Effects of heterogeneity in hosts and pathogens on effectiveness of vaccination Mirjam Kretzschmar RIVM, Department of Infectious Diseases Epidemiology The Netherlands Populations are heterogeneous ... Why do we have to think about heterogeneity? Measles outbreak (almost 3000 cases) despite coverage of 96% Host heterogeneity • Disease independent (can be measured also for non-infected individuals): – Age, sex, other demographic variables – Behaviour (e.g. number of contacts, compliance with vaccination) • Disease dependent (only for infected individuals): – Transmission route – Disease stage; primary versus secondary infection – Clininal symptoms or asymptomatic Pathogen heterogeneity • Heterogeneity between strains: – Virulence (defined as host mortality or severity of disease) – Vulnarability to host immune response – Competition via cross-immunity • Within host heterogeneity: – Immunogenic variability (HIV) – Different location within host leads to different effects (invasive infection versus carrier) Effects of heterogeneity on vaccination depend on vaccination strategy • Universal vaccination – Rationale: create herd immunity such that unvaccinated individuals are also protected • pockets of not vaccinated persons (MMR vaccination in the Netherlands) • core groups of individuals with very many contacts (STD, Hepatitis B) • non-homogeneous contact patterns, i.e. household contacts, spatial patterns, • repeated contacts with same individuals (long term partnerships, networks) • Targetted vaccination of risk groups – Rationale: protect those individuals who are at greatest risk for being infected • Takes heterogeneity in risk into account, but what are effects of mixing between risk groups? • Critical coverage per risk group? • Effects of vaccination on non-risk groups? • Ring vaccination – Rationale: vaccinate direct contacts of infected individuals to interrupt transmission chain • Heterogeneity in contact patterns is taken into account, only persons at risk are included in vaccination. • But: how to estimate fraction of contacts that have been found and vaccinated? • Modelling: Contact tracing requires a network modelling approach Contact patterns: who has contact with whom? Contact and transmission route • Influenza (airborne infection): – talking with each other at close distance – coughing at each other • Gonorrhoea (sexually transmitted dis.): – sexual intercourse • Hepatitis C (bloodborne infection): – sharing contaminated needles – blood transfusion Knowledge about contact patterns leads to insight into transmission routes • Contact network AIDS cases (Auerbach et al. 1984) – Probability that cluster of cases is connected by contact on the basis of random events – timing of contacts and onset of disease • Hypothesis: AIDS is transmitted by homosexual contact Cluster of AIDS patients number: order of diagnosis 0 index case A.S. Klovdahl. Social networks and the spread of infectious diseases: The AIDS example. Soc. Sci. Med. 21 (1985): 1203-1216. Contacts are non-random • Population heterogeneity – Age structure, social economic structure, education • Social grouping – Families, working environment, recreation • Geographical distribution – Cities, rural areas, mobility between regions People are not the same and they choose contacts with certain preferences these choices influence the way infectious diseases spread Influence of contact patterns on epidemiological outcome • • • • Age distribution of cases in STDs for men and women Biannual measles epidemics in prevaccination era High prevalence of STDs in high activity core groups Widespread heterosexual transmission of HIV in subSaharan Africa • Hepatitis A outbreaks in day care centers • Increasing incidence of HIV in monogamous married women in Thailand • Increasing incidence of malaria in Western Europe Modelling heterogeneity • Heterogeneity in number of contacts – Core groups – Stratification by activity – Mixing? • Local/global contacts – Households – Metapopulation models • Partnership duration: pair formation models, pair approximation models • Networks Vaccination in a population stratified by households local contacts global contacts Equalizing strategy: Choose individuals for vaccination sequentially from those households with largest number of susceptibles. Minimizes the number of vaccinations needed to reduce R to below 1. Ball, Mollison & Scalia-Tomba.Ann. Appl. Prob. 7 (1997) 46. The basic reproduction number R0 The expected number of secondary cases caused by one index case during his entire infectious period in a completely susceptible population. homogeneous population: R0=cD heterogeneous population: number of secondary cases has to be averaged in the right way. Heterogeneous population Diekmann, Heesterbeek, Metz. J. Math. Biol. 1990; 28:365-382 Diekmann, Heesterbeek. Mathematical Epidemiology of Infectious Diseases, Wiley, 2000. Next generation operator Number of cases in the (n+1)-th generation of infections given the distribution of infectious individuals (with respect to population structure) in the n-th generation. Basic reproduction number Dominant eigenvalue of the next-generation operator Explicit calculation of R0 for separable mixing Contact funtion c(a,b)=f(a)g(b) Host heterogeneity: example Hepatitis B vaccination • Background: – Introduction of universal infant vaccination in the Netherlands? – Low prevalence, high costs of vaccination – How many cases of chronic hepatitis B infection can potentially be prevented? • Project including case-control study, modelling and cost-effectiveness analysis Hepatitis B: many types of heterogeneity • Transmission routes: – Sexual transmission – Vertical to babies at birth – Horizontal close contact (household) • Age: – Age dependent immune response (clinical symptoms and development of chronic carrier state) – Age dependent sexual activity level • Behaviour: – High versus low activity within age groups • Disease states: – Latent (1-2 months), acute (3-4 months), and chronic stages Model structure Williams et al. (1996), Epidemiol & Infect. 116: 71-89 Kretzschmar et al. (2002) Epidemiol & Infect. 128: 229-244. • Population stratified by age and sexual activity (6 activity classes) • Two transmission routes (vertical and sexual) • Different stages of infection (acute, chronic carrier) susceptible latent acute vaccinated carrier immune Model • System of partial differential equations (age structure) • Proportionate mixing • Separate models for hetero/homosexual populations • Included immigration and age dependence in probability to become carrier • Explicit formula for R0 Calculation of R0 Individuals can be infected via two routes. R0 is the dominant eigenvalue of next generation matrix R R ss 0 vs 0 s v sexual transmission vertical transmission sv 0 vv 0 R R Calculation of R0ss 6 L R0ss k 1 0 with k ck (a) La LK c (a )( P ( ) P ( , a))dda k 1 Y 2 C 0 6 L LK k ck (a)da k 1 0 L k ck(a) i maximum lifetime fraction in activity class k =1,...6 age dependent contact rate in activity class k time since infection transmission probability per partnership Age-dependence of becoming carrier PC(,a) has factor p(a), the probability of becoming carrier when infected at age a 0 for a=0 p(a) 2 exp( 1a ) for a>0 Edmunds et al. 1993: Point estimate of parameters 1 and 2 from data from 29 different studies Probability of becoming carrier after infection 1 ml estimate R0 minimal in 95% conf region R0 maximal in 95% conf region observations probability 0.8 0.6 0.4 0.2 0 0 10 20 age (years) 30 Partner change rates from sexual behaviour surveys (UK and NL) new partners per year 1,2 heterosexual homosexual 1 0,8 0,6 0,4 0,2 0 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 age Estimates for R0 (heterosexual population) constant age-dependent R0 1 0 1.11 0.79 estimates UK 0.69 0.53 estimates NL Estimate R0 • Homosexual men R0>1: – Hepatitis B virus can persist – Immigration of infected persons has little influence • Heterosexual population R0<1: – short transmission chains – immigration of infected persons determines prevalence Compare with data • Case control study: – heterosexual cases (N=41): 60% of cases infected by immigrant from high endemic country – homosexual cases (N=44): 16% infected by immigrant from medium or high endemic country Conclusions for vaccination • Vaccinating general population can reduce incidence of new infections within the country, but has little influence on prevalence of carriers. • Vaccination of risk groups is being intensified • Vaccination is offered to children of whom at least one parent is an immigrant from country with higher prevalence Vertical transmission • In highly endemic countries it is believed that vertical transmission and horizontal transmission to children are the most important transmission routes. • In low endemic countries the role of horizontal transmission to children is not known. • Can we use R0 to analyse importance of those transmission routes? • Assume sexual behaviour comparable to UK data vv Consider R0 L R0vv ( )(b1PY ( ) b2 PC ( ,0)) d 0 with () bi fertility rate at age transmission probabilities per offspring (b1=0.724, b2=0.115) Fertility distributions 0.14 0.12 UK NL live births / woman / year 0.1 0.08 0.06 0.04 0.02 0 0-5 5-10 10-15 15-20 20-25 25-30 30-35 Age (years) 35-40 40-45 45-50 50-55 Horizontal transmission • Horizontal transmission in households can be approximately described by increasing bi • The fertility function can vary in age distribution and total number of offspring during lifetime. • Example: mean offspring number 3, b2=0.5 R0ss vs R 0 sv 0 vv 0 R R 0.75 0.05 4.17 0.89 R0 = 1.29 Neither of the transmission modes alone could sustain endemic prevalence, together they can Horizontal transmission as main transmission mode? 0.9 transmission probability b 2 0.8 0.7 Rvv>1 0.6 0.5 R0>1, 0.4 Rvv<1 0.3 R0<1 0.2 0.1 0 0 1 2 3 4 mean offspring number 5 6 7 Conclusions • Explicit expression for R0 in heterogeneous populations can help to get insight into influence of different types of heterogeneity on transmission dynamics and their interaction • Drawback: proportionate mixing assumption • How does R0 depend on underlying model assumptions? • How well does R0 reflect heterogeneity? • Hepatitis B: different vaccination strategies depending on population heterogeneity? Heterogeneity in the pathogen population and vaccination • When can serotype replacement occur? • Indirect effects of serotype replacement: partial immunisation by replacing strains? • Optimal composition of vaccine (trade-off between breadth and effectiveness)? • Evolution to higher virulence? • Vaccination against disease or against infection? Competition of 2 strains Model McLean: Assumptions: – 2 strains, total cross-immunity – Vaccinated individuals can become infected with a small probability, vaccine efficacy differs between strains – after infection permanent immunity – Strain 1 outcompetes 2 in absence of vaccination A.R. McLean. Proc R Soc Lond B (1995) 261: 389-393. McLean model vaccination birth S transmission V 2I2S death 1I1S I1 death (1-r)2I2V (1s)1I1V recovery + death I2 Effects of vaccination • Vaccination reduces competitive pressure on weaker strain 2 -> outbreaks • indirect effect: more herd immunity against strain 1 Dynamicsof 2 competing strains 500 400 300 200 100 0 20 40 60 80 100 Superinfection Model Lipsitch: Assumptions: – no immunity, after recovery susceptible again – Individual can be infected by 2 strains simultaneously – Cross-immunity – vaccine 100% effectiv for target strain M. Lipsitch. Emerging Infectious Diseases (1999) 5: 336-345 Model with superinfection birth death transmission recovery S vaccination V 2(I2+I12)S 1(I1+I1v+I12)S (1s)1(I1+I1v+I12)V I2 I1 c22(I2+I12)I1 c11(I1+I1v+I12)I2 aI1I2 I12 I1v I1v Effects of vaccination • Vaccination enables coexistence of strains • serotype replacement can occur • If vaccine is also effective for other than the target strains, higher coverage is needed for eradication Model Lipsitch 1 prevalence 0.8 0.6 0.4 0.2 15 20 25 HL 30 time years 35 40 45 50 Example pertussis • Since middle of the 90‘s increase in incidence of pertussis in NL • Increase in incidence among vaccinated children • Large incidence of subclinical infections in adults • Hypothesis: vaccine not as effective against presently circulating strains Model with 2 strains Birth Birth I1 S J1 R I2 red = transmission green = recovery blue = loss of immunity W V J2 Assumptions • Full cross immunity after natural infection • vaccine protects fully against strain 1, partly against strain 2 • vaccine induced immunity lasts shorter than natural immunity Equilibrium under vaccination primary infection more transmissible, R01>R02 0.005 I1 0.004 I2 J1 J2 prevalence 0.003 total 0.002 0.001 0 0 0.2 0.4 0.6 -0.001 vaccination coverage 0.8 1 Dynamics with vaccination 0.003 I1 I2 J1 J2 0.0025 prevalence 0.002 0.0015 0.001 0.0005 0 0 100 200 300 time 400 500 600 Conclusions • • • • Strains can coexist for certain range of vaccination coverage for high coverages strain 2 is dominant total prevalence of infection decreases with increasing coverage elimination for p larger than critical vaccination coverage Equilibrium under vaccination secondary infection more transmissible 0.005 0.004 I1 prevalence 0.003 I2 J1 0.002 J2 total 0.001 0 0 0.2 0.4 0.6 -0.001 vaccination coverage 0.8 1 Dynamics with vaccination 0.0045 I1 I2 J1 J2 0.004 0.0035 prevalence 0.003 0.0025 0.002 0.0015 0.001 0.0005 0 0 100 200 300 time 400 500 600 Conclusions • For high coverages coexistence of both strains • total prevalence of infection increases when 2 strains are present • infection remains present even with 100% vaccination coverage Summary • • • • • Vaccination can lead to coexistence of strains Contribution of secondary infections determines success of vaccination Even very high coverage might not suffice for elimination Changes of transmission rate of primary infections may lead to sudden shifts in prevalence Need more empirical data about secondary infections