Studies selected
After excluding 47048 duplicates, the two searches yielded a total of 27151 unique records from a total of three major databases - Ovid MEDLINE, Ovid Embase, and Web of Science - in addition to the Cinahl database (EBSCO) (Fig. 1). Title screening for relevance to both antimicrobial resistance and cancer excluded 25341 records, where most excluded records were only about cancer, e.g. basic research on the drug resistance of the cancer disease. Abstract screening for a non-observational study design excluded a further 845 records, where 103 had a review design, 50 had an interventional design, 454 were case reports, and 238 were miscellaneous, mostly commentary articles and conference abstracts. Finally, after full text screening; 821 records were excluded, of which 165 included other diagnoses than cancer (including healthy individuals), 197 did not include an infection or mortality outcome (often microbial distribution or endpoints like length of stay), 130 did not explicitly include antimicrobial resistance in the model, 311 did not include a multivariable model, 16 were not in English, and we failed to gain access to two leaving 144 studies included.
Of the 144 studies, 55/144 (38%) had an infection outcome, 66/144 (46%) had a mortality outcome, and 23/144 (16%) had both outcomes, all of which are listed in detail and cited in tables with all extracted data in Supplementary Material S3, S4, and S5, respectively. In total, there were 23/144 (16%) studies of patients with solid cancers (18–40), 65/144 (45%) studies of patients with haematological cancers (41–105), and 56/144 (39%) studies with patients of both or unspecified cancer types (106–161). Most studies selected (39/144, 27%) reported and modelled several bacteria and/or fungi that were tested for resistances towards several antimicrobials (18–33, 41–53, 106–115). Eight of 144 (6%) reporting and modelling several microorganisms focused only on Gram-negative bacteria (54–58, 116–118) and 1/144 (1%) focused only on fungi (59). Twelve of 144 (8%) studies studied the family Enterobacterales (especially Escherichia coli and Klebsiella pneumoniae combined), either ESBL- or carbapenemase-producing (34, 60–64, 119–124). The most commonly studied single organism was Clostridioides difficile with 27/144 (19%) studies (35–37, 65–77, 125–135). Also studied were three non-fermenters with 5/144 (3%) studies about Acinetobacter baumannii (78, 136–139), 4/144 (3%) studies about Pseudomonas aeruginosa (79, 80, 140, 141), and 5/144 (3%) studies about Stenotrophomonas maltophilia (81–84, 142), often specified as either multidrug- or extensively drug resistant. We also found several studies focusing on four well-known healthcare-associated bacteria, with 11/144 (8%) focusing on VRE (85–94, 143, 144), 7/144 (5%) focusing on carbapenemase-producing K. pneumoniae (KPC) (95–98, 145–147), 2/144 (1%) focusing on methicillin-resistant Staphylococcus aureus (MRSA) (38, 148), and 6/144 (4%) focusing on extended-spectrum beta-lactamase-producing Escherichia coli (ESBL-E) (99, 100, 149–152). We also found that fungi typically resistant to antifungals were studied, 3/144 (2%) studies about Aspergillus spp. (39, 101, 102) and 9/144 (6%) studies about Candida non-albicans (40, 103, 153–159), respectively. Finally, there was one of 144 (1%) studies for each of the microbes Staphylococcus epidermidis (with linezolid-resistance) (104), Bacillus spp. (160), Streptococcus pneumoniae (several resistance mechanisms) (161), and Aeromonas sobria (105).
Summary of findings and in-depth example
The most common microbial aetiologies were several bacteria/fungi (39, 25%), followed by C. difficile (27, 19%) and Enterobacteriaceae (12, 8%) (Table 2). The most common country for the study setting was the United States of America (USA) (40, 28%), followed by China (15, 10%), and Italy (11, 8%). The selected studies included a median of 200 patients (IQR 102–338) and 46 events (IQR 25-83.5). 103 (72%) studies used a p-value-based variable selection, either bivariable screening or stepwise regression. These studies screened a median of 16 variables (IQR 9–27), and included a median of 7 variables in the final (and largest) model. The median events per variable was close to 7.
Table 2
Characteristics
|
Frequencies
|
Outcome
|
% (n)
|
Infection
|
38 (55)
|
Death
|
45.8 (66)
|
Both
|
15.9 (23)
|
Country/setting
|
% (n)
|
USA
|
28 (40)
|
China
|
10 (15)
|
Italy
|
8 (11)
|
Spain
|
7 (10)
|
Germany
|
6 (9)
|
South Korea
|
6 (9)
|
Microbial aetiology
|
% (n)
|
Several bacterias and / or fungi
|
27 (39)
|
C. difficile
|
19 (27)
|
Enterobacteriaceae
|
8 (12)
|
Vancomycin-resistant enterococci
|
8 (11)
|
Carbapenemase-producing K. pneumoniae
|
5 (7)
|
Patients (median)
|
200 (IQR 102–338)
|
Events (median)
|
46 (IQR 25–84)
|
Events per variable (median)
|
6.9 (IQR 4.2–12.7)
|
Variables screened (median)
|
16 (IQR 9–27)
|
Variables in the final model (median)
|
7 (IQR 4–10)
|
Stepwise regression
|
% (n)
|
|
72 (103)
|
Total
|
144
|
As a model organism, we explored how risk factors for and of VRE in cancer patients was described and modelled in the selected studies. Enterococcus is a genus of Gram-positive cocci, and often refer to the two commensal E. faecium and E. faecalis. These bacteria may acquire resistance towards the antibiotic vancomycin. Of our selection, 46/144 (32%) of the studies mention either Enterococcus spp., E. faecium, or E. faecalis, which is either described as vancomycin-resistant or “multidrug-resistant” (sometimes not further specified), or where the resistance mechanism is not described at all. Most of the studies (30/46, 65%) that mention the organism include it in a larger group, like Gram-positive bacteria or multidrug-resistant organisms, which hampers any specific focus on VRE (18–30, 41–49, 106–113).
Eight studies model the risk of being either colonised or infected with VRE. The number of patients included ranged from 72 to 342, and the variables screened ranged from 6 to 46. One study focused on patients undergoing liver transplantation for hilar cholangiocarcinoma, and seven focused on haematological cancer patients. What risk factors the authors investigated for VRE infection or colonisation in a multivariable model, and what patient population they included, can be found in Table 3. Although a plethora of risk factors were investigated, five of the studies selected variables based on their corresponding p-values in bivariable analyses, and as such dropped several variables from the analysis due to their failure to achieve the p-value criterion. Authors did not necessarily conclude that all risk factors investigated in the multivariable model were of importance. The risk of bias was rated low for three, medium for five, while none had a “fatal flaw”. However, five of eight studies had less than ten events per variable in the final model, which may lead to a lack of statistical power. Among the common conclusions from the studies of VRE infection, four concluded that antibiotic exposure was a risk factor for VRE infection or colonisation, and two highlighted neutropenia. None of the antibiotic exposure risk factors was the same (one was vancomycin, one was carbapenem, one was general antibiotic exposure and one was daptomycin). When looking at the findings in detail, Aktürk et al. found that severe neutropenia and previous bacteraemia with another pathogen may increase the risk of progressing from VRE colonisation to VRE infection in paediatric haematological cancer patients (143). In 2015, Ford et al. found that severe neutropenia and number of stools per day was associated with VRE bloodstream infections in leukaemia patients, and in 2019 some of the same authors concluded that VRE colonisation rates fell when the hospital started using less carbapenems (85, 86). Herc et al reported that only previous daptomycin exposure was associated with daptomycin-resistant VRE infections in patients with haematological malignancies (87). Hefazi et al found that VRE colonisation is associated with VRE infection in stem cell transplantation patients and Ramanan et al found that VRE colonisation pre-transplantation was associated with any infection post-transplantation (31, 88). Heisel et al found that cephalosporin use and intravenous vancomycin were associated with VRE infections in patients with acute myeloid leukaemia or myelodysplastic syndrome undergoing intensive induction therapy, and finally Klein et al found that in multiple myeloma patients, granulocyte-colony stimulating factor was associated with fewer VRE cases than antibiotic prophylaxis (89, 90).
Table 3
Risk factors for VRE infection or colonisation investigated in multivariable models.
Authors and publishing year
|
Patient population
|
Risk factors investigated
|
Akturk, Sutcu, Somer et al 2016
|
All patients admitted to the paediatric haematology/oncology ward with a documented Vancomycin-resistant enterococci colonisation 48–72 h after admission.
|
Severe neutropenia (< 100/mm3), Previous bacteremia with another pathogen
|
Ford, Coombs, Stofer et al 2019
|
Patients admitted with newly diagnosed acute myelogenous leukaemia, acute lymphoblastic leukaemia, biphenotypic leukaemia, or chronic myelogenous leukaemia in blast phase between September 2011 and August 2017
|
Age, Days of empiric antibiotics, Length of stay, Cycling vs carbapenem period, Days of carbapenem, Days of cefepime or piperacillin/tazobactam
|
Ford, Lopansri, Haydoura et al 2015
|
Patients with newly diagnosed acute leukaemia between 2006 and 2012
|
Severe neutropenia, Number of stools/day
|
Hefazi, Damlaj, Alkhateeb et al 2016
|
All consecutive patients who underwent their first allogeneic hematopoietic cell transplantation for acute myelogenous leukaemia at a clinic between April 2004 and December 2014
|
Age (≥ 60 years vs < 60), Hematopoietic cell transplantation-specific comorbidity index (≥ 3 vs 0–2), Karnofsky Performance Scale (> 80 vs ≤ 80), Conditioning (Myeloablative Versus Reduced-Intensity), VRE colonisation (within 30 days of hematopoietic cell transplantation)
|
Heisel, Sutton, Mascara et al 2017
|
Adult patients who received intensive induction chemotherapy with the standard 7 + 3 regimen of cytarabine and idarubicin for newly diagnosed acute myelogenous leukaemia or myelodysplastic syndrome from January 2012 to December 2015
|
Gender (male), Vancomycin intravenous, Cephalosporin
|
Herc, Kauffman, Marini et al 2017
|
Adult patients treated on an inpatient haematology service from January 2011 to December 2015 with a haematological malignancy and a blood culture positive for an enterococcus species.
|
Congestive heart failure, Typhlitis, Gastrointestinal bleeding, Daptomycin, Cefepime received within 90 days, Cumulative Daptomycin days within 90 days, Cumulative Fluoroquinolone days within 90 days, Prior enterococcal infection, Hospital length of stay prior to positive culture, Stage of malignancy, Number induction chemotherapy regimens, Clofarabine within 30 days
|
Ramanan, Cummins, Wilhelm et al 2017
|
Adult patients who underwent liver transplantation for biopsy-proven hilar cholangiocarcinoma at a clinic between January 2004 and March 2013
|
VRE colonisation pre-transplant, Living donor transplant, Cold ischemia time, cytomegalovirus donor seropositivity, Hepatic artery thrombosis post- transplant, Biliary stricture post- transplant, Intra-abdominal fluid collection post-transplant, Number of re- operations in 1st month post- transplant.
|
Klein, Sauer, Klein et al 2021
|
Multiple myeloma inpatients who received high-dose therapy with melphalan followed by autologous stem cell transplantation at a hospital between March 2016 and July 201
|
Antibiotic prophylaxis (vs. granulocyte-colony stimulating factor support), No prophylaxis (vs. granulocyte-colony stimulating factor support), Age (per ten years), Autologous stem cell transplantation at relapse (vs. first-line treatment), ≥very good partial response before Autologous stem cell transplantation (vs. ≤partial response), Stem cell amount ≥ 2.5 * (vs. <2.5)
|
Another eight studies analyse the deaths associated with VRE in cancer patients. The number of patients included ranged from 95 to 1424, and the variables screened ranged from 11 to 56, although the exact number was indeterminable for one of the studies. Six studies selected variables based on a specified p-value threshold, but for one study the method for variable selection could not be determined. A full list of the different patient populations and risk factors investigated can be found in Table 4. We assessed the risk of bias among these studies and found four studies at low risk and four studies at a medium risk of bias. Of the eight studies, six had less than 10 events per variable in the final model. The studies had few conclusions about risk factors for mortality among cancer patients associated with VRE in common. However, three studies found that VRE bacteremia was a risk factor for death and two studies found no risk factors after running their model. When looking at the findings in detail, Akhtar et al. found that only shock (not further specified) was associated with difference in mortality between VRE and vancomycin-sensitive enterococcus bacteraemia in cancer patients (144). Kamboj et al did not find any factors that were associated with higher mortality in stem cell transplantation patients with a VRE bacteraemia (91). Kern et al modelled the mortality associated with an enterococcal bacteraemia in haematological cancer patients, but did not specify vancomycin-resistance (50). Kirkizlar et al found that in leukaemia patients colonised with VRE, a low neutrophil count and coinfection were associated with increased mortality (92). Mendes et al included VRE in a bivariable screening, but discarded the factor as it did not meet the criterion of p < 0.1 (51). Ornstein et al found that leukaemia patients with a VRE bacteraemia at the induction of chemotherapy had poorer survival than patients with other bloodstream infections (93). Papanicolaou et al found that VRE bacteraemia increased the mortality in patients receiving their first stem cell transplantation, but did not disclose how variables were selected for the multivariable model (94). Finally, Pugliese et al modelled the risk of mortality associated with several bacteria in leukaemia inpatients, among them Enterococcus spp. (no vancomycin-resistance mentioned), but did not find an association (52).
Table 4
Risk factors for death associated with VRE investigated in multivariable models.
Authors and publishing year
|
Patient population
|
Risk factors investigated
|
Akhtar, Sultan, Nizamuddin et al 2016
|
Cancer patients whose blood culture grew either vancomycin-sensitive enterococcus or VRE from January 2012 to December 2014
|
Paediatric (Compared to adults), Male (Compared to females), Haematological malignancies (Compared to solid ones), Location in hospital at onset of bacteraemia (Compared to patients in intensive care), Inpatient, Others (outpatient/ emergency room), APACHE-II score at onset of bacteraemia, Length of hospital stay in days before onset of bacteraemia, Shock (Compared to patients not in shock at onset of bacteraemia), disseminated intravascular coagulation score, Received vancomycin within 4 weeks prior to bacteraemia
|
Kamboj, Cohen, Huang et al 2019
|
Adults 18 years and older who underwent allogeneic haematopoietic stem cell transplantation and developed VRE bacteraemia between January 1, 2005 and December 31, 2014, at this cancer centre
|
Prior VRE bacteremia, Hypotension (< 90/50), Persistent VRE bacteraemia > 48 hours
|
Kern, Roth, Bertz et al 2019
|
Patients with bloodstream infection during neutropenia subsequent to high-dose chemotherapy for acute leukaemia or autologous or allogeneic haematopoietic stem cell transplantation from the beginning of 2002 through April 2015
|
Age, Sex, Antineoplastic treatment, Admitted with neutropenia, Days of neutropenia before bloodstream infection onset, bloodstream infection pathogen
|
Kirkizlar, Akalin, Kirkizlar et al 2020
|
Adult acute leukaemia patients with VRE colonised who had febrile neutropenia
|
Age, Female, < 0,5x109/L neutrophil count while VRE (+), Coinfection
|
Mendes, da Silva, da Costa Bandeira de Melo et al 2021
|
Patients aged > = 16 years with newly diagnosed acute myelogenous leukaemia who started any regimen of intensive treatment
|
Age > 60 years, Gram-negative bacteria colonisation, C-reactive protein levels > 15 mg/dL, monocytic acute myelogenous leukaemia, albumin–globulin ratio
|
Ornstein, Mukherjee, Keng et al 2015
|
Patients with de novo and secondary acute myelogenous leukaemia who received cytarabine-based induction chemotherapy at Cleveland Clinic between 1 January 2000 and 1 April 2008
|
Age (> 60 vs. <60), Cytology (Poor vs. Favourable, Poor vs. Intermediate, Poor vs. Unknown), VRE bacteraemia, Other bloodstream infections, Aetiology (secondary vs. de novo), white blood cell at diagnosis, Year of diagnosis
|
Papanicolaou, Ustun, Young et al 2019
|
Patients who received their first allogeneic haematopoietic cell transplantation for acute myelogenous leukaemia, acute lymphoblastic leukaemia, or myelodysplastic syndrome between January 2008 and December 2012
|
VRE bloodstream infection, Age (21–40, 41–50, 51–60, > 60), acute myelogenous leukaemia/acute lymphoblastic leukaemia intermediate, acute myelogenous leukaemia/acute lymphoblastic leukaemia/myelodysplastic syndrome advanced, myelodysplastic syndrome advanced, Cord blood allograft, Human leukocyte antigen 7/8, Karnofsky performance scale < 90, Cytomegalovirus donor or receptor positive, haematopoietic cell transplantation during 2010–2012
|
Pugliese, Salvatore, Iula et al 2017
|
Adult patients with newly diagnosed acute myelogenous leukaemia hospitalised at a haematological department from 1 January 2002 to 31 December 2012 in order to receive cytotoxic agent induction courses for haematological remission
|
High-dose cytarabine-containing chemotherapy, Ultrasonography-driven neutropenic enterocolitis therapy with antibiotic regimens including tigecycline.
|