Pneumonia caused by extensive drug-resistant Acinetobacter baumannii among hospitalized patients: genetic relationships, risk factors and mortality

Acinetobacter baumannii (AB) is one of the most important and common pathogens causing nosocomial outbreaks worldwide, especially in intensive care units (ICUs). The most common bodily site of A. baumannii infection is the respiratory tract, particularly in cases of hospital-acquired pneumonia (HAP) [1, 2]. A. baumannii is also notorious for its remarkable ability to acquire antibiotic resistance. Data from the CHINET surveillance system demonstrated that A. baumannii resistance to many important antimicrobial agents has increased, especially imipenem and meropenem, which increased from 31% in 2005 to 62.4% in 2014 and from 39% in 2005 to 66.7% in 2014, respectively [3]. Recently, the rise in the frequency of nosocomial infections caused by extremely drug resistant (XDR) A. baumannii strains (defined as resistance to all available antibiotics except colistin and tigecycline) has been of great concern because XDR resistance has been associated with high mortality and treatment failure [2, 4‐7]. According to our previous study [5], most of the isolates (76.2%,32/42) were XDR strains, mostly recovered from the respiratory system, but at present little research concerning extensive drug-resistant A. baumannii (XDRAB) pneumonia has been reported.

Currently, A. baumannii is recognized as one of the most difficult health care-associated infections to control and treat, and the optimal treatment of infections caused by XDRAB has not been established [6]. Surveillance of A baumannii isolates may inform prevention and control measures for these infections. Additionally, determining the process of disease spread by routine surveillance can abrogate routes of bacterial transmission [2]. Multilocus sequence typing (MLST) is a widely used technique for bacterial typing. MLST provides a portable method that is suitable for global epidemiological studies and monitoring of the national and international spread of bacteria [8, 9]. Currently, two large national studies [10, 11] have confirmed that CC92 represents the most epidemic sequence type (ST) in China. ST92, which is the founder of CC92, is the predominant ST, whereas other STs belonging to CC92 vary by area. ST75 may be the most common epidemic ST in eastern China [12], whereas ST138 may be the most common ST in western China [13]. Our previous study discovered that ST195 and ST208 belonged to CC92 were the major clone spreading in our hospital [5]. We assumed that ST195 and ST208 may be more common in southern China Guangzhou area, but this needed to be confirmed further. These differences may be due to the different antibiotic usage habits, which possibly influenced the evolution of ST92. However, little is known about the relationship between antibiotic resistance and certain STs. Although MLST has many advantages, it is a robust scheme that is often time-consuming, expensive, and labour-intensive [14]. Currently, several studies have reported that sequence-based typing (SBT) of bla _OXA-51-like gene variants has potential for application to assess the epidemiological characterization of A. baumannii [14‐16], but more data are needed.

In this study, we examined the molecular typing and characteristics of A. baumannii at three teaching hospitals. In the MLST, CC92 was the most prevalent clonal complex. However, ST92, which was the predicted founder of CC92 and was reported to be one of the most epidemic STs in multiple provinces in China [1, 10, 11], was not detected in our study. ST92 was also not detected in other studies from China [1, 5] and South Africa [24]. The lack of ST92 detection may be due to the relatively small sample size in our study, which may not accurately represent the diversity and relative abundance of A. baumannii STs. However, some studies [1, 5, 25] have shown that ST195, ST208, ST365, and ST191 (but not ST92) are the most common STs discovered in different hospitals across China. In this study, ST195 was the most commonly observed ST, accounting for 24/52 (46.1%) of the isolates, follow by ST208. These findings are similar to the findings of Zhou et al. [25], who found that ST195 accounted for 31/57 (54.4%) isolates. These data collectively suggest that ST195 is the most common ST and may serve as a severe epidemic marker in the Guangzhou area.

Some data are available to clarify why CC92 is predominant. Zhong et al. [12] found that ST75, which is the single locus variant (SLV) of ST92, differs in its gpi loci and belongs to CC92, has more severe imipenem resistance that might improve its chances of survival. In our study, CC92 accounted for 82.7% (43/52) of the isolates. Most of the XDR isolates (29/34, 85.3%) were from CC92. However, CC92 included 14 MDR isolates with less severe resistance, and non-CC92 had similar results. No significant differences with respect to the resistance phenotype were detected between the CC92 and non-CC92 strains (P = 0.767). Runnegar et al. [26] revealed that CC92 had been spread in the hospital for 9 years with variable antibiotic susceptibility. These findings may suggest that both adaptation to the hospital environment and antibiotic resistance have been important for the success of CC92.

The bla _OXA-51-like genes are unique to A. baumannii and immobile and thus may be used as markers for the identification of this species [21]. Recently, sequence-based typing (SBT) of bla _OXA-51-like gene variants was reported to have potential for use in the epidemiological characterization of A. baumannii isolates obtained from various locations in Europe. bla _OXA-69 , bla _OXA-66 , and bla _OXA-71 are the predominant members of closely related bla _OXA-51-like subgroups, which are associated with European clone I (EUI), EUII, and EUIII, respectively [14‐16]. However, few studies have investigated the correlation between bla _OXA-51-like variants with MLST typing.

According to a study by Hamouda et al. [14], MLST data showed that all isolates harbouring the major bla _OXA-51-like alleles (OXA-66, OXA-69, and OXA-71) fell within the three major European clonal lineages. The SBT-bla _OXA-51-like gene scheme produced results comparable to those produced by the Bartual MLST for the identification of the major epidemic lineages. Pournaras et al. [27] evaluated the SBT-bla _OXA-51-like gene scheme in parallel with Pasteur’s MLST. In the study, according to the SBT-bla _OXA-51-like gene, all 585 A. baumannii isolates from a large international collection were typed and assigned correctly to the nine CCs and the singleton ST78. Zhou et al. [25] revealed that 52 isolates from Guangzhou city in China carrying bla _OXA-66 were assigned to six distinct STs, which clustered into CC92; the remaining isolates belonged to four singletons that each carried a single bla _OXA-51-like allele. The results of these studies indicate that the SBT-bla _OXA-51-like gene scheme has the advantage of a significantly reduced sequencing cost and assay time and may be effective for the rapid typing of A. baumannii strains. However, in our study, only one isolate carried OXA-199, whereas the remaining 51 A. baumannii isolates carried OXA-66. The SBT-bla _OXA-51-like gene scheme failed to discriminate strains carrying the same OXA-66 allele. Similar findings reported by Wang et al. [28] showed that 18 representative isolates from different hospitals across China carried the same OXA-66 allele.

In Hamouda et al.’s study [14], the SBT-bla _OXA-51-like method was evaluated in a large international collection of A. baumannii isolates including 22 countries. Similarly, Pournaras et al. [20] typed isolates obtained from various locations in Europe (Italy, Greece, Turkey, and Lebanon). Environmental differences among countries may act as natural selection forces [12] to introduce diversity among bla _OXA-51-like genes. When confined to one city or hospital, the controlled environment and antibiotic usage habits might influence the convergent evolution of the bla _OXA-51-like genes. As a result, single-locus sequence-based typing of the bla _OXA-51-like genes may not be effective at distinguishing isolates in a small-scale, geographically restricted study. However, further investigation is needed to confirm this hypothesis.

Some studies have reported risk factors for antibiotic resistance in A. baumannii infection, which include the length of stay in an ICU, severity of the underlying disease, mechanical ventilation, invasive procedures, and prior antibiotic use [29‐32]. However, the clinical characteristics of XDRAB pneumonia have rarely been reported. In our study, univariate analysis was used to identify patients at risk of acquiring XDRAB pneumonia and showed that patients with XDRAB were older and had higher APACHE II scores than patients without XDRAB, which was similar to the findings of Özgür et al. [33]. Additionally, patients who had COPD and cardiac diseases were more likely to acquire XDRAB. Multivariate analysis using a logistic regression model showed that the APACHE II score, COPD and cardiac disease were independent risk factors for XDRAB pneumonia development. Thus, XDRAB may be particularly pathogenic in patients who are immunocompromized. Our study found that carbapenem use was an important risk factor for XDRAB pneumonia (P = 0.04). Previous studies [29, 31, 34] have also shown that selective pressure exerted by the use of carbapenem leads to the emergence of multidrug resistant (MDR) and XDR A. baumannii isolates. Therefore, rational use of carbapenem is necessary to reduce the risk of generating resistant mutants.

High mortality rates have been reported for nosocomial pneumonia caused by A. baumannii (ranging from 28.1 to 85.3%). The independent risk factors included the severity of illness (e.g., severe sepsis, septic shock and APACHE II score) and empiric inappropriate therapy [31, 33, 35, 36]. In one study, patients with carbapenem-resistant A. baumannii (CRAB) pneumonia had a higher mortality rate than patients with carbapenem-susceptible A. baumannii (CSAB) pneumonia based on the survival analysis (29.9% vs 45.6%, respectively, P = 0.02) [31]. However, another study reported mortality rates for patients with MDR A. baumannii infections that were not significantly higher than the mortality rates in patients without MDR A. baumannii infections [37]. Furthermore, in a study on ventilator-associated pneumonia (VAP) due to XDRAB, the mortality rates were not significantly higher than those of non-XDRAB VAP in ICU patients [33]. Our study showed a high hospital mortality rate in patients with XDRAB pneumonia, but A. baumannii resistance was not associated with mortality (P = 0.582).

ST195 may be the most common ST in the Guangzhou region of China and may serve as a severe epidemic marker in this region. SBT of bla _OXA-51-like gene variants may not be able to sufficiently distinguish isolates obtained from a small-scale, geographically restricted study. XDRAB pneumonia was strongly related to systemic illnesses and the APACHE II score but was not associated with in-hospital mortality.