By Jodie Glenn
What is this paper about/background?
Alzheimer’s disease (AD) primarily affects memory, thinking and behaviour. The disease progresses over time and that progression is very serious as AD is the leading cause of dementia. This progression begins decades before diagnosis and so it would be very beneficial to be able to predict the risk of AD in young populations to support preventative healthcare. This study is a baby step towards this as the average age of the people participating in this study was 56 years old which is significantly younger than other study groups. Another unique part of this study is that cognitive ability scores were available for all participants from when they were around 20 years old and so could be included as input in the study’s model. This is useful because usually education is used as an approximate for cognitive ability but this is unreliable. This study sort to determine if a model using what we know on the genetic risk for AD as the predicator variable can be used to detect early cognitive impairment before the typical age at which AD even develops?
So, how did the researchers do this?
This study used a genetic technique called, polygenic risk scoring (PRS). To do this, they looked at specific locations in the genes of people that harbour AD-associated risk variants (deviations from the “healthy” DNA). These risk variants have different weightings based on how much they contribute to the genetic component of the disease. PRS takes this data along with an individual’s specific DNA and calculates their genetic liability to a trait. To use an analogy, PRS is guessing how much water a person has in their bucket based on how many drops of water (risk variants) they have and how big each drop of water is (the risk variant’s weighting). The results are a spectrum and ancestry is very important in making the guessing accurate. In this case, European ancestry data was used. The risk scores were used to predict those with mild cognitive impairment (MCI) which is related to AD but importantly does not guarantee development of AD.
What were their results?
Importantly all these analyses are summaries of using risk variants for AD that have different levels of evidence. The aMCI/naMCI vs CN odds ratio compares the odds of having MCI with being cognitively normal (CN) for people with a higher AD PRS. There is a lot of nuance to the results of the study which is captured in even more complex statistics but the main finding is that people with higher PRSs were more likely to have amnestic MCI than cognitively normal individuals. The trend was similar when predicting individuals with non-amnestic MCI (naMCI) but there weren’t enough participants with naMCI to be confident that the association was real. The value of this result is that PRS improved prediction for MCI beyond risk factors like advanced age, the APOE gene and head injuries. The model used in this study was good at classifying people as cognitively normal who were actually cognitively normal but had a moderate positive predictive value (PPV) meaning that some people classified as having MCI were actually cognitively normal (false positives).The model could be refined to avoid false positives but would mean losing sensitivity. Making this adjustment could potentially make the model useful for recruitment in clinical trials.
What should we do next?
The study doesn’t know which participants will actually have AD because they didn’t keep checking on the participants over many years. So it’s a good idea to maybe use the same AD PRS method in a study of people over many decades.
Reference:
Logue, M.W., Panizzon, M.S., Elman, J.A., Gustavson, D.E., Andreassen, O.A., Gillespie, N.A et al. (2019) Use of an Alzheimer’s disease polygenic risk score to identify mild cognitive impairment in adults in their 50s. Molecular Psychiatry, 24(3), pp.421–430. https://doi.org/10.1038/s41380–018–0030–8
Leave a Reply