Association of bacterial genotypes and epidemiological features with treatment failure in hemodialysis patients with methicillin-resistant Staphylococcus aureus bacteremia

Objectives Methicillin-resistant Staphylococcus aureus (MRSA) infections in the hemodialysis (HD) population are epidemiologically classified as healthcare-associated infections. The data about the clinical impact and bacterial characteristics of hospital-onset (HO)- and community-onset (CO)-MRSA in HD patients are scarce. The current study analyzed the difference in the clinical and molecular characteristics of HO-MRSA and CO-MRSA. Methods We performed a retrospective review and molecular analysis of clinical isolates from 106 HD patients with MRSA bacteremia from 2009 to 2014. CA genotypes were defined as isolates carrying the SCCmec type IV or V, and HA genotypes were defined as isolates harboring SCCmec type I, II, or III. Results CO-MRSA infections occurred in 76 patients, and 30 patients had HO-MRSA infections. There was no significant difference in the treatment failure rates between patients with CO-MRSA infections and those with HO-MRSA infections. CA genotypes were associated with less treatment failure (odds ratio [OR]: 0.18; 95% confidence interval [95% CI], 0.07–0.49; p = 0.001). For isolates with a vancomycin minimum inhibitory concentration (MIC) < 1.5 mg/L, the multivariate analysis revealed that HA genotypes and cuffed tunneled catheter use were associated with treatment failure. For isolates with a vancomycin MIC ≥1.5 mg/L, the only risk factor for treatment failure was a higher Pitt score (OR: 1.76; 95% CI, 1.02–3.05; p = 0.043). Conclusion CA genotypes, but not the epidemiological classification of CO-MRSA, impacted the clinical outcome of MRSA bacteremia in the HD population.


Data collection and case definitions
We collected data on patient demographics, clinical characteristics, co-morbidities, the Charlson co-morbidity index, healthcare-associated (HA) risk factors in the 12 months preceding the MRSA-positive culture, disease severity, length of hospitalization, infection foci, empiric and definitive therapies, vancomycin trough level (if available), laboratory drug susceptibility results and outcomes.
The HA risk factors for MRSA infections are defined as follows: (1) prior hospitalization, (2) indwelling percutaneous devices or catheters at the present admission, (3) receiving surgery, (4) long-term care facility or nursing home residence, and (5) MRSA colonization or infection within one year before bacteremia onset [12]. The severity of bacteremia was graded as follows: no sepsis or simple sepsis, severe sepsis, and septic shock [13]. Thrombocytopenia was defined as a platelet count <100,000 platelets/mm 3 , as determined by a complete blood cell count. The severity of sepsis at initial presentation was assessed by the Pitt score. The empirical antibiotic was defined as the antibiotic administered within the first 48 h of index blood culture collection. Antibiotic therapies were considered effective if the isolate was susceptible to at least one of the antibiotics administered according to the 2010 CLSI guidelines [14]. The vancomycin trough level was measured after at least 2 doses of vancomycin [15], and inadequate vancomycin therapy was defined as a vancomycin trough level of <15 mg/L [16]. The concomitant use of β-lactam was defined as β-lactams administration for !24 h concurrent with intravenous vancomycin [17].
The primary outcome was treatment failure, which was defined in terms of the following events: (1) 30-day mortality; (2) persistent bacteremia, defined as a positive blood culture for MRSA obtained after >7 days of effective antibiotic therapy; or (3) recurrent MRSA bacteremia within 30 days of discontinuation of anti-MRSA therapy.

Classification of MRSA bloodstream infections
Patients with at least one positive MRSA blood culture were included in the study. Data collection and genotyping were conducted only on the first blood isolate from the same patient. Cases were identified as HO by the CDC epidemiologic definitions if the culture-positive sample was obtained > 48 h after admission and as CO if the culture-positive sample was obtained 48 h after admission [3,4]. Strains carrying SCCmec element type I, II or III are defined as HA genotypes, and those harboring SCCmec type IV or V are defined as CA genotypes [6,7]. The MRSA isolates were classified into two subgroups according to vancomycin minimal inhibitory concentrations (MICs): low (<1.5 mg/L) or high (!1.5 mg/L).

Microbiological and molecular characteristics of MRSA isolates
MRSA isolates were collected from the microbiology laboratory of Kaohsiung Medical University Hospital at the time of blood culture identification and stored at -80˚C. All S. aureus isolates were identified and tested for susceptibility to antimicrobial agents using the Vitek 2 system (bioMeriéux, Marcyl'Etoile, France). The interpretation of susceptibility followed Clinical and Laboratory Standards Institute (CLSI) guidelines [14]. Vancomycin MICs were determined by the Vancomycin E-test (AB Biodisk, Solna, Sweden) using a 0.5 McFarland inoculum streaked evenly with a swab onto Mueller-Hinton agar plates [18].
A multiplex polymerase chain reaction (PCR) method was used for SCCmec typing, and multilocus sequence typing (MLST) was performed as described in previous studies [19,20]. MLST and SCCmec types were further inferred for all strains. The PCR method described by Gilot P. et al. was used to determine accessory gene regulator (agr) typing [21].

Statistical analysis
Categorical variables were compared by a chi-square test or Fisher's exact test. Continuous variables were compared by Student's t-test or the Mann-Whitney U-test, as appropriate. The proportions of CO-MRSA and CA genotypes over 6 years were examined using the linear regression model for linear trends. Variables with a p-value less than 0.05 in the univariate analysis were incorporated into a multivariate analysis. Of the many parameters that represent the initial clinical severity of illness, only the Pitt score was included in the multivariable analysis to avoid multicolinearity. A multivariate backward logistic regression was performed to evaluate the independent effects on the primary outcome after adjustments for baseline discrepancies between groups. A p-value less than 0.05 was considered significant. The analyses were performed using SPSS v19 (SPSS Inc., Chicago, IL, USA).  Table 1. Age, weight, and the percentages of co-morbidities were similar in both groups. A significantly higher percentage of patients with HO-MRSA infections had received a shorter duration of HD at the time of bacteremia. Most of the underlying diseases among both groups were similar, but a significantly higher percentage of patients with CO-MRSA bacteremia had indwelling percutaneous devices or catheters at admission. Most catheters were non-cuffed tunneled catheters. The initial presentation in patients with HO-MRSA bacteremia was more severe (higher percentages of severe sepsis/shock, ICU admission, Pitt score) than in patients with CO-MRSA bacteremia. The majority of infection foci in both groups were catheterrelated infections, arteriovenous fistula/graft infections and skin and soft tissue infections. No significant difference was observed with regard to antimicrobial treatment, including the percentage of effective empiric, definitive antibiotics and concomitant therapies, between the groups. Catheter removal was performed in 39 of 51 (76.5%) patients with catheter-related infection. Arteriovenous fistulas or grafts were removed in 16 of 23 (69.6%) patients with an arteriovenous fistula/graft infection. The rates of catheter and arteriovenous fistula or graft removal were similar in both groups (S1 Table). No significant difference was observed with regard to the rate of source removal and time to remove an infected device for most infection foci between CO-and HO-MRSA groups, except for skin and soft tissue infection (S1 Table). However, there was no significant difference in treatment failure rates between the two groups.

Molecular characteristics of the CO-and HO-MRSA groups
SCCmec genotyping and MRSA MLST typing were available for 96 MRSA isolates (90.6% of cases). Among the 96 isolates, thirty-eight (39.6%) isolates were SCCmec type III, twenty-five (26.0%) were SCCmec type V, twenty-three (24.0%) were SCCmec type IV, six (6.3%) were SCCmec type II, one (1.0%) was SCCmec type I and three (3.1%) isolates were untypable. The molecular characteristics of CO-and HO-MRSA isolates are presented in Table 1. Significantly different proportions of SCCmec genotypes were observed between HO-and CO-MRSA isolates. Most HAHO-MRSA isolates were HA-genotypes (76.0%), and most HACO-MRSA isolates were CA-genotypes (61.8%). The proportions of HO-and CO-MRSA infections for each year of the study are presented in Fig 1A. The differences in SCCmec types by year are shown in Fig 1B. The proportion of CO-MRSA increased over time (p = 0.062). The proportion of CA genotypes significantly increased with time (p = 0.018).  ST239 was the most common MLST type. It accounted for 23 (23.9%) isolates. ST59 accounted for 17 (17.7%) isolates, and ST45 accounted for 13 (13.5%) isolates. The remaining isolates belonged to various other ST types. ST239 was identified in 33.3% of HO-MRSA cases and in 17.1% of CO-MRSA cases. ST45 and ST59 were identified in 20.0% of HO-MRSA cases and in 31.5% of CO-MRSA cases. The genotype analysis of the agr gene revealed that most isolates belonged to agr type 1 (Table 1).

Outcome associations with the SCCmec genotypes
The main clinical characteristics of the MRSA bacteremia cases infected with HA and CA genotypes were not significantly different (S2 Table). The numbers of HA risk factors were not significantly different for cases of both groups (p = 0.243). Compared with cases with CA genotypes, cases with HA genotypes showed more severe manifestations relative to the proportions of severe sepsis/shock and higher Pitt scores. There was no significant difference in antimicrobial treatment administration for cases of both groups, including the percentage of effective empiric, definitive antibiotics and concomitant therapies. No significant difference was observed with regard to the rate of source removal and time to remove infected devices between CA and HA genotypes (S3 Table).

Predictors of treatment failure
In the study, 49 (46.2%) cases had treatment failure outcomes. There was no significant difference in the primary outcome between patients with HO-MRSA infections versus CO-MRSA infections (OR, 1.80; 95% CI, 0.77-4.22; p = 0.176). In the univariate analysis, cuffed tunneled catheter use, underlying cardiovascular diseases, a history of MRSA colonization or infection, initially severe clinical presentation, infective endocarditis, high vancomycin MICs and SCCmec type were associated with treatment failure ( Table 2). The multivariate analysis revealed that a higher Pitt score was associated with treatment failure (OR:1.81; 95% CI, 1.23-2.67; p = 0.003) and CA genotypes (OR: 0.18; 95% CI, 0.07-0.49; p = 0.001) were associated with better primary outcomes (Table 3).
We further explored whether the vancomycin MIC level differences between different SCCmec type MRSA isolates accounted for the disparity in clinical impacts on MRSA bacteremia patients. A subgroup analysis stratified by vancomycin MIC was conducted to clarify this possible confounding effect. In the group with a vancomycin MIC <1.5 mg/L (n = 57), the multivariate analysis showed that cuffed tunneled catheter use was associated with treatment failure (OR: 8.09; 95% CI, 1.32-49.73; p = 0.024) and that CA genotypes (OR: 0.03; 95% CI, 0.003-0.26; p = 0.001) were associated with better primary outcomes. In the group with a vancomycin MIC !1.5 mg/L (n = 47), the only risk factor for treatment failure was Pitt score (OR: 1.76; 95% CI, 1.02-3.05; p = 0.043) ( Table 3). The association between CA genotypes and treatment failure was consistently seen in subgroup analyses, except for patients with Pitt score !2 (OR: 0.20; 95% CI, 0.03-1.19; p = 0.076) ( Table 4).

Discussion
In this study, we identified an increasing trend of CO-MRSA in the HD population with MRSA bacteremia in a Taiwanese medical center during 2009-2014. Increasingly, MRSA bacteremia cases in the HD population in Taiwan are due to CA-genotypes, indicating the successful spread of CA genotypes in the HD population. This finding that CA-and HA-MRSA strains infected HD patients in healthcare facilities is similar to the situation of the co-existence of CA and HA genotypes in hospital settings [1,22,23]. Our results revealed that the CA genotype is a better predictor of clinical outcome than the epidemiologic classification of CO-or HO-infections. This strain-specific virulence factor effect was an independent risk factor for treatment outcomes in a subgroup analysis for patients with a vancomycin MIC < 1.5 mg/L, but this effect was attenuated with a vancomycin MIC !1.5 mg/L.
In Taiwan, ST59 is the major CA MLST type, and ST239 and ST45 are the most common MLST types for hospital-acquired infections and healthcare-associated infections, respectively [24,25]. Our results revealed the existence of these three major MLST types in the HD population. We found increasing infections from CA genotypes in the HD population in Taiwan, a trend that matches the increase in CA genotypes observed in the HD population in USA [3].
It has been reported in the USA and in Taiwan that CA genotypes were independently associated with better treatment outcomes than HA genotypes in the general population [26,27]. In the early 2000s, 80% of HD patients with MRSA bacteremia were infected with HA genotypes [2,28,29]. Only a few studies have evaluated the impact of CA or HA genotypes in the HD population. Earlier reports revealed that CA genotypes were not significantly associated with 30-day mortality [28,29]. In this study, the characteristics of patients with HO-MRSA bacteremia included a shorter duration of HD, more severe clinical presentation, and more catheter-related/vascular access infections, but this epidemiological classification of MRSA infections (HO-or CO-MRSA) was not associated with treatment failure. Strain differences were observed between HO-MRSA (76% HA genotypes and 24% CA genotypes) and CO-MRSA (38% HA genotypes and 62% CA genotypes). A study in the USA showed that patients who resided in long-term care facilities were less likely to harbor CA genotypes [30], but the current study did not demonstrate the importance of this factor in Taiwan. It has been reported that CA genotypes were more prevalent in patients with a lower number of HA risk factors [31]. However, the numbers of HA risk factors were not significantly different between groups with CA-or HA genotypes in the HD population in this study (p = 0.243). In Taiwan, CA genotypes have circulated in the healthcare setting; traditional distinctions between HA-MRSA and CA-MRSA strains based on clinical characteristics are becoming less useful [6,7]. Our results support that the approach that differentiated CA-or HA-MRSA strains based on SCCmec type is of clinical significance. We evaluated multiple factors, including patient comorbidities, clinical syndromes, clinical severity and therapeutic strategies, and demonstrated that CA genotypes were significantly associated with better clinical outcomes in the HD population (OR: 0.18; 95% CI, 0.07-0.49; p = 0.001). In this study, the vancomycin MICs were lower for CA-MRSA strains than for HA-MRSA strains. In the sub-analysis according to different vancomycin MICs, CA genotypes represented an independent risk factor for clinical outcomes among isolates with vancomycin MICs < 1.5 mg/L (OR: 0.03; 95% CI, 0.003-0.26; p = 0.001). The strain-specific differences in clinical outcomes have been suggested to be due to different intrinsic virulence factors [26, 27, 32]. Among isolates with vancomycin MICs !1.5 mg/L, the only significant factor for clinical outcome was clinical severity (OR: 1.76; 95% CI, 1.02-3.05; p = 0.043). The effect of CA genotypes on clinical outcomes was not observed with vancomycin MICs !1.5 mg/L.
Maintaining a vancomycin trough level within 15-20 mg/L to treat severe MRSA infections is recommended because a ratio of the vancomycin level area under the curve to the minimum inhibitory concentration of the S. aureus isolate (AUC/MIC) of > 400 can be targeted [16]. However, inadequate vancomycin therapy was not associated with treatment failure in this study. Because the value of the AUC may be significantly increased in the HD population, aggressively targeting a pre-HD level of 15-20 mg/L may not be needed to achieve an AUC/ MIC ratio > 400 [33]. This may be the reason that 12% of the patients with low vancomycin trough levels in this study did not have poor clinical outcomes.
The limitations of this study include the following: first, an information bias may have been introduced due to the nature of the retrospective design. Second, this study was conducted in a single center in Taiwan, and the predominant CA genotypes in this study were ST59-SCCmec type IV and ST45-SCCmec type V. Whether our study findings can be generalized to different CA genotypes circulating in other HD populations in different geographic regions should be further evaluated.
In conclusion, these findings indicate that CA genotypes, but not the epidemiologic classification of CO-MRSA, impact the clinical outcome of MRSA bacteremia in the HD population. The difference in clinical impact between CA and HA genotypes was significant for isolates with vancomycin MICs < 1.5 mg/L.