The role of diabetes mellitus in the treatment of skin and skin structure infections caused by methicillin-resistant Staphylococcus aureus: results from three randomized controlled trials

International Journal of Infectious Diseases 15 (2011) e140–e146

Contents lists available at ScienceDirect

International Journal of Infectious Diseases journal homepage: www.elsevier.com/locate/ijid

The role of diabetes mellitus in the treatment of skin and skin structure infections caused by methicillin-resistant Staphylococcus aureus: results from three randomized controlled trials§ Benjamin A. Lipsky a,b,*, Kamal M.F. Itani c, John A. Weigelt d, Warren Joseph e, Christopher M. Paap f, Arlene Reisman f, Daniela E. Myers f, David B. Huang f a

Medical Service, VA Puget Sound Health Care System, 1880 S. Columbian Way (S-111-GIMC), Seattle, WA 98108-1587, USA Division of General Internal Medicine, Department of Medicine, University of Washington School of Medicine, Seattle, Washington, USA Boston VA Health Care System and Boston University, Boston, Massachusetts, USA d Medical College of Wisconsin, Milwaukee, Wisconsin, USA e Roxborough Memorial Hospital, Philadelphia, Pennsylvania, USA f Pfizer Inc., New York, New York, USA b c

A R T I C L E I N F O

S U M M A R Y

Article history: Received 25 May 2010 Received in revised form 11 October 2010 Accepted 14 October 2010

Objective: To compare outcomes of treating complicated skin and skin structure infections (cSSSI) caused by methicillin-resistant Staphylococcus aureus (MRSA) with linezolid versus vancomycin in diabetic and non-diabetic patients. Methods: We pooled data from three prospective clinical trials in which 1056 patients were randomized to receive either linezolid (intravenous (IV) or oral) or vancomycin (IV) every 12 h, for 7–28 days. Results: Diabetic (n = 349) and non-diabetic patients (n = 707) had comparable demographics and comorbidities. Clinical success rates were lower in diabetic than in non-diabetic patients (72.3% and 85.8%, respectively). Overall, non-diabetic patients had a shorter adjusted mean length of stay (LOS) compared with diabetic patients (8.2 and 10.7 days, respectively; p < 0.0001). Among diabetic patients, rates were comparable with linezolid and vancomycin treatment for clinical success (74% and 71%, respectively) and microbiological success (60% and 54%, respectively). Among non-diabetic patients, clinical and microbiological success rates were higher in linezolid- than in vancomycin-treated patients (90% and 81%, respectively, and 78% and 65%, respectively). Rates of drug-related adverse events were comparable in diabetic and non-diabetic patients and with linezolid and vancomycin treatment. Adjusted mean LOS was shorter with linezolid than with vancomycin treatment in diabetic patients (9.5 and 11.7 days, respectively; p = 0.03) and non-diabetic patients (7.6 and 8.9 days, respectively; p = 0.02). Conclusions: Clinical success rates were lower in diabetic than non-diabetic patients with cSSSI caused by MRSA. Comparing linezolid and vancomycin, clinical and microbiological success rates were comparable in diabetic patients, but were better for linezolid than for vancomycin in non-diabetic patients. ß 2010 Published by Elsevier Ltd on behalf of International Society for Infectious Diseases.

Corresponding Editor: Hubert Wong, Vancouver, Canada Keywords: Complicated skin and skin structure infection Diabetes mellitus Linezolid Methicillin-resistant Staphylococcus aureus Vancomycin

1. Introduction Complicated skin and skin structure infections (cSSSI) are common in persons with diabetes1 and may be more difficult to treat than in those without diabetes. However, few studies have

§ This study was presented in part as a poster at the 47th Annual Meeting of the Infectious Diseases Society of America (IDSA), October 2009, Philadelphia, Pennsylvania, USA (poster #877). * Corresponding author. Tel.: +1 206 277 1640; fax: +1 206 764 2849. E-mail address: [email protected] (B.A. Lipsky).

compared outcomes of these infections in diabetic and nondiabetic populations. Staphylococcus aureus is consistently reported as the most commonly isolated pathogen in cSSSI, as well as in diabetic foot infections,1,2 and methicillin-resistant S. aureus (MRSA) strains have been isolated more frequently in many communities worldwide.1,3,4,5 Some, but not all, studies indicate that compared with other pathogens, MRSA infections are associated with lower infection cure rates, higher amputation rates, and increased mortality.5,6,7,8,9,10 The standard therapy for MRSA infections is vancomycin; however, because of limitations of this agent, several new anti-MRSA antibiotics have been developed. Among these, only linezolid can be administered orally as

1201-9712/$36.00 – see front matter ß 2010 Published by Elsevier Ltd on behalf of International Society for Infectious Diseases. doi:10.1016/j.ijid.2010.10.003

B.A. Lipsky et al. / International Journal of Infectious Diseases 15 (2011) e140–e146

well as parenterally. Almost all mild and many moderate cSSSI, including diabetic foot infections, can be treated on an outpatient basis; for this situation an oral agent would be preferable. To compare the efficacy and safety of linezolid and vancomycin in diabetic and non-diabetic patients with cSSSI caused by MRSA, we conducted a pooled analysis of data from three randomized controlled trials.11,12,13 Our secondary objective was to compare clinical and microbiological success rates and length of stay (LOS) in hospitalized patients with and without diabetes, who were treated with one of these two antibiotics for culture-confirmed MRSA cSSSI. 2. Patients and methods 2.1. Study design We analyzed data from three prospective randomized, openlabel, multicenter, phase 3b/4 clinical studies that examined the efficacy and safety of linezolid and vancomycin in diabetic and nondiabetic patients hospitalized for cSSSI (including foot infections), culture-confirmed to be caused by MRSA.11,12,13 The assessments of outcome in the original studies were blinded. The methods and overall results of the individual studies have been published in detail.11,12,13 In brief, eligible adults who provided institutional review board-approved informed consent were randomized in a 1:1 ratio to receive from 7 to 28 days (based on the judgment of the investigator) of linezolid 600 mg every 12 h (either intravenous (IV) or oral)11,12,13 or IV vancomycin (either 15 mg/kg every 12 h with dosages adjusted for creatinine clearance13 or 1 g every 12 h11,12). Investigators were given permission to adjust vancomycin dosages according to normal standards of care. In either treatment group, investigators were allowed to add treatment with aztreonam or gentamicin for suspected or proven infection with Gram-negative pathogens. Similarly, for coverage of obligate anaerobic pathogens, the investigators could add metronidazole. The duration of treatment with aztreonam, gentamicin and metronidazole was determined by the investigator. Investigators did not allow patients to receive any type of local antimicrobial therapy. 2.2. Patients In each of the three trials we identified enrolled patients with a diagnosis of diabetes mellitus based on their medical history and a review of their medications (seeking evidence of use of insulin or oral hypoglycemic agents). For each enrolled patient we also obtained selected data from their medical history, physical examination, vital signs, and baseline laboratory tests. Investigators obtained skin or soft tissue specimens for culture by incision and drainage (n = 281), fine-needle aspiration (n = 125), tissue biopsy (n = 76), tissue scraping (n = 67), tissue debridement (n = 45), wound swabs of purulent lesions (n = 424), or other (unspecified) methods (n = 38). The study protocols disallowed culture specimens obtained by a superficial swab from a relatively dry lesion with no purulence. Investigators obtained other types of culture specimens (e.g., blood) at baseline if they believed they were indicated. All specimens for culture were sent to the local microbiology laboratory for a Gram stained smear, aerobic and anaerobic cultures, and antibiotic susceptibility testing of all recovered pathogens. Among the patients with diabetes, we specifically identified the subset with a diabetic foot infection, as defined by the Infectious Diseases Society of America guidelines, i.e., purulence or at least two classical signs of inflammation of any inframalleolar lesion.1 2.3. Safety and efficacy evaluations To examine the safety of treatment with the two antibiotic agents we pooled data from patients in the intent-to-treat (ITT)

e141

populations, defined as patients with a cSSSI caused by any pathogen who received at least one dose of study medication (either linezolid or vancomycin). To examine the clinical and microbiological efficacy of the treatments, we pooled data from all patients in the modified ITT (MITT) populations, defined as patients in the ITT populations who had MRSA isolated as a causative pathogen at baseline. We assessed efficacy outcomes at the test-of-cure (TOC) visit, defined as 6–28 days after the last dose of study drug. We compared clinical and microbiological success rates in patients with and without diabetes and by treatment with linezolid versus vancomycin. We categorized clinical response at TOC in each study as one of three categories: ‘clinical success’, defined as resolution of all clinical signs and symptoms of infection that were identified at baseline; ‘failure’, defined as persistence or progression of clinical signs and symptoms of infection after receiving at least two days of treatment or development of new clinical findings consistent with active infection; or, ‘unknown’, defined as extenuating circumstances that precluded any other classification. We categorized microbiological outcome at TOC as one of the following: ‘documented eradication’, defined by a repeat culture from the original infection site with no MRSA isolated; ‘presumed eradication’, defined by having no culture data, but a clinical outcome defined as success; ‘documented persistence’, defined as a repeat positive culture for MRSA from the original infection site; ‘presumed persistence’, defined as no culture data available, but the clinical outcome was a failure; or ‘recurrence’, defined by a positive culture for MRSA from the original infection site at TOC when a culture was negative at the end of treatment. We excluded any patient who could not be classified from the efficacy analyses. A central laboratory provided the final pathogen identification and determined susceptibility to study drugs by a broth microdilution method according to the Clinical and Laboratory Standards Institute guidelines. 2.4. Length of stay evaluation We calculated LOS for patients in the MITT populations by the number of days in the hospital from the date of first dose of study drug to the date of discharge, independent of route of administration of study drug. Prior to this analysis, we determined that for patients who were hospitalized longer than the study period of 35 days, their LOS would be set to 35 days. We did this because the observation period of the clinical trial design was 35 days, and truncation enabled the analysis to be based on uniformly monitored data for all groups in all trials. As a result of truncation, the calculated mean LOS did not reflect the actual mean LOS, but was used solely to compare groups. Patients were excluded from this analysis if clinical outcome was unknown. Patients with missing clinical outcomes were excluded from the LOS analysis to maintain a uniform data analysis set consistent with the clinical portion of the ad hoc analysis. In the one study in which hospitalization was not required, LOS was set to 0 when patients were not hospitalized.12 2.5. Statistical analysis We conducted all analyses on final locked databases employing the Statistical Analysis System (SAS Institute, Cary, NC, USA) and used the Chi-square test or Fisher’s exact test to assess the association between treatment and categorical variables, and the F statistic to assess the association between treatment or diabetes category and continuous variables. To estimate the effect of study, we constructed separate logistic regression models for the diabetic and the non-diabetic populations. A logistic regression model was run that contained only indicator variables for treatment and

e142

Table 1 Baseline characteristics of diabetic and non-diabetic patients Diabetic LZD (n = 171)

VAN (n = 178)

59.8 (14.6) 103 (60.2) 68 (39.8)

58.6 (16.0) 101 (56.7) 77 (43.3)

127 (74.3) 23 (13.5) 3 (1.8) 10 (5.8) 8 (4.7) 93.9 (31.5) 12.7 (4.8)

LZD (n = 356)

VAN (n = 351)

Total (n = 707)

59.2 (15.3) 204 (58.5) 145 (41.5)

47.4 (18.8) 213 (59.8) 143 (40.2)

47.3 (18.6) 229 (65.2) 122 (34.8)

47.4 (18.7) 442 (62.5) 265 (37.5)

11.8 (9.5, 14.1) 4.0 ( 10.4, 2.2) 4.0 ( 2.2, 10.4)

133 (74.7) 23 (12.9) 7 (3.9) 7 (3.9) 8 (4.5) 90.3 (28.5) 12.0 (5.2)

260 (74.5) 46 (13.2) 10 (2.9) 17 (4.9) 16 (4.6) 92.1 (30.0) 12.3 (5.0)

257 (72.2) 53 (14.9) 9 (2.5) 25 (7.0) 12 (3.4) 80.8 (21.0) 11.7 (4.7)

234 (66.7) 61 (17.4) 8 (2.3) 35 (10.0) 13 (3.7) 79.9 (25.6) 9.9 (4.6)

491 (69.5) 114 (16.1) 17 (2.4) 60 (8.5) 25 (3.5) 80.3 (23.4) 10.8 (4.8)

5.1 ( 0.7, 10.8) 2.9 ( 7.4, 1.5) 0.5 ( 1.6, 2.5) 3.6 ( 6.7, 0.6) 1.1 ( 1.5, 3.6) 11.7 (8.4, 15.0) 1.5 (0.9, 2.2)

114 (66.7)

1814.0 (724.1) 111 (62.4)

N/A 225 (64.5)

169 (47.5)

1950.8 (636.1) 171 (48.7)

N/A 340 (48.1)

136.8 ( 287.5, 13.9) 16.4 (10.2, 22.6)

80 (46.8)

67 (37.6)

147 (42.1)

112 (31.5)

119 (33.9)

231 (32.7)

9.4 (3.2, 15.7)

21 (12.3) 83 (48.5) 47 (27.5) 20 (11.7) 178 (91.5)

27 (15.2) 70 (39.3) 61 (34.3) 20 (11.2) 188 (92.6)

48 (13.8) 153 (43.8) 108 (30.9) 40 (11.5) 183 (92.0)

0 0 0 356 (100) 106 (28.8)

0 0 0 351 (100) 105 (24.0)

0 0 0 707 (100) 105 (26.6)

21 (12.3) 45 (26.3) 58 (33.9) 37 (21.6) 10 (5.9)

19 (10.7) 58 (32.6) 47 (26.4) 39 (21.9) 15 (8.4)

40 (11.5) 103 (29.5) 105 (30.1) 76 (21.8) 25 (7.2)

51 (14.3) 171 (48.0) 31 (8.7) 71 (19.9) 32 (9.0)

53 (15.1) 170 (48.4) 27 (7.7) 66 (18.8) 35 (10.0)

104 (14.7) 341 (48.2) 58 (8.2) 137 (19.4) 67 (9.5)

3.3 ( 7.5, 0.1) 18.7 ( 24.8, 1.3) 21.9 (16.7, 27.1) 2.4 ( 2.8, 7.6) 2.3 ( 5.8, 11.5)

68 (39.8) 1 (0.6) 8 (4.7) 46 (26.9) 95 (55.6)

68 (38.2) 3 (1.7) 5 (2.8) 48 (27.0) 128 (71.9)

136 (39.0) 4 (1.2) 13 (3.7) 94 (26.9) 223 (63.9)

71 (19.9) 6 (1.7) 13 (3.7) 36 (10.1) 114 (32.0)

58 (16.5) 6 (1.7) 15 (4.3) 27 (7.7) 97 (27.6)

129 (18.3) 12 (1.7) 28 (4.0) 63 (8.9) 211 (29.8)

20.7 (14.9, 26.6) 0.6 ( 2.0, 0.9) 0.2, ( 2.7, 2.2) 18.0 (12.9, 23.1) 34.1 (28.0, 40.1)

178 (100)

266 (76.2) 83 (23.8)

196 (55.1) 160 (44.9)

351 (100)

547 (77.4) 160 (22.6)

88 (51.5) 83 (48.5)

0

Total (n = 349)

CI, confidence interval; LZD, linezolid; N/A, not applicable; SD, standard deviation; VAN, vancomycin.

0

Difference (95% CI) (diabetic vs. non-diabetic)

N/A

77.9 (68.5, 87.4)

N/A

B.A. Lipsky et al. / International Journal of Infectious Diseases 15 (2011) e140–e146

Age, years, mean (SD) Male, n (%) Female, n (%) Race, n (%) White Black Asian Other Not listed Weight, kg, mean (SD) Treatment duration, days, mean (SD) Mean (SD) VAN dose, mg/day Concomitant Gram-negative coverage, n (%) Concomitant anaerobic coverage, n (%) Diabetic medication, n (%) Oral hypoglycemics Insulin Oral hypoglycemics + insulin None Baseline serum glucose level, mg/dl, mean (SD) Baseline diagnosis, n (%) Cellulitis Abscess Infected skin ulcer Infected surgical incision Other Co-morbidities, n (%) Cardiac Hepatobiliary Oncologic Renal Vascular Route of initial treatment, n (%) Intravenous Oral

Non-diabetic

B.A. Lipsky et al. / International Journal of Infectious Diseases 15 (2011) e140–e146

e143

Table 2 Isolated pathogens from diabetic and non-diabetic patients Isolated pathogens, n (%)

Diabetic patients LZD

VAN

LZD

VAN

MRSA MRSA MRSA MRSA MRSA

119 (69.6) 30 (17.5) 14 (8.2) 6 (3.5) 2 (1.2)

124 (69.7) 26 (14.6) 14 (7.9) 13 (7.3) 1 (0.6)

285 (80.1) 34 (9.6) 27 (7.6) 8 (2.2) 2 (0.6)

294 (83.8) 21 (6.0) 26 (7.4) 9 (2.6) 1 (0.3)

+ + + +

other Gram-positive Gram-negative other Gram-positive + Gram-negative anaerobes

Non-diabetic patients

LZD, linezolid; MRSA, methicillin-resistant Staphylococcus aureus; VAN, vancomycin.

study. A sensitivity analysis consisted of a full logistic regression model that included variables for treatment as well as factors that were imbalanced by treatment group at baseline. To account for differences by study, we used the adjusted odds ratio (OR) to compare clinical and microbiological success rates between the two treatment groups, for diabetic and non-diabetic patients separately and for the total population. We estimated 95% confidence intervals (CIs) for the adjusted OR and considered intervals that excluded 1.00 as statistically significant. For categorical analyses we considered p < 0.05 as statistically significant. We made no statistical adjustment for multiple comparisons. While we excluded all outcomes classified as missing, unknown, or indeterminate in the calculations, we present them for information purposes. Patients with missing, unknown, or indeterminate clinical outcomes were excluded from the LOS analysis to maintain a uniform data analysis set with the clinical portion of the ad hoc analysis. The number of MITT patients who had clinical outcome missing, unknown, or indeterminate and were therefore not included in any of the LOS analyses (which was the same data set for clinical and microbiological analyses) were: 57 diabetic patients (linezolid 31 and vancomycin 26) and 138 non-diabetic patients (linezolid 62 and vancomycin 76). The outcomes research analysis of LOS incorporated study as a class variable using the same coding as used in the clinical and microbiological analyses in order to adjust for study effect. LOS was assessed using analysis of covariance in a model that compared treatment effect in the diabetic and non-diabetic patients separately, as well as for the total population, adjusted for the study effect. Least square (LS) means were used to evaluate the treatment or diabetes status effect size. 3. Results 3.1. Patient characteristics Combining the three studies, a total of 868 diabetic patients received at least one dose of either linezolid (n = 451) or vancomycin (n = 417), and a total of 1688 non-diabetic patients received at least one dose of either linezolid (n = 841) or vancomycin (n = 847). These patients combined constituted the ITT population. In the ITT population, MRSA was isolated on culture at baseline in 349 (40.2%) of the diabetic patients, including 171 treated with linezolid and 178 treated with vancomycin. Similarly, MRSA was isolated on culture at baseline in 707 (41.9%) of the non-diabetic patients, including 356 treated with linezolid and 351 treated with vancomycin. These patients constituted the MITT population and we have summarized their baseline characteristics in Table 1. Diabetic patients in the linezolid and vancomycin treatment groups were comparable with respect to the distribution of their age, sex, race, weight, diabetic medications, baseline cSSSI diagnosis, and co-morbidities (with the exception of vascular disease, which was less common in the linezolid treatment group).

Compared with non-diabetic patients, diabetic patients were: older; heavier; more frequently had underlying cardiac, renal, and vascular co-morbidities; presented more commonly with an infected skin ulcer; and more often received additional Gramnegative coverage. Approximately 90% of the diabetic patients were receiving oral hypoglycemic medications, an insulin-containing regimen, or both. In the pooled analyses of patients in the MITT population who received linezolid, 49% of diabetic patients and 45% of non-diabetic patients received the drug by the oral route as initial treatment. There was no significant difference in the vancomycin daily dosing between diabetic and non-diabetic patients. 3.2. Microbiological characteristics Table 2 summarizes the baseline pathogens isolated in the MITT population with and without diabetes. MRSA was identified as the sole pathogen less frequently in patients with diabetes than in those without diabetes. Mixed infections with other Gram-positive and Gram-negative pathogens were more frequent in patients with diabetes. Anaerobes were identified infrequently in both treatment groups, regardless of diabetes status. Overall, there were no significant differences in distributions of baseline pathogens between patients in the linezolid and vancomycin groups when stratified by presence or absence of diabetes. 3.3. Efficacy Table 3 summarizes the clinical and microbiological outcomes in the MITT population by patient diabetes status. Overall, the rates of clinical success were lower for diabetic patients compared with non-diabetic patients (72.3% and 85.8%, respectively; absolute difference 13.5%; adjusted OR 0.4, 95% CI 0.3–0.6). In diabetic patients, those treated with linezolid and vancomycin had similar rates of clinical success (73.6% and 71.1%, respectively; absolute difference 2.5%; adjusted OR 1.2, 95% CI 0.7–2.0) and microbiological success (60.4% and 54.2%, respectively; absolute difference 6.2%; adjusted OR 1.3, 95% CI 0.8–2.0). In a sensitivity analysis of the diabetic patients in the MITT population in which all missing, unknown, and indeterminate clinical outcomes were set to failure, linezolid was found to have a clinical success rate comparable to vancomycin (59% vs. 60%; 95% CI –11.7% to 8.6%) and linezolid was found to have a microbiological success rate comparable to vancomycin (54% vs. 51%; 95% CI – 7.3% to 14.2%). In patients without diabetes, linezolid treatment resulted in significantly higher clinical success rates (90.1% and 81.1%, respectively; absolute difference 9.0%; adjusted OR 2.1, 95% CI 1.3–3.5) and microbiological success rates (77.8% and 64.9%, respectively; absolute difference 12.9%; adjusted OR 1.9, 95% CI 1.3–2.7) compared with vancomycin. In a sensitivity analysis of the non-diabetic patients in the MITT population in which we set all missing, unknown, and indeterminate clinical outcomes to failure, treatment with linezolid compared to vancomycin had a significantly higher clinical success rate (74% vs. 62%; 95% CI 4.7–18.1%)

47 46 11 55 13 42 14 45 24 15 11 40 7 16 14 30 23 31 NA 1 6 26 NA 0 NA NA NA NA

LZD, linezolid; MITT, modified intent-to-treat; MRSA, methicillin-resistant Staphylococcus aureus; NA, not applicable; VAN, vancomycin. a MITT population, defined as the ITT population who had isolated MRSA at baseline. b One study did not report microbiological outcomes.11 c Excluded from analysis, includes missing or non-classifiable data.

NA NA NA NA

NA NA NA

16 12 9 13 18 11 NA 0 8 12 NA 4

21 15 5 12

NA NA NA NA 39 26 5 25 24 24 9 27

NA NA NA NA

56 136 104 (35) 75 167 69 (22) NA NA NA

2 (22) 1 NA

30 22 37 (42) 6 30 29 (45) 9 26 20 (36)

21 28 41 (46)

NA NA NA 27 56 70 (46) 39 48 57 (40)

NA NA NA

29 60 32 (26)

15 51 54 (45)

46 107 37 (20)

41 85 50 (28)

52 (19) 76 192 (65) 29 (10) 62 242 (78) 0 3 NA

7 (78)

Clinical success, n (%) Failure, n (%) Unknown, nc Microbiological success, n (%) Documented eradication Presumed eradication Microbiological failure, n (%) Documented persistence Presumed persistence Recurrence Indeterminatec

29 (33) 14 52 (58) 10 (19) 11 36 (55) 6 (14) 16 35 (64)

34 (38) 12 49 (54)

0 4 NA 44 (29) 26 83 (54) 37 (26) 31 87 (60)

0 4 NA

5 (5) 26 89 (74)

16 (18) 32 66 (55)

24 (13) 32 153 (81)

36 (20) 40 126 (72)

223 (81) 265 (90) 108 (71) 8 (100)

LZD (n = 10) MITT populationa,b

59 (67) 44 (82) 37 (86)

56 (62)

103 (74)

11 (100)

10 (100)

90 (95)

73 (82)

164 (87)

140 (80)

VAN (n = 351) LZD (n = 220) LZD (n = 121) VAN (n = 14) LZD (n = 15) VAN (n = 178) LZD (n = 171) LZD (n = 102) VAN (n = 65) LZD (n = 59)

VAN (n = 102)

Ref. 11 Total Ref. 13

VAN (n = 11)

and microbiological success rate (71% vs. 57%; 95% CI 6.5–20.7%). A sensitivity analysis that adjusted efficacy outcomes for factors that were imbalanced at baseline gave similar results. 3.4. Safety

Ref. 12 Ref. 11

Diabetics

Table 3 Clinical and microbiological outcomes among patients by diabetes status and type of antibiotic therapy

Non-diabetics

Ref. 12

VAN (n = 121)

Ref. 13

VAN (n = 216)

LZD (n = 356)

B.A. Lipsky et al. / International Journal of Infectious Diseases 15 (2011) e140–e146

Total

e144

Table 4 summarizes adverse events that occurred among diabetic and non-diabetic patients in the ITT population. The percentage of patients with 1 adverse event was nearly identical in both treatment groups in the diabetic patients, and the rates were higher than in non-diabetic patients. The trend was the same in patients with 1 serious adverse event. The percentage of patients with 1 drug-related adverse event was nearly identical for those treated with linezolid and vancomycin. The percentage of diabetic patients who discontinued study treatment due to an adverse event was similar with linezolid and vancomycin therapy. Table 5 lists the most commonly reported drug-related adverse events (DRAE) by diabetes status and type of antibiotic therapy. Among diabetic patients receiving linezolid, the most common reported DRAEs were diarrhea (5%), nausea (5%), and vomiting (2%), while the most commonly reported DRAEs among diabetic patients receiving vancomycin were diarrhea (2%), nausea (2%), kidney failure or impairment (2%), and IV catheter site complications (2%). Among non-diabetic patients, the most commonly reported DRAEs for linezolid were diarrhea (5%), nausea (5%), headache (2%), vomiting (2%), and anorexia (2%), and the most commonly reported DRAEs for vancomycin were rash (5%), IV catheter site complications (3%), and nausea (2%). We compared changes in laboratory values from baseline in patients in the ITT population (not shown). The percentages of diabetic patients who developed thrombocytopenia or anemia (defined as values < 75% of the lower limit of normal) were 6% and 1%, respectively, with linezolid, and 1% and 1%, respectively, with vancomycin. Among non-diabetic patients, the percentages with thrombocytopenia and anemia were 3% and 1% with linezolid, and 1% and 1% with vancomycin, respectively. 3.5. Subset analysis of diabetic foot infections A total of 97 patients (57 in the linezolid group and 40 in the vancomycin group) met classification criteria for a diabetic foot infection and had a skin or soft tissue culture that yielded MRSA. Baseline demographic and clinical characteristics of these patients were similar to those of the cohort of diabetic patients with a nonfoot skin infection. For patients with a diabetic foot infection, clinical success rates were 76% with linezolid and 73% with vancomycin (adjusted OR 1.3, 95% CI 0.5 3.6), and microbiological success rates were 60% in each group (adjusted OR 1.0, 95% CI 0.4 2.5). 3.6. Length of stay There were 16 patients (five treated with linezolid and six with vancomycin among diabetic patients; two treated with linezolid and three with vancomycin among non-diabetic patients) who had a LOS greater than 35 days, but who were assigned a LOS of 35 days for analysis purposes. Overall, non-diabetic patients had a shorter adjusted mean LOS compared with diabetic patients (8.2 vs. 10.7 days, respectively; p < 0.0001; adjusted median LOS 5.0 vs. 8.0 days; respectively; p < 0.0001). Among diabetic patients, the adjusted mean LOS was shorter in linezolid-treated patients than in vancomycin-treated patients (9.5 vs. 11.7 days, respectively; p = 0.03; adjusted median LOS 7.0 vs. 8.0 days, respectively; p = 0.03). Similarly, among nondiabetic patients, the adjusted mean LOS was shorter in linezolidtreated patients than in vancomycin-treated patients (7.6 vs. 8.9

B.A. Lipsky et al. / International Journal of Infectious Diseases 15 (2011) e140–e146

e145

Table 4 Safety outcomes among patients by diabetes status and type of antibiotic therapy ITT populationa

1 1 1 1

Diabetic patients

AE, n (%) serious AE, n (%) drug-related AE, n (%) AE leading to study drug discontinuation, n (%)

Non-diabetic patients

LZD (n = 451)

VAN (n = 417)

LZD (n = 841)

VAN (n = 847)

256 (57) 79 (18) 88 (20) 26 (6)

239 (57) 57 (14) 83 (20) 19 (5)

421 (50) 69 (8) 191 (23) 21 (3)

414 (49) 55 (7) 188 (22) 40 (5)

AE, adverse event; cSSSI, complicated skin and skin structure infection; ITT, intent-to-treat; LZD, linezolid; VAN, vancomycin. a ITT population was defined as all patients with cSSSI who received at least one dose of study medication.

Table 5 Reported adverse events among patients by diabetes status and type of antibiotic therapy Adverse events, n (%)

Diabetic patients LZD (n = 451)

Diarrhea Headache Nausea Vomiting Anorexia Kidney failure or impairment Rash IV catheter site complications

23 (5) 6 (1) 22 (5) 8 (2) 2 (0.4) 2 (0.4) 1 (0.2) 1 (0.2)

Non-diabetic patients VAN (n = 417) 8 (2) 4 (1) 9 (2) 5 (1) 2 (0.5) 7 (2) 4 (1) 7 (2)

LZD (n = 841) 45 (5) 13 (2) 45 (5) 13 (2) 14 (2) 0 9 (1) 6 (1)

VAN (n = 847) 6 (1) 9 (1) 17 (2) 4 (0.5) 3 (0.4) 7 (1) 38 (5) 22 (3)

IV, intravenous; LZD, linezolid; VAN, vancomycin.

days, respectively; p = 0.02; adjusted median LOS 4.0 vs. 6.0 days, respectively; p = 0.02). Because six patients (three on vancomycin and three on linezolid) from one study12 did not require hospitalization, we reran the LOS analyses excluding these patients. Non-diabetic patients had a shorter adjusted mean LOS compared with diabetic patients (8.2 vs. 10.8 days, respectively; p < 0.0001). Among diabetic patients, the adjusted mean LOS was shorter in linezolidtreated patients than in vancomycin-treated patients (9.6 vs. 11.8 days, respectively; p = 0.02). Similarly, among non-diabetic patients, the adjusted mean LOS was shorter in linezolid-treated patients than in vancomycin-treated patients (7.6 vs. 9.6 days, respectively; p = 0.03). 4. Discussion The availability of data from three randomized controlled trials allowed us to explore the important question of whether outcomes of treatment of cSSSI caused by MRSA were different in patients with diabetes compared with non-diabetic patients. These data also enabled us to compare the clinical and microbiological outcomes of these infections when treated with IV vancomycin and either IV or oral linezolid. In this large group of patients we found that both the clinical and microbiological success rates were significantly lower in diabetic patients than in non-diabetic patients, despite the fact that diabetic patients more often received the optional additional antibiotic coverage for Gram-negative pathogens (diabetic patients 64% and non-diabetic patients 48%) and for anaerobic pathogens (diabetic patients 42% and nondiabetic patients 33%). While the clinical and microbiological success rates for patients treated with vancomycin and linezolid were similar for the patients with diabetes, they were significantly better for non-diabetic patients who were treated with linezolid than with vancomycin. It is also noteworthy that in almost half of enrolled patients randomized to linezolid therapy, investigators elected to prescribe initial therapy with the drug orally, rather than intravenously. There are several possible explanations for the outcome differences in diabetic versus non-diabetic patients. To begin

with, diabetes is associated with various immune defects,14,15 including decreased T-cell-mediated immunity and neutrophil dysfunction.16 Furthermore, as noted in our study, people with diabetes have a higher rate of co-morbidities, including cardiac abnormalities, renal insufficiency, and peripheral vascular disease. Many studies have shown that diabetic patients are prone to peripheral arterial atherosclerotic disease, small-vessel dysfunction, and, perhaps, platelet dysfunction, each of which may result in decreased blood flow to peripheral tissues. This decreased perfusion could lead to decreased drug penetration to infected sites. Some studies have reported decreased tissue penetration of both linezolid and vancomycin in patients with diabetes.17,18,19 The higher body weight of patients with diabetes, as noted in our study, may result in lower relative doses of an antibiotic that is not adjusted for body weight. Another factor we noted is that patients with diabetes had more infected ulcers and fewer abscesses than non-diabetic patients. The latter are generally easier to cure, since surgical drainage is a major part of their treatment. In this study, the percentage of patients with adverse events, including those that were serious, was higher in diabetic than in non-diabetic patients in both treatment groups. The overall safety profiles of linezolid and vancomycin were similar by diabetes status. The percentage of diabetic patients with adverse events, including serious and drug-related, was comparable for the two treatment groups. Thrombocytopenia was uncommon and occurred most often in diabetic and non-diabetic patients treated with linezolid. Renal failure or impairment and catheter-related adverse events were observed more frequently in vancomycintreated patients. The adjusted mean LOS for treating these cSSSI was approximately 2–3 days longer for patients with diabetes compared with non-diabetic patients. This difference may be a result of the higher failure rate of treatment or the increased rate of complications and co-morbid diseases associated with diabetes. The adjusted mean LOS was significantly shorter among patients treated with linezolid compared with vancomycin, regardless of their diabetes status. This difference is likely because linezolid has an oral stepdown option, which may allow earlier patient discharge from the hospital.

e146

B.A. Lipsky et al. / International Journal of Infectious Diseases 15 (2011) e140–e146

The limitations of this evaluation include those inherent in all retrospective pooled analyses. Simple pooling of patients from the three trials could lead to underestimation of standard errors because it ignores potential heterogeneity across studies (when generalizing beyond these three studies); however, our analyses were based on three prospective randomized clinical trials with predetermined primary endpoints of clinical and microbiological cure rates. The studies had a different mix of infection types, with cellulitis more common in the earlier studies11,12 and abscesses more prevalent in the more recent study.13 Vancomycin dosing was different in the most recent study (15 mg/kg every 12 h) compared with the other two (1 g every 12 h). We lacked data on glycemic control for the diabetic patients, a potentially important confounder, although 90% of patients were receiving a diabetes medication and 40% to 50% were using insulin. In summary, in this pooled analysis of three randomized controlled trials of patients with cSSSI caused by MRSA, clinical and microbiological success was significantly lower in patients with diabetes compared with those without diabetes. Among patients with diabetes, clinical and microbiological success rates were comparable for diabetic patients treated with either linezolid or vancomycin, but both were significantly higher with linezolid treatment among non-diabetic patients. LOS was shorter among non-diabetic patients, and was statistically significantly shorter among both diabetic and non-diabetic patients who received linezolid compared with vancomycin. Despite the limitations of our study, this analysis provides useful data on the treatment of cSSSI caused by MRSA by patient diabetes mellitus status. Conflict of interest statement B.A.L. has been a consultant to, a speaker for, or has received research support from Pfizer, Ortho-McNeil Janssen, Cubist, and Merck. K.M.F.I. has received research support from Pfizer, Merck, and Sanofi-aventis, and has been a consultant to and speaker for Pfizer. J.A.W. has been a consultant to, a speaker for, and has received research support from Pfizer. W.J. is a consultant to and speaker for Pfizer and Merck. C.M.P., A.R., D.E.M., and D.B.H. are employees of Pfizer Inc. Acknowledgements This study was sponsored by Pfizer Inc. All authors collaborated in the development of the study design, in the collection, analysis, and interpretation of study data, in the writing of the manuscript, and in the decision to submit the manuscript for publication. Editorial support for this manuscript (technical writing) was provided by E.M. Wells of PAREXEL and was funded by Pfizer Inc.

References 1. Lipsky BA, Berendt AR, Deery HG, Embil JM, Joseph WS, Karchmer AW, et al. Infectious Diseases Society of America. Diagnosis and treatment of diabetic foot infections. Clin Infect Dis 2004;39:885–910. 2. American Diabetes Association. Diabetes basics: diabetes statistics; 2007. Available at: http://www.diabetes.org/diabetes-basics/diabetes-statistics/. (accessed May 5, 2010). 3. Dang CN, Prasad YD, Boulton AJ, Jude EB. Methicillin-resistant Staphylococcus aureus in the diabetic foot clinic: a worsening problem. Diabet Med 2003;20:159–61. 4. Tentolouris N, Jude EB, Smirnof I, Knowles EA, Boulton AJ. Methicillin-resistant Staphylococcus aureus: an increasing problem in a diabetic foot clinic. Diabet Med 1999;16:767–71. 5. Mantey I, Hill RL, Foster AV, Wilson S, Wade JJ, Edmonds ME. Infection of foot ulcers with Staphylococcus aureus associated with increased mortality in diabetic patients. Commun Dis Public Health 2000;3:288–90. 6. Martı´nez-Go´mez Dde A, Ramı´rez-Almagro C, Campillo-Soto A, Morales-Cuenca G, Paga´n-Ortiz J, Aguayo-Albasini JL. [Diabetic foot infections. Prevalence and antibiotic sensitivity of the causative microorganisms.]. Enferm Infecc Microbiol Clin 2009;27:317–21. 7. Cosgrove SE, Sakoulas G, Perencevich EN, Schwaber MJ, Karchmer AW, Carmeli Y. Comparison of mortality associated with methicillin-resistant and methicillin-susceptible Staphylococcus aureus bacteremia: a meta-analysis. Clin Infect Dis 2003;36:53–9. 8. Lodise TP, McKinnon PS. Clinical and economic impact of methicillin resistance in patients with Staphylococcus aureus bacteremia. Diagn Microbiol Infect Dis 2005;52:113–22. 9. Arago´n-Sa´nchez J, La´zaro-Martı´nez JL, Quintana-Marrero Y, Herna´ndez-Herrero MJ, Garcı´a-Morales E, Cabrera-Galva´n JJ, et al. Are diabetic foot ulcers complicated by MRSA osteomyelitis associated with worse prognosis? Outcomes of a surgical series. Diabet Med 2009;26:552–5. 10. Falanga V. Wound healing and its impairment in the diabetic foot. Lancet 2005;366:1736–43. 11. Stevens DL, Herr D, Lampiris H, Hunt JL, Batts DH, Hafkin B. Linezolid versus vancomycin for the treatment of methicillin-resistant Staphylococcus aureus infections. Clin Infect Dis 2002;34:1481–90. 12. Weigelt J, Itani K, Stevens D, Lau W, Dryden M, Knirsch C, Linezolid cSSTI Study Group. Linezolid versus vancomycin in treatment of complicated skin and soft tissue infections. Antimicrob Agents Chemother 2005;49:2260–6. 13. Itani KM, Dryden MS, Bhattacharyya H, Kunkel MJ, Baruch AM, Weigelt JA. Efficacy and safety of linezolid versus vancomycin for the treatment of complicated skin and soft-tissue infections proven to be caused by methicillinresistant Staphylococcus aureus. Am J Surg 2010;199:804–16. 14. Joshi N, Caputo GM, Weitekamp MR, Karchmer AW. Infections in patients with diabetes mellitus. N Engl J Med 1999;341:1906–12. 15. Geerlings SE, Hoepelman AIM. Immune dysfunction in patients with diabetes. FEMS Immunol Med Microbiol 2000;26:259–65. 16. Goren I, Ka¨mpfer H, Mu¨ller E, Schiefelbein D, Pfeilschifter J, Frank S, Oncostatin M. expression is functionally connected to neutrophils in the early inflammatory phase of skin repair: implications for normal and diabetes-impaired wounds. J Invest Dermatol 2006;126:628–37. 17. Skhirtladze K, Hutschala D, Fleck T, Thalhammer F, Ehrlich M, Vukovich T, et al. Impaired target site penetration of vancomycin in diabetic patients following cardiac surgery. Antimicrob Agents Chemother 2006;50:1372–5. 18. Stein GE, Schooley S, Peloquin CA, Missavage L, Havlichek DH. Linezolid tissue penetration and serum activity against strains of methicillin-resistant Staphylococcus aureus with reduced vancomycin susceptibility in diabetic patients with foot infections. J Antimicrob Chemother 2007;60:819–23. 19. Majcher-Peszynska J, Haase G, Sass M, Mundkowski R, Pietsch A, Klammt S, et al. Pharmacokinetics and penetration of linezolid into inflamed soft tissue in diabetic foot infections. Eur J Clin Pharmacol 2008;64:1093–100.