Seeing Multiple Sclerosis Coming: What Early-Life Health Clues Reveal About Future MS Risk

For decades, multiple sclerosis (MS) has been framed as a disease that appears “out of the blue” when clear neurological symptoms—such as visual loss or limb weakness—prompt a diagnosis. Yet many patients, when looking back, recall years of unexplained pain, fatigue, mood changes, or odd sensory symptoms that never quite made sense at the time. The new study by Nova and colleagues in Annals of Clinical and Translational Neurology asks a straightforward but powerful question: can we systematically detect these early signals of MS in routine health records, long before the first formal MS diagnosis is made? Using the UK Biobank, they mapped the full clinical histories of nearly half a million people and searched for conditions that tend to appear more often—and earlier—in those who eventually develop MS. Their findings provide some of the clearest evidence so far that MS is usually preceded by a measurable “prodromal” phase, spanning multiple organ systems and often many years.

Mining a Lifetime of Health Data for Early Clues
To tackle this, the authors turned to the UK Biobank, a large, population-based cohort of 477,421 participants recruited at ages 40–70 and followed for up to several decades. For each person, they reconstructed a lifetime timeline of diagnoses using ICD-10 codes from hospital records, primary care data, death registries, and self-reports. MS was defined by code G35, yielding 2,463 individuals with an MS diagnosis. From an initial set of 1,785 ICD-10 codes, they focused on 600 clinically meaningful conditions after excluding childbirth and congenital codes and aggregating overlapping diagnoses. Each condition was treated as a time-varying exposure in a Cox proportional hazards model: a participant was considered “unexposed” until the diagnosis date and “exposed” thereafter. Observation time began at birth and ended at MS diagnosis, death, loss to follow-up, or 31 December 2022, as illustrated schematically in the timeline graphic on page 3 (Figure 1). The models were adjusted for sex, year and place of birth, ethnicity, smoking, and a multiple sclerosis polygenic risk score (MS-PRS), ensuring that associations reflected more than just demographic or genetic risk differences.

A Long List of Warning Lights—Many Years Before Diagnosis
Out of the 371 conditions that appeared in at least five MS cases, 192 were significantly associated with future MS after correcting for multiple testing. As expected, classical onset symptoms such as demyelinating events, optic neuritis, and gait or mobility problems showed the strongest associations. Yet these “typical” MS presentations accounted for only about 20% of the MS-associated conditions. The majority were either prodromal conditions (around 56%)—health issues that likely reflect early disease biology—or conditions with an unclear relationship to MS. Timing analysis, summarized in the scatterplot on page 5 (Figure 3), showed that roughly one-third of these conditions were diagnosed more than five years before MS, and about half appeared at least three years prior. Childhood and adolescence were marked by infections such as measles and mononucleosis; in adulthood, early signals clustered in the genitourinary, nervous, cardiovascular, and musculoskeletal systems, and prominently included mental health disorders such as depression, anxiety, and bipolar disorder. Importantly, 60% of people with MS had at least one MS-associated condition before diagnosis (median two conditions), and those with such a history were diagnosed with MS at a later age than those without—suggesting a longer, more clinically visible prodromal phase.

From Association to Prediction: Can We Flag High-Risk Individuals?
Association alone is not enough; clinicians need to know whether these patterns actually help predict who will develop MS. To address this, the authors built time-to-event prediction models. They split the cohort into a 70% training/validation set and a 30% test set and used LASSO Cox regression to select the most informative predictors. A baseline model that included only demographics and smoking achieved a Harrell’s C-index of 0.653 in the test set—better than chance, but modest. When they added 62 conditions that typically occur more than five years before MS, the model selected 43 key conditions and the C-index rose to 0.685. Trigeminal neuralgia, an MS onset symptom that can appear well before a formal diagnosis, emerged as the single most predictive early condition, followed by prodromal features such as dorsalgia (back pain), bladder and prostate problems, depression, brain disorders, anemia, hypertension, pneumonia, diabetes, certain infections (e.g., mononucleosis, viral CNS infection, measles), and pregnancy. As additional conditions diagnosed closer to MS (3–5 years and 1–3 years before) were included, the C-index improved further to 0.713 with a final set of 30 predictors. About one quarter of these were clearly prodromal rather than onset symptoms, including neuromuscular diseases, thromboembolism, and depression, which tended to appear more than five years before MS. The performance curves in Figure 4 show that adding clinical history improved age-dependent discrimination, particularly from age 30 onwards. When MS-PRS was included, the overall C-index rose to 0.783 and the age-specific AUC exceeded 0.80 from about age 50, highlighting the complementary value of genetic and clinical information.

Common Diagnostic Journeys: Migraine, Hypertension, Depression, Dorsalgia
Beyond individual risk factors, the authors asked whether there are typical trajectories—chains of diagnoses—that commonly lead to MS. Among the 1,485 MS cases with at least one MS-associated condition, the earliest recorded diagnoses were most often from the nervous (17%) and circulatory (17%) systems. Conditions such as hypertension, depression, migraine, dorsalgia, and hypothyroidism frequently appeared first (see page 7 and the supplementary tables). In those with at least two MS-associated conditions, the last diagnosis before MS was most often a symptom or abnormal clinical finding (23%) or a nervous system disorder (15%), including CNS demyelination, urinary disorders, depression, and dorsalgia. By testing thousands of ordered diagnosis pairs and then assembling them into longer paths, the team identified 282 significant trajectories. Two main clusters stood out. In one, a relatively direct cardiometabolic pathway, early conditions like hypertension, diabetes, dyslipidemia, and dorsalgia connected via lipid metabolism and renal disorders to MS. In the other, more complex cluster, early hypertension, dorsalgia, and hypothyroidism were followed by intestinal, musculoskeletal, cardiovascular, and visual problems, then by increased healthcare utilization and subtle neurological symptoms—fatigue, abnormal movements, dysphagia, sensory disturbances—before culminating in CNS demyelination. Migraine occupied a particularly central role, often preceding depression, hypertension, dorsalgia, and hypothyroidism, while trigeminal neuralgia frequently preceded demyelination and seemed to anchor a distinct pathway.

What This Could Mean for Patients and Clinicians
Taken together, these results suggest that for many individuals, MS does not arrive as an abrupt catastrophe but as the end point of a long, often confusing journey through multiple specialties—neurology, psychiatry, cardiology, rheumatology, urology, and primary care. For a mid-life patient with migraine, long-standing depression, intermittent urinary problems, and a history of mononucleosis, this study suggests that their pattern of diagnoses carries more information about future MS risk than any single symptom alone. In practice, a risk model that combines demographics, smoking, genetic risk, and a structured summary of clinical history could, in principle, flag such individuals as “MS-at-risk” long before a first relapse or overt demyelinating event. The age-dependent performance is particularly relevant: genetic risk (MS-PRS) plays a dominant role for adult-onset MS in younger patients, whereas clinical history becomes a stronger predictor after age 50, when late-onset MS is more common and often more difficult to recognize. Used carefully, such tools could prompt earlier neurological evaluation, MRI imaging, or biomarker testing (e.g., serum neurofilament light chain) in high-risk patients, potentially shortening the diagnostic delay and allowing earlier initiation of disease-modifying therapies.

Disclaimer: This blog post is based on the provided research article and is intended for informational purposes only. It is not intended to provide medical advice. Please consult with a healthcare professional for any health concerns.

References:
Nova, A., Fazia, T., Di Caprio, G., Gentilini, D., Bernardinelli, L., & Bergamaschi, R. (2025). Timing and Predictive Value of Clinical Conditions Preceding Multiple Sclerosis in the UK Biobank. Annals of Clinical and Translational Neurology.