Parvovirus B19 in five European countries: disentangling the sociological and microbiological mechanisms underlying infectious disease transmission

Transcription

1 Parvovirus B19 in five European countries: disentangling the sociological and microbiological mechanisms underlying infectious disease transmission Nele Goeyvaerts 1, Niel Hens 1, John Edmunds 2, Marc Aerts 1, Philippe Beutels 3 1 Interuniversity Institute for Biostatistics and statistical Bioinformatics, Universiteit Hasselt & Katholieke Universiteit Leuven, Belgium 2 Infectious Diseases Epidemiology Unit, London School of Hygiene & Tropical Medicine, University of London, United Kingdom 3 Centre for Health Economics Research and Modeling Infectious Diseases & Centre for the Evaluation of Vaccination, Vaccine & Infectious Disease Institute, University of Antwerp, Belgium

2 Outline 1 Introduction Parvovirus B19 Serological Data Objectives 2 Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates 3 4

3 Parvovirus B19 Parvovirus B19 Serological Data Objectives causes range of diseases, e.g. erythema infectiosum = fifth disease or slapped cheek syndrome primarily spread by infected respiratory droplets transmission requires sufficiently close contact mean infectious period of 6 days incubation period > latent period in children and teenagers: usually mild during pregnancy: complications may occur (0 9% miscarriage) currently no vaccine available

4 Serological Data Parvovirus B19 Serological Data Objectives Belgium, England and Wales, Finland, Italy and Poland serum samples collected between 1995 and 2004 tested for IgG antibodies against parvovirus B19 Country Year of collection Age range Sample size BE EW FI IT PL

5 Objectives Introduction Parvovirus B19 Serological Data Objectives For each country: estimate age-specific transmission rates using data on social contacts are the contact data fully explaining the observed seroprofile? need for an additional age-specific factor? Comparative study to assess country-specific differences in: effectiveness of a close contact the shape of an additional age-specific factor

6 Objectives Introduction Parvovirus B19 Serological Data Objectives For each country: estimate age-specific transmission rates using data on social contacts are the contact data fully explaining the observed seroprofile? need for an additional age-specific factor? Comparative study to assess country-specific differences in: effectiveness of a close contact the shape of an additional age-specific factor understanding age-specific transmission of parvovirus B19 may help to understand transmission of other airborne infections

7 The Mass Action Principle Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates Mass Action Principle, Anderson and May (1991) λ(a) = D 0 β(a, a )λ(a )X(a )da λ(a) = rate at which susceptibles acquire infection FOI β(a, a ) = transmission rate, i.e. per capita rate at which an individual of age a makes an effective contact with a person of age a, per year X(a ) = number of susceptibles of age a D = mean duration of infectiousness

8 The Mass Action Principle Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates Mass Action Principle, Anderson and May (1991) λ(a) = D 0 β(a, a )λ(a )X(a )da λ(a) = rate at which susceptibles acquire infection FOI β(a, a ) = transmission rate, i.e. per capita rate at which an individual of age a makes an effective contact with a person of age a, per year X(a ) = number of susceptibles of age a D = mean duration of infectiousness

9 Social Contact Hypothesis Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates Assume transmission rates are proportional to rates of making a certain type of contact: Social Contact Hypothesis q(a, a ) proportionality factor β(a, a ) c(a, a ) contact rate

10 Social Contact Hypothesis Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates Assume transmission rates are proportional to rates of making a certain type of contact: Social Contact Hypothesis β(a, a ) q(a, a ) c(a, a ) proportionality factor contact rate estimation serological data social contact data

11 Contact Surveys Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates multi-country population-based survey in Europe (POLYMOD) diary-based questionnaires, May September 2006 non-close contact: two-way conversation of at least three words in each others proximity close contact: involving any sort of physical skin-to-skin touching diary weights based on age and household size Recruitment # participants # contacts # close > 15 min BE random digit dialling (44%) FI population registers (38%) GB face-to-face interview (42%) IT random digit dialling (47%) PL face-to-face interview (49%)

12 Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates Estimating Mean Number of Contacts Y ij = number of contacts in age class j during one day as reported by a respondent in age class i model: Y ij NegBin(m ij, k)

13 Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates Estimating Mean Number of Contacts Y ij = number of contacts in age class j during one day as reported by a respondent in age class i model: Y ij NegBin(m ij, k) m ij estimated using bivariate smoothing approach: two-dimensional continuous function over age of respondent and age of contact ( contact surface ) tensor-product spline basis: log( ˆm ij ) = K l=1 p=1 K ˆδ lp b l (a [i] )d p (a [j] ), basis dimension K, parameters δ lp, known basis functions b l and d p for thin plate regression splines

14 Estimating Contact Rates Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates Smooth-then-constrain reciprocal nature of contacts taken into account: m ij w i = m ji w j w i denotes population size in age class i (demography)

15 Estimating Contact Rates Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates Smooth-then-constrain reciprocal nature of contacts taken into account: m ij w i = m ji w j w i denotes population size in age class i (demography) Focus on type of contact with high transmission potential for parvovirus B19: close contacts longer than 15 minutes K = 13 seems satisfactory basis dimension choice

16 Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates Estimated Daily Contact Rates ˆm ij /w j BE EW FI age contact age contact age contact age participant age participant age participant IT PL age contact age contact age participant age participant

17 Estimating Transmission Rates Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates IgG antibodies against parvovirus B19 are thought to induce lifelong immunity MSIR-model

18 Estimating Transmission Rates Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates IgG antibodies against parvovirus B19 are thought to induce lifelong immunity Further assumptions: time equilibrium closed population of size N no mortality due to infection type I mortality age L MSIR-model type I maternal antibodies age A

19 Estimating q(a, a ), R 0 and v Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates π(a) = { 1 if a < A 1 exp ( a A λ(s)ds) if a A,

20 Estimating q(a, a ), R 0 and v Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates π(a) = { 1 if a < A 1 exp ( a A λ(s)ds) if a A, λ(a) = ND L L A q(a, a )ĉ(a, a )λ(a ) exp ( a A λ(s)ds ) da, if a A.

21 Estimating q(a, a ), R 0 and v Transmission Dynamics Estimating Contact Rates Estimating Transmission Rates π(a) = { 1 if a < A 1 exp ( a A λ(s)ds) if a A, λ(a) = ND L L A q(a, a )ĉ(a, a )λ(a ) exp ( a A λ(s)ds ) da, if a A. move to discrete age framework solve equations iteratively estimate q(a, a ) from serological data using maximum likelihood estimate: basic reproduction number R 0 critical immunization level v

22 Application Settings 1-year age intervals: [A, 1), [1, 2),..., [L 1, L) Demographical parameters: Country Year L N BE EW FI IT PL Age of waning maternal antibodies A = 0.5 years Mean duration of infectiousness D = 6/365 years

23 Two Parametric Models for q(a, a ) Two parametric models for q(a, a ): constant proportionality: q(a, a ) = q age-dependent proportionality: log{q(a, a )} = log{q(a)} = γ 0 + γ 1 a allows for age-specific differences in characteristics related to susceptibility

24 Results Country ˆq 95% CI for q R0 ˆv AIC BIC BE [0.064, 0.069] EW [0.057, 0.061] FI [0.052, 0.055] IT [0.026, 0.028] PL [0.047, 0.050]

25 Results Country ˆq 95% CI for q R0 ˆv AIC BIC BE [0.064, 0.069] EW [0.057, 0.061] FI [0.052, 0.055] IT [0.026, 0.028] PL [0.047, 0.050]

26 Results Country ˆq 95% CI for q R0 ˆv AIC BIC BE [0.064, 0.069] EW [0.057, 0.061] FI [0.052, 0.055] IT [0.026, 0.028] PL [0.047, 0.050] Country Parameter 95% CI R0 ˆv AIC BIC BE ˆγ [-2.321, ] ˆγ [-0.050, ] EW ˆγ [-2.489, ] ˆγ [-0.069, ] FI ˆγ [-3.117, ] ˆγ [-0.004, 0.014] IT ˆγ [-3.312, ] ˆγ [-0.059, ] PL ˆγ [-2.674, ] ˆγ [-0.091, ]

27 Results Country ˆq 95% CI for q R0 ˆv AIC BIC BE [0.064, 0.069] EW [0.057, 0.061] FI [0.052, 0.055] IT [0.026, 0.028] PL [0.047, 0.050] Country Parameter 95% CI R0 ˆv AIC BIC BE ˆγ [-2.321, ] ˆγ [-0.050, ] EW ˆγ [-2.489, ] ˆγ [-0.069, ] FI ˆγ [-3.117, ] ˆγ [-0.004, 0.014] IT ˆγ [-3.312, ] ˆγ [-0.059, ] PL ˆγ [-2.674, ] ˆγ [-0.091, ]

28 Results Country ˆq 95% CI for q R0 ˆv AIC BIC BE [0.064, 0.069] EW [0.057, 0.061] FI [0.052, 0.055] IT [0.026, 0.028] PL [0.047, 0.050] Country Parameter 95% CI R0 ˆv AIC BIC BE ˆγ [-2.321, ] ˆγ [-0.050, ] EW ˆγ [-2.489, ] ˆγ [-0.069, ] FI ˆγ [-3.117, ] ˆγ [-0.004, 0.014] IT ˆγ [-3.312, ] ˆγ [-0.059, ] PL ˆγ [-2.674, ] ˆγ [-0.091, ]

29 Estimated Prevalence and FOI for Belgium BE BE prevalence force of infection prevalence force of infection age age q(a, a ) = q q(a, a ) = exp(γ 0 + γ 1 a)

30 Estimated Prevalence and FOI for England & Wales EW EW prevalence force of infection prevalence force of infection age age q(a, a ) = q q(a, a ) = exp(γ 0 + γ 1 a)

31 Estimated Prevalence and FOI for Finland FI FI prevalence force of infection prevalence force of infection age age q(a, a ) = q q(a, a ) = exp(γ 0 + γ 1 a)

32 Estimated Prevalence and FOI for Italy IT IT prevalence force of infection prevalence force of infection age age q(a, a ) = q q(a, a ) = exp(γ 0 + γ 1 a)

33 Estimated Prevalence and FOI for Poland PL PL prevalence force of infection prevalence force of infection age age q(a, a ) = q q(a, a ) = exp(γ 0 + γ 1 a)

34 ˆq(a) as a function of a ˆq(a) = exp(ˆγ 0 + ˆγ 1 a) q(a) BE EW FI IT PL Italy has an overall smaller proportionality factor Interpretation for Italy? lower effectiveness rate compared with other countries (climate... )? overreporting close contacts?...? a

35 ˆq(a) as a function of a ˆq(a) = exp(ˆγ 0 + ˆγ 1 a) q(a) BE EW FI IT PL a Finland: no need for an additional age-dependent factor Interpretation for other countries? children are more susceptible than adults? underreporting of close contacts in children?...?

36 ˆq(a) as a function of a ˆq(a) = exp(ˆγ 0 + ˆγ 1 a) q(a) BE EW FI IT PL a Finland: no need for an additional age-dependent factor Other factors involved? Country FI PL EW IT BE Effective parental leave 99 weeks 53 weeks 25 weeks 24 weeks 18 weeks Source: Plantenga J. & Siegel M. (2004), Position Paper childcare in a changing world, RuG, The Netherlands

37 Conclusion and Discussion Social contact surveys are useful to gain more insight in the transmission process of airborne infections For BE, EW, IT and PL there is need for an additional age-specific factor to explain the observed seroprofile for parvovirus B19 Country-specific differences in B19 transmission are observed (How) can we interpret the age-dependent proportionality factor and country-specific differences?

38 Discussion and Further Research Formal inference on country-specific differences joint modeling Other functional forms for proportionality factor: gamma, normal, Weibull,... Allow proportionality factor to vary with age of infected a : bivariate model for q(a, a ) sensible? serological data don t provide direct information related to infectiousness Other work: analyzing pre-vaccination data on a range of different airborne infections from EW

39 Acknowledgements ECDC funding support SBO-project SIMID funded by the IWT-institute in Flanders all co-authors