Congenital Cytomegalovirus Infection in Pediatric Hearing Loss

Objectives  To compare the prevalence of congenital cytomegalovirus (CMV) infection in Washington State in children with hearing loss (HL) and the general population and to compare the characteristics of HL in children with and without congenital CMV infection.

Design  Matched case-control; case cohort.

Setting  Regional pediatric hospital, Washington State Department of Health (WSDOH).

Patients  Cases were children 4 years and older with HL born in Washington State. Control individuals matched for demographic characteristics were identified at random through the WSDOH.

Main Outcome Measures  Congenital CMV status determined using quantitative polymerase chain reaction testing on newborn heel stick blood spots archived by the WSDOH. Audiologic data were used to characterize HL.

Results  Congenital CMV testing was performed for 222 matched cases and controls. Congenital CMV infection was detected in 1.4% of controls and in 9.9% of cases (odds ratio, 10.5; 95% confidence interval, 2.6-92.4). An estimated 8.9% of HL in children in Washington can be attributed to CMV infection. After inclusion of an additional 132 children with HL (for a total of 354 cases in the case cohort), we observed that children with congenital CMV had more severe HL (P < .001) and higher proportions of progressive (P = .02) and unilateral (P = .002) HL compared with children without congenital CMV infection. In the 35 children with congenital CMV infection, there was no relationship between neonatal CMV load and severity of HL.

Conclusions  In Washington State, children with HL had a far higher prevalence of congenital CMV viremia than did the general pediatric population, and CMV infection seems to be responsible for an appreciable fraction of pediatric HL in Washington State.

Our understanding of the causes of hearing loss (HL) has grown considerably in the past 30 years. More than 300 mutations have been associated with childhood HL. The main nongenetic causes of HL include environmental exposures and infections. With successful immunization campaigns, many of the historically common infectious causes of HL, such as rubella and measles, have been nearly eradicated in the United States.

One of the most common causes of pediatric HL is congenital cytomegalovirus (CMV) infection. Previous studies^1-4 have suggested that 10% to 60% of children with HL had congenital CMV infection compared with a prevalence in the general population of 0.18% to 2.1% depending on the geographic location. In children with congenital CMV infection, it has been estimated that 35% to 65% of symptomatic and 7% to 15% of asymptomatic newborns will ultimately develop HL, with a median age at onset of 44 months.^1,5

In landmark studies performed at the University of Alabama, in which the prevalence of congenital CMV in the general population was 1.3%,⁶ 860 children with congenital CMV infection were identified and observed across time using audiologic testing. In this population, 15% developed sensorineural HL. Of the children with HL, 40% had unilateral HL and 60% bilateral HL; 54% had progressive HL.⁵

One of the challenges in diagnosing congenital CMV infection is the requirement to initiate diagnostic testing before 4 weeks of age because postnatal exposure to CMV is not associated with HL. Even if newborn hearing screening is used, by the time an infant is referred to an otolaryngologist for evaluation of HL, the opportunity to diagnose congenital CMV has often passed. The importance of identifying CMV as an etiology of HL in infants has become clinically relevant with the availability of oral antiviral agents that may prevent the progression of CMV-related HL.⁷ Furthermore, children with congenital CMV are at risk for progressive HL that may not be present until several years of age, at a time when the ability to diagnose congenital CMV using routine methods has passed.⁸

Investigation of congenital CMV infection has been facilitated by the development of techniques that allow detection of CMV from archived newborn heel stick blood spots,^9,10 which are collected routinely in many parts of the United States as well as in Europe and Asia. In addition to the ability to identify CMV in archived blood specimens, quantitative polymerase chain reaction (PCR) methods have allowed measurement of viral load in these blood spots, even when they have been stored for many years.¹¹

In areas of the United States such as the Pacific Northwest, where the ethnic and racial distribution is markedly different from that in Alabama, the prevalence of congenital CMV infection is unknown, as is the effect of congenital CMV infection in children with HL. The objectives of this study were to compare the magnitude of the association between congenital CMV infection and HL in children in Washington State, to describe characteristics of HL associated with congenital CMV infection, and to determine whether CMV viral load at birth was associated with severity of HL.

This retrospective study was approved by the institutional review boards of Seattle Children's Hospital and the University of Washington, Seattle, as well as by the Washington State Department of Health (WSDOH), Olympia. Written informed consent was obtained accordingly. Candidate cases were identified through Seattle Children's Hospital, which serves as a referral center for children with HL throughout the state, using International Classification of Diseases, Ninth Revision, codes for sensorineural HL. Sensorineural HL was confirmed by reviewing audiologic data. Recruitment was performed between June 1, 2007, and April 30, 2009. The inclusion criteria included age older than 4 years to allow for the development of delayed-onset HL,¹² birth in Washington State to allow analysis of the archived newborn heel stick blood spot, and the presence of bone conduction thresholds of 25 dB or greater at 2 or more frequencies. In Washington State, newborn heel stick blood spot cards are collected for all live births except those at military hospitals. Control individuals were identified at random through the WSDOH and were matched for date of birth, sex, race, and hospital of birth, as recorded on the newborn dried blood spot cards. Because the identity of controls was not made available to the investigators, the requirement for informed consent was waived for these individuals. The hearing status of the controls was unknown, but they were conservatively assumed to have normal hearing.

Medical records were reviewed and data were collected using a standardized abstraction form before CMV testing results were known. Information was collected about demographic characteristics, relevant neonatal and ongoing medical problems, and characteristics of HL, including audiologic data. Sensorineural, progressive, and fluctuating HL were defined as previously described.^5,13-16 Pure-tone averages were calculated using air conduction thresholds at 1000, 2000, and 4000 Hz. For cases with multiple audiograms in the medical records, the most recent audiogram was used for calculation of the pure-tone average.

Archived dried blood spots were retrieved through the WSDOH, and quantitative PCR amplification of UL55/UL123-exon 4 was performed in triplicate.¹⁷ The cell number in each blood spot was measured by beta-globin PCR¹⁸ and was used to standardize titers of CMV. Specimens in which all 3 runs demonstrated the presence of CMV were considered positive to have the most conservative interpretation of the results. Therefore, the 9 specimens with trace positivity in 1 or 2 runs were considered negative.²

The McNemar test was used to compare matched cases and controls. The case cohort was then expanded to allow comparisons of descriptive characteristics between hearing-impaired children with and without congenital CMV infection. These analyses were performed using χ² and 2-sided t tests as appropriate. Logistic regression analysis was used to evaluate the relationship between viral load and characteristics of HL in children with congenital CMV infection. Statistical analysis was performed using a software program (STATA, version 9.2; StataCorp LP, College Station, Texas).

Of 645 eligible patients identified through Seattle Children's Hospital, 354 (54.9%) completed the study: 222 patients for whom matched controls were identified to allow for case-control analysis and an additional 132 patients who were included for the remainder of the analysis.

The remaining eligible patients were excluded because of lack of telephone contact (n = 128, 19.8%), no interest in participation (n = 117, 18.1%), incomplete or unattainable consent (n = 14, 2.2%), death (n = 13, 2.0%), unattainable blood spots due to military hospital birth or inability to locate blood spots in the WSDOH archives (n = 10, 1.6%), or other reasons (n = 9, 1.4%).

A total of 222 cases were matched with controls identified through the WSDOH as described previously herein. Of these, 50.5% were male and most were white, similar to the Washington State population.¹⁹ In cases, the prevalence of congenital CMV infection was 22 of 222 (9.9%). In matched controls, the prevalence of congenital CMV infection was 3 of 222 (1.4%).

Of the 222 matched case-control pairs, 1 pair had a CMV-positive case and a CMV-positive control, 2 pairs had a CMV-positive control and a CMV-negative case, 21 pairs had a CMV-positive case and a CMV-negative control, and 198 pairs had a CMV-negative case and a CMV-negative control (Table 1). Using the results of the discordant pairs, as is customary in analyzing matched case-control data, these results gave an odds ratio of 10.5 (95% confidence interval [CI], 2.6-92.4) for the presence of congenital CMV infection in children with HL. Population-attributable risk (defined as [(odds ratio − 1) ÷ odds ratio] × prevalence among cases) was 8.9%, meaning that 8.9% of HL in this population was attributable to congenital CMV infection.²⁰

Recruitment of the additional 132 cases led to a total enrollment of 354 children with HL, as described previously herein. In these cases, the mean age was 11.5 years, and the sexes were equally represented. Most cases were white (74.6%). Among respondents, the most commonly reported household income was $50 000 to $120 000 per year. Mean maternal age at the child's birth was 30.7 years (Table 2).

Of the 354 cases, congenital CMV infection was detected in 35. Comparison of hearing-impaired children with and without congenital CMV infection demonstrated no appreciable differences for the demographic characteristics examined except for birth weight, which was slightly lower in children with congenital CMV infection (Table 2).

In CMV-positive and CMV-negative cases, the mean age at which HL was first suspected was slightly older than 3 years. Cases with congenital CMV infection tended to have a higher proportion of unilateral HL with a contralateral normal ear than did cases without congenital CMV infection (P = .002) and rarely had mixed HL (P = .03). The contribution of Eustachian tube dysfunction or otitis media with effusion to the cases of mixed HL was not specifically assessed. Cases with congenital CMV infection also more commonly had progressive HL (P = .02). On average, cases with congenital CMV infection had worse pure-tone averages in their worse-hearing ear (mean, 97.2; 95% CI, 90.7-103.7) compared with cases without congenital CMV infection (67.9; 64.6-71.2) (P < .001). Current educational settings were similar between congenital CMV-positive and CMV-negative cases, with the largest subgroups in mainstream schools and the remaining distributed among special education, deaf/hard-of-hearing programs, homeschool, other, and unknown (Table 3).

The clinically attributed etiology of HL or risk factors identified by the Joint Committee on Infant Hearing²¹ were gleaned from the medical records. Of patients with congenital CMV infection detected via the newborn blood spot, 28.6% had their HL attributed to a congenital infection. An additional 14.3% were associated with other Joint Committee on Infant Hearing risk factors. However, more than half of the cases with congenital CMV infection had no clinically presumed etiology or risk factors for HL identified (Table 4). Of CMV-positive individuals, 16 were symptomatic in the neonatal period and 13 were asymptomatic (determined using criteria described by Fowler et al¹⁴), and there was insufficient information in the medical record to determine the neonatal status of the remaining 6 individuals. Relatively little information on other associated symptoms was available in the medical records, but few of the individuals were documented as having intracranial calcifications, hepatosplenomegaly, or thrombocytopenia. Of individuals whose clinically attributed etiology of HL was congenital infection, 1 had rubella and 1 had syphilis. The remainder were either unspecified in the medical record or were attributed to congenital CMV infection.

A total of 131 individuals in the cohort underwent clinical genetic testing for GJB2 and GJB6 or SLC26A4 mutations; of those, 15 had mutations in GJB2/6 and 12 had mutations in SLC26A4. Testing for CMV on the dried blood spots from this group identified 11 subjects who had congenital CMV infection. No patients had both congenital CMV infection and GJB2/6 or SLC26A4 mutations.

Examination of CMV load in the blood spots gave a wide range of titers, from 46.9 to 297 131.4 copies of CMV per million cells. Most CMV-positive cases had severe to profound HL in their worse-hearing ear. The relationship between CMV load and hearing status as represented by a pure-tone average taken from the worse ear at the most recent audiogram is presented in the Figure. Bivariate analysis did not demonstrate a relationship between viral titers and the presence of maximal vs partial HL. Similarly, using a logistic regression model, no significant relationship was demonstrated between viral load at birth and presence of symptoms in the neonatal period, progression of HL, or pure-tone averages in cases with congenital CMV infection. There was no evidence of interaction with age or sex.

The present finding of a 1.4% prevalence of congenital CMV infection in children in Washington State falls within a range of previously presented percentages ranging from 0.18% in Italy² to 2.1% in India³ and is similar to the prevalence of 1.3% described in Alabama despite several different population characteristics.⁶ In children with HL, the prevalence of congenital CMV infection was almost 10%, which is somewhat lower than the prevalence described elsewhere.^2,22 The population-attributable risk of 8.9% was likewise toward the lower end of the estimates provided by other research groups.¹ Variable audiologic inclusion criteria and even more widely variable methods of diagnosis of congenital CMV infection across different studies may limit direct comparison of prevalence or risk figures. Finally, recent work by Boppana et al²³ demonstrated that detection of congenital CMV infection using PCR methods from dried blood spot testing had low sensitivity but high specificity. The techniques used in that study were similar but not identical to those used in this study. Nevertheless, it is highly likely that the prevalences reported in this study are underestimates given the low sensitivity of this type of approach.

With those caveats, the somewhat lower prevalence of congenital CMV infection in this study may be related, in part, to demographic or socioeconomic characteristics. Whereas children with CMV-related HL in the University of Alabama cohort were mostly black with young (<20 years old) single mothers,²⁴ most children who developed congenital CMV-associated HL in this study were born to white mothers with a relatively high household income. Although this may represent a selection bias due to greater willingness to participate in research in certain subsets of the population or more stable living arrangements allowing contact for recruitment, these findings may represent patterns of CMV infection in this population. This information may be applicable to populations in parts of the United States whose ethnic and racial profiles are similar to that of the Pacific Northwest. The observation of lower birth weights in babies with congenital CMV infection is consistent with other studies.^22,25

A large fraction of children with congenital CMV infection in this study had unilateral HL, falling approximately in the middle of a spectrum of percentages from several studies.^5,22,26 Many cases with bilateral asymmetric HL and severe to profound HL were also identified, similar to that in other populations.^22,27 Because children 4 years and older were recruited, it is also possible that there was enrichment for unilateral HL because of diagnosis of previously undetected unilateral HL by school hearing screenings.

The present findings are concordant with those of other studies demonstrating that progressive HL is more common in children with congenital CMV infection, although the incidence seen herein is lower than that in other studies.^5,12,14,28 In contrast to another study,¹⁴ fluctuating HL was not observed in this population. The ability to demonstrate fluctuation and progression in this population may have been limited by the availability of follow-up data.

Among individuals in this cohort who underwent genetic testing, there was no overlap between those with GJB2/6 or SLC26A4 mutations and congenital CMV infection. Other research groups have identified patients with congenital CMV who also had genetic mutations associated with HL,^4,29 but this has not been universal.³⁰ Larger numbers of subjects would likely further elucidate this issue.

Examination of a subset of the children in the University of Alabama cohort demonstrated that when stratified according to viral load, patients with the highest viral loads had a higher incidence of HL. When infants with and without symptoms of CMV infection were compared, the symptomatic infants had higher viral loads,^24,30 but no difference was observed in the viral loads in peripheral blood when children with overt symptoms of CMV infection with and without HL were compared, suggesting that viral load did not necessarily predict degree of HL.²⁴ In the present data, no obvious relationship was seen between viral load and hearing outcome or between viral load and the presence of neonatal symptoms.

The lack of a detectable relationship between viral load and hearing outcomes may be partially due to the large proportion of subjects with severe to profound HL, who were distributed across the range of viral loads. In part, this phenomenon may be secondary to clinical patterns at Seattle Children's Hospital, which, although it is one of the main referral centers for HL in Washington State, may be disproportionately selected for follow-up visits by families whose children have more severe HL. It is also possible that factors other than congenital viral load, such as timing of the peak of viral load or duration of viral shedding, may play an important role in determining hearing outcomes.

Another potential limitation is the possibility of deterioration in the accuracy of CMV load from blood spot cards that have been stored for many years.³¹ However, the results of a study by Atkinson et al³¹ suggest that if there is deterioration, it may level out and stabilize with time, and regression analysis of the data presented herein did not demonstrate a relationship between the age of the blood spot card and viral load. Nevertheless, if there is a differential degree of deterioration of measurable CMV across blood spot cards of varying ages, it is possible that a relationship between CMV viral load at birth and pure-tone average was present but not detectable in this study.

We did not study the relationship between congenital viral load and the likelihood of developing clinically relevant symptoms because this study, by definition, included children with symptomatic HL. However, in the future, if relatively nontoxic treatments become available for congenital CMV infection and if low-cost and high-quality assays allow efficient identification of congenital infection, widespread screening for congenital CMV infection could lead to the prevention of a considerable fraction of pediatric HL. Future investigations could also determine whether treatment with antiviral medications at the time of diagnosis of congenital CMV-associated HL could slow or halt the progression of HL.

In conclusion, in this retrospective study, a baseline prevalence of congenital CMV infection in children in Washington State was determined. The prevalence of congenital CMV infection in children with HL was 10-fold higher. Children with CMV-associated HL more commonly had unilateral and progressive HL, and congenital CMV infection was nearly as common as were GJB2/6 and SLC26A4 mutations. There was no clear relationship between viral load and hearing outcomes. Several groups have advocated consideration of newborn CMV screening,^4,32 particularly given the advent of antiviral therapy.³³ Determination of whether screening would be cost-effective will require further study, but testing for congenital CMV infection could be a meaningful addition to a standard battery of tests in the workup of HL.

Correspondence: Stephanie Misono, MD, MPH, Department of Otolaryngology/Head and Neck Surgery, University of Washington, 1959 NE Pacific St, Seattle, WA 98195 (smisono@u.washington.edu).

Submitted for Publication: April 2, 2010; final revision received June 29, 2010; accepted October 12, 2010.

Author Contributions: Drs Misono, Sie, Huang, Boeckh, Norton, and Yueh had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Misono, Sie, Boeckh, Norton, and Yueh. Acquisition of data: Misono, Huang, Norton, and Yueh. Analysis and interpretation of data: Misono, Sie, Weiss, Norton, and Yueh. Drafting of the manuscript: Misono and Norton. Critical revision of the manuscript for important intellectual content: Misono, Sie, Weiss, Huang, Boeckh, Norton, and Yueh. Statistical analysis: Misono, Weiss, and Yueh. Obtained funding: Misono, Norton, and Yueh. Administrative, technical, and material support: Sie, Huang, Norton, and Yueh. Study supervision: Sie, Weiss, Norton, and Yueh.

Financial Disclosure: Dr Boeckh has received research support from Roche, Vical Inc, Viropharma Inc, Theraclone Sciences, and Chimerix Inc and consultant payments from Vical Inc, Novartis, Boehringer Ingelheim, Genentech, Theraclone Sciences, Viropharma Inc, Chimerix Inc, and Astellas.

Funding/Support: This study was funded by the Gustavus and Louise Pfeiffer Research Foundation and by grant K24 HL93294 from the National Institutes of Health (Dr Boeckh).

Role of the Sponsor: The Gustavus and Louise Pfeiffer Research Foundation had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript.

Previous Presentation: This study was presented at the American Society for Pediatric Otolaryngology Annual Meeting; May 2, 2010; Las Vegas, Nevada.

Additional Contributions: We are deeply grateful to Marie Lutton for her coordination of this study. We also thank all the families who participated, members of the CMV research team, and the Seattle Children's Hospital clinic staff for their invaluable assistance. A collaboration with the WSDOH and its Director of Newborn Screening, Michael Glass, MS, made the study possible.