Reduced Mortality of Cytomegalovirus Pneumonia After Hematopoietic Cell Transplantation Due to Antiviral Therapy and Changes in Transplantation Practices

Outcome of cytomegalovirus pneumonia has improved moderately over the past 25 years. Advances in outcome seem to be due to antiviral treatment and changes in transplant practices rather than use of adjunct intravenous immunoglobulin treatment.

Abstract

Background. Despite major advances in the prevention of cytomegalovirus (CMV) disease, the treatment of CMV pneumonia in recipients of hematopoietic cell transplant remains a significant challenge.

Methods. We examined recipient, donor, transplant, viral, and treatment factors associated with overall and attributable mortality using Cox regression models.

Results. Four hundred twenty-one cases were identified between 1986 and 2011. Overall survival at 6 months was 30% (95% confidence interval [CI], 25%–34%). Outcome improved after the year 2000 (all-cause mortality: adjusted hazard ratio [aHR], 0.7 [95% CI, .5–1.0]; P = .06; attributable mortality: aHR, 0.6 [95% CI, .4–.9]; P = .01). Factors independently associated with an increased risk of all-cause and attributable mortality included female sex, elevated bilirubin, lymphopenia, and mechanical ventilation; grade 3/4 acute graft-vs-host disease was associated with all-cause mortality only. An analysis of patients who received transplants in the current preemptive therapy era (n = 233) showed only lymphopenia and mechanical ventilation as significant risk factors for overall and attributable mortality. Antiviral treatment with ganciclovir or foscarnet was associated with improved outcome compared with no antiviral treatment. However, the addition of intravenous pooled or CMV-specific immunoglobulin to antiviral treatment did not seem to improve overall or attributable mortality.

Conclusions. Outcome of CMV pneumonia showed a modest improvement over the past 25 years. However, advances seem to be due to antiviral treatment and changes in transplant practices rather than immunoglobulin-based treatments. Novel treatment strategies for CMV pneumonia are needed.

Although the development of effective cytomegalovirus (CMV) prevention strategies has significantly reduced the need to treat established CMV disease after hematopoietic cell transplantation (HCT) [1–3], the management of CMV pneumonia remains a formidable challenge [4–6]. Furthermore, late CMV disease may occur after both myeloablative and nonmyeloablative HCT [5–7].

Although there have been several studies evaluating the risk of developing CMV pneumonia [7–10], there is very limited information on factors that determine outcome of CMV pneumonia [11]. There have been no randomized trials evaluating the treatment for CMV pneumonia. Open-label studies of CMV pneumonia in the late 1980s showed improved outcome with the combination of ganciclovir and high-dose intravenous immunoglobulin (IVIG) compared with historical survival rates [12–14]. Since then, the combination of ganciclovir and high-dose IVIG became the standard of care for CMV pneumonia [15]. However, CMV pneumonia continued to be a major clinical problem with poor survival despite this combination treatment strategy [5, 11, 16].

METHODS

Study Patients and Setting

All patients with CMV pneumonia as defined in international guidelines [17] between 1986 and 2011 were included. The study was approved by the Fred Hutchinson Cancer Research Cancer institutional review board. Strategies for prevention and treatment of CMV pneumonia have varied over time (see Supplementary Figure 1). Statistical models accounted for these changes. From January 1986 to June 1992, all CMV-seropositive recipients of allogeneic HCT received high-dose intravenous acyclovir [18], from 5 days before to 30 days after transplantation. Since 1986, all HCT recipients who were seropositive for herpes simplex virus and seronegative for CMV received the standard low-dose acyclovir prophylaxis from day 5 before until day 30 after transplantation. Between 1990 and June 1994, 2 randomized trials of ganciclovir prophylaxis at engraftment vs ganciclovir preemptive therapy were conducted [1, 19]. Thereafter, a preemptive therapy strategy based on pp65 antigenemia or CMV DNAemia detection was used [20]. The CMV prevention strategy was verified for each patient and included in the analysis.

Treatment of established CMV disease consisted of 3 different strategies. In the 1980s, no effective treatment was available (rare cases received high-dose intravenous acyclovir or IVIG monotherapy; these cases were analyzed in the “no treatment” category). Once ganciclovir became available in the mid-1980s, it was used as first-line treatment (5 mg/kg every 12 hours [induction therapy] for 7–21 days followed by 5 mg/kg/day [maintenance therapy] for at least 3 weeks); in case of marrow suppression, foscarnet was administered instead (90 mg/kg every 12 hours [induction therapy] for 7–14 days followed by 90 mg/kg/day [maintenance therapy] for at least 3 weeks). After phase 2 results from 3 studies were published in the late 1980s, pooled IVIG or CMV-specific immune globulin (CMV-Ig) was given as adjunct therapy to the majority of patients (IVIG 500 mg/kg or CMV-Ig 150 mg/kg was administered every other day for 2 weeks followed by a weekly dose for an additional 4 weeks).

Laboratory Testing and Surveillance Cultures

CMV identification by cell culture, shell vial (early antigen detection by monoclonal antibody) techniques, or fluorescent antibody assays was performed as previously described [21]. Available frozen bronchoalveolar lavage (BAL) specimens (−80°C) obtained at the time of CMV diagnosis were retrospectively tested for CMV DNA. DNA was extracted from 200 µL of BAL and eluted into 100 µL of AE buffer using a Qiagen 96 DNA blood kit; 20 µL of DNA was then used for CMV polymerase chain reaction (PCR) using a quantitative PCR assay [22]. In a subset of patients with available serum samples at CMV pneumonia diagnosis, immunoglobulin G (IgG) serum levels were determined using standard laboratory methods.

Definitions

CMV pneumonia was defined according to published criteria; it required documentation of the virus in BAL, in biopsy or in autopsy by culture, cytology, immunohistology, or direct fluorescent antibody methods (in BAL specimens) in conjunction with radiological pulmonary infiltrate [17]. Patients who had a concomitant viral, fungal, bacterial (>10⁴ colony-forming units/mL), or parasitic pathogen diagnosed in BAL or lung biopsy were categorized as having coinfections. Attributable mortality was defined as death that occurred within 6 months after the day of CMV disease onset with active CMV disease at autopsy and/or respiratory failure, with no other cause directly contributing to the death.

Statistical Analysis

The primary and secondary outcome of this study was defined as overall and CMV-attributable mortality within 6 months following CMV pneumonia onset, respectively. Cumulative incidence curves were used to estimate overall survival and CMV-attributable mortality from time of first CMV pneumonia. Death not attributed to CMV was treated as a competing risk for CMV-attributable mortality [23].

The associations between candidate risk factors and the outcomes were estimated by means of Cox regression models. Covariates included were age at transplantation, recipient sex, donor sex, race, number and type of transplant procedure (allogeneic vs autologous), human leukocyte antigen (HLA) matching status, conditioning regimen, cell source, underlying disease prognosis, recipient/donor CMV serostatus, pretransplant pulmonary function, and anti–T-cell therapy in the 6 months preceding diagnosis of CMV pneumonia. Pulmonary functions include forced expiratory volume and carbon monoxide diffusion capacity. Other covariates included were diagnostic test for CMV pneumonia (BAL vs biopsy), time of CMV pneumonia from transplantation, the CMV pneumonia treatment regimen used, maximum creatinine and bilirubin values, and lymphopenia within 2 weeks preceding CMV pneumonia onset, presence of respiratory copathogens at the time of diagnosis, the requirement for mechanical ventilation at diagnosis, acute and chronic graft-vs-host disease (GVHD), and lung viral load at time of CMV pneumonia. GVHD indicators were entered as time-dependent covariates, with the time of occurrence set to zero if GVHD was diagnosed before the onset of CMV pneumonia. Variables with >10% of missing value were not entered in the initial multivariable model. For variables with <10% of missing values, a separate category was fitted for missing data. All covariates with univariate P values <.1 or factors of particular interest (cell source, HLA matching status, conditioning regimen, time of CMV diagnosis, CMV treatment, anti–T-cell therapy in the 6 months preceding diagnosis, maximum creatinine and bilirubin values, and lymphopenia in the 2 weeks preceding diagnosis and mechanical ventilation) were considered for inclusion in the multivariable model.

A subset analysis among patients who survived for at least 3 weeks after the diagnosis of CMV pneumonia was conducted to determine the risk of death associated with the duration of anti-CMV induction treatment (<14 days vs ≥14 days) and different strategies of corticosteroid treatment dictated by CMV diagnosis (no corticosteroid treatment; increasing, decreasing, or unchanged corticosteroid dose). Another subset analysis specifically explored the role of the use of immunoglobulin products in the treatment of CMV pneumonia in the overall and a more contemporary subset. The analysis included only patients who received ganciclovir or foscarnet with or without CMV-Ig or IVIG. To explore whether immunoglobulin products were beneficial in specific subgroups of patients, we made unadjusted comparisons of survival according to cell sources (peripheral blood stem cells [PBSCs] vs bone marrow), year of transplantation, high bilirubin value (>1 mg/dL; >17.1 mmol/L), lymphopenia (<300 cells/µL) within the 2 weeks preceding CMV pneumonia diagnosis, and mechanical ventilation at time of diagnosis.

A 2-sided P value of <.05 was considered statistically significant. No adjustments were made for multiple comparisons. Analysis was performed using Stata Intercooled 9 statistical software (StataCorp LP, College Station, Texas) and SAS software, version 8.1 (SAS Institute, Cary, North Carolina).

RESULTS

CMV pneumonia occurred in 421 HCT recipients a median of 67 days (range, 0–2650 days; interquartile range [IQR], 46–134 days) after transplantation.

Patient Characteristics

Demographics and transplant characteristics are summarized in Table 1. Clinical and biological characteristics as well as management strategies at onset of CMV pneumonia are displayed in Table 2.

Table 1.

Characteristics of the Study Cohort of 421 Hematopoietic Cell Transplant Recipients With Cytomegalovirus Pneumonia

Characteristic	No. (%)
Age, y
Median (IQR)	39 (26–49)
0–19	70 (17)
20–49	251 (60)
50–73	100 (24)
Sex
Female	188 (45)
Male	233 (55)
Donor sex
Female	185 (44)
Male	231 (55)
Unknown	5 (1)
Ethnicity
Nonwhite	70 (17)
White	348 (83)
Unknown	3 (1)
Transplant year
Median (IQR)	1995 (1990–2001)
1986– July 1992	161 (38)
August 1992–1999	128 (30)
2000–2011	132 (31)
No. of transplants
1	394 (94)
2	23 (5)
2 in tandema	4 (1)
HLA donor status
HLA-matched donor	170 (40)
HLA-mismatched donor	55 (13)
Unrelated donor	168 (40)
Autologous	28 (7)
Conditioning regimen
Myeloablative with combination chemotherapy	101 (24)
Myeloablative with total body irradiation	273 (65)
Nonmyeloablativeb	47 (11)
Stem cell graft source
PB or PB + bone marrow	129 (31)
Bone marrow or cord blood	292 (69)
Disease risk at transplantationc
Low	166 (39)
Intermediate	46 (11)
High	206 (49)
Unknown	3 (1)
Recipient CMV serostatus pretransplant
Negative	39 (9)
Positive	381 (91)
Unknown	1 (<1)
Donor CMV serostatus pretransplant
Negative	198 (47)
Positive	221 (52)
Unknown	2 (<1)

Abbreviations: CMV, cytomegalovirus; HLA, human leukocyte antigen; IQR, interquartile range; PB, peripheral blood.

^a Tandem regimens combine 2 transplant procedures: autologous transplantation with subsequent autologous or nonmyeloablative allogeneic transplantation (auto–auto or auto–allo).

^b Nonmyeloablative regimen includes low-dose total body irradiation (200 cGy) alone or combined with fludarabine (90 mg/m²).

^c Risk of death associated with the underlying hematologic disease as previously defined [24].

Table 2.

Characteristics of Cytomegalovirus (CMV) Pneumonia and Characteristics of Patients at Time of CMV Pneumonia Diagnosis and Patient Outcomes After Hematopoietic Cell Transplantation

Characteristic	No. (%)
CMV pneumonia diagnosis, d after transplanta
0–30	70 (17)
>30–100	208 (49)
>100	143 (34)
Copathogens isolated from respiratory specimen at time of CMV pneumonia
None	234 (56)
Fungi or viruses (± bacteria)	41 (10)
Bacteria only	105 (25)
Unknown	41 (10)
Treatment of CMV pneumonia
Ganciclovir or foscarnet	73 (17)
Ganciclovir or foscarnet plus CMV-Ig/IVIGb	326 (77)
No treatment or CMV-Ig alone or high-dose intravenous acyclovir alone	22 (5)
Bilirubin (highest value) within 2 wk prior to CMV pneumonia
Normal range (0.1–1 mg/dL; 1.7–17.1 µmol/L)	72 (17)
>ULN (1 mg/dL; 17.1 µmol/L)	168 (40)
>5× ULN (5 mg/dL; >85.5 µmol/L)	78 (19)
Unknown	103 (24)
Creatinine (highest value) within 2 wk prior to CMV pneumonia
Normal range (0.3–1.2 mg/dL; 23–90 µmol/L)	116 (28)
>ULN (1.2 mg/dL; 90 µmol/L)	173 (41)
>2× ULN (2.4 mg/dL; 180 µmol/L)	40 (10)
Unknown	92 (22)
Lymphopenia (lowest value) within 2 wk prior to CMV pneumonia
>300 cells/µL	90 (21)
100–300 cells/µL	115 (27)
<100 cells/µL	182 (43)
Unknown	34 (8)
Mechanical ventilation at diagnosis of CMV pneumonia
No	271 (64)
Yes	146 (35)
Unknown	4 (1)

Abbreviations: CMV, cytomegalovirus; CMV-Ig, CMV-specific immunoglobulin; IVIG, pooled immunoglobulin; ULN, upper limit of normal.

^a The day of CMV pneumonia onset was the day when clinical manifestation attributed to CMV pneumonia was reported in the medical records. Three cases that were identified by autopsy report were included because they were managed empirically as CMV pneumonia (cases identified on the autopsy report that were never treated for CMV were not included in this study).

^b Includes 10 patients who received combination antiviral therapy (ganciclovir + foscarnet or ganciclovir + cidofovir). These patients did not differ in their outcome in univariate analysis and were therefore included in this group.

Outcome of CMV Pneumonia

Of 421 HCT recipients with CMV pneumonia, 368 (87%) died a median of 25 days (range, 1–8331 days; IQR, 10–105 days) after the onset of CMV pneumonia. Two hundred ninety-six deaths occurred within 6 months following CMV pneumonia onset; of these, 186 were attributed to CMV pneumonia. The overall survival estimate at 6 months after CMV pneumonia onset was 30% (95% confidence interval [CI], 25%–34%). Changes in overall and attributable mortality over time are displayed in Figure 1A and 1B.

Figure 1.

A–D, Time-to-event results of overall and cytomegalovirus (CMV)-attributable mortality over time (A and B) and by antiviral treatment regimen (C and D). E, Changes in the prevalence of important risk factors before and after the year 2000. For creatinine and bilirubin units, see Table 2. Abbreviations: CMV-Ig, CMV-specific immunoglobulin; IVIG, intravenous immunoglobulin; TTT, treatment.

Risk Factors Associated With Overall Mortality

Recipient sex, grade 3–4 acute and chronic GVHD, type of antiviral treatment, need for mechanical ventilation, lymphocyte count, bilirubin value, and creatinine value were associated with an increase (P < .1) in the risk of 6-month overall mortality in univariate analysis (Supplementary Table 1). Presence of lung copathogens at CMV pneumonia onset and time period of transplantation were kept in the multivariable model because of their confounding role between antiviral treatment and outcome. In the multivariable model, female sex, grade 3–4 acute GVHD, mechanical ventilation, and lymphopenia were independently associated with a higher risk of 6-month overall mortality (Table 3).

Table 3.

Multivariable Analysis of Overall and Attributable Mortality Within 6 Months From First Cytomegalovirus Pneumonia (n = 417)^a

Variable	Overall Mortality			CMV-Attributable Mortality
	HR	95% CI	P Value	HR	95% CI	P Value
Treatment of CMV pneumonia
Ganciclovir or foscarnet	1.0	…	…	1.0
CMV-Ig or IVIG + antiviral drug(s)	0.8	.6–1.1	.13	1.1	.7–1.7	.68
No treatment or CMV-Ig alone or high-dose acyclovir aloneb	1.7	1.0–3.0	.06	1.6	.7–3.3	.24
Sex
Male	1.0			1.0
Female	1.3	1.1–1.7	.01	1.4	1.1–1.9	.01
Year of transplantation
1986–July 1992	1.0			1.0
August 1992–1999	0.9	.7–1.2	.59	0.9	.6–1.2	.41
2000–2011	0.7	.5–1.0	.06	0.6	.4–.9	.01
Acute GVHD grade
0–2	1.0
3–4	1.4	1.1–1.7	.01
Copathogens at time of CMV pneumonia
None	1.0
Fungal or viral (± bacteria)	1.3	1.0–1.8	.05
Unknown	0.8	.5–1.3	.36
Lymphocyte count within 2 wk prior pneumonia
≥300 cells/µL	1.0	…	…	1.0
<300 cells/µL	1.8	1.3–2.5	<.001	1.7	1.1–2.7	.01
Unknown	1.6	.9–2.8	.13	1.7	.9–3.3	.09
Mechanical ventilation at diagnosis
No	1.0	…	…	1.0
Yes	3.1	2.4–3.9	<.001	4.0	2.9–5.4	<.001

Abbreviations: CI, confidence interval; CMV, cytomegalovirus; CMV-Ig, CMV-specific immunoglobulin; GVHD, graft-vs-host disease; HR, hazard ratio; IVIG, intravenous immunoglobulin.

^a Four patients were excluded from the analysis because of missing data on mechanical ventilation.

^b Includes patients who received IVIG/CMV-Ig or high-dose acyclovir alone for treatment during the early years of the study period.

Risk Factors Associated With Attributable Mortality

Female sex, white race, year of transplantation (before 2000), marrow/cord blood as stem cell graft source, use of anti–T-cell drugs within 6 months before CMV pneumonia onset, lack of antiviral monotherapy, hyperbilirubinemia, use of mechanical ventilation, and lymphocytopenia met the P < .1 threshold for association with the risk of 6-month attributable mortality in the univariate analysis (data not shown). Female sex, lymphopenia, and use of mechanical ventilation at time of CMV onset were independently associated with an increased risk of 6-month CMV-attributable mortality (Table 3); treatment strategies for CMV did not reach statistical significance (Tables 3 and 4; Supplementary Tables 2 and 3).

Table 4.

Multivariable Analysis of Overall and Cytomegalovirus (CMV)–Attributable Mortality Within 6 Months From First CMV Pneumonia, Including Only Patients Who Received Antiviral Agents Alone or No Treatment for CMV (n = 95)

Variable	Overall Mortality			CMV-Attributable Mortality
	HR	95% CI	P Value	HR	95% CI	P Value
Treatment of CMV pneumonia
Ganciclovir or foscarnet	1.0	…	…	1.0	…	…
No treatment or CMV-Ig alone or high-dose acyclovir alone	1.7	1.0–2.8	.05	1.6	.7–3.4	.23
Lymphocyte counts within 2 wk prior to pneumonia
≥300 cells/µL	1.0	…	…	1.0	…	…
<300 cells/µL	2.8	1.6–5.1	<.001	3.6	1.4–9.5	.01
Unknown	1.6	.6–4.5	.35	2.7	.6–11.6	.17

Patients who received pooled intravenous immunoglobulin or CMV-specific IgG were excluded.

Abbreviations: CI, confidence interval; CMV, cytomegalovirus; CMV-Ig, CMV-specific immunoglobulin; HR, hazard ratio; IgG, immunoglobulin G.

Effect of Treatment and Effect Modifiers on Overall and CMV-Attributable Death

Antiviral monotherapy (ganciclovir or foscarnet) significantly decreased overall mortality in the univariate analysis (hazard ratio [HR], 1.9 [95% CI, 1.2–3.2] for mortality for lack of treatment) (Figure 1C; Supplementary Table 1) but did not reach the level of significance for attributable mortality (HR, 1.8 [95% CI, .9–3–8]; Figure 1D). In a multivariate model that included all patients (Table 3), the protective effect showed a strong trend. To avoid overadjustment due to the small number of events, we compared monotherapy with no treatment in a bivariate model (Table 4), which confirmed the protective effect.

Adjunctive immunoglobulin treatment was not associated with a benefit in outcome (Table 3; Supplementary Tables 2 and 3). We explored whether the effect of transplant period and addition of immunoglobulin products (either CMV-Ig or pooled IVIG) to antiviral drug could be beneficial in specific subgroups of patients, identified by factors independently associated with an increased risk of death. These variables were high bilirubin value (>1 mg/dL; 17.1 mmol/L), lymphopenia (<300 cells/µL), mechanical ventilation at time of diagnosis, and transplantation before and after the year 2000 (Supplementary Figures 2 and 3). When stratified by ventilation status, survival was improved in both strata after the year 2000 (Supplementary Figure 2A and 2B). With regard to lymphopenia and hyperbilirubinemia at diagnosis, improvement in outcome in recent years was more apparent in patients without lymphopenia or hyperbilirubinemia (Supplementary Figure 2C–F). No consistent patterns were seen for adjunctive immunoglobulin treatment between subgroups and for overall and CMV-related mortality. Although there appeared to be trends toward improved overall survival among lymphopenic and hyperbilirubinemic patients, no such trends were seen with regard to CMV-related mortality (Supplementary Figure 3C–F).

Subset Analyses

Additional models of patients (1) with available data for bilirubin and creatinine (n = 305); (2) who survived at least 21 days following CMV pneumonia onset (n = 344); (3) who had viral load determined in BAL (n = 105); and (4) with available IgG levels prior to CMV pneumonia (n = 49) are shown in the Supplementary Data.

DISCUSSION

This study shows that the outcome of CMV pneumonia has improved modestly over the past few decades (Figure 1). The effect of time appeared to be consistent among the different risk subgroups and may be due to changes in the prevalence of important risk factors over time (Supplementary Figure 2). Contrary to earlier reports [12, 25], this analysis suggests a significant effect of antiviral monotherapy with ganciclovir or foscarnet, whereas no major overall benefit of immunoglobulin therapy could be demonstrated in multivariable models.

We found that the presence of fungal and viral copathogens, bone marrow or cord blood as stem cell source, concomitant liver and renal failure [26], and respiratory failure were all important factors associated with poor outcome. Not surprisingly, viral or fungal coinfections were associated with overall but not attributable mortality, probably due to the high mortality associated with these pathogens [27]. The effect was modest (Table 3) but clearly showed an additive effect of these pathogens to poor outcome. Somewhat unexpectedly, female sex was an important risk factor for poor outcome (Table 3). We were unable to find a statistically significant effect of viral load in the BAL on outcome (Supplementary Data).

Lymphopenia independently increased the risk of attributable and overall mortality. Moreover, in univariate analysis, anti–T-cell treatment within 6 months prior to pneumonia and bone marrow or cord blood as cell source vs PBSCs were associated with an increased risk of attributable mortality. Thus, our results are in accordance with previous reports showing that recovery of functional cytotoxic and helper T lymphocytes as well as natural killer cells are central in controlling active CMV infection [28]. To date, CMV-specific T cells have only rarely been used for the treatment of established CMV disease [29]. Our results might provide the rationale for adjunctive cellular therapy [28, 30–33].

The established practice for treatment of CMV pneumonia is to add high-dose IVIG or CMV-Ig to antiviral therapy, usually with ganciclovir. This recommendation is based on several uncontrolled studies showing improved outcome compared with historical controls [12–14]. We attempted to determine the impact of antiviral treatment and the adjunct use of immunoglobulin (CMV-Ig or IVIG) in our study. Contrary to early studies of ganciclovir as monotherapy [25], our data provide the best evidence to date that monotherapy with ganciclovir or foscarnet had a beneficial effect vs no therapy. Our analysis included the earlier reported patients [34], but the number of patients receiving monotherapy was larger, thereby permitting statistical adjustment for lymphopenia, an important risk factor for overall mortality. Notably, 21 of 22 patients without effective therapy died shortly after diagnosis, confirming the devastating effect of CMV pneumonia. The effect of antiviral monotherapy on overall survival compared with no treatment was significant in both univariate and adjusted models (Table 4; Supplementary Table 1). The effect size was similar for attributable mortality but statistical significance was not reached, likely due to the smaller number of events.

We were unable to demonstrate that combination therapy with antiviral therapy and pooled IVIG or CMV-Ig decreased the risk of overall and attributable mortality in the overall cohort and a more contemporary subset (Table 3; Supplementary Tables 2 and 3). The observed trend toward improved overall survival may support previous speculations that the use of pooled IVIG may indirectly increase survival of patients with CMV pneumonia by improving the control of bacterial or fungal infection and GVHD, 2 conditions modulated by CMV infection [35, 36]. Although a conclusive statement about adjuvant immunoglobulin in CMV pneumonia cannot be drawn from our analysis, the data suggest that, if there is a potential beneficial effect of the addition of immunoglobulins, the effect size is significantly smaller than previously believed, and very large cohort or randomized trials would be required to demonstrate statistical significance. We explored whether there were specific subgroups of patients who might benefit from immunoglobulin therapy. Analyzing specific subsets of patients with increased risk factors (Supplementary Figure 3), we found possible trends in beneficial effects on overall survival in patients with concomitant lymphopenia and liver failure. Notably, no beneficial effects or even trends in the opposite direction were seen in terms of CMV-attributable mortality.

Strengths of this study include the largest sample size of any CMV pneumonia outcome study to date and a highly guideline-driven treatment approach throughout the study period. The latter point is important, as treatment was generally applied according to the standards of the given time period rather than by attending physician preference. Limitations include the retrospective nature and an incomplete data set for some parameters (ie, BAL viral load, IgG levels). We do not view the long study period as a limitation, as one aim of this study was to analyze changes of outcome of CMV pneumonia over time. However, while we attempted to examine the impact of changes in transplant methodology over time, there may have been some supportive care factors that we were unable to analyze. Our study was nonrandomized but used statistical methodology that accounted for several of the pitfalls of the studies that were used to establish the current standard of care (ie, small sample size, lack of controls, lack of multivariable modeling).

In conclusion, the results of this study suggest a modest improvement in outcome of CMV pneumonia over the past 25 years, but mortality continues to be unacceptably high. Our results suggest that the introduction of effective antiviral drugs in the late 1980s was an important advance. This study adds to the controversy regarding the benefit of using immunoglobulin products as adjunct therapy for CMV pneumonia after HCT [11, 37], as it suggests that the benefits are much less impressive than previously reported. Validation of our findings are needed using another larger data set and to help provide definitive guidance for the standard of care of CMV pneumonia after HCT. The findings from this study also provide the rationale for novel treatment strategies, such as T-cell therapy and combination antiviral therapy. The continued high mortality and lack of significant progress over the past few decades emphasize the need for randomized controlled trials for this disease. Although the frequency of CMV pneumonia has declined, large networks such as the National Heart, Lung, and Blood Institute/National Cancer Institute's Blood and Marrow Transplant Clinical Trials Network and/or international networks could conduct such a trial. With new antiviral drugs in advanced stages of development [38, 39], combination therapy regimens will likely soon be available for evaluation. Also, advances in CMV-specific T-cell therapy may now facilitate using cellular therapy as a treatment option [33]. The experience presented here reemphasizes the known limitations of using phase 2 data for establishing a standard of care, which then makes randomized trials difficult, if not impossible, to conduct due to perceived lack of clinical equipoise. Optimized prevention strategies for CMV pneumonia after HCT are needed.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.