In this session, our learning objectives are to understand how the use of vaccines may produce indirect effects in non-vaccinees. And how we can use both epidemiological analyses and mathematical models to understand these effects. Vaccines may not only protect vaccinees from disease, they may also prevent vaccinees from transmitting infection to non-vaccinees. This effect, called herd immunity, reduces the risk in non-vaccinees from the same population. In fact, if vaccine coverage is high enough, then enough people are protected in the population that the infection cannot spread at all, and hence will eventually be eliminated. The level of vaccine coverage for this to occur is called the herd immunity threshold. The herd immunity threshold depends on how transmissible an infection is. The more transmissible an infection is, the higher the herd immunity threshold. So higher vaccine coverage will need to be achieved to eliminate the infection. As you can see, the herd immunity threshold for smallpox is relatively low. Which helped to make it the first disease to be eradicated through a vaccination program. Some vaccines only have direct effects because the risk of infection does not depend on person to person transmission. Examples are vaccines against tetanus and rabies. However, most vaccines have both direct and indirect effects because they prevent transmission from the vaccinated person either to other people or to an intermediate vector like mosquitoes. If vaccination coverage is not high enough to eliminate the infection, then the force of infection, the rate type which susceptible people become infected may still decrease. As a result people who have not been vaccinated will be infected at a slower rate than before vaccination. The average age, at which infection occurs, in an vaccinated people, may hence increase. This can be beneficial, if adults experience less severe disease, than children as is the case for whooping cough or pertussis. However, sometimes this can actually be a bad thing because the serious consequences of a disease get worse with increasing age. For instance, rubella is usually a mild disease but if a pregnant is infected with rubella, then her fetus may not be born alive or it may be born with congenital abnormalities. This is called congenital rubella syndrome. If rubella vaccination is introduced but coverage is not high enough to reach the herd immunity threshold, the average age at which people get rubella may shift from early childhood to the childbearing ages. This may result in more pregnant women acquiring rubella, and more cases of congenital rubella syndrome. In Greece, a rubella vaccine was introduced in 1975 and became available in the private sector. However, vaccination was not part of a nationally funded program. So there was only partial uptake of the vaccine. As a result, the number of pregnant women who were susceptible to rubella increased leading to many cases of congenital rubella syndrome during an epidemic in 1993. Besides herd immunity, another potential indirect effect of vaccines is replacement. Replacement occurs when there is competition between related types of a pathogen to infect their host. Hence if a vaccine reduces infection by some types it can lead to an increase in infection rates by the competing types. An example is pneumococcal conjugate vaccines which protect against several serotypes of a type of bacteria called Streptococcus pneumoniae. Countries such as the United Kingdom that have introduced this vaccine have seen a decrease in disease due to the pneumococcal serotypes that are protected against by the vaccine. But they have sometimes also seen an increase due to other pneumococcal serotypes not in the vaccine. This replacement effect has diminished although not eliminated completely the total disease reduction as a result of vaccination. So how can we measure the extent of heard immunity and other ecological indirect effects in vaccinated populations? One way is by conducting a cluster randomized trial, a clinical trial in which instead of allocating individuals to be vaccinated or not, we allocate Trial clusters. These clusters may be households, villages, communities, regions, subsections of the population where we may expect disease to be transmitted. Within a cluster, we may expect the incidence of disease to fall, not only because we vaccinate some members of the cluster, But because even non-vaccinated members are protected by herd immunity. A second way is by making ecological observations. When human papilloma virus vaccination was introduced in Australia, declines of genital warts, which are caused by human papilloma virus was seen in women in the age groups that were vaccinated. But declines were also seen in heterosexual men in the same age groups even though most of them had not been vaccinated likely as a result of herd immunity. Patterns of infection transmission and interactions between pathogens and their hosts in a population can be very complex. Sometimes, mathematical models are fitted to surveillance data and used to predict how much herd immunity is likely to occur following a vaccination program. Models that take these indirect effects into account are called transmission dynamic models. These models can suggest the level of vaccine coverage needed to control and to eliminate an infection. As well as to avoid detrimental effects such as increases in the rate of adult infection for diseases like rubella. They can also indicate which are the best groups to target with vaccination. For instance, Dr Marc Baguelin and his coworkers recently used transmission dynamic models fitted to many seasons of influenza surveillance data from the United Kingdom to suggest that vaccinating children against seasonal influenza would have a large impact on mortality among people of all ages, even though influenza mortality in children is lower than in some other age groups. However, models usually require good data on the rates of infection and disease in different age groups, as well as the rate at which different people mix with each other. Building a mathematical model is not a substitute for adequate data collection.