This chart shows the relationship between the population distribution of DBP and the incidence (number of new events / unit of time) of stroke. As you can see, there is no point at which stroke suddenly becomes much more likely. Instead it increases in a smooth geometric fashion, but the higher DBP is the greater the incidence, and therefore individual risk, that stroke will occur. This isn't really very surprising and is the case with many different biological measures used n medicine.

However, the problem arises when, as a doctor, you have to make a decision to treat or not, the raised blood pressure. This is a problem because most treatments are not benign, that is, they carry a risk (of complications, side-effects, loss of quality of life, and in some cases even death). The doctor is then faced with the complex decision of determining the point at which treatment risk is less than risk from stroke. You can probably prevent almost all cases of stroke by screening all people and treating anyone with a DBP >60 mmHg. However, most of those people you treat would never have had a stroke any way. The number needed to treat (NNT) (number of people that are needed to be treated to prevent one occurrence of the problem) would be very high. The higher the NNT, the less cost-effective is the treatment. It would cost a great deal, both financially and in terms of loss of quality of life (QoL) that it would be quite unacceptable to do so. Instead, an arbitrary decision is made to choose the cut-off point at 90mmHg because thereafter risk of stroke rises increasingly steeply and only a relatively small proportion of patients would have a DBP of this value or greater.