Professional Documents
Culture Documents
M E T A - A N A L Y S I S
F
oot infections in patients with dia- for several reasons, including inadequate
(10), including those with foot infections
betes can be difficult to treat, and surgical interventions, suboptimal
(18), several investigators have explored
therapeutic failure often leads to a wound care, or severe limb ischemia (3).
using it as an adjunct to treating diabetic
lower-extremity amputation (1,2). These All infected foot lesions require antibiotic
foot infections. Unfortunately, there have
infections may be refractory to treatment therapy, but their penetration to infected
only been a few studies and each enrolled
a relatively small numbers of subjects.
From the 1Center of Preventive Medicine, Verona, Italy; the 2General Internal Medicine Clinic, Veterans Furthermore, the available studies have
Affairs Puget Sound Health Care System, Seattle, Washington; the 3School of Medicine, University of come to different conclusions regarding
Washington, Seattle, Washington; the 4Departments of Histology, Microbiology, and Medical Biotechnol- the usefulness of G-CSF. In such situa-
ogy, University of Padua, Padua, Italy; and the 5Department of Infectious Diseases and Tropical Medicine,
San Bortolo Hospital, Vicenza, Italy. tions, meta-analysis is a useful tool to de-
Address correspondence and reprint requests to Benjamin A. Lipsky, MD, FACP, FIDSA, Professor of termine the potential benefit of a
Medicine, University of Washington, School of Medicine, Director, General Internal Medicine Clinic, VA therapeutic intervention (19 22). Thus,
Puget Sound Health Care System (S-111-GIMC), 1660 South Columbian Way, Seattle, WA 98108-1597. to define the effectiveness of G-CSF as an
E-mail: benjamin.lipsky@med.va.gov.
Received for publication 28 September 2004 and accepted in revised form 31 October 2004.
adjunctive therapy for treating diabetic
Abbreviations: G-CSF, granulocyte colonystimulating factor. foot infections, we thoroughly searched
2005 by the American Diabetes Association. the literature for all prospective studies of
this issue then subjected these to a sys- differs from those shown in the published about using G-CSF for diabetic foot prob-
tematic review and meta-analysis. articles. lems. These included one systematic re-
view, seven traditional reviews, seven
RESEARCH DESIGN AND Statistical analysis clinical studies, a comment letter on one
METHODS We searched the medi- We conducted a conventional meta- of these studies, and one case report. The
cal literature, using Medline, Embase, analysis using the Mantel-Haenszel fixed- systematic review of treating foot ulcers in
LookSmarts Find Articles, and the Co- effects model (24), applying the Der diabetic patients was published in 1999
chrane Library, for relevant studies pub- Simonian and Laird random-effects (34) and only included one study (pub-
lished between January 1990 and July model only in cases where the homogene- lished in 1997) using subcutaneous G-
2003. MeSH terms used were granulo- ity hypothesis was rejected (25,26). We CSF. One placebo-controlled trial (48)
cyte colonystimulating factor (or calculated both the study-specific and with granulocyte-macrophage colony
G-CSF) and diabetic foot. We supple- common 95% CIs by the method of stimulating factor examined its effect on
mented the computer search by reviewing Woolf (27) and used risk ratio (RR) as a healing uninfected ulcers. Thus, there
as many diabetic foot online websites and measure of the effect size. To calculate the were five prospective randomized studies
published bibliographies as we could number of patients who needed to be (8,3538) using G-CSF for infected dia-
find, hand searching the bibliographies treated to prevent one event, we assessed betic foot lesions that met the predefined
from the articles retrieved, reviewing rel- the pooled risk differences (28). inclusion criteria.
evant meeting abstracts, and asking study For continuous variables (e.g., neu- Table 1 summarizes the main ele-
authors and other experts in the field trophil count and duration of antibiotic ments of the protocol, patient character-
about any additional published or unpub- treatment), we used the weighted mean istics, and major outcomes of the five
lished studies on this topic. difference. The weight assigned to each included studies. In each study, the en-
study (i.e., how much influence it had on rolled patients were hospitalized for treat-
Study selection, quality assessment, the overall meta-analysis results) was de- ment. The details provided by the authors
and data extraction termined by the precision of its estimate on the clinical characteristics of the infec-
We included in our analysis only prospec- of effect, which is equal to the inverse of tions varied, but the described severity
tive randomized studies whose main pur- the variance. This method assumes that among the studies ranged from relatively
pose was to investigate the therapeutic all of the trials have measured the out- mild (36,38) to severe (35). Patients with
effects of G-CSF in diabetic foot infec- come on the same scale. sepsis, gangrene, or deep soft tissue infec-
tions. Studies were included only if they tion were generally not enrolled. Initial
compared the efficacy of standard treat- Assessment of publication bias and antibiotic therapy was apparently mostly
ment alone with that of standard treat- heterogeneity parenteral and in most studies not modi-
ment plus adjunctive G-CSF therapy. We To visually inspect for publication bias, fied by culture results. The specific regi-
assessed the quality of each trial with a we generated graphical funnel plots (29). mens and duration of therapy varied, but
scale developed by Jadad et al. (23) that The statistical methods used to detect in four studies, it consisted of a fluoro-
scores (from a low of zero to a high of five) funnel-plot asymmetry were the rank cor- quinolone (ciprofloxacin or ofloxacin)
the randomization, double blinding, and relation test of Begg and Mazumdar (30) combined with an antianaerobic drug
reports of dropouts and withdrawals. and the regression asymmetry test of (clindamycin or metronidazole). The in-
Data extracted from each study in- Egger et al. (29). Because the relative mer- clusion and exclusion criteria also varied,
cluded the following: clinical outcomes its of the two available methods are not but all required that the infections were
related to both curing the infection (reso- well established, we used both. severe enough to warrant hospitalization,
lution of cellulitis or other signs and We assessed the heterogeneity of re- and they were usually classified as Wag-
symptoms of infection) and healing of any sults of the studies by using the Cochrans ner grade 2 or 3 (49). Patients receiving
foot ulcer, the duration of antibiotic ther- Q test (31,32). This test, however, has a immunosuppressive therapy or with im-
apy (by any route) provided, the duration low power for detecting heterogeneity munosuppressive disorders, critical limb
of hospitalization, the need for any type of when the number of studies included in ischemia, hepatic or renal insufficiency,
lower-extremity amputation or other ma- the meta-analysis is small. Thus, we also or hematological disorders were excluded
jor surgical procedures, and the need for used the recently introduced quantity, I2 in each study.
vascular (arterial) surgery or angioplasty. (33), which is calculated from the usual The G-CSF preparation used was fil-
We also sought information on the test statistics and provides a less-biased grastim in four studies and lenograstim in
changes in blood leukocyte count and any measure of the degree of inconsistency one. It was administered subcutaneously
side effects of G-CSF treatment. Two re- across studies in a meta-analysis. There in four studies and intravenously in one.
viewers (M.C. and F.D.L.) independently was neither external funding nor any In each study, the G-CSF preparation was
examined the data and resolved disagree- sponsorship for this study. held, or its dose reduced, if the neutrophil
ments of interpretation by discussion. count increased beyond a previously set
When a publication had missing or RESULTS value. The duration of G-CSF therapy var-
incomplete information, we attempted to ied from 3 to 21 days.
contact the author(s). In three instances, Description of studies and The Jadad scores for quality of the
they provided additional data that we methodological quality studies ranged from 1 to 5; the mean was
added to our tables. Thus, in some in- Our literature search uncovered 17 arti- 3.4, and four trials had a score 3. Four
stances, the results presented in our tables cles (8,18,34 48) with information studies (8,35,37,38) reported conceal-
therapy. C, control patients; G, G-CSFtreated patients; Hosp, number of days hospitalized; Inf, number in whom infection resolved; NS, not statistically significant; Surg, number who required amputation
*Number of patients and number of hospital days. Sponsored by Amgen. Author provided additional unpublished details. Withdrawal of intravenous antibiotics, time to negative swab culture, diminution
of foot temperature, and number of vascular procedures. Number who completed the study in parentheses. Percent reduction in infection summary score. #Number with improved cellulitis on day 7 of
ment of treatment allocation, but only
Duration of intravenous
two described employing a double-blind
Number of vascular
vascular disease.
antibiotic NS
were similar.
Main results
eradication
A total of 167 patients were included in
the five randomized studies, 85 in the
G-CSFtreated group and 82 in the con-
trol group. There was no evidence of an
imbalance in baseline patient demo-
Inf: G 9, C 3, P NS#;
NS; Surg: G 2, C 3, P
P 0.02; Surg: G 0,
NS; Surg: G 1, C
C 4, P NS; Hosp:
Surg: G 3, C 9, P
Surg: G 0, C 3, P
Inf: G 77%, C 66%, P
8.8, P 0.02
Inf: G 7, C 12,
28, P NS
outcomes of interest, we can only present
1, P NS
them descriptively.
The meta-analysis showed that add-
0.02
0.04
G 15, C 15
G 10, C 10
Not blinded or
(randomized,
Study design
not placebo
Double blind,
Patient blind,
controlled
controlled
controlled
controlled
controlled
placebo
placebo
placebo
placebo
for 10 days
for 7 days
for 7 days
Wagner grades 23
Wagner grades 23
grade 2 lesion
(38), India
Viswanathan
Ynem (36),
Austria
Gough (8),
Turkey
U.K.
country
Figure 1 Pooled RR estimates and their 95% CIs for the outcomes amputation (A) and overall surgery (B). Studies are identified by name of
the first author. Size of squares is proportional to Mantel-Haenszel weighted risk ratio. *Cannot be computed because the presence of frequencies
equals zero.
25.24 109/l (9.57 40.92, P 0.002). tertrial heterogeneity for the outcomes the Beggs funnel plots for both outcomes
None of the studies reported any signifi- analyzed. Values of I2 (with their 95% un- of interest. The Begg-Mazudmar and
cant adverse effects of G-CSF therapy. certainty intervals) were 0% (0 53%) for Egger tests also showed no evidence of
amputation and 0% (0 30%) for overall publication bias (Beggs test: adjusted
Heterogeneity and publication bias surgical interventions, indicating no ob- Kendalls score 0, SD of score 2.94,
assessment served heterogeneity. There was also no z 0, p[z] 1, continuity corrected z
With the exception of the neutrophil evidence for publication bias, as shown in 0.34, p[z] 1; Eggers test: t 0.30,
count data, there was no evidence of in- Fig. 2, by the symmetrical appearance of p[t] 0.790).
Table 2Need for lower-extremity amputation and for overall invasive interventions (including amputations) for patients with diabetic foot
infections treated with standard treatment plus G-CSF versus standard treatment alone (control subjects)
CONCLUSIONS Conducting ther- major benefit to diabetic patients and to As with all meta-analyses, our conclu-
apeutic trials for a complex and serious their health care systems. sions can only be as accurate as the studies
problem like diabetic foot infections is This analysis has several limitations. from which they were based. Based on the
difficult, especially when investigating a
new adjunctive technology like G-CSF.
While our literature search uncovered
five randomized trials addressing this is-
sue, it is not surprising that none of the
studies enrolled more than 40 patients.
Considering the heterogeneous nature of
diabetic foot infections and the varied re-
search methods employed, it is difficult to
interpret the results of these individual
studies. Meta-analysis is the best way to
try to determine from the available data if
G-CSF therapy can help avert a poor out-
come in a diabetic patient with a foot
infection.
Our analysis found that adding G-
CSF to standard therapy did not appear to
benefit the primary outcome of interest,
i.e., increasing the likelihood of or hasten-
ing the time until resolution of infection.
Nor did G-CSF improve the healing of
foot wounds. It did, however, have other
beneficial effects. Not surprisingly, G-
CSF increased the leukocyte count in
each of the studies in which this was ex-
amined, but the clinical significance of
this finding is unknown. Treatment with
G-CSF was also associated with a ten-
dency toward a shorter duration of paren-
teral antibiotic therapy. If true, this could
help constrain antibiotic-associated ad-
verse effects, costs, and the development
of resistant organisms. More importantly,
G-CSF therapy was associated with a sta-
tistically significantly reduced risk of re-
quiring lower-extremity amputation as
well as other foot infectionrelated inva-
sive interventions. Because amputations
are among the most feared and expensive Figure 2 Beggs funnel plot with pseudo 95% CIs for the outcomes amputation (A) and
consequences of diabetic foot infections overall surgery (B). For each study (E), the natural logarithm of the odds ratio is plotted against
(50), reducing their incidence would be a its SE.
J Med 318:1728 1733, 1988 betic foot infection. Diabetes Obes Metab 45. Hafner J, Burg G:Dermatological aspects
29. Egger M, Davey Smith G, Schneider M, 3:332337, 2001 in prevention and treatment of the dia-
Minder C: Bias in meta-analysis detected 37. Kastenbauer T, Hornlein B, Sokol G, Irsi- betic foot syndrome. Schweiz Rundsch Med
by a simple, graphical test. BMJ 315:629 gler K: Evaluation of granulocyte-colony Prax 88:1170 1177, 1999
634, 1997 stimulating factor (filgrastim) in infected 46. Chantelau E, Kimmerle R: Granulocyte-
30. Begg CB, Mazumdar M: Operating char- diabetic foot ulcers. Diabetologia 46:27 colony-stimulating factor in diabetic foot
acteristics of a rank correlation test for 30, 2003 infection (Letter). Lancet 351:370, 1998
publication bias. Biometrics 50:1088 38. Viswanathan V, Mahesh U, Jayaraman M, 47. Baddour LM: Randomized prospective
1101, 1994 Shina K, Ramachandram A: Beneficial controlled trial of recombinant granulo-
31. Cochran WG: The effectiveness of adjust- role of granulocyte colony stimulating cyte colony-stimulating factor as adjunc-
ment by subclassification in removing factor in foot infection in diabetic pa- tive therapy for limb-threatening diabetic
bias in observational studies. Biometrics tients. J Assoc Physicians India 51:90 91, foot infection. Curr Infect Dis Rep 4:413
24:295313, 1968 2003 414, 2002
32. Wermuth N, Cochran WG: Detecting sys- 39. Bennett SP, Griffiths GD, Schor AM, Leese 48. Agrawal R, Agrawal S, Beniwal S, Joshi C,
tematic errors in multi-clinic observa- GP, Schor SL: Growth factors in the treat- Kochar D: Granulocyte-macrophage col-
tional data. Biometrics 35:683 686, 1979 ment of diabetic foot ulcers. Br J Surg 90: ony-stimulating factor in foot ulcers. The
33. Higgins J, Thompson S, Deeks J, Altman 133146, 2003 Diabetic Foot Summer:1 8, 2003
D: Measuring inconsistency in meta-anal- 40. Lazareth I: X. Local care and medical 49. Oyibo SO, Jude EB, Tarawneh I, Nguyen
yses. BMJ 327:557560, 2003 treatment of ischemic trophic disorders in HC, Harkless LB, Boulton AJ: A compari-
34. Mason J, OKeeffe C, Hutchinson A, the diabetic patients. J Mal Vasc 27:287 son of two diabetic foot ulcer classifica-
McIntosh A, Young R, Booth A: A system- 288, 2002; discussion 27:301283, 2002 tion systems: the Wagner and the
atic review of foot ulcer in patients with 41. Dale DC: Colony-stimulating factors for University of Texas wound classification
type 2 diabetes mellitus. II: treatment. the management of neutropenia in cancer systems. Diabetes Care 24:84 88, 2001
Diabet Med 16:889 909, 1999 patients. Drugs 62 (Suppl. 1):115, 2002 50. Lipsky BA: A report from the international
35. de Lalla F, Pellizzer G, Strazzabosco M, 42. Kreyden OP, Hafner J, Burg G, Nestle FO: consensus on diagnosing and treating the
Martini Z, Du Jardin G, Lora L, Fabris P, Case report on therapy with granulocyte infected diabetic foot. Diabetes Metab Res
Benedetti P, Erle G: Randomized prospec- stimulating factor in diabetic foot. Hau- Rev 20 (Suppl. 1):S68 S77, 2004
tive controlled trial of recombinant gran- tarzt 52:327330, 2001 51. Ashry HR, Lavery LA, Armstrong DG, La-
ulocyte colony-stimulating factor as 43. Edmonds M, Bates M, Doxford M, Gough very DC, van Houtum WH: Cost of dia-
adjunctive therapy for limb-threatening A, Foster A: New treatments in ulcer heal- betes-related amputations in minorities. J
diabetic foot infection. Antimicrob Agents ing and wound infection. Diabetes Metab Foot Ankle Surg 37:186 190, 1998
Chemother 45:1094 1098, 2001 Res Rev 16 (Suppl. 1):S51S54, 2000 52. Tennvall GR, Apelqvist J, Eneroth M:
36. Yonem A, Cakir B, Guler S, Azal OO, 44. Bakker K, Schaper NC: New developments Costs of deep foot infections in patients
Corakci A: Effects of granulocyte-colony in the treatment of diabetic foot ulcers. Ned with diabetes mellitus. Pharmacoeconom-
stimulating factor in the treatment of dia- Tijdschr Geneeskd 144:409 412, 2000 ics 18:225238, 2000