Serum Inflammatory Markers in the Diagnosis of Diabetic Foot Osteomyelitis: A Meta-Analytic Review

Introduction
:

The

aim

of

this

meta
-
analytic

review

was

to

help

establish

optimal

cut
-
off

levels

when

using

erythrocyte

sedimentation

rate

(ESR)

and

C
-
reactive

protein

(CRP)

in

the

diagnosis

of

diabetic

foot

osteomyelitis

(OM)
.

A

secondary

aim

was

to

determine

whether

these

simple

diagnostic

tests

demonstrate

more

of

a

“threshold

effect”

or

“dose
-
response”

when

discriminating

between

patients

with

and

without

OM
.

Methods
:

Six

eligible

articles

were

identified

through

a

search

of

Medline

and

EMBASE

electronic

databases

from

1966

to

2010
.

Summary

likelihood

ratios

for

predicting

the

presence

of

OM

in

the

case

of

both

a

positive

test

result

(LR+,

sensitivity/
1
-
specificity)

and

negative

test

result

(LR
-
,

1
-
sensitivity/specificity)

were

calculated

along

with

95
%

confidence

intervals

(CIs)

using

random
-
effects

models

for

several

commonly

reported

thresholds

for

ESR

and

CRP
.

Mantel
-
Haenszel

test

was

used

to

test

for

dose
-
response

as

thresholds

were

increased

given

a

positive

result

(LR+)

and

decreased

for

a

negative

result

(LR
-
)
.

Results/Conclusion
:

ESR

>

80
mm/h

proved

to

be

the

best

threshold

and

test

for

ruling

in

OM

(LR+

9
.
40

[
95
%

CI

0
.
65

-

135
)
.

Furthermore,

there

was

an

obvious

dose
-
response

with

increasing

ESR

levels,

and

a

three
-
fold

increase

in

the

likelihood

of

OM

as

ESR

levels

increased

from

60
mm/h

to

80
mm/h

(p

=

0
.
04
)
.

CRP

≤

3
.
2
mg/dl

proved

to

be

the

best

blood

test

for

ruling

out

OM

(LR
-

0
.
23

[
95
%

CI

0
.
17
-
0
.
52
]),

however,

unlike

ESR,

CRP

demonstrated

more

of

a

threshold

effect

in

the

diagnosis

of

OM

as

increasing

and

decreasing

levels

did

not

appear

to

alter

the

likelihood

of

having

underlying

OM

significantly

(p

>

0
.
05
)
.


Abstract:

Serum Inflammatory Markers
in
t
he
Diagnosis of
Diabetic
Foot Osteomyelitis
: A Meta
-
Analytic Review

Adam Fleischer, DPM, MPH, Maureen
Allanson
, BS, James
Wrobel
, DPM, MS

Department of Surgery, Advocate Illinois Masonic Medical Center

Introduction:

Diabetes

mellitus

(DM)

is

a

significant

public

health

concern,

with

an

estimated

23
.
6

million

people

affected

in

the

US

alone
.

The

US

spends

approx
.

$
174

billion,

or

~
1

out

of

every

5

healthcare

dollars,

on

patients

with

diabetes

annually
.

Foot
-
related

complications

in

the

form

of

ulceration

and

infection

are

the

leading

cause

of

hospitalization

and

lower

extremity

amputation

(LEA)

in

patients

with

diabetes,

costing

the

US

healthcare

system

$
43
,
000

to

$
63
,
000

per

event
.

Early

recognition

of

diabetic

OM

may

curtail

the

need

for

LEA,

however,

confirmatory

tests

such

as

biopsy

are

invasive
.

Diagnostic

tests

such

as

MRI

are

expensive

and

not

always

available

and

standard

radiographs

lack

sufficient

sensitivity

to

accurately

diagnose

early

OM
.

Therefore,

simple

laboratory

tests

of

common

serum

inflammatory

markers

(e
.
g
.
,

ESR,

CRP)

have

been

proposed

as

diagnostic

aids,

but

there

is

uncertainty

regarding

their

usefulness

in

DM

OM,

and

it

is

unclear

whether

or

not

these

tests

demonstrate

either

a

dose

response

or

threshold

effect
.




To

better

characterize

how

ESR

and

CRP

might

be

used

in

the

diagnosis

of

diabetic

foot

OM
.

Methods:



We

searched

MEDLINE

and

EMBASE

databases

with

the

following

search

terms
:

“diabetic”

and

“osteomyelitis”

or

“diabetic”

and

“ulcer”,

in

addition

to

each

of

the

following

words
:

C
-
reactive

protein

and

erythrocyte

sedimentation

rate
.


The

search

was

limited

to

studies

involving

adults

and

English

language

articles

from

the

period

1966

through

27

June

2010
.



We

identified

additional

articles

through

a

hand

search

of

references

from

retrieved

articles,

previous

reviews,

and

polling

experts
.

The

titles

and

abstracts

of

the

articles

retrieved

were

evaluated

to

determine

their

eligibility

for

our

review
.

Publications

in

abstract

and

letter

form

were

included

to

minimize

publication

bias
.



Conclusions:

1.
ESR

demonstrates

a

dose
-
response

likelihood

of

having

underlying

OM

with

increasing

levels

(p=
0
.
04
)
;

however

CRP

does

not

show

this

same

pattern

(
p=
0
.
266
)
.


2.
The

likelihood

of

having

OM

increased

three
-
fold

as

the

ESR

threshold

was

increased

from

70

mm/h

to

80

mm/h
.


3.
CRP,

on

the

other

hand,

demonstrates

more

of

a

threshold

effect

in

the

diagnosis

of

OM,

with

3
.
2

mg/dl

being

the

optimal

cut
-
point
.

4.
ESR

appears

to

be

a

specific

marker

for

diabetic

OM
,

with

increasingly

higher

LR+

values

at

higher

threshold

levels

(p<
0
.
05
)
.

An

elevated

ESR

above

80

mm/h

strongly

favors

the

diagnosis

of

OM

(LR+

9
.
40
)
.


5.
CRP,

conversely,

appears

to

be

a

sensitive

marker

for

diabetic

OM,

with

values

lower

than

3
.
2

effectively

ruling

out

underlying

OM

(LR
-

0
.
23
)
.


6.
Because

serum

inflammatory

markers,

by

themselves,

lack

accuracy,

we

believe

they

will

be

most

useful

in

modulating

our

pre
-
test

probability

for

OM

prior

to

considering

our

more

formal

diagnostic

(MRI,

radiographs)

or

confirmatory

tests

(biopsy/culture)
.


References:

Ertugrul, BM, Savk, O, Ozturk, MC, Oncu, S, Sakarya, S. The diagnosis of diabetic foot osteomyelitis: Examination findings an
d l
aboratory values. Med Sci Monit 15: 307
-
312, 2009.

Fleischer, AE, Didyk AA, Woods, JB, Burns, SE, Wrobel, JS, Armstrong, DG. Combined clinical and laboratory testing improves d
iag
nostic accuracy for osteomyelitis in the
diabetic foot. J Foot Ankle Surg 48:39
-
46, 2009.

Kaleta, JL, Fleischli, JW, Reilly, CH. The diagnosis of osteomyelitis in diabetes using erthorcyte sedimentation rate: A pilo
t s
tudy. J Am Pod Med Assoc. 91:445
-
450, 2001.

Malabu, UH, Al
-
Rubeaan, KA, Al
-
Derewish, M. Diabetic foot osteomyelitis: Usefulness of erythrocyte sedimentation rate in its dia
gnosis. West Afr J Med 26:113
-
116, 2007.

Newman, LG, Waller J, Palestro, CJ, Schwartz, M, Klein, MJ, Hermann, G, Harrington, E, Harrington, M, Roman, SH, Stagnaro
-
Green,

A. Unsuspected osteomyelitis in
diabetic foot ulcers: Diagnosis and monitoring by leukocyte scanning with indium in 111 oxyquinoline. J Am Med Soc 266:1246
-
1251
, 1991.

Rabjohn, L, Roberts, K, Troiano, M, Schoenhaus, H. Diagnostic and prognostic value of erythrocyte sedimentation rate in conti
guo
us osteomyelitis of the foot and ankle. J
Foot Ankle Surg. 46:230
-
237, 2007.


Purpose

Background

Search

strategy



Articles

were

included

for

review

if

they

fulfilled

all

of

the

following

criteria
:



They

were

original

studies

describing

CRP

or

ESR

in

the

diagnosis

of

lower

extremity

osteomyelitis

in

patients

with

DM


Data

could

be

extracted

into

2
x
2

tables


The

diagnostic

test

was

compared

with

a

reference

standard
.



In

studies

that

included

patients

with

and

without

DM,

only

those

patients

with

DM

were

included

in

the

analysis
.



Statistical

Analysis




Summary

likelihood

ratios

for

predicting

the

presence

of

OM

in

the

case

of

both

a

positive

test

result

and

negative

test

result,

as

well

as

sensitivities

and

specificities

were

calculated

using

random
-
effects

models

for

several

commonly

reported

thresholds

for

ESR

and

CRP
.


The

Mantel
-
Haenszel

test

for

trend

was

used

to

test

for

dose
-
response

as

thresholds

were

increased

given

a

positive

result

and

decreased

for

a

negative

result
.


P

values

less

than

0
.
5

were

considered

significant
.


All

analyses

were

performed

using

SAS

version

9
.
2

for

Windows
.


Results:



Our

original

search

yielded

152

potential

articles

of

which

19

were

selected

for

review

by

the

authors
.

Fourteen

articles

were

excluded
:

9

articles

did

not

use

ESR

or

CRP

in

the

diagnosis

of

osteomyelitis

and

the

data

could

not

be

extracted

in

the

remaining

5
.

Our

reference

search

provided

us

with

an

additional

35

articles
;

however

only

1

of

these

articles

fulfilled

all

of

the

inclusion

criteria

mentioned

previously
.


In

total,

six

articles

were

included

in

our

meta
-
analysis

(Figure

1
)
.

Figure

1
.

Search

strategy



All

six

articles

examined

ESR’s

relation

to

diabetic

osteomyelitis

however,

only

one

article,

Fleischer

et

al

2009
,

described

CRP

thresholds
.


The

articles

we

examined

were

of

fair

to

poor

quality

generally
.



Two
-
thirds

of

the

studies

were

prospective
.



None

of

the

article

displayed

level

one

evidence

and

only

two

studies

reported

consecutive

patient

enrollment
.



All

but

one

of

the

studies

used

bone

histopathology

to

diagnose

osteomyelitis
;

however,

Kaleta

et

al

and

Ertrugal

et

al

utilized

the

histopathology

gold
-
standard

inconsistently

at

about

55
%

of

the

time
.



There

were

a

total

of

305

patients

represented

in

this

meta
-
analysis
.



The

majority

of

these

people

were

inpatients

with

a

mean

prevalence

of

OM

of

59
%
.


Table 1.

Studies Describing CRP or ESR in the Diagnosis of
Lower Extremity Osteomyelitis in Patients with Diabetes
Mellitus





Source

Level of
Evidence*

Study
Type

Consecutive
Enrollment

Population

Age

Prevalence of
OM, %

Biopsy,
%

Method of Diagnosis of OM

Newman
et al 1991

II

Prospe
ctive

Yes

Inpatients
and
outpatients
with foot
ulcer
(n=35)

55.0

68†

100†

Culture and/or pathologic criteria
(osteonecrosis, marrow fibrosis,
and/or presence of inflammatory
cells)

Kaleta et
al 2001

IV

Retros
pective

No

Inpatients
with
osteomyeli
tis or
cellulitis
(n=29)

62.0

66

53

Pathology , positive results of at
least two imaging modalities (bone
scan, MRI, radiographs), or positive
probe to bone

Malabu et
al 2007

IV

Prospe
ctive

NS

Patients
with foot
ulcer
(n=43)

56.3

54

0

Positive results of two imaging
modalities (bone scan, MRI,
radiographs) or positive probe to
bone

Rabjohn
et al 2007

III

Prospe
ctive

NS

Inpatients
with
clinically
suspected
osteomyeli
tis (n=95)

61.0

69

100

Pathology or radiographs or
Technitium
-
99 three phase bone
scans

Ertugrul
et al 2009

II

Prospe
ctive

Yes

Inpatients
with
diabetic
foot lesions
(n=46)

64.0

52

54

Histopathology or microbiological
culture or T1 weighted MRI scan

Fleischer
et al 2009

III

Retros
pective

No

Inpatients
with a
single
forefoot
ulceration
(n=54)

61.5

63

100

Histopathology (focal necrosis or
intramedullary fibrosis with
infiltration of neutrophils)

Table 1
.
Abbreviations: CRP, C
-
reactive protein, ESR, Erythrocyte sedimentation rate, MRI, magnetic
resonance imaging, NS, Not specified, OM, osteomyelitis

*Level I: independent, blind evaluation of test with gold standard among a large sample of consecutive patients
with suspected target condition.. Level II: independent, blind evaluation of test with gold standard among a
small sample of consecutive patients with suspected target condition. Level III: nonindependent, blind
evaluation of test with gold standard among a nonconsecutive group of patients with suspected condition.
Level IV: nonindependent evaluation of test with standard of uncertain validity among a sample of patients
who have the target condition. Level V: nonindependent evaluation of test with standard of uncertain validity
among a sample of patients

† Calculations based on number of ulcers




ESR

>

80
mm/h

was

identified

as

the

single

most

specific

laboratory

test

for

diagnosing

OM

(spec=
94
.
2
%
)
.

There

was

an

obvious

dose
-
response

observed

for

the

likelihood

of

underlying

OM

with

increasing

ESR

thresholds,

and

a

three
-
fold

increase

in

the

likelihood

of

OM

as

ESR

threshold

was

increased

from

70
mm/h

to

80
mm/h

(
3
.
11

-
>

9
.
40
)
.



This

trend

was

statistically

significant

using

the

Mantel
-
Hanszel

test

with

a

p

value

of

0
.
04

(Table

2
)
.


Table 2. Summary Characteristics
for Erythrocyte
Sedimentation Rate for Diagnosing Osteomyelitis





ESR threshold

n

Sensitivity

Specificity

LR (+)*

LR (
-
)†

>60 mm/h

129

80.5 (74.0
-
85.8)

73.1 (63.4
-
80.9)

2.89 (0.38
-
21.5)

0.245 (0.24
-
2.56)

>70 mm/h

267

64.7 (60.6
-
68.0)

82.5 (75.3
-
88.2)

3.11 (1.21
-
7.99)

0.424 (0.178
-
1.01)

>80 mm/h

129

62.3 (56.8
-
64.9)

94.2 (86.1
-
98.0)

9.40 (
0.65
-
135)

0.365 (
0.03
-
5.26
)

Table 2.
Data in parentheses represent 95% confidence intervals. Abbreviations: LR (+), positive likelihood
ratio; LR (
-
), negative likelihood ratio; ESR, erythrocyte sedimentation rate

*Mantel Haenszel test for trend with increasing threshold levels = 4.041, p = 0.044

†Mantel Haenszel test for trend with decreasing threshold levels = 0.636, p = 0.425





Neither

increasing

nor

decreasing

levels

of

CRP

demonstrated

a

dose

response

for

diagnosing

OM

as

evident

by

the

nonsignifcant

Mantel
-
Haenszel

test
.

CRP

3
.
2

mg/dl

was

the

best

threshold

for

diagnosing

OM

with

the

highest

combined

total

or

sum

for

sensitivity

and

specificity

of

85
.
3

and

65
.
0

respectively
.

Table 3. Summary Characteristics for C
-
Reactive Protein for
Diagnosing Osteomyelitis





CRP threshold

n

Sensitivity

Specificity

LR (+)*

LR (
-
)†

2.3 mg/dl

54

88.0 (79.6
-
94.7)

40.0 (25.3
-
51.0)

1.47 (0.10
-
2.84)

0.29 (0.21
-
0.49)

3.2 mg/dl

54

85.3 (75.7
-
92.2)

65.0 (48.7
-
76.8)

2.44 (1.12
-
3.76)

0.23 (0.17
-
0.52)

8.4 mg/dl

54

55.9 (45.9
-
62.5)

80.0 (63.0
-
91.3)

2.79 (1.51
-
4.08)

0.55 (0.32
-
1.90)

Table 3.
Data in parentheses represent 95% confidence intervals

Abbreviations: LR (+) , positive likelihood ratio; LR (
-
), negative likelihood ratio; CRP, C
-
reactive protein

*Mantel Haenszel test for trend with increasing threshold levels = 1.236, p = 0.266

†Mantel Haenszel test for trend with decreasing threshold levels = 1.381, p = 0.240



As

CRP

threshold

(
green

arrow
)

for

establishing

OM

is

raised

from

2
.
3

to

3
.
2

to

8
.
4
,

the

likelihood

of

getting

a

true

positive

compared

to

a

false

positive

remains

essentially

the

same,

indicating

that

CRP

may

demonstrate

more

of

a

threshold

effect

for

diagnosing

underlying

OM

(Figure

2
)
.


However,

this

is

not

true

of

ESR
.

As

the

ESR

threshold

(
blue

arrow
)

for

diagnosing

OM

is

increased

from

60

to

70

to

80
,

the

likelihood

of

getting

a

false

positive

decreases

significantly

as

the

likelihood

of

getting

a

true

positive

remains

relatively

constant
.

This

trend

suggests

that

ESR

demonstrates

more

of

a

dose

response

likelihood

of

OM

with

increasing

levels

(Figure

3
)
.


Figure

2
.

Summary

receiver

operating

characteristic

curves

for

CRP

(from

Fleischer

et

al
.

2009
,

unpublished

figure,

n=
54
)
.



Figure

3
.

Summary

receiver

operating

characteristic

curves

for

ESR

(from

Fleischer

et

al
.

2009
,

unpublished

figure,

n=
54
)
.