Study design
A retrospective chart review of eligible patients was conducted. Patients were identified for eligibility using the Stewardship Program Integrating Resource Information Technology (SPIRIT) database of our Antimicrobial Stewardship Program [6]. A SPIRIT query generated a list of inpatients at SHSC between March 12th, 2010 and December 31st, 2015, inclusive, with at least one steady state vancomycin trough concentration. Hospital charts for the patients identified by this query were retrieved from Health Data Resources (HDR) and reviewed to confirm patient eligibility for this study.
Patient eligibility
Adult inpatients (aged ≥18 years) were eligible for study inclusion if they were admitted to SHSC between March 12th, 2010 and December 31st, 2015, received a minimum of 48 h and a maximum of 14 days of vancomycin for a presumed or confirmed non-deep seated infection (defined below) associated with any bacteria for which vancomycin may be indicated, and for whom at least one steady state vancomycin trough concentration was available. A steady state vancomycin trough concentration was defined as a level obtained at the earliest prior to the 3rd dose for ≥ every 12 h dosing, and prior to the 4th dose for ≤ every 8 h dosing; or a random vancomycin concentration obtained at least 24 h after vancomycin initiation in patients receiving continuous infusion. An exception was made for inclusion of patients receiving a maximum of 15 days of vancomycin, if the treating physician’s intent was to treat for 14 days. Patients were eligible for inclusion if their infection-related diagnosis was:
-
Uncomplicated skin and soft tissue infections (SSTI): folliculitis, carbuncle(s), surgical site infections, wound infections, non-suppurative cellulitis or erysipelas;
-
Urinary tract infections (UTI) in the absence of anatomic abnormalities, renal abscess, or renal stones;
-
Bacteremia without any evidence of seeding to heart, bone/joint, brain, or lungs; or
-
Coagulase negative Staphylococci (CNST) line-related infections (excluding tunnel infection or vascular graft infections).
If patients received more than one course of vancomycin during their hospital stay, only data from their first vancomycin course was included, with documentation of any subsequent courses of vancomycin within 14 days of vancomycin discontinuation or patient readmission within 30 days of completion of their initial vancomycin treatment course. Subsequent vancomycin treatment courses and patient readmission were considered markers of potential relapsing infection (clinical failure) with a risk of development of a vancomycin resistant bacterial isolate. Patients were excluded if there was a presumed or documented diagnosis of abscess at any site, endocarditis, meningitis, osteomyelitis, joint infection, febrile neutropenia, pneumonia, or sinusitis; were switched to an alternate antibiotic on day 2 or 3 of vancomycin due to culture and sensitivity results; received antibiotics for no documented infection (e.g. prophylactic antibiotics); received renal replacement therapy; or if vancomycin was discontinued as a result of a patient care plan that was changed to palliative.
Data collection and definitions
Required patient clinical and laboratory data were extracted from the hospital’s electronic databases (SPIRIT, electronic patient records [EPR/Sunnycare]) and patient medical records (HDR/Sovera) and entered into a Microsoft Excel workbook. Patients were stratified to vancomycin low or high trough cohorts based on their final documented vancomycin steady-state trough concentration (≤10 mg/L or > 10 mg/L, respectively). The primary outcome of this study was clinical cure, defined as all of the following:
-
Resolution of all presenting signs and symptoms of infection within ≤14 days therapy with vancomycin; whereby patients could be discharged home on a short course of vancomycin (total duration ≤14 days) with documentation of improvement during hospital stay;
-
Maintained resolution of all presenting signs and symptoms of infection for 14 days following vancomycin discontinuation;
-
No additional course of antibiotics within 14 days with the same indication as initial vancomycin treatment course;
-
No documentation of hospital readmission requiring antibiotic therapy within 30 days of completion of their initial vancomycin treatment course.
Secondary outcomes:
-
Identification of clinical, demographic or microbiological factors associated with clinical cure;
-
Kidney injury outcomes as per RIFLE (Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease) criteria [7];
-
Workload based on number of dose adjustment(s) during the vancomycin course; and
-
Development of microbiological resistance during the vancomycin treatment course or within 30 days of completion of the initial vancomycin treatment course.
Severity of illness was measured with the Acute Physiology and Chronic Health Evaluation (APACHE) II score for intensive care unit (ICU) patients and Pitt Bacteremia score for ward patients [8, 9]. As there are no other validated measures of severity of illness in hospital ward patients, the Pitt Bacteremia score was used in these patients, recognizing that the Pitt Bacteremia score has only been validated in patients with bacteremia. Hospital location (ICU versus ward) at time of vancomycin initiation was identified and ICU patients included those with admission or transfer to a critical care bed within 48 h of vancomycin initiation. Concomitant antibiotics were defined as any antibiotic whose administration overlapped that of vancomycin by at least 48 h and was administered for at least 48 h. The time required for each vancomycin dose adjustment was estimated from typical nursing, physician, pharmacist and pharmacy technician practices at our institution (Appendix).
Sample size calculation
The literature was reviewed to provide an estimate of the vancomycin treatment failure rate in non-deep seated infections. Although three studies provided a treatment failure rate in SSTIs (ranging from 12 to 38% for all organisms; and 17–35% for MRSA only), the SSTIs included in these studies were complicated skin and soft tissue infections (cSSTIs) that involved deep soft tissue or required significant surgical intervention (infected ulcers, burns, and major abscesses) [10,11,12]. No reliable treatment failure estimates were available for UTIs or isolated uncomplicated bacteremia associated with bacteria for which vancomycin may be used. Since only patients with non-deep seated infection(s) were eligible for this study, the treatment failure rates were predicted to be lower than those reported in the literature for cSSTIs.
Based on the only available published literature, we estimated our rate of treatment failure to be between 10 and 25%; 10% being a hypothesized estimate, and 25% being the midpoint of treatment failure rates observed in the literature for cSSTIs. Thus, using a standard sample size equation for dichotomous data, 93 to 348 patients per group (based upon treatment failure rates of 10 to 25% in the high trough cohort) would be required to detect a minimal difference in failure (∆) of 10 percentage points between patients with a trough of ≤10 mg/L versus > 10 mg/L (2-tailed, p = 0.05, power = 0.80).
Statistics & data analysis
Comparison of the study cohorts for interval, nominal and ordinal data were analyzed using GraphPad InStat (version 3.05, 32-bit for Win95/NT; GraphPad Software Inc., La Jolla, California). Nominal data were expressed as total number (proportion) of patients; and low trough and high trough groups were compared using the Fisher Exact test (odds ratio, 95% confidence interval and p-value). Interval data were expressed as mean ± standard deviation (and range) and the Kolmogorov Smirov test was used to evaluate the data for normality. If interval data passed the test for normality and standard deviations were not significantly different then a two-sided unpaired t-test was used to compare the study cohorts. If interval data were normally distributed, but standard deviations between groups were significantly different then the cohorts were compared using the two-sided unpaired t-test with Welch correction. If interval data did not pass the test for normality then a two tailed Mann-Whitney test was used. A p-value of < 0.05 was considered as statistically significant.
A Pearson’s Correlation matrix (univariable analysis) (SPSS version 13.0 for Windows, created September 1, 2004) was completed to identify patient clinical, microbiological, laboratory and vancomycin dosing related parameters (independent variables) associated with clinical cure (dependent variable). Patient related parameters that were evaluated by univariable analysis as independent variables were: sex, age, hospital location, length of stay at vancomycin initiation, comorbidities, immunosuppressive medications, nephrotoxic medications, severity of illness, baseline serum creatinine, final serum creatinine, serum creatinine change, risk of renal injury, kidney injury, Gram positive bacterial species identified in culture, and infectious diagnosis for which vancomycin was prescribed. Vancomycin dosing related factors that were evaluated by univariable analysis were initial dose, final dose, initial steady state vancomycin trough, final steady state vancomycin trough, and high versus low trough cohort. Any independent variables that were available for > 20% of patients and had a p value < 0.05 with Pearson’s Correlation were maintained in the multivariable analysis of binary logistic regression (SPSS version 13.0 for Windows, created September 1, 2004) to identify the existence of a statistically significant model (p < 0.05) in which all independent variables remaining in the model had an odds ratio of > 1.