Need our growing knowledge of the human genome and its role in disease susceptibility spawn a 'genetic underclass' forced to pay crippling insurance terms or refused insurance altogether? As far as life insurance goes, apparently not, according to mathematician Dr Angus McDonald of Herriot-Watt University, Edinburgh, UK, who models genetic risk, although health and income-protection insurance could be a different matter.

McDonald has been studying the possible insurance costs connected with the link between Alzheimer's disease and certain variants, or alleles, of genes for proteins called apolipoproteins, connected with fat transport around the body. McDonald feels that there is a need for a lot more mathematical research before the ethics, economics and policy making surrounding this issue can be reconciled. "Genetic tests are likely to have a much smaller impact on insurance premiums than is commonly believed; possibly no impact at all in most cases," he said at the British Association Festival of Science in Sheffield, UK. However, he did add that for about 2% of the population, the presence of one of the apolipoprotein alleles associated with Alzheimer's disease would make long-term insurance care between 10% and 30% more expensive.

Ironically, many of the issues thrown up by today's genetic discoveries have, as far as insurers are concerned, been known for decades. Namely, that there are conditions that run in families. Thus the 'genetic prejudice' suffered by those with many of the single gene conditions such as Huntington's disease and familial breast cancer is really nothing new.

However, in mathematical terms, McDonald assures us, these special cases are not necessarily relevant to good risk prediction for similar ailments in the general population.

So what of the myriad multifactorial disorders whose precise genetic origins research will soon resolve? "These may only be responsible for a 10% extra chance of dying every year," McDonald comments, "and this kind of figure would be lost in the noise of insurance underwriting [as their] outcomes are eminently modifiable by disease and lifestyle". This makes sense in the light of the fact that at present, those with a medical condition that makes them 30-50% more likely to die, year on year, are only in the borderline category for an increased premium.

The mathematics used to make these predictions is a set of probabilistic tools known as 'Markov multiple state models', that are also commonly used in engineering and epidemiology. These are based on representing life spans in terms of states and transitions between those states. The simplest example of this, then, is a two-state model: life and death. A slightly more complex one might be an Alzheimer's model which divides life up into several more states such as health, early disease at home, severe disease needing full-time care or hospitalization, and then death.

Herriot-Watt University is currently setting up a centre for genetics and insurance research. One of the first things they plan to do is a statistical review of critical-illness insurance and income-protection insurance which are more sensitive to genetic risk but have not, as yet, been mathematically modelled.