CMV indicates cytomegalovirus; FFP, fresh frozen plasma; NAT, nucleic acid testing; RBC, red blood cell; and WBCs, white blood cells.
aThe weight inclusion criterion was changed in November 2010 to increase enrollment.
A and B, The cumulative incidence of CMV infection and mortality during the study period, including 95% CIs at weeks 4, 8, and 12. C, Log10 CMV viral load values were analyzed with a repeated-measures model; mean estimates and their 95% CIs were back transformed to the original scale and are reported as the geometric mean with 95% CIs. BM indicates breast milk; NAT, nucleic acid testing; and limit lines, 95% CI.
eMethods. Additional methods for laboratory testing and statistical analysis
eFigure 1. Percentage of longitudinal blood and urine samples with CMV NAT expression
eFigure 2. ROC curve for maximum log10 CMV NAT expression in breast milk as a predictor of mother-to-infant CMV transmission among 189 CMV seropositive mothers who fed their infants CMV-positive maternal breast milk
eResults. Results for CMV-infected infant laboratory abnormality, statistical model adequacy
eTable. Factors associated with mother-to-infant CMV transmission by breast milk among CMV seropositive mothers
Customize your JAMA Network experience by selecting one or more topics from the list below.
Josephson CD, Caliendo AM, Easley KA, et al. Blood Transfusion and Breast Milk Transmission of Cytomegalovirus in Very Low-Birth-Weight Infants : A Prospective Cohort Study. JAMA Pediatr. 2014;168(11):1054–1062. doi:10.1001/jamapediatrics.2014.1360
Postnatal cytomegalovirus (CMV) infection can cause serious morbidity and mortality in very low-birth-weight (VLBW) infants. The primary sources of postnatal CMV infection in this population are breast milk and blood transfusion. The current risks attributable to these vectors, as well as the efficacy of approaches to prevent CMV transmission, are poorly characterized.
To estimate the risk of postnatal CMV transmission from 2 sources: (1) transfusion of CMV-seronegative and leukoreduced blood and (2) maternal breast milk.
Design, Setting, and Participants
A prospective, multicenter birth-cohort study was conducted from January 2010 to June 2013 at 3 neonatal intensive care units (2 academically affiliated and 1 private) in Atlanta, Georgia. Cytomegalovirus serologic testing of enrolled mothers was performed to determine their status. Cytomegalovirus nucleic acid testing (NAT) of transfused blood components and breast milk was performed to identify sources of CMV transmission. A total of 539 VLBW infants (birth weight, ≤1500 g) who had not received a blood transfusion were enrolled, with their mothers (n = 462), within 5 days of birth. The infants underwent serum and urine CMV NAT at birth to evaluate congenital infection and surveillance CMV NAT at 5 additional intervals between birth and 90 days, discharge, or death.
Blood transfusion and breast milk feeding.
Main Outcomes and Measures
Cumulative incidence of postnatal CMV infection, detected by serum or urine NAT.
The seroprevalence of CMV among the 462 enrolled mothers was 76.2% (n = 352). Among the 539 VLBW infants, the cumulative incidence of postnatal CMV infection at 12 weeks was 6.9% (95% CI, 4.2%-9.2%); 5 of 29 infants (17.2%) with postnatal CMV infection developed symptomatic disease or died. A total of 2061 transfusions were administered among 57.5% (n = 310) of the infants; none of the CMV infections was linked to transfusion, resulting in a CMV infection incidence of 0.0% (95% CI, 0.0%-0.3%) per unit of CMV-seronegative and leukoreduced blood. Twenty-seven of 28 postnatal infections occurred among infants fed CMV-positive breast milk (12-week incidence, 15.3%; 95% CI, 9.3%-20.2%).
Conclusions and Relevance
Transfusion of CMV-seronegative and leukoreduced blood products effectively prevents transmission of CMV to VLBW infants. Among infants whose care is managed with this transfusion approach, maternal breast milk is the primary source of postnatal CMV infection.
clinicaltrials.gov Identifier: NCT00907686
Transfusion-transmitted cytomegalovirus (TT-CMV) and breast milk–transmitted CMV (BM-CMV) infections can cause serious morbidity and mortality in immunologically immature, very low-birth-weight (VLBW) infants (birth weight, ≤1500 g). Transfusion of CMV-seronegative and/or leukoreduced blood components is a common strategy used to prevent TT-CMV; however, studies1-3 conducted to validate this approach were small and yielded imprecise estimates of TT-CMV risk. Many of these studies1-3 did not address factors associated with breakthrough cases of TT-CMV including leukoreduction quality control (linked to white blood cell filter failures and CMV transmission) and donor window period infections (when immunologically based assays may not detect CMV viremia).4 Additionally, studies of TT-CMV have not systematically evaluated BM-CMV, which may confound identification of the source of the infection. The burden of BM-CMV in VLBW infants has not been well quantified.5 Other, less common, sources of CMV in this population are genital secretion from CMV-seropositive mothers and community-acquired transmission.6,7
We performed a multicenter, prospective, birth-cohort study to quantify the risk of CMV infection from transfusion of CMV-seronegative and leukoreduced blood components. We also evaluated CMV transmission from maternal breast milk among infants who were fed breast milk and applied CMV nucleic acid testing (NAT) to transfused blood products and breast milk samples to determine the source of CMV in cases of postnatal transmission.
Infants born at 3 Atlanta-area hospitals (2 academically affiliated institutions [Emory University Hospital–Midtown and Grady Memorial Hospital] and 1 private hospital [Northside Hospital]) were screened (Figure 1). The institutional review boards of all centers approved the study. The participants did not receive financial compensation. Infants meeting the study criteria and whose parent or guardian gave written informed consent were enrolled and monitored from birth to 90 postnatal days, hospital discharge, or death. Infants transferred to Children’s Healthcare of Atlanta hospitals were monitored at that hospital. Race and/or ethnicity, known to be associated with CMV infection, was determined by maternal report from options defined by federally funded study guidelines.8
Maternal serum at study entry was tested with a CMV IgG/IgM assay. If the result of the serology test was positive, the sample was retested by an IgM-specific assay. For seronegative mothers, CMV NAT was performed on a maternal blood sample at study entry and conclusion to exclude infection that developed during the study.
Cytomegalovirus infection was prospectively evaluated in all infants through CMV NAT of residual blood samples and urine. Congenital CMV infection was defined as a positive result of CMV NAT (or positive viral culture obtained from clinician-ordered testing) in blood or urine samples within 2 weeks of life. Postnatally acquired CMV infection was defined as a positive CMV NAT or viral culture result in blood or urine after 2 weeks of life with a previously documented negative result.5 Transfusion-transmitted CMV was defined as a positive result of CMV NAT performed on blood products that were transfused to the infant combined with a positive result of CMV NAT detected in the infant’s blood or urine after transfusion. Blood was tested on the day of birth and at 4 time windows (±4 days) near days of life 21, 40, 60, 90, and at discharge; urine was collected on the day of birth and at discharge if blood was not available. In the event of clinical or laboratory suspicion of CMV infection, results from clinician-initiated CMV testing were included in the study data. At least 2 times per week, screening for CMV-associated disease, defined as pneumonitis, hepatitis, abnormal hematologic indices, or fever in the setting of CMV infection, was performed by study personnel.
All transfused red blood cell and apheresis platelet units were CMV seronegative, leukoreduced before storage, and irradiated (some before and some after storage); residual leukocyte quantitation and CMV NAT were performed on samples from these products (Figure 1). Breast milk samples were obtained from lactating mothers during weeks 1, 3, and 4, as well as days 34 to 40.9 If postnatal CMV infection was detected, CMV NAT was immediately performed on available milk samples. Otherwise, breast milk was stored and batch tested once an infant reached the study end point. Positive results of CMV tests were reviewed immediately by the study investigators (C.D.J. and A.M.C.) and reported to the patient’s treating neonatologist who determined further evaluation and/or treatment.
The presence of IgG/IgM-polyspecific CMV antibodies in maternal blood was determined by a US Food and Drug Administration–approved commercial serology assay (Immucor). Serum samples were tested for CMV IgM by enzyme-linked immunosorbent assay (Bio-Quant). Nucleic acid extraction for CMV NAT was performed using a commercial product (EZ1 Virus Mini Kit, version 2.0; Qiagen, Inc) All assays were performed following the manufacturers’ protocols. Nucleic acid testing was performed with a polymerase chain reaction kit (Artus CMV TM, using the Roto-Gene instrument; Qiagen, Inc).10 The polymerase chain reaction assay was validated on whole blood, urine, and breast milk samples, calibrated to the first World Health Organization international standard.11 Newly diagnosed CMV in infants with a viral load of greater than 300 IU/mL was verified by repeating the first extraction as well as a new extraction. If there was an insufficient amount of a specimen for a second test, it was diluted 1:2. Any specimen that tested positive with a viral load of less than 300 IU/mL was repeated in duplicate and reported as low positive (<300 IU/mL). Specimens discordant on a second test were reported as indeterminate. Prior to testing breast milk samples were stored at 4°C for up to 7 days and at −20°C for long-term storage. Blood samples were stored at 4°C and tested within 7 days of collection. To quantify residual white blood cells in leukoreduced cellular blood products, a 100-µL volume of blood was added to 400 µL of propidium iodide/RNase reagent (Leucocount; BD Biosciences) and analyzed by flow cytometry (3Ti).
The onset of CMV infection time was estimated as the midpoint between the last negative result of CMV NAT and the first positive NAT result in blood or urine. The incidence of first-time CMV infection and death was estimated by the cumulative incidence function.12 A competing risk analysis was done to estimate the cause-specific hazard ratio (CSHR) and the subdistribution hazard ratio for CMV and mortality using a Cox regression model. The 95% CIs were calculated using the Wilson score method and were 2-sided except in cases where the incidence was zero. In those cases a 1-sided upper limit confidence boundary is reported.10 The cumulative incidence of CMV infection at given time points was estimated from the CMV cumulative incidence function. The CIs were estimated using bootstrapping by mother as the clustering unit (1000 bootstrap samples). The competing risk model for the CSHR was implemented with SAS PHReg, version 9.3 (SAS Institute Inc), using robust sandwich covariance estimates to account for within-mother correlation that may occur in outcomes of multiple-birth infants.13 Additional methods are contained in the eMethods in the Supplement.
From January 16, 2010, to June 11, 2013, a total of 541 VLBW infants born to 462 mothers were enrolled; 2 were excluded from follow-up (Figure 1). Three hundred fifty-two of the mothers (76.2%) tested positive for the CMV IgG/IgM combination test and, of these women, 11 (3.1%) tested positive for CMV IgM antibody. Infants born to CMV-seropositive or CMV-seronegative mothers did not differ significantly in baseline characteristics except for race and Apgar score (Table 1). The maternal groups did not differ significantly except in receipt of prenatal care and isolated spontaneous labor as an indication for premature delivery.12 Three hundred seventy-one of the mothers (80.3%) fed breast milk to at least 1 of their infants, and the median duration of breastfeeding was 38 days (interquartile range, 19-56 days).
Cytomegalovirus infection was detected in 29 infants (5.4% of the cohort) (Table 2). The cumulative incidence of postnatal CMV infection at 12 weeks was 6.9% (95% CI, 4.2%-9.2%) (Figure 2A). Five of 29 (17.2%) CMV-infected infants developed CMV disease and/or death (Figure 2B). All 29 infants with CMV infection had blood or urine CMV NAT performed within the first 5 days of life. Twenty-seven infants (93.1%) had CMV NAT performed on their blood and 25 infants (86.2%) had CMV NAT performed on their urine. One infant had positive blood and urine results, consistent with congenital infection, and the results of initial CMV NAT were negative in all remaining infants.
The percentage of longitudinal blood and urine samples from 539 VLBW infants with detectable CMV increased from 0.5% at 1 to 3 weeks to 3.2% at 4 to 6 weeks. By 10 to 12 weeks, 9.1% (95% CI, 4.9%-16.8%) of the samples had detectable virus (eFigure 1 and eMethods in the Supplement). Of 29 infants with CMV infection, virus was detected in blood samples of 26 infants and in urine samples of 16 infants (eFigure 1 in the Supplement). With mixed linear models used to account for multiple tests from each infant, the geometric mean viral load detected in the infants was estimated to be 2887 IU/mL (95% CI, 1462-5703) in blood and 133 783 IU/mL (95% CI, 23 922-748 170) in urine. Of the 27 mothers with infants who developed postnatal CMV infection, only 2 women (7.4%) had a positive CMV IgM test. Furthermore, of the 11 mothers who tested positive for IgM antibody, only 2 (18.2%) had an infant with CMV infection.
Five of 29 infants (17.2%) with CMV infection had abnormal laboratory values at the time of initial detection of CMV (details available in eResults in the Supplement). Among the 24 (82.8%) infants determined to have asymptomatic CMV infection, including 1 infant with a congenital infection, no laboratory abnormalities associated with CMV were detected up to 10 days before diagnosis of CMV infection (details available in eResults in the Supplement). Furthermore, no clinical suspicion of disease occurred for these 24 infants, and no further investigation or antiviral treatment was pursued. However, 5 CMV-infected infants developed disease or died. Infants with CMV disease or associated mortality had viral loads similar to those of infants with asymptomatic CMV infection. One infant died of pneumonia following the development of necrotizing enterocolitis (NEC). This infant had a maximum CMV viral load of 13 000 IU/mL. Two other infants died of NEC with viral loads at death of 8000 and 4000 IU/mL. The 2 surviving infants who developed CMV disease, one with punctate densities in the basal ganglia consistent with early signs of CMV infection and the other with a sepsis-like syndrome, were the only infants who received ganciclovir and/or valganciclovir treatment. Both patients had clinical improvement with treatment. All infants with CMV disease or associated mortality received only frozen/thawed breast milk and had negative initial CMV testing in the first 2 weeks of life.
A total of 310 infants (57.5%) received 1 or more transfusions. A total of 2061 transfusions were administered from 1038 cellular blood components during the study (1545 red blood cell transfusions from 703 units, 379 platelet transfusions from 251 units, 129 fresh frozen plasma transfusions from 76 units, and 8 cryoprecipitate transfusions from 8 units). The overall TT-CMV incidence for infants was 0.0% (95% CI, 0.0%-0.9%); similarly, the TT-CMV incidence from 880 CMV seronegative and leukoreduced cellular blood components was 0.0% (95% CI, 0.0%-0.3%) (Table 2). One platelet unit had a leukoreduction failure (5.2 × 106 residual leukocytes), for an overall failure incidence of 0.11% (95% CI, 0.02%-0.6%). All blood components showed negative results on CMV NAT. The unit that failed leukoreduction was not associated with CMV transmission.
All 28 infants with postnatal CMV infection were fed maternal breast milk from CMV-seropositive mothers. Twenty-seven of these infants (96.4%) received maternal breast milk with positive CMV NAT before BM-CMV from 26 mothers (1 set of twins). The mean (SD) time from the first detection of CMV in maternal breast milk to the first detection of postnatal CMV infection in the infants was 36 (22) days. The source of CMV infection for the 28th infant, born to a CMV-seropositive mother, could not be identified. This infant’s CMV infection was detected by NAT on day of life 25, prior to any blood transfusion and after receipt of breast milk with negative CMV NAT (tested in week 1). The 12-week incidence of CMV infection among infants fed CMV-positive breast milk was 15.3% (95% CI, 9.3%-20.2%) (n = 221) (Table 2 and Figure 2B).
Overall, 74.1% (95% CI, 69.7%-80.3%) of CMV-seropositive mothers had CMV DNA lactia in their expressed breast milk, compared with 0% (95% CI, 0.0%-4.5%) of CMV-seronegative mothers (Table 2). Once CMV was initially detected in breast milk, all subsequent breast milk samples contained CMV DNA. Of 189 mothers with CMV-positive breast milk, 26 women (13.8%) were CMV transmitters and 163 (86.2%) were CMV nontransmitters. Mean breast milk CMV viral loads according to NAT were similar for transmitting and nontransmitting mothers at week 1 (1306 vs 664 IU/mL; P = .13) but became significantly higher in CMV-transmitting mothers during postpartum weeks 2 to 3 (9129 vs 2033 IU/mL; P < .001) and in weeks 4 to 5 (20 421 IU/mL vs 3064 IU/mL; P < .001) (Figure 2C).
Most breast milk–fed infants (78.2%) received exclusively frozen/thawed milk. The CMV transmission rate from breast milk for 221 infants fed CMV-positive breast milk did not differ significantly between infants who were fed some fresh breast milk vs those fed exclusively frozen/thawed milk (12-week CMV incidence, 17.6% vs 11.6%; HR, 0.55; 95% CI, 0.19-1.56; P = .26).
Factors that increased the risk for postnatal CMV infection included a higher number of breast milk–feeding days, higher breast milk CMV viral load, and premature rupture of membranes (PROM) (Table 3). The adjusted hazard of CMV infection increased as the breast milk CMV viral load increased and the hazard was more than 3 times higher for infants born to mothers with PROM prior to delivery compared with infants born to mothers with other indications for preterm delivery (Table 4). PROM was also associated with an increase in the cumulative incidence of CMV infection (subdistribution hazard rate, 3.07; 95% CI, 1.31-7.18; P = .01).
Furthermore, PROM was an independent predictor of mother-to-infant CMV transmission among 189 CMV-seropositive mothers, whereas the mode of delivery was not associated with mother-to-infant transmission (eTable and eMethods in the Supplement). In addition, maximum log10 CMV expression in breast milk was associated with mother-to-infant CMV transmission among CMV-seropositive mothers, although the accuracy of CMV viral expression in breast milk to identify postnatal CMV infection was poor, as reflected by the receiver operating characteristic curve (eFigure 2 and eMethods in the Supplement).
To our knowledge, the present prospective, multicenter, birth-cohort study is the largest reported evaluation of both blood transfusion and breast milk sources of postnatal CMV infection in VLBW infants. Prior to our study, the residual risks of TT-CMV with CMV-seronegative or leukoreduced transfusions were estimated to be 1% to 3%.2-4,16 Furthermore, the efficacy of combining both approaches had not been rigorously examined.17 The present results demonstrate that the exclusive use of blood components that are both CMV-seronegative and leukoreduced is effective in preventing TT-CMV. We believe that this approach should be adapted as a standard of care when administering transfusions to VLBW infants until the comparative effectiveness of alternative transfusion strategies to prevent TT-CMV can be evaluated.
Historically, failure to prevent TT-CMV with CMV-seronegative units of blood components was attributed to donors in the window phase of an infection,14 whereas leukoreduced units were believed to transmit CMV if the leukoreduction filters failed.15,18 In our study, only 1 unit had a filter failure and no donor-window–phase infections were identified. Thus, recent advances in serologic and leukoreduction methods may account for the effectiveness of the combined approach to prevent TT-CMV.
The American Academy of Pediatrics19 states that, “the value of routinely feeding [fresh] human milk from [CMV] seropositive mothers to preterm infants outweighs the risks of clinical disease, especially because no long-term neurodevelopmental abnormalities have been reported.” Given the benefits of breastfeeding, new strategies to prevent BM-CMV are needed, because freezing and thawing breast milk did not completely prevent transmission in the present study. Alternative approaches may include routine CMV-serologic testing of pregnant women to enable counseling regarding the risk of CMV infection and risk stratification of infants.20 For feeding breast milk to VLBW infants born to seropositive mothers, pasteurization of breast milk until a corrected gestational age of 34 weeks, as is recommended by the Austrian Society of Pediatrics,21,22 and routine screening for postnatal CMV infection may be warranted. Given the toxicity of antiviral therapy,23 further research is needed to determine whether antiviral treatment in infants with asymptomatic CMV infection is beneficial, especially because it is unclear which infants will progress to CMV disease. Although we found an association between CMV DNA levels in breast milk and BM-CMV, we could not identify a viral load cut-off below which BM-CMV did not occur. Thus, any level of CMV DNA in breast milk should be considered potentially infectious until more detailed investigations can be performed. We also found that PROM and the amount of CMV virus in breast milk were independently associated with an increased risk of postnatal CMV infection. The role of PROM in postnatal CMV infection is unclear. Two studies24,25 have reported that intrauterine CMV infection is not associated with PROM. Furthermore, vaginal delivery was not associated with postnatal infection in our study, making intrapartum acquisition an unlikely source of postnatal CMV infection.
Our study has several limitations. We did not compare the relative risk of TT-CMV between CMV-seronegative and CMV-leukoreduced units and blood components that were leukoreduced from only CMV untested donors. Therefore, we could not determine the relative safety of the latter approach in VLBW infants. Furthermore, we were unable to test all breast milk for enrolled infants since samples were not available or mothers were not breastfeeding during the evaluation period. Also, we were unable to ascertain with certainty whether CMV infection caused NEC or was simply a co-occurrence in the 3 infants with CMV infection who died, although CMV infection is a reported cause of NEC.26 We were also unable to test genital tract secretions at delivery to identify this potential source of CMV infection owing to the complexity of our enrollment at sites involving numerous obstetrical practices. However, most infants in this study were delivered by cesarean section, and we did not detect an association between mode of delivery and mother-to-infant transmission of CMV infection. Furthermore, the results of all urine and blood CMV NAT of infants, with the exception of 1 infant with congenital CMV, were negative in the first 2 weeks of life. This makes the possibility that we misclassified infants with congenital CMV infection as having postnatal CMV infection unlikely. Finally, we did not perform systematic hearing assessments or long-term neurodevelopmental assessments.
The effect of asymptomatic postnatal CMV infection on long-term neurodevelopmental outcomes is unclear, with some studies demonstrating an increased risk of adverse neurologic outcomes27 and others revealing no difference in long-term outcomes28 or suspected sensorineural hearing loss.28,29
Transfusion of CMV-seronegative and leukoreduced blood products effectively prevents transmission of CMV to VLBW infants. Among infants whose care is managed with this transfusion approach, maternal breast milk from CMV seropositive mothers, is the primary source of postnatal CMV infection. The frequency of CMV infection in our cohort raises significant concern regarding the potential burden of CMV infection among VLBW infants and potential sequelae. This concern necessitates large, long-term follow-up studies of neurodevelopmental outcomes in infants with postnatal CMV infection.
Accepted for Publication: June 18, 2014.
Corresponding Author: Cassandra D. Josephson, MD, Department of Pathology, Children’s Healthcare of Atlanta, 1405 Clifton Rd NE, Atlanta, GA 30322 (firstname.lastname@example.org).
Published Online: September 22, 2014. doi:10.1001/jamapediatrics.2014.1360.
Author Contributions: Dr Josephson had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Josephson, Caliendo, Knezevic, Shenvi, Hillyer, Roback.
Acquisition, analysis, or interpretation of data: Josephson, Caliendo, Easley, Knezevic, Shenvi, Hinkes, Patel, Roback.
Drafting of the manuscript: Josephson, Easley, Knezevic, Shenvi, Patel, Hillyer.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Josephson, Easley, Knezevic, Shenvi, Patel.
Obtained funding: Josephson, Roback.
Administrative, technical, or material support: Josephson, Caliendo, Patel, Hillyer, Roback.
Study supervision: Josephson, Caliendo, Knezevic, Shenvi, Hillyer, Roback.
Conflict of Interest Disclosures: Dr Caliendo’s laboratory used Qiagen products during the study that were supported, in part, by the company. Dr Roback has partial ownership in 3Ti, whose instrument was used to count residual white blood cells for the donor blood products.
Funding/Support: The study was supported by grant P01 HL086773 from the National Heart, Lung, and Blood Institute (NHLBI) of the National Institutes of Health to Emory University School of Medicine.
Role of the Funder/Sponsor: The NHLBI program officers provided advice on the study design, but had no role in the conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We thank all of the families who volunteered to participate in this study and the nurses, laboratory technologists, and physicians for their dedication, critical thinking, and kindness. Deborah Abdul-Ali, BS, MT(ASCP), Jessica Ingersoll, BS, MS, MT(ASCP), and Doris Igwe, BS, MLS(ASCP) (Emory University Department of Pathology), performed CMV NAT testing; Jane Skvarich, BSN, MN (Emory University, Department of Pathology), Katrina H. Grier, BSN (Northside Hospital), and Janna M. Benston, BSN (Northside Hospital), served as research nurses; Shieghla Barclay, BS, MT(ASCP) (Emory University, Department of Pathology), performed residual white blood cell counting; Natia Saakadze (Emory University, Department of Pathology) performed serology testing and specimen processing; and Martha Sola-Visner, MD (Harvard University Medical School), provided editorial comments on the manuscript. No contributors received financial compensation.