J Infect Dis.-2010-Lau-1509-16

Viral Shedding in Influenza Infections • JID 2010:201 (15 May) • 1509

M A J O R A R T I C L E

Viral Shedding and Clinical Illness in NaturallyAcquired Influenza Virus Infections

Lincoln L. H. Lau,1 Benjamin J. Cowling,1 Vicky J. Fang,1 Kwok-Hung Chan,2 Eric H. Y. Lau,1 Marc Lipsitch,3

Calvin K. Y. Cheng,1 Peter M. Houck,4 Timothy M. Uyeki,5 J. S. Malik Peiris,2 and Gabriel M. Leung1

1Infectious Disease Epidemiology Group, School of Public Health, and 2Department of Microbiology, University of Hong Kong, Hong Kong, China;3Harvard School of Public Health, Boston, Massachusetts; 4Seattle Quarantine Station, Division of Global Migration and Quarantine, Centersfor Disease Control and Prevention, National Center for Preparedness, Detection and Control of Infectious Diseases, Seattle, Washington;5Influenza Division, National Center for Immunization and Respiratory Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia

Background. Volunteer challenge studies have provided detailed data on viral shedding from the respiratorytract before and through the course of experimental influenza virus infection. There are no comparable quantitativedata to our knowledge on naturally acquired infections.

Methods. In a community-based study in Hong Kong in 2008, we followed up initially healthy individualsto quantify trends in viral shedding on the basis of cultures and reverse-transcription polymerase chain reaction(RT-PCR) through the course of illness associated with seasonal influenza A and B virus infection.

Results. Trends in symptom scores more closely matched changes in molecular viral loads measured withRT-PCR for influenza A than for influenza B. For influenza A virus infections, the replicating viral loads determinedwith cultures decreased to undetectable levels earlier after illness onset than did molecular viral loads. Most viralshedding occurred during the first 2–3 days after illness onset, and we estimated that 1%–8% of infectiousnessoccurs prior to illness onset. Only 14% of infections with detectable shedding at RT-PCR were asymptomatic, andviral shedding was low in these cases.

Conclusions. Our results suggest that “silent spreaders” (ie, individuals who are infectious while asymptomaticor presymptomatic) may be less important in the spread of influenza epidemics than previously thought.

Influenza virus is associated with substantial mortality

and morbidity worldwide through seasonal epidemics

and occasional emergence of novel strains that lead to

pandemics [1]. Volunteer challenge studies have pro-

vided detailed data on viral shedding from the respi-

ratory tract through the course of experimental influ-

Received 12 November 2009; accepted 11 December 2009; electronicallypublished 12 April 2010.

Potential conflicts of interest: none reported.Financial support: US Centers for Disease Control and Prevention (grant 1 U01

CI000439-02), the Research Fund for the Control of Infectious Disease, Food andHealth Bureau, Government of the Hong Kong SAR (grant 08070632), the HarvardCenter for Communicable Disease Dynamics from the US National Institutes ofHealth Models of Infectious Disease Agent Study program (grant 1 U54 GM088558),and the Area of Excellence Scheme of the Hong Kong University Grants Committee(grant AoE/M-12/06). The funding agencies had no role in data collection andanalysis, or the decision to publish, but the CDC was involved in study designand preparation of the manuscript. This work represents the views of the authorsand not their institutions, including the Centers for Disease Control and Prevention.

Reprints or correspondence: Dr Benjamin J Cowling, School of Public Health,University of Hong Kong, Units 624-7, Cyberport 3, Pokfulam, Hong Kong ([email protected]).

The Journal of Infectious Diseases 2010; 201(10):1509–1516� 2010 by the Infectious Diseases Society of America. All rights reserved.0022-1899/2010/20110-0010$15.00DOI: 10.1086/652241

enza virus infections and the time lines and duration

of illness after infection [2]. However, in volunteer chal-

lenge studies, the participants are typically young adults

who have been screened to ensure that they have low

levels of preexisting immunity to the influenza strain

of interest; thus, the findings may not generalize to the

overall population annually at risk for naturally ac-

quired infections [2]. There are few data on viral shed-

ding after naturally acquired influenza virus infection

in a community setting. The proportion of infections

that are asymptomatic or subclinical and the degree to

which these are contagious, as well as the proportion

of shedding that occurs prior to onset of symptoms,

affects the potential impact of control measures. We

studied trends in clinical illness and viral shedding as-

sociated with naturally acquired influenza virus infec-

tions in a community setting.

METHODS

We conducted a community-based study of the effec-

tiveness of face masks and hand hygiene to prevent

influenza transmission in households during 2008 [3].

by guest on August 30, 2015

http://jid.oxfordjournals.org/D

ownloaded from

http://jid.oxfordjournals.org/

1510 • JID 2010:201 (15 May) • Lau et al

As part of that study, index cases in patients of all ages who

presented with �2 of a range of signs and symptoms consistent

with influenza virus infection were recruited from outpatient

clinics and private hospital emergency rooms from January

through September 2008. A positive result for influenza virus

infection on a rapid antigen test was used to determine eligi-

bility of the index cases’ households for additional follow-up.

Index cases and household contacts were eligible regardless of

preexisting conditions or treatment prescribed, but index cases

who had household members with concurrent influenza-like

illness were excluded [3, 4]. There was no recorded use of

antiviral prophylaxis. Symptom diaries were provided to all

household members at an initial home visit within 24–48 h of

index case recruitment to record respiratory and systemic

symptoms daily; 82% of initial home visits were conducted

within 12 h of index case recruitment [3]. Compliance with

symptom reporting was high; all symptom diaries were com-

plete with the exception of 12 (0.6%) of 2170 entries. We pro-

vided each household with a digital tympanic thermometer and

asked all household members to record their body temperature

daily. Household members were followed up for ∼7 days to

observe secondary infections. Pooled nose and throat swab

(NTS) specimens were collected from all household members

regardless of illness at the initial home visit and at 2 follow-

up visits 3 (�1) and 6 (�2) days later. Because index cases

are likely to have higher viral loads, given their positive result

on a rapid test [5], and are less likely to have mild illness given

that they sought outpatient care, all analyses here focus only

on the household contacts to avoid selection bias.

Laboratory methods. NTS specimens from household con-

tacts were tested by quantitative reverse transcription poly-

merase chain reaction (RT-PCR) to detect influenza A or B

virus infection and determine molecular viral loads. Total nu-

cleic acid was extracted from specimens by using the NucliSens

easyMAG extraction system (bioMerieux, Boxtel, the Nether-

lands) according to the manufacturer’s instructions. Twelve mi-

croliters of extracted nucleic acid was used to prepare com-

plementary DNA by using an Invitrogen Superscript III kit

(Invitrogen) with random primer as described elsewhere [6].

For detection of influenza A virus, 2 mL of complementary

DNA was amplified in a LightCycler 2.0 system (Roche Di-

agnostics), with a total reaction-mix volume of 20 mL reaction

containing FastStart DNA Master SYBR Green I Mix reagent

kit (Roche Diagnostics), 4.0 mmol/L MgCl2 and 0.5 mmol/L

of each primer. The forward primer (5′-CTTCTAACCGAGGT-

CGAAACG-3′) and the reverse primer (5′-GGCATTTTGGAC-

AAAKCGTCTA-3′) were used for amplification of the matrix

gene of influenza A virus [7]. Cycling conditions were as fol-

lows: initial denaturation at 95�C for 10 min, followed by 40

cycles of 95�C for 10 s, 60�C for 3 s, and 72�C for 12 s, with

ramp rates of 20�C/s. At the end of the assay, polymerase chain

reaction (PCR) products were subjected to a melting-curve

analysis to determine the specificity of the assay. The lower

limit of detection of the RT-PCR assay was 23 virus gene copies

per reaction (ie, ∼900 copies/mL).

For detection of influenza B virus, forward (5′-GCATCTTT-

TGTTTTTTATCCATTCC) and reverse (5′-CACAATTGCCT-

ACCTGCTTTCA) primers and 5′ nuclease probe (Fam-TGCT-

AGTTCTGCTTTGCCTTCTCCATCTTCT-TAMRA) were used

for amplification of the matrix gene [8]. Testing was performed

by using the TagMan EZ RT-PCR system. Core reagent kit

(Applied Biosystems), with 0.8 mmol/L of forward and reverse

primers and 0.2 mmol/L of probe in a total reaction volume of

25 mL, comprising 4 mL of nucleic acid extract. Amplification

and detection was performed on an ABI StepOneTM real-time

PCR system (Applied Biosystems) under the following con-

ditions: initial hold at 50�C for 20 min and 95�C for 15 min,

followed by 45 cycles at 95�C for 15 s and 60�C for 1 min.

NTS specimens were additionally tested by quantitative vi-

ral dilutions to detect median tissue culture infectious dose

(TCID50) and determine replicating viral load. Madin-Darby

canine kidney (MDCK) cells were grown on microtiter plates.

The cells were rinsed in serum-free minimum essential medium

(Gibco). The NTS specimen was diluted initially by 1 in 5 and

then in 10-fold steps in serum-free minimum essential medium

containing tosylsulfonyl phenylalanyl chloromethyl ketone-

treated trypsin (2 mg/mL) (Sigma Aldrich). One hundred mi-

croliters of the undiluted NTS specimen as well as each of the

specimen dilutions was added in quadruplicate to the MDCK

cell monolayers. Medium alone was added to control cells. An

additional 100 mL of serum-free minimum essential medium

with 2 mg/mL trypsin was added to each well, and the plates

were incubated at 33�C for 7 days. The plates were evaluated

for cytopathic effect daily. The TCID50 was determined ac-

cording to the Reed and Muench Method. The lower limit of

detection was ∼100.3 TCID50.

Statistical analysis. We calculated daily scores grouped as

systemic signs and symptoms (fever of �37.8�C, headache, and

myalgia), upper respiratory symptoms (sore throat and runny

nose), and lower respiratory symptoms (cough and phlegm)

by summing the presence versus absence of each symptom or

sign and dividing by 3, 2, and 2, respectively [2]. We define

acute respiratory illness (ARI) as �2 of the 7 symptoms listed

above [9]. We plotted average symptom scores by time since

ARI onset, which was defined as the first day when the subject

reported �2 of the 7 symptoms. We plotted quantitative viral

loads by time since ARI onset and calculated daily geometric

means, imputing half the lower limit of detection (ie, 450 cop-

ies/mL and 0.15 log10 TCID50) for the undetectable values. As

a sensitivity analysis, we plotted viral shedding by time since

onset of fever of �37.8�C. We investigated the associations



ownloaded from



Table 1. Comparison of Characteristics of 59Secondary Influenza Virus Infections with 76 Ex-cluded Infections

This table is available in its entirety in the onlineversion of Journal of Infectious Diseases.

Table 2. Initial Symptoms and Signs of Naturally Acquired In-fluenza A and B Virus Infections Reported at Acute RespiratoryIllness (ARI) Onset

Symptom or sign

No. (%) of symptomsreported at illness onset

Influenza A(n p 26)

Influenza B(n p 18)

Runny nose or nasal congestion 19 (73) 11 (61)Cough 18 (69) 14 (78)Sore throat 14 (54) 7 (39)Headache 14 (54) 5 (28)Phlegm 12 (46) 5 (28)Myalgia 9 (35) 6 (33)Fever �37.8�C 8 (31) 8 (44)

NOTE. ARI onset is defined as the first day with �2 of the 7 signs orsymptoms listed.

between tympanic temperature and viral shedding, and between

symptom scores and viral shedding, by day since ARI onset.

We used a Bayesian approach to fit alternative parametric

forms to the viral shedding trajectories and selected between

models using the Bayesian information criterion [10]. We used

the models to quantify the proportion of infectiousness re-

maining at ARI onset and 1, 2, or 3 days after onset, assuming

lognormal, Weibull, or gamma-form associations between mo-

lecular viral shedding and infectiousness. Additional technical

details are provided in the Appendix, which is not available in

the print edition of the Journal. All analyses were conducted

with R software (version 2.7.1; R Development Core Team)

[11] and WinBUGS software (version 1.4) [12]. Additional

information about the study design, raw data from the study,

and R syntax to permit reproducible statistical analyses are avail-

able on the authors’ Web site at http://www.hku.hk/bcowling/

influenza/HK_NPI_study.htm.

RESULTS

We followed up 1015 household contacts in 322 households.

A total of 135 (13%) contacts were confirmed with RT-PCR

to have influenza virus infection. Because we recorded symp-

toms prospectively and were particularly interested in patterns

of viral shedding and symptoms around illness onset, for this

analysis we excluded 59 contacts for whom the first NTS spec-

imens collected were RT-PCR positive regardless of whether

ARI was reported at the first home visit, and an additional 17

contacts who met the criteria for ARI onset at the first home

visit with influenza virus infection subsequently confirmed with

RT-PCR. Characteristics of the excluded 76 contacts were sim-

ilar to those of the 59 retained contacts (Table 1). Of the 59

secondary infections analyzed here, 16 were caused by influenza

A/H1N1 viruses, 17 by influenza A/H3N2 viruses, and 1 by

both influenza A/H1N1 and A/H3N2 confection, with the re-

mainder by influenza B virus. Forty-six (78%) of 59 subjects

were persons aged 16 years or older. At ARI onset, the most

common symptoms were cough, nasal congestion, runny nose,

and sore throat (Table 2). Of the 34 subjects infected with

influenza A viruses, 15 (44%) reported a temperature of

�37.8�C on at least 1 day over the course of illness, whereas

21 (62%) reported having a temperature of �37.8�C on at least

1 day, taking antipyretic medication, or both. The correspond-

ing proportions for influenza B virus infections were 8 (32%)

of 25 subjects and 13 (52%) of 25 subjects. Thirty (51%) of

59 subjects reported seeking medical care.

Peak molecular viral shedding for influenza A virus infection

as assessed with RT-PCR occurred on the day of ARI onset,

followed by a steady decrease in viral shedding through the

following 7 days (Figure 1). The pattern of viral shedding for

A/H1N1 and for A/H3N2 subtypes were similar (data not

shown). Viral shedding was detected with RT-PCR in 4 (27%)

of 15 subjects (95% confidence interval [CI], 8%–55%), with

a specimen collected 1 day before ARI onset. Influenza B viral

shedding was more variable over time and a clear peak was not

observed; initial shedding was detected at 1–2 days before ARI

onset in 4 (29%) of 14 subjects (95% CI, 8%–58%) and per-

sisted for ∼6 days before subsiding. We did not have sufficient

sample size to explore differences in viral shedding between

children and adults.

In influenza A virus infections the replicating viral load de-

termined by viral culture peaked on the day of ARI onset and

steadily decreased over ∼5 days (Figure 1). For influenza virus

B infections, the TCID50 levels initially peaked on the day of

ARI onset and were more variable over time. Molecular viral

shedding was most closely correlated with replicating viral load

on the day after ARI onset. As replicating viral load decreased

more rapidly with time, the association between molecular viral

shedding and replicating viral load became weaker later in the

course of illness (Figures 2 and 3). Replicating viral load de-

termined with viral culture was detected in 1 adult with influ-

enza A infection and 2 adults and 3 children with influenza B

infection before ARI onset.

We found that a modified lognormal form provided a good

fit to the molecular viral shedding patterns for influenza A virus

infections. If infectiousness were proportional to molecular vi-

ral shedding, most infectiousness would occur within 2–3 days

of ARI onset, whereas if infectiousness were proportional to



ownloaded from


1512 • JID 2010:201 (15 May) • Lau et al

Figure 1. Patterns of viral shedding and symptoms and signs in naturally acquired influenza A and B virus infections by day relative to acuterespiratory illness (ARI) onset (day 0). Top row: Viral shedding for children (plus signs) and adults (circles), and the geometric mean viral shedding(solid lines) of (A) 26 individuals with influenza A and (B ) 18 individuals with influenza B virus infections. The lower limit of detection of the reverse-transcription polymerase chain reaction assay was ∼900 copies/mL (gray line). Second row: Median tissue culture infectious dose (TCID50) of specimenscollected from children (plus signs) and adults (circles), and the geometric mean TCID50 of (C ) influenza A and (D ) influenza B. Third row: Symptomsand signs for subjects with (E ) influenza A and (F ) influenza B virus infection are split into lower respiratory (dotted line), upper respiratory (dashedline), and systemic (solid line). Bottom row: Mean tympanic temperature associated with (G ) influenza A and (H ) influenza B virus infections. ARIonset is defined as at the first day with �2 of the 7 signs or symptoms listed in Table 2. Individuals with asymptomatic or subclinical infectionswere excluded. NTS, nose and throat swab specimen.

log10 molecular viral shedding or presence of detectable viral

RNA, infectiousness would persist for longer periods (Figure

4). We did not find any parametric forms that provided a good

fit to the influenza B molecular viral shedding patterns, or

replicating viral load for influenza A or B measured by TCID50.

In both influenza A and B virus infections, mean tympanic

temperature and trends in average scores for systemic signs and

symptoms peaked at the day of ARI onset and then steadily

subsided to baseline after 3–5 days (Figure 1). On average,

upper respiratory and lower respiratory symptom scores peaked

at ARI onset; however, respiratory symptoms resolved less

quickly than systemic symptoms and signs (Figure 1). For in-

fluenza A virus infections. tympanic temperature was signifi-

cantly positively correlated with viral RNA shedding on the day

following ARI onset ( ), although correlations were lessP p .01

clear later in the course of illness (Figure 5). There was also a

positive correlation between number of symptoms and molec-

ular viral shedding (data not shown). Data were insufficient to

repeat these analyses for influenza B virus infections.

We detected viral shedding with RT-PCR in the absence of

any reported signs or symptoms in 8 (14%) of 59 subjects (95%

CI, 6.0%–25%), 5 of whom were infected with influenza A

virus. In 2 of the 8 subjects, only NTS samples collected at the

final home visit were positive with RT-PCR; we may have iden-



ownloaded from



Figure 2. Association between replicating (by median tissue culture infectious dose [TCID50]) and molecular (by reverse-transcription polymerasechain reaction) influenza A viral shedding by day since acute respiratory illness (ARI) onset for children (plus signs) and adults (circles). Linear regressionlines are plotted where there are �3 points. ARI onset is defined as at the first day with �2 of the 7 symptoms or signs listed in Table 2. NTS,nose and throat swab specimen.

Figure 3. Association between replicating (by median tissue cultureinfective dose [TCID50]) and molecular (by reverse-transcription polymerasechain reaction) influenza B viral shedding by day since acute respiratoryillness (ARI) onset for children (plus signs) and adults (circles). Linearregression lines are plotted where there are �3 points. ARI onset isdefined as at the first day with �2 of the 7 symptoms or signs listedin Table 1. NTS, nose and throat swab specimen.

tified presymptomatic rather than asymptomatic shedding be-

cause these subjects may have subsequently developed symp-

toms. The geometric mean of peak viral titers in the NTS

specimens of these 8 asymptomatic subjects was cop-33.2 � 10

ies/mL compared with copies/mL in symptomatic73.6 � 10

subjects. Among the 8 individuals with asymptomatic infec-

tions, 3 had specimens that were positive on quantitative viral

culture with a geometric mean TCID50 of . Subclinical in-0.710

fections were identified in an additional 7 of 59 subjects who

reported at most 1 of 7 signs or symptoms on each day of the

study period, and viral shedding in the NTS specimens of these

7 subjects had geometric mean copies/mL. Three of33.4 � 10

7 specimens from subjects with subclinical infection were pos-

itive on viral culture with a mean TCID50 of .0.510

Viral shedding and replication trends assessed with RT-PCR

and TCID50, respectively, were also assessed relative to time

since fever resolution, along with symptom scores and tympanic

body temperature time lines for influenza A (data not shown).

Viral shedding measured with RT-PCR peaked 1 day before

fever resolution. Shedding was detected in 4 (44%) of 9 subjects

(95% CI, 14%–79%) after fever resolution. Symptom trends,

mean tympanic temperature, and replicating viral load mea-

sured by TCID50 all followed a similar trend.

In a sensitivity analysis, we investigated viral shedding relative

to onset of fever of �37.8�C in the subset of 22 subjects who

reported a fever and found that trends in viral shedding were

generally consistent with the main analysis based on ARI onset

(data not shown).

DISCUSSION

There are few data in the literature on the patterns in infec-

tiousness in influenza virus infections over time, and how in-

fectiousness may be related to viral shedding or symptoms. We

investigated 3 alternative simple models for infectiousness in

influenza A virus infections and found that if infectiousness is

proportional to viral RNA shedding (or log10 viral shedding or

presence/absence of shedding), then the majority of infectious-

ness occurs within 1–2 days (or 3–4 days) after ARI onset

(Figure 4). Given the more rapid decrease in replicating viral

load compared with molecular viral shedding (Figure 1), we

may have overestimated the duration of infectiousness. In pre-

vious work, we estimated the mean serial interval of influenza

to be 3.6 days [9], given an incubation period of 1.5–2 days

[13]. This implies that the average time between ARI onset and

subsequent transmission to household contacts was ∼2 days,



ownloaded from


1514 • JID 2010:201 (15 May) • Lau et al

Figure 4. Proportion of infectiousness remaining by time since acuterespiratory illness (ARI) onset in the course of naturally acquired influenzaA virus infection. Infectiousness assumed proportional to (A) molecularviral shedding, (B ) log10 molecular viral shedding, and (C ) presence ofmolecular viral shedding in excess of 900 copies/mL under the fittedcurve.

Figure 5. Association between tympanic temperature and molecularinfluenza A viral shedding by reverse-transcription polymerase chain re-action by day since acute respiratory illness (ARI) onset for children(plus signs) and adults (circles). Linear regression lines are plotted wherethere are �3 points. ARI onset is defined as at the first day with �2of the 7 signs or symptoms listed in Table 1. NTS, nose and throat swabspecimen.

which is consistent with infectiousness profiles in between a

proportional relationship with log10 viral shedding and viral

shedding. From a detailed model of influenza transmission in

households in a French household study [14], Ferguson et al

estimated that infectiousness was highest around the time of

illness onset, and that the infectiousness profile correlated

closely with viral shedding [15], whereas in our data viral shed-

ding peaked within 1–2 days of symptom onset (Figure 1).

Similarly, Pitzer et al [16] found that moderate increases in

transmissibility of severe acute respiratory syndrome 5–10 days

after illness onset appeared more consistent with changes in

log viral shedding rather than viral shedding over the same

period [17]. Because the majority of viral shedding measured

with RT-PCR or TCID50 occurred within 2–3 days of ARI onset,

isolation of patients or other interventions must be applied

very soon after ARI onset; otherwise, they could not have any

substantial effectiveness in reducing onwards transmission.

Our data are consistent with data from previous studies that

documented the time lines of influenza viral shedding from

volunteer challenge studies [2]. We found that influenza A viral

shedding measured with RT-PCR peaked around the same day

as ARI onset before decreasing. Whereas some viral shedding

was detectable with RT-PCR before ARI onset, this occurred

in a minority of cases. Daily replicating viral shedding measured

with TCID50 for influenza A virus infections decreased faster

than viral shedding measured with RT-PCR. Patterns in viral

shedding associated with influenza B virus infections were more

variable and followed a pattern similar to that of data observed

in volunteer challenge studies, with shedding at substantial lev-

els from ARI onset through to 5 days after ARI onset [2]. We

found that levels of replicating virus measured by TCID50 had

high correlation with molecular viral loads measured with RT-

PCR soon after ARI onset but diverged as the infection pro-

gressed when TCID50 decreased more rapidly (Figures 2 and

3). A previous study also found that viral culture is less sensitive

than RT-PCR later in the course of illness [18], perhaps because

inactivated virus or viral DNA persist in the respiratory tract

towards the end of illness.

In both influenza A and B virus infections, systemic symp-

toms and signs subsided more rapidly than respiratory symp-

toms [2]. The decrease of systemic signs and symptoms, pri-

marily fever, correlated closely with the decrease in viral shed-

ding, most likely due to the subsiding of immune response with

the gradual clearing of virus from the body [19]. Resolution

of fever could also be affected by use of antipyretic medication.

The correlation between higher tympanic temperature and

higher viral shedding (Figure 5) suggests that individuals with

a higher fever could be more infectious in general.

Eight (14%) of the 59 individuals (95% CI, 6.0%–25%) with

RT-PCR–confirmed secondary infections did not report any

clinical signs or symptoms, and in total 15 (25%) of 59 subjects

(95% CI, 15%–38%) were asymptomatic (reporting 0 of 7 signs

or symptoms) or subclinical (reporting 1 of 7 signs or symp-

tom). Our upper bound for the frequency of asymptomatic

infection is somewhat lower than that identified in volunteer

challenge studies [2] or from longitudinal studies including pre-

and post-season serology with illness recall [20]. It is possible

that we failed to detect infections associated with lower levels

of viral shedding or shedding for a shorter period, because

swab samples were collected on average only at 3-day intervals,

and proportionally more such infections may be asymptomatic

[2, 21]. Some individuals may have been infected without shed-

ding virus. It is still unknown whether or to what degree asymp-

tomatic individuals could transmit infection to others [22],

although mathematical models typically assume that 33%–50%

of infections are asymptomatic or subclinical, and these indi-

viduals are approximately half as infectious as symptomatic

persons [23, 24]. Our results suggest the possibility that “silent



ownloaded from



spreaders” (ie, individuals who are infectious while asymptom-

atic or presymptomatic) may be less important in the spread

of epidemics than was previously thought.

US pandemic guidelines suggest sick individuals should re-

main home for a minimum of 24 h after fever resolution in

the absence of fever-reducing medication [25]. In our study of

seasonal influenza predominantly among adults, viral shedding

resolved within the recommended exclusion period for 10

(83%) of 12 individuals with febrile influenza A virus infection,

suggesting that the exclusion period covers the majority of

infectiousness for febrile infections if the patterns of fever and

viral shedding are similar to that in pandemic influenza.

A strength of our study is that we analyzed laboratory-con-

firmed, naturally acquired influenza virus infections in a com-

munity setting, which should allow broad generalizability. It is

important also to note the limitations of our study. First, be-

cause of sample-size limitations, our study was underpowered

to explore in detail the differences by age and other charac-

teristics. Second, because our study design tended to exclude

households with 11 case at the recruitment stage, subjects in

our study could be biased toward those having a lower risk of

infection or illness [3]. Third, although we have investigated

alternative models for the relationship between infectiousness

and viral shedding, it is possible that specific symptoms also

play an important role in infectiousness. Fourth, we have a

relatively small sample size, and more data would be valuable

to more precisely characterize patterns in viral shedding and

illness, and particularly viral shedding associated with asymp-

tomatic infections and before illness onset. Finally, we did not

record symptoms in household contacts retrospectively, so we

were unable to include in this analysis household contacts with

evidence of influenza virus infection at the initial home visit;

however, the general characteristics of excluded contacts were

similar to those of contacts who were included (Table 1).

One research gap highlighted by our work is the need for

studies that collect paired serologic data as well as detailed viral

shedding data. This would allow more accurate estimates of

the proportion of infections that are asymptomatic or subclin-

ical, and the characteristics of viral shedding in asymptomatic

infections. Future studies of naturally acquired infections with

increased frequency of clinical specimens and a longer follow-

up period would also contribute valuable knowledge about the

shape of the peak and duration of viral shedding. If combined

with data on transmission to contacts, these studies may be

able to provide additional information on the relations between

viral shedding, illness, and infectiousness, which would facilitate

optimal application of interventions and control measures. Vol-

unteer challenge studies can provide useful data, but large de-

tailed studies of naturally acquired infections are essential to

inform differences in time lines and trends by age and other

characteristics.

Acknowledgments

We thank all the physicians, nurses, and staff of participating centers forfacilitating recruitment and Rita Fung, Lai-Ming Ho, Winnie Wai, andEileen Yeung for research support.

References

1. Nicholson KG, Wood JM, Zambon M. Influenza. Lancet 2003; 362:1733–1745.

2. Carrat F, Vergu E, Ferguson NM, et al. Time lines of infection anddisease in human influenza: a review of volunteer challenge stud-ies. Am J Epidemiol 2008; 167:775–785.

3. Cowling BJ, Chan KH, Fang VJ, et al. Facemasks and hand hygieneto prevent influenza transmission in households: a cluster random-ized trial. Ann Intern Med 2009; 151:437–446.

4. Cowling BJ, Fung RO, Cheng CK, et al. Preliminary findings of arandomized trial of non-pharmaceutical interventions to preventinfluenza transmission in households. PLoS One 2008; 3:e2101.doi:10.1371/journal.pone.0002101. Published 7 May 2008.

5. Cheng CK, Cowling BJ, Chan KH, et al. Factors affecting QuickVueInfluenza A + B rapid test performance in the community setting. Di-agn Microbiol Infect Dis 2009; 65:35–41.

6. Peiris JS, Tang WH, Chan KH, et al. Children with respiratory dis-ease associated with metapneumovirus in Hong Kong. Emerg InfectDis 2003; 9:628–633.

7. Chan KH, Peiris JSM, Lim W, Nicholls JM, Chiu SS. Comparison ofnasopharyngeal flocked swabs and aspirates from pediatric patients forrapid diagnosis of respiratory viruses. J Clin Virol 2008; 42:65–69.

8. Lambert SB, Whiley DM, O’Neill NT, et al. Comparing nose-throatswabs and nasopharyngeal aspirates collected from children with symp-toms for respiratory virus identification using real-time polymerasechain reaction. Pediatrics 2008; 122:e615–e620. doi:10.1542peds.2008-0691. Published 25 August 2008.

9. Cowling BJ, Fang VJ, Riley S, Peiris JSM, Leung GM. Estimation ofthe serial interval of influenza. Epidemiology 2009; 20:344–347.

10. Schwarz GE. Estimating the dimension of a model. Ann Stat 1978; 6:461–464.

11. R Development Core Team. R: a language and environment for sta-tistical computing. Vienna, Austria: R Foundation for Statistical Com-puting, 2005.

12. Spiegelhalter DJ, Thomas A, Best NG, Lunn D. WinBUGS user manual,version 1.4. Cambridge, England: MRC Biostatistics Unit; 2003.

13. Lessler J, Reich NG, Brookmeyer R, Perl TM, Nelson KE, CummingsDA. Incubation periods of acute respiratory viral infections: a system-atic review. Lancet Infect Dis 2009; 9:291–300.

14. Viboud C, Boelle PY, Cauchemez S, et al. Risk factors of influenzatransmission in households. Br J Gen Pract 2004; 54:684–689.

15. Ferguson NM, Cummings DA, Cauchemez S, et al. Strategies for con-taining an emerging influenza pandemic in Southeast Asia. Nature 2005;437:209–214.

16. Pitzer VE, Leung GM, Lipsitch M. Estimating variability in the trans-mission of severe acute respiratory syndrome to household contactsin Hong Kong, China. Am J Epidemiol 2007; 166:355–363.

17. Peiris JS, Chu CM, Cheng VC, et al. Clinical progression and viral loadin a community outbreak of coronavirus-associated SARS pneumo-nia: a prospective study. Lancet 2003; 361:1767–1772.

18. Lee N, Chan PK, Hui DS, et al. Viral loads and duration of viral shed-ding in adult patients hospitalized with influenza. J Infect Dis 2009; 200:492–500.

19. Hayden FG, Fritz R, Lobo MC, Alvord W, Strober W, Straus SE. Localand systemic cytokine responses during experimental human influenzaA virus infection: relation to symptom formation and host defense. JClin Invest 1998; 101:643–649.



ownloaded from


1516 • JID 2010:201 (15 May) • Lau et al

20. Carrat F, Sahler C, Rogez S, et al. Influenza burden of illness: estimatesfrom a national prospective survey of household contacts in France.Arch Intern Med 2002; 162:1842–1848.

21. Foy HM, Cooney MK, Allan ID, Albrecht JK. Influenza B in house-holds: virus shedding without symptoms or antibody response. Am JEpidemiol 1987; 126:506–515.

22. Patrozou E, Mermel LA. Does influenza transmission occur fromasymptomatic infection or prior to symptom onset? Public Health Rep2009; 124:193–196.

23. Ferguson NM, Cummings DA, Fraser C, Cajka JC, Cooley PC, BurkeDS. Strategies for mitigating an influenza pandemic. Nature 2006; 442:448–452.

24. Longini IM, Jr., Halloran ME, Nizam A, Yang Y. Containing pandemicinfluenza with antiviral agents. Am J Epidemiol 2004; 159:623–633.

25. Centers for Disease Control and Prevention. CDC recommendationsfor the amount of time persons with influenza-like illness should beaway from others. http://www.cdc.gov/h1n1flu/guidance/exclusion.htm. Published 5 August 2009. Accessed 19 August 2009.



ownloaded from


Documents

J Infect Dis.-2010-Lau-1509-16