Cytomegalovirus pneumonia in patients with lymphoma

Abstract

BACKGROUND

Even when treated with antiviral therapy, cytomegalovirus pneumonia (CMVp) is associated with high morbidity and mortality in immunocompromised patients. CMVp has been rarely reported in patients with lymphoma.

METHODS

The authors reviewed the records of patients treated at The University of Texas M. D. Anderson Cancer Center (Houston, TX) between 1997 and 2003. Collected information included demographics, use of chemotherapy, or corticosteroids, concomitant infections, and outcome.

RESULTS

Thirty-one patients with lymphoma with 36 episodes of CMVp were identified. The incidence of CMVp increased between 1997 and 2003 (0 of 1000 treated patients vs. 9 of 1000 treated patients; P = 0.07). Most episodes occurred in patients with non-Hodgkin lymphoma (89%). Most of the patients (92%) had received chemotherapy and corticosteroids (89%) before the onset of CMVp. Concomitant CMV antigenemia was detected in 11 (41%) of the 27 episodes in which testing was performed. In 19 episodes (53%), patients had coinfections within 90 days of the episode of CMVp. Coinfections were present at the onset of CMVp in 11 episodes (31%). The yield for CMV in bronchoalveolar lavage (BAL) specimens was higher with culture methods than with cytologic evaluation or immunohistochemical staining (P < 0.001). The number of CMV antigenemia tests performed increased fourfold over the study period. The CMV-attributed mortality rate was 30% (9 of 30 patients). Independent predictors of death by multivariate Cox regression analysis were high APACHE II score (> 16) at onset of CMVp (P = 0.02, hazards ratio [HR] = 15.5, 95% confidence interval [CI], 1.5–163.7), and development of toxicity to antivirals (P = 0.04, HR = 14.03, 95% CI, 1.2–169.1).

CONCLUSIONS

The incidence of CMVp in patients with lymphoma is increasing. CMV detection in BAL specimens was better with culture methods than with cytologic or immunohistochemical methods. High APACHE II score and development of antiviral toxicity were associated with a fatal outcome. Cancer 2005. © 2005 American Cancer Society.

Cytomegalovirus (CMV) is an opportunistic pathogen for a variety of patients with impaired cellular immunity, including patients with hematologic malignancies, transplant recipients, and therapeutically immunosuppressed patients. CMV pneumonia (CMVp) is a common presentation of CMV disease in immunocompromised patients.1–3 The incidence of CMVp can be as high as 90% in lung transplant recipients.4 CMVp is also an important cause of morbidity and mortality in patients with leukemia.2 However, the incidence of CMVp in patients with lymphoma is not known.5–8

New chemotherapeutic drugs such alemtuzumab, rituximab, and fludarabine have been associated with the development of CMV infection and CMV end-organ disease in patients with hematologic malignancies.2, 9, 10 Patients with lymphoma, who are commonly treated with these drugs, might be at risk for developing CMV infection or CMV disease. In addition, more sensitive culture methods for CMV infection have improved the diagnosis of such an infection. Therefore, we sought to determine the incidence of and risk factors for CMVp among patients with lymphoma treated recently at The University of Texas M. D. Anderson Cancer Center (MDACC; Houston, TX) and to describe the treatment and outcomes of affected patients.

MATERIALS AND METHODS

We reviewed the microbiology, pathology, and autopsy reports for all patients with lymphoma treated at MDACC between January 1997 and December 2003. Patients with lymphoma and documented CMVp were included in our analysis unless they had undergone hematopoietic stem cell transplantation (HSCT) before the diagnosis of CMVp. Information on demographics, type of lymphoma, predisposing conditions of CMVp, use of chemotherapy, corticosteroids, concomitant infections, clinical and radiologic signs of CMVp, antiviral treatment, and outcome was collected from patient records. The current study was approved by the MDACC institutional review board.

Definitions

Definite CMVp was defined according to standardized criteria,11 specifically described as the presence of both clinical and radiographic evidence of pneumonia and isolation of CMV in bronchoalveolar lavage (BAL) fluid or lung tissue specimens by culture, cytology, immunohistochemical staining, or histopathologic evaluation. Probable CMVp was defined as signs and/or symptoms of pulmonary infection combined with the detection of CMV in respiratory fluids (e.g., sputum or nasal wash specimens) in the absence of pulmonary copathogens at diagnosis of CMVp. CMV shedding from the respiratory tract was defined as the isolation of CMV in BAL fluid and/or others respiratory fluids (e.g., sputum or nasal wash specimens) in the absence of clinical and radiographic evidence of infection. The date of onset of pneumonia was defined as the day of the first confirmation of CMV infection by either culture or histopathology. Coinfections were defined as other potentially life-threatening infections that occurred within 90 days of the episode of CMVp. Neutropenia was defined as an absolute neutrophil count (ANC) < 500 neutrophils per milliliter. Lymphopenia was defined as an ANC < 1000 lymphocytes per milliliter. Hypogammaglobulinemia was defined as immunoglobulin (Ig) G levels < 624 mg/dL.

Response to treatment was defined as the complete resolution of clinical and radiologic manifestations of CMVp. Antiviral-associated hematologic toxicity was defined as a decrease of ≥ 50% in the baseline neutrophil count, and antiviral-associated renal toxicity was defined as a ≥ 50% increase in the baseline serum creatinine level during therapy in the absence of other causes for the hematologic and renal toxicities observed.

Patients were considered to have recurrent CMVp if they had previously documented pneumonia with initial improvement after antiviral therapy and have not had virus detected from a respiratory sample for an interval of ≥ 4 weeks followed by reappearance of clinical symptoms and radiographic findings and the concomitant isolation of CMV.

Microbiology

Testing for CMV infection in blood specimens is not a part of routine surveillance in our patients with lymphoma. CMV antigenemia was performed at the discretion of the primary physician.

Respiratory samples (e.g., sputum specimens, BAL specimens, or nasal wash specimens) were assayed routinely for CMV infection by shell vial and/or conventional viral cultures as reported previously.1 Immunohistochemical studies in BAL specimens were performed as previously described.6

Treatment and Outcome

In general, treatment consisted of intravenous administration of ganciclovir 5 mg/kg every 12 hours, with or without intravenous administration of standard immunoglobulin (IVIG) 500 mg/kg every other day or CMV Ig 100 mg/kg 2 times weekly. Foscarnet was substituted for ganciclovir in patients who experienced severe myelosuppression during ganciclovir therapy. The foscarnet dosage was 60 mg/kg given intravenously every 8 hours. Valganciclovir was given at a dosage of 900 mg orally every 12 hours. Ganciclovir, foscarnet, and valganciclovir dosage was adjusted as needed depending on the creatinine clearance. The response to treatment was considered evaluable if treatment was administered for ≥ 48 hours.

Death was attributed to CMVp if it was due to progressive respiratory failure of a patient who had such an infection diagnosed within 3 weeks of death or when CMV infection was found postmortem in lung tissue specimens. Coinfections could have contributed to death in patients with fatal CMVp.

Statistical Analysis

The Fisher exact test was used to evaluate differences in the distribution of clinical or demographic prognostic factors for fatal outcome of CMVp.12 The survival rate was estimated by using the Kaplan–Meier nonparametric method, and the log-rank test was used to determine statistical significance.

Univariate Cox proportional hazards analysis including each potential risk factor was conducted to evaluate the crude effect of each variable in the progression to death. Multivariate Cox proportional hazards analysis was performed after adjustment for potential confounding variables, including median days between diagnosis of lymphoma and onset of CMVp (as a continuous variable), Acute Physiology and Chronic Health Evaluation (APACHE) II score at onset of CMVp (variable dichotomized as ≤ 16 or > 16), antiviral treatment (dichotomized variable), and toxicity related to antiviral treatment (dichotomized variable). All statistical tests were two sided. Differences were considered statistically significant when P < 0.05. Statistical analysis was performed using SPSS version 11.0 statistical software (SPSS Inc., Chicago, IL).

RESULTS

Patient Characteristics

Of 3091 patients with lymphoma (2478 patients with non-Hodgkin lymphoma [NHL] and 613 patients with with Hodgkin disease) cared for at MDACC between January 1997 and December 2003, 31 had CMVp, for an overall incidence of 1.1% (an incidence of 1.2% among patients with NHL and an incidence of 0.6% among those with Hodgkin disease). Isolates from an additional five patients were considered to represent shedding of the virus, and these patients were excluded from the analysis. These 5 isolates were cultured for BAL samples (4 isolates) or sputum samples (1 isolate) from patients without clinical or radiologic evidence of pneumonia. Of the 31 patients with documented CMVp, 3 had 2 separate episodes of pneumonia and 1 had 3 separate episodes of pneumonia during the study period, for a total of 36 episodes. Thirty (83%) of these 36 episodes were definite CMVp, whereas 6 (17%) were probable.

There was a trend toward a higher incidence of cases in 2003 compared with 1997 (0 vs. 9 cases per 1000 patients; P = 0.07). Most episodes occurred in patients with NHL (89%), active disease (86%), Stage III–IV disease (92%), or recurrent disease (55%) (Table 1).

Table 1. Episodes Characteristics of Patients with Lymphoma and CMV Pneumonia

Characteristics	No. of episodes (%)a
CMV: cytomegalovirus. a Data refer to the number and percentage of episodes (series included 36 episodes in 31 patients) unless otherwise indicated. b Administered within 6 months before diagnosis. c Chemotherapy was administered within 1 month in 18 of 36 episodes (50%), with salvage chemotherapy in 10 (56%) of them. d Chronic pulmonary disease in 6; end-stage renal disease in 3; diabetes mellitus, hepatic failure, and acquired immunodeficiency syndrome each in 1. e Data were available in 15 episodes. CD4 and CD8 T-lymphocyte counts < 200 cells/mm^3 were each observed in 7 (47%) of 15 episodes with information available.
Type of lymphoma
Non-Hodgkin	32/36 (89)
B-cell non-Hodgkin	24/32 (75)
T-cell non-Hodgkin	8/32 (25)
Hodgkin disease	4/36 (11)
Chemotherapy administeredbc	33/36 (92)
Salvage chemotherapy administeredbc	20/33 (61)
Corticosteroids administeredb	32/36 (89)
≥ 600 mg of prednisone equivalent	25/31 (81)
Comorbiditiesd	12/36 (33)
Hypogammaglobulinemia	10/36 (28)
Median APACHE II score at onset of CMV pneumonia (range)	13 (7–28)
Absolute neutrophil count < 500/mm^3	3/34 (9)
Absolute lymphocyte count < 1000/mm^3	17/34 (50)
Median lymphocyte count (range)	1,000 (20–7370)
Median CD4 T-lymphocyte count (range)e	238 (4–5728)
Median CD8 T-lymphocyte count (range)e	208 (33–5502)

The patients with CMVp comprised 9 women and 27 men. Their median age was 61 years (range, 17–78 years). The median time from diagnosis of lymphoma to onset of CMVp was 469 days (range, 27–4682 days) in patients with NHL and 135 days (range, 40–275 days) in patients with Hodgkin disease (P =0.02). All 16 patients in whom CMV serologic tests were performed before the onset of pneumonia were seropositive for CMV infection.

In 33 of 36 episodes (92%), chemotherapy had been administered to the patient within 6 months before the onset of CMVp, in 20 cases as salvage therapy mainly with a rituximab-containing regimen (50%). In general, 15 patients were treated with a regimen that contained monoclonal antibodies (MoAbs) (rituximab or alemtuzumab), and 5 patients were treated with a regimen that contained purine analogs (e.g., fludarabine, 2-chlorodeoxyadenosine, 2-deoxycoformycin). General characteristics and outcome were similar in patients who did or did not receive MoAbs (data not shown). In 32 episodes, corticosteroids had been administered within 6 months before the onset of CMVp. In 25 of these cases, a high dose (≥ 600 mg) of a prednisone equivalent was given.

The median APACHE II score at onset of CMVp was 13 (range, 7–28). Although the median APACHE II score was higher in patients with T-cell NHL compared with those with B-cell NHL, the outcome was comparable in both groups.

In 19 episodes (53%), patients had coinfections within 90 days of the episode of CMVp. Coinfections were present at the onset of CMVp in 11 episodes (31%). In 9 of these cases, ≥ 1 concurrent pulmonary infection was identified: 4 had infections due to bacteria, including Staphylococcus aureus (n = 1), Alcaligenes xylosoxidans (n = 1), Mycobacterium kansasii (n = 1), and Nocardia spp. (n = 1); 1 patient had infection due to viruses, parainfluenza virus (n = 1); and 4 had polymicrobial infection with Pneumocystis jiroveci/Aspergillus versicolor (n = 1), Penicillium marneffei/Mycobacterium tuberculosis/Actinomyces spp. (n = 1), S. aureus/Aspergillus spp. (n = 1), and Cryptococcus neoformans/Aspergillus flavus (n = 1). No patients with probable CMVp had pulmonary copathogens at diagnosis of infection.

All patients were symptomatic at the time of diagnosis, mainly with fever (81% of episodes), cough (69%), and dyspnea (58%). In most cases (83%), there were bilateral infiltrates on chest X-ray or chest computed tomographic scans at presentation. In 15 episodes, patients were admitted to the intensive care unit (ICU). In 9 of these episodes, mechanical ventilation was required.

Diagnosis

In 31 episodes (86%), the diagnosis of CMVp was based on the isolation of CMV in only 1 respiratory sample. In these patients, the sample was mainly obtained by BAL (in 24 patients), followed by sputum samples (in 4 patients), nasal wash samples (in 2 patients), and lung tissue samples (in 1 patient).

The yield for CMV in BAL specimens was significantly higher with culture methods than with cytologic evaluation or immunohistochemical staining (P < 0.001) (Table 2). In two patients, CMVp was diagnosed after death.

Table 2. Diagnostic Yield of Different Tests for CMV Pneumonia in Patients with Lymphomaa

Mode of evaluation of specimens
Type of specimen	No. of cases identified/no. of cases tested (%)
	Immunohistochemical staining for CMV	Cytology	Conventional culture	Shell vial assay
CMV: cytomegalovirus; ND: not done. a Some patients had more than one diagnostic method. b P < 0.001. c P < 0.001.
Bronchoalveolar lavage (n = 30)	0/29 (0)b	2/26 (8)c	17/27 (63)	23/30 (77)bc
Sputum (n = 8)	ND	0/1 (0)	3/8 (38)	7/8 (88)
Lung biopsy (n = 2)	ND	ND	2/2 (100)	1/2 (50)

CMV antigenemia was performed in 27 episodes, and results were positive in 11 (41%) of 27 cases. The number of CMV antigenemia tests performed in patients with lymphoma changed over the study period, reflecting a greater awareness of CMV infection/disease in this patient population over time (11 of 1000 patients from 1997 to 1999 vs. 41 of 1000 patients from 2000 to 2002; P < 0.01). The onset of antigenemia preceded the diagnosis of pneumonia in 5 episodes (median, 2 days; range, 1–3 days). In two of these five episodes, antigenemia was intermittent. CMV infection was detected in urine specimens in all three episodes in which urine cultures were performed. CMV excretion in a urine specimen preceded the onset of CMVp by 24 days in 1 patient tested before the onset of pneumonia.

Therapy

Antiviral therapy was administered in 33 episodes (92%). The antivirals used as initial monotherapy were ganciclovir (20 episodes [61%]), foscarnet (9 episodes [27%]), and valganciclovir (4 episodes [12%]). The median duration of antiviral therapy was 19 days (range, 4–63 days). Information on the dose and duration of valganciclovir therapy was available for 2 patients (1 received 450 mg every 12 hours for 14 days and the other patient received 900 mg orally every 12 hours for 21 days). Twenty patients received IVIG, with a median dose of 243 g (range, 30–980 g). Only one patient received CMV Ig.

Of the three patients who did not receive antiviral therapy, two patients died with progressive pneumonia. These two patients with recurrent NHL were not given antiviral therapy because the diagnosis of CMVp was made postmortem. The third patient had Hodgkin disease with pulmonary involvement and squamous cell lung carcinoma. He was not treated because the respiratory illness had substantially improved by the time the diagnosis became available and the radiographic picture was not considered to be compatible with viral pneumonia. He was followed at MDACC for 47 days and transferred to a hospice, with no additional follow-up data.

Predictors of Death Due to Cytomegalovirus Pneumonia

The CMV-attributed mortality rate was 30% (9 of 30 patients). Using univariate analysis, predictors of death caused by CMVp included a high APACHE II score (> 16) at onset of CMVp, admission to the ICU, lack of antiviral therapy, and development of toxicity to antivirals (Table 3). Using multivariate analysis, predictors of death due to CMVp were a high APACHE II score (> 16) at onset of CMVp and development of toxicity to antivirals (Table 3). Patients with high APACHE II scores (> 16) at onset of CMVp had 15.5 times the risk of dying of CMVp compared with patients with lower APACHE II scores. Patients with antiviral toxicity had 14.03 times the risk of dying of CMVp compared with patients without such a side effect. The presence or absence of these 2 variables significantly affected survival (Fig. 1).

Table 3. Risk Factors for Fatal Outcome in Patients with Lymphoma with CMV Pneumonia (n = 30)a

Factor	Patients who died (n = 9)	Patients who survived (n = 21)	P value (Fisher exact test)	Univariate HR (95% CI)	P value	Multivariate HR (95% CI)b	P value
CMV: cytomegalovirus; HR: hazards ratio; 95% CI: 95% confidence interval; M: male; F: female; NS, not significant; yrs: years; ICU: intensive care unit; IVIG: intravenous immunoglobulin. a Data were based on the 30 patients for whom outcome data were available. One patient was transferred to another hospital and his outcome was unknown. b Obtained from a Cox survival analysis. c dministered within 6 months of the onset of CMV pneumonia. d Fludarabine, 2-chlorodeoxyadenosine, or 2-deoxycoformycin. e Not included in the multivariate analysis, based on time frame. f Degrees of freedom reduced because of constant or linearity-dependent covariates. g Presence of either antiviral-associated hematologic toxicity or nephrotoxicity.
Gender	8 M/1 F	13 M/8 F	NS
Median age in yrs (range)	61 (40–76)	56 (17–78)	NS
Non-Hodgkin lymphoma	9/9 (100)	17/21 (81)	NS
Active lymphoma	8/9 (89)	18/21 (86)	NS
Recurrence of lymphoma	6/9 (67)	9/18 (50)	NS
Median days between lymphoma diagnosis and CMV pneumonia diagnosis (range)	650 (197–4682)	275 (27–3107)	0.09
Corticosteroids ≥ 600 mg prednisone equivalentc	8/9 (89)	13/21 (62)	NS
Salvage chemotherapyc	5/8 (63)	11/20 (55)	NS
Previous use of rituximabc	3/8 (38)	9/20 (45)	NS
Previous use of alemtuzumabc	0/8 (0)	1/20 (5)	NS
Previous use of purine analogscd	1/8 (13)	2/20 (10)	NS
APACHE II score at onset of CMV pneumonia, median (range)	18 (10–24)	13 (7–19)	0.009)
APACHE II score	6/9 (67)	2/20 (10)	0.003	1.00	0.001	1.00	0.02
>16				11.33 (2.72–47.25)		15.5 (1.5–163.7)
Admission to ICUe	7/9 (78)	4/21 (19)	0.004	1.00	0.006
				9.21 (1.88–45.00)
Mechanical ventilation	7/9 (78)	2/21 (10)	0.0005
Median absolute lymphocyte count (range)	540 (20–3540)	1040 (30–7370)	NS
Absolute lymphocyte count < 1000/mm^3	4/7 (57)	10/21 (48)	NS
Concomitant antigenemia	3/6 (50)	7/16 (44)	NS
Fungal infection at onset of CMV pneumonia	3/9 (33)	1/21 (5)	0.06
Antiviral treatmentf	7/9 (78)	20/21 (95)	NS	1.00	0.04
Combination therapy	6/9 (67)	11/21 (52)	NS	0.19 (0.03–0.95)
Median days of antiviral therapy (range)	17 (4–25)	20 (7–63)	NS
Toxicity related to antiviralg	5/7 (71)	3/20 (15)	0.01	1.00	0.04	1.00	0.04
				5.16 (1.04–25.53)		14.03 (1.2–169.1)
Recurrence of CMV pneumonia	3/9 (33)	2/21 (10)	NS

Figure 1. Kaplan–Meier survival curves for patients with lymphoma and cytomegalovirus pneumonia (CMVp). (A) Patients with (solid line) and without (dotted line) toxicity due to antivirals (P = 0.009). (B)Patients with APACHE II scores ≤ to 16 (dotted line) or > 16 (solid line) at onset of CMVp (P < 0.001).

DISCUSSION

To our knowledge, this is the largest reported series of patients with lymphoma and CMVp.5–8, 13

CMV infection has emerged as an important cause of life-threatening pneumonia among patients with lymphoma. Its incidence is increasing. A steady increase in the rate of CMVp in patients with lymphoma since the early 1960s was noted in an earlier autopsy study from our institution.6 This increase might reflect an increase in the number of patients who were heavily treated, a greater clinical suspicion of infection, and/or more sensitive culture methods for diagnosing CMVp. Nevertheless, CMVp is less common among patients with lymphoma (1%) than among patients with leukemia (2.9%),2 or patients who have undergone autologous HSCT (2%),1 solid organ transplantation (17–90%),4, 14 or allogeneic HSCT (7–20%).15, 16 The gender discrepancy in our series may be explained by a male predominance among patients with lymphoma.17

Suppression of a host's cellular immune response is the major underlying predisposing factor for CMV disease.18 Lymphocytes, particularly CMV-specific major histocompatibility complex-restricted cytotoxic T cells, play a crucial role in maintaining CMV in the latent stages and in controlling CMV infection.3 Their impairment may contribute to the development of CMVp in patients who have undergone HSCT.19 Patients with Hodgkin disease exhibit a persistent defect in cellular immunity, which includes a decrease in natural killer cell cytotoxicity.20 Similarly, patients with T-cell NHL may have diminished T-cell function with the disappearance of the CD8 cytotoxic population.21

It is noteworthy that the majority of the patients in our series had non-Hodgkin B-cell lymphoma. In contrast, an autopsy study identified Hodgkin disease as the most common underlying cancer and the sole type of lymphoma in 17 cases of CMVp.7 The increasing incidence of CMVp in patients with NHL, particularly lymphoma derived from a B-cell lineage, might be due primarily to the exposure to more selective suppressive chemotherapy. CMVp has been associated with the use of traditional cytotoxic immunosuppressants and depressors of T-cell function such as methotrexate, corticosteroids, and cyclosporine.2, 18 Recently, more selective chemotherapeutic drugs such as alemtuzumab (anti-CD52 MoAb), rituximab (anti-CD20 MoAb), and fludarabine (a purine analog) have also been associated with the development of CMV infection or CMVp.2, 9, 10 A number of patients with lymphoma in our series were pretreated with either MoAbs or purine analogs.

In our series, the clinical manifestations and radiologic appearances of CMVp varied widely, with no pattern sufficiently characteristic to allow differentiation of CMVp from other causes of pneumonia in patients with lymphoma.

A definitive diagnosis of CMVp depends on the detection of the virus in lung tissue specimens by immunohistochemical staining, histopathologic evaluation for viral inclusions, or culture.18 Immunohistochemical staining of lung tissue specimens is a sensitive method for diagnosing CMVp.6, 22 However, lung biopsies are rarely performed in patients with cancer because of the increased risk of bleeding and other complications from this procedure.18, 23 In contrast, bronchoscopy is a safe, effective, and rapid method for investigating pulmonary infiltrates in immunocompromised patients.24 We found that the diagnostic yield for CMV with immunohistochemical staining or cytologic evaluation of BAL specimens was extremely low, as reported by other investigators.25

In our series, the diagnostic yield with shell vial culture and conventional viral culture of BAL specimens was high. Conventional viral culture has been the gold standard for diagnosing CMV infection, but its shortcomings include a poor predictive value, lack of quantitation, false-negative results, and 1–6-week turnaround time.18, 26 Rapid viral culture (shell vial method) is a modification of conventional culture that has reduced the turnaround time to approximately 48 hours and that has a sensitivity of 68–100% compared with conventional culture.15, 18, 24 Nevertheless, shell vial or traditional cultures may be too sensitive in the detection of CMV infection in BAL specimens, as reflected by the presence of viral shedding without disease in immunocompromised patients.1, 2, 16, 27 In fact, a proportion (14%) of the CMV isolates identified in our series were of dubious clinical significance. Therefore, a cautious approach to the clinical interpretation of CMV infection identified from such diagnostic methods is recommended. Recent data showed that quantitation of CMV (i.e., determination of the viral load) may be useful in differentiating between viral shedding and CMVp in lung transplant recipients.27 Additional studies evaluating the quantitation of CMV in BAL samples for diagnosing CMVp in patients with cancer are warranted.

In the current study, CMV antigenemia testing was performed at the discretion of the treating physician with significant variations in the time of testing and lack of testing. Therefore, prospective studies may be helpful to determine the role of CMV antigenemia in patients with lymphoma with or without CMVp, as in HSCT patients, in whom CMV antigenemia is a predictor for the development of CMVp.18, 28

Coinfections secondary to hyaline molds were a major contributor to death in our patients, as we also found in a previous study.6 The immunomodulatory role of CMV infection and its effect on susceptibility to mold infections has previously been described.29 HSCT and lung transplant recipients who develop CMVp or infection have an increased risk for subsequent invasive aspergillosis.30, 31

Combination therapy with ganciclovir and high-dose intravenous Ig appears to be more effective than ganciclovir alone in the treatment of CMVp in HSCT patients.14, 32 However, no randomized trials have been done comparing these two treatment modalities. Similarly, combinations of antivirals against CMVp have not been evaluated in patients with lymphoma.

Although the mortality rate from CMVp in our patients was high (30%), it is still lower than the mortality rates from CMVp in patients with leukemia and patients who have undergone HSCT (mortality rates of 57% and ≤ 100%, respectively).1, 2, 16, 32 We described several risk factors associated with fatal outcome. Although the range of values for the risk factors of APACHE II score and toxicity related to antivirals was very wide on the multivariate analysis suggesting very low precision of the test, the impact of these variables on patient survival was significant (Fig. 1). These various predictors of mortality may be used to identify high-risk patients with lymphoma in whom early and aggressive interventions might improve outcome. Moreover, changing to a different antiviral therapy after the first evidence of toxicity might benefit these patients.

Alternative approaches need to be explored in patients with lymphoma, including a combination of antiviral drugs and immunotherapy, as do preemptive or prophylactic strategies for high-risk patients.

Our study is limited by both the small sample of patients and its retrospective nature. In addition, most of our cases were not confirmed by lung tissue biopsy, a diagnostic method replaced by BAL in most cancer centers.16 Nonetheless, all our cases were diagnosed on the basis of both clinical and radiographic evidence of pneumonia in association with CMV isolated from a lower respiratory tract sample. An important confounder of our analysis is the finding that one-half of the CMVp episodes were associated with serious coinfections, including one-third of the episodes with pulmonary coinfections at the onset of CMVp. The indicators of severe infection such as a high APACHE II score and the presence of hematologic or renal toxicities may also be related to these coinfections.

In conclusion, CMV infection has emerged as an important cause of life-threatening pneumonia among patients with lymphoma. Its incidence is increasing. The diagnostic yield for CMV with culture methods of BAL specimens is high. Predictors of death due to CMVp include a high APACHE II score and development of antiviral toxicity.