Staphylococcus aureus Bacteremia in Diabetic Patients
Endocarditis and Mortality
G. COOPER, M.D.
R. PLATT, M.D.
Boston, Massachusetts
From ihe Hospital Epidemiology Unit and De-
partment of Medicine, New England Deaconess
Hospital and Channing Laboratory, Brigham and
Women’s Hospital, Harvard Medical School.
Boston, Massachusetts. Requests for reprints
should be addressed to Dr. Platt, Channing Labo-
ratory, 180 Longwood Avenue, Boston, Massa-
chusetts 02115. Manuscript accepted March 22,
1982.
658 November 1982 The American Journal of Medicine
The presentation and course of Staphylococcus aureus bacteremia
in 27 diabetic patients (18 insulin-dependent) were compared with
those in 34 nondiabetic patients. The groups were comparable in
age, proportion with pre-existing cardiac valvular disease, com-
munity-acquired bacteremia, fever, and leukocytosis. Endocarditis
(vegetation or new regurgitant murmur) was present in eight (30
percent) diabetics and four (12 percent) nondiabetics (p = 0.16).
A primary focus of infection was present in 67 percent of diabetics
and 65 percent of nondiabetics. Among those with a focus, six of 18
diabetics and none of 22 nondiabetics had endocarditis (p <O.OO$).
Fifteen of 54 (28 percent) patients who received appropriate anti-
biotic therapy died. After stratification for underlying illness, there
was no mortality difference between those with and without endo-
carditis (three endocarditis deaths versus 1.78 expected), or be-
tween those with and without diabetes (four diabetic deaths versus
4.8 expected). Diabetics with staphylococcai bacteremia were more
likely than nondiabetics to have endocarditis in the presence of a
primary focus. They had no Increase in mortality.
Clinical studies have suggested that diabetics with staphylococcal
bacteremia and endocarditis have a higher mortality than nondiabetics
[ 1,2]. No investigation has specifically examined the characteristics
of staphylococcal bacteremia in diabetics. We reviewed the course
of staphylococcal bacteremia in diabetic and nondiabetic patients to
compare the rate of endocarditis and factors associated with endo-
carditis and mortality.
METHODS
Patient Population. Records were sought of all patients with Staph. aureus
bacteremia at the New England Deaconess Hospital between August 1977
and August 1980. We excluded patients with polymicrobial bacteremia.
Because the New England Deaconess Hospital is the principal admitting fa-
cility for the Joslin Diabetes Foundation, approximately 40 percent of all in-
patients have diabetes mellitus.
Definition ol Terms. We defined bacteremia as definite if the same organism
was isolated from two or more blood cultures within a 24-hour period. Bac-
teremia was considered probable if there was a single positive culture and
no other was obtained during the same 24-hour period. We considered an
isolate to be a contaminant if only one blood culture showed growth and there
were additional negative culture specimens taken in the same 2-w period,
prior to initiation of specific antibiotic therapy. These criteria are similar to
those that have been used previously [3].
Volume 73
 STAPHYLOCOCCUS AUREUS BACTEREMIA IN DIABETES-COOPER AND PLAIT

TABLE I Characteristics of Bacteremic Patients

Diabetics Nondiabelics
(N = 27) (N = 34)
Endocarditis NoEndocarditis EndocardHls NoErnbcarditis
(N = 8) (N = 19) (N = 4) (N = 30)
Meanage (range) 59 (35-74) 60 (22-87) 58 (43-72) 58 (30-87)
Community-acquired
infection 7 (88). 7 (37) 4 (100) 10 (33)
Insulin use 7 (88) 11 (59) - -
Pre-existing valvular
disease 0 (0) 3 (16) 1 (25) 3 (10)
Fever 8 (100) 19 (100) 4 (100) 30 (100)
Leukocytosis
(>lO,OOO WBC/mm3) 8 (100) 10 (53) 4 (100) 23 (77)
Left shift
(> 10 bands) 8 (100) 19 (100) 3 (75) 23 (77)
Identifiable primary
focus of infection 6 (75) 12 (63) 0 (0) 22 (73)
l Percentages are shown in parentheses.

Bacteremia was classified as community+3cquiredif it was
detected within 72 hours of admission and was the first
bacteremia for the current hospitalization [4]. Endocarditis
was said to be present if (1) two or more blood cultures
showed growth and a new regurgitant heart murmur devel-
oped, or (2) a vegetation was found at surgery or at au-
topsy.
A primary focus of infection was said to be present if
Staph. aureus with an antibiogram identical to that of the
blood isolate was reoovered from a pefipherai site antedating
the onset of bacteremia.
Resufb of Therapy. Therapy was considered appropriate
if an intravenous antibiotic was administered to which the
Staph. aweus was susceptible by antimicrobial susceptibility
testing.
Patients were stratified on the basis of severity of under-
lying m. Patiis were classified as having rapidly fatal,
ultimately fatal, or nonfatal disease using criteria formulated
by McCabe and Jackson (51. Diabettcs without other serious
medical problems were said to have nonfatal disease.
StatIstIca. Comparisonsbetween simple proportions were
made using the Yates’ corrected chi-square test, or Fisher’s
exact test, when appropriate [6]. Confidence limits for pro-
portions were calculated by standard methods [6]. Stratified
comparisons were performed by the log-rank method [7].
The rxdichce limits of difference between proportions were
calculated by the method of Fleiss [8], and the confidence
limits for the relative risk were derived by the delta
method.
RESULTS
Patient Population. Sixty-nine of 92 patients with
Staph. aureus blood isolates met the inclusion criteria
noted earlier. All 69 had definite bacteremia. Twenty-
three patients were excluded from analysis, six with
polymicrobiai bacteremia and 17 in whom isolates were
defined as contaminants. The charts of 6 I(88 percent)
were available for review. Twenty-seven patients had
diabetes and 34 patients did not. The ciini/cal features
of the patients are summarized in Table I.1Eighteen of
the 27 diabetic patients received daify insulin injections.
At the time of bacteremia, all diabetics Were hyper-
glycemic on their usual diabetic therapy (blood sugar
range 190 to 490 mg/di). None of the pati&tts had ke-
toacidosis or hyperosmoiar coma. There ‘were no in-
travenous drug abusers among the study; population.
Diabetic and nondiabetic patients did not differ in age,
proportion with community-acquired infection, pre-
existing vaivuiar disease, primary focus of infection,
fever, leukocytosis, or presence of a left shift in pe-
ripheral leukocytes. Sixty (98 percent) of the 61 patients
studied had an underlying disease antding the ac-
quisition of staphylococcal infection. Nond/abetics had
a greater proportion of rapidly or ultimatdly fatal dis-
eases than diabetics (47 percent versus 22 percent, p
= 0.08). Among the nondiabetics, 12 patients had
metastatic cancer, two had hepatic faiture, and two had
cardiogenic shock. Among the diabetics, tht-ee patients
had rnetastatic cancer, two had hepatic failtie, and two
had cardiogenic shock.
Twelve patients (20 percent) met the ‘criteria for
endocarditis. Seven had a clinical diagnosis, whereas
five had a tissue diagnosis. Ail patients had left-sided
endocarditis. Endocarditis was present in eight diabetics
(30 percent) and four nondiabetics (12 pei-cent). This
difference was not statistically significant (p = 0.16).
The relative risk of endocarditis in diabeitics versus
nondiabetics was 2.5; the 95 percent confidence limits
were 0.8 to 7.5.
Among the diabetic patients with endocarditis, seven
of eight (88 percent) acquired the infection in the
community. Only seven of 19 (39 percent) without en-

November 1982 The American Journal of Medlcim Volumr 73 859

STAPHYLOCOCCUS AUREUS BACTEREMIA IN DIABETES-COOPER AND PLATT

TABLE II Primary FOCIof Staphylococcal Infection

Diabetics Nondiabetics
Endocardltis No Endocarditis Endocarditis No Endocarditis

Primary focus
present 6 12 0 22
Foot ulcer 4 1 - 2
Intravascular catheter 1 4 - a
Surgical wound - 4 - 4
Osteomyelitis 1 - - 2
Dialysis shunt - 2
Pneumonia - 1 - -
Skin lesion - 3
Urinary tract - 2
Decubitus ulcer - 1
Primary focus
absent 2 7 4 a
Totals 6 19 4 30

docarditis had community-acquired infection (p <0.05).
The situation was similar among nondiabetics. Four of
four (100 percent) nondiabetics with endocarditis and
10 of 30 (33 percent) without endocarditis acquired the
infection outside the hospital (p X0.05). Diabetics with
endocarditis were no more likely to use insulin than
those without (88 percent versus 59 percent, p >OS).
None of the eight diabetics with endocarditis had evi-
dence of prior cardiac valvular disease. Two of four
nondiabetics with endocarditis had known rheumatic
valvulitis. Six diabetic and nondiabetic patients without
endocarditis had known valvular disease (rheumatic or
fibrocalcific disease, prosthetic valves, mitral valve
prolapse) but did not meet the clinical criteria for
staphylococcal endocarditis.
Primary Foci of Staphylococcal Infection. A primary
source of staphylococcal bacteremia or endocarditis
was identified in 18 (67 percent) diabetics and 22 (65
percent) of nondiabetics (Table II). Thirty-three percent
(six of 18) of diabetics with a primary focus had endo-
carditis. In contrast, none of the 22 nondiabetics with
a primary focus had endocarditis (p = 0.005). Staphy-
lococcal infection of an ischemic or neuropathic foot
was a primary source among five of the eight diabetics
with endocarditis. The most frequent portals of staph-
ylococcal entry into the bloodstream among all patients
without endocarditis were infected surgical wounds
(eight) and intravascular catheters (12).
Mortality. Appropriate antimicrobial therapy was ad-
ministered to 54 of the 61 patients studied. When ap-
propriate therapy was not given, the outcome was
uniformly fatal. Three diabetics and four nondiabetics
were in this group and rapidly died of the infection. The
overall mortality was 36 percent (22 of 61). Among
appropriately treated patients, 26 percent (15 of 54)
died. The median age among surviving diabetics and
nondiabetics was 56 and 58 years, respectively. Among
those patients who died, the median age was 57 and 60
years. Table III stratifies the mortality among these
patients by the severity of underlying illness. All seven
patients (one diabetic and six nondiabetic) with rapidly
fatal diseases died despite therapy. Three of 11 patients
with endocarditis died, as did 12 of 43 without endo-
carditis. After stratification for underlying illness, there
was a nonsignificant increase in mortality among pa-
tients with endocarditis (three deaths observed versus
1.8 expected). Four of 24 diabetics died, as did 11 of 30
nondiabetics. Comparison of mortality between dia-
betics and nondiabetics after stratification for underlying
illness indicated that the four observed diabetic deaths
were not significantly different from the expected 4.8
deaths.

TABLE Ill Mortality among Patients Receiving Appropriate Therapy

Diabetics Nondiabetics
Endocardltls No Endocardftis Endocardltls No Ertdocardltis

Underlying disease
Rapidly fatal - l/l’ - 6/6
Ultimately fatal - l/3 - I/?
Nonfatal/none l/7 l/13 214 2113

Total l/7 3/17 214 9/26

l Number of fatalities/number of patients.

660 November 1982 The American Journal of Medicine Volume 73

 STAPHYLOCOCCUS
COMMENTS
Several aspects of cellular and humorai immunity have
been shown to be impaired in diabetic subjects [9- 1 I] d
Despite these host abnormalities, patients with weii-
controlled diabetes are probably no more susceptible
to infection than are persons in the general population
[9,1 I]. Although certain infections appear to occur
more frequently in diabetics (rhinocerebral mucormy-
cosis, emphysematous pyeionephritis and M&cystitis,
malignant otitis external, cutaneous candidiasis, syn-
ergistic bacterial gangrene), ail but candidiasis are
unusual infections that occur in the presence of ke-
toacidosis or poorly controlled disease.
An unresolved question is whether diabetic patients
tolerate infection less well than nondiabetic patients do.
Staph. aureus bacteremia and endocarditis are said to
have a poor prognosis in diabetics. Ciuff et al. [I]
studied a group that included diabetics. The overall
mortality in the series was 42 percent but 69 percent
among diabetics. Forty patients with Staph. aureus
endocarditis reviewed by Watanakunakorn and Baird
[2] had a mortality of 40 percent. The nine diabetics
among them had a 56 percent mortality, and underlying
diabetes was deemed a poor prognostic factor in pa-
tients with staphyiococcai endocarditis.
This study was undertaken to determine whether
underlying diabetes is actually a risk factor in patients
with staphyiococcai bacteremia and endocardiiis. in this
retrospective analysis, comparisons were made be-
tween diabetic and nondiabetic patients who were cared
for in the same hospital during the same time period.
Although the two groups were clearly drawn from dif-
ferent populations, they were nearly identical in age and
sex distribution, and proportions with community-ac-
quired bacteremia and pre-existing vaivuiar disease.
The overall prevalence of endocarditis, 20 percent, was
compatible with that of recent studies [I 2,131 that re-
ported rates of 16 and 25 percent. This series failed to
demonstrate a significant difference between diabetics
and nondiabetics in the occurrence of endocarditis in
the presence of staphyiococcal bacteremia (30 percent
versus 12 percent). The 95 percent confidence limits
for the difference between these two percentages
however is -6 percent to 42 percent.
in agreement with other studies [ 12-141, staphylo-
coccal endocarditis among both diabetic and nondia-
betic patients was usually a community-acquired in-
fection. In fact, only one of 12 patients with endocarditis
in this series became infected within the hospital. in
contrast, in both patient groups, bacteremia without
endocarditis was usually hospital-acquired. in this study,
none of the diabetics in whom endocarditis developed
had known vaivuiar disease; two of four nondiabetics
had known rheumatic heart disease. Diabetics and
nondiabetics were equally likely to have fever and
November 1982
AUREUS BACTEREMIA IN DIABETES-COOPER AND PLATT
ieukocytosis. Neither of these signs was useful in in-
ferring the presence of endocarditis.
The 36 percent overall mortality in this series is
higher than that reported in recent studies (21 percent,
30 percent) [ 12,131 and is due to the relatively large
number of patients with cancer and the absence of
narcotic addicts, a group with a favorable prognosis
[ 151. The 27 percent mortality of staphyiococcai en-
docarditis in this group was comparable to the rates, 28
percent and 40 percent, reported elsewhere [2,12] and
was slightly, although not significantly, greater than the
adjusted mortality of patients without endocarditis. Di-
abetics had a slightly, although insignificantly, lower
mortality than nondiabetics despite the fact that a
greater proportion of diabetics had endocarditis. Ai-
though advanced age has been considered a risk factor
[ 21, the median ages of those who died and survived in
our group were nearly equal.
in summary, the diabetic and nondiabetic patients
with staphyiococcaf bacteremia in this series were in
many ways indistinguishable. They had the same clinical
presentation, the same prevalence of endocarditis, the
same distribution of community- and hospital-acquired
infection, and the same mortality rates. Therefore, it
does not appear that underlying diabetes meiiitus
confers an unfavorable prognosis in staphylococcai
bacteremia.
The diagnosis of endocarditis is important because
of the prolonged treatment it requires. Current diag-
nostic methods underestimate the presence of endo-
carditis by at least 15 percent [ 147. Staph. aureus en-
docarditis has been reported to be likely in the absence
of an identifiable primary focus of infection [ 131.
Conversely, endocarditis was unusual in the presence
of a primary focus. The reliability of these correlations
has, however, been challenged by well-documented
cases of endocarditis accompanying primary foci of
infection [ 161. Our results found no endooarditis in the
presence of a primary focus in nondiabetics (none of
22) but clearly reject such an association in diabetics.
One third of the diabetic patients (six of 18) with a pri-
mary focus had endocarditis (none of 22 versus six of
18, p <O-05).
Why should diabetics be so much more kkeiy to nave
endocarditis in the presence of primary foci than non-
diabetics? Five of the six diabetics with endocarditis and
a primary focus had chronic infections of the soft tissue
or bone. These infections may smouider and provide
a source of recurrent staphyiococcai access to the
circulation. in contrast, primary sources of infection like
intravascular catheters or surgical wounds are usually
quickly recognized and promptiy removed and/or
treated, thus decreasing the duration of bacteremia.
Therefore, it may be the nature of the infection (long-
standing, unrecognized) and not any defect in host re-
The American Journalof Medlclne Volume 73 661
STAPHYLOCOCCUS AUREUS BACTEREMIA IN DIABETES-COOPER
sponse that differentiates diabetics from nondiabetics
in the correlation between primary sites of infection and
the presence of endocarditis. It is also possible that the
primary foci in diabetic patients deliver larger inocula
of organisms to the blood. Although diabetics who inject
insulin are known to have a higher mucocutaneous
carriage rate of Staph. aureus than either nondiabetics
or diabetics who do not use insulin [ 171, insulin usage
was not a risk factor for endocarditis among bacteremic
diabetics.
REFERENCES
1. Cluff LE, Reynolds RC, Page DL, et al.: Staphylococcal bac-
teremia and altered host resistance. Ann Intern Med 1968;
69: 859.
2. Watanakunakorn C, Baird IM: Prognostic factors in Staphy-
lococcus aureus endocarditis and results of therapy with
a penicillin and gentamicin. Am J Med Sci 1977; 273:
133.
3. MacGregor RR, Beaty HN: Evaluation of positive blood cul-
tures. Arch Intern Med 1972; 130: 84.
4. Center for Disease Control: Algorithms for diagnosing infec-
tions. Am J Epidemiol 1980; 111: 635-643.
5. McCabe WR, Jackson GG: Cram-negative bacteremia. Arch
Intern Med 1962; 110: 847.
6. Armitage P: Statistical methods in medical research. Oxford:
Blackwell Scientific, 1973; 129-138.
7. Peto R, Pike MC, Armitage P, et al.: Design and analysis of
randomized clinical trials requiring prolonged observation
of each patient. Br J Cancer 1975; 35: l-39.
8. Fleiss JL: Statistical methods for rates and proportions, 2nd
ed. New York: John Wiley & Sons, 1981; 29-30.
9. Editorial: Infection and diabetes. Br Med J 1974; 2: 76.
10. Robertson HD, Polk HC: The mechanism of infection in pa-
662 November 1982 The American Journal of Medicine
AND PLATT
The presence of primary foci may at times play a role
in the decision concerning duration of treatment of
nondiabetic patients with staphylococcal bacteremia.
Our data, however, indicate that in the diabetic popu-
lation, the presence of such foci is not helpful in making
therapeutic decisions. Primary infecting foci, particularly
chronic infections of ischemic or neuropathic areas,
often coexist with endocarditis. When appropriately
treated, diabetics and nondiabetics had similar survival
rates.
tients with diabetes mellitus: a review of leukocyte mal-
function. Surgery 1974; 75: 123.
11. Allen JC: The diabetic as a compromised host. In: Allen JC,
ed. Infection and the compromised host. 2nd ed. Baltimore:
Williams & Wilkins, 1981; 229-270.
12. Shah M, Watanakunakorn C: Changing patterns of Staphylo-
coccus aureus bacteremia. Am J Med Sci 1979; 278:
115.
13. Nolan CM, Beaty HN: Staphylococcus aureus bacteremia:
current clinical patterns. Am J Med 1976; 60: 495.
14. Watanakunakorn C, Tan JS, Phair JP: Some salient features
of Staphylococcus aureus endocarditis. Am J Med 1973;
54: 473.
15. Menva KB, Gorbach SL: Favorable experience with bacterial
endocarditis in heroin addicts. Ann Intern Med 1973; 78:
25.
16. Watanakunakorn C, Baird IM: Staphylococcus aureus bac-
teremia and endocarditis associated with a removable in-
fected intravenous device. Am J Med 1977; 63: 253.
17. Tuazon CU, Perez A, Kishaba T, et al.: Staphylococcus aureus
among insulin-injecting diabetic patients: an increased
carrier rate. JAMA 1975; 231: 1272.
Volume 73