The impact of hemodialysis on mortality risk and cause of death in Staphylococcus aureus endocarditis

The risk of infective endocarditis (IE) is high in patients with end-stage kidney disease (ESKD) [1]. Overall, the most common microbial cause of IE in Denmark is streptococcal species [2]. However, hemodialysis is associated with frequent staphylococcus (S.) aureus bacteremia. The risk factors include repeated access to the vascular system required for hemodialysis, a high frequency of underlying heart valve disease and uremia related immune impairment [3‐7]. The all-cause in-hospital mortality has been reported as high as 52% in hemodialysis patients with IE and up to 56.3% at one-year follow-up [8‐11]. Mitral valve disease and septic embolism have been identified as mortality risk factors in hemodialysis patients with IE [12].

The all-cause in-hospital mortality of S. aureus endocarditis in the general population ranges between 15 and 22%, depending on the geographic region and study population [13‐15]. Heart failure, age and cerebrovascular events are reported as mortality risk factors in the general IE population [13, 15].

In Denmark, all residents are provided with a permanent personal identification number that allows linkage between nationwide administrative registries on an individual level. Four of these registries were utilized to obtain data. The Danish National Patient Registry includes information on all outpatient appointments and hospital admissions including diagnoses and procedural codes in Denmark, since 1978. Each admission and outpatient appointment is at discharge and at end of consultation coded with one primary diagnosis, and if appropriate one or more secondary diagnoses, according to the International Classification of Diseases - until 1994 the 8th revision (ICD-8) was used and from 1994 the 10th revision (ICD-10) has been applied [16]. The codes used to retrieve comorbidities are considered valid [17]. All deaths in Denmark are registered in The National Civil Registry within 2 weeks from time of death. The Danish Registry of Cause of Death holds information on cause of death among deceased Danish residents in Denmark on an individual level since 1970 and is coded according to ICD-10, since 1994 [18]. The Danish National Registry on Regular Dialysis and Transplantation was established in 1990 and contains data on all Danish patients receiving renal replacement therapy, including changes in treatment modality and is considered valid [19].

Data in The East Danish Database on Endocarditis was prospectively collected with consecutive enrolment of patients diagnosed with IE from October 1st, 2002 to December 31st, 2012, at two tertiary referral heart centers in Copenhagen, Denmark. The centers cover a catchment area of more than 2.4 million people. The diagnosis of IE was based on clinical, microbiological and echocardiographic findings evaluated according to the revised Duke criteria [20]. The population has been described in more details previously [21].

There were two major findings in this study. First, in patients with S. aureus endocarditis in-hospital mortality was high, but there was no difference between hemodialysis and non-ESKD patients. Second, in patients discharged alive, mortality remained high in the one-year follow-up period. The one-year mortality in hemodialysis patients with S. aureus IE remained at the same level as in-hospital and higher than the annual mortality in the general Danish dialysis population [24, 25]. In non-ESKD patients with S. aureus IE, mortality declined by approximately one third at one-year follow-up compared with the in-hospital mortality.

Overall, the distribution of basic characteristics in hemodialysis- and non-ESKD patients was similar, except from the prevalence of diabetes and peripheral vascular disease, which were significantly higher in hemodialysis- compared with non-ESKD patients. These differences in baseline characteristics are common features of hemodialysis patients [1]. It is observed that the women in the hemodialysis population are younger than the women in the non-ESKD population, which remains to be explored further.

The present all-cause in-hospital mortality rate in hemodialysis patients was comparable to findings in other studies on hemodialysis and IE. In a large series of 13,130 hemodialysis patients with IE, Shroff et al. found an in-hospital mortality rate of 23.5% [26]. In smaller observational studies in-hospital- and 30 days mortality were reported to be 14.3–31% [3, 8, 27, 28]. In a recent Danish study, Ludvigsen et al. included 9392 hemodialysis patients, of these 150 patients were diagnosed with IE [29]. The thirty-day mortality was 15% in IE patients undergoing hemodialysis, which is slightly lower than the in-hospital mortality in the present study. The reported thirty-day mortality in the study by Ludvigsen et al. is based on IE regardless of bacterial etiology, which might explain the observed difference.

The in-hospital mortality varied between 15 and 22% in other recently studied populations [13‐15], which is in accordance with our current observation. However, patients in renal replacement therapy were not excluded from previous studies.

In the present study, the in-hospital mortality was similar for hemodialysis- and non-ESKD patients. Likewise, Hsiao et al. found evenly distributed in-hospital mortality in 39 hemodialysis patients (46.2%) and 51 non-ESKD patients (51%) with IE [30]. However, these in-hospital mortality rates were markedly higher than those in the present study. This difference might be caused by non-comparable study populations. In our study, 70% of hemodialysis patients underwent TEE. The sensitivity of TEE is superior to transthoracic echocardiography (TTE) not only in detecting perivalvular abscesses, −fistulas and perforations, but also in detecting small vegetations [31, 32]. The outcome of uncomplicated IE patients with small vegetations is often good, and patients with complicated IE who receive surgery have a definite better outcome than complicated IE treated conservatively [33]. In the study by Hsiao et al., the percentage of TEE is unclear and only TTE data is referred. Thus, a lower proportion of TEE may have led to selection bias in disease severity, by including patients with only larger vegetations and unrecognized perivalvular complications, which in part may explain their higher in-hospital mortality rates.

The all-cause one-year mortality in hemodialysis- and non-ESKD patients was similar to other studies on IE. In a single center study of 2239 hemodialysis patients from the period 1990–2000, Maraj et al. identified 24 incident cases of S. aureus endocarditis among 32 cases of IE [34]. They found a one-year all-cause mortality of 56.3% among all IE cases. In a study from the International Collaboration on Endocarditis-cohort, Lauridsen et al. reported the all-cause one-year mortality among patients with left-sided native valve S. aureus endocarditis to 43%, which is consistent with the current non-ESKD population [35]. Hemodialysis patients were not excluded in the study by Lauridsen et al., but only accounted for 12% of their population.

Patients undergoing dialysis have remarkably high one-year mortality. Cardiovascular disease is recognized as the leading cause of death in dialysis patients [36] followed by infections frequently related to the hemodialysis vascular access [37]. Recurrent bacteremia due to impaired immune function second to uremia [4] and the reduced defense against bacteria following repeated access to the vascular system dispose to increased risk of infections [38]. The all-cause one-year mortality was 16–20% in the Danish dialysis population in the period 2014–2015 [24, 25]. In a recent study on a large European end-stage kidney disease population from 1993 to 2007 with follow-up until 2012, Vogelzang et al. demonstrated an all-cause long-term mortality of 67% in the studied period with a median follow-up of 3.7 years. Cardiovascular death accounted for 24.4%, whereas infections were ascribed as the cause of death in 11% [39]. However, hemodialysis patients with S. aureus endocarditis have an even higher one-year cardiovascular mortality and a higher long-term cardiovascular and all-cause mortality.

Left-sided involvement in IE is most common in hemodialysis patients [11, 27, 28]. Calcification of the aortic- and mitral valve is common in end-stage kidney disease, which leads to altered local hemodynamics, increasing the susceptibility for IE [40, 41]. The right sided heart valves were involved in nine cases only in the present hemodialysis population, which is consistent with previous published data [11, 12, 27, 28]. It has been reported that the mitral valve is most commonly involved in left-sided IE in hemodialysis patients [11, 12, 27, 28], which may be due to annular calcification of the mitral valve [42]. In accordance with this observation we found mitral valve endocarditis in 51.2% of the hemodialysis patients.

Heart valve surgery was performed twice as often in the non-ESKD population as in the hemodialysis population. This difference may be explained by the high number of comorbidities leading to increased risk of heart valve surgery in hemodialysis patients. The in-hospital mortality following surgery was similar in both populations, but differed at one-year follow-up. Restriction in selection of hemodialysis patients eligible for surgery may explain the similar in-hospital mortality. In other studies, in-hospital mortality following heart valve surgery in hemodialysis patients has been varying (7–73%) [8, 12, 27, 43]. Differences in patient selection and disease severity at the time of intervention may be part of the explanation.

Raza et al. [44] found a higher mortality at one-year among surgically treated hemodialysis patients with IE compared with a propensity matched group of non-ESKD population and compared with the general hemodialysis population. This is consistent with our data. However, Raza et al. also found that mortality at one-year was higher in the non-surgically treated hemodialysis population (70%) with IE than in the surgically treated hemodialysis population (44%). We found no difference in the one-year mortality in the operated- (50%) and the non-operated (48.5%) hemodialysis patients.

The mortality rate related to mitral valve endocarditis has recently been reported higher than in aortic valve endocarditis. Surgical challenges related to invasive mitral endocarditis compared with aortic valve surgery complicate the surgical treatment and outcome of mitral valve endocarditis [45]. The proportion of mitral valve endocarditis was higher in the non-surgically treated hemodialysis population reported by Raza et al. (62%) compared with the present study (48.5%), which may explain a lower mortality rate in our study.

In non-ESKD patients, age, heart failure and cerebrovascular event have been described as risk factors of all-cause death in IE [14, 15, 46]. In hemodialysis patients, age, cerebrovascular event and diabetes mellitus have been related to all-cause mortality [12, 26]. In line with these previous data, diabetes mellitus and age were also related to both all-cause- and cardiovascular mortality in the current study. In addition, hemodialysis was associated with increased risk of all-cause mortality and cardiovascular death 70- and 81 days after admission with hazard ratios of 2.64 (95% CI 1.70–4.10) and 3.20 (95% CI 1.78–5.77) compared with non-ESKD patients, respectively. These comparisons have not been reported before and might be explained by a combination of frequent and continuous calcification of heart valves and high prevalence of mitral valve involvement in infective endocarditis, leading to progressive heart failure in the hemodialysis population with IE [40].