In today’s JAMA Neurology, researchers led by Oskar Hansson, Lund University, Sweden, report how a fully automated immunoassay for plasma Aβ performed when they put it through its paces. Roche Diagnostic’s Elecsys system predicted Aβ-positive individuals with about 80 percent accuracy. That number improved by 5 percent when the researchers took ApoE genotype into consideration. Alzforum first reported on this data when Hansson presented it at AAIC in 2018 (Aug 2018 conference news). Since then, first author Sebastian Palmqvist and colleagues have tested samples in a validation cohort in Germany. Here the test was more accurate, at 86 percent. “I think we are not that far away from a blood-based biomarker that can be implemented in primary care to improve diagnostics of AD,” Hansson told Alzforum.
- Roche Diagnostic’s Elecsys tested in Swedish and German cohorts.
- Fully automated plasma test predicts AD with 85 percent accuracy.
- The immunoassay could offer a screen for people at risk.
The Elecsys system is one of a handful of plasma immunoassays that are vying with mass spectrometry-based methods for use as a blood test in the clinic. Each has its pros and cons. The immunoassays are easier to implement, though not quite as accurate as the mass-spec approaches. Being fully automated, the Elecsys system comes with the added advantage of cutting down on variability that can be introduced by manual steps that are part of mass spec, ELISA, and Simoa immunoassay approaches. It can be scaled up easily for use in primary-care settings, which makes it attractive to Hansson. “A lot of patients with Alzheimer’s disease are not diagnosed, or they get an incorrect diagnosis,” he noted. “A plasma assay could improve this situation and boost the number of people who get proper symptomatic treatments,” he said.
Palmqvist and colleagues first tested the assay using samples from 446 women and 396 men in the Swedish BioFINDER cohort. These volunteers ranged in age from 59 to 88, with an average age of 72. Sixty-four had Alzheimer’s disease, 265 mild cognitive impairment, 513 were healthy controls. All agreed to lumbar taps for cerebrospinal fluid collection, and the researchers used their CSF Aβ42/Aβ40 ratio to determine their brain-amyloid status. All 64 AD patients tested positive, as did 157 of those with MCI and 147 of the controls. Blood samples were collected at the same time.
The plasma Aβ42/Aβ40 ratio tracked lower in amyloid-positive people, in keeping with the premise that the stickier Aβ42 gets trapped in the brain and no longer clears into the blood. The lower plasma ratios predicted amyloid status with an AUC of 0.8. This is not good enough for a stand-alone diagnostic test, the authors note.
To see if they could improve accuracy, they combined this analysis with plasma tests for tau and neurofilament light, and they also took ApoE status into consideration. Individually, the two plasma markers improved the AUC slightly. ApoE brought it to 0.85.
However, while these additional parameters increased the sensitivity of detection, they decreased the specificity. This is not surprising. Tau and NfL levels increase in other neurodegenerative diseases, and ApoE4 is a risk variant for other forms of neurodegeneration as well. Still, the researchers got the best predictive value by adding all three markers to the Aβ ratio, improving accuracy to 0.87. The Aβ assay worked as well in cognitively impaired as in healthy people, but plasma tau and NfL added no value in the latter. Importantly, however, the assay worked as well in younger people as in older. “The data suggest the assay performs equally well in preclinical as in prodromal AD,” said Hansson.
For validation, the researchers turned to a prospective study running in the German cities of Ulm and Hanover. Among 94 volunteers with AD, 109 with MCI, and 34 healthy controls, the Elecsys immunoassay performed slightly better than in the Swedish cohort, predicting amyloid positivity with an AUC of 0.86. Adding plasma tau brought that down to 0.84. Plasma NfL and ApoE genotype were unavailable for this cohort. “Overall, the data show good reproducibility,” said Hansson.
The authors hope better plasma-tau assays will make the test more accurate. Hansson told Alzforum that Roche already has upped the sensitivity of the Aβ assay beyond that reported in this paper; he hopes the improved assay will be finalized within the year.
Sid O’Bryant, University of North Texas Health Science Center, Fort Worth, thinks the work represents a significant advance. “Overall, the findings […] are promising and demonstrate that the field is rapidly moving from ‘if’ blood biomarkers can be used in AD to ‘how’ they can be used,” he wrote in a JAMA Neurology editorial.
These assays could facilitate both primary-care diagnosis and recruitment into clinical trials. In a cost analysis, Palmqvist and colleagues estimated that a plasma test could take $3.2 million off the $9.2 million price tag for screening 1,000 trial volunteers by amyloid PET, assuming they could reduce the number of scans to 200.
O’Bryant believes that there is room for improvement. If these tests are to be used in the clinic, scalability will be a huge concern, he wrote. “The blood collection and processing procedures are not applicable to standard clinical lab practice and will cause substantial barriers to clinical application.”
This point was echoed by Hugo Vanderstichele, ADx Neurosciences, Zwijnaarde, Belgium. He also thinks it is important to understand how assays relate to biology. “Now we need to develop Precision Qualified Assays,” he said. These marry assay analytical performance to biology that’s relevant for specific contexts of use. They demand better standardization than currently available, he said. He also thinks Aβ plasma assays could be improved by taking isoforms besides Aβ40 and Aβ42 into account, by incorporating other markers, and by better understanding biological factors that might shift plasma Aβ levels, such as exercise, the body-mass index, and medications (see comment below).
Could other assay modalities do better than Elecsys? Perhaps. The system falls short of the accuracy of recently described mass-spec assays, which claim AUCs of 0.84 to 0.97 depending on assay and cohort (Feb 2018 news; Jul 2017 conference news). “We really need head-to-head comparisons running the different assays on the same set of samples,” said Hansson.
Such studies are underway. Co-author Kaj Blennow, University of Gothenburg, Sweden, has completed a round-robin study. It collected one set of samples and shared them among labs for analysis on different assay platforms. Part of the difficulty with such studies is collecting enough material to share, but Blennow told Alzforum they have 80 individual samples, all treated exactly the same way and divided into identical aliquots. They have been tested in 10 different laboratories using 11 different plasma assays, including three SIMOA immunoassays, two ELISAs, four immunoprecipitation/mass spec methods, one immunomagnetic reduction method, and the fully automated Elecsys immunoassay.
All have measured plasma Aβ42, Aβ40, and the ratio. Blennow and colleagues are currently analyzing the data and will present them in July at AAIC.—Tom Fagan
No Available Further Reading
- Palmqvist S, Janelidze S, Stomrud E, Zetterberg H, Karl J, Zink K, Bittner T, Mattsson N, Eichenlaub U, Blennow K, Hansson O. Performance of Fully Automated Plasma Assays as Screening Tests for Alzheimer Disease-Related β-Amyloid Status. JAMA Neurol. 2019 Jun 24; PubMed.