices in cirrhotic Postshunt and accelerated are hepatic failure are the major causes of the poor results.1 Many investigators believe that these complications arise from acute deprivation of portal blood flow from the liver after the standard portacaval or mesocaval shunts."' Thus, newer variceal decompressive procedures that maintain hepatic portal inflow have been investigated in attempts to improve the surgical results."'7·" Some cirrhotic patients,
mortality portal Surgihigh.1cal morbidity bleeding esophageal patients encephalopathy
'
publication May 8, 1979. Department of Surgery, Albany (NY) Medical College. Reprint requests to Department of Surgery, Albany Medical College, Albany, NY 12208 (Dr Sarfeh). Accepted
From the
for
SUBJECTS AND METHODS
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Arizona Health Sciences Library User on 06/09/2015
Table 1.—Clinical Data of Patients With Cirrhosis and Variceal Hemorrhage
68 55 51 14-80
patients
Mean age, yr Median age, yr
Age range, yr
Portal Vein 18
Total No. of patients No. of deaths
53 51 41-68
Mortality, %_
Thrombosed Portal Vein
68 26 38
56
10
Causes of Death, No. (%) Liver failure/coagulopathy_19(73)_2 (20)
14
4(15) 3(12)
Other
Preoperative Hepatic
2(11) 4(22) 12(67)
16(24) 30(44)
and Causes of Death
Portal Vein
Rebleeding
Cause of cirrhosis, No. 53 Laennec's Other 15 Operative circumstances, No. (%) Elective 22(32)
Urgent Emergency
Mortality
Patent
Thrombosed
Patent Portal Vein No. of
Table 3—Assessment of
8(80)
Function
not significantly different in either but a group (Table 2), greater percentage of patients without portal vein thrombosis had albumin levels less than 3.0 g/dL ( 4.3; < .05, Yates' correction). The inci¬ of moderate or severe preoperative encephalopathy dence and ascites was less in patients with portal vein thrombo¬ sis, but the differences were not statistically significant. A higher proportion of patients with thrombosed portal veins were Child's class A, while fewer were class C when compared with patients who had patent portal veins ( - 4.07; < .05, Yates' correction).
Bilirubin levels
were
=
Table
2.—Preoperative
Assessment of Liver Function Patent Portal Vein No. of Patients 36
3.0
20
% 53 18 29
7
12
Albumin, g/dL
Thrombosed Portal Vein No. of Patients 7 7
39
4
39 22
10
3
17
29 32
43 47
12 3
66 17
34 16 18
50 24 26
11
61
5
28
39 21 8
57 31 12
14 3
78
25 24 19
37
10 6 2
56 33
=
Mortality
"
>3.5 3.0-3.5 .05, Yates' correction). =
Postoperative Encephalopathy in Survivors Of the 42 surviving patients without portal vein throm¬ bosis, 14 had substantial postoperative encephalopathy,
requiring continued therapy
after hospital discharge or readmission for treatment, whereas none of the eight surviving patients who had portal vein thrombosis had such severe encephalopathy. To make more valid comparisons,
31
10 32
!'MCS indicates mesocaval shunt; PCS, portacaval shunt.
all
surviving
class C patients with shunts other than mesocaval anastomoses have been excluded in the following analysis. Thus, when these patients are excluded, 32% of survivors without portal vein thrombosis had significant encephalopathy compared with none of the survivors with portal vein thrombosis ( 2 4.95; < .05, Yates' correction). Of the five class A and survivors with patent portal veins included in the above comparison, severe encephalopathy developed in two after mesocaval
portacaval
or
=
shunting.
COMMENT
This study has confirmed that portal vein thrombosis is frequently present in patients with bleeding eosphageal varices and cirrhosis. Several findings in the study would suggest that clinically these patients represent a different subpopulation of cirrhotics. The most significant of these findings was the absence of postshunt encephalopathy in the surviving cirrhotics with portal vein thrombosis when
compared with the 32% incidence in those with patent portal veins. This difference was present despite omission
of Child's class C survivors (mostly in the group without portal vein thrombosis) as well as those who had other than portacaval or mesocaval shunts, thus allowing for more valid comparisons. One interpretation of this finding might be that a higher proportion of surviving patients with portal vein thrombosis underwent mesocaval interpo¬ sition operations, so that lack of encephalopathy might be attributable to this operation. A functioning and effective mesocaval shunt, however, is hemodynamically similar to a side-to-side portacaval shunt in that both splanchnic as well as hepatic portal decompression are achieved.-" A more likely explanation for the absence of postshunt encephalopathy in the group of cirrhotics with portal vein thrombosis is that in response to loss of portal flow, long-term compensatory hepatic arterial changes had taken place. Thus, the subsequent portal decompressive operations did not alter the preexisting hepatic hemody¬ namics, sparing further deterioration of hepatic function.
This observation confirms that of Warren and others "" that patients whose portal flows are acutely reduced or reversed after total or "nonselective" portal decompressive
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Arizona Health Sciences Library User on 06/09/2015
operations have a high incidence of postshunt encephalopa¬ thy and liver failure. When compared with cirrhotics with patent portal veins, a significantly greater proportion of those with portal vein thrombosis had variceal hemorrhage of such magnitude that emergency operation was necessitated. This would suggest that compensatory collateral flow in these patients is less likely to prevent variceal hemorrhage or that bleeding occurs before adequate collaterals can develop. Along similar lines a greater proportion of patients with portal vein thrombosis had less advanced stages of liver disease, possibly because of the total portal obstruction resulting in earlier hemorrhage. In this series, variceal rebleeding after portal decom¬ pression in patients without preexisting portal vein throm¬ bosis occurred in 10% of patients. In this group, shunt failure occurred in a higher percentage (33%) of patients who had mesocaval interposition grafts as their primary procedure. Since in this procedure the length of prosthetic graft may be substantial, and kinking of the graft by adjacent structures (such as the third portion of the duodenum) can occur readily, shunt thrombosis is a distinct possibility as observed. Furthermore, in one such patient in the series, variceal bleeding recurred despite a patent mesocaval interposition graft. This phenomenon has been observed by others" and has been ascribed to altered flow patterns as a result of "jet streaming," resulting in diversion of gastrosplenic venous blood aw-ay from the graft. In the group of cirrhotics w-ith preexisting portal vein thrombosis, the prohibitively high (50%) variceal rebleeding rate attests to the difficulties in the surgical management of such patients. In this group, contrary to the group without portal vein thrombosis, the mesocaval interposition shunt was slightly, although not significant¬ ly, superior to the other procedures employed. Portal vein thrombectomy followed by portacaval shunt would appear unsatisfactory, as observed in this series, because of the chronicity of the thrombotic process, as well as the frequent extension of thrombus to areas in which extrac¬ tion of the thrombotic material is difficult. An alternative procedure in such patients would be proximal splenorenal shunt following splenectomy but the technical difficulties in the performance of this operation compared with the mesocaval interposition shunt would suggest that for many surgeons the latter should be the procedure of choice in cirrhotic patients with portal vein thrombosis. Variceal
rebleeding may occur in a few of these patients after the procedure, and angiography should then be performed since mesocaval graft thrombosis may not be the course. As demonstrated in one such patient in the present study, thrombus may have extended to the confluence of the splenic and superior mesenteric veins, resulting in segmentalization of the left and right sides of the splanchnic '-·''' venous system.1 The interposition graft would thus be prevented from providing adequate variceal decompres¬ sion. Under such circumstances splenectomy should be curative. In conclusion, this retrospective study would suggest that cirrhotics with portal vein thrombosis may well repre¬ sent a subpopulation of patients different from those without portal thrombosis. Variceal hemorrhage may occur at an earlier stage of progression of liver disease, and the likelihood of nonoperative control is less. If after operation adequate variceal decompression is achieved in cirrhotics with portal vein thrombosis, it may be anticipated that the incidence of postshunt encephalopathy will be far less than in cirrhotics with patent portal veins who undergo total
portal decompressive operations.
This study AM-18765.
was
supported
in part bv Public Health Service grant
References 1. Malt RA: Portasystemic shunts: I. N Engl Med 295:24-29, 1976. 2. Malt RA: Portasystemic shunts: II. N Engl J Med 295:80-86, 1976. 3. Resnick RH, Iber FL, Ishihara A, et al: A controlled study of the therapeutic portacaval shunt. Gastroenterology 67:843-857, 1974. 4. Jackson FC, Perrin EB, Felix WR, et al: A clinical investigation of the portacaval shunt. Ann Surg 174:672-701, 1971. 5. Britton RC: The clinical effectiveness of selective portal shunts. Am J Surg 133:506-511, 1977. 6. Salam AA, Warren WD, LePage JR, et al: Hemodynamic contrasts between selective and total portal-systemic decompression. Ann Surg 173:827-844, 1971. 7. Adamson RJ, Butt K, Iyer S, et al: Portacaval shunt with arterialization of the portal vein by means of a low flow arteriovenous fistula. Surg Gynecol Obstet 146:869-876, 1978. 8. Warren WD, Salam AA, Hutson D, et al: Selective distal splenorenal shunt. Arch Surg 108:306-314, 1974. 9. Child CG III: The Liver and Portal Hypertension. Philadelphia, WB Saunders Co, 1964, vol 1, p 50. 10. Brownlee AK: Statistical Theory and Methodology: Science in Engineering. New York, John Wiley & Sons Inc, 1965. 11. Rikkers LF, Rudman D, Galambos JT, et al: A randomized, controlled trial of the distal splenorenal shunt. Ann Surg 188:271-282, 1978. 12. Stone R, Wilson S, Passro E: Gastric portal hypertension. Am J Surg 136:73-79, 1978. 13. Witte CL, Ovitt TW, Witte MH, et al: Left-sided segmental portal hypertension following interposition mesocaval shunt. Surg Gynecol Obstet 145:169-174, 1977.
Downloaded From: http://archsurg.jamanetwork.com/ by a University of Arizona Health Sciences Library User on 06/09/2015
J
