The study was a single-center case-controlled retrospective study including 186 patients with BSIs in solid tumors who were treated in Zhongshan Hospital affiliated to Xiamen University from January 2017 to December 2021. This is a tertiary-level hospital integrating medical treatment, teaching, prevention and health care. It admits more than 60,000 inpatients every year, including over 10,000 patients with malignant tumors.

This study included only solid tumors with BSIs. Eligible participants included all patients ≥ 18 years old who had suffered at least one episode of BSIs. Their cancers were in the active stage. Based on the definition of BSIs by the Centers for Disease Control and Prevention (Garner et al., 1988), the inclusion criteria of the BSIs group were as follows: (i) The patient’s body temperature was more than 38.0 °C or below 36.0 °C, or patient was accompanied by chills; (ii) Blood culture was positive (but normal skin flora, such as diphtheria, Bacillus spp., Propionibacterium spp., and micrococci need to two blood cultures at different times); (iii) At least one positive blood culture obtained at the same hospitalization and the first infection was recorded; (iv) The pathological diagnosis was identified as a malignant tumor. The above four criteria need to be met at the same time in the study. Finally, there were 186 patients of BSIs and 416 of non-BSIs enrolled in this study. Figure 1 showed the procedure of the study.

According to CDC definitions for BSIs (Garner et al., 1988), laboratory-confirmed BSIs were defined as primary BSIs if the pathogen isolated from the blood culture was not associated with other sites of infection, otherwise they were secondary BSIs. According to the definition of catheter-related BSI (CRBSI) in the Infectious Diseases Society of America (IDSA) (Mermel et al., 2009), we defined CRBSI as follows: (i)Semi-quantitative or quantitative culture of catheter colonies was positive; (ii)Semi-quantitative culture of strains was positive, or the count of cultured strains in blood samples taken from the central vein and peripheral vein at a ratio of 5:1 increased obviously; (iii)Both blood cultures from central vein and peripheral vein at different times were positive. Pathogen which showed insensitivity to 3 or more types of antibacterial drugs was defined as multi-drug resistance (MDR) (Kadri et al., 2018). Chemotherapy and surgery history were all traced back to one month before BSIs. The history of intensive care unit (ICU) admission meant transferring to ICU due to BSIs during the same hospitalization. Neutrophil-to-lymphocyte (NLR) referred to the ratio of neutrophil count to lymphocyte count. Inappropriate initial antibiotic therapy (IIAT) was defined as the antibiotics used within 72 hours after suspecting BSIs but proved to be inactive against the pathogen in vitro culture and susceptibility testing (Tang et al., 2020).

The information of all participants was collected from the electronic medical record system and the doctors’ order system. The demographic and clinical features of the patients included age, sex, body mass index (BMI), length of stay in hospital, laboratory examination, characteristics of infections, comorbidities and use of antibiotics, etc. The susceptibility results of antibiotics were collected from the laboratory information system. The incidence was calculated from annual admission data, which was displayed as the number of BSIs episodes per 1000 hospitalized patients.

The matching of the BSIs and control group was done by the Propensity Score Matching (PSM) plugin. All analyses were carried out by SPSS26.0 (IBM SPSS Statistics, IBM Corp., Armonk, NY, US). The continuous variables which were normally distributed were described as mean ± standard deviation, and differences between the BSIs and control group were tested by T-test. While those that did not follow normal distribution were represented by median (25-75% quartile ranges), using the Manne Whitney U test. The classification variables were expressed as frequency and percentage, and the differences between the two groups were compared by the chi-square test or Fisher exact test. The Cox regression model was used to determine the hazard factors for death of BSIs in solid tumors, and constructed by R. Statistics 4.2.3 software. The statistical difference was defined as P<0.05 (double tails).

From 2017 to 2021, there were 102,704 admissions in total with patients who visited Zhongshan Hospital affiliated to Xiamen University for solid tumors. A total of 303 patients with BSIs in malignant tumors, of which 97 were excluded for the following reasons: (i) There was no definite pathological diagnosis of tumor (n=8); (ii) The patient was diagnosed as hematologic malignancies (n=85); (iii) Cancer had been cured clinically(n=4). Finally, the number of patients included in the study was 206. The BSIs and control cohorts were matched in a ratio of 1:3 by PSM. The matching conditions involved age, sex and tumor types. The demographic and clinical features of BSIs and control cohorts were displayed in Table 1 . The median age was 62 and 61 years old, while males accounted for 72.0% and 72.6%, respectively. There were no significant differences in age, sex and tumor types between the two cohorts.

Compared with the control group, the patients with BSIs were more likely to have 3 or more underlying diseases (35.5% vs. 20.0%), organ failure (10.5% vs. 2.4%) or renal diseases (19.6% vs. 3.8%), and die within 30 days (18.8% vs. 4.3%) (all P<0.001). There were significantly more patients in the BSIs group who suffered from hepato-biliary-pancreatic infections (15.6% vs. 5.3%, P<0.001) and peritoneal cavity infections (7.5% vs. 2.9%, P=0.01). During hospitalization, there were significantly more cases experiencing mechanical ventilation, urinary catheter indwelling, hemodialysis, and surgery in the BSIs group (all P<0.01). The length of hospitalization in the BSIs group was eminently longer than that in the control (24.5 days vs. 20.0 days, P<0.001). Baseline laboratory data before blood culture (including NLR, lymphocyte count, serum albumin [ALB], lactate dehydrogenase [LDH], blood urea nitrogen [BUN], creatinine, potassium ions, and sodium ions) were not statistically different between the two cohorts.

In the BSIs group, there were 124 cases of Gram-negative bacteria (GNB) BSIs, 53 cases of Gram-positive bacteria (GPB) BSIs and 15 cases of fungus BSIs, which involved 180 cases of monomicrobial infections and 6 cases of polymicrobial infections. The six polymicrobial infections included Extended-Spectrum-Beta-Lactamases (ESBL) -producing Escherichia coli plus Candida tropicalis, ESBL-producing Escherichia coli plus Candida albicans, ESBL-producing Escherichia coli plus Enterococcus faecalis, Pseudomonas aeruginosa plus Enterococcus avium, Acinetobacter baumannii plus Staphylococcus haemolyticus, ESBL-negative Escherichia coli plus Aeromonas hydrophila, respectively. Table 2 showed the distribution of microorganism of 192 episodes and the important drug resistance bacteria that causing BSIs in solid tumors. In 192 episodes of BSIs, the GNB BSIs, GPB BSIs and fungus BSIs accounted for 64.6%, 27.6% and 7.8% respectively. The most predominant GNB were Escherichia coli (n=67) and Klebsiella pneumoniae (n=27), while the most common GPB were Staphylococcus aureus (n=16) and Enterococcus spp.(n=14). Klebsiella pneumoniae had the highest 30-day mortality (25.9%, 7/27), followed by Enterococcus spp. (21.4%, 3/14). In terms of drug susceptibility testing, multi-drug resistant Gram-negative bacteria (MDR-GNB) accounted for 46.0% (57/124). There was no significant difference in outcomes caused by multi-drug resistant bacteria in the mortality and survival group (40.0% vs. 32.2%, P = 0.381) ( Table 3 ). In Escherichia coli isolates, the sensitivities to cefotaxime were 35.8%. The Enterobacteriaceae producing ESBL accounted for 53.8%. The isolates of Klebsiella pneumoniae were 100.0% sensitive to imipenem and meropenem. About 12.5% of Staphylococcus aureus were resistant to methicillin.

The follow-up time lasted 30 days after the occurrence of BSIs. The patients in the BSIs group were divided into the mortality group (n=35) and the survival (n=151). Table 3 showed the clinical characteristics and laboratory data at the time of blood cultures of both groups. The percentage of liver cancer patients was significantly larger in the mortality group than that in the survival (34.3% vs. 16.6%, P=0.018). Compared with the survival group, there was markedly a larger portion of patients in the mortality group with stage III/IV (93.3% vs. 76.1%, P=0.035), lung/thorax infections (42.9% vs. 21.2%, P=0.008), peritoneal cavity (17.1% vs. 5.3%, P=0.042), digestive diseases (48.6% vs. 23.8%, P=0.003), ICU admission (34.3% vs. 13.9%, P=0.004) and septic shock (34.3% vs. 4.6%, P<0.001). As to laboratory findings, serum ALB (28.87g/L vs. 32.53g/L, P=0.001) level in mortality was significantly lower than that in survival, whereas serum C-reactive protein (CRP) (125.17mg/L vs. 90.94mg/L, P=0.009), LDH (323.00U/L vs. 208.30U/L, P=0.005) and BUN (8.78mmol/L vs. 5.58mmol/L, P=0.001) levels were relatively higher. Patients who received initial empirical antibiotic treatment (IEAT) accounted for 94.3% of patients in the mortality group and inappropriate initial antibiotic therapy (IIAT) accounted for 11.4%. The duration of antibiotic use in the mortality group was not noticeably different from that of the survival (12.0 days vs. 13.0 days, P=0.285). The total incidence of CRBSI was 12.9% (24/186). But no significant differences in the incidences of CRBSI (17.1% vs. 11.9%, P=0.406), IEAT (94.3% vs. 92.7%, P=0.743) and IIAT (11.4% vs. 9.9%, P=0.792) were discovered between the two groups.

Multivariate logistic regression analysis revealed that organ failure (odds ratio [OR]: 3.16; 95% confidence interval [CI]: 1.15~8.68; P=0.026), nephropathy (OR: 4.83; 95% CI: 2.09~11.17; P<0.001), hepato-biliary-pancreatic infections (OR: 4.00; 95% CI: 2.18~7.33; P<0.001) were independent sensitive factors for patients with BSIs in solid tumours ( Figure 2A ). While liver cancer (hazard ratio [HR]: 3.114; 95% CI: 1.360~7.126; P=0.007), septic shock (HR: 7.095; 95% CI: 2.764~18.212; P<0.001), organ failure (HR: 2.464; 95%CI: 1.053~5.767; P=0.038) and LDH≥288.1U/L (HR: 3.002; 95% CI: 1.341~6.716; P=0.007) were independent risk factors for mortality after BSIs in patients with solid tumours according to multivariate Cox regression analysis ( Figure 2B ).

Based on these risk factors including liver cancer, LDH≥288.1U/L, organ failure and septic shock from multivariate Cox regression analysis, we established a nomogram that could predict the probability of 30-day survival of BSIs in solid tumors ( Figure 3A ). In the nomogram, we could find the score of each item through the vertical line in “Points” and calculate the total scores. Then the probability could be found by the total scores corresponding to the vertical line. The length of each variable was represented for its contribution degree to the outcome. In the calibration curve ( Figure 3B ), the slope =1 straight light indicated an assumption that the real world is in line with the predicted value. The red line deviated from the dotted line slightly. The ROC curve ( Figure 3C ) showed the AUC in the model was 0.854, which illustrated the prediction model has an excellent predictive ability.