In order to make accurate projections, epidemiologists need to take into account the following variables. They are closely interrelated and it is not possible to single out one as more important than the others.

Virulence of HIV and infectivity of HIV-infected persons

It is likely that individuals are most infectious in the six to eight weeks after infection until antibodies are produced to suppress HIV in the blood, semen and vaginal and cervical fluid. Thereafter very low levels of HIV are present in these body fluids for much of the time that individuals are asymptomatic. Virtually all the viral burden is confined to the lymph tissue (the lymph nodes, the spleen etc.). Increasing viral load tends to correspond with the development of symptoms, and individuals are once again highly infectious. For example, one study has shown that semen is most likely to contain HIV in individuals with CD4 counts below 100, and it was very difficult to find HIV in the semen of men with CD4 counts above 200 (for further discussion of factors which influence infectivity see HIV transmission).

These fluctuations of infectivity may help to explain the shape of the epidemic. For instance, reductions in viral load in infected gay men, together with the effectiveness of community–based safer sex education, probably account for the apparent fall–off in new infections amongst gay men in the mid–1980s in Europe and North America.

Because the periods of high infectivity are relatively short, the levels of unprotected sex or needle sharing with different partners (and the choice and location of those partners) during those periods are very important in determining how many people become infected and where the virus spreads. This explains why gay men and injecting drug users have been seriously affected; members of these groups with large numbers of contacts are more likely to come into contact with people who are going through a period of high infectiousness.

This is not to say that HIV is not passed on at other times by HIV–positive people. But the current distribution of HIV infection in the UK and other countries in the developed world does suggest that much HIV infection can be accounted for by contacts which take place during those periods of high infectivity. If HIV–positive people were very infectious much of the time, the virus would have spread even more widely in the core groups, far more frequently to contacts of the core groups, and far more frequently to their other partners in turn.

The frequency of partner change amongst different groups in the population:

As noted above, the more contacts take place when an individual is infectious, the more infections will result. If there is a low level of partner change this will tend to minimise the spread of HIV. Levels of partner change amongst heterosexuals have been overestimated in the past.

Extent of mixing between different groups

If there is little mixing between bisexual men and female partners, there will be little spread of HIV by this route. But if many bisexual men have large numbers of female partners whilst highly infectious, HIV will be widely disseminated amongst women. The same is true with injecting drug users, and with people from high prevalence locations.

Extent of risk-taking behaviour amongst different groups

Even if many injecting drug users in a locality are infected with HIV, there will be little HIV transmission unless needle-sharing takes place. HIV will not spread if a large proportion of individuals have adopted behaviour which minimises the risk of infection, or if they don't engage in risk behaviour in the first place.

Prevalence of HIV in different groups and locations

Glasgow is a good example of a city which has been insulated from an HIV epidemic amongst injecting drug users by a low initial prevalence of HIV amongst injecting drug users and the adoption of behaviour which minimises the risk of HIV infection.

The levels of HIV infection amongst migrants to the UK from Africa and the European Union

This has turned out to be the biggest single influence on HIV diagnoses in the last five years. The largest proportion of AIDS and HIV cases amongst the heterosexual population is found in people who are migrants to the UK. If increasing numbers of people entering the UK are HIV-positive, this will affect the size of the UK epidemic.

The effect of treatment more alive, but less infectious

Improvements in treatment mean that people with HIV will live longer without symptoms, and that people will live longer with AIDS. Antiretroviral treatment which reduces viral load will play an especially significant role in limiting transmission, rather than increasing the potential `pool' of infectious persons by keeping people alive.

Underestimates of the size of the key risk groups

Predictions are partly based on estimates of the frequency of homosexual and drug injecting behaviour, yet no one really knows how frequently these things happen. Epidemiologists have to use indirect indicators such as the incidence of hepatitis B infection amongst injecting drug users, the incidence of rectal gonorrhoea amongst gay and bisexual men and the size of the population of GUM clinic attenders as surrogates for measuring the level of risk-taking behaviour.

The emergence of increasingly virulent strains of HIV

Although this variable is unlikely, we cannot predict with any certainty how HIV will evolve. It may become more infectious or more deadly, or less so.

Understanding of the UK HIV epidemic relies on being able to establish how the individuals reported to the data set probably acquired infection, and if the route is heterosexual, how or where they or their partners are likely to have become infected. All reports of new diagnoses of HIV infection that are received with insufficient information to ascertain the probable route of infection are subject to follow-up. Where necessary this follow-up may include interview by a research nurse, subject to the agreement of the diagnosing clinician and patient. The time taken to complete this process means that the proportion of reports for which infection route is unresolved is higher for recent time periods than for the data set overall.