Characteristics and Outcomes of Acinetobacter baumannii Infections in Patients with HIV: A Matched Case-Control Study

Acinetobacter baumannii (AB) infection is an increasing global threaten to hospitalized patients, especially those with impaired immune function. Still, few studies addressed the disease burdens and outcomes of AB infection in HIV patients. We aimed to describe characteristics and outcomes of AB infections in patients with HIV, measure the impact of AB infection on 28-day mortality in HIV patients, as well as assess the predictors of 28-day survival among HIV patients with AB pneumonia. A retrospective study with HIV/AB co-infected patients was conducted at Shanghai Public Health Clinical Center (SPHCC), China. Patients with AB pneumonia were further analyzed for predictors of mortality, as well as an additional 1:1 case-control study to determine the fatality of AB pneumonia compared with pneumonia of other pathogens. We found the incidence of AB infection was 17.4 cases per 100 person-years among all hospitalized HIV patients. Hospital mortality rate was 37.5% (21/56). There was a higher 28-day mortality rate in HIV patients with pneumonia due to AB than other pathogens (34% vs 16%, P = 0.03). APACHE II score was independently associated with 28-day survival by multivariate logistic regression (P = 0.031). Our findings indicate that AB infection is incident and can be fatal in HIV seropositive population. AB infection is an independent risk factor of mortality in patients with HIV and pneumonia. A lower APACHE II score on admission predicts a higher 28-day survival rate among HIV/AB co-infected patients.


Patients. Inclusion Criteria: i)
A positive result of anti-HIV by ELISA and Western Blot; ii) positive AB cultures in any a biological specimen iii) clinical manifestation with or without radiological findings in adherence with AB infection of the culture site. For AB pneumonia, additionally require: iv) AB pure or predominant culture in at least 2 eligible sputum.
Exclusion Criteria: i) Growth of AB and other bacteria together in one culture, with another bacterium as the predominant strain; ii) AB colonization without relevant clinical manifestation; iii) patients who did not receive antimicrobial therapy.
Demographic data of patients, including age, sex, dates of admission and discharge, was collected from the electronic medical record system. So was the routine blood test, biochemistry data, antibiotic treatment, usage of tigecycline, CD4 + T lymphocyte count, CD4/CD8 ratio, microbiological culture, antibiotic resistance and other potential predictors. APACHE-II score was retrospectively calculated making use of data from the medical record system, including vital signs, fraction of inspired oxygen, and consciousness on admission, as well as age and need for surgery. Immune dysfunction was default for each a patient. Patients was recorded until death in hospital or discharge. Those discharged within 28 days after admission was followed by telephone call to determine the outcome.
Microbiology. Blood specimen were collected from at least 2 sites and then incubated in BacT/ALERT3D ® system (bioMérieux, Marcy l'Etoile, France) for 7 days. Sputum, urine, stool, or other specimens were cultured on blood plate, bloodthirsty chocolate plate, and/or China blue agar plate, respectively. The identification of species was made by standard phenotypic tests including API 20 NE (bioMe'rieux, Marcy l'Etoile, France). Compared with PCR assays, phenotypic methods have limitation in distinguish AB from other species of the AB complex, such as A. pittii, A. nosocomialis and A. seifertii 16 . Thus, AB in this article refer to the AB complex. Antimicrobial susceptibility testing was performed manually with disk diffusion method. All the procedures were in adherence to the NCCLS standards.
We classified the AB isolates into multidrug-resistant (MDR), extensively drug-resistant (XDR), pandrug-resistant (PDR) and non drug-resistant (NDR). MDR was defined as resistant to all penicillins, cephalosporins, fluoroquinolones, and aminoglycosides. XDR was defined as resistant to the three classes listed above and resistant to carbapenems. PDR was defined as resistant to polymyxins and tigecycline in addition to XDR. NDR hereby referred to susceptible to at least 2 classes out of penicillins, cephalosporins, fluoroquinolones, aminoglycosides and carbapenems. Statistical analysis. Demographic data were presented as means with standard deviation, medians with interquartile ranges or percentages, respectively. The t test or Mann-Whitney U test was used to compare the continuous data, and Pearson's χ 2 test or Fisher's exact test was used to compare the categorical data in the two groups. Kaplan-Meier survival analysis was employed to compare survival between the two groups. Logistic regression was performed to explore the risk factors for mortality at 28 days in patients with AB pneumonia. All the statistics were done with SPSS (version 19.0), and P < 0.05 was considered statistically significant for all tests. Characteristics and outcomes of HIV/AB co-infected patients were described in Table 1. Of 56 HIV/AB co-infected patients, the mean age was 42.2 ± 13.5 years, and 52 (92.9%) patients were male. The median CD4 + T lymphocyte countwas 19.0 (6.0-59.5) cells/μL. The median APACHE II score was 17.5 (14.3-22.0). 24 (42.9%) patients required mechanical ventilation during hospital stay.

Characteristics and outcomes of AB infections in patients with HIV. During the period from
The AB positive cultures were primarily obtained from the respiratory tract (89.3%). The rest were from blood (3.6%), urinary tract (3.6%), intestinal tract (1.8%) and needle biopsies for abscess (1.8%), respectively.
Fourteen (25.0%) patients had received ART before admission, and the rest of them started ART during hospitalization.

Mortality and survival at 28 days in the case and control groups. Given the predominant infection
site was the lung, 50 patients with AB pneumonia were selected for the matched case-control study to control the bias. Control group consisted of 50 patients with pneumonia due to other nosocomial bacterial, including Klebsiella pneumonia (n = 13), Coagulase negative Staphylococcus (n = 9), Pseudomonas aeruginosa (n = 9), Enterococcus Escherichia coli (n = 5), Staphylococcus aureus (n = 5), Enterobacter cloacae (n = 3), Enterococcus faecalis (n = 2), Proteus mirabilis (n = 1), Indolonus indolens (n = 1), Stenotrophomonas maltophilia (n = 1) or Burkholderia cepacia (n = 1), respectively. The case and control group are comparable in age, gender, length of hospital stay, mechanical ventilation, history of ART, CD4 + T lymphocyte count and APACHE II score, as presented in Table 2.
The mortality rate of the case group was significantly higher than the control group (34% vs 16%, P = 0.03). We also conducted a survival curve at 28 days for the two groups (Fig. 1). The 28-day survival rate for the case group and the control group was 66% (33/50) and 84% (42/50), respectively. The control group achieved a significantly higher survival rate compared with the case group (P = 0.021). Predictors of 28 day survival among patients with AB infections. Among the 50 HIV patients with AB pneumonia, the predictors of 28 day survival were given special attention. Logistic regressions of univariate analysis and multivariate analysis were performed to sort out these risk factors (Table 3). Thirty-four percent (17/50) of the patients died within 28 days after admission. The potential variables included age, gender, duration of hospitalization, mechanical ventilation, AB resistance pattern, CD4 + T lymphocyte count, CD4/CD8 ratio, APACHE II score, ART experienced or naiive, as well as therapy including tigecycline.
By univariate logistic regression, survival rate at 28 days was significantly lower in patients using mechanical ventilation (P = 0.004), however, there was no significant difference by multivariate logistic regression (P = 0.59). We did not find an association between 28-day survival and factors such as CD4 + T lymphocyte count, usage of tigecycline, history of ART.
Overall, the results indicated that the APACHE II score was independently associated with 28 day survival by multivariate logistic regression [P = 0.031, OR = 0.82 (0.69-0.98)]. Twenty-eight day survival was significantly higher for patients who had lower APACHE II score compared to those with higher scores.

Discussion
AB can cause a broad range of severe nosocomial infections, including skin and soft tissue infections, wound infections, urinary tract infections and secondary meningitis 17 . Still, there is little knowledge about the disease burden, characteristics and outcomes of AB infections in patients with HIV.
The incidence rate of AB infection varies in different settings. Van den Broek, et al. reported that the incidences of Acinetobacter isolates ranged from 1.7 to 3.7 per 10,000 patient-days in a Dutch university hospital 18   person-years 14 . A meta-regression concluded that the worldwide incidence of Acinetobacter associated ventilator associated pneumonia ranged from 12-88 per 10,000 mechanical ventilation days 19 .
In this study among HIV seropositive inpatients, we reported an incidence rate of 4.78 cases per 10,000 person-days, in other words, 17.4 per 100 person years. Although the incidence rate is not so high as in mechanical ventilated population, it is higher than the number in general patients and comparable with that in ICU. This indicates that immune dysfunction in HIV seropositive patients may facilitate the occurrence of AB infection. Given the huge amount of HIV/AIDS patients over the world, their co-infection due to AB may be an alarming public health issue. This highlights the urgency of infection prevention and control against AB, especially in medical facilities serving for patients including those infected with HIV 14 .
A systematic review of data up to May 2013 concluded a 33% (850/2545) mortality rate among AB infections in 16 observational studies 8 . In our study, hospital mortality rate and 28-day mortality rate for 56 patients with HIV/AB co-infection was 37.5% and 35.6%, respectively. The data is high, and comparable with previous study as well. In 1:1 case-control study, after controlling potential risk factor factors, we found a higher 28-days mortality rate in HIV patients with pneumonia due to AB than other nosocomial bacteria (34% vs 16%, P = 0.03). This calls for further insight into factors influencing the outcome of AB infection in HIV patients.
Studies have shown that the immunosuppression can lead to an increase in 30-day mortality of AB infection 7 . We did not find an association between CD4 + T lymphocyte count and mortality. This may be partially restricted by low level of CD4 + T lymphocyte counts in a majority of the subjects. Out of a bundle of potential risk factors, APACHE-II score was positively associated with mortality rate by multivariate analysis. This give prominence to intensive care in treating patients with HIV and AB infection.
Among 27 European countries, 12 reported that over 50% of AB isolates were resistant to carbapenems 5 . WHO claimed that CRAB is one of potential threatens to the stability of the health care system in both the short and long term 15 . In our study, XDR-AB and PDR-AB was isolated in 16 and 17 patients, respectively, implying that no less than 58.9% of AB strains were resistant to carbapenems. This favors the coverage of carbapenem resistant strains in empirical treatment for HIV patients at risk of AB infection.
In vivo study and murine model has shown satisfactory activity of tigecycline against AB 20,21 . However, in a guideline published on 2016, IDSA recommend against the use of tigecycline in patients with HAP/VAP caused by Acinetobacter species 22 . The recommendation was mainly based on 2 matched cohort studies, showing that tigecycline was non-superior than imipenem and inferior to colistin, respectively 23,24 . In our study, tigecycline therapy failed to reduce the mortality of HIV/AB co-infection. We should be cautious when interpreting the data, given the relative small sample size and huge variation in drug-resistance of AB among different settings. Decision of tigecycline usage in HIV/AB co-infections should be cautious, considering drug-resistance data, individual factors, as well as drug availability and affordability.
There are certain limitations in our study. First, the sample size is not so large compared to studies in the general population. However, this is the largest to date in HIV/AB co-infection. Second, data was collected retrospectively and bias were inevitable. Future prospective studies may better control the bias. Third, colistin and polymyxin B was not available in China during the study period, thus we failed to provide data about these drugs. Future usage of polymyxin B should follow literatures such as the IDSA guidelines 22 . Also, the commercial phenotypic systems used in this study could not differentiate A. baumannii from several closely related species, such as from several close strains of Acinetobacter such as A. pittii, A. nosocomialis and A. seifertii 19,25 . Thus AB in this article actually refers to the Acinetobacter baumannii complex. The hospital has recently applied MALDI-TOF MS assay to better identify the species. Future study should present data from state-of-the-art molecular techniques.  In conclusion, the study reveals the burden and fatality of AB infection among HIV seropositive population, highlighting the need for infection prevention and control against the bacterial. The data may benefit therapeutic decision making for HIV/AB co-infection, including empirical treatment and the usage of tigecycline. Rather than indicators of immune suppression, APACHE-II score was found to be associated with mortality, giving prominence to intensive care when managing patients with HIV and AB infection. We expect larger multi-center prospective study covering a wider range of sub-populations and employing more state-of-the-art molecular techniques in the future.