Background

Adenovirus (Ad) infections account for about 4–10 % of viral pneumonia and bronchiolitis in children [6], and an alarming number of fatal illnesses and long-term pulmonary complications have been reported worldwide. Some particular serotypes, especially 3, 7, and 21, have been reported to cause either epidemics of respiratory illnesses or fulminant events, including lethal pneumonia [3, 15, 18]. Infection by these serotypes can also result in long-term pulmonary complications, e.g., bronchiolitis obliterans, bronchiectasis, and the Swyer James syndrome [8, 25, 29, 31, 36]. Apart from specific serotypes, some host factors have also been found to be associated with development of severe Ad infection, such as young age (less than 7 years old), underlying chronic disease, post-transplant or immunocompromised state [13, 24, 30].

In the past decade, there were increasing numbers of reports on Ad serotype analysis, in part because of the more readily available polymerase chain reaction sequencing approach for typing Ad as compared to the traditional neutralization method [5, 11, 20, 21]. However, the exact association of serotype with clinical phenotype is still unclear. So far, there have been no reports on the Ad serotype pattern in Hong Kong, and their relationships with disease manifestations and complications are unknown. The few local studies that related to Ad infection only focused on the overall epidemiology of viral pneumonia or acute respiratory tract infections in hospitalized children [7, 32, 33].

Hence, we conducted a retrospective review to study the epidemiology, clinical features, and outcome of Ad respiratory tract infections in children admitted to our hospital. We focused on those pre-school children under the age of 6 years old as nearly 90 % of Ad infections occur in this age group [5, 6, 17]. We also aimed to explore for any serotype–clinical syndrome association and to identify the at-risk group for “lower” versus “upper” respiratory tract infections. The study was approved by the Institutional Review Board of our Hospital Authority. The need for informed consent was waived.

Patients/methods

Case definition and identification

The study was carried out in Queen Mary Hospital, a university affiliated hospital, and one of the two public hospitals in Hong Kong Island serving a population of more than 100,000 children under 18 years old. All preschool children under 6 years of age admitted with acute respiratory symptoms and with a virologic diagnosis of Ad from January 2001 to December 2004 were included in the study. These cases were identified retrospectively from laboratory surveillance database, and their clinical records were reviewed for demographic and clinical details. Based on our admission policy, nasopharyngeal aspirates were obtained routinely in all patients who admitted with respiratory symptoms for isolation purposes and tested for five common viruses—Ad, influenza A and B, parainfluenza virus, and respiratory syncytial virus (RSV)—by direct antigen detection by immunofluorescence (IF) and viral culture.

Children with acute respiratory illnesses were categorized into those with upper respiratory tract infection (URTI) and those with lower respiratory tract infection (LRTI). LRTI included pneumonia and acute bronchiolitis. The medical records were reviewed carefully and the clinical diagnosis of each case was classified cautiously: Acute bronchiolitis was defined as the presence of viral upper respiratory prodrome including rhinorrhea and cough with any signs of increased respiratory effort and wheezing in young children [9]. The diagnosis of pneumonia was based on the presence of cough with tachypnea [28] or positive auscultative findings (bronchial breath sound or crackles) in addition to abnormal radiological features (consolidation/interstitial infiltrates) that were consistent with pneumonia, as interpreted by the attending pediatrician or by the radiologist. Patients with co-infection of other known respiratory viruses or bacteria in the same course of hospital stay were excluded.

Adenovirus isolation and serotyping

Virus isolation was performed by inoculating HEp-2C cell monolayers in tissue culture tubes with 150 μl of the nasopharyngeal aspirate in viral transport medium and incubated at 37 °C for up to 7 days. The culture maintenance fluid, minimum essential medium supplemented with 2 % fetal bovine serum (GibcoBRL, Grand Island, NY, USA), l-glutamine (2 mM), and antimicrobials (final antimicrobial concentration per milliliter of medium: vancomycin, 50 μg; amikacin, 20 μg; nystatin, 20 units), was changed every 2 to 3 days or twice during the incubation period. Ad was identified initially by characteristic cytopathic effect and direct immunofluorescence staining (Adenovirus DFA Reagent, EMD Millipore Corporation, USA) of the infected cells. Definitive typing of the viral isolate was performed by micro-neutralization assay with HEp-2C cells for serotyping as described previously [16]. The isolate was passed to a new HEp2-C tube until ≥75 % cytopathic effect was evident and diluted the culture at 1:10 with 2 % minimum essential medium after freezing and thawing of the culture once before carrying out the neutralization test. Briefly, 0.05 ml of each Ad typing serum type 1 to 7 (Denka Seiken, Japan) at 10–20 antibody unit per 0.05 ml was added to 0.05 ml of each isolate at a viral concentration of 100 TCID50 (50 % tissue culture infective dose) per 0.05 ml and incubated at 37 °C in 2.5 % CO2 for 1 h, followed by incubation with 0.1 ml HEp-2C cells (150,000 cells per ml) for up to 7 days. The serotypes were determined by inhibition of cytopathic effect by corresponding type-specific antisera. End results were read when the virus control showed 100 % cytopathic effect. Back-titration of the input virus dose and cells without virus inoculation as negative control was run in parallel. In case where the virus could not be neutralized by the antisera used, the test was repeated at 1:20, 1:50, or higher viral dilution if necessary. Isolates that did not type after repeated tests were considered as untypeable.

Statistical analysis

Descriptive statistics were used to describe the epidemiological data of the disease and clinical characteristics of patients. Chi-square test was used to test if there was any significant relationship between individual serotype and each disease category (pneumonia versus acute bronchiolitis versus URTI). Demographic characteristics were compared between the groups of children with LRTI and URTI using Pearson's chi-square test (with Yates’ correction) for categorical variables and unpaired t-test for continuous variables for any risk factors that might predispose a child to lower respiratory tract involvement. Significant risk factors identified in the univariate analysis were then adjusted using multi-variate logistic regression to determine their independent effects. All analyses were done by using SAS software (version 9.12). A p-value of <0.05 was considered as statistically significant.

Results

Demographic characteristics and clinical presentation

A total of 287 children fulfilled the study criteria during the specified period, and their demographic characteristics were shown (Table 1). There was a slight male predominance. The mean age was 2.82 ± 1.61 years old. The majority (85.4 %) of them were previously healthy. Thirty children were admitted from institutions (i.e., residential facilities for children with social problems). Twenty-nine cases were excluded from the study due to co-infection with other pathogens: nine cases with influenza virus A or B, two cases with RSV, 14 cases with parainfluenza virus 1, 2, or 3, one case with Coxsackie virus A, one case with human metapneumovirus, one case with rotavirus, and one case with both Stapylococcus aureus and Moraxella catarrhalis.

Table 1 Demographic characteristics and clinical features among hospitalized children with adenovirus respiratory infections
Full size table

Common presenting symptoms included fever (97.9 %), cough (74.9 %), and rhinitis (73.9 %). A total of 107 (37.3 %) patients had one or more associated extra-pulmonary manifestations: gastrointestinal symptoms (n = 39), febrile seizure (n = 36), conjunctivitis (n = 26), and skin rash (n = 14). URTI was the most common presentation (245 cases, 85.4 %). Among them, four cases presented as croup-like illness and 14 cases were complicated with otitis media. LRTI was diagnosed in 42 patients (14.6 %) with 28 cases of pneumonia and 14 cases of acute bronchiolitis.

Laboratory and radiological findings

Blood tests with differential white cell counts (WCC) were available in 255 patients. Among them, 52.2 % had neutrophilia as defined by absolute neutrophil count (ANC) of ≥8.5 × 109/L according to the laboratory reference. The mean total WCC, ANC, and lymphocyte count did not differ significantly between the URTI and LRTI groups (Table 1). C-reactive protein level was not routinely performed in our setting. Among those with LRTI, interstitial pulmonary infiltrates were the most common abnormal radiological findings (73.3 %), while the remaining ones showed lobar-type consolidation.

Clinical course and outcome

Both URTI and LRTI patients suffered from high fever (mean peak temperature of 39.6 ± 0.6 and 39.5 ± 0.7 °C, respectively), and the overall mean duration of fever prior to attendance of the Accident and Emergency Department was 3.4 ± 2.5 days. Those with LRTI had a slightly longer hospital stay than those with URTI (4.6 ± 2.9 versus 2.9 ± 1.9 days). The whole febrile illness usually lasted 5 to 6 days. While awaiting for the rapid IF result, 30.3 % of the patients were empirically treated with antibiotics in view of the high-grade fever together with the finding of neutrophilia. Indeed 43.9 % of the admitted children had received antibiotics from their primary care physicians before hospital admissions.

Four out of 42 patients (9.5 %) with LRTI had hypoxia, defined as oxygen saturation below 92 % on admission, and required oxygen supplement for 1 to 5 days. Two patients needed intensive care support due to associated febrile status epilepticus which were not directly related to the severity of respiratory illness. One had acute bronchiolitis and the other had URTI. There was no mortality.

Among those 42 patients with LRTI, about 60 % of them were offered follow-up as they still had considerable symptoms or significant chest X-ray abnormalities at the time of discharge. Most of them had symptoms or radiological improvement within a few weeks’ time and did not require long-term follow-up. Only one patient developed chronic pulmonary sequelae. He was a 3-year-old child with good past health who suffered from right lower lobe pneumonia caused by Ad serotype 7. Serial follow-up chest X-rays showed persistent right lower lobe collapse consolidation even after several months. Initial computer tomography scan of the thorax revealed lobar collapse, but repeated scan after 1 year showed bronchiectatic changes in the corresponding lobe. Workup for other common causes of bronchiectasis was negative.

Risk factors for LRTI

Among the risk factors selected for analysis, age between 1 and 2 years (p = 0.0153), history of prematurity (p = 0.0018), history of ventilator assistance at birth (p = 0.0215), and institutionalization (p = value 0.0009) were found to be associated with LRTI in univariate analysis (Table 2). After multiple logistic regression, only age between 1 and 2 years remained as the statistically significant risk factor for LRTI (adjusted p = 0.0165, odds ratio 2.48, 95 % confidence interval 1.18 to 5.22).

Table 2 Identification of risk factor(s) for lower versus upper respiratory tract infection by adenovirus
Full size table

Serotype pattern analysis

In our study, rapid detection of Ad by IF had a sensitivity of 73.8 % (212/287) in comparison to cell culture. Serotyping by microneutralization assay was achieved for 93.7 % of the culture samples. Among the seven serotypes (type 1–7) that were isolated during the 4-year period, serotypes 2, 3, and 7 accounted for most of the respiratory tract infections. The prevalent serotype and the peak season varied in each year (Table 3; Fig. 1). Serotype 7 was the predominant type (40.4 %) in 2001. However, it was no longer detected in the years 2003 and 2004. Instead serotype 3 was found in the majority in these 2 years, accounting for 43.5 and 64.1 % of the admissions, respectively.

Table 3 Serotype distribution among 287 children hospitalized for adenovirus infection between 2001 and 2004
Full size table
Fig. 1
figure 1

Monthly distribution of hospitalized cases of adenovirus respiratory infection

Full size image

The serotype distribution among different respiratory illnesses is shown in Table 4. Being more common, serotypes 2, 3, and 7 proportionally accounted for most of the pneumonia cases. Serotype 3 infection accounted for over one third of the pneumonia cases. Interestingly, the majority of acute bronchiolitis (71.4 %) were associated with serotype 2 infection, and this association was statistically significant (p < 0.0001). Infection with serotype 5 was only associated with URTI in our series.

Table 4 Serotype distribution in different types of adenovirus respiratory illnesses
Full size table

On stratifying patients according to their age groups and types of respiratory illness (Table 5), we found that serotype 2 accounted for the majority of acute bronchiolitis and half of the pneumonia cases in those aged below 2 years, while it did not cause any LRTI in those aged above 3 years old. The prevalence of serotype 2 in causing URTI also significantly decreased after the first 3 years of age. Serotype 3, despite being the most prevalent serotype, did not play any significant role for LRTI in those aged below 3 years. It was, however, the predominant etiological agent of pneumonia in those aged between 3–4 and 4–5 years old, accounting for 75 and 57.1 % of pneumonia cases, respectively.

Table 5 Serotype distribution stratified according to age groups and types of respiratory illness
Full size table

Serotype 7, on the contrary, did not seem to have a particular age group predilection. Comparison between serotypes 3 and 7 revealed no significant difference between their clinical and laboratory characteristics, but those patients infected with serotype 7 required a slightly longer mean hospital stay (3.7 ± 3.3 versus 2.8 ± 1.2 days, p = 0.0207) and were more likely to have a history of prematurity (p = 0.0407) or were living in institutions (p = 0.0025) (Table 6). The latter was likely contributed by the fact that an outbreak of serotype 7 occurred between April and May in 2001 which resulted in nine consecutive admissions from an institution.

Table 6 Comparison of demographic and clinical features between adenovirus serotypes 3 and 7 respiratory infections
Full size table

Discussion

Adenovirus has been known as a common cause of childhood respiratory illness since it was first described in 1950s. By now, more than 50 serotypes have been recognized, but the exact association of serotype with clinical phenotype is still unclear. It is likely that different serotypes display different tissue tropisms and clinical manifestations of infection. As the predominant serotypes differ among countries and change over time, the epidemiological and clinical pattern of Ad infection will vary accordingly. However, there is paucity of local data on the epidemiological characteristics, serotype pattern, and clinical outcome of Ad infection in Hong Kong children.

We showed that serotypes 2 and 3 were the most prevalent serotypes among hospitalized children with Ad respiratory infections during this 4-year study period. There was a small epidemic due to serotype 7 in 2001. The number of admissions due to Ad was markedly reduced in 2003, during the SARS outbreak. Reduced infection of other respiratory viruses in Hong Kong during that time had also been documented and might be partly explained by increased social distancing and improved community hygiene during that period [22]. Unlike influenza virus and RSV, no consistent seasonal pattern of Ad respiratory infection was observed, which was similarly seen in some other areas [1, 19].

The clinical picture of Ad respiratory infection was characterized by coryzal symptoms with prolonged and high fever. Our children suffered from a mean peak temperature of 39.6 ± 0.6 °C, and the mean duration of this febrile illness lasted 5.2 ± 2.7 days. The presence of extra-pulmonary manifestations (gastrointestinal symptoms, febrile seizure, conjunctivitis, and skin rash) might potentially serve as a hint for Ad infection as these symptoms were seen in over one third of cases.

Some studies have reported leukocytosis or elevated C-reactive protein with Ad infections [5, 8]. We have also found neutrophilia in over half of our cases as opposed to lymphopenia or lymphocytosis seen in most other viral infections. Because of the high fever and neutrophilic blood picture, the disease presentation often mimicked bacterial disease and led to unnecessary antibiotic prescription. The fact that Ad infections mimicked bacterial infection and parental anxiety due to persistent high fever might lead to a higher likelihood of hospital admission and hence explained why a considerable number of children were admitted to the hospital for URTI in our locality.

There was a relatively even distribution in age among children who suffered from URTI, but over one third of the children suffered from LRTI aged between 1 and 2 years old. We have confirmed from the multi-variate analysis that children in this age group were more predisposed for “lower” versus “upper” respiratory tract infection by Ad. Acute bronchiolitis is generally more common in those aged below 2 years old but possible explanation that these children (1 to 2 years old), rather than the even younger age group, were at higher risk for developing LRTI might include waning of protective antibodies at this age.

It is well reported that a subgroup of patients with Ad infection can progress to severe pneumonia with long-term complications. Serotypes 3, 7, and 21 have been previously identified as the culprits [3, 15, 18, 29]. Apart from susceptible host risk factors, severe infection can also occur in previously healthy person [14, 26]. Interestingly, there was a marked variation in the morbidity and mortality rates of adenovirus LRTI in different studied populations. Ad respiratory infection had a profound impact in some countries. Mortality up to 16 % and permanent pulmonary sequelae ranging from 14 to 60 % were reported in children in some South American countries following Ad pneumonia [3, 4, 24, 34]. Lower yet still significant mortality and morbidity rates were reported in some neighboring Asian countries. In Korea, Hong et al. reported that Ad LRTI-related mortality in a children hospital was 12 % overall and 19 % among patients who were infected with serotype 7 [17]. Residual sequelae were identified in 50 % of the patients who were infected with serotype 3 and in 25 % of those infected with serotype 7. In a Taiwanese study, two out of 48 patients with Ad pneumonia died and five had permanent lung damage, while there was no mortality nor long-term sequelae found in the non-Ad group [12]. While traditionally serotypes 3 and 7 were well known to cause more severe respiratory infections, there were occasional reports on other serotypes which could be potentially associated with severe infections. For example, in the review by Lee et al. [20], serotype 8 was one risk factor for severe respiratory infection in Korea, while serotype 4 was identified as an important pathogen responsible for the fatal childhood pneumonia in South China [27]. The reason why a certain strain of Ad in some populations was more virulent and damaging to the growing lung was not understood clearly. It was speculated that host variation in immune response and the local virulence of the specific serotype might both play a part [10, 23, 35].

However, the disease severity of Ad respiratory infections in our locality appeared to be less alarming than many of our neighbouring countries. In our study, 14.6 % of those admitted were classified to have lower respiratory tract complications, either acute bronchiolitis or pneumonia, based on rather stringent diagnostic criteria, yet only four patients with LRTI suffered from mild hypoxemia, without severe acute respiratory compromise or mortality. As our study was conducted in a hospital providing both secondary and tertiary care in the region, it should have covered the whole spectrum of patients with severe illness.

Still one of our patients developed bronchiectasis after serotype 7 infection, a subtype notorious in causing long-term pulmonary sequelae. This serotype was indeed well known to be associated with severe pneumonia, but it only ranked third in causing pneumonia in our hospitalized cases, probably because it was only present in two cases out of the 4-year study period. It was no longer detected in 2003–2004 among our hospitalized children after the apparent small epidemic in 2001. We could not ascertain whether this might reflect an increasing trend of herd immunity in the community, resulting in lesser case severity. Lee et al. also observed a decreasing size of subsequent epidemics after a large outbreak of serotype 7 [20]. This could partly account for the milder disease phenotype associated with Ad infection during the study period in our locality.

As shown, the most prominent serotype–clinical syndrome correlation appeared to be serotype 2 with acute bronchiolitis. While Ad has been known to be occasionally associated with acute bronchiolitis, the role of serotype 2 has not been previously reported. One study from Cuba reported that serotype 5 accounted for 71 % of the 14 cases of acute bronchiolitis identified, but serotype 2 was not tested for in that study [2]. Their sample size was smaller (49 cases) and with a lower serotyping rate (63 %). The study was not confined to the pediatric age group and there was no clear case definition of acute bronchiolitis versus pneumonia.

It would be difficult to clearly delineate the complex relationship between serotype virulence, host factor, and resultant disease phenotype. For example, serotype 2 seemed to infect the youngest children predominantly (mean age 1.94 ± 1.29 years old), but “young age” could also be the risk factor for developing acute bronchiolitis. On the other hand, serotype 3 displayed predilection for the slightly older age groups (mean age 3.43 ± 1.42 years old). It was the leading cause for pneumonia among those aged between 3 and 5 years old, but it did not play any significant role for LRTI in those younger children. Hence, host factor might also be a potential confounding factor for the serotype–respiratory illness association. Further epidemiological studies would be needed to better support our observation.

The potential limitations of our study must be addressed. Firstly, there was no assurance of standardized documentation relating to the history, physical findings, and radiological interpretation because this was a retrospective study. Secondly, our study was confined to a single centre and the result might not be extrapolated beyond our local community. Thirdly, the neutralization test only routinely tested for serotypes 1 to 7 as these serotypes were considered to be most frequently associated with respiratory disease. It meant that other serotypes (constituted 6.3 % of the total cases) were not serotyped, though the majority of them only caused URTI. The neutralization test also might not be able to type virus isolates containing coinfections with multiple Ad serotypes. Such coinfections, even by serotypes 1–7, might appear non-neutralizable by the serotype-specific sera and would be falsely categorized as non-1–7 serotype. Fourthly, although we had tried to exclude the possible impact of co-infections by other pathogens in the analysis of the clinical disease spectrum, viruses such as human metapneumovirus, coronavirus, and rhinovirus and atypical bacteria such as mycoplasma and chlamydia were not tested for. Lastly, it was postulated that Ad might persist and be shed for prolonged periods and thus positive cell cultures for Ad might not represent an etiological link with disease. However, direct immunofluorescence assay from nasopharyngeal aspirate was known to have lower sensitivity than the cell culture method. Currently, there is no good diagnostic test to ascertain acute Ad infection as seroconversion test is not available in most clinical settings.

Conclusion

The current study was the first comprehensive review of Ad respiratory infections among hospitalized children in Hong Kong with serotype data presented. Apart from revealing the local serotype pattern and epidemiology, we have shown that Ad generally did not cause severe acute respiratory disease in our young children and long-term pulmonary sequelae was comparatively infrequent, but as the clinical manifestations of Ad infections mimicked bacterial infections, it often led to unnecessary hospital admissions and antibiotic use. Children aged between 1 and 2 years were most at risk for LRTI. Serotypes 2 and 3 were the most prevalent serotypes among our hospitalized children, and they seemed to have a predilection for different age groups and with different resultant respiratory illness association. A longer study period over time is needed to monitor the serotype pattern and disease complication for better understanding and prognostication of Ad infections in the region and to better delineate the potential complex relationship between serotype virulence, host factor, and resultant disease phenotype.