Cytomegalovirus: An Unexpected Cause of Ventilator-associated Pneumonia

This study was conducted in our 15-bed medico-surgical intensive care unit (ICU) from June 1, 1989 to May 31, 1994 and was approved by the local Committee for the Protection of Human Subjects. Autopsies were performed on patients who died with acute respiratory failure and/or symptoms suggestive of ventilator-associated pneumonia (VAP) with negative antemortem bacteriologic cultures (protected specimen brush, bronchoalveolar lavage, blind bronchial sampling). Open-lung biopsy was performed when a patient's respiratory status was worsening but bacteriologic cultures were negative (protected specimen brush, bronchoalveolar lavage, blind bronchial sampling) and no explanation for the impaired respiratory status could be found. Ventilator-associated CMV pneumonia was diagnosed when these two criteria were met: (1) mechanical ventilation for at least 7 days; (2) histologically proven CMV pneumonia. This delay virtually guaranteed that all patients with histologically proven CMV pneumonia were classified as such. Immunocompromised hosts were excluded from the study. Patients were classified as immunocompromised if there was an established diagnosis of immunodeficiency (i.e., acquired immunodeficiency syndrome), or if the patient was receiving immunosuppressive agents for an underlying medical problem (i.e., lymphoma, prior organ transplantation, etc.). Patients receiving steroids were also considered to be immunocompromised. [4]The following data were prospectively recorded by a single physician: age; sex; Acute Physiology and Chronic Health Evaluation II score on admission [5]; hemoglobin; white blood cell differential count; presence of atypical lymphocytosis; serum concentrations of creatinine, bilirubin, aspartate aminotransferase, and alanine aminotransferase; maximal abnormalities on the chest roentgenogram using Weinberg's radiologic score [6]; underlying diagnoses; use of 5.0 mg *symbol* kg sup -1 ganciclovir (Cytovene, Syntex Pharmaceuticals, Palo Alto, CA) twice a day for at least 2 weeks; duration of mechanical ventilation; duration of hospitalization; and final outcome.

Histologic assessment was performed on pulmonary specimens obtained by open-lung biopsy or postmortem histologic examination of the whole lung. In such cases, four or more biopsy specimens from each lobe were taken for histologic examination. They were fixed in 10% buffered formalin for 24 h at room temperature. Then samples were dehydrated in a modified alcohol series: 95% for 15 min, 100% for 15 min, and xylene for 15 min. After dehydration, samples were embedded in a single paraffin block and serially cut to 4-micro meter thick with standard microtomes with disposable blades. Slides were stained with hematoxylin-eosin-safran. Cytomegalovirus pneumonia was diagnosed on pulmonary samples by the identification of large cells with large nuclei containing a basophilic or eosinophilic inclusion surrounded by a light halo. [7,8]Multiple cytoplasmic granular inclusions were often present. These typical findings were always associated with a diffuse interstitial pneumonitis characterized by the presence of inflammatory cells (predominantly lymphocytes), thickened alveolar septi, and an interstitial inflammation. A fibrinous alveolitis and a mild or moderate fibrosis were often associated. Bacterial pneumonia was defined by the presence of scattered neutrophilic infiltrates localized to terminal bronchioles and surrounding alveoli with evident confluence of infiltrates between adjacent lobules. [9,10]Bacteriologic investigation was performed on open-lung biopsy or autopsy and included Gram and Ziehl-Neelsen staining and culture for bacteria, mycobacteria, and fungi.

The CMV antibody status of patients was determined on admission to the ICU by an enzyme-linked immunosorbent assay (Behring, Marburg, Germany). Bronchoalveolar lavage (BAL), blood, or urine cultures were performed when the diagnosis of CMV pneumonia was suspected clinically. Bronchoalveolar lavage was performed by wedging the bronchoscope into a subsegment of the area with greatest radiologic abnormality, or when disease was diffuse, into the lingula or right middle lobe. Normal saline was sequentially instilled in 20-ml aliquots (total, to 100 ml) and suctioned into sterile traps. The first aliquot was discarded. Bronchoalveolar lavage specimens were sent for conventional microbiologic processing. In addition, BAL, blood, and urine samples were sent for shell-vial culture technique. Specimens for these cultures were inoculated onto MRC-5 cells in tissue culture. [11]Monoclonal antibodies directed against immediate early antigen (E 13, Biosoft, Clonatec, Paris) were applied 48 h after inoculation for detection of viral antigen expression. [12]Only BAL, blood, and urine cultures performed within a 1-week period preceding histologic assessment were considered.

There were 2,785 admissions to the ICU during the study period. During this period, 60 autopsies and 26 open-lung biopsies were performed in nonimmunocompromised patients. In all, we observed 25 cases of CMV pneumonia diagnosed by histologic examination (Figure 1). In the remaining 61 cases, no histologic sign of CMV pneumonia was observed. Ventilator-associated CMV pneumonia was diagnosed by histologic examination 22.4+/-8.8 days after ICU admission (median, 18 days; range, 10-40 days). Seventeen of these cases of CMV pneumonia were diagnosed on autopsy and eight cases by open-lung biopsy. Histologic examination showed that CMV was the sole respiratory pathogen in 88% of the cases. In three patients, lung cultures were positive with one microorganism (Staphylococcus aureus, Serratia marcescens, and Streptococcus pneumoniae) and histologic results were consistent with a bacterial pneumonia associated with the CMV pneumonia.

Patient serologic status was determined in 18 cases. Thirteen of these 18 patients were seropositive for CMV on admission to the ICU. During the same study period, 58% of the 2,785 patients admitted to the ICU were seropositive for CMV. Bronchoalveolar lavage was performed in 17 of the 25 study patients whereas blood and urine cultures were performed in 15 of the 25 study patients. Bronchoalveolar lavage, blood, and urine cultures were performed during the week preceding histologic assessment in 43 of the 61 patients without histologic signs of CMV pneumonia. The sensitivity of BAL, blood, and urine cultures for the diagnosis of histologically proven ventilator-associated CMV pneumonia was 53%, 20%, and 13%, respectively. The specificity of these tests for the diagnosis of histologically proven ventilator-associated CMV pneumonia was 92%, 83%, and 62%, respectively.

Characteristics of the 25 patients with ventilator-associated CMV pneumonia are outlined in Table 1. The onset of the clinical deterioration of respiratory status occurred at 17.4+/-8.8 days after ICU admission (median, 14 days; range, 8-38 days). The presence of fever > 38 degrees C was observed in all but 6 patients (mean, 38.5 degrees C; range, 37.0-40.1 degrees C). Chest roentgenograms did not show any specific signs; Weinberg's score was 9.2+/-1.9 (median, 9; range, 4-12). Evidence of CMV infection of the retina was never found. Involvement of the gastrointestinal tract was noted in one of the 25 patients with CMV pneumonia. At the moment of diagnosis of CMV pneumonia, the minute ventilation used was 13.2+/-3.0 l *symbol* min sup -1 (range, 9.0-20.0), and the highest FIO2 used was 0.68 +/-0.17 (range, 0.45-1.0). All patients were ventilated with positive end-expiratory pressure. Laboratory data are summarized in Table 2. Hypoxemia and moderate hypercapnia were observed on arterial blood gases at the moment of diagnosis of VAP. The CD4+:CD8+ ratio was assessed in 17 patients and was found reversed in 5 of them.

Comparison of the 25 patients with VAP and the 61 patients without histologic sign of CMV pneumonia showed a more severe hypoxemia in the CMV group (72+/-16 mmHg vs. 95+/-41 mmHg, P < 0.05) and a higher Weinberg's score (9.2+/-1.9 vs. 7.4+/-2.7, P < 0.05) whereas age, Acute Physiology and Chronic Health Evaluation II score on admission, duration of mechanical ventilation, PaCO², white blood cell count, and temperature were not statistically different.

Only the eight patients who underwent open-lung biopsy were treated with ganciclovir for their ventilator-associated CMV pneumonia (Table 3). The pulmonary status of four of these eight patients was improved within the first 5 days after the onset of ganciclovir therapy permitting an increase of more than 75% of their PaO²/FIO2 ratio. Of the four patients who died, we observed an initial good response to antiviral therapy in two cases, with an increase of the PaO²/FIO2 ratio of 25-30%.

The purpose of this study was to describe and elaborate on the previously unreported problem of ventilator-associated CMV pneumonia in the ICU. We used histologic signs, the "gold standard," for the diagnosis of CMV pneumonia. According to the literature, histologic examination has proved to be a highly specific tool for the diagnosis of CMV pneumonia. [11,13-15]In fact, CMV induces highly specific cytologic modifications (i.e. cytomegaly, intranuclear inclusions, intracytoplasmic inclusions) that are totally different from those produced by other Herpesviridae. [16,17]We considered the presence of pathognomonic cells with intranuclear inclusions to be a necessary criterion for the diagnosis of CMV pneumonia. From a practical point of view, it would appear necessary to examine several sections because cytologically altered cells are present only in some sections. Distribution of infected cells within a tissue may be highly variable, thus reducing the chance of finding them in a given section.

Oda et al. [18]have noted that CMV pneumonia could be observed in patients without hematologic malignancies or acquired immunodeficiency syndrome. They also observed histologic CMV pneumonia in patients with cancer. In most cases, documented infection with CMV in nonimmunosuppressed hosts produces little, if any, disease. [19]Although one could consider that the frequency of CMV infection fluctuates depending on immunologic disturbances observed in ICU patients, we were unable to find any published data on the incidence of CMV pneumonia in mechanically ventilated patients. It is noteworthy that ICU patients are considered hosts susceptible to bacterial or fungal infections (linked to impaired host defenses) but not viral infections. The absence of previous publications on the existence of ventilator-associated CMV pneumonia may reflect the lack of a standardized approach to the documentation of CMV pneumonia. Our diagnostic approach, which uses frequent virologic assessment of blood, urine and BAL fluid, suggests that CMV pneumonia is not an exceptional cause of VAP. In our experience, CMV pneumonia could develop in all categories of patients (multiple trauma, chronic obstructive lung disease, and surgical patients). It must be noted that the true incidence of CMV pneumonia could not be drawn from our study. Indeed, open-lung biopsy and autopsies were performed essentially to locate the cause of a respiratory failure when all bacteriologic cultures were negative. Indirect findings suggest that CMV infection could play an important role in the morbidity and mortality of ICU patients. For example, Domart et al. [20]have shown that, in patients with mediastinitis after cardiac surgery, mortality was higher for patients with viral shedding than for patients without it.

The mechanisms of acquisition of CMV pneumonia during ICU stays remain unknown. Human CMV is thought to infect 70% of adults but generally remains asymptomatic and silent, although it may be reactivated and produce lesions at any time during the life of the host. There is also a growing body of evidence for exogenous transmission of CMV even in seropositive patients, with the risk of acquiring viral infection linked to the number of transfused units. [21]Although a lack of evidence for nosocomial transmission of CMV has been demonstrated, [22]Faix [23]showed that contaminated objects may retain live virus and therefore potential infectious ability for several hours after contamination. This possibility stresses the need for hand washing before and after contact with CMV-infected subjects or environmental objects possibly contaminated by infected secretions.

We found that clinical features were not useful in differentiating CMV from bacterial infection. It should be noted, however, that the cases of CMV pneumonia were diagnosed after a longer period of mechanical ventilation (median, 18 days) than cases of bacterial VAP as reported in the literature. [24]Chest radiograph played only a small role in the diagnosis of CMV pneumonia. All ICU patients had some type of radiographic abnormality at baseline, probably reflecting interstitial edema, atelectasis, scarring, or other postoperative changes. Episodes of CMV pneumonia were superimposed upon baseline radiographic abnormalities, making it difficult to identify subtle changes. However, we found that radiographic infiltrates were generally bilateral (median Weinberg's score, 9) associating interstitial and alveolar infiltrates.

As clinical and radiologic signs lack specificity and even sensitivity, it would seem necessary to use virologic methods to diagnose CMV pneumonia. Rapid diagnostic techniques (e.g., centrifugation cultures, which are read within 16-48 h) could be performed on BAL, blood, and urine products. These techniques are more useful in the treatment of ICU patients than conventional cultures, which require 1-4 weeks to diagnose CMV. The sensitivity of these techniques varies in the literature and has not been evaluated in mechanically ventilated patients. Thus, comparing a shell vial culture technique with conventional viral culture of lung tissue in marrow transplant patients, Crawford et al. [25]found a sensitivity of 96% with this rapid culture technique. Other authors have found that between 60% and 70% of the specimens positive by shell vial assay yield CMV by conventional culture method. [26-28]It has even been reported that shell vial assay could be negative in 59% of the blood specimens that are positive with conventional tube cultures. [28]Our results showed that sensitivity of shell vial assay was low, 53% for BAL culture and 20% or less for blood and urine cultures. Conversely, and as published previously, [25]we found a good specificity for the shell vial assay (92%). Some recent diagnostic procedures are available for clinicians. For example, Erice et al. [29]reported that CMV antigenemia assay was significantly more sensitive than shell vial cultures of CMV in the polymorphonuclear leukocyte fraction of blood leukocytes. The CMV antigenemia assay is relatively simple to perform and may be completed in 5-6 h. One of the most important features of the antigenemia assay is that it can be quantified and that high levels of antigenemia appear to correlate with CMV disease. [30]Polymerase chain reaction has found an ever-increasing number of clinical applications, including CMV diagnosis. The method has some major pitfalls. Owing to its high sensitivity, even a small trace of contaminating DNA can cause false-positive results. False-negative results may be caused by the genetic variability of clinical strains of CMV, because altered nucleotide sequence may prevent annealing of the primers. Development of a rapid and sensitive double polymerase chain reaction to detect conserved sequences from the immediate early gene of human CMV could help to differentiate latency from active infection. [31]We are currently evaluating antigenemia and double polymerase chain reaction for the diagnosis of CMV pneumonia in ICU patients.

Is CMV infection of the lung a clinically significant process justifying specific antiviral therapy? If so, are conventional pulmonary diagnostic techniques sufficiently sensitive and do they provide for early detection of CMV pneumonia to initiate an optimal treatment program? Discrepancies between our results and those of other investigators concerning the existence of CMV as a causative agent of VAP indicate that additional studies are needed to determine the incidence and mortality of CMV pneumonia in patients whose lungs are mechanically ventilated. These studies will no doubt use shell vial culture techniques, antigenemia, and polymerase chain reaction assay. The main problem is that none of these techniques has been validated in the diagnosis of ventilator-associated CMV pneumonia. Nevertheless, these new techniques may not necessarily increase specificity in diagnosing CMV pneumonia. Under these circumstances, precise diagnosis of CMV pneumonia continues to rely on documenting CMV inclusions on tissue specimens obtained by open-lung biopsy or transbronchial lung biopsy and improved methods for diagnosing CMV pneumonia are required.