The Journal of Foot & Ankle Surgery 50 (2011) 541–546
Contents lists available at ScienceDirect
The Journal of Foot & Ankle Surgery
journal homepage: www.jfas.org
Post-treatment Leukocytosis Predicts an Unfavorable Clinical Response
in Patients with Moderate to Severe Diabetic Foot Infections
Adam E. Fleischer, DPM, MPH, FACFAS 1, James S. Wrobel, DPM, MS 2, Andrea Leonards, DPM 3,
Scott Berg, BS 4, Daniel P. Evans, DPM 5, Robert L. Baron, DPM 6, David G. Armstrong, DPM, MD, PhD 7
1
Assistant Professor, Departments of Radiology and Surgery, Dr. William M. Scholl College of Podiatric Medicine at Rosalind Franklin University of Medicine and Science, North Chicago, IL;
Director of Research, Podiatric Surgical Residency Training Program, Advocate Illinois Masonic Medical Center, Chicago, IL
2
Associate Professor, Department of Medicine and Director, Scholl’s Center for Lower Extremity Ambulatory Research (CLEAR), Rosalind Franklin University of Medicine and Science, North
Chicago, IL
3
Chief Resident, Podiatric Surgical Residency Training Program, Advocate Illinois Masonic Medical Center, Chicago, IL
4
Fourth Year Podiatric Medical Student, Dr. William M. Scholl College of Podiatric Medicine at Rosalind Franklin University of Medicine and Science, North Chicago, IL
5
Professor and Chair, Department of Radiology, Dr. William M. Scholl College of Podiatric Medicine at Rosalind Franklin University of Medicine and Science, North Chicago, IL
6
Professor and Former Chair, Department of Radiology, Dr. William M. Scholl College of Podiatric Medicine at Rosalind Franklin University of Medicine and Science, North Chicago, IL
7
Professor of Surgery and Director of Southern Arizona Limb Salvage Alliance (SALSA), University of Arizona College of Medicine, Tucson, AZ
a r t i c l e i n f o
a b s t r a c t
Level of Clinical Evidence: 2
Keywords:
diabetes mellitus
foot
length of stay
logistic regression
prognosis
surgery
white blood cell count
Our aim was to determine whether post-treatment laboratory values could help to predict the clinical
response in patients with advanced diabetic foot infections. One hundred and three consecutive patients
hospitalized for moderate or severe diabetic foot infections at a large, university-afﬁliated hospital were
identiﬁed and their records retrospectively reviewed. Deﬁnitive therapy during each patient’s hospital course
was deﬁned as any foot surgery when additional major surgery was not anticipated or when a course of deep
soft tissue and/or bone culture-speciﬁc antibiotics had been initiated. The clinical response was assessed at 90
days after the start of deﬁnitive therapy. A poor response was recognized as persistent infection at the initial or
a contiguous site or when unplanned revision surgery or amputation was subsequently required. The
peripheral white blood cell count, neutrophil count, erythrocyte sedimentation rate, and C-reactive protein
levels measured shortly after initiating deﬁnitive therapy (i.e., post-treatment) were examined for their
association with the clinical response using logistic regression models. A total of 38 patients with the complete
compliment of laboratory and clinical follow-up aged 59.7 12.3 years with a diabetes duration of 13.3 9.1
years were included. Leukocytosis, deﬁned as a white blood cell count >11,000 cells/mL, observed an average of
3 1.4 days after treatment, was the single most important marker for predicting a poor clinical response, and
the only signiﬁcant study variable in both univariate and multivariate analyses (multivariate odds ratio 9.7,
95% conﬁdence interval 1.0 to 92, p ¼ .048). We conclude that leukocytosis observed shortly after initiating
deﬁnitive therapy is predictive of an unfavorable clinical response by 90 days.
Ó 2011 by the American College of Foot and Ankle Surgeons. All rights reserved.
Foot ulceration and infection is the most common reason for
admission among those with diabetes in the United States and is the
leading cause of lower extremity amputations in developed countries
(1–3). Much of the cost of treating infected foot ulcers results from the
complex management they require and prolonged number of days
Financial Disclosure: None reported.
Conﬂict of Interest: None reported.
Address correspondence to: Adam E. Fleischer, DPM, MPH, FACFAS, Assistant
Professor, Departments of Radiology and Surgery, Dr. William M. Scholl College of
Podiatric Medicine at Rosalind Franklin University of Medicine and Science, 3471
Green Bay Road, North Chicago, IL 60064.
E-mail address: adam.e.ﬂ[email protected] (A.E. Fleischer).
Audio ﬁle online only at http://www.jfas.org
spent in the hospital determining treatment and obtaining specialty
consultations. In the United States, the average length of stay (LOS) for
patients with diabetic foot infections that require amputation is
longer than 10 days (4). This is a greater time period than for those
admitted for myocardial infarction, congestive heart failure, cerebrovascular accident, and, even, septicemia and is twice as long as the
average LOS in general (5). Identifying variables that may be predictive of the treatment response could help us to better moderate the
LOS by distinguishing between patients who require an extended stay
and those who could be safely discharged earlier.
Recently, several admission characteristics have been found to be
associated with poorer short-term outcomes in patients with diabetic
foot infections (6–13). The most signiﬁcant predictors of treatment
failure are an elevated white blood cell (WBC) count and deep foot
1067-2516/$ - see front matter Ó 2011 by the American College of Foot and Ankle Surgeons. All rights reserved.
doi:10.1053/j.jfas.2011.04.023
542
A.E. Fleischer et al. / The Journal of Foot & Ankle Surgery 50 (2011) 541–546
ulceration at the time of hospitalization (6,7,13). However, little is
known of the value of using laboratory test results observed immediately after initiating treatment to predict the clinical response in
this population. In the present analysis, we sought to determine
whether the laboratory values commonly obtained after the start of
medical and/or surgical intervention (e.g., peripheral WBC count,
neutrophil count, erythrocyte sedimentation rate [ESR], and C-reactive protein [CRP]) could be predictive of short-term clinical outcomes
in patients admitted with moderate and severe diabetic foot infections. It was our belief that the laboratory values observed after
initiating treatment would be even more valuable in predicting the
clinical course than those from the time of admission.
Patients and Methods
Our retrospective cohort consisted of 103 consecutive patients with diabetes
hospitalized because of an infected neurotrophic foot ulceration from January 2002 to
December 2006 at a large, urban university-afﬁliated hospital. The local ethics
committee reviewed and approved the study protocol, and the requirement for patient
consent was waived. In the present analysis, we included only those patients with
moderate (grade 3) or severe (grade 4) foot infections, as determined using the Infectious Diseases Society of America-International Working Group on the Diabetic Foot
scoring system (14). The ﬁrst encounter only was used for patients with multiple
admissions. Of the 103 patients, 62, who did not have serial CRP blood measurements
or who did not have a minimum of 90 days of follow-up, were excluded from the
present study. Three additional patients with either active cardiac disease or another
active infection, known confounders of the CRP level, were also excluded. The ﬁnal
number of patients thus included in our study was 38.
Deﬁnitive treatment was deﬁned as any foot surgery (e.g., debridement or amputation) when additional surgery was not anticipated, except in the case of delayed
primary closure. In those situations when no surgery was planned, deﬁnitive treatment
was deﬁned as the point when a course of deep soft tissue and/or bone culture-speciﬁc
antibiotics had been initiated and continued for an appropriate period (2 to 3 weeks for
soft tissue infections or 6 to 8 weeks for bone infections). Without knowledge of the
outcome status, 2 independent raters (A.L., A.F.) reviewed the records to determine the
date at which deﬁnitive treatment for each subject occurred. No conﬂicts were found
regarding the date of deﬁnitive treatment among the 38 subjects included, and all
included patients were thought to have had an appropriate length of antibiotic treatment. For the purposes of our report, we referred to the point immediately after the
start of deﬁnitive treatment as “post-treatment.”
The primary outcome of interest in the present analysis was the treatment
response, evaluated at 90 days after the start of deﬁnitive treatment. This interval is
a reasonable period to expect resolution of an infectious process within the lower
extremity and is comparable to the periods used in similar studies (13,15). A favorable
response was deﬁned as the absence of clinical signs of infection (2) at the initial and
contiguous sites and when revision surgery was not required. A poor response was
deﬁned as anything other than this, including death.
The primary predictor variables in the present study were the post-treatment WBC
count, neutrophil count, ESR level, and CRP level. All post-treatment values were
obtained at least 2 but no more than 6 days after the patient’s deﬁnitive treatment
began. The laboratory variables selected for inclusion were determined from previous
investigations (16–22). Because ESR does not demonstrate rapid changes after treatment (16,17,21), the post-treatment ESR values were not always available for every
patient. In addition to considering the absolute values, each of the post-treatment
markers were also examined for their trend (either increasing or decreasing from
baseline), the percentage of reduction (deﬁned as 1 [post-treatment value/baseline
value]), and the rate of reduction (deﬁned as [post-treatment value baseline value/
number of days between values]) after deﬁnitive treatment.
Numerous covariates were also examined for their association with the clinical
response, including patient demographics, baseline clinical ﬁndings, and the laboratory
values at admission (Table 1). Because we included only patients with infection
stemming from a diabetic neurotrophic ulceration, all study patients were believed to
have adequate arterial inﬂow for healing with either a palpable pedal pulse or an
Table 1
Baseline variables analyzed for association with clinical response (N ¼ 38 patients)
Baseline Predictor
Patient characteristics
Age (y)
Male gender
White ethnicity
Federal insurance
Previous or current smoking
Previous DFU hospitalization
History of foot/toe amputation
Deﬁnitive treatment involved surgery
Mild or moderate PAD
Infection characteristics
Duration of foot infection (d)
Antibiotic interval before admission (d)
Oral temperature ( F)
Hemodynamically unstable (IDSA 4)
Ulcer characteristics
Duration of primary ulcer (d)
Ulcer necrosis
Ulcer purulence
Probe to bone
Malodorous ulcer
Presence of osteomyelitis
Severe UT wound grade
Laboratory markers
Serum creatinine (mg/dL)
Serum glucose (mg/dL)
Hemoglobin (g/dL)
Platelets (103 cells/mL)
Serum albumin (g/dL)
Alkaline phosphatase (U/L)
Hemoglobin A1c (%)
White blood cell count (103 cells/mL)
Neutrophil count (103 cells/mL)
Erythrocyte sedimentation rate (mm/h)
C-reactive protein (mg/dL)
p Value*
Favorable Response (n ¼ 13)
Poor Response (n ¼ 25)
61.7 10.9
77
23
54
62
54
31
77
23
58.7 13.1
84
40
48
52
52
40
68
36
.481
.672
.473
.736
.580
.915
.581
.714
.486
13.4 15.4
9.8 16.0
98.7 1.8
8
10.6 11.9
5.4 12.4
99.1 1.3
20
.545
.358
.445
.643
43.5 52.1
23
54
62
70
85
62
73.0 214
44
44
64
64
88
92
.519
.294
.569
.883
.750
1.000
.034
.039
.556
.093
.353
.187
.073
.949
.047
.059
.035
.665
1.4
208
12.2
303
4.0
105
8.7
11.3
8.6
84.2
13.3
0.7
75
2.0
139
0.7
32
2.2
3.8
3.8
37.4
11.8
2.8
229
11.1
341
3.6
139
8.6
14.7
11.7
108
14.8
3.0
114
1.8
105
0.9
74
2.8
5.3
5.1
29.8
9.6
Abbreviations: DFU, diabetic foot ulceration; IDSA, Infectious Diseases Society of America score; PAD, peripheral arterial disease.
Data presented as percentage of column total or mean standard deviation.
Baseline variables were obtained within 24 hours of admission, except for PAD, which was conﬁrmed with ankle-brachial indexes within 6 months of admission.
* Student’s t test or Wilcoxon’s rank-sum test used for comparisons of continuous variables and chi-square test or Fisher’s exact test for comparisons of categorical variables.
A.E. Fleischer et al. / The Journal of Foot & Ankle Surgery 50 (2011) 541–546
Table 2
Post-treatment laboratory variables analyzed for association with clinical response
(N ¼ 38 patients)
Post-treatment Variable
White blood cells
Absolute count (103 cells/mL)
Trending upward
Percentage of reduction (%)
Rate of reduction (103 cells/mL/d)
Neutrophils
Absolute count (103 cells/mL)
Trending upward
Percentage of reduction (%)
Rate of reduction (103 cells/mL/d)
Erythrocyte sedimentation ratey
Absolute level (mm/h)
Trending upward
Percentage of reduction (%)
Rate of reduction (mm/h/d)
C-reactive protein
Absolute level (mg/dL)
Trending upward
Percentage of reduction (%)
Rate of reduction (mg/dL/d)
Favorable Response
(n ¼ 13)
Poor Response
(n ¼ 25)
p Value*
8.48 2.0
7.7
18.4 27.1
0.71 0.57
12.2 5.5
20
16.5 24.9
0.55 0.99
.005
.643
.822
.539
5.84 1.83
7.7
23.7 27.6
0.69 0.57
9.07 4.4
20
19.6 30.1
0.60 0.89
.016
.643
.686
.759
81.7 41.8
25
10.7 22.5
1.35 4.32
105 26.6
41
6.3 21.8
5.15 13.0
.079
.450
.602
.276
5.75 5.88
9.73 6.94
20
24.7 48.6
1.93 3.37
.086
.144
.039
.819
0
50.6 26.1
1.74 1.80
Data presented as mean standard deviation, except for “Trending upward,” given as
the percentage of the column total.
Post-treatment variables obtained an average of 3 1.4 days after the start of deﬁnitive
treatment.
* Student’s t test or Wilcoxon’s rank-sum test used for comparisons of continuous
variables and chi-square or Fisher’s exact test for comparisons of categorical variables.
y
Post-treatment erythrocyte sedimentation rate values determined for 29 patients,
17 with a poor response and 12 with a favorable response.
ankle-brachial index greater than 0.45 within the affected extremity. Patients with
nonpalpable pedal pulses and an ankle-brachial index greater than 0.45 but less than
0.9 or an ankle-brachial index greater than 1.2 were classiﬁed as having mild or
moderate peripheral arterial disease. Osteomyelitis was determined by histopathologic
examination or x-ray evidence of osteolysis, periostitis, and/or cortical destruction of
the underlying bone. The wound depth was deﬁned using the University of Texas (UT)
diabetic wound classiﬁcation (23,24) and categorized as either moderate (e.g., UT grade 1),
with extension to the subcutaneous tissue only, or severe (e.g., UT grade 2 or 3), with
extension to the tendon, capsule, bone, or joint (13).
Statistical Analysis
The predictor variables were initially stratiﬁed according to the clinical responsedfavorable or poor. Comparisons of the continuous variables were performed using
a 2-sample t test or Wilcoxon’s rank-sum test. Differences in proportions were
compared using either a chi-square test or Fisher exact test. Logistic regression models
were used to assess the univariate associations between the predictor variables and the
treatment response. In an effort to minimize the covariate-to-observation ratio (25), we
adopted a relatively conservative threshold for entry into the multivariate analysis.
Therefore, only variables with p < .05 on univariate analysis were included in the
multivariate model-building process. For predictors measured on a continuous scale
(e.g., WBC count, serum creatinine), both the original continuous handling and
important binary cutpoints determined using receiver operating characteristic curves
were tested in the multivariate models. The cutpoints were considered important if
they maximized the test’s sensitivity, maximized the test’s speciﬁcity, or maximized the
Table 3
Final multivariate model for predicting poor clinical response (N ¼ 38 patients)
Risk Factor
Regression Coefﬁcient
Odds Ratio
p Value
Intercept
Severe UT wound grade*
(versus moderate)
Post-treatment leukocytosisy
1.308
1.785
d
6.0
.160
.078
2.270
9.7
.048
Abbreviation: UT, University of Texas.
* Severe UT wounds involved extension to tendon, capsule, bone, or joint; moderate
UT wounds involved extension to subcutaneous tissue only.
y
Post-treatment leukocytosis deﬁned as white blood cell count >11,000 cells/mL
shortly after initiating deﬁnitive therapy, drawn, on average, for the cohort 3 1.4 days
after starting deﬁnitive treatment.
543
sum of the test’s sensitivity plus speciﬁcity. The ﬁnal multivariate model was determined using stepwise logistic regression analysis, with p < .10 as the criterion for model
entry and p > .10 for removal. The ﬁnal model was tested for goodness of ﬁt using the
Hosmer–Lemeshow test (26). We also performed a Greenland sensitivity analysis (27)
to account for the potential inﬂuence that an unmeasured variable might have had on
our effect estimates. Statistical analyses were conducted using SAS software, version 9.2
(SAS Institute, Cary, NC, and Microsoft, Redmond, WA). All tests were 2-tailed, with
p < .05 considered statistically signiﬁcant.
Results
A total of 38 patients with complete laboratory and clinical followup data were included. The average age and diabetes duration was 59.7
12.3 years and 13.3 9.1 years, respectively. Of the 38 patients, 27
(71%) were treated with a combination of surgery and antibiotics, and
11 (29%) received antibiotics only. The average hospital LOS was 12.3 7.9 (range 2 to 35) days. Deﬁnitive treatment was generally initiated
within the ﬁrst 3 days of admission (84%, 32 of 38 patients), and more
than half (55%, 21 of 38) of the study population underwent deﬁnitive
treatment during the ﬁrst 24 hours. Nearly two thirds of the cohort
(25 of 38) was classiﬁed as having had a poor clinical response by
90 days after starting treatment. Eight patients had persistent infection at the initial or a contiguous site and 17 required revision surgery
or amputation after what was believed to be deﬁnitive in-house
treatment. No deaths occurred during the 90-day follow-up period.
The population’s baseline characteristics stratiﬁed by clinical
response are described in Table 1. The baseline factors associated
(p < .05) with a poor clinical response in the univariate analysis
included UT wound grade, serum creatinine level, WBC count, and
ESR. The post-treatment laboratory variables stratiﬁed by the clinical
response are given in Table 2. The post-treatment values were
obtained an average of 3 1.4 (range 2 to 6) days after deﬁnitive
treatment. For individual patients, the post-treatment laboratory
values were always recorded within 24 hours of each other. Posttreatment ESR values were available for 76% of the cohort (29 of
38), and the values for the other post-treatment laboratory predictors
were available for every patient in the study population. The posttreatment variables associated (p < .05) with a poor clinical
response included the absolute WBC count, absolute neutrophil
count, and percentage of CRP reduction from baseline.
The study’s ﬁnal multivariate model, which included only posttreatment WBC count (using a binary handling) and severe UT
wound grade at admission, is shown in Table 3. The strongest
predictor of treatment failure was post-treatment leukocytosis, which
conferred an almost 10 times greater risk of experiencing a poor
clinical response (multivariate odds ratio [OR] 9.7, 95% conﬁdence
interval 1.0 to 92, p ¼ .048). Patients presenting with severe (versus
moderate) UT wounds and with leukocytosis after the start of treatment were at the greatest risk of treatment failure at 90 days
(combined OR 58). The Hosmer–Lemeshow goodness of ﬁt test was
not signiﬁcant for the ﬁnal model (p ¼ .7452), indicating that the
model was well ﬁt throughout the spectrum of predicted risk.
Furthermore, with just 2 predictor variables, the ﬁnal model was able
to discriminate between those with a poor and favorable response
nearly 80% of the time (area under the receiver operating characteristic curve 0.782). Despite the relatively small numbers in our study,
the results of the Greenland sensitivity analysis also revealed that the
primary effect estimate was fairly resistant to the potential inﬂuence
of an unmeasured variable. The estimated OR for post-treatment
leukocytosis varied only 3% to 19% as the OR of an unmeasured
confounder was arbitrarily varied from 2.4 to 10 on the outcome.
Fig. 1 contains the receiver operating characteristic curve for
post-treatment WBC count. This graph displays the sensitivity and
speciﬁcity levels associated with each unique post-treatment WBC
value in the study. A cutpoint of 11,000 cells/mLda commonly
544
A.E. Fleischer et al. / The Journal of Foot & Ankle Surgery 50 (2011) 541–546
Fig. 1. Receiver operating characteristic curve demonstrating ability of post-treatment WBC count to identify patients with unfavorable clinical response. Values along curve represent
potential cutpoints of absolute WBC count (103 cells/mL) observed w3 days after treatment. A threshold of 11 103 cells/mL corresponded to speciﬁcity of 0.92 and sensitivity of 0.48.
Univariate diagnostic odds ratios provided for important cutpoints, with 95% conﬁdence intervals in parentheses (N ¼ 38 patients). Area under curve of 0.712.
accepted threshold for leukocytosis, and the threshold used at our
institutiondproved to be the most discriminating test in the univariate and multivariate analyses. Because this cutpoint falls far to the left
on the receiver operating characteristic curve, post-treatment
leukocytosis was also a highly speciﬁc indicator of a poor clinical
response (speciﬁcity 0.92). This threshold, however, was not a very
sensitive indicator of a poor response, because 12 of the 25 patients
with a poor clinical response also had a normal WBC count at their
post-treatment blood draw (sensitivity 0.48).
Discussion
In the present study, we found that an elevated WBC count, shortly
after initiating deﬁnitive therapy, is predictive of an unfavorable
clinical response in patients with moderate to severe diabetic foot
infections. Even after adjusting for wound severity on admission,
post-treatment leukocytosis was associated with an almost 10 times
greater risk of experiencing a poor clinical response by 90 days. Posttreatment leukocytosis was also the only signiﬁcant study variable
(baseline or post-treatment) in both univariate and multivariate
analyses. To our surprise, the WBC trend, magnitude of change after
treatment, and the rate of change after treatment were not at all
associated with the patient’s short-term clinical outcome. Rather, our
ﬁndings indicated that it was the absolute WBC count after the start of
deﬁnitive treatment that carried the most prognostic weight.
Not surprisingly, we also found that the post-treatment WBC count
appeared to be slightly better at predicting the clinical response than
the WBC count at admission (area under the curve 0.712 versus 0.658,
respectively)da previously recognized predictor of poor outcomes in
patients with advanced diabetic foot infections (6,7,13). This makes
sense because we would expect that laboratory values that change as
a result of medical and/or surgical intervention should be more
representative of the patient’s prognosis than the laboratory values
the patient originally presents with.
Many of our ﬁndings are also supportive of previous work. In the
univariate analysis and without considering the effects of the other
covariates, we found that patients experiencing a poor clinical
response generally had higher WBC counts on admission (mean WBC
count 14.7 versus 11.3 103 cells/mL, p ¼ .047). Although many
patients with serious foot infections will not present with leukocytosis (28), the presence of a high WBC count at the time of admission
appears to carry a rather ominous prognosis. Ours is now the fourth
such study to suggest that leukocytosis, at or around admission, may
be associated with a greater rate of treatment failure among patients
with diabetic foot infections (6,7,13). This observation could seem
rather counterintuitive and probably warrants additional investigation. After all, initial leukocytosis should signify a greater degree of
immunocompetence at a point when patients are faced with the
challenge of a serious limb and/or life-threatening infection. We also
observed that patients with deep ulcerations fared far worse in their
response to treatment than those with superﬁcial wounds. In fact, 92%
of patients experiencing a poor clinical response in our population
had an ulceration that penetrated to the tendon, bone, joint, or
capsule at admission compared with only 62% in the group experiencing a favorable clinical response (univariate OR 7.2, 95% conﬁdence
interval 1.1 to 44, p ¼ .034). The concept of deeper, more complicated
wounds generally doing worse is fairly well supported in published
reports. Armstrong et al (24), for example, ﬁrst demonstrated that the
ulceration depth correlates highly with amputation risk in 1998. Pittet
et al (8) also found that gangrenous wounds were associated with
A.E. Fleischer et al. / The Journal of Foot & Ankle Surgery 50 (2011) 541–546
a much greater rate of treatment failure than wounds free of tissue
necrosis. Finally, using a large cohort of 402 diabetic patients with
infected foot ulcerations but without underlying osteomyelitis, Lipsky
et al (13) found that patients with deeper wounds were more than
twice as likely to experience antibiotic therapy failure.
Our results also suggest that the role of obtaining CRP levels in
those with serious diabetic foot infections probably requires additional clariﬁcation. CRP is an acute phase protein, perhaps best known
for its ability to predict heart disease risk, but might have similar
utility in diabetic foot syndrome as well. CRP has proved helpful in
differentiating infected from noninfected diabetic foot ulcers (29) and
for differentiating cellulitis from underlying osteomyelitis in patients
with mild to moderately infected diabetic foot ulcers (30). However,
its greatest strength might lie in its ability to rapidly monitor and
predict treatment response, just as has been described for other
disease states (17–20,22). In the present study, we observed that
patients with a favorable clinical response generally had a greater
percentage of reduction in their serum CRP levels after treatment than
those experiencing a poor response (50% versus 25%, respectively, p ¼
.039), and no one with a favorable response had increasing CRP levels
post-treatment. This univariate association between the CRP level and
treatment response, however, did not remain in the multivariate
analysis when the effects of the other laboratory predictors (i.e., posttreatment WBC count) were also considered. Still, Akinci et al (31)
recently found CRP levels were actually the most predictive laboratory marker of amputation risk using a large diabetic cohort with 165
infected ulcer episodes. Clearly, additional prospective study in this
area is needed to better elucidate the role of using CRP levels to
monitor the clinical response and predict outcomes in this unique
patient population.
The present study had several limitations. First, the retrospective
nature of the study might have allowed for a potential selection bias. It
is reasonable to believe that the included patients with serial CRP data
available, for example, could have been more complicated and
potentially sicker than those not undergoing serial blood draws. This
would help explain why 82% of the cohort had a severe UT wound and
why almost two thirds of the cohort experienced a poor clinical
response, a much greater rate than in previous reports (13). This may
make the results slightly less generalizable to more typical patients
hospitalized with moderate to severe foot infections. Also, our results
might have suffered some from residual confounding, potentially
resulting in inﬂated risk estimates for the predictor variables.
However, our sensitivity analysis suggested adequate stability for the
study’s primary effect estimate (i.e., post-treatment leukocytosis)
when an unmeasured confounder was factored into the analysis.
Finally, with such a large percentage of the study population also
experiencing concomitant bone infection (87%), we were limited in
our ability to determine whether, and to what extent, the presence of
osteomyelitis contributed to an unfavorable clinical response after
deﬁnitive treatment.
In conclusion, health care costs in the U.S. now exceed 16% of the
national GDP, and the current rate of growth has been described by
most economists as unsustainable. U.S. hospitals can expect to come
under increased ﬁnancial pressure during the foreseeable future. The
efﬁcient use of hospital beds can reduce health care expenditures and
improve the quality of health care delivery by limiting practice variationda recognized short-coming among diabetic foot providers
(32). Therefore, having insight into which patients require acute beds
versus those who could be safely discharged earlier in their hospital
course will become increasingly more important in the years ahead. In
the present study, we found that leukocytosis, observed shortly after
initiating deﬁnitive therapy, was a highly speciﬁc indicator of treatment failure, and conferred an almost 10 times greater risk of experiencing an unfavorable clinical response by 90 days. Perhaps just as
545
important, normalization of the WBC count and a decreasing WBC
count did not guarantee treatment success. A signiﬁcant reduction in
the CRP level after initial treatment might be our strongest indicator
of a favorable short-term clinical outcome; however, our study failed
to fully validate this ﬁnding. These results might aid physicians when
deciding whether additional in-patient management is necessary,
and, in turn, better help us to moderate the LOS for this challenging
patient population. Future work could focus on investigating the
reason leukocytosis at admission and after treatment appears to carry
such an ominous prognosis. Additional work is also needed to clarify
the role of using CRP to monitor and predict treatment response in
patients with serious foot infections.
References
1. Cavanagh P, Lipsky BA, Bradbury A. Treating diabetic foot ulcers. Lancet 366:1725–
1735, 2005.
2. Lipsky BA, Berendt AR, Deery HG, Embil JM, Joseph WS, Karchmer AW, LeFrock JL,
Lew DP, Mader JT, Norden C, Tan JS. Diagnosis and treatment of diabetic foot
infections. Clin Infect Dis 39:885–910, 2004.
3. Singh N, Armstrong DG, Lipsky BA. Preventing foot ulcers in patients with diabetes. JAMA 293:217–228, 2007.
4. US Centers for Disease Control Prevention. Average length of stay (LOS) in days of
hospital discharges for nontraumatic lower extremity amputation with diabetes as
a listed diagnosis, United States, 1980–2005 [cited 11 September 2008]. Available
at: http://www.cdc.gov/diabetes/statistics/lea/ﬁg2.htm. Accessed 24 July 2010.
5. Kozak LJ, DeFrances CJ, Hall MJ. National Hospital Discharge Survey: 2004 annual
summary with detailed diagnosis and procedure data. National Center for Health
Statistics. Vital Health Stat Series 13, Issue 162, 2006.
6. Eneroth M, Apelqvist J, Stenstrom A. Clinical characteristics and outcome in 223
diabetic patients with deep foot infections. Foot Ankle Int 18:716–722, 1997.
7. Akanji AO, Famuyiwa OO, Adetuyibi A. Factors inﬂuencing the outcome of treatment of foot lesions in Nigerian patients with diabetes mellitus. Q J Med 73:1005–
1014, 1989.
8. Pittet D, Wyssa B, Herter-Clavel C, Kursteiner K, Vaucher J, Lew PD. Outcome of
diabetic foot infections treated conservatively: a retrospective cohort study with
long-term follow-up. Arch Intern Med 159:851–856, 1999.
9. Benotmane A, Mohammedi F, Ayad F, Kadi K, Azzouz A. Diabetic foot lesions:
etiologic and prognostic factors. Diabetes Metab 26:113–117, 2007.
10. Diamantopoulos EJ, Haritos D, Yfandi G, Grigoriadou M, Margariti G, Paniara O,
Raptis SA. Management and outcome of severe diabetic foot infections. Exp Clin
Endocrinol Diabetes 106:346–352, 1998.
11. Grayson ML, Gibbons GW, Habershaw GM, Freeman DV, Pomposelli FB,
Rosenblum BI, Levin E, Karchmer AW. Use of ampicillin/sulbactam versus imipenem/cilastatin in the treatment of limb-threatening foot infections in diabetic
patients. Clin Infect Dis 18:683–693, 1994.
12. Criado E, De Stefano AA, Keagy BA, Upchurch GR Jr, Johnson G Jr. The course of severe
foot infection in patients with diabetes. Surg Gynecol Obstet 175:135–140, 1992.
13. Lipsky BA, Sheehan P, Armstrong DG, Tice AD, Polis AB, Abramson MA. Clinical
predictors of treatment failure for diabetic foot infections: data from a prospective
trial. Int Wound J 4:30–38, 2007.
14. Lipsky BA. A report from the international consensus on diagnosing and treating
the infected diabetic foot. Diabetes Metab Res Rev 20:S68–S77, 2004.
15. Malay DS, Margolis DJ, Hoffstad OJ, Bellamy S. The incidence and risks of failure to
heal after lower extremity amputation for the treatment of diabetic neuropathic
foot ulcer. J Foot Ankle Surg 45:366–374, 2006.
16. Crosby LA, Powell DA. The potential value of the sedimentation rate in monitoring
treatment outcome in puncture-wound-related Pseudomonas osteomyelitis. Clin
Orthop Rel Res 188:168–172, 1983.
17. Unikila-Kallio L, Kallio MJT, Peltola H. The usefulness of C-reactive protein levels in
the identiﬁcation of concurrent septic arthritis in children who have acute
hematogenous osteomyelitis: a comparison with the usefulness of the erythrocyte
sedimentation rate and the white blood-cell count. J Bone Joint Surg 76:848–853,
1994.
18. Khan MH, Smith PN, Rao N, Donaldson WF. Serum C-reactive protein levels
correlate with clinical response in patients treated with antibiotics for wound
infections after spinal surgery. Spine J 6:311–315, 2006.
€m O,
19. Nilsdotter-Augustinsson A, Briheim G, Herder A, Ljunghusen O, Wahlstro
Ohman L. Inﬂammatory response in 85 patients with loosened hip prostheses:
a prospective study comparing inﬂammatory markers in patients with aseptic and
septic prosthetic loosening. Acta Orthop 78:629–639, 2007.
20. Owens CD, Ridker PM, Belkin M, Hamdan AD, Pomposelli F, Logerfo F, Creager MA,
Conte MS. Elevated C-reactive protein levels are associated with postoperative
events in patients undergoing lower extremity vein bypass surgery. J Vasc Surg
45:2–9, 2007.
21. Rabjohn L, Roberts K, Troiano M, Schoenhaus H. Diagnostic and prognostic value of
erythrocyte sedimentation rate in contiguous osteomyelitis of the foot and ankle.
J Foot Ankle Surg 46:230–237, 2007.
546
A.E. Fleischer et al. / The Journal of Foot & Ankle Surgery 50 (2011) 541–546
22. Dupont C, Rodenbach J, Flachaire E. The value of C-reactive protein for postoperative monitoring of lower limb arthroplasty. Ann Readapt Med Phys 51:348–
357, 2008.
23. Lavery L, Armstrong DG, Harkless LB. Classiﬁcation of diabetic foot wounds. J Foot
Ankle Surg 35:528–531, 1996.
24. Armstrong DG, Lavery LA, Harkless LB. Validation of a diabetic wound classiﬁcation system: the contribution of depth, infection, and ischemia to risk of amputation. Diabetes Care 21:855–859, 1998.
25. Concato J, Feinstein AR, Holford TR. The risk of determining risk with multivariable
models. Ann Int Med 118:201–210, 1993.
26. Hosmer DW, Lemeshow S. Applied Logistic Regression, John Wiley & Sons, New
York, 1989.
27. Greenland S. Basic methods for sensitivity analysis of biases. Int J Epidemiol
25:1107–1116, 1996.
28. Armstrong DG, Lavery LA, Sariaya M, Ashry H. Leukocytosis is a poor indicator of acute
osteomyelitis of the foot in diabetes mellitus. J Foot Ankle Surg 35:280–283, 1996.
29. Jeandrot A, Richard JL, Combescure C, Jourdan N, Finge S, Rodier M, Corbeau P,
Sotto A, Lavigne JP. Serum procalcitonin and C-reactive protein concentrations to
distinguish mildly infected from non-infected diabetic foot ulcers: a pilot study.
Diabetologia 51:347–352, 2008.
30. Fleischer AE, Didyk AA, Woods JB, Burns SE, Wrobel JS, Armstrong DG. Combined
clinical and laboratory testing improves diagnostic accuracy for osteomyelitis in
the diabetic foot. J Foot Ankle Surg 48:39–46, 2009.
31. Akinci B, Yener S, Yesil S, Yapar N, Kucukyavas Y, Bayraktar F. Acute phase reactants predict
the risk of amputation in diabetic foot infection. J Am Podiatr Med Assoc 101:1–6, 2011.
32. Wrobel JS, Mayﬁeld JA, Reiber GE. Geographic variation of lower-extremity major
amputation in individuals with and without diabetes in the Medicare population.
Diabetes Care 24:860–864, 2001.