CSF protein levels at diagnosis may help predict MS long-term outcomes
Findings could help development of individualized treatment protocols
Levels of proteins in the liquid that surrounds the brain and spinal cord, called the cerebrospinal fluid (CSF), can help predict disease activity and disability worsening for people with multiple sclerosis (MS), a study reports.
“We identified several promising protein biomarkers which could be used to predict short-term activity and long-term disease progression in newly diagnosed MS,” the study’s researchers wrote in “Proteomics reveal biomarkers for diagnosis, disease activity and long-term disability outcomes in multiple sclerosis,” which was published in Nature Communications.
Researchers and clinicians are working to develop more personalized treatment strategies for MS. But this requires being able to accurately predict the risk of future disease activity and identifying reliable markers has been challenging.
A research team led by scientists in Sweden analyzed levels of 1,463 proteins in the CSF and plasma (the liquid component of blood) from 143 people with early-stage MS and 43 people without MS to find new disease markers. The samples were collected from patients at two centers in Sweden. The researchers analyzed data from MS patients and healthy controls at one center (the discovery cohort), then tested whether the same patterns were present in people at the other center (the replication cohort).
“We consistently used the same … MS cohorts from two different sites and allowed no retraining of any prediction model parameters in the replication cohort, thus increasing the generalizability and ability for successful replication of models also in other cohorts,” the researchers said.
Analyzing protein levels
Fifty-two proteins were found at significantly different levels in CSF between MS patients and healthy people in the discovery cohort, and 23 of them showed consistent results in the replication cohort. Many of these proteins have been linked with MS, said the researchers, who then conducted statistical tests to see how well protein levels could differentiate between people with or without MS.
The researchers used a statistical measure called the area under the receiver operating characteristic curve, or AUC, which assesses how well a particular test can discriminate between two groups. AUC values range from 0.5 to 1, with higher scores reflecting a better ability to accurately differentiate the groups.
Based on CSF levels of the identified proteins, people with MS could be distinguished from healthy people with an AUC of up to 0.99 in the discovery cohort and 0.87 in the replication cohort.
Predicting disease activity
Many of the proteins with the clearest discriminatory power are related to the activation of B-cells, a type of immune cell known to be involved in MS, said the researchers, who then looked for protein markers that could identify the risk of short-term disease activity, namely whether patients achieved no evidence of disease activity-3 (NEDA-3) at two years after the samples were taken. Achieving NEDA-3 means a person has had no relapses, no substantial disability worsening, and no new signs of MS activity on MRI scans.
To calculate the accuracy of markers for predicting disease activity, the researchers again used the AUC, but this time to see if patients achieved NEDA-3.
A well-established marker of nerve damage called neurofilament light chain (NfL) showed the highest accuracy for detecting NEDA-3, with an AUC of 0.75 in the discovery cohort and 0.77 in the replication cohort.
Certain immune molecules in the CSF accurately detected people with and without NEDA-3, but no blood proteins could significantly distinguish the two groups.
The researchers also looked at longer-term disability progression activity, as measured by an age-adjusted disability measure called the age-related MS score (ARMSS) over at least three years of follow-up.
A mathematical model using patients’ age and levels of NfL and 10 other proteins that were differentially produced in patients and healthy individuals — namely CXCL13, LTA, FCN2, ICAM3, LY9, SLAMF7, TYMP, CHI3L1, FYB1, and TNFRSF1B — showed statistically significantly associations with ARMSS scores in both cohorts.
While protein levels in CSF were largely accurate for identifying MS and predicting disease activity, plasma protein levels generally didn’t show statistically meaningful associations, except for a moderate association between serum NfL levels and long-term disability scores.
If these results can be verified, having a set of CSF proteins that can predict long-term disease disability progression could help doctors develop “individualized treatment protocols to avoid both over- and under-treatment in terms of drug efficacy, which is highly relevant with respect to risk of side-effects as well as costs,” the researchers said.
About the Author
