Blood test for Alzheimer’s 'no better than coin toss'
“Research in more than 1,000 people has identified a set of proteins in the blood which can predict the start of the dementia with 87% accuracy,” BBC News reports.
The primary goal of the test was to predict whether people with mild cognitive impairments (usually age-related memory problems) would go on to develop “full-blown” Alzheimer’s disease over approximately a year.
There is currently no cure for Alzheimer’s, so people may question whether an early warning system for the disease is of any practical use.
However, having a relatively reliable method of identifying high-risk people who will develop Alzheimer’s could be useful in recruiting suitable candidates for clinical trials investigating future treatments.
An important point is that, while the test accuracy rate of 87% sounds impressive, it may be a large overestimate of what would happen in reality.
Given real world assumptions on the proportion of people who have mild cognitive impairment that progress to Alzheimer’s disease (10-15%), the predictive ability of the test falls to around 50% – no better than a coin toss.
Furthermore, the test only made predictions from around a year earlier.
Consequently, on its own, this test is unlikely to be much good. However, refining this test and combining it with other methods (such as a lipid test we discussed in March) might improve accuracy rates, making it a viable predictive tool in the future.
What is Alzheimer’s disease?
Alzheimer’s disease is the most common type of dementia, affecting almost 500,000 people in the UK. The term “dementia” describes a loss of mental ability associated with the gradual death of brain cells. Alzheimer’s disease is a progressive condition, which means the symptoms develop gradually and become more severe over the course of several years.
The first sign of Alzheimer’s disease is usually minor memory problems. This could be forgetting recent conversations or events, and the names of places and objects.
As the condition develops, memory problems become more severe, and further symptoms can develop.
For more information, visit the NHS Dementia Guide.
Where did the story come from?
The study was led by researchers from Kings College London and was funded by the Medical Research Council, Alzheimer’s Research, The National Institute for Health Research (NIHR) Biomedical Research Centre and various European Union (EU) grants.
Some of the researchers reported potential conflicts of interest, as they had patents filed with, or work for, Proteome Sciences plc. Proteome Sciences is a life sciences company with a commercial interest in biomarking testing. Another researcher works for the pharmaceutical company GlaxoSmithKline (GSK). No other conflicts of interest were reported.
The media coverage was broadly accurate, but none reported the positive predictive value of the test. This reduces the impressive sounding 87% accurate figure to around the 50% level, depending on the prevalence assumptions, giving the test the same predictive value as a coin toss.
This important information should have been highlighted to avoid overstating the utility of the test on its own.
What kind of research was this?
This study used information from three existing cohorts of people, to study the prognostic value of a new blood test in predicting people’s progress from mild cognitive impairment to Alzheimer’s disease.
There are currently no drug treatments that cure Alzheimer’s, although there are some that can improve symptoms or temporarily slow down progression of the disease in some people.
Some believe that many new clinical trials fail because drugs are given too late in the disease process.
A blood test could be used to identify patients in the early stages of memory loss, who could then be used in clinical trials to find drugs to halt the disease’s progression.
What did the research involve?
The researchers studied the blood plasma of 1148 elderly people – 476 with clinically diagnoses of Alzheimer’s disease, 220 with mild cognitive impairment (a mild form of dementia) and 452 with no signs of dementia. They then studied how differences in proteins correlated with disease progression and severity over a period of between one and three years.
Diagnosis of Alzheimer’s disease was made using established criteria, but three groups were used and combined, so the diagnosis tool used in each was actually different.
Other standardised clinical assessment included the Mini-Mental State Examination (MMSE) for measuring general cognition and cognitive decline, as well as the Clinical Dementia Rating (ANM and KHP-DCR only) for measuring dementia severity.
Participants’ brains were also scanned using an MRI scanner, to measure the volume and thickness of the brain to look for further signs of Alzheimer’s or brain deterioration.
The researchers started with 26 candidate proteins they thought might be useful to predict progression and severity. These were tested in different combinations and reduced to the best 10, based on specificity and sensitivity of the test.
What were the basic results?
The team identified 16 proteins in participants’ blood that correlated with disease severity and cognitive decline.
The strongest associations predicting progression from mild cognitive impairment to Alzheimer’s disease were formed of a panel of 10 proteins. Depending on different threshold inputs, this test had an accuracy of between 72.7% and 87.2%, and a positive predictive value of between 47.8% and 57.1%.
The positive predictive value of a test is the proportion of positive and negative results that are true positive and true negative results. That is the test’s ability to both correctly identify people with a specific condition, and not misdiagnose people who don’t have the condition.
The accuracy of the protein test was improved when it was combined with a test for gene variant associated with increased amyloid protein in the brain (APOE ε4 allele).
This combined test predicted progression from mild cognitive impairment to Alzheimer’s disease over a year, with an accuracy of 87% (sensitivity 85%, and specificity 88% – the positive predictive value not reported in the publication).
Figures from the Alzheimer’s Society estimate that between 10% and 15% of people with mild cognitive impairment progress to Alzheimer’s disease each year. Based on this assumption, the test has a positive predictive value of between 44% and 56%. This means that the combined test will only identify people correctly in around half of all tests, and potentially less, similar to the predictive ability of flipping a coin.
The average time for mild cognitive impairment to develop into Alzheimer’s in the study was around one year.
How did the researchers interpret the results?
The study authors concluded they had, “identified 10 plasma proteins strongly associated with disease severity and disease progression” and that, “such markers may be useful for patient selection for clinical trials and assessment of patients with predisease subjective memory complaints”.
This research developed and tested a new blood test that predicted the progression from mild cognitive impairment to Alzheimer’s disease, with an accuracy of 87% approximately a year before development.
However, in a non-experimental setting, the test may be much less effective than the 87% figure suggests. Based on figures from the Alzheimer’s society indicating that 10-15% of people progress from mild cognitive impairment to Alzheimer’s disease each year, the test is only correct around 50% of the time – no better than a coin toss. However, the test is unlikely to be used by itself, so its predictive ability may be improved if used in combination with other tests in development. The predictive ability of the test would also improve if the 10-15% assumptions turned out to be an underestimate.
A further limitation is that it only made its predictions a year in advance of Alzheimer’s diagnosis. This is certainly better than nothing, but Alzheimer’s disease is often diagnosed at a later stage, with the disease having already caused damage for many years (the exact time is variable). A test that predicted Alzheimer’s disease using a 5 or 10-year period would be a much bigger advancement.
As there is currently no cure for Alzheimer’s, there is likely to be a debate about whether patients would want to know this information if the test was successfully developed further and made available in mainstream medicine.
Some people may prefer to know their prognosis, as it may influence what they do or the way they live.
Others may prefer not to know, given that current drug treatments can only slow the progression of the disease in some people, and do not improve the symptoms in all of those affected.
However, as the researchers point out, the test has an important potential use. If confirmed to be effective in further studies, the test could be used to recruit people into clinical trials, testing new drugs or treatments to help future generations.
Promising Alzheimer’s drugs are reported to have a high failure rate in human clinical trials.
Many researchers believe this to be because by the time a person is diagnosed with Alzheimer’s, it is too late to do anything about it, with medicine unable to reverse the brain damage that has already been caused.
Therefore, scientists are looking for ways to intervene earlier.
Knowing who will likely develop Alzheimer’s in a year is a step forward in this effort, as researchers can test different drugs and treatments, and would be able to see if they are preventing the progression from mild cognitive decline to Alzheimer’s disease. This isn’t currently possible with existing diagnostic tools and approaches.
One of the limitations of this research is that it did not use post-mortem assessments to diagnose Alzheimer’s and assess its severity. Instead, it relied on clinical diagnosis, severity scores and MRI scans. While these are practical and valid measures, the gold standard for Alzheimer’s diagnosis is a post-mortem examination of the brain. This could be corroborated with the test results in future studies.
This is the first research group to test the predictive ability of this specific panel of proteins.
Interestingly, a previous small study found 10 other blood lipid biomarkers predicted, with 90% accuracy, 28 cognitively normal participants who progressed to have either mild cognitive impairment or mild Alzheimer’s disease within two to three years, compared to those who did not.
It will be important for future research groups to confirm and replicate the findings, to see if the results are the same, or if a combination of these approaches improves the predictive values in larger trials.