Septicemia in Patients on Chronic Hemodialysis JAY F. DOBKIN, M.D.; MICHAEL H. MILLER, M.D.; and NEAL H. STEIGBIGEL, M.D., F.A.C.P.; Bronx, New York
Bacterial sepsis, a major complication of chronic hemodialysis, is due mainly to infections of the vascular access site despite increasing use of internal fistulas. Sixty episodes of septicemia occurred in two chronic dialysis centers, with an incidence of 0.15 episodes of significant bacteremia per patient-dialysis-year in each. Forty-four of the 6 0 episodes were judged to be due to vascular access site infection by clinical, bacteriologic, and histologic criteria. Seventy percent ( 3 1 of 4 4 ) of the vascular access siterelated episodes were due to staphylococci and 2 5 % ( 1 1 of 4 4 ) to Gram-negative bacilli; nonvascular access site-related episodes were often due to transplant site infections caused by Gram-negative bacilli or streptococci. Mortality was about 1 8 % in both vascular access site-related and nonrelated septic episodes. Bovine heterograft arteriovenous fistulas more often led to sepsis than did Brescia arteriovenous fistulas. Treatment with appropriate antibiotics was successful in most cases. Routine removal or ligation of the vascular access site was not necessary.
I N F E C T I O N accounts for much of the morbidity and mortality associated with acute and chronic renal failure ( 1 3). Hemodialysis and renal transplantation, which have been successful in prolonging the lives of patients with chronic renal insufficiency, are often complicated by infection leading to significant mortality (4, 5). The first successful vascular access for chronic dialysis, the Quinton-Scribner external shunt, was frequently associated with complications such as bleeding, thrombosis, and local and systemic infection (6-8). The internal arteriovenous fistula was introduced by Brescia and colleagues (9) with the expectation of fewer complications (10); comparisons of internal fistulas with external shunts suggested that fistulas had greater longevity and fewer local complications (11-15). Therefore, the internal fistula has become the predominant vascular access site in the chronic hemodialysis population. Reports describing bacteremic infection in chronic hemodialysis patients subsequent to the introduction of internal arteriovenous fistulas are few and have involved only a small number of patients (14, 16-18). However, in view of a substantial experience at our institutions with vascular access site-related sepsis in a large population of chronic hemodialysis • F r o m the Division of Infectious Diseases, D e p a r t m e n t of Medicine, Montefiore Hospital a n d Medical Center, and t h e D e p a r t m e n t of Medicine, Albert Einstein College of Medicine; Bronx, N e w York. 28
© 1 9 7 8 American College of Physicians
p a t i e n t s , m a i n l y u s i n g i n t e r n a l fistulas, w e h a v e u n d e r t a k e n a r e t r o s p e c t i v e s t u d y of t h e e p i d e m i o l o g y , b a c t e r i o l o g y , clinical c h a r a c t e r i s t i c s , a n d p r o g n o s i s of t h i s e n t i ty. V a s c u l a r a c c e s s site-related sepsis w a s c o m p a r e d t o s e p t i c e p i s o d e s a r i s i n g from o t h e r foci in t h e s a m e h e m o dialysis p o p u l a t i o n . Materials and Methods Records of all patients with positive blood cultures hemodialyzed from 1972 to 1975 at the chronic dialysis facilities (Center A and Center B) affiliated with Montefiore Hospital and the Albert Einstein College of Medicine were reviewed. T h e two centers have separate patient populations and medical staffs. Positive cultures were considered to be the result of contamination if one bottle of a set yielded a typically nonpathogenic organism and if chart review gave no indication of an infectious process and the patient recovered without antimicrobial therapy. All bacteremias not excluded by these criteria were considered significant and consisted of 60 episodes in 43 patients. All of these episodes were associated with signs and symptoms of sepsis, and all but two had at least two positive blood cultures. The vascular access site was an internal fistula in 38 patients and an external shunt in four while one patient had both. Of the 39 patients with internal fistulas, 30 were Brescia arteriovenous fistulas and nine were bovine heterografts. The 60 episodes of significant bacteremia were classified by chart analysis as definitely, probably, or possibly vascular access site-related, or as nonvascular access site-related. T h e following criteria were used. A. Vascular access site-related (44 episodes) 1. Definite (16 episodes) Histologic or bacteriologic proof of local infection at the vascular access site along with signs of infection at the vascular access site and without another apparent primary source of bacteremia. 2. Probable (13 episodes) No apparent primary source for the bacteremia other than the vascular access site, and either local signs of infection at the vascular access site (without histologic or bacteriologic proof) (12 episodes) or signs of sepsis developing in close temporal relation to manipulation of the vascular access site at the time of dialysis (one episode). 3. Possible (15 episodes) No apparent source for the bacteremia other than the vascular access site. B. Nonvascular access site-related (16 episodes) No signs of infection at the vascular access site; evident focus of infection at another site. Signs taken to signify local infection at the vascular access site included purulent discharge, fluctuance, or severe tenderness. Erythema, warmth, and induration, which are signs often found normally at the vascular access site, were not considered signs of local vascular access site infection. In 32 of 44 episodes classified as vascular access site-related septicemia, there was evidence of prolonged bacteremia, suggesting endovascular infection: 24 episodes with multiple positive blood cultures four or more h apart; and eight episodes with multiple positive blood cultures (at least six bottles posiAnnals of Internal Medicine 8 8 : 2 8 - 3 3 , 1 9 7 8
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19529/ by a University of California San Diego User on 06/12/2017
Table 1 . Incidence of Septicemia in Chronic Hemodialysis Patients
Center
Patients
Dialyses
178 72 250
24 856 7629 32 485
Bacteremic Episodes
Vascular Access Site-Related Episodes
32 11 43
23 (72) 9(82) 32 (74)
-> %~
no. A*
Bt
Total
Risk of Bacteremic Infection per PatientDialysis-Year 0.15 0.15 0.15
* Twenty two-month period (April 1973-January 1975). t Twenty-month period (August 1973-March 1975).
tive) without recorded times. Incidence data were compiled for a representative period at each center during which complete information was available. The dialysis record for each patient during this period was reviewed. At Center A this period covered 22 months (April 1973-January 1975) and at Center B 20 months (August 1973March 1975). A chronic dialysis patient was defined as one hemodialyzed at least once a week for 3 weeks in at least 2 months. One patient-dialysis-month was defined as dialysis of one chronic hemodialysis patient in 3 different weeks in a given month. A patient-dialysis-year was defined as 12 patient-dialysis-months. Data were analyzed for statistical significance by the chisquare, Fisher's exact, or Student's t test. Results INCIDENCE AND CLINICAL SETTING
Forty-three of the 60 episodes of significant bacteremia took place during the periods studied for incidence (Table 1). Incidence was identical in the two independent populations studied: 0.15 episodes of bacteremia per patient-dialysis-year. At Center A 13.5% of patients and at Center B 12.5% had one or more bacteremias during the period of analysis. About 7 5 % of the episodes of septicemia were vascular access site-related (0.11 episodes per patient-dialysis-year). Many patients on chronic hemodialysis had several revisions and replacements of their vascular access sites due to malfunctioning, bleeding, infection, or other reasons. However, 4 5 % (20 of 44) of vascular access site-related septic episodes occurred in patients who were still being dialyzed through their original vascular access site. Bovine heterograft fistulas were surgically placed for only a limited period during the study. Of the 67 bovine grafts placed, eight were associated with sepsis. One patient with sepsis who had a bovine heterograft placed at another institution was excluded from this analysis. During this same period, 236 Brescia fistulas were placed, eight of which were associated with sepsis. The greater incidence of sepsis with bovine heterografts was significant (P < 0.05). This occurred despite a shorter period of chronic hemodialysis before infection for patients with bovine heterografts (mean duration, 9.5 months) compared to those with Brescia arteriovenous fistulas (mean duration, 17.8 months). The mean duration of time between the construction of the most recent vascular access and the onset of infection was significantly less in patients with bovine heterografts (1.6 months) than in patients with Brescia fistulas (15.0 months), P < 0.01. Septic episodes occurred throughout the period of
study at both centers without clustering suggestive of epidemic spread. Staphylococcal phage typing or antibiotic susceptibility data available for 20 episodes in 18 patients at Center A showed each bacterial isolate to be a different strain. In four of five patients who had multiple episodes of infection, different species were involved in each episode. The fifth patient had four episodes of Staphylococcus aureus sepsis; two of the organisms were phage-typed and were different. Ten of 16 nonvascular access site-related episodes were attributed to a genitourinary source, five to a pulmonary source, and one to a soft tissue infection. Six of the 10 episodes related to a genitourinary source involved patients with perinephric abscesses or nephrectomy-site abscesses after unsuccessful renal transplantation. Seven of the 16 episodes of nonvascular access site-related sepsis involved patients with recent renal transplantation who were being treated with immunosuppressive therapy. Five of the six perinephric abscesses occurred in the latter group. In contrast, only three of 44 vascular access siterelated episodes took place in patients receiving immunosuppressive therapy. Patients with vascular access site-related sepsis did not differ significantly from those with nonvascular access site-related sepsis in age, sex, type of renal disease, or presence of other underlying chronic disease. Similarly these factors did not differ significantly between the septic patients with bovine heterografts and those with Brescia arteriovenous fistulas. BACTERIOLOGY A N D O T H E R L A B O R A T O R Y D A T A
Vascular access site-related cases were predominantly due to staphylococci (31 of 44 episodes, or 70%), while septicemias unrelated to the vascular access site were mostly due to streptococci or Gram-negative bacilli (13 of 16 episodes, or 81%) (Table 2). Staphylococcus aureus was a significantly more common pathogen in the groups classified as definitely or probably related to the vascular access site (21 of 29 episodes) than in the possible group (five of 15 episodes), P < 0.05, or the nonvascular access site-related group (two of 16 episodes), P < 0.005 (Table 2). Of the six episodes of vascular access site-related sepsis that occurred in three heroin users, five were due to 5. aureus. Among the Gram-positive cocci other than staphylococci, enterococci were isolated in three episodes. Among the Gram-negative bacilli other than Enterobacteriaceae, Pseudomonas species were isolated in five episodes (four Dobkinetal.
• Septicemia and Hemodialysis
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19529/ by a University of California San Diego User on 06/12/2017
29
Table 2. Bacterial Isolates in 60 Episodes of Septicemia
Bacterial Isolate
Relation of 1Infection to the Vascular Access Site Definite or Probable no.
Staphylococcus aureus Staphylococcus epidermidis Other Grampositive cocci Enterobacteriaceae Other Gramnegative rods
%
21 (70.0)
Possible no.
%
Nonvascula Access Site Related
no.
%
5 (31.2)
2 (12.5)
2
(6.7)
3* (18.8)
1 (6.2)
2 2t
(6.7) (6.7)
1 (6.2) 5* (31.2)
7 (43.8) 3 (18.8)
3 f (10.0)
2 (12.5)
3 (18.8)
* Includes one mixed bacteremia: Proteus mirabilis and S. epidermidis. t Includes one mixed bacteremia: Enterobacter and Pseudomonas species.
were vascular access site-related), Bacteroides species in two episodes, and Hemophilus influenzae in one episode. Purulent material obtained from the vascular access site in 15 episodes of vascular access site-related sepsis yielded the same organism on culture as that obtained from blood cultures. Gram stain of purulent material accurately predicted the class of organisms obtained by culture. Total and differential leukocyte counts at the onset of sepsis were normal in most cases, and there was no significant difference between the vascular access site-related and nonrelated groups in this regard. SIGNS A N D SYMPTOMS IN VASCULAR ACCESS SITER E L A T E D EPISODES
About 8 0 % of episodes in both the vascular access siterelated and nonrelated groups had accompanying fever (38 °C or more) at the time of presentation. Rigors occurred in 4 5 % (13 of 29) of those cases in the definite and probable groups, but in none of the 15 cases classified as possibly vascular access site-related {P < 0.01) and none of those cases considered unrelated to the vascular access site. Significant local signs and symptoms of infection at the vascular access site were more common with staphylococcal infections than with those due to organisms other than staphylococci. Severe pain was noted in eight of the 30 cases due to staphylococci alone and in none of 13 nonstaphylococcal cases, P < 0.05. Fifty-three percent (16 of 30) of staphylococcal episodes were characterized by at least one of the following findings: severe pain, purulent discharge, skin necrosis, or abscesses. Twentythree percent (three of 13) of the nonstaphylococcal cases had one or more of these findings. The physical finding of a purulent discharge was more common in the small number of patients with external shunts (three of five) than in those patients with arteriovenous fistulas (five of 39). In two patients, development of vascular access site-related septicemia followed the surgical removal or revision of a locally infected arteriovenous fistula. Changing signs, symptoms, and laboratory data in each case suggested 30
that manipulation had caused dissemination of a previously localized process. METASTATIC INFECTION
Sixteen metastatic infections occurred in 13 of 44 vascular access site-related episodes: septic pulmonary emboli, six; endocarditis, four; meningitis, two; metastatic urinary tract infection, two; empyema, one; and myocardial abscesses, one. One of the patients with endocarditis and the patient with myocardial abscesses were detected only at autopsy. Fourteen of the 16 metastatic infections were due to S. aureus. Metastatic infections were much more frequent in staphylococcal vascular access site-related sepsis, occurring in 10 of 31 episodes compared to one of 13 episodes in nonstaphylococcal sepsis. The four cases of endocarditis among those with vascular access site-related sepsis were all associated with 5. aureus. Among the episodes of vascular access site-unrelated sepsis, one patient had endocarditis due to S. aureus. Evidence of endocarditis diagnosed during life included the development of acute aortic insufficiency in all four patients. Septic pulmonary emboli were detected in six of 44 vascular access site-related cases (14%). Repeated episodes of septic pulmonary emboli were not observed. T R E A T M E N T A N D COURSE
All but three of 60 episodes were treated with appropriate antibiotics (based on culture and sensitivity data) within 48 h of the onset of signs and symptoms of sepsis. Most patients received 4 to 6 weeks of antibiotic treatment. Data regarding antibiotic therapy are not adequate to allow comparison of different regimens. Defervescence occurred within 3 days of the start of therapy in 7 1 % of cases. Surgery was done on the vascular access site in the management of eight episodes (seven in internal fistulas) of vascular access site-related sepsis. Clinical indications for surgery included appearance of an abscess, thrombosis, or bleeding. The findings at surgery were abscesses in five, thrombosis in five, and pseudoaneurysm in four. There was one death among these eight episodes. None of the four patients with clinical evidence of endocarditis had surgery performed on the vascular access site. Only one of the six patients with septic pulmonary emboli underwent surgery of the vascular access site, for thrombectomy. There was no evidence that continued hemodialysis through an infected arteriovenous fistula predisposed patients to septic pulmonary embolization. Several patients with vascular access site-related sepsis were dialyzed through their arteriovenous fistulas without evident complications. Three of the six cases of septic pulmonary emboli occurred in patients who were not being dialyzed through the infected fistula at the time of embolization, but who were being dialyzed through another vascular access site or femoral vein puncture. MORTALITY
The case fatality rate for both the vascular access site-
January 1978 • Annals of Internal Medicine • Volume 88 • Number 1
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19529/ by a University of California San Diego User on 06/12/2017
One (17) reported 0.18 episodes per patient-dialysis-year for all bacteremic events, while the other (16) noted 0.20 episodes per patient-dialysis-year for vascular access siterelated septicemias. Therefore, although local infections appear to occur less frequently with internal arteriovenous fistulas than with external vascular shunts, the incidence of septicemia in patients with these two types of vascular access sites is similar. Epidemiologic findings, including phage typing, antibiotic sensitivity data, and the lack of case clustering, suggest autoinfection in the pathogenesis of vascular access site-related sepsis in our series. The same route of spread was found in a prospective study of local infections of external shunts (20). Sepsis occurred earlier and with greater frequency in our patients with bovine heterograft arteriovenous fistulas than in those with Brescia fistulas, despite a significantly shorter history of chronic hemodialysis in the former group. Bovine grafts have been recommended as the alternative vascular access for patients with vessels inadequate for the Brescia fistula (24). The higher incidence of septicemia with bovine heterograft fistulas has not been previously reported. Along with observations by others (25, 26) that local infection may be common with bovine heterografts, our findings prompt caution against unnecessary use of these prostheses. Our report shows that vascular access site-related sepsis in patients with arteriovenous fistulas is predominantly caused by staphylococci; however, an appreciable number of cases were due to enteric Gram-negative bacilli and Pseudomonas species. In contrast, 29 of the 31 cases of sepsis with arteriovenous fistulas reported in three studies (14, 16, 18) were due to staphylococci. Local infection and septicemia in patients with external shunts are also predominantly due to staphylococci, although Pseudomonas species have been frequently reported as well (8, 11). Sepsis unrelated to the vascular access site occurred most often as a complication of previous renal transplantation. Sepsis in these patients arose particularly from surgical wound infection, nephrectomy site abscess, or urinary tract infection and was due predominantly to Gram-negative rods and streptococci. This clinical and bacteriologic pattern has been noted in several major reviews of infection in transplant patients (27-30). The correlation of sepsis with renal transplant rejection has been documented, although as yet the possible predisposing role of increased immunosuppressive therapy has not
related and nonvascular access site-related episodes was about 18% (Table 3). Mortality for vascular access siterelated cases due to Gram-negative bacilli (five of 10) was significantly greater than that for staphylococci (three of 30), P < 0.05 (Table 3). Increased age (60 years or older) carried a fourfold increased death rate in vascular access site-related cases and a ninefold increase in nonvascular access site-related cases. Sepsis in patients dialyzed for 12 or more months had a mortality rate twice that of those dialyzed for less than 12 months. Mortality rates in vascular access site-related episodes with metastatic complications (three of 13) or in those with surgical procedures performed on the vascular access site (one of eight) were not significantly different from episodes without such complications (five of 31) or procedures (seven of 36). None of the four patients with clinical evidence of endocarditis died. Discussion
Published studies of the epidemiology of infection in chronic hemodialysis patients have emphasized local infection at the vascular access site. Reported rates of local infection in external shunts range from about one to three episodes per patient-dialysis-year (6, 7, 14, 19-21). Reported rates of local infection with arteriovenous fistulas are lower, occurring at 0.15 episodes per-patient-dialysisyear or less (14, 22). However, in our experience bacteremic infection continues to be a major complication of chronic renal failure despite the use of maintenance hemodialysis and the development of the internal fistula to replace the external shunt for vascular access. Clinical reports of disseminated sepsis and endocarditis in hemodialysis patients are primarily of patients with external shunts (23). Septicemia secondary to local vascular access site infection in patients with external shunts has been reported at rates from 0.10 to 0.30 episodes per patientdialysis-year (7, 14, 21). In our study of patients on hemodialysis mainly with arteriovenous fistulas for vascular access, sepsis occurred at a rate of 0.15 episodes per patient-dialysis-year at each of two centers. Two clinical entities accounted for most of the septic episodes: bacteremic vascular access site infections, occurring at a rate of 0.11 episodes per patient-dialysis-year; and bacteremic transplant site or genitourinary tract infections in patients with rejected renal transplants. The incidence figures in our study are similar to those of two French reports (16, 17) involving a total of 34 episodes of sepsis in patients with arteriovenous fistulas. Table 3. Mortality of Septicemia in Chronic Hemodialysis Patients
Mortality by Bacterial Cause Relation of Septicemia to the Vascular Access Site
Overall Mortality
Definite or probable Possible Total vascular access site-related Vascular access site-unrelated Total
no. % 3/29 (10) 5/15 (33) 8/44 (18) 3/16(19) 11/60(18)
Staphylococci no. 2/23 1/7 3/30 0/3 3/33
% (9) (14)* (10) (9)
Gram-Negative Rods no. 1/4 4/6 5/10 0/6 5/16
% (25) (67)* (50) (31)
Other Organisms no. % 0/2 0/1 0/3 3/7 (43) 3/10 (30)
'• Nonfatal case with mixed staphylococcal and Gram-negative bacteremia excluded. Dobkin et al. • Septicemia and Hemodialysis
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19529/ by a University of California San Diego User on 06/12/2017
31
been definitely established (30,31). Our experience indicates that the clinical entity of vascular access site-related sepsis with the internal fistula differs from that with the external shunt in terms of local signs, clinical course, and complications. Local inflammation with purulent discharge, pseudoaneurysm formation, thrombosis of the shunt, bleeding around the shunt, and even extrusion of the cannula are frequent signs in external shunt infection (6, 7). In contrast, nearly one third of our cases of vascular access site-related sepsis had no local findings distinguishable from those reflecting the usual inflammation accompanying regular use of the fistula. Severe pain and tenderness were the most prominent local findings with septicemic arteriovenous fistula infection. Local signs of inflammation at the vascular access site were most prominent with staphylococcal vascular access site-related sepsis compared to sepsis due to other organisms. Our 15 episodes classified as possibly vascular access site-related had several characteristics pointing to the vascular access site as the source despite absence of local findings: there were no other evident primary sources for sepsis in these patients; prolonged bacteremia in most patients indicated endovascular infection, while the absence of findings of endocarditis combined with the frequent Gram-negative bacteriology made the diagnosis of bacterial endocarditis unlikely; and the group's high rate of mortality and prolonged bacteremia are inconsistent with contamination or transient bacteremia. Although bacterial endocarditis or transient bacteremia cannot be ruled out in all of these cases, there is strong evidence for occult vascular access site infection in most of them. Septicemia due to clinically inapparent infection of the arteriovenous fistula has been suggested by Meyrier and colleagues (16) in nine of 22 episodes of staphylococcal septicemia in patients with arteriovenous fistulas and in six of seven cases of arteriovenous fistula-related sepsis reported by Levi, Robson, and Rosenfeld (18), although data on the exclusion of other infected foci, and the duration of bacteremia, are not given. Metastatic complications associated with vascular access site-related sepsis strongly correlate with infection by 5. aureus, which was responsible for 14 of 16 incidents in our series. Septic pulmonary emboli have been noted as sporadic or recurrent complications of vascular access site infection with both external shunts and arteriovenous fistulas (16, 18, 32) and may occur without local signs of infection at the vascular access site. Removal of an external shunt is frequently necessary to prevent recurrent septic emboli (32), but this was not necessary in our patients with arteriovenous fistula infections or in those reported by others (16, 18). Bacterial endocarditis, diagnosed in five of our patients, has been associated with chronic hemodialysis and is usually staphylococcal or streptococcal in origin (23). Endocarditis must be considered as a diagnostic possibility in septic hemodialysis patients, especially those with prolonged bacteremia. However, the absence of the usual clinical findings of bacterial endocarditis or the isolation of Gram-negative organisms should suggest vascular ac32
cess site infection as a stronger possibility. The concurrent presence of subclinical endocarditis and infection of the vascular access site cannot be ruled out in patients who have staphylococcal bacteremia and strong evidence of infection of the vascular access site. Empiric antimicrobial therapy of sepsis of vascular access site origin should include coverage for S. aureus and Gram-negative bacilli. Our experience, and that of others (33, 34), suggests that surgical treatment is indicated for drainage of abscesses or hematomas, resection of aneurysms, and control of bleeding in patients with arteriovenous fistula infections. Routine removal or ligation of a vascular access site that was a likely source of sepsis was not mandatory in our patients even when complicated by endocarditis or septic emboli. This experience, along with other reports of successful treatment of arteriovenous fistula infections with antibiotics alone (14, 16, 18, 35) and the need to preserve vascular access for hemodialysis, is in contrast to the recent suggestion of routine vascular access site removal in hemodialysis patients with endocarditis (23). ACKNOWLEDGMENTS: The authors thank Philip Lief, M.D., of the Renal Division for review of the manuscript and helpful suggestions; Robert Soberman, M.D., Edward Weinstein, M.D., and Vivian Tellis, M.B., for permission to study their hemodialysis records; and Ms. Ronni Jeser for secretarial support. Grant support: During the course of this study Dr. Dobkin was supported by U.S. Public Health Service Training Grant T01-AI00405-05 from the National Institute of Allergy and Infectious Diseases. This paper was presented in part at the 16th Interscience Conference on Antimicrobial Agents and Chemotherapy, Chicago, Illinois, 1976. Received 17 March 1977; revision accepted 29 August 1977. • Requests for reprints should be addressed to Neal H. Steigbigel, M.D.; Division of Infectious Diseases, Montefiore Hospital and Medical Center, 111 East 210th St.; Bronx, NY 10467
References 1. LUNDING M, STEINESS I, THAYSEN JH: Acute renal failure due to tubular necrosis: immediate prognosis and complications. Acta Med Scan d 176:103-119, 1964 2. M A H E R JF, SCHREINER GE: Cause of death in acute renal failure. Arch Intern Med 110:493-504, 1962 3. M O N T G O M E R I E JZ, K A L M A N S O N GM, G U Z E LB: Renal failure and
infection. Medicine (Baltimore) 47:1-32, 1968 4. SIDDIQUI JY, F I T Z AE, L A W T O N RL, K I R K E N D A L L WM: Causes of
death in patients receiving long-term hemodialysis. JAMA 1354, 1970
212:1350-
5. T A P I A HR, H O L L E Y KE, W O O D S JE, JOHNSON WJ: Causes of death
after renal transplantation. Arch Intern Med 131:204-210, 1973 6. FAROOKI MS, G E R G E L Y N F , M C A N I N C H LN, C O A T E S RK, K I M B E R
RW: External arteriovenous shunts for intermittent long-term hemodialysis: five years' experience. Can Med Assoc J 103:1371-1374, 1970 7. F O R A N R F , G O L D I N G AL, T R E I M A N
RL, D E P A L M A JR: Quinton-
Scribner cannulas for hemodialysis. Calif Med 112:8-13, 1970 8. K A S L O W RA, ZELLNER SR: Infection in patients on maintenance haemodialysis. Lancet 2:117-118, 1972 9. BRESCIA MJ, C I M I N O JE, A P P E L K, H U R W I C H BJ: Chronic hemodialy-
sis using venipuncture and a surgically created arteriovenous fistula. TV Engl J Med 275:1089-1092, 1966 10. N O L P H KD: External shunts and internal fistulas (editorial). Ann Intern Med 74:1008-1009, 1971 11. K U R U V I L A KC, BEVEN EG: Arteriovenous shunts and fistulas for hemodialysis. Surg Clin North Am 51:1219-1234, 1971 12.
B A I L L O D RA, K N I G H T AH, C R O C K E T T RE, N A I S H PF: Comparative
assessment of arteriovenous shunts and Cimino-Brescia fistulae. Proc Eur Dial Transplant Assoc 6:65-72, 1969 13. B Y R N E JP, S T E V E N S LE, W E A V E R D H , M A X W E L L JG, R E E M T S M A K:
Advantages of surgical arteriovenous fistulas for hemodialysis. Surg 102:359-362, 1971 14.
Arch
R A L S T O N AJ, H A R L O W GR, J O N E S DM, D A V I S P: Infections of Scrib-
ner and Brescia arteriovenous shunts. Br Med J 3:408-409, 1971 15. SHALDON S: Infection risks of haemodialysis. Br Med J 3:614-615, 1968 16.
M E Y R I E R A, C H E V E T D, M A R S A C J, L E R O U X - R O B E R T C, SRAER J D ,
January 1978 • Annals of Internal Medicine • Volume 88 • Number 1
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19529/ by a University of California San Diego User on 06/12/2017
SCETBON V: Staphylococcemies chez les malades hemodialyses. Nouv PresseMed 2:2379-2383, 1973 17. FOISSAC-GEGOUX P, D U M O N T A: Septicemics au cours du traitement par hemodialyse periodique. Nouv Presse Med 3:151, 1974 18. LEVI J, ROBSON M, ROSENFELD JB: Septicaemia and pulmonary embolism complicating use of arteriovenous fistula in maintenance haemodialysis. Lancet 2:288-290, 1970 19. D A T H A N JRE, F R A N K E L RJ, G O O D W I N FJ, M A R S H FP, M U R R A Y M,
26. P A Y N E JE, C H A T T E R J E E SN, BARBOUR BH, B E R N E TV: Vascular ac-
cess for chronic hemodialysis using modified bovine arterial graft arteriovenous fistula. Am J Surg 128:54-57, 1974 27. A N D E R S O N RJ, SCHAFER LA, O L I N DB, E I C K H O F F TC: Septicemia in
renal transplant recipients. Arch Surg 106:692-694, 1973 28. K Y R I A K I D E S GK, SIMMONS RL, N A J A R I A N JS: Wound infection in
renal transplant wounds: pathogenetic and prognostic factors. Ann Surg 182:770-775, 1975
THOMPSON JMA, YOUNGS GR: One year's experience in a ministry of health dialysis centre. Br Med J 2:102-105, 1970
29. M C D O N A L D JC, R I T C H E Y RJ, F U S E L I E R PF, LINDSEY ES, M C C R A C K -
M A R T I N AM, C L U N I E GJA, T O N K I N RW, ROBSON JS: The aetiology
30. M U R P H Y JF, M C D O N A L D F D , D A W S O N M, R E I T E A, T U R C O T T E J,
and management of shunt infections in patients on intermittent haemodialysis. Proc Eur Dial Transplant Assoc 4:67-72, 1967
FEKETY R: Factors affecting the frequency of infection in renal transplant recipients. Arch Intern Med 136:670-677, 1976
21. O Z E R A N RS, G R A L T, SOKOL A, G O R D O N HE: Long-term experience
31. R I F K I N D D, M A R C H I O R O TL, W A D D E L L WR, STARZL TE: Infectious
20.
with arteriovenous shunts for hemodialysis. Am Surg 38:259-267, 1972 22.
L U M L E Y JSP, C A T T E L L WR, B A K E R LRI: Access to the circulation for
EN BH: Sepsis in human renal transplantation. Surgery 69:189-193, 1971
diseases associated with renal homotransplantation. JAMA 407, 1964
189:397-
regular haemodialysis. Lancet 1:510-513, 1973 23. CROSS AS, STEIGBIGEL RT: Infective endocarditis and access site infections in patients on hemodialysis. Medicine (Baltimore) 55:453-466, 1976
32. G O O D W I N NJ, CASTRONUOVO JJ, F R I E D M A N EA: Recurrent septic
24. C H I N I T Z JL, YOKOYAMA T, B O W E R R, S W A R T Z C: Self-sealing pros-
ence with arteriovenous fistulas and bypasses for hemodialysis. Surgery 76:1018-1023, 1974
thesis for arteriovenous fistula in man. Trans Am Soc Artif Intern Organs 18:452-455, 1972 25. M E R I C K E L JH,
A N D E R S O N RC,
KNUTSON
R,
L I P S C H U L T Z ML,
HITCHCOCK CR: Bovine carotid artery shunts in vascular access surgery. Arch Surg 109:245-250, 1974
pulmonary embolization complicating maintenance hemodialysis. Ann Intern Med 71:29-38, 1969 33. Z E R B I N O VR, T I C E DA, K A T Z LA, N I D U S BD: A 6 year clinical experi-
34. C H E E K RC, MESSINA JJ, A C C H I A R D O SR, B R I T T LG: Arteriovenous
fistulas for hemodialysis: experience with 100 cases. Am Surg 42:386389, 1976 35. SHERRARD DJ: Infections in haemodialysis patients. Lancet 2:880, 1970
Dobkinetal.
• Septicemia and Hemodialysis
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19529/ by a University of California San Diego User on 06/12/2017
33
Bacterial sepsis, a major complication of chronic hemodialysis, is due mainly to infections of the vascular access site despite increasing use of internal fistulas. Sixty episodes of septicemia occurred in two chronic dialysis centers, with an incidence of 0.15 episodes of significant bacteremia per patient-dialysis-year in each. Forty-four of the 6 0 episodes were judged to be due to vascular access site infection by clinical, bacteriologic, and histologic criteria. Seventy percent ( 3 1 of 4 4 ) of the vascular access siterelated episodes were due to staphylococci and 2 5 % ( 1 1 of 4 4 ) to Gram-negative bacilli; nonvascular access site-related episodes were often due to transplant site infections caused by Gram-negative bacilli or streptococci. Mortality was about 1 8 % in both vascular access site-related and nonrelated septic episodes. Bovine heterograft arteriovenous fistulas more often led to sepsis than did Brescia arteriovenous fistulas. Treatment with appropriate antibiotics was successful in most cases. Routine removal or ligation of the vascular access site was not necessary.
I N F E C T I O N accounts for much of the morbidity and mortality associated with acute and chronic renal failure ( 1 3). Hemodialysis and renal transplantation, which have been successful in prolonging the lives of patients with chronic renal insufficiency, are often complicated by infection leading to significant mortality (4, 5). The first successful vascular access for chronic dialysis, the Quinton-Scribner external shunt, was frequently associated with complications such as bleeding, thrombosis, and local and systemic infection (6-8). The internal arteriovenous fistula was introduced by Brescia and colleagues (9) with the expectation of fewer complications (10); comparisons of internal fistulas with external shunts suggested that fistulas had greater longevity and fewer local complications (11-15). Therefore, the internal fistula has become the predominant vascular access site in the chronic hemodialysis population. Reports describing bacteremic infection in chronic hemodialysis patients subsequent to the introduction of internal arteriovenous fistulas are few and have involved only a small number of patients (14, 16-18). However, in view of a substantial experience at our institutions with vascular access site-related sepsis in a large population of chronic hemodialysis • F r o m the Division of Infectious Diseases, D e p a r t m e n t of Medicine, Montefiore Hospital a n d Medical Center, and t h e D e p a r t m e n t of Medicine, Albert Einstein College of Medicine; Bronx, N e w York. 28
© 1 9 7 8 American College of Physicians
p a t i e n t s , m a i n l y u s i n g i n t e r n a l fistulas, w e h a v e u n d e r t a k e n a r e t r o s p e c t i v e s t u d y of t h e e p i d e m i o l o g y , b a c t e r i o l o g y , clinical c h a r a c t e r i s t i c s , a n d p r o g n o s i s of t h i s e n t i ty. V a s c u l a r a c c e s s site-related sepsis w a s c o m p a r e d t o s e p t i c e p i s o d e s a r i s i n g from o t h e r foci in t h e s a m e h e m o dialysis p o p u l a t i o n . Materials and Methods Records of all patients with positive blood cultures hemodialyzed from 1972 to 1975 at the chronic dialysis facilities (Center A and Center B) affiliated with Montefiore Hospital and the Albert Einstein College of Medicine were reviewed. T h e two centers have separate patient populations and medical staffs. Positive cultures were considered to be the result of contamination if one bottle of a set yielded a typically nonpathogenic organism and if chart review gave no indication of an infectious process and the patient recovered without antimicrobial therapy. All bacteremias not excluded by these criteria were considered significant and consisted of 60 episodes in 43 patients. All of these episodes were associated with signs and symptoms of sepsis, and all but two had at least two positive blood cultures. The vascular access site was an internal fistula in 38 patients and an external shunt in four while one patient had both. Of the 39 patients with internal fistulas, 30 were Brescia arteriovenous fistulas and nine were bovine heterografts. The 60 episodes of significant bacteremia were classified by chart analysis as definitely, probably, or possibly vascular access site-related, or as nonvascular access site-related. T h e following criteria were used. A. Vascular access site-related (44 episodes) 1. Definite (16 episodes) Histologic or bacteriologic proof of local infection at the vascular access site along with signs of infection at the vascular access site and without another apparent primary source of bacteremia. 2. Probable (13 episodes) No apparent primary source for the bacteremia other than the vascular access site, and either local signs of infection at the vascular access site (without histologic or bacteriologic proof) (12 episodes) or signs of sepsis developing in close temporal relation to manipulation of the vascular access site at the time of dialysis (one episode). 3. Possible (15 episodes) No apparent source for the bacteremia other than the vascular access site. B. Nonvascular access site-related (16 episodes) No signs of infection at the vascular access site; evident focus of infection at another site. Signs taken to signify local infection at the vascular access site included purulent discharge, fluctuance, or severe tenderness. Erythema, warmth, and induration, which are signs often found normally at the vascular access site, were not considered signs of local vascular access site infection. In 32 of 44 episodes classified as vascular access site-related septicemia, there was evidence of prolonged bacteremia, suggesting endovascular infection: 24 episodes with multiple positive blood cultures four or more h apart; and eight episodes with multiple positive blood cultures (at least six bottles posiAnnals of Internal Medicine 8 8 : 2 8 - 3 3 , 1 9 7 8
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19529/ by a University of California San Diego User on 06/12/2017
Table 1 . Incidence of Septicemia in Chronic Hemodialysis Patients
Center
Patients
Dialyses
178 72 250
24 856 7629 32 485
Bacteremic Episodes
Vascular Access Site-Related Episodes
32 11 43
23 (72) 9(82) 32 (74)
-> %~
no. A*
Bt
Total
Risk of Bacteremic Infection per PatientDialysis-Year 0.15 0.15 0.15
* Twenty two-month period (April 1973-January 1975). t Twenty-month period (August 1973-March 1975).
tive) without recorded times. Incidence data were compiled for a representative period at each center during which complete information was available. The dialysis record for each patient during this period was reviewed. At Center A this period covered 22 months (April 1973-January 1975) and at Center B 20 months (August 1973March 1975). A chronic dialysis patient was defined as one hemodialyzed at least once a week for 3 weeks in at least 2 months. One patient-dialysis-month was defined as dialysis of one chronic hemodialysis patient in 3 different weeks in a given month. A patient-dialysis-year was defined as 12 patient-dialysis-months. Data were analyzed for statistical significance by the chisquare, Fisher's exact, or Student's t test. Results INCIDENCE AND CLINICAL SETTING
Forty-three of the 60 episodes of significant bacteremia took place during the periods studied for incidence (Table 1). Incidence was identical in the two independent populations studied: 0.15 episodes of bacteremia per patient-dialysis-year. At Center A 13.5% of patients and at Center B 12.5% had one or more bacteremias during the period of analysis. About 7 5 % of the episodes of septicemia were vascular access site-related (0.11 episodes per patient-dialysis-year). Many patients on chronic hemodialysis had several revisions and replacements of their vascular access sites due to malfunctioning, bleeding, infection, or other reasons. However, 4 5 % (20 of 44) of vascular access site-related septic episodes occurred in patients who were still being dialyzed through their original vascular access site. Bovine heterograft fistulas were surgically placed for only a limited period during the study. Of the 67 bovine grafts placed, eight were associated with sepsis. One patient with sepsis who had a bovine heterograft placed at another institution was excluded from this analysis. During this same period, 236 Brescia fistulas were placed, eight of which were associated with sepsis. The greater incidence of sepsis with bovine heterografts was significant (P < 0.05). This occurred despite a shorter period of chronic hemodialysis before infection for patients with bovine heterografts (mean duration, 9.5 months) compared to those with Brescia arteriovenous fistulas (mean duration, 17.8 months). The mean duration of time between the construction of the most recent vascular access and the onset of infection was significantly less in patients with bovine heterografts (1.6 months) than in patients with Brescia fistulas (15.0 months), P < 0.01. Septic episodes occurred throughout the period of
study at both centers without clustering suggestive of epidemic spread. Staphylococcal phage typing or antibiotic susceptibility data available for 20 episodes in 18 patients at Center A showed each bacterial isolate to be a different strain. In four of five patients who had multiple episodes of infection, different species were involved in each episode. The fifth patient had four episodes of Staphylococcus aureus sepsis; two of the organisms were phage-typed and were different. Ten of 16 nonvascular access site-related episodes were attributed to a genitourinary source, five to a pulmonary source, and one to a soft tissue infection. Six of the 10 episodes related to a genitourinary source involved patients with perinephric abscesses or nephrectomy-site abscesses after unsuccessful renal transplantation. Seven of the 16 episodes of nonvascular access site-related sepsis involved patients with recent renal transplantation who were being treated with immunosuppressive therapy. Five of the six perinephric abscesses occurred in the latter group. In contrast, only three of 44 vascular access siterelated episodes took place in patients receiving immunosuppressive therapy. Patients with vascular access site-related sepsis did not differ significantly from those with nonvascular access site-related sepsis in age, sex, type of renal disease, or presence of other underlying chronic disease. Similarly these factors did not differ significantly between the septic patients with bovine heterografts and those with Brescia arteriovenous fistulas. BACTERIOLOGY A N D O T H E R L A B O R A T O R Y D A T A
Vascular access site-related cases were predominantly due to staphylococci (31 of 44 episodes, or 70%), while septicemias unrelated to the vascular access site were mostly due to streptococci or Gram-negative bacilli (13 of 16 episodes, or 81%) (Table 2). Staphylococcus aureus was a significantly more common pathogen in the groups classified as definitely or probably related to the vascular access site (21 of 29 episodes) than in the possible group (five of 15 episodes), P < 0.05, or the nonvascular access site-related group (two of 16 episodes), P < 0.005 (Table 2). Of the six episodes of vascular access site-related sepsis that occurred in three heroin users, five were due to 5. aureus. Among the Gram-positive cocci other than staphylococci, enterococci were isolated in three episodes. Among the Gram-negative bacilli other than Enterobacteriaceae, Pseudomonas species were isolated in five episodes (four Dobkinetal.
• Septicemia and Hemodialysis
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19529/ by a University of California San Diego User on 06/12/2017
29
Table 2. Bacterial Isolates in 60 Episodes of Septicemia
Bacterial Isolate
Relation of 1Infection to the Vascular Access Site Definite or Probable no.
Staphylococcus aureus Staphylococcus epidermidis Other Grampositive cocci Enterobacteriaceae Other Gramnegative rods
%
21 (70.0)
Possible no.
%
Nonvascula Access Site Related
no.
%
5 (31.2)
2 (12.5)
2
(6.7)
3* (18.8)
1 (6.2)
2 2t
(6.7) (6.7)
1 (6.2) 5* (31.2)
7 (43.8) 3 (18.8)
3 f (10.0)
2 (12.5)
3 (18.8)
* Includes one mixed bacteremia: Proteus mirabilis and S. epidermidis. t Includes one mixed bacteremia: Enterobacter and Pseudomonas species.
were vascular access site-related), Bacteroides species in two episodes, and Hemophilus influenzae in one episode. Purulent material obtained from the vascular access site in 15 episodes of vascular access site-related sepsis yielded the same organism on culture as that obtained from blood cultures. Gram stain of purulent material accurately predicted the class of organisms obtained by culture. Total and differential leukocyte counts at the onset of sepsis were normal in most cases, and there was no significant difference between the vascular access site-related and nonrelated groups in this regard. SIGNS A N D SYMPTOMS IN VASCULAR ACCESS SITER E L A T E D EPISODES
About 8 0 % of episodes in both the vascular access siterelated and nonrelated groups had accompanying fever (38 °C or more) at the time of presentation. Rigors occurred in 4 5 % (13 of 29) of those cases in the definite and probable groups, but in none of the 15 cases classified as possibly vascular access site-related {P < 0.01) and none of those cases considered unrelated to the vascular access site. Significant local signs and symptoms of infection at the vascular access site were more common with staphylococcal infections than with those due to organisms other than staphylococci. Severe pain was noted in eight of the 30 cases due to staphylococci alone and in none of 13 nonstaphylococcal cases, P < 0.05. Fifty-three percent (16 of 30) of staphylococcal episodes were characterized by at least one of the following findings: severe pain, purulent discharge, skin necrosis, or abscesses. Twentythree percent (three of 13) of the nonstaphylococcal cases had one or more of these findings. The physical finding of a purulent discharge was more common in the small number of patients with external shunts (three of five) than in those patients with arteriovenous fistulas (five of 39). In two patients, development of vascular access site-related septicemia followed the surgical removal or revision of a locally infected arteriovenous fistula. Changing signs, symptoms, and laboratory data in each case suggested 30
that manipulation had caused dissemination of a previously localized process. METASTATIC INFECTION
Sixteen metastatic infections occurred in 13 of 44 vascular access site-related episodes: septic pulmonary emboli, six; endocarditis, four; meningitis, two; metastatic urinary tract infection, two; empyema, one; and myocardial abscesses, one. One of the patients with endocarditis and the patient with myocardial abscesses were detected only at autopsy. Fourteen of the 16 metastatic infections were due to S. aureus. Metastatic infections were much more frequent in staphylococcal vascular access site-related sepsis, occurring in 10 of 31 episodes compared to one of 13 episodes in nonstaphylococcal sepsis. The four cases of endocarditis among those with vascular access site-related sepsis were all associated with 5. aureus. Among the episodes of vascular access site-unrelated sepsis, one patient had endocarditis due to S. aureus. Evidence of endocarditis diagnosed during life included the development of acute aortic insufficiency in all four patients. Septic pulmonary emboli were detected in six of 44 vascular access site-related cases (14%). Repeated episodes of septic pulmonary emboli were not observed. T R E A T M E N T A N D COURSE
All but three of 60 episodes were treated with appropriate antibiotics (based on culture and sensitivity data) within 48 h of the onset of signs and symptoms of sepsis. Most patients received 4 to 6 weeks of antibiotic treatment. Data regarding antibiotic therapy are not adequate to allow comparison of different regimens. Defervescence occurred within 3 days of the start of therapy in 7 1 % of cases. Surgery was done on the vascular access site in the management of eight episodes (seven in internal fistulas) of vascular access site-related sepsis. Clinical indications for surgery included appearance of an abscess, thrombosis, or bleeding. The findings at surgery were abscesses in five, thrombosis in five, and pseudoaneurysm in four. There was one death among these eight episodes. None of the four patients with clinical evidence of endocarditis had surgery performed on the vascular access site. Only one of the six patients with septic pulmonary emboli underwent surgery of the vascular access site, for thrombectomy. There was no evidence that continued hemodialysis through an infected arteriovenous fistula predisposed patients to septic pulmonary embolization. Several patients with vascular access site-related sepsis were dialyzed through their arteriovenous fistulas without evident complications. Three of the six cases of septic pulmonary emboli occurred in patients who were not being dialyzed through the infected fistula at the time of embolization, but who were being dialyzed through another vascular access site or femoral vein puncture. MORTALITY
The case fatality rate for both the vascular access site-
January 1978 • Annals of Internal Medicine • Volume 88 • Number 1
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19529/ by a University of California San Diego User on 06/12/2017
One (17) reported 0.18 episodes per patient-dialysis-year for all bacteremic events, while the other (16) noted 0.20 episodes per patient-dialysis-year for vascular access siterelated septicemias. Therefore, although local infections appear to occur less frequently with internal arteriovenous fistulas than with external vascular shunts, the incidence of septicemia in patients with these two types of vascular access sites is similar. Epidemiologic findings, including phage typing, antibiotic sensitivity data, and the lack of case clustering, suggest autoinfection in the pathogenesis of vascular access site-related sepsis in our series. The same route of spread was found in a prospective study of local infections of external shunts (20). Sepsis occurred earlier and with greater frequency in our patients with bovine heterograft arteriovenous fistulas than in those with Brescia fistulas, despite a significantly shorter history of chronic hemodialysis in the former group. Bovine grafts have been recommended as the alternative vascular access for patients with vessels inadequate for the Brescia fistula (24). The higher incidence of septicemia with bovine heterograft fistulas has not been previously reported. Along with observations by others (25, 26) that local infection may be common with bovine heterografts, our findings prompt caution against unnecessary use of these prostheses. Our report shows that vascular access site-related sepsis in patients with arteriovenous fistulas is predominantly caused by staphylococci; however, an appreciable number of cases were due to enteric Gram-negative bacilli and Pseudomonas species. In contrast, 29 of the 31 cases of sepsis with arteriovenous fistulas reported in three studies (14, 16, 18) were due to staphylococci. Local infection and septicemia in patients with external shunts are also predominantly due to staphylococci, although Pseudomonas species have been frequently reported as well (8, 11). Sepsis unrelated to the vascular access site occurred most often as a complication of previous renal transplantation. Sepsis in these patients arose particularly from surgical wound infection, nephrectomy site abscess, or urinary tract infection and was due predominantly to Gram-negative rods and streptococci. This clinical and bacteriologic pattern has been noted in several major reviews of infection in transplant patients (27-30). The correlation of sepsis with renal transplant rejection has been documented, although as yet the possible predisposing role of increased immunosuppressive therapy has not
related and nonvascular access site-related episodes was about 18% (Table 3). Mortality for vascular access siterelated cases due to Gram-negative bacilli (five of 10) was significantly greater than that for staphylococci (three of 30), P < 0.05 (Table 3). Increased age (60 years or older) carried a fourfold increased death rate in vascular access site-related cases and a ninefold increase in nonvascular access site-related cases. Sepsis in patients dialyzed for 12 or more months had a mortality rate twice that of those dialyzed for less than 12 months. Mortality rates in vascular access site-related episodes with metastatic complications (three of 13) or in those with surgical procedures performed on the vascular access site (one of eight) were not significantly different from episodes without such complications (five of 31) or procedures (seven of 36). None of the four patients with clinical evidence of endocarditis died. Discussion
Published studies of the epidemiology of infection in chronic hemodialysis patients have emphasized local infection at the vascular access site. Reported rates of local infection in external shunts range from about one to three episodes per patient-dialysis-year (6, 7, 14, 19-21). Reported rates of local infection with arteriovenous fistulas are lower, occurring at 0.15 episodes per-patient-dialysisyear or less (14, 22). However, in our experience bacteremic infection continues to be a major complication of chronic renal failure despite the use of maintenance hemodialysis and the development of the internal fistula to replace the external shunt for vascular access. Clinical reports of disseminated sepsis and endocarditis in hemodialysis patients are primarily of patients with external shunts (23). Septicemia secondary to local vascular access site infection in patients with external shunts has been reported at rates from 0.10 to 0.30 episodes per patientdialysis-year (7, 14, 21). In our study of patients on hemodialysis mainly with arteriovenous fistulas for vascular access, sepsis occurred at a rate of 0.15 episodes per patient-dialysis-year at each of two centers. Two clinical entities accounted for most of the septic episodes: bacteremic vascular access site infections, occurring at a rate of 0.11 episodes per patient-dialysis-year; and bacteremic transplant site or genitourinary tract infections in patients with rejected renal transplants. The incidence figures in our study are similar to those of two French reports (16, 17) involving a total of 34 episodes of sepsis in patients with arteriovenous fistulas. Table 3. Mortality of Septicemia in Chronic Hemodialysis Patients
Mortality by Bacterial Cause Relation of Septicemia to the Vascular Access Site
Overall Mortality
Definite or probable Possible Total vascular access site-related Vascular access site-unrelated Total
no. % 3/29 (10) 5/15 (33) 8/44 (18) 3/16(19) 11/60(18)
Staphylococci no. 2/23 1/7 3/30 0/3 3/33
% (9) (14)* (10) (9)
Gram-Negative Rods no. 1/4 4/6 5/10 0/6 5/16
% (25) (67)* (50) (31)
Other Organisms no. % 0/2 0/1 0/3 3/7 (43) 3/10 (30)
'• Nonfatal case with mixed staphylococcal and Gram-negative bacteremia excluded. Dobkin et al. • Septicemia and Hemodialysis
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19529/ by a University of California San Diego User on 06/12/2017
31
been definitely established (30,31). Our experience indicates that the clinical entity of vascular access site-related sepsis with the internal fistula differs from that with the external shunt in terms of local signs, clinical course, and complications. Local inflammation with purulent discharge, pseudoaneurysm formation, thrombosis of the shunt, bleeding around the shunt, and even extrusion of the cannula are frequent signs in external shunt infection (6, 7). In contrast, nearly one third of our cases of vascular access site-related sepsis had no local findings distinguishable from those reflecting the usual inflammation accompanying regular use of the fistula. Severe pain and tenderness were the most prominent local findings with septicemic arteriovenous fistula infection. Local signs of inflammation at the vascular access site were most prominent with staphylococcal vascular access site-related sepsis compared to sepsis due to other organisms. Our 15 episodes classified as possibly vascular access site-related had several characteristics pointing to the vascular access site as the source despite absence of local findings: there were no other evident primary sources for sepsis in these patients; prolonged bacteremia in most patients indicated endovascular infection, while the absence of findings of endocarditis combined with the frequent Gram-negative bacteriology made the diagnosis of bacterial endocarditis unlikely; and the group's high rate of mortality and prolonged bacteremia are inconsistent with contamination or transient bacteremia. Although bacterial endocarditis or transient bacteremia cannot be ruled out in all of these cases, there is strong evidence for occult vascular access site infection in most of them. Septicemia due to clinically inapparent infection of the arteriovenous fistula has been suggested by Meyrier and colleagues (16) in nine of 22 episodes of staphylococcal septicemia in patients with arteriovenous fistulas and in six of seven cases of arteriovenous fistula-related sepsis reported by Levi, Robson, and Rosenfeld (18), although data on the exclusion of other infected foci, and the duration of bacteremia, are not given. Metastatic complications associated with vascular access site-related sepsis strongly correlate with infection by 5. aureus, which was responsible for 14 of 16 incidents in our series. Septic pulmonary emboli have been noted as sporadic or recurrent complications of vascular access site infection with both external shunts and arteriovenous fistulas (16, 18, 32) and may occur without local signs of infection at the vascular access site. Removal of an external shunt is frequently necessary to prevent recurrent septic emboli (32), but this was not necessary in our patients with arteriovenous fistula infections or in those reported by others (16, 18). Bacterial endocarditis, diagnosed in five of our patients, has been associated with chronic hemodialysis and is usually staphylococcal or streptococcal in origin (23). Endocarditis must be considered as a diagnostic possibility in septic hemodialysis patients, especially those with prolonged bacteremia. However, the absence of the usual clinical findings of bacterial endocarditis or the isolation of Gram-negative organisms should suggest vascular ac32
cess site infection as a stronger possibility. The concurrent presence of subclinical endocarditis and infection of the vascular access site cannot be ruled out in patients who have staphylococcal bacteremia and strong evidence of infection of the vascular access site. Empiric antimicrobial therapy of sepsis of vascular access site origin should include coverage for S. aureus and Gram-negative bacilli. Our experience, and that of others (33, 34), suggests that surgical treatment is indicated for drainage of abscesses or hematomas, resection of aneurysms, and control of bleeding in patients with arteriovenous fistula infections. Routine removal or ligation of a vascular access site that was a likely source of sepsis was not mandatory in our patients even when complicated by endocarditis or septic emboli. This experience, along with other reports of successful treatment of arteriovenous fistula infections with antibiotics alone (14, 16, 18, 35) and the need to preserve vascular access for hemodialysis, is in contrast to the recent suggestion of routine vascular access site removal in hemodialysis patients with endocarditis (23). ACKNOWLEDGMENTS: The authors thank Philip Lief, M.D., of the Renal Division for review of the manuscript and helpful suggestions; Robert Soberman, M.D., Edward Weinstein, M.D., and Vivian Tellis, M.B., for permission to study their hemodialysis records; and Ms. Ronni Jeser for secretarial support. Grant support: During the course of this study Dr. Dobkin was supported by U.S. Public Health Service Training Grant T01-AI00405-05 from the National Institute of Allergy and Infectious Diseases. This paper was presented in part at the 16th Interscience Conference on Antimicrobial Agents and Chemotherapy, Chicago, Illinois, 1976. Received 17 March 1977; revision accepted 29 August 1977. • Requests for reprints should be addressed to Neal H. Steigbigel, M.D.; Division of Infectious Diseases, Montefiore Hospital and Medical Center, 111 East 210th St.; Bronx, NY 10467
References 1. LUNDING M, STEINESS I, THAYSEN JH: Acute renal failure due to tubular necrosis: immediate prognosis and complications. Acta Med Scan d 176:103-119, 1964 2. M A H E R JF, SCHREINER GE: Cause of death in acute renal failure. Arch Intern Med 110:493-504, 1962 3. M O N T G O M E R I E JZ, K A L M A N S O N GM, G U Z E LB: Renal failure and
infection. Medicine (Baltimore) 47:1-32, 1968 4. SIDDIQUI JY, F I T Z AE, L A W T O N RL, K I R K E N D A L L WM: Causes of
death in patients receiving long-term hemodialysis. JAMA 1354, 1970
212:1350-
5. T A P I A HR, H O L L E Y KE, W O O D S JE, JOHNSON WJ: Causes of death
after renal transplantation. Arch Intern Med 131:204-210, 1973 6. FAROOKI MS, G E R G E L Y N F , M C A N I N C H LN, C O A T E S RK, K I M B E R
RW: External arteriovenous shunts for intermittent long-term hemodialysis: five years' experience. Can Med Assoc J 103:1371-1374, 1970 7. F O R A N R F , G O L D I N G AL, T R E I M A N
RL, D E P A L M A JR: Quinton-
Scribner cannulas for hemodialysis. Calif Med 112:8-13, 1970 8. K A S L O W RA, ZELLNER SR: Infection in patients on maintenance haemodialysis. Lancet 2:117-118, 1972 9. BRESCIA MJ, C I M I N O JE, A P P E L K, H U R W I C H BJ: Chronic hemodialy-
sis using venipuncture and a surgically created arteriovenous fistula. TV Engl J Med 275:1089-1092, 1966 10. N O L P H KD: External shunts and internal fistulas (editorial). Ann Intern Med 74:1008-1009, 1971 11. K U R U V I L A KC, BEVEN EG: Arteriovenous shunts and fistulas for hemodialysis. Surg Clin North Am 51:1219-1234, 1971 12.
B A I L L O D RA, K N I G H T AH, C R O C K E T T RE, N A I S H PF: Comparative
assessment of arteriovenous shunts and Cimino-Brescia fistulae. Proc Eur Dial Transplant Assoc 6:65-72, 1969 13. B Y R N E JP, S T E V E N S LE, W E A V E R D H , M A X W E L L JG, R E E M T S M A K:
Advantages of surgical arteriovenous fistulas for hemodialysis. Surg 102:359-362, 1971 14.
Arch
R A L S T O N AJ, H A R L O W GR, J O N E S DM, D A V I S P: Infections of Scrib-
ner and Brescia arteriovenous shunts. Br Med J 3:408-409, 1971 15. SHALDON S: Infection risks of haemodialysis. Br Med J 3:614-615, 1968 16.
M E Y R I E R A, C H E V E T D, M A R S A C J, L E R O U X - R O B E R T C, SRAER J D ,
January 1978 • Annals of Internal Medicine • Volume 88 • Number 1
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19529/ by a University of California San Diego User on 06/12/2017
SCETBON V: Staphylococcemies chez les malades hemodialyses. Nouv PresseMed 2:2379-2383, 1973 17. FOISSAC-GEGOUX P, D U M O N T A: Septicemics au cours du traitement par hemodialyse periodique. Nouv Presse Med 3:151, 1974 18. LEVI J, ROBSON M, ROSENFELD JB: Septicaemia and pulmonary embolism complicating use of arteriovenous fistula in maintenance haemodialysis. Lancet 2:288-290, 1970 19. D A T H A N JRE, F R A N K E L RJ, G O O D W I N FJ, M A R S H FP, M U R R A Y M,
26. P A Y N E JE, C H A T T E R J E E SN, BARBOUR BH, B E R N E TV: Vascular ac-
cess for chronic hemodialysis using modified bovine arterial graft arteriovenous fistula. Am J Surg 128:54-57, 1974 27. A N D E R S O N RJ, SCHAFER LA, O L I N DB, E I C K H O F F TC: Septicemia in
renal transplant recipients. Arch Surg 106:692-694, 1973 28. K Y R I A K I D E S GK, SIMMONS RL, N A J A R I A N JS: Wound infection in
renal transplant wounds: pathogenetic and prognostic factors. Ann Surg 182:770-775, 1975
THOMPSON JMA, YOUNGS GR: One year's experience in a ministry of health dialysis centre. Br Med J 2:102-105, 1970
29. M C D O N A L D JC, R I T C H E Y RJ, F U S E L I E R PF, LINDSEY ES, M C C R A C K -
M A R T I N AM, C L U N I E GJA, T O N K I N RW, ROBSON JS: The aetiology
30. M U R P H Y JF, M C D O N A L D F D , D A W S O N M, R E I T E A, T U R C O T T E J,
and management of shunt infections in patients on intermittent haemodialysis. Proc Eur Dial Transplant Assoc 4:67-72, 1967
FEKETY R: Factors affecting the frequency of infection in renal transplant recipients. Arch Intern Med 136:670-677, 1976
21. O Z E R A N RS, G R A L T, SOKOL A, G O R D O N HE: Long-term experience
31. R I F K I N D D, M A R C H I O R O TL, W A D D E L L WR, STARZL TE: Infectious
20.
with arteriovenous shunts for hemodialysis. Am Surg 38:259-267, 1972 22.
L U M L E Y JSP, C A T T E L L WR, B A K E R LRI: Access to the circulation for
EN BH: Sepsis in human renal transplantation. Surgery 69:189-193, 1971
diseases associated with renal homotransplantation. JAMA 407, 1964
189:397-
regular haemodialysis. Lancet 1:510-513, 1973 23. CROSS AS, STEIGBIGEL RT: Infective endocarditis and access site infections in patients on hemodialysis. Medicine (Baltimore) 55:453-466, 1976
32. G O O D W I N NJ, CASTRONUOVO JJ, F R I E D M A N EA: Recurrent septic
24. C H I N I T Z JL, YOKOYAMA T, B O W E R R, S W A R T Z C: Self-sealing pros-
ence with arteriovenous fistulas and bypasses for hemodialysis. Surgery 76:1018-1023, 1974
thesis for arteriovenous fistula in man. Trans Am Soc Artif Intern Organs 18:452-455, 1972 25. M E R I C K E L JH,
A N D E R S O N RC,
KNUTSON
R,
L I P S C H U L T Z ML,
HITCHCOCK CR: Bovine carotid artery shunts in vascular access surgery. Arch Surg 109:245-250, 1974
pulmonary embolization complicating maintenance hemodialysis. Ann Intern Med 71:29-38, 1969 33. Z E R B I N O VR, T I C E DA, K A T Z LA, N I D U S BD: A 6 year clinical experi-
34. C H E E K RC, MESSINA JJ, A C C H I A R D O SR, B R I T T LG: Arteriovenous
fistulas for hemodialysis: experience with 100 cases. Am Surg 42:386389, 1976 35. SHERRARD DJ: Infections in haemodialysis patients. Lancet 2:880, 1970
Dobkinetal.
• Septicemia and Hemodialysis
Downloaded From: http://annals.org/pdfaccess.ashx?url=/data/journals/aim/19529/ by a University of California San Diego User on 06/12/2017
33
