Search results, quality assessment, and studies characteristics
As shown in Figure 1, a total of 831 articles were retrieved from PubMed and EMBASE together with additional sources. 577 records were screened by title and abstract following the removal of duplicates, and 429 articles were crossed out. After a full-text review of the remaining 148 articles, 121 articles were excluded. Finally, 27 studies were considered eligible, with 24 included in the meta-analysis for MRSA prevalence (5, 13, 29-50), three in the potential MRSA predictors (51-53), and two in the risk factors for mortality in patients with MRSA BSIs (54). The selected studies scored five or six according to the Newcastle-Ottawa scale and were considered of high or moderate quality (Supplementary Table S1).
The characteristics of the studies concerning MRSA prevalence are shown in Table 1. The studies collected BSI data in cancer patients between 1995 and 2017 and were from 19 countries covering all the six WHO geographical stratifications. All studies were observational; 12 were retrospective cohorts, 10 were prospective cohorts, and two were cross-sectional studies. Most of the studies were single-center studies except for two multicenter studies. More than half of the studies included only adults with a proportion of males varying between 39.7% and 69.9%. In total, 10,550 BSI isolates were included in the meta-analysis for MRSA prevalence. The characteristics of the studies included in the analysis of MRSA predictors and mortality were summarized in Supplementary Table S2.
Prevalence of MRSA-BSIs
Overall, the 24 data sets presented a pooled MRSA rate of 3% (95% CI 2–5%) (Fig. 2). High heterogeneity (I2 = 91%) was present among the studies. The funnel plot seemed approximately symmetrical (Supplementary Figure S1) and no significant publication bias was detected by further Egger’s test (p = 0.12).
Among a subset of 23 studies that contained data on 3690 Gram-positive BSI cases, the MRSA isolation rate in Gram-positive bacteria was 10% (95% CI 6–14%) (5, 13, 29-46, 48-50). In 15 studies whose number of S. aureus was more than 10, the pooled prevalence of MRSA among S. aureus was 44% (95% CI 32–57%) (5, 13, 29, 33, 35, 37, 39, 42-49).
Subgroup analyses of MRSA prevalence were conducted by region, age group, type of cancer, and the presence of febrile neutropenia. The prevalence of MRSA was 5% in Africa (95% CI 1–14%), 1% in Americas (95% CI 1–2%), 2% in Europe (95% CI 1–4%), 4% in Western Pacific (95% CI 2–7%), 8% in South-east Asia (95% CI 4–14%) and 0% in Eastern Mediterranean (95% CI 0–3%). All 6 regions showed significantly different MRSA rates when compared to the rest of the regions (all P<0.05). Ghana, an African country, saw the greatest MRSA rate (28%) which was nearly 10 times higher than the overall prevalence (49). On the contrary, four studies from different regions (Africa, Europe, Western Pacific, and Eastern Mediterranean) did not detect MRSA in BSI cases (32, 34, 41, 50). In addition, studies demonstrate no statistically significant changes through the examination of the temporal trend of the MRSA prevalence rates (R2 = 0.06; P=0.24) (Fig. 3).
Based on data from 16 studies on 8461 adult BSI isolates, the MRSA isolation rate among BSIs in adults was 4% (95% CI 3–6%) (5, 13, 29, 32, 33, 36, 37, 39, 40, 42, 44-49). Besides, five studies reported data on 927 BSI isolates from children and the MRSA prevalence in this population was 2% (95% CI 0–5%) (31, 34, 38, 43, 50). When comparing the two groups, the difference in the MRSA rates was statistically significant (c2 = 12.7; P = 0.0004).
To discriminate the MRSA prevalence in BSI cases with hematological malignancy and solid tumors, a Spanish study (55) which was previously excluded for duplicate analysis was retrieved and included in the latter group. In this way, 685 BSI cases among patients with solid tumors from two studies were identified, and the prevalence of MRSA in this subgroup was 4% (95% CI 0–12%) (42, 55). Seven studies provided data on 3569 BSI isolates from patients with hematological malignancies and the pooled prevalence of MRSA was 4% (95% CI 1–7%) (32, 33, 39, 40, 44-46). The difference between the two groups did not reach statistical significance (c2 = 0.03; P = 0.86).
In the specific subgroup of patients with febrile neutropenia, a total of 1351 BSI isolates from eight studies saw a pooled MRSA rate of 2% (95% CI 1–5%) (31, 32, 34, 36, 37, 40, 46, 50). Half of the studies were conducted in Europe (34, 36, 40, 46), two studies in the Western Pacific (31, 32), one in South-East Asia (37), and one in the Eastern Mediterranean (50).
Predictors for MRSA-BSIs
Risk factors for the development of MRSA bacteremia among cancer patients with BSIs were summarized in Table 2. All the three relevant studies (13, 51, 53) compared the demographic and clinical characteristics of cancer patients with MRSA BSIs to their counterparts with methicillin-susceptible S. aureus (MSSA) BSIs. Hospital-acquired BSI was found to be more prevalent during MRSA BSIs than MSSA BSIs in two studies, although with slightly different definitions (13, 53). Consistently, Indwelling urinary catheter, the presence of nasogastric tube, the need for catheter removal, and healthcare-associated pneumonia as origin were shown to be related to a higher possibility for MRSA development (13, 51). In the specific population with hematologic malignancies, hospital-acquired infection was also identified as a predisposing factor for MRSA, and primary bacteremia was found to be a protective factor (53).
Mortality of MRSA-BSIs
The 60-day mortality in adult cancer patients with MRSA BSIs was reported to be 12% (54), and the 6-month overall mortality was 43.2% (53). In those who were treated with vancomycin, the treatment failure rate was 52% with the failure defined as death, uncontrolled infection, rapid relapse, or severe adverse events (54). Concerning the risk factors for mortality in cancer patients with MRSA BSIs, multivariate analysis revealed that community-onset infection, secondary BSI, and vancomycin MIC≥2g/mL were significantly related to 60-day mortality (54).