This observational, retrospective, multicenter study was performed at 4 referral centers for IE: Hospital Universitari Vall d’Hebron (HUVH), a 1000-bed teaching hospital in Barcelona (Spain), Complejo Hospitalario Universitario de Vigo (CHUV), a 1200-bed teaching hospital in Vigo, Pontevedra (Spain), Presidio Ospedaliero Universitario Santa Maria della Misericordia (POUSMdM), a 1000-bed teaching hospital in Udine (Italy), and Hospital de Barcelona (HdB), a 250-bed private hospital in Barcelona (Spain). All 4 centers are referral centers for cardiac surgery.

Considering the importance of controlling the portal of entry to avoid recurrences, since January 2014 a colonoscopy has been systematically performed in EFIE patients at POUSMdM, since July 2014 at HUVH, and since January 2015 at CHUV, and HdB. From the moment that a colonoscopy was systematically performed in each participating hospital until October 2018, all consecutive episodes of definite EFIE in adult patients (aged ≥ 18 years) were included in the study. Only the first episode of EFIE in each patient was recorded. Patients were retrospectively identified from the Infectious Diseases Registry of each participating hospital, in which all consecutive episodes of IE are prospectively recorded.

The endpoint was an endoscopic finding of colorectal disease that could potentially cause bacteremia.

The definition of variables related to IE can be found in detail in the methods of the supplementary data.14–20

We included all colonoscopies performed after the onset of IE symptoms or 6 months before EFIE diagnosis, as well as those performed during treatment and 6-month follow-up. Assessment of bowel preparation was performed according to gastroenterologist reports and was classified as good (no fecal remains or very few that allowed an adequate surface examination), average (some semisolid stool that could be suctioned or washed away and allowed adequate surface examination) or poor (abundant fecal remains that could not be suctioned or washed away and that prevented adequate surface examination),21 and the possibility of cecal intubation was recorded.

Endoscopic findings of the colonoscopy and adverse events were also recorded. Adverse events related to the procedure included colonic perforation, lower gastrointestinal bleeding, and allergic reactions to sedative medication.

We recorded all endoscopic findings. We did not include uncomplicated diverticula and uncomplicated hemorrhoids to be a potential cause of bacteremia. Among any endoscopic findings that could potentially cause bacteremia, we classified colonic lesions into neoplastic and nonneoplastic diseases according to the histopathological report. We conducted the classification based on the most advanced lesion identified. If no histopathological report was available, we classified the lesion as nonneoplastic.

Colorectal neoplasms included colorectal adenomas and colorectal carcinomas. Adenomas were divided into nonadvanced colorectal adenomas (tubular adenomas with a diameter <10 mm) or advanced colorectal adenomas (adenomas measuring ≥ 10 mm, with villous architecture, high-grade dysplasia, or intramucosal carcinoma). The criterion for colorectal carcinoma was the presence of malignant cells beyond the muscularis mucosae.22 Nonneoplastic diseases included colonic mucosal inflammation, angiodysplasias, ulcers, and nonneoplastic polyps. All patients with endoscopic findings were referred to the gastroenterologist for follow-up.

Demographic, clinical, diagnostic, treatment, outcome and follow-up data were obtained from the prospective IE registry of each center. Data on colonoscopy (date of performance, pathological findings, and adverse events) were retrospectively collected from the patients’ medical charts and entered in a database created specifically for this study.

Quantitative variables are reported as the median and interquartile range [IQR], and qualitative variables as number and percentage. Differences between patients according to the source of infection or the presence or absence of endoscopic findings in colonoscopy were assessed by the chi-square test or Fisher exact test for categorical variables, as appropriate, and the 2-sample Wilcoxon rank-sum (Mann-Whitney) test for continuous variables. Statistical analyses were performed using STATA software, version 15. A 2-tailed P value <.05 was considered statistically significant.

The study was approved by the hospital ethics committee of Hospital Universitari Vall d’Hebron, Barcelona (Spain), (approval PR(AG)332/2018) and the remaining participating centers. Informed consent from patients was not required.

A total of 103 episodes of definite EFIE were included in the study. Of these cases, 78 (76%) patients underwent a colonoscopy (figure 1). Of those with a colonoscopy, 47 (60%) had endoscopic findings indicating a potential source of bacteremia. The epidemiological, clinical, and outcome characteristics are shown in table 1 of the supplementary data.

Figure 1.

Flowchart. a The reasons were acute pulmonary edema, bronchoaspiration, and hematologic malignancy, respectively. b One patient had a previous urinary tract infection due to E. faecalis, 1 patient was a 32-year-old woman with recurrent urinary tract infections, and 1 patient had undergone a hemicolectomy for a colon adenocarcinoma 8 months previously.

(0.23MB).

Table 1 shows the endoscopic findings in all patients undergoing a colonoscopy and divided by groups depending on whether the source of infection was known (excluding patients with previously diagnosed colorectal disease) or unknown. For more detailed information see table 2 of the supplementary data.

Of the 78 patients who underwent a colonoscopy, 47 (60%) had endoscopic findings indicating a potential source of bacteremia. Thirty-nine patients (83%) had a colorectal neoplastic disease, and 8 (17%) had a nonneoplastic disease. Of the 39 patients with a colorectal neoplasm, 19 had a nonadvanced colorectal adenoma, 18 had an advanced colorectal adenoma, and 2 had a colorectal carcinoma. Of the 8 patients with a nonneoplastic disease, 3 patients had nonmalignant ulcers, 2 patients had colorectal mucosa inflammation (1 due to ischemic colitis and the other to radiation proctitis), 2 patients had angiodysplasias, and 1 patient had a polyp without a histopathological report.

A colonoscopy was available in 45 (71%) of the 63 patients with an unknown source of infection, 25 (78%) of the 32 patients with a known source of infection (including urinary tract, hepatobiliary source, catheter-related bacteremia source and other sources), and all 8 patients with a previously diagnosed colorectal disease as the presumed source of infection. Of 45 patients with an unknown portal of entry who underwent a colonoscopy, a potential gastrointestinal origin was identified in 29 (64%). Of 25 patients with a known source of infection and a colonoscopy, excluding those with previously diagnosed colorectal disease, 11 (44%) patients had colorectal disease (table 1 and table 2 of the supplementary data).

Bowel preparation was considered good in 27 colonoscopies, average in 42 colonoscopies, poor in 6 colonoscopies and was not available in 2 colonoscopies. Cecal intubation was feasible in 74 patients (95%), and unknown in 1 patient. A polypectomy was performed in all 38 patients with polyps, and a biopsy was performed in 3 patients (1 of the 3 ulcers and in 2 suspected colorectal carcinomas).

Regarding complications, 4 (5%) patients had lower gastrointestinal bleeding. All bleedings occurred after polypectomy for advanced adenomas at a median of 10 days, including 2 cases in patients receiving acenocoumarin (both patients required red cell transfusion) and the other 2 in patients receiving enoxaparin (1 patient also required red cell transfusion). A second colonoscopy was performed in the 4 patients with bleeding, and endoscopic hemoclips were placed in 3 patients. There were no reported colonic perforations or allergic reactions to sedative medication.

Table 2 shows the baseline characteristics of the EFIE patients according to the presence or absence of relevant endoscopic findings. Age and comorbidities were similar. There were no significant differences between the presumed sources of infection, although an unknown origin and gastrointestinal source had more endoscopic findings than the urinary tract. Complications, surgical treatment, and outcomes were similar in the 2 groups, with the exception of the only 2 relapses that both occurred in patients with relevant endoscopic findings and neither had previous surgical indication. One patient received 26 days of intravenous antibiotic therapy and relapsed at day 23 after treatment completion. Colonoscopy was performed during treatment of the first episode of EFIE, and small ulcers were identified in the sigmoid colon. The other patient received 44 days of treatment and relapsed at day 48 after completing antibiotic treatment with negative blood cultures in between. Colonoscopy was performed 14 days after relapse, not during the treatment of the first episode of EFIE, and 3 nonadvanced adenomas and 1 advanced adenoma were identified.

This study has several limitations. The main limitations are its retrospective nature and the sample size, inherent in EFIE as it is a serious but rare disease. The source of infection is typically presumed; however, it is very difficult to determine the exact moment in which bacteremia occurs. Although patients with EFIE were included from the moment that a colonoscopy was systematically performed in each participating center, colonoscopy was not performed in all EFIE cases, mainly due to the patient's rapid transfer to another center after surgery was refused or due to comorbidities or poor patient clinical course. Moreover, some colorectal lesions could have gone unnoticed due to inadequate bowel preparation. On the other hand, we compared the percentage of adenomas found in our cohort characterized mainly by men in their late seventies with a noncontemporaneous cohort of patients from a different country undergoing colonoscopy as colorectal screening,9 which can lead to a bias given that some patients in our cohort had previous colorectal disease, and some had symptoms attributable to the gastrointestinal tract. In this regard, due to the retrospective design of the study, we have no information regarding specific risk factors for colon cancer in our cohort.