influenza and rsv make a modest contribution to invasive pneumococcal disease incidence in the uk

Journal of Infection (2013) 66, 512e520

www.elsevierhealth.com/journals/jinf

Influenza and RSV make a modest contribution toinvasive pneumococcal disease incidence in the UK

Emily J. Nicoli a,*, Caroline L. Trotter a, Katherine M.E. Turner a,Caroline Colijn b, Pauline Waight c, Elizabeth Miller c

a School of Social and Community Medicine, University of Bristol, Canynge Hall, 39 Whatley Road, Bristol BS8 2PS, UKbDepartment of Mathematics, Imperial College London, South Kensington Campus, London SW7 2AZ, UKc Immunisation Department, Health Protection Agency, Colindale Avenue, London NW9 5EQ, UK

Accepted 6 February 2013Available online 5 March 2013

KEYWORDSInvasive pneumococcaldisease;Influenza;Respiratory syncytialvirus;Seasonality

* Corresponding author. Tel.: þ44 1E-mail address: [email protected]

0163-4453/$36 ª 2013 The British Infhttp://dx.doi.org/10.1016/j.jinf.2013

Summary Objectives: The common seasonality of incidence of invasive pneumococcal dis-ease (IPD) and viral respiratory infections has long been recognized, however, the extent towhich this affects the association between the pathogens is unknown. We have analysedweekly surveillance data of IPD, influenza and respiratory syncytial virus (RSV), using ambienttemperature and hours of sunshine as measures of seasonality.Methods: Reported cases of influenza, IPD and RSV, were collected in England and Wales, fromweek 1 (January) 1996 to week 23 (June) 2009. The associations between IPD and respiratoryviral infections were analysed using several statistical methods, including correlation coeffi-cients and both additive and multiplicative regression models.Results: 6e7.5% of cases of IPD are attributable to influenza and 3e4% attributable to RSV.Correlation coefficients reported considerably stronger associations between IPD and the viralinfections compared to regression models.Conclusions: A small but potentially important percentage of IPD may be attributable to influ-enza and RSV when adjusted for seasonality by temperature. Jointly these viral infections maylead to over 10% of IPD cases. Therefore, prevention of viral respiratory infections may offersome additional benefit in reducing invasive pneumococcal infections.ª 2013 The British Infection Association. Published by Elsevier Ltd. All rights reserved.

Introduction

The synergism between viral and bacterial infectionshas been widely reported, particularly for respiratory

17 928 7265..uk (E.J. Nicoli).

ection Association. Published by E.02.007

viruses and secondary bacterial pneumonia1; however, theunderlying mechanisms are complex and remain largely un-known.2,3 Respiratory infections, such as influenza, respira-tory syncytical virus (RSV) and Streptococcus pneumoniae,

lsevier Ltd. All rights reserved.

mailto:[email protected]

http://dx.doi.org/10.1016/j.jinf.2013.02.007

http://www.elsevierhealth.com/journals/jinf



Associations between viral infections and IPD 513

show strong seasonal patterns, each having increased inci-dence in winter in temperate areas of the world. Tempera-ture, humidity, pollution, light intensity and increasedcrowding in winter4e7 have all been suggested as factorsin causing the annual fluctuations in disease incidence. De-spite many studies and the use of multiple statistical tech-niques, the strength of association between invasivepneumococcal disease (IPD) and respiratory viral infectionsremains unclear.

There has been a recent resurgence in interest in therelationship between IPD and influenza in the context ofcontemporary pandemic influenza preparedness and theuse of the pneumococcal vaccines as an additional mea-sure to prevent mortality.8,9 At a population level, severalstudies of surveillance data, outside of influenza pan-demics, have sought to measure the associations betweeninfluenza, RSV and IPD.4,5,10e19 The reported strength ofthese associations varies between the studies, and ap-pears to depend, at least partially, on the quantity ofdata available as well as the methods used. Even withinthe same data sample, the use of different statisticalmethods can lead to wildly different results.10 The associ-ations are particularly difficult to measure because thecommon seasonality of the pathogens causes an overesti-mation of the result. A review of studies that have re-ported associations between IPD and influenza or RSVand their results can be found in the SupplementaryMaterial.

We have conducted a novel analysis of IPD, influenza andRSV surveillance data from England and Wales, using a rangeof statistical methods, in order to estimate the proportion ofIPD cases that are attributable to respiratory viruses, whilstattempting to account for the common seasonality of thepathogens.

Methods

Source of data

Clinically significant isolates of influenza,20 invasive pneu-mococcal disease (IPD)21 and respiratory syncytial virus(RSV) are recorded by microbiology laboratories in En-gland and Wales. These are reported on a weekly basisto the Health Protection Agency (HPA) as part of the na-tional surveillance system. We used data extracted fromthe HPA national surveillance database22 for influenzaand RSV, and for IPD used a reconciled dataset as previ-ously described.21 In brief, microbiology laboratories inEngland and Wales report all clinically significant pneu-mococcal isolates to the HPA through a computerized sys-tem (CoSurv). These isolates are often referred to theRespiratory and Vaccine Preventable Bacteria ReferenceUnit, HPA Microbiology Services for serotyping. Thesetwo datasets are then combined and any duplicates areremoved.

Weekly counts of cases, between 1st January 1996 to 7thJune 2009, stratified by age group (0e4 years, 5e14 years,15e64 years and 65 years and over) were analysed. Averageweekly temperatures (in degrees Celsius) and monthlysunshine (in hours) for the UK over the same period weresourced from the UK MET office information.23

Statistical analyses

The relationship between the weekly incidence of IPD andviral infections was initially analysed by calculating thePearson and Spearman’s correlation coefficients, for theoriginal and standardized datasets. The data were standard-ized in order to crudely remove the effect of the concurrentseasonality of the pathogens. For each weekly count, thedata were standardized by subtracting the mean and di-viding by the standard deviation of the counts for that weekover all of the years of the study period (13 years), thusproviding a measure of how the incidence for a particularweek deviates from the average for that time of year.

Three different regression models were investigated(Table 2). Two were additive models (a basic linear regres-sion and an identity-linked negative binomial regression)and one was a multiplicative model (a log-linked negativebinomial regression). The negative binomial regressionmodels were applied to account for over-dispersion of thedataset. The dependent variable was the incidence ofIPD, with explanatory variables, the incidence of influenzaand of RSV. Two additional explanatory variables, the UKmean weekly temperature23 and monthly hours of sunshine,were investigated in the models to adjust for the commonseasonality of the pathogens. The models were applied toall ages and then to each age group individually, as wellas to a range of lags (0e4 weeks).

We estimated the percentage of IPD cases that could beattributable to influenza and RSV. For the additive models,this was estimated by multiplying the virus’ case count withits regression coefficient. This determined the estimatednumber of cases of IPD attributable to the virus and fromwhich a percentage could be calculated. For the multipli-cative model, the percentages of IPD cases attributable toinfluenza and RSV were estimated by multiplying the virus’case count with its fitted rate ratio (RR), (attributablepercentage Z case count � (RR-1) � 100).24

All analyses were carried out with STATA version 11.2(StataCorp. 2009. Stata Statistical Software: Release 11.College Station, TX: StataCorp LP).

Results

The common seasonal incidence of all three diseases, IPD,influenza and RSV can be clearly seen in this dataset(Fig. 1). Whilst IPD cases are reported all year round, thereare distinct increases during winter months. For influenza,there are similarly timed peaks in reported incidence, butwith fewer cases out of season. The same is true for RSV,with very few cases reported in the summer months andwith large numbers of cases being reported, mainly in in-fants (Table 1), in the winter. Fig. 2a also displays thestrong shared seasonality of the different pathogens. How-ever, in Fig. 2b, where the data have been standardized toremove the effect of the common seasonality, the correla-tion in the scatter plot is weakened considerably comparedto Fig. 2a. The age group with the largest reported inci-dence of IPD is the over 65 year olds (49.0%). The 0e4year age group reported the most RSV infections (94.8%).For influenza, most cases were reported in the 15e64 yearsage group (49.6%).

Figure 1 Time series plots of incidence per 100 000 population of IPD, influenza and RSV, the mean weekly temperature (�C) andmonthly sunshine (h), from week 1 (January), 1996 to week 23 (June), 2009.

514 E.J. Nicoli et al.

Pearson and Spearman’s correlation coefficients betweenIPD and both respiratory viruses found strong, significantassociations for all age groups (Fig. 3): all coefficients havea P-value<0.001. In most age groups, the correlation coeffi-cients are higher for RSV than for influenza. Both coefficientsare highest in the older age groups, with the 65 years andover having the strongest correlation for IPD and influenzaand similarly strong associations for IPD and RSV.

In the multivariate regression analyses, the factor re-sponsible for the strongest associations with IPD is found tobe the average temperature as opposed to either of theviral infections or hours of sunshine (Tables 3 and 4). Therewas no evidence of an association between IPD and hours ofsunshine (results not shown). There was, however, some ev-idence of an association between IPD and one month laggedhours of sunshine (Table 4). For the age group of all ages,the strongest viral association is with influenza, followedby RSV, for all of the regression techniques. There is no ev-idence of any significant time lags in the incidence data(i.e. model fit did not improve with the introduction ofany lags of 1e4 weeks).

The linear regression model adjusted by weekly tem-perature indicates that 6.9% of IPD cases are attributableto influenza and 3.9% attributable to RSV, for all ages

Table 1 Number (percentage), by age group, of cases of each

Age group IPD

0e4 yrs 8153 (11.43%)5e14 yrs 1787 (2.51%)15e64 yrs 26 457 (37.09%)65þ yrs 34 936 (48.98%)Total 71 333

(Table 5). The results using the additive negative binomialmodel are similar (7.5% attributable to influenza and 3.5%attributable to RSV) and the results from the multiplicativenegative binomial model are slightly lower than the addi-tive models (5.6% attributable to influenza and 2.9% attrib-utable to RSV). For the linear model adjusted by laggedmonthly sunshine, 6.1% of IPD cases were attributable to in-fluenza and 3.8% attributable to RSV, for all ages (Table 6).The percentage is higher for the additive negative binomialmodel (9.2% attributable to influenza and 4.1% attributableto RSV) and lower for the multiplicative negative binomialmodel (5.7% attributable to influenza and 3.4% attributableto RSV). The multiplicative model tends to predict a lowerpercentage of attributable IPD cases to influenza and RSV inall of the age groups.

For RSV, the lowest percentage of attributable cases is inthe 0e4 year olds (1e2%, dependent on the model) and thehighest percentage is in the 15e64 year olds (15e25%). Thepercentages of attributable IPD cases increase across allage groups and in all models. The percentage of influenza-attributable cases increased with age from 0 to 6%. Thepercentage of cases attributable to RSV ranged from 3 to28% with the highest percentage being for the 15e64 yearsgroup.

disease over the study period.

Influenza RSV

4510 (18.30%) 107 525 (94.81%)2393 (9.71%) 1190 (1.05%)12 209 (49.55%) 3079 (2.72%)5528 (22.44%) 1616 (1.43%)24 640 113 410

Table 2 Description of regression models, method, log- and identity-linked and their equation.

Model Method Link Equation

1 Linear Identity Y Z b0 þ b1$XFLU þ b2$XRSV þ b3$XTEMP2 Negative binomial Identity Y Z b0 þ b1$XFLU þ b2$XRSV þ b3$XTEMP3 Negative binomial Log log(Y ) Z b0 þ b1$XFLU þ b2$XRSV þ b3$XTEMP


Discussion

We have found a small but statistically significant associ-ation between invasive pneumococcal disease and viralinfections after accounting for the common seasonality ofthe infections. Influenza-attributable IPD accounted forbetween 0 and 9.2% of cases of IPD according to age,meteorological variable and regression method used. Inthe additive negative binomial regression model, 7.5% ofIPD is attributable to influenza, for all ages, when adjustedby average temperature (best-fitting model). The percent-age of RSV-associated IPD accounted for between 1.5 and25% of all IPD cases, with 3.5% of IPD attributable to RSV,for all ages, when adjusted by average temperature in theadditive negative binomial regression model. Our resultsfor influenza are in line with those of other studiesapplying similar techniques. They found influenza wasassociated with 6e10%11 and 5e6%17 of IPD cases. Ourstudy is the first, to our knowledge, to estimate the IPDcases attributable to both influenza and RSV, in differentage groups and including average temperature and hoursof sunshine to allow for the seasonal characteristics ofthe data.

Our study has looked inmore detail at the influence of agein associations between IPD and viral infections. We found

Figure 2 Scatter plot of (a) weekly incidence of IPD against w[RIGHT]) and (b) the equivalent standardized results; with linear t

that for influenza the attributable percentage of IPD cases islowest in the 0e4 years group for both meteorologicalvariables (w0%) and highest in the over 65 years groupwhen adjusted by temperature (3.2e4.8%, dependent on themodel) or highest in the 5e14 years group when adjusted byhours of sunshine (5.7e6.9%). For RSV, the attributablepercentage of IPD cases was again lowest in the 0e4 yearsgroup for both meteorological variables (1e2%) and highestin the 15e64 years group for both variables (14.5e25%). Inprevious studies, evidence of associations between influenzaand IPD has been more consistently reported inadults11,13,14,17 compared to children where the associationsare weaker or non-existent.4,5,12,15,16 We also found that theassociations between IPD and influenza were stronger inolder age groups when adjusted by temperature. This wasnot the case when adjusted by hours of sunshine. Howeverthe data on hours of sunshine is only available at monthlytime periods as opposed to weekly temperature measure-ments and the association between IPD and temperaturewas found to be stronger than that between IPD and sunshine(where all data was converted to monthly time periods).

In the case of IPD and RSV in children, most studies4,5,18

have found the association between IPD and RSV was stron-ger than that of IPD and influenza, with only Talbot et al.15

finding the reverse result. Our study also estimates that

eekly incidence of viral infection (influenza [LEFT] and RSVrend lines and 95% confidence intervals.

Figure 3 Pearson and Spearman’s correlation coefficients forall ages and each age group for IPD and influenza and IPD andRSV (all coefficients have P-values <0.001).


more cases of IPD in children are attributable to RSV thaninfluenza; however the strength of the statistical evidenceof our results for influenza is weak. We also found a similarresult for adults.

Whether the average air temperature and hours ofsunshine are merely markers of some other underlyingprocess or whether they do truly affect the transmissibility,severity or duration of infection is still debated. However,due to the relative strength of the evidence that average

Table 3 Linear regression, negative binomial regression with anperature as a meteorological variable.

Age Variable Additive

Linear Reg. N. Binomial Re

Coefficient P-value Coefficient

All ages ‘Flu 0.262 <0.001 0.285RSV 0.032 <0.001 0.029Temp. �4.711 <0.001 �4.664Constant 136.102 <0.001 135.279

0e4 yrs ‘Flu �0.004 0.771 0.004RSV 0.002 0.046 0.002Temp. �0.613 <0.001 �0.630Constant 17.746 <0.001 17.848

5e14 yrs ‘Flu 0.033 0.023 0.032RSV 0.132 <0.001 0.109Temp. �0.084 <0.001 �0.094Constant 3.081 <0.001 3.227

15e64 yrs ‘Flu 0.092 0.002 0.071RSV 1.853 <0.001 1.640Temp. �1.158 <0.001 �1.401Constant 40.006 <0.001 43.771

65þ yrs ‘Flu 0.340 <0.001 0.406RSV 2.647 <0.001 2.502Temp. �2.407 <0.001 �2.298Constant 65.960 <0.001 64.553

temperature18,25e30 and hours of sunshine,6,14,18,31e35 areassociated with IPD and viral infections it was determinedthat they were a logical choice to account for seasonality.We found a 1 month lag in the association between IPDand hours of sunshine, consistent with three other studiesreporting lags of 2e5 weeks6,14,33 though no lag was re-ported in 2 other studies.31,32 This may be related to thestrong, positive effects of sunlight on the immune systemdue to increased 1,25-(OH)2-vitamin-D metabolism.35,36

Other meteorological factors such as rainfall and relativehumidity were not included in the models as associationswith IPD and viral infections are less consistent.14,18,27,31

It may be that the use of average temperature as an adjust-ment for seasonality has led to slightly lower percentages,for some age-groups, of influenza-attributable IPD whencompared to previous studies which included seasonal har-monic curves.11,17 However, the use of harmonic curvesdoes not allow for annual variations.

From our results using Pearson and Spearman’s correla-tion coefficients, we could conclude that there is a verystrong association between IPD and the viral infections;however these are rather crude measures of associationthat cannot be seasonally adjusted, and so are likely tooverestimate any association in our data. Further analysis,beyond the use of correlation coefficients, should beconsidered in similar studies of seasonal diseases in orderto formulate more robust conclusions. We investigateda range of regression models; looking at both additive andmultiplicative models. It is considered that the additivemodel is a more plausible fit for this biological data,37

identity link and with a log link for each age group with tem-

Multiplicative

g. with identity link N. Binomial Reg. with log link

P-value Coefficient P-value

<0.001 0.002 <0.001<0.001 0.000 0.001<0.001 �0.060 <0.001<0.001 5.095 <0.001

0.787 �0.001 0.5770.032 0.000 0.135

<0.001 �0.060 <0.001<0.001 3.024 <0.001

0.076 0.009 0.0900.001 0.034 0.001

<0.001 �0.043 <0.001<0.001 1.246 <0.001

0.064 0.001 0.098<0.001 0.032 <0.001<0.001 �0.048 <0.001<0.001 3.896 <0.001

<0.001 0.004 <0.001<0.001 0.034 <0.001<0.001 �0.062 <0.001<0.001 4.366 <0.001

Table 4 Linear regression, negative binomial regression with an identity link and with a log link for each age group withmonthly lagged hours of sunshine as a meteorological variable.

Age Variable Additive Multiplicative

Linear Reg. N. Binomial Reg. with identity link N. Binomial Reg. with log link

Coefficient P-value Coefficient P-value Coefficient P-value

All ages ‘Flu 0.226 <0.001 0.349 <0.001 0.000 <0.001RSV 0.030 0.006 0.033 0.005 0.000 0.013Sun �1.655 <0.001 �1.320 <0.001 �0.004 <0.001Constant 577.816 <0.001 519.261 <0.001 6.456 <0.001

0e4 yrs ‘Flu �0.004 0.879 0.013 0.571 0.000 0.832RSV 0.002 0.293 0.002 0.255 0.000 0.415Sun �0.213 <0.001 �0.213 <0.001 �0.005 <0.001Constant 74.342 <0.001 73.834 <0.001 4.427 <0.001

5e14 yrs ‘Flu 0.060 0.014 0.066 0.029 0.004 0.017RSV 0.149 0.008 0.147 0.005 0.010 0.010Sun �0.015 0.235 �0.018 0.084 �0.002 0.036Constant 10.881 <0.001 11.165 <0.001 2.500 <0.001

15e64 yrs ‘Flu 0.043 0.414 0.094 0.157 0.000 0.407RSV 2.396 <0.001 2.752 <0.001 0.011 <0.001Sun �0.237 0.050 �0.168 0.059 �0.003 <0.001Constant 142.349 <0.001 125.034 <0.001 5.105 <0.001

65þ yrs ‘Flu 0.204 0.010 0.447 0.006 0.001 0.079RSV 3.800 <0.001 4.509 <0.001 0.013 <0.001Sun �0.692 <0.001 �0.503 <0.001 �0.004 <0.001Constant 251.428 <0.001 215.461 <0.001 5.633 <0.001


a multiplicative relationship between the independent vari-able terms in the model would be hard to substantiate.However, it is difficult to firmly conclude which model isthe best as we have no gold standard for comparison (seeref and below).10

The ecological nature of this study restricts theconclusions that can be drawn. Research at an individuallevel may be more revealing with respect to the trueincidence of virus-attributable IPD, but will be morechallenging. Potential study designs that could be em-ployed include case-control studies of IPD with serologicalinvestigations of recent viral infections. There are furtherlimitations in the use of surveillance systems for the datain this study, under-reporting and changes over time in thereporting thresholds cannot be ruled out. The lack ofdistinction between the strains and serotypes of thediseases could mask any true associations between IPDand the viral infections, if the association only applies toa subgroup of the strains or serotypes.38 There are impor-tant issues with the statistical methods used to gauge theassociations, other studies10,38 have found, as we have,that the different methods produce variable results.Moreover, negative binomial regression requires assump-tions to be made about the data; the observations shouldbe independent and the virulence of the viruses should re-main constant.

The possible mechanisms underlying the interactionbetween S. pneumoniae and influenza and RSV have beenreviewed by Bosch et al.39 A primary host defence to infec-tion is the secretion of a mucus layer in the upper

respiratory tract. Bacteria bind to the mucus40,41 enablingthem to be cleared by the action of cilia cells. However,primary viral infection destroys these epithelial cellsthrough metabolic exhaustion or lysis39 reducing mucusand bacterial clearance.42 This enables bacteria to prog-ress further into the respiratory tract by inhalation or ad-herence to exposed cell surface receptors.43,44 Viralfactors produced by influenza and RSV also increase bacte-rial adherence. Influenza produces neuraminidase (NA),which cleaves sialic acids exposing bacterial receptorsand thus increasing adherence.45 RSV expresses RSV-protein G which acts directly as a bacterial receptor.46 Vi-ral infection may alter behaviour of the immune system, bymodifying the expression of antimicrobial peptides47 andadhesion proteins, these act as receptors for immune cells,however S. pneumoniae and other bacteria have beenshown to adhere to these proteins as well.48,49 Influenza vi-rus is also known to impair neutrophil function and in-crease apoptosis,50 decrease oxidative burst51 and reduceproduction and activity of cytokines.39

The time period of our analysis covers only seasonalinfluenza and excludes the H1N1 ‘swine flu’ pandemic.We censored our dataset at the week preceding theWorld Health Organization’s (WHO) declaration of thepandemic on 11th June 2009 because the UK surveillancesystems were modified and enhanced during the pan-demic, making direct comparisons with previous timeperiods difficult. During the second wave of the pan-demic in winter 2010/2011, linkage between influenzaand invasive bacterial infection surveillance reports

Table 6 Percentage of IPD cases attributable to influenza and RSV for the two additive models; linear regression model and negative binomial regression model with identity-link (additive) and with log-link (multiplicative) with monthly lagged hours of sunshine as a meteorological variable.

Age Attributable to Linear regression Additive negative binomial regression Multiplicative negative binomial regression

% Std deva 95% CIb % Std deva 95% CIb % Std deva 95% CIb

All ages Influenza 6.127 6.925 0.723 19.907 9.233 9.907 1.172 29.012 5.740 8.462 0.414 23.312RSV 3.751 5.284 0.193 16.457 4.122 5.824 0.213 17.789 3.393 5.220 0.096 15.328

0e4 yrs Influenza �0.184 0.348 �0.717 0.000 0.562 1.030 0.000 2.210 �0.221 0.423 �0.917 0.000RSV 1.580 2.368 0.052 7.066 1.612 2.409 0.055 7.215 1.213 1.913 0.024 5.638

5e14 yrs Influenza 6.377 8.142 0.000 23.730 6.899 8.673 0.000 25.393 5.686 8.849 0.000 23.850RSV 7.944 9.531 0.000 29.074 7.724 9.267 0.000 28.360 7.600 10.624 0.000 34.290


65þ yrs Influenza 2.533 3.295 0.242 9.416 5.322 6.429 0.545 19.129 2.213 3.985 0.129 8.642RSV 14.120 10.792 0.000 35.426 16.574 12.330 0.000 40.386 12.977 14.289 0.000 42.692

a Standard deviation.b 95% confidence interval.

Table 5 Percentage of IPD cases attributable to influenza and RSV for the two additive models; linear regression model and negative binomial regression model with identity-link (additive) and with log-link (multiplicative) with temperature as a meteorological variable.

Age Attributable to Linear regression Additive negative binomial regression Multiplicative negative binomial regression

% Std deva 95% CIb % Std deva 95% CIb % Std deva 95% CIb

All ages Influenza 6.888 8.114 0.461 26.312 7.461 8.689 0.503 28.131 5.557 8.745 0.158 23.806RSV 3.892 5.734 0.167 17.731 3.525 5.235 0.150 16.175 2.940 4.779 0.054 14.208

0e4 yrs Influenza �0.198 0.410 �0.932 0.000 0.184 0.374 0.000 0.868 �0.373 0.785 �1.793 0.000RSV 1.764 2.766 0.038 8.338 1.956 3.052 0.044 9.175 1.418 2.364 0.018 6.949



65þ yrs Influenza 4.047 5.593 0.000 16.328 4.781 6.502 0.000 19.176 3.226 6.306 0.000 14.687RSV 9.514 9.365 0.000 28.468 9.053 8.962 0.000 27.222 7.905 10.236 0.000 27.356

a Standard deviation.b 95% confidence interval.

518E.J.

Nico

lietal.


suggested that between 6 and 11% (depending on age,with the highest percentage in the 15e44 year age group)of IPD cases had concurrent influenza.52 This is broadlyconsistent with our findings.

We suggest that there is a small, but measurable associ-ation between IPD and RSV and influenza. These results arerelevant for public health policy decisionmaking. Preventionof viral respiratory infections may offer some additionalbenefit in terms of reducing invasive pneumococcal infec-tions53 and prevention of pneumococcal infections during,say, influenza pandemics could see a reduction in hospitali-zations andmortality.8,9 Therewould bemerit in consideringboth IPD and viral infections when looking at interventionsfor any one of these infections and this could be explored us-ing modelling techniques.54

Acknowledgements

This work was supported by a Medical Research Council PhDstudentship to EN.

Appendix A. Supplementary data

Supplementary data related to this article can be found athttp://dx.doi.org/10.1016/j.jinf.2013.02.007.

References

1. McCullers JA. Insights into the interaction between influenzavirus and pneumococcus. Clinical Microbiology Reviews 2006;19(3):571e82.

2. Lipsitch M, Viboud C. Influenza seasonality: lifting the fog. Pro-ceedings of the National Academy of Sciences of the UnitedStates of America 2009;106(10):3645e6.

3. Altizer S, Dobson A, Hosseini P, Hudson P, Pascual M, Rohani P.Seasonality and the dynamics of infectious diseases. EcologyLetters 2006;9(4):467e84.

4. Ampofo K, Bender J, Sheng X, Korgenski K, Daly J, Pavia AT,et al. Seasonal invasive pneumococcal disease in children:role of preceding respiratory viral infection. Pediatrics 2008;122(2):229e37.

5. Kim PE, Musher DM, Glezen WP, Rodriguez-Barradas MC,Nahm WK, Wright CE. Association of invasive pneumococcaldisease with season, atmospheric conditions, air pollution,and the isolation of respiratory viruses. Clinical Infectious Dis-eases 1996;22(1):100e6.

6. Dowell SF, Whitney CG, Wright C, Rose Jr CE, Schuchat A. Sea-sonal patterns of invasive pneumococcal disease. Emerging In-fectious Diseases 2003;9(5):573e9.

7. Lofgren E, Fefferman NH, Naumov YN, Gorski J, Naumova EN.Influenza seasonality: underlying causes and modeling theo-ries. Journal of Virology 2007;81(11):5429e36.

8. Rubin JL, McGarry LJ, Klugman KP, Strutton DR, Gilmore KE,Weinstein MC. Public health and economic impact of vaccina-tion with 7-valent pneumococcal vaccine (PCV7) in the contextof the annual influenza epidemic and a severe influenza pan-demic. BMC Infectious Disease 2010;10:14.

9. Simonsen L, Taylor RJ, Young-Xu Y, Haber M, May L,Klugman KP. Impact of pneumococcal conjugate vaccinationof infants on pneumonia and influenza hospitalization and mor-tality in all age groups in the United States. MBio 2011;2(1):e00309e10.

10. Gilca R, De SG, Skowronski D, Boivin G, Buckeridge DL. Theneed for validation of statistical methods for estimating respi-ratory virus-attributable hospitalization. American Journal ofEpidemiology 2009;170(7):925e36.

11. Grabowska K, Hogberg L, Penttinen P, Svensson A, Ekdahl K.Occurrence of invasive pneumococcal disease and number ofexcess cases due to influenza. BMC Infectious Disease 2006;6:58.

12. Jansen AG, Sanders EA, Van Der Ende A, van Loon AM, Hoes AW,Hak E. Invasive pneumococcal and meningococcal disease: as-sociation with influenza virus and respiratory syncytial virus ac-tivity? Epidemiology and Infection 2008;136(11):1448e54.

13. Kyaw MH, Clarke S, Jones IG, Campbell H. Incidence of invasivepneumococcal disease in Scotland, 1988e99. Epidemiology andInfection 2002;128(2):139e47.

14. Murdoch DR, Jennings LC. Association of respiratory virus activ-ity and environmental factors with the incidence of invasivepneumococcal disease. Journal of Infection 2009;58(1):37e46.

15. Talbot TR, Poehling KA, Hartert TV, Arbogast PG, Halasa NB,Edwards KM, et al. Seasonality of invasive pneumococcal dis-ease: temporal relation to documented influenza and respira-tory syncytial viral circulation. American Journal of Medicine2005;118(3):285e91.

16. Toschke AM, Arenz S, von KR, Puppe W, Weigl JA, Hohle M,et al. No temporal association between influenza outbreaksand invasive pneumococcal infections. Archives of Disease inChildhood 2008;93(3):218e20.

17. Walter ND, Taylor TH, Shay DK, Thompson WW, Brammer L,Dowell SF, et al. Influenza circulation and the burden of inva-sive pneumococcal pneumonia during a non-pandemic periodin the United States. Clinical Infectious Diseases 2010;50(2):175e83.

18. Watson M, Gilmour R, Menzies R, Ferson M, McIntyre P. The as-sociation of respiratory viruses, temperature, and other cli-matic parameters with the incidence of invasivepneumococcal disease in Sydney, Australia. Clinical InfectiousDiseases 2006;42(2):211e5.

19. Kuster SP, Tuite AR, Kwong JC, McGeer A, Fisman DN. Evalua-tion of coseasonality of influenza and invasive pneumococcaldisease: results from prospective surveillance. PLoS Medicine2011;8(6):e1001042. [Epub 2011/06/21].

20. HPA, Goddard NL, Joseph CA, Zambon M, Nunn M, Fleming D,et al. Influenza surveillance in the United Kingdom: October2000 to May 2001. Communicable Disease Report CDR Weekly2001;11:1e7 [serial online].

21. Trotter CL, Waight P, Andrews NJ, Slack M, Efstratiou A,George R, et al. Epidemiology of invasive pneumococcal dis-ease in the pre-conjugate vaccine era: England and Wales,1996e2006. Journal of Infection 2010;60(3):200e8.

22. Laboratory Reporting to the Health Protection Agency. Guidefor diagnostic laboratories. Version 3.0. London: Health Pro-tection Agency. Available from: http://www.hpa.org.uk/web/HPAwebFile/HPAweb_C/1194947381307; August 2010.

23. Parker DE, Legg TP, Folland CK. A new daily central Englandtemperature series, 1772e1991. International Journal of Cli-matology 1992;12(4):317e42.

24. Mangtani P, Hajat S, Kovats S, Wilkinson P, Armstrong B. The as-sociation of respiratory syncytial virus infection and influenzawith emergency admissions for respiratory disease in London:an analysis of routine surveillance data. Clinical Infectious Dis-eases 2006;42(5):640e6.

25. Braga AL, Zanobetti A, Schwartz J. The effect of weather onrespiratory and cardiovascular deaths in 12 U.S. cities. Envi-ronmental Health Perspectives 2002;110(9):859e63.

26. Fisman D. Seasonality of viral infections: mechanisms and un-knowns. Clinical Microbiology and Infection: The Official Pub-lication of the European Society of Clinical Microbiology andInfectious Diseases 2012;18(10):946e54. [Epub 2012/07/24].


http://www.hpa.org.uk/web/HPAwebFile/HPAweb_C/1194947381307

http://www.hpa.org.uk/web/HPAwebFile/HPAweb_C/1194947381307


27. Tang JW, Lai FY, Nymadawa P, Deng YM, Ratnamohan M,Petric M, et al. Comparison of the incidence of influenza in re-lation to climate factors during 2000e2007 in five countries.Journal of Medical Virology 2010;82(11):1958e65. [Epub2010/09/28].

28. du Prel JB, Puppe W, Grondahl B, Knuf M, Weigl JA, Schaaff F,et al. Are meteorological parameters associated with acute re-spiratory tract infections? Clinical Infectious Diseases: An Offi-cial Publication of the Infectious Diseases Society of America2009;49(6):861e8. [Epub 2009/08/12].

29. Lowen AC, Steel J, Mubareka S, Palese P. High temperature(30 �C) blocks aerosol but not contact transmission of influ-enza virus. Journal of Virology 2008;82(11):5650e2. [Epub2008/03/28].

30. Tang JW. The effect of environmental parameters on the sur-vival of airborne infectious agents. Journal of the Royal Soci-ety, Interface/the Royal Society 2009;6(Suppl. 6):S737e46.[Epub 2009/09/24].

31. Yusuf S, Piedimonte G, Auais A, Demmler G, Krishnan S, VanCaeseele P, et al. The relationship of meteorological conditionsto the epidemic activity of respiratory syncytial virus. Epidemi-ology and Infection 2007;135(7):1077e90. [Epub 2007/03/10].

32. White AN, Ng V, Spain CV, Johnson CC, Kinlin LM, Fisman DN.Let the sun shine in: effects of ultraviolet radiation on invasivepneumococcal disease risk in Philadelphia, Pennsylvania. BMCInfectious Diseases 2009;9:196. [Epub 2009/12/08].

33. Paynter S, Weinstein P, Ware RS, Lucero MG, Tallo V,Nohynek H, et al. Sunshine, rainfall, humidity and child pneu-monia in the tropics: time-series analyses. Epidemiology andInfection 2012:1e9. [Epub 2012/08/14].

34. Dowell SF. Seasonal variation in host susceptibility and cyclesof certain infectious diseases. Emerging Infectious Diseases2001;7(3):369e74. [Epub 2001/06/01].

35. Cannell JJ, Vieth R, Umhau JC, Holick MF, Grant WB,Madronich S, et al. Epidemic influenza and vitamin D. Epidemi-ology and Infection 2006;134(6):1129e40. [Epub 2006/09/09].

36. Hayes CE, Nashold FE, Spach KM, Pedersen LB. The immunolog-ical functions of the vitamin D endocrine system. Cellular andMolecular Biology 2003;49(2):277e300. [Epub 2003/07/31].

37. Gay NJ, Andrews NJ, Trotter CL, Edmunds WJ. Estimatingdeaths due to influenza and respiratory syncytial virus.JAMA: The Journal of the American Medical Association2003;289(19):2499 [author reply 500e2. Epub 2003/05/22].

38. Thompson WW, Weintraub E, Dhankhar P, Cheng PY,Brammer L, Meltzer MI, et al. Estimates of US influenza-associated deaths made using four different methods. Influ-enza Other Respi Viruses 2009;3(1):37e49.

39. Bosch AA, Biesbroek G, Trzcinski K, Sanders EA, Bogaert D. Vi-ral and bacterial interactions in the upper respiratory tract.PLoS Pathogens 2013;9(1):e1003057. [Epub 2013/01/18].

40. Plotkowski MC, Beck G, Puchelle E. A new model for studyingbacterial adherence to the respiratory epithelium. BrazilianJournal of Medical and Biological Research Z Revista brasi-leira de pesquisas medicas e biologicas/Sociedade Brasileirade Biofisica [et al] 1988;21(2):285e8. [Epub 1988/01/01].

41. Feldman C, Read R, Rutman A, Jeffery PK, Brain A, Lund V,et al. The interaction of Streptococcus pneumoniae with intacthuman respiratory mucosa in vitro. The European Respiratory

Journal: Official Journal of the European Society for ClinicalRespiratory Physiology 1992;5(5):576e83. [Epub 1992/05/01].

42. Pittet LA, Hall-Stoodley L, Rutkowski MR, Harmsen AG. Influ-enza virus infection decreases tracheal mucociliary velocityand clearance of Streptococcus pneumoniae. American Jour-nal of Respiratory Cell and Molecular Biology 2010;42(4):450e60. [Epub 2009/06/13].

43. van der Flier M, Chhun N, Wizemann TM, Min J, McCarthy JB,Tuomanen EI. Adherence of Streptococcus pneumoniae to im-mobilized fibronectin. Infection and Immunity 1995;63(11):4317e22. [Epub 1995/11/01].

44. Plotkowski MC, Puchelle E, Beck G, Jacquot J, Hannoun C. Ad-herence of type I Streptococcus pneumoniae to tracheal epi-thelium of mice infected with influenza A/PR8 virus. TheAmerican Review of Respiratory Disease 1986;134(5):1040e4. [Epub 1986/11/01].

45. Peltola VT, McCullers JA. Respiratory viruses predisposing tobacterial infections: role of neuraminidase. Pediatric Infec-tious Disease Journal 2004;23(1 Suppl.):S87e97.

46. Hament JM, Aerts PC, Fleer A, Van Dijk H, Harmsen T, Kimpen JL,et al. Enhanced adherence of Streptococcus pneumoniae to hu-man epithelial cells infected with respiratory syncytial virus. Pe-diatric Research 2004;55(6):972e8. [Epub 2004/04/23].

47. McGillivary G, Mason KM, Jurcisek JA, Peeples ME, Bakaletz LO.Respiratory syncytial virus-induced dysregulation of expressionof a mucosal beta-defensin augments colonization of the upperairway by non-typeable Haemophilus influenzae. Cellular Mi-crobiology 2009;11(9):1399e408. [Epub 2009/06/09].

48. Cundell DR, Gerard NP, Gerard C, Idanpaan-Heikkila I,Tuomanen EI. Streptococcus pneumoniae anchor to activatedhuman cells by the receptor for platelet-activating factor. Na-ture 1995;377(6548):435e8. [Epub 1995/10/05].

49. Avadhanula V, Rodriguez CA, Devincenzo JP, Wang Y,Webby RJ, Ulett GC, et al. Respiratory viruses augment the ad-hesion of bacterial pathogens to respiratory epithelium in a vi-ral species- and cell type-dependent manner. Journal ofVirology 2006;80(4):1629e36. [Epub 2006/01/28].

50. Shahangian A, Chow EK, Tian X, Kang JR, Ghaffari A, Liu SY,et al. Type I IFNs mediate development of postinfluenza bacte-rial pneumonia in mice. The Journal of Clinical Investigation2009;119(7):1910e20. [Epub 2009/06/03].

51. McNamee LA, Harmsen AG. Both influenza-induced neutrophildysfunction and neutrophil-independent mechanisms contrib-ute to increased susceptibility to a secondary Streptococcuspneumoniae infection. Infection and Immunity 2006;74(12):6707e21. [Epub 2006/09/20].

52. Zakikhany K, Degail MA, Lamagni T, Waight P, Guy R, Zhao H,et al. Increase in invasive Streptococcus pyogenes and Strepto-coccus pneumoniae infections in England, December 2010 toJanuary 2011. Euro Surveillance 2011;16(5).

53. O’Brien KL, Walters MI, Sellman J, Quinlisk P, Regnery H,Schwartz B, et al. Severe pneumococcal pneumonia in previ-ously healthy children: the role of preceding influenza infec-tion. Clinical Infectious Diseases 2000;30(5):784e9.

54. Chien YW, Levin BR, Klugman KP. The anticipated severity ofa “1918-like” influenza pandemic in contemporary popula-tions: the contribution of antibacterial interventions. PLoSOne 2012;7(1):e29219. [Epub 2012/02/01].

influenza and rsv make a modest contribution to invasive pneumococcal disease incidence in the uk

Documents