key: cord-354974-bh2expef
authors: Peterson, Ingrid; Bar-Zeev, Naor; Kennedy, Neil; Ho, Antonia; Newberry, Laura; SanJoaquin, Miguel A.; Menyere, Mavis; Alaerts, Maaike; Mapurisa, Gugulethu; Chilombe, Moses; Mambule, Ivan; Lalloo, David G.; Anderson, Suzanne T.; Katangwe, Thembi; Cunliffe, Nigel; Nagelkerke, Nico; McMorrow, Meredith; Widdowson, Marc-Allain; French, Neil; Everett, Dean; Heyderman, Robert S.
title: Respiratory Virus–Associated Severe Acute Respiratory Illness and Viral Clustering in Malawian Children in a Setting With a High Prevalence of HIV Infection, Malaria, and Malnutrition
date: 2016-09-13
journal: Journal of Infectious Diseases
DOI: 10.1093/infdis/jiw426
sha: 
doc_id: 354974
cord_uid: bh2expef

BACKGROUND: We used data from 4 years of pediatric severe acute respiratory illness (SARI) sentinel surveillance in Blantyre, Malawi, to identify factors associated with clinical severity and coviral clustering. METHODS: From January 2011 to December 2014, 2363 children aged 3 months to 14 years presenting to the hospital with SARI were enrolled. Nasopharyngeal aspirates were tested for influenza virus and other respiratory viruses. We assessed risk factors for clinical severity and conducted clustering analysis to identify viral clusters in children with viral codetection. RESULTS: Hospital-attended influenza virus–positive SARI incidence was 2.0 cases per 10 000 children annually; it was highest among children aged <1 year (6.3 cases per 10 000), and human immunodeficiency virus (HIV)–infected children aged 5–9 years (6.0 cases per 10 000). A total of 605 SARI cases (26.8%) had warning signs, which were positively associated with HIV infection (adjusted risk ratio [aRR], 2.4; 95% confidence interval [CI], 1.4–3.9), respiratory syncytial virus infection (aRR, 1.9; 95% CI, 1.3–3.0) and rainy season (aRR, 2.4; 95% CI, 1.6–3.8). We identified 6 coviral clusters; 1 cluster was associated with SARI with warning signs. CONCLUSIONS: Influenza vaccination may benefit young children and HIV-infected children in this setting. Viral clustering may be associated with SARI severity; its assessment should be included in routine SARI surveillance.

It is estimated that, worldwide, the case-fatality rate of severe pneumonia in children aged <5 years is 8.9%, which, in 2011, amounted to 1.26 million deaths [1] . Much of this burden falls on sub-Saharan Africa, where severe acute respiratory infection (SARI), including pneumonia, is a leading cause of childhood hospital attendance and death [2] . Although laboratory diagnostic facilities are rarely available in such settings, sentinel surveillance using multiplex molecular diagnostic assays has recently provided considerable insight into the true burden of disease and the complexity of SARI etiology. Respiratory syncytial virus (RSV), parainfluenza viruses, rhinoviruses, influenza viruses, and adenovirus have been commonly detected in SARI surveillance across the African continent [3] [4] [5] [6] [7] [8] . While there are a few viruses for which detection in respiratory disease cases is likely causal (eg, influenza virus and RSV) [9, 10] , for other commonly identified viruses causality has been difficult to determine. Use of multiplex assays has led to an increasing realization that children with SARI commonly carry multiple viral pathogens that may potentially contribute to disease.

In the context of a low-income population with multiple drivers of immune compromise (eg, human immunodeficiency virus [HIV] infection, malnutrition, and malaria) [11] , we conducted active surveillance at a large urban teaching hospital in Malawi to estimate the incidence of childhood SARI and explore the association of SARI clinical severity with HIV infection and clustering of respiratory viral coinfection. While previous studies have focused on children aged <5 years, we included children aged 3 months to 14 years in our analysis, to better capture the total burden and identify age groups particularly at risk. it offers care free at the point of delivery. Overall, 13% of children aged <5 years in Malawi are moderately to severely underweight, and 4% are experiencing wasting; 80.9% of children aged 12-23 months have received all Expanded Program on Immunization vaccinations [12] . There is no national routine influenza vaccination in Malawi. In 2010, a monovalent vaccine campaign targeting 2009 pandemic influenza A(H1N1) virus (A[H1N1]pdm09) achieved 74% coverage in pregnant women and 7% of the overall population [13] . An estimated 2.5% of children aged <15 years are HIV infected [14] ; the HIV prevalence in children aged <5 years on QECH nonsurgical pediatric wards is estimated at 6%. Blantyre has 2 distinct weather seasons, a rainy season (January-April) and a cool dry season (May-August). Overall, 25.2% of Paediatric Accident and Emergency Unit (PAEU) patients have a malaria parasitepositive blood slide; malaria presentations to the PAEU peak from December to May.

Patients aged 3 months to 14 years presenting during surveillance hours (weekdays, from 8:00 AM to 1:00 PM) from January 2011 through December 2014 were screened. Consecutive patients fulfilling the SARI case definition were recruited (maximum, 5 per day). Demographic and clinical data were captured through an electronic data collection system [15] . Nasopharyngeal aspirates (NPAs) were obtained and tested for influenza viruses; from 2011 to 2013, NPAs were also tested by multiplex assay for respiratory pathogens. Thick blood films for detection of malaria parasites were performed for all children.

SARI was defined as (1) an acute illness with symptom onset <7 days and (2) a reported or recorded fever of ≥38°C (or hypothermia in children <6 months). Additional criteria for SARI varied by age. In children aged <6 months, additional criteria were (3) cough or apnea or (4) any respiratory symptom requiring hospitalization. In children aged 6-59 months, an additional criterion was (3) clinician-diagnosed lower respiratory infection. In children aged 6-14 years, additional criteria were (3) cough or sore throat and (4) shortness of breath or difficulty breathing. SARI with warning signs was considered clinically more severe and defined as the occurrence of one of the following: admission to the hospital, chest recession, or blood oxygen saturation of ≤90%. In this resource-limited setting, some patients with severe illness requiring admission were sent home. Thus, hospital attendance (not admission) was required for study enrollment.

NPAs were stored at −80°C in Universal Transport Medium (Copan, Brescia, Italy) [16] and tested in batches for influenza viruses by real-time reverse transcription-polymerase chain reaction (RT-PCR). Total nucleic acids were extracted from 300-µL aliquots of each specimen with the Qiagen BioRobot Universal System, using the QIAamp One-For-All nucleic acid kit (Qiagen, Manchester, United Kingdom). The quantity of nucleic acid used per reaction was 5 µL for the Centers for Disease Control and Prevention (CDC) Human Influenza real-time RT-PCR diagnostic panel (CDC Influenza Division), which detects influenza A and B viruses and influenza A subtypes H1, H3, 2009H1, and H5N1, and 10 µL for the FTD respiratory pathogens 33 kit (Fast-track Diagnostics, Luxembourg). Details on sample processing with by FTD real-time RT-PCR are provided in Appendix 1. HIV serostatus was assessed by the rapid test (Alere Determine HIV-1/2 and Trinity Biotech Uni-Gold HIV) according to World Health Organization guidelines [17] . PCR for detection of HIV RNA was performed in children aged 3-11 months who had a positive HIV rapid test. HIV infection was defined on the basis of positive results of an HIV rapid test (in the absence of an HIV-negative PCR); data were not collected on HIV exposure.

Ethics approval for this study was obtained from the Liverpool School of Tropical Medicine Research Ethics Committee (approval RETH000790), the University of Malawi College of Medicine Research Ethics Committee (COMREC; approval 958), and the CDC through reliance on the COMREC. Informed consent was obtained from guardians of all study participants.

Numerators for minimum SARI incidence estimates were generated by summing the number of cases resident in Blantyre within strata of age category and HIV status. Numerators were adjusted by multiplying by the reciprocal of the daily proportion of recruited cases among all SARI cases attending the PAEU. Denominators for HIV and age strata were derived by applying age-specific HIV prevalence estimates to census figures for Blantyre District's population aged 0-14 years [18] . The former were obtained by apportioning the total HIV prevalence among Malawian children aged <15 years [14] according to the age distribution of pediatric HIV infections in Mozambique, which borders Malawi and has a similarly severe HIV epidemic [19, 20] . Estimates of age-specific HIV prevalence were unavailable for Malawi for the study period. The incidence was obtained by dividing numerators by denominators and multiplying by 10 000; HIV-associated incidence rate ratios (IRRs) were calculated by dividing the incidence in HIV-infected strata by the incidence in HIV-uninfected strata; 95% confidence intervals (CIs) of incidence and HIV-associated IRRs were generated with 1000 bootstrap samples.

Data analysis was performed using SAS 9.3 (SAS Institute, Cary, North Carolina). Temporal trends in weekly sample counts for SARI cases were assessed by plotting 5-week moving averages of sample counts by recruitment week. We developed 2 logistic regression models with a binary outcome factor for the child's clinical status. The first outcome represented SARI with warning signs (ie, clinical markers of very severe illness) versus SARI without warning signs. The second outcome represented influenza virus-positive SARI versus influenza virus-negative SARI. Autoregressive correlation of residuals was removed by introducing a patient-level kernel weighted moving average of the prior probability of case status. Parsimonious models were developed by stepwise logistic regression, retaining age, sex a priori, and explanatory factors with a 2-sided P value of <.05. Adjusted relative risk ratios for factors associated with the outcomes were derived from these models.

Detection of multiple viruses in SARI is common, with many possible combinations of viral carriage. Conventional statistical techniques (eg, regression models, covariance matrices, and temporal plots) have limited capacity to quantify, characterize or identify factors associated with viral carriage groupings. To assess multiple virus carriage clusters in our setting, we performed nearest-neighbor discrete hierarchical cluster analysis in patients with viral codetection, using the Gower distance [21] . Distance was based on similarity of viral pathogens detected in the nasopharynx of patients with SARI; each patient was a member of only 1 cluster. We defined clusters as those that increased the R 2 value by ≥0.05 (using the Ward method); to improve precision, 10% of observations with the lowest densities were discarded. Using univariate logistic regression, we identified factors associated with cluster membership.

From Table 2) .

Plots of weekly influenza virus-positive SARI cases suggest both unimodal and bimodal (2 peaks per year) seasonality. Weekly influenza virus-positive SARI cases increased during the rainy Figure 1 ).

Incidence Estimates for SARI and Respiratory Virus-Associated SARI SARI incidence was 17.5 cases per 10 000 children annually, with the highest incidence in children aged 3-11 months Table 5 ), as well as an increased incidence of SARI, SARI with warning signs, and influenza virus-positive SARI (Table 4) . HIV-associated IRRs rose with increasing age. The HIV-associated IRRs for SARI with warning signs was 2.6 in children aged 3-11 months as compared to 37.7 in children aged 10-14 years. In children aged >5 years, the incidence of hospital-attended influenza virus-positive SARI was at least 8-fold higher in HIV-infected children as compared to HIV-uninfected children. There was no difference in the incidence of RSV-positive SARI between HIV-infected and HIVuninfected children.

In multivariable analysis controlling for etiology, patients with SARI recruited during the rainy season (January-April) were more than twice as likely to have warning signs, compared with patients enrolled during September-December (aRR, 2.4; 95% CI, 1.6-3.8; Table 5 ). Peaks in RSV and influenza virus activity corresponded to peaks in the occurrence of SARI with warning signs (Figure 1 ). Detection of RSV in cases of SARI warning signs was much higher during the rainy season (39.8%) as compared to other times of year (5.9%).

The aRR for a positive results of an influenza virus test in patients with SARI increased with older age and rainy season of recruitment (Table 3 ). After adjustment for age, sex, and HIV status, rainy season recruitment was significantly associated with SARI with warning signs in influenza virus-positive patients with SARI (aRR, 3.42; 95% CI, 1.37-8.53; analysis not shown). In adjusted analysis, A(H1N1)pdm09 was associated with double the risk of SARI with warning signs, compared with other influenza virus subtypes (aRR, 2.10; 95% CI, .98-4.53; analysis not shown).

Detection of ≥2 viral pathogens by multiplex PCR occurred in 362 of 1835 SARI cases (19.7%). Viral codetection was highest in SARI cases positive for coronavirus 229 (70.6%) and enterovirus (79.7%). Viral codetection was least common in SARI cases testing positive for A(H1N1)pdm09 (27.3%), influenza A(H3N2) virus (29.0%), and RSV (29.5%) ( Table 2) . Viral codetection per se was not associated with warning signs in SARI (Table 5) . We used discrete hierarchical cluster analysis based on similarity of viral pathogens detected by multiplex PCR assay in SARI cases to explore whether particular groupings of viruses were associated with warning signs, host factors, or seasonal factors. We identified 6 clusters, which accounted for 48.3% of the total variation in viral pathogen test results in children with viral codetection. Cluster size ranged from 23 to 96 members; smaller clusters had fewer viral pathogens and lower within-cluster heterogeneity. Clusters were distinguishable by the type of viral pathogens detected. For example, 80% of influenza A(H3N2) viruses detected were found in cluster A; >65% of bocaviruses detected were found in cluster E (Appendix 3).

Cluster membership was significantly associated with clinical and temporal factors (Figure 2 ). Among children with viral codetection, membership in cluster D (characterized by Abbreviations: CI, confidence interval; IRR, HIV-associated incidence rate ratio; RSV, respiratory syncytial virus. a Inestimable. 

Hospital-attended SARI was common in this urban sub-Saharan African setting, particularly in infants aged 3-11 months, in whom the incidence was 91.7 cases per 10 000 children annually. Similar to studies from other settings, influenza viruses and RSV were important SARI-associated pathogens [5-8, 22, 23] , with prevalence rates of 11% and 12%, respectively. As elsewhere, HIV infection increased the risk of SARI and the presence of warning signs in SARI cases [24] [25] [26] . Among older children, HIV infection greatly increased the risk of influenza virus-positive SARI, consistent with data from South Africa [25] . Viral coinfection occurred in almost 20% of SARI cases, highlighting its potential impact in the development or clinical worsening of SARI [27] . Although viral codetection per se was not associated with clinical severity or season, we found 1 viral cluster, characterized by a high proportion of RSV and A(H1N1)pdm09 infection, which was significantly associated with clinical warning signs and rainy season recruitment. Cluster members coinfected with RSV and A(H1N1)pdm09 had a higher rate of warning signs, but the number of coinfected individuals (within the cluster and the entire sample) was too small to formally test for interaction. It is unclear therefore whether clinical severity in this cluster resulted from biological interaction of pathogens, additive risks from each pathogen, or other underlying factors. Clusters clearly mapped to peaks and troughs in individual pathogen activity. We suggest that this viral clustering, which was associated temporal dynamics of pathogen activity, may have arisen from complex virus-virus and host-virus pathogen interactions.

Clinical severity in SARI demonstrated seasonal peaks, coinciding with rainy season peaks in RSV activity. RSV was detected in 40% of SARI cases with warning signs recruited during the rainy season, compared with 6% recruited at other times of the year. Thus, RSV may drive rainy season increases in clinical severity in pediatric SARI in our setting, consistent with studies elsewhere in sub-Saharan Africa [28, 29] . Nevertheless, the rainy season remained independently associated with an increased risk of warning signs in SARI in Figure 2 . Dendrogram of coviral clusters. Six coviral clusters (A-F) were identified in 362 pediatric SARI cases, in whom >2 viral pathogens were detected in the nasopharynx. Each severe acute respiratory infection (SARI) case is a member of only one cluster; clusters membership is based on similarity of viral pathogens detected. As shown here, characteristics such as SARI severity, number of viruses detected per child, and particular season and year of recruitment are more common in some clusters than others. Green bars denote SARI without warning signs, red bars denote SARI with warning signs, bluish-gray bars denote detection of <3 viruses detected, orange bars denote detection of ≥3 viruses, lavender bars denote recruitment in the rainy season, yellow bars denote recruited outside of the rainy season, gray bars denote recruitment in 2011, blue bars denote recruitment in 2012, pink bars denote recruitment in 2013, and light green bars denote recruitment in 2014. multivariable analysis controlling for RSV, HIV, and other viral pathogens. Therefore, the observed rainy season excess of clinical severity in SARI is in part attributable to unmeasured factors. We speculate that these factors include other intervening illnesses and seasonal malnutrition (in Malawi, the rainy season coincides with the so-called lean season after harvest [30] ). However, we cannot exclude seasonal differences in healthcare utilization.

We acknowledge that our study has limitations. We did not recruit children aged <3 months, in whom the frequency of SARI-related deaths is known to be elevated [31] . We were unable to determine the role of bacterial pathogens in SARI, as we lacked laboratory data and systematic radiological data to identify probable infection in the context of a very high prevalence of bacterial carriage. Our estimates of SARI incidence by HIV strata were based on Mozambican pediatric HIV prevalence rates because we lacked data from Malawi. Nevertheless, Malawi and Mozambique have similar rates of antenatal HIV prevalence [12, 32, 33] and similarly high rates of HIV-infected pregnant women accessing antiretroviral treatment [34] . We did not assess the impact of HIV exposure on SARI risk in HIV-uninfected children. HIV exposure was associated with higher SARI incidence and greater SARI severity in HIV-uninfected South African children [35] .

In conclusion, SARI is common in this setting of high HIV prevalence, where influenza viruses, rhinoviruses, and RSV were the most prevalent viruses detected. HIV greatly increased the risk of influenza virus-associated SARI in children, and therefore yearly influenza vaccination should be considered in routine pediatric HIV clinical care. Influenza vaccination in HIV-infected children is safe, but it has low efficacy (<20%) and may only be immunogenic in older children and adolescents with virological suppression [36] [37] [38] . Viral coinfection was common, with 1 coviral cluster associated with clinical severity in SARI cases. In this context, there is considerable potential for the use of multiplex respiratory virus assays in tandem with cluster analysis to reveal multiplepathogen-associated outbreaks and disease burden. This approach may expose the potential for synergistic effects of vaccine strategies that disrupt viral clusters. Vaccine probe studies to assess the impact of viral coinfection on clinical severity could clarify complex pathogen and host interrelationships and reveal the true burden of disease.

Global burden of childhood pneumonia and diarrhoea

Global and regional burden of hospital admissions for severe acute lower respiratory infections in young children in 2010: a systematic analysis

Identification of viral and bacterial pathogens from hospitalized children with severe acute respiratory illness in Lusaka, Zambia

Incidence of respiratory virus-associated pneumonia in urban poor young children of

Viral and bacterial causes of severe acute respiratory illness among children aged less than 5 years in a high malaria prevalence area of western Kenya

Viral and bacterial etiology of severe acute respiratory illness among children < 5 years of age without influenza in Niger

Influenza sentinel surveillance among patients with influenza-like-illness and severe acute respiratory illness within the framework of the national reference laboratory

Severe acute respiratory infection in children in a densely populated urban slum in Kenya

Respiratory viral detection in children and adults: comparing asymptomatic controls and patients with community-acquired pneumonia

Viral etiology of severe pneumonia among Kenyan infants and children

Do multiple concurrent infections in African children cause irreversible immunological damage?

Malawi Demographic and Health Survey

Pandemic influenza A virus subtype H1N1 vaccination in Africa-successes and challenges

HIV and Syphilis Sero-Survey and National HIV Prevalence and AIDS Estimates Report for

Surveillance Programme of INpatients and Epidemiology (SPINE): implementation of an electronic data collection tool within a large hospital in Malawi

Universal Transport Medium

Rapid HIV tests: guidelines for use in HIV testing and counselling services in resource-constrained settings

Malawi Population and Housing Census

Inquérito Nacional de Prevalência, Riscos Comportamentais e Informação sobre o HIV e SIDA em Moçambique

HIV prevalence estimates from the demographic and health surveys

Metric and euclidean properties of dissimilarity coefficients

Epidemiology of respiratory syncytial virus infection in rural and urban Kenya

Results from the first six years of national sentinel surveillance for influenza in Kenya

Increased burden of respiratory viral associated severe lower respiratory tract infections in children infected with human immunodeficiency virus type-1

Severe influenza-associated respiratory infection in high HIV prevalence setting, South Africa

Epidemiology of severe acute respiratory illness (SARI) among adults and children aged >/=5 years in a high HIV-prevalence setting

Mixed respiratory virus infections

Mortality associated with seasonal and pandemic influenza and respiratory syncytial virus among children <5 years of age in a high HIV prevalence setting-South Africa

Global burden of acute lower respiratory infections due to respiratory syncytial virus in young children: a systematic review and meta-analysis

Malawi: current issues and what the World Food Programme is doing

Global, regional, and national causes of child mortality in 2000-13, with projections to inform post-2015 priorities: an updated systematic analysis

Mapping HIV prevalence using population and antenatal sentinel-based HIV surveys: a multi-stage approach

Routine data from prevention of mother-to-child transmission (PMTCT) HIV testing not yet ready for HIV surveillance in Mozambique: a retrospective analysis of matched test results

Lessons learned from early implementation of option B+: the Elizabeth Glaser Pediatric AIDS Foundation experience in 11 African countries

Epidemiology of acute lower respiratory tract infection in HIV-exposed uninfected infants

Efficacy and immunogenicity of influenza vaccine in HIV-infected children: a randomized, double-blind, placebo controlled trial

Shedding of live vaccine virus, comparative safety, and influenza-specific antibody responses after administration of live attenuated and inactivated trivalent influenza vaccines to HIV-infected children

HIV virological suppression influences response to the AS03-adjuvanted monovalent pandemic influenza A H1N1 vaccine in HIV-infected children

Description of implementation of FTF multiplex assay FTD rRT-PCR assay was used in combination with the AgPath one-step qRT-PCR reagents according to the manufacturer's instructions (Applied Biosystems, Carlsad, California, USA)

Consolidated Standards of Reporting Trials diagram of data analyses. Abbreviations: FTD, FTD respiratory pathogens 33 kit; HIV, human immunodeficiency virus

SARI, severe acute respiratory infection

Disclaimer. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention (CDC).Financial support. This work was supported by with the CDC through a cooperative agreement (grant 5U01CK000146-04).Potential conflicts of interest. All authors: No reported conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.