Clinical and prognostic differences between methicillin-resistant and methicillin-susceptible Staphylococcus aureus infective endocarditis

Inclusion criteria were: age ≥ 18 years with “definite” or “possible” S. aureus IE according to modified Duke Criteria [11], which were retrospectively applied to patients enrolled before their publication.

The age-adjusted Charlson comorbidity index was used to estimate the 10-year life expectancy of our patients as a function of age and comorbidities [12], determined at admission to hospital for the endocarditis episode.

Surgical risk was estimated at admission in all patients using EuroSCORE I or II (European System for Cardiac Operative Risk Evaluation), which predicts early mortality after cardiac valve surgery [13, 14]. We used both scores in order compare their capacity to predict the prognosis. In cases of IE between 1998 and 2017, this scale was calculated prospectively by the attending physician. In cases of IE before 1998, the EuroSCORE was calculated retrospectively from data in the clinical records.

IE relapse was defined by an episode of S. aureus endocarditis within 12 months of a first episode that had met cure criteria.

Mortality rates considered deaths for any cause during hospitalization or the first 30 days post-discharge (hospital mortality) as well as the IE-related mortality [e.g., heart failure due to valve dysfunction] and the non-IE-related mortality [e.g., cancer] at 1 year post-discharge.

Severe sepsis was defined by ≥2 criteria of systemic inflammatory response syndrome with organ dysfunction; and Septic shock by sepsis with refractory hypotension and end-organ perfusion dysfunction despite adequate fluid resuscitation [15].

Early prosthetic IE was defined by onset during the first year post-surgery and Late prosthetic IE by onset after more than 1 year post-surgery [16].

Nosocomial IE was defined by symptom onset more than 48 h after hospital admission [17].

Healthcare-related IE was defined by symptom onset more after medical manipulation in the 3 months preceding the diagnosis (intravenous treatments, wound healing, hemodialysis, and stays at care home or assisted centers) [17].

The indication for surgery was initially assessed by the attending physician, based on universally accepted criteria at the time [18‐20], and was confirmed by consensus of a multidisciplinary team that included attending physicians and heart surgeons, who also considered the quality of life, comorbidities, surgical prognosis, and life expectancy of patients.

Indications for surgery were divided into five groups: a) Surgery not indicated; b) Surgery indicated and performed without delay; c) Surgery indicated and performed with delay > 72 h in grade IV left ventricular failure (LVF) or delay > 1 week in progressive LVF; d) Surgery indicated but not performed for any cause (e.g., technical impossibility, neurological complication, death before surgery, patient refusal, etc.); and e) Surgery indicated but not proposed by the attending physician due to the condition of patients (e.g., comorbidity with low life expectancy, critical status, etc.).

Methicillin resistance was defined by an inhibition halo for oxacillin of ≤10 mm or oxacillin MIC ≥4 mg/L. The E-test with oxacillin strip was used in some centers [21] and automated microdilution systems in others [22].

In a descriptive analysis, central tendency and dispersion measures (mean, standard deviation, median, percentiles) were calculated for quantitative variables and absolute frequencies with 95% confidence interval (CI) for qualitative variables. In bivariate analyses, prognostic, clinical, epidemiological, and therapeutic variables were compared between patients with MRSA IE versus MSSA IE, and mortality rates for SA endocarditis were compared with prognostic factors. The Student’s t-test for independent samples was used for quantitative variables with a normal distribution and the Mann-Whitney U test for those with non-normal distribution. Qualitative variables were analyzed using Pearson’s or Fisher’s chi-square test, as appropriate. The normality of variable distribution was checked with the Kolmogorov-Smirnov test. Two multivariate logistic regression models were developed according to Freeman’s formula [n = 10*(k + 1)] [23], one for differences between MRSA IE versus MSSA IE and the other for factors related to mortality in patients due to SA IE. The models included variables found to be statistically significant in bivariate analyses or considered clinically relevant. A stepwise procedure was used, considering an entry probability of 0.05 and exit probability of 0.10. Goodness of fit was evaluated with the Hosmer-Lemeshow test. The regression model for differences between MRSA versus MSSA IE included the following variables: history of myocardiopathy, congenital heart disease, hemodialysis, chronic obstructive pulmonary disease (COPD); intravenous drug addiction; Charlson index; early prosthetic IE; perivalvular involvement diagnosed by echocardiography; previous invasive procedure or focus of infection; place of IE acquisition; interval between symptom onset and hospital admission; cutaneous manifestations (Osler’s nodes, Janeway lesions); Duke vascular or embolic phenomena; and adequate treatment administration. The regression model for risk factors associated with mortality due to IE included: hospital where IE was treated, decade of IE onset, previous valve disease, early prosthetic IE, IE on pacemaker or defibrillator lead, IE on mitral valve, onset of IE as severe sepsis/septic shock, CNS involvement, kidney failure, heart failure, infectious osteoarticular involvement (arthritis/osteomyelitis), surgical risk (EuroSCORE, logistic EuroSCORE), heart surgery indicated without delay and performed during hospitalization, and surgery indicated but not performed. IBM SPSS Statistics 20.0 software was used for data analyses. The level of significance was 0.05 for all tests.

We included 378 patients with MSSA IE and 59 with MRSA IE from 1984 through January 2017 (15.8% 1984–1999; 35.7% 2000–2009, and 48.5% 2010–2017).

According to bivariate analyses, the main epidemiological and clinical differences between IE by MRSA versus MSSA were age (62.5 vs. 58 years, p = 0.048); nosocomial acquisition (71.2 vs. 41.8%; p = 0.001); history of COPD (30.5 vs. 12.2%; p = 0.0001), elevated Charlson index (3 [1‐4] vs. 2 [0–3]; p = 0.006), history of previous invasive procedure or focus (79.7 vs. 57.1%; p = 0.001), and more frequent diagnostic delay (57.7 vs. 39.5%, p = 0.05). In comparison to patients with MRSA IE, higher percentages of those with MSSA IE had congenital heart disease (7.1 vs. 0%; p = 0.037), were in a hemodialysis program (7.6 vs. 0%; p = 0.02), and had infectious perivalvular involvement during IE (28.3 vs. 15.3%; p = 0.043). The remaining results are listed in Tables 1 and 2.

With respect to outcomes, bivariate analysis showed a significantly longer hospitalization in patients with MRSA versus MSSA endocarditis (37 vs. 30 days; p = 0.019) but no significant difference in mortality from IE at 1 year (49.1 vs. 43.7%, p = 0.33) or in the percentage of patients undergoing surgery (25.4 vs. 30.9%; p = 0.13). Among the patients with MRSA IE, there was a larger percentage for whom surgery was indicated but not performed (18.6 vs. 10.1%; p = 0.05) and a higher relapse rate (10 vs. 3%; p = 0.05) (Table 3).

The risk factors associated with MRSA in the multivariate analysis were: history of COPD (OR 3.19; 95% CI 1.41–7.23), previous invasive procedure or recognized focus of infection in the three-month period before IE onset (OR 2.9; 95% CI 1.14–7.65), and a delay of ≥7 days between symptom onset and hospital admission (OR 3.94; 95% CI 1.64–9.5) (Table 4).

MIC values were determined by E-test in 37 (62.7%) of MRSA cases and by microdilution in 22 (37.3%). Vancomycin MICs for MRSA were available in 74.6% (44/59) of cases and were > 1 μg/mL in 38.6% (17/44). There were 4 relapses, observing 3 (75%) in strains with MIC > 1 μg/mL and 1 (25%) with MIC ≤1 (p = 0.7). The Mortality attributable to MRSA IE in hospital or during the first 30 days post-discharge was 49.1% (29/59) of which Vancomycin AUC:CMI was available in 18 patients. The hospital mortality rate was 61.1% (11/18) in strains with MIC ≤1 μg/mL versus 38.8% (7/18) with MIC> 1 μg/mL, also a non-significant difference (p = 0.847).

During the hospital stay or within 30 days post-discharge, 182 (44.8%) patients died from SA endocarditis-specific mortality. The mortality rate was 56.9% (37/65) in 1985–1999; 53.8% (78/145) in 2000–2009; and 34.2% (67/196) in 2010–2017. During the one-year follow-up, 13 (2.9%) died from a cause other than IE, 6 (1.4%) died from a new IE episode due to a resistant strain, and 16 (3.7%) were lost to the follow-up. Data were available on the treatment of 378 patients (86.5%), and 342 (90%) of these received adequate initial antibiotic treatment according to the antibiogram and clinical practice guidelines.

The mortality from SA in our cohort was associated in bivariate analyses with: the decade of endocarditis onset (1985–1999: mortality rate of 20.3% (37/182) vs. survival rate of 12.5% (28/224), p = 0.032; 2000–2010: 42.8% (78/182) vs. 36.8% (67/224), p = 0.007; 2010–2017: 36.8% (67/182) vs 57.6% (129/224); p = 0.0001); older age (61.5 vs. 56.4 years; p = 0.004); active neoplasm (11.2 vs. 5.8%; p = 0.05); early prosthetic IE (9.4 vs. 4%; p = 0.028); mitral valve involvement (64.6 vs. 49.5%; p = 0.03); sepsis/septic shock (46.7 vs. 27.8%; p = 0.0001); kidney failure during IE episode (52.2 vs. 33.9%; p = 0.0001); de novo heart failure (60.2 vs. 33.8%; p = 0.0001); CNS involvement (encephalopathy 37.6 vs. 22.6%; p = 0.006; embolic stroke 29.6 vs. 18.7%; p = 0.033; meningitis 8.9% vs 5.8%; p = 0.035); high surgical risk (median EuroSCORE of 13 vs. 9; p = 0.0001; median logistic EuroSCORE of 30.76 vs. 15.4; p = 0.0001; and indication but non-performance of surgery (16.5 vs. 4.9%; p = 0.0001). MRSA itself did not emerge as a risk factor for mortality in our cohort (29% vs. 11.8%, p = 0.22) Protective factors were: IE on pacemaker lead or automatic implantable cardioverter-defibrillator (AICD) (2.8 vs. 12.9%; p = 0.0001); osteoarticular spread of the infection (9.6 vs. 17.3%; p = 0.026); heart surgery conducted when indicated without delay (23.1 vs. 36.2%; p = 0.004). According to the multiple logistic regression analysis, poor prognostic factors for SA endocarditis were: decade of endocarditis onset in 1985–1999 (OR 8.391; 95% CI (2.82–24.9) or 2000–2009 (OR 6.4; 95% CI 2.92–14.06); active neoplasm (OR 6.63; 95% CI 1.7–25.5); and sepsis/shock (OR 2.28; 95% CI 1.053–4.9) (Table 5).