Predictors of influenza a molecular viral shedding in Hutterite communities

Background Patterns of influenza molecular viral shedding following influenza infection have been well established; predictors of viral shedding however remain uncertain. Objectives We sought to determine factors associated with peak molecular viral load, duration of shedding, and viral area under the curve (AUC) in children and adult Hutterite colony members with laboratory‐confirmed influenza. Methods A cohort study was conducted in Hutterite colonies in Alberta, Canada. Flocked nasal swabs were collected during three influenza seasons (2007‐2008 to 2009‐2010) from both symptomatic and asymptomatic individuals infected with influenza. Samples were tested by real‐time reverse‐transcription polymerase chain reaction for influenza A and influenza B, and the viral load was determined for influenza A‐positive samples. Results For seasonal H1N1, younger age was associated with a larger AUC, female sex was associated with decreased peak viral load and reduced viral shedding duration, while the presence of comorbidity was associated with increased peak viral load. For H3N2, younger age was associated with increased peak viral load and increased AUC. For pandemic H1N1, younger age was associated with increased peak viral load and increased viral AUC, female sex was associated with reduced peak viral load, while inapparent infection was associated with reduced peak viral load, reduced viral shedding duration, and reduced viral AUC. Conclusions Patterns of molecular viral shedding vary by age, sex, comorbidity, and the presence of symptoms. Predictor variables vary by influenza A subtype.

Background: Patterns of influenza molecular viral shedding following influenza infection have been well established; predictors of viral shedding however remain uncertain.
Objectives: We sought to determine factors associated with peak molecular viral load, duration of shedding, and viral area under the curve (AUC) in children and adult Hutterite colony members with laboratory-confirmed influenza.
Methods: A cohort study was conducted in Hutterite colonies in Alberta, Canada. Results: For seasonal H1N1, younger age was associated with a larger AUC, female sex was associated with decreased peak viral load and reduced viral shedding duration, while the presence of comorbidity was associated with increased peak viral load.
For H3N2, younger age was associated with increased peak viral load and increased AUC. For pandemic H1N1, younger age was associated with increased peak viral load and increased viral AUC, female sex was associated with reduced peak viral load, while

| INTRODUCTION
Influenza is a major cause of annual epidemics of respiratory illness, with the potential to lead to hospitalizations and death. 1 (pH1N1) viral shedding in either community or institutional settings. [5][6][7][8][9][10][11] Few studies characterize viral shedding patterns following naturally acquired seasonal influenza and pH1N1 and most are only descriptive. 4,[12][13][14] Factors associated with influenza viral loads, duration of viral shedding, and viral area under the curve (AUC) remain uncertain. 15,16 In our previous study, 17 we described molecular viral shedding patterns in naturally infected children and adults over multiple seasons in Hutterite communities. The objective of this study was to examine predictors of viral shedding patterns in these participants.

| Study population and surveillance
The study population and surveillance methods are described in detail in our previous study. 17 To summarize, children and adults residing in 10 Hutterite colonies within 150 km from the city of Red Deer, Alberta, were enrolled. The surveillance period was from December 29, 2007, until June 15, 2010. All study participants were assessed twice weekly by a research nurse using a standardized self-reported symptom or sign checklist, which was completed by a representative of a family for all family members and was provided when the research nurse made a site visit. Study participants with any new symptoms reported were interviewed by the research nurse while onsite at the colony. Children aged between 36 months and 15 years in colonies were vaccinated against influenza in the three flu seasons during the study period with inactivated influenza vaccine.
One nasopharyngeal specimen and one flocked nasal specimen were collected from participants who had two or more of the following symptoms: fever (≥38°C), cough, nasal congestion, sore throat, headache, sinus problems, muscle aches, fatigue, ear ache, and chills. Flocked nasal swabs were collected for up to 8 weeks from the participants whose nasopharyngeal swab tested positive for influenza by reverse-transcription polymerase chain reaction (RT-PCR). Specimens were collected daily for 7 days and then every two days for up to 8 weeks or when two consecutive specimens tested negative.
We also collected specimens from asymptomatic participants in colonies when it was established that the colonies had an outbreak, which we defined as ≥2 positive specimens within any 48-hour period. Specimens from asymptomatic participants were collected for up to 3 weeks: daily for the first week and then every second day until two consecutive specimens tested negative or for a maximum

| Statistical analysis
We performed descriptive statistical analyses to summarize the characteristics by age, sex, comorbidity, and vaccination status of the study participants over the three seasons. We also used mixed-effect univariate logistic regression to test the association between these predictors and the development of influenza.
We then analyzed factors associated with peak viral load, duration of viral shedding, and viral AUC in this study. Factors considered in our analyses were age, sex, comorbidity, vaccination status, and asymptomatic status. Univariate analysis was performed using Student's t test. Factors with a P-value <.2 in univariate analysis were considered for multivariable analysis. We used backwards selection method to build our final models and retained variables with a P-value <.05. For all multivariable analyses, we used mixed-effect models to account for the hierarchical nature of the data (ie, the effect of colony and household). 18,19 We defined peak viral load (expressed in copies/mL and then logtransformed) as the maximum viral RNA concentrations in all specimens collected from a participant during an episode of infection and used mixed-effect regression models to identify independent predictors. An episode of infection began upon the initial detection of influenza by RT-PCR and ended when two consecutive RT-PCR tests were negative. We defined prolonged shedding as detectable viral RNA at >5 days after the episode start date. 20 We analyzed duration of viral shedding using mixed-effect Cox proportional hazard. Viral AUC was calculated to assess both the level and duration of viral shedding with inapparent infection was associated with reduced peak viral load, reduced viral shedding duration, and reduced viral AUC.

Conclusions:
Patterns of molecular viral shedding vary by age, sex, comorbidity, and the presence of symptoms. Predictor variables vary by influenza A subtype.

K E Y W O R D S
influenza viral shedding, predictors, viral AUC, viral load, viral shedding duration a single parameter [20]. We adopted the definition of viral AUC from Shaefer et al. 16 as in Equation (1).
where t 1 ,t 2 ,…,t n were the days on which samples were collected; y 1 , y 2 ,…..,y n were corresponding viral concentration levels measured on the days. Viral AUC (expressed in copies/mL-days and then log-transformed) was modeled using mixed-effect regression models.
In all analyses, a P-value of <.05 was considered to indicate statistical significance. All probabilities were two-tailed. Statistical analyses were performed with R (version 3.2.2) software. 21  Nine (15%) of the 62 seasonal H1N1 cases were asymptomatic (Table 2). Of these, seven (78%, 7/9) cases had other seasonal H1N1 cases in their household. 31 (50%) of the 62 seasonal H1N1 cases had received the seasonal vaccine in 2007-2008 season ( Table 2).

| Baseline characteristics
Of these, 26 (84%, 26/31) cases had other seasonal H1N1 influenza cases in their household. There were six seasonal H1N1 cases that were both asymptomatic and vaccinated with the seasonal vaccine.
Of these, two (66%, 4/6) had other seasonal H1N1 cases in their household. Figure 1 compares peak viral load, viral shedding duration, and viral AUC for seasonal H1N1 between asymptomatic and symptomatic cases and between non-vaccinated and vaccinated cases.
There were no asymptomatic seasonal H3N2 cases ( Table 2). Five  Table 2). Of these, two (40%, 2/5) cases had other seasonal H3N2 cases in their household. Figure 2 compares peak viral load, viral shedding duration, and viral AUC for seasonal H3N2 between non-vaccinated and vaccinated cases.

| Peak load analysis
For the 2007-2008 season (seasonal H1N1), we found that the mean peak viral load (±SD) among males was significantly greater than in females (5.87±0.96 vs 5.37±0.85 log 10 copies/mL, P=.04) ( was associated with a decreased peak viral load for pH1N1 (Table 4).

| DISCUSSION
We found that predictor variables varied by influenza A subtype.
For seasonal H1N1, younger age was associated with a larger AUC, while female sex was associated with decreased peak viral load and reduced viral shedding duration while the presence of comorbidity was associated with increased peak viral load. For H3N2, younger age was associated with increased peak viral load and increased AUC. For pandemic H1N1, younger age was associated with increased peak viral load and increased viral AUC, female sex was associated with reduced peak viral load, while inapparent infection was associated with reduced peak viral load, reduced viral shedding duration, and reduced viral AUC.
Age was significantly associated with viral AUC for all the three subtypes of influenza A virus. Our analyses suggested younger cases had significantly higher viral AUC for all the three influenza A subtypes. As no data are available with respect to influenza viral AUC, our findings contribute to the literature.
Our results demonstrate that H1N1 viral shedding in males tends to reach higher levels and lasts longer than in females. One possible explanation is that females typically generate higher innate and adaptive immune responses accelerating viral clearance and reducing virus load. 22,23 However, our findings were not consistent for all three subtypes of influenza A virus, and further investigations to explain this discrepancy are needed. 24 Comorbidity significantly increased peak viral load for H1N1, but not for pH1N1. For pH1N1, Lee et al. 25 also reported that comorbidity was not significantly associated with viral loads. We were unable to assess comorbidity associations for H3N2 from our data set, as the sample size was too small. However, Lee et al. 25  T A B L E 3 (Continued) H3N2 cases. Another study 15 also suggested that the presence of major comorbidities was a significant factor that affected initial viral load in adult patients hospitalized with H3N2.
Asymptomatic status was a significant variable in all the three analyses for pH1N1, but not for seasonal H1N1. We were unable to assess association for H3N2 from our data set, as we did not detect any asymptomatic H3N2 cases in our study. Although it has been reported that asymptomatic shedding is common for both seasonal influenza A 26 and pandemic strain, 10  The reduced risk associated with vaccination in 2008-09 season could be explained by the good match between influenza vaccine antigens and circulating strains in this season. 28 However, there has been controversy about an increased risk of pH1N1 associated with seasonal influenza (TIV) vaccination. Two studies from Canada 29,30 reported an increased risk of pH1N1 associated with receipt of seasonal TIV.
Conversely, there also were studies suggesting cross-protection of TIV against pH1N1. 31 A systematic review and meta-analysis by Yin et al. 32 showed that moderate cross-protection might exist with no overall increased risk of infection from TIV. 32 Our finding of pH1N1 risk associated with TIV is consistent with other findings in Canada. 30 T A B L E 4 Explanatory variables in final models for peak viral load, viral shedding duration, and viral AUC a For the analysis of peak viral load and viral AUC, we used mixed-effect regression model to account for the hierarchy structure of the data (colony and household); for the analysis of viral shedding duration, we used mixed-effect Cox proportional hazards regression model to account for the hierarchy structure of the data set. b We defined prolonged shedding as detectable viral RNA at>5 d after the episode start date.
Strengths of this study include assessment over multiple seasons, inclusion of both adults and children, analysis by influenza A subtypes, active follow-up for a relatively long period, and systematic swabbing of asymptomatic participants. One limitation is that we were not able to collect specimens daily from some participants (46%). Therefore, the value of peak viral load for these participants might not be their real peak value. Another limitation is that we did not assess VL for influenza B. Lastly, we conducted the study in Hutterite communities, which might limit generalizability of our results.
We conclude that patterns of influenza A molecular viral shedding vary by age, sex, comorbidity, and the presence of symptoms and predictor variables vary by influenza A subtype. The findings may inform interpretation of epidemiologic patterns of influenza for different influenza A subtypes.