Blood test could diagnose ALS years before symptoms appear: Study
Method based on unique protein signature shows high accuracy
Through analyses of blood samples from hundreds of patients, researchers have developed a highly accurate test for diagnosing amyotrophic lateral sclerosis (ALS).
The test is based on a unique protein signature in the blood that may be able to detect ALS-related changes up to 10 years before symptoms of the disease become evident. The blood proteins could also help estimate when someone asymptomatic may experience their first ALS symptoms.
“We see the light at the end of the tunnel here, and that target is an approved and available blood test for ALS,” Alexander Pantelyat, MD, co-author of the study at Johns Hopkins University School of Medicine, said in a university news story. “With a test that allows for earlier detection of ALS, we have opportunities to enroll people in observational studies, and by extension, offer promising disease-modifying — and hopefully disease-stopping — medications, before ALS becomes debilitating.”
The study, “A plasma proteomics-based candidate biomarker panel predictive of amyotrophic lateral sclerosis,” was published in Nature Medicine.
Definitive diagnosis can take months or years
ALS was first formally described more than 150 years ago. However, methods for diagnosing ALS have not substantially changed in that time — the diagnosis still relies on a doctor examining a patient’s symptoms and ruling out other conditions. It often takes months or years for patients to get a definitive diagnosis.
“This uncertainty can be unsettling for patients and their families. Moreover, it causes delays in entering clinical trials and limits the number of eligible participants due to enrollment restrictions for patients with more severe disability,” the researchers wrote. They noted that delays in diagnosis also mean delays in starting treatment, which is associated with worse outcomes.
Aiming to create a more objective diagnostic test, scientists analyzed blood samples from 231 people with ALS, 170 people with other neurological diseases, and 214 people without neurological disease.
Of a total of 2,886 proteins, the researchers identified 33 blood proteins that were present at significantly altered levels in the ALS patients. Two of these proteins, neurofilament light chain (NfL) and leukemia inhibitory factor (LIF), have been implicated in ALS previously, but the other 31 have not.
Most of the proteins were associated with biological processes needed for the function of nerve cells and skeletal muscles, which are involved in voluntary movements.
Blood test challenges belief that ALS onset occurs rapidly
The researchers then created a machine learning model that looked at levels of these 33 proteins alongside factors including age and biological sex to identify ALS. The model was created using data from most of the patients, and the remaining patients’ data was used to test the model.
Results showed the model was 98.3% accurate. Analyses using other independent datasets likewise showed very high accuracy.
“We identified a panel of 33 differentially abundant proteins in ALS and validated our findings across multiple independent datasets, showing that the same protein profile can differentiate ALS from other neurological diseases,” the researchers concluded.
Additional tests showed that the model used 20 factors to classify samples as ALS or control, including levels from 17 proteins, sex, age at sample collection, and the type of tube used to collect the blood sample.
We had always assumed that ALS was a rapid disease that starts 12 to 18 months before symptom onset. But when we look at our findings, we see this has been a process that goes on for a decade or so before the patient ever steps into the doctor’s office or clinic.
In some datasets, blood samples collected before the onset of ALS symptoms were available for some patients. The researchers showed that their machine learning model could identify ALS signatures in samples taken up to 10 years before the onset of symptoms, and could even predict when symptoms were likely to appear with high accuracy.
“We demonstrated that the ALS risk score computed from our machine learning model could be used as a proxy for forecasting the age at which symptoms may appear in individuals,” the scientists wrote, adding that these data challenge a longstanding belief that the onset of ALS occurs very rapidly.
Added Pantelyat: “We had always assumed that ALS was a rapid disease that starts 12 to 18 months before symptom onset. But when we look at our findings, we see this has been a process that goes on for a decade or so before the patient ever steps into the doctor’s office or clinic.”
The analyses also indicated that the protein signature identified was not driven by inherited genetic mutations, implying that this signature may apply to all ALS patients.
The researchers noted, however, that most of the patients included in these analyses were of European descent, adding that “a larger case-control cohort with greater ancestral diversity is necessary for additional validation to assess the global applicability of our predictive model.”
Some of the researchers have a patent pending on a blood test based on this study’s results, which was partly funded by the National Institutes of Health and the pharma company Merck, among others.
“Fifteen years of cross-institutional collaboration went into this work,” Pantelyat said. “Large-scale partnerships are the lifeblood of research. They’re what will lead to effective diagnostics and, ultimately, effective treatments for devastating diseases like ALS.”
About the Author
