Glaucoma risk factors

The next step is to establish whether the identified risk factor is a causal risk factor that leads directly to the disease. If the risk factor is causal, then removing that risk factor should reduce or remove the disease itself, said Dr Healey.

“Risk factors are not always causal, however. They may be surrogates for true causal risk factors that we have not been able to discover or do not have the technology to discover, or they could be epi-phenomena on the causal pathway. These are reliable features that are strongly associated with the glaucoma, but the removal of which does not reduce the risk of disease progression†he said.

Proving a causal risk factor is a complex task, with the first step being to rule out biases or possible confounders, said Dr Healey. This needs robust study design and standardised statistical methodology.

“It really requires a number of high-quality studies that are well conducted and independent from each other that show similar risk factors and a similar magnitude of risk. A good example is IOP, which has been very well described and its association with glaucoma by many independent studies,†he said.

The next step is to identify a plausible biological link between the risk factor and the disease, said Dr Healey.

“This is the role of much of the basic laboratory science, to describe how what we find in a clinical or epidemiological study is linked to the disease, and it usually entails understanding the cellular mechanisms involved. We also need to show a reduction in disease progression in people not exposed to the risk factor and that of course requires randomised controlled interventional trials or sometimes cohort studies,†he said.

Other factors which need to be borne in mind when identifying risk factors are the varying definitions of glaucoma across different studies, taking account of whether the study was looking at early onset or more advanced disease, and whether patient groups had diagnosed or undiagnosed glaucoma, said Dr Healey.

While multivariate modelling can help in adjusting the correlation to risk factors, some interactions may still persist, said Dr Healey.

“A perfect example of that are vascular risk factors. There are so many different measurements that can be made relating to vascular risk, heart attack, stroke status, systolic blood pressure, perfusion pressure, mean blood pressure and so forth, that it is very hard to tease out which one or group of these is actually more important,†he said.