Amyotrophic lateral sclerosis is a relentlessly progressive neurodegenerative disease in which motor neurons in the brain and spinal cord die, producing progressive weakness, respiratory failure and, typically, death within two to three years of symptom onset. Because the disease moves at strikingly different speeds in different people, and even within the same person over time, clinicians have long sought a metric that captures an individual’s progression rate early enough to guide counseling, treatment planning and trial enrollment. The instrument they rely on most is the revised ALS Functional Rating Scale, or ALSFRS-R, a 48-point questionnaire covering speech, swallowing, fine motor tasks, walking, breathing and self-care that is administered at virtually every clinic visit.

From this scale, researchers derive what is known as the pre-diagnostic ALSFRS-R slope, often called the preslope. It is calculated by subtracting the patient’s ALSFRS-R score at diagnosis from the maximum of 48 points and dividing by the number of months between symptom onset and diagnosis. The resulting figure, expressed in points lost per month, has become embedded in ALS practice and research: it has repeatedly been shown to predict survival and later functional decline, and it has been written into the design of randomized clinical trials both as an eligibility criterion and as a stratification variable intended to balance treatment groups. In the clinic, many neurologists use it to give newly diagnosed patients a rough forecast of what lies ahead.

The implicit assumption behind all of this is that the rate of decline before diagnosis continues, more or less, after diagnosis. The new study, led by Rosario Vasta and Adriano Chiò of the ALS Center at the University of Turin, together with colleagues in Modena, Novara and across the Emilia-Romagna region, put that assumption to a rigorous test. Their primary cohort came from the Piemonte and Valle d’Aosta Register for ALS, known as PARALS, a prospective population-based registry covering roughly 4.3 million people in northwestern Italy. Of 2,895 patients diagnosed there between 2000 and 2022, 2,222 had an ALSFRS-R assessment at diagnosis and entered the analysis. Every finding was then independently replicated in the Emilia-Romagna Registry for ALS, or ERRALS, which covers about 4.5 million people in northeastern Italy and contributed 1,543 patients diagnosed between 2009 and 2023.

The results were consistent across both cohorts and across several complementary statistical approaches. The correlation between the pre-diagnostic slope and the rate of decline measured 6, 12 or 18 months after diagnosis was modest at best, with Spearman’s rank correlation coefficients ranging from 0.19 to 0.30 in PARALS and from 0.23 to 0.24 in ERRALS. More striking was how little variance the preslope explained: 6.1 percent of the variability in the post-diagnostic slope at 6 months in PARALS, falling to just 0.4 percent by 18 months, and only 3.2 percent at 6 months in ERRALS, dropping to 0.4 percent thereafter. Scatter plots of pre- versus post-diagnostic slopes showed wide dispersion across the entire range, a visual confirmation that knowing how fast a patient declined before diagnosis leaves the future largely undetermined.

The team then asked whether the preslope added anything beyond established prognostic factors such as sex, age at onset and site of onset, and whether it outperformed other clinical measures. In regression models, adding the pre-diagnostic slope to a base model raised the explained variance by 6.1 percentage points at 6 months, but only 2.8 points at 12 months and a statistically insignificant 0.6 points at 18 months. When baseline lung function, measured as forced vital capacity, and the percentage of body mass lost since the premorbid state were also included, the preslope’s unique contribution shrank further, to 5.2, 1.9 and 0.08 percentage points at the three time points. Notably, at 6 months the preslope still outperformed both respiratory function and weight loss, but by 12 and 18 months forced vital capacity had become the stronger predictor, suggesting that a global rate reconstructed from symptom onset loses value faster than a direct snapshot of respiratory reserve.

Because patients who die or undergo tracheostomy drop out of follow-up, and because that attrition is itself tied to faster progression, the researchers also fitted sophisticated joint longitudinal-survival models that explicitly link a patient’s underlying functional trajectory to their risk of death. These models confirmed the overall pattern. In PARALS, accounting for informative attrition barely changed the estimates, while in ERRALS it attenuated the large late-follow-up differences, indicating that some of the apparent long-term predictive power of the preslope in that cohort reflected the selective disappearance of the sickest patients rather than genuine trajectory prediction.

The mixed-effects analyses, which used all 9,670 ALSFRS-R assessments recorded during the first 18 months after diagnosis in 1,743 PARALS patients, added an important nuance. A higher pre-diagnostic slope was associated with faster early decline, but also with a subsequent deceleration: the disease’s pace after diagnosis was not linear, and the shape of that curvature depended on the preslope itself. Patients in the highest quartile of pre-diagnostic slope declined about 2.04 points more on the ALSFRS-R at 12 months than those in the lowest quartile in PARALS, with a comparable 2.47-point gap in ERRALS. Yet even with both linear and quadratic interaction terms included, the preslope-related terms raised the marginal R-squared by only 0.43 percentage points in PARALS and 1.68 percentage points in ERRALS. In other words, the average difference between groups was meaningful, but the metric was nearly useless for distinguishing one individual’s future from another’s.

The authors are careful to frame this as a refinement rather than a refutation. The pre-diagnostic slope remains a legitimate group-level prognostic marker, consistent with earlier work using the PRO-ACT database, which found a moderate correlation of about 0.40 between pre- and post-baseline slopes, and with a smaller Japanese registry study that found an even weaker association. The new findings also dovetail with the recent PRECISION-ALS study, which characterized the natural history of ALSFRS-R trajectories and included many of the same PARALS patients. What this study adds is a quantification of just how little the preslope contributes to explaining the heterogeneity of individual futures, a distinction with direct consequences for how the metric should be used.

The implications ripple outward in two directions. For clinical trials, using the preslope as a stratification or eligibility variable may balance groups on average but does little to reduce the individual-level heterogeneity that inflates sample size requirements, and trial designers may want to weigh respiratory function and other measures more heavily at later time points. For patients and clinicians, the message is one of calibrated humility: a fast preslope does signal a worse average outlook, but it cannot tell any single person whether their own decline will track that average or diverge from it. The study also carries the usual caveats, including reliance on patient-reported symptom onset dates, the ordinal structure of the ALSFRS-R itself, and the fact that both cohorts were Italian, which may limit generalizability. Still, the replication across two independent, population-based registries gives the central conclusion unusual weight: in ALS, the past rate of decline is a weak prophet of the future, and the search for biomarkers that truly predict individual trajectories remains wide open.

Subject of Research: Predictive value of the pre-diagnostic ALSFRS-R slope for subsequent functional decline in amyotrophic lateral sclerosis

Article Title: Pre-diagnostic ALSFRS-R slope provides limited incremental information about the variability in subsequent functional decline in amyotrophic lateral sclerosis: a population-based study

Article References: Vasta, R., Calvo, A., Moglia, C., Canosa, A., Manera, U., Gianferrari, G., Zucchi, E., Martinelli, I., Mora, G., Cabras, S., Palumbo, F., Matteoni, E., Maccabeo, A., Pellegrino, G., Minerva, E., Pascariu, D., Callegaro, S., D’Ovidio, F., Mazzini, L., … Chiò, A. (2026). Pre-diagnostic ALSFRS-R slope provides limited incremental information about the variability in subsequent functional decline in amyotrophic lateral sclerosis: a population-based study. Journal of Neurology, 273(10), Article 651. https://doi.org/10.1007/s00415-026-14177-2

Image Credits: AI Generated

DOI: 10.1007/s00415-026-14177-2

Keywords: amyotrophic lateral sclerosis, ALSFRS-R, disease progression, prognosis, population-based registry, clinical trials, predictive markers, forced vital capacity, longitudinal modeling, neurodegeneration, PARALS, ERRALS

The study, published in the Journal of Neurology, drew on the PARALS registry, a prospective, population-based register of every incident case of ALS in the Piedmont and Aosta Valley regions between 2000 and 2022. From 2,808 patients with non-cognitive onset, the researchers assembled an analytic cohort of 2,738 individuals, of whom 2,561, or 93.5 percent, had reached the composite endpoint of death or tracheostomy by the end of 2025. This near-complete ascertainment of outcomes is what gives the analysis its power: rather than a selected clinic population, it captures the full demographic and clinical spectrum of ALS as it actually occurs in a defined geographic area, with a median age at onset of 67.8 years.

The classification under scrutiny is the ALS-OPM 3.3 system, a three-axis framework that summarises each patient at first observation along three dimensions. The O axis records the anatomical region of symptom onset across eight categories, from head and bulbar onset through distal, proximal and mixed arm and leg presentations to trunk-respiratory onset. The P axis measures propagation time, the interval in months from first symptoms to involvement of a second, distant body region. The M axis captures the motor-neuron pattern: whether upper and lower motor neurons are affected in balance, as in the classical Charcot form of the disease, or whether one population predominates. The Italian team operationalised this axis as a three-level variable, M0 for balanced involvement, M1d for upper motor neuron predominance and M2d for lower motor neuron predominance, which together covered 99 percent of the non-PLS cohort.

The survival differences across the motor-neuron axis were stark and highly significant. Patients with the balanced M0 pattern, roughly two-thirds of the cohort, had the shortest median survival at 27.0 months from onset. Those with lower motor neuron predominance, M2d, which includes the scapulohumeral flail-arm and flail-leg variants, survived a median of 36.0 months. The upper motor neuron predominant group, M1d, fared best at 41.9 months. A multivariate log-rank test confirmed that these three populations are genuinely distinct, with all pairwise comparisons highly significant. In adjusted Cox models, both M1d and M2d carried independently protective hazard ratios relative to M0, at 0.59 and 0.83 respectively, effects that persisted regardless of how the other axes were handled.

Propagation time emerged as the single most powerful prognostic ingredient of the framework. Each additional month of delay before disease spread to a second region reduced the adjusted hazard by roughly 3.3 percent, so that a twelve-month difference in propagation time corresponded to an approximate hazard ratio of 0.66. Median survival climbed from 21.0 months in the fastest tertile of propagation to 43.1 months in the slowest. Onset anatomy alone, by contrast, discriminated only modestly, raising the C-index from chance levels to about 0.57, while adding propagation lifted discrimination to 0.68 and the motor-neuron pattern and demographics brought it to roughly 0.71. The authors recommend the linear specification of propagation capped at 60 months as the primary model, since a more flexible spline added no significant gain.

The decisive experiment concerned the arm-proximal onset category, the topographic signature of the flail-arm syndrome, long regarded as an indolent form of the disease. In the cohort, 95.7 percent of arm-proximal patients belonged to the lower motor neuron predominant M2d class, making this onset region a near-deterministic proxy for that motor-neuron pattern. When the researchers modelled survival without accounting for motor-neuron pattern, arm-proximal onset appeared protective, with a hazard ratio of 0.86 relative to distal arm onset. But once the M axis entered the model, the protective effect of arm-proximal onset vanished entirely, rising to a hazard ratio of 0.99 with a confidence interval spanning unity, while the M2d term retained its significant protective effect of 0.83. The favourable prognosis attributed to the anatomical site was, in fact, a property of the underlying motor-neuron biology that the anatomy happened to signal.

This decomposition is the intellectual heart of the paper. The authors report that ALS-OPM 3.3 did not materially improve prognostic discrimination over the established classical phenotype classification, with optimism-corrected C-indices of 0.703 versus 0.701, and it did not surpass clinical staging systems or El Escorial diagnostic certainty. Rather than treating this as a failure, the researchers interpret it as the central finding: the new framework does not predict better than existing taxonomies, but it reveals what those taxonomies were implicitly measuring all along. Classical composite labels such as flail-arm bundle together onset topography and motor-neuron pattern into a single descriptor; OPM disaggregates that signal into three separately interpretable components, of which propagation kinetics is the dominant contributor and motor-neuron pattern the operative biological variable.

The framework also proved complementary to, rather than redundant with, the tools already in clinical use. Combining OPM with King’s staging, the Milano-Torino staging system, or the more recent Fine’til 9 system significantly improved discrimination over OPM alone, and the motor-neuron hazard ratios remained stable across all three staging frameworks. Even among patients at identical King’s baseline stages, those with upper or lower motor neuron predominance transitioned more slowly to severe clinical milestones than the balanced M0 group, with median time to King’s stage 4 reaching 64.4 and 60.4 months respectively against 40.5 months for M0. El Escorial diagnostic certainty, which largely reflects baseline multifocality, also added independent prognostic information, and the combination of El Escorial with the OPM axes achieved the highest discrimination of any model tested, at a corrected C-index of 0.7128.

The study is not without caveats, which the authors confront directly. The cohort, though population-based, comes from a single Italian region, and external validation in other incident registries will be needed. The propagation axis depends partly on patients’ recollection of when symptoms spread, though concordance with prospective functional-rating data was high and a landmark analysis confirmed the robustness of the findings. The primary lateral sclerosis spectrum subgroup was analysed separately to avoid immortal-time bias, and sensitivity analyses showed that its inclusion or exclusion left the principal motor-neuron effects unchanged. The classification is also purely motor, and does not capture the cognitive and behavioural dimensions of the disease that independently influence prognosis.

The practical implications reach from the neurologist’s consulting room to the design of clinical trials. A patient presenting with arm-proximal onset can now be counselled that the relatively favourable trajectory reflects the lower motor neuron predominant pattern rather than the site of onset itself, a distinction that purely anatomical classifications obscure. In trials, balancing treatment arms on motor-neuron pattern rather than onset topography would remove a hidden source of prognostic imbalance that composite labels leave undetected. More broadly, the study offers a template for how the field should evaluate any proposed classification: not by asking whether it predicts better, but by asking what, exactly, it measures. In the case of ALS motor phenotypes, the answer is now clear: the motor neurons themselves, and the speed at which the disease escapes its point of origin, matter far more than the map coordinates of the first symptom.

Subject of Research: Prognostic decomposition of ALS motor phenotype classifications in a population-based cohort

Article Title: What do ALS motor phenotypes measure? A population-based decomposition of their prognostic dimensions

Article References: Calvo, A., Moglia, C., Callegaro, S., Canosa, A., Manera, U., Mora, G., Cabras, S., Palumbo, F., Matteoni, E., Maccabeo, A., Pellegrino, G., Minerva, E., Pascariu, D., D’Ovidio, F., Vasta, R., Mazzini, L., De Marchi, F., & Chiò, A. (2026). What do ALS motor phenotypes measure? A population-based decomposition of their prognostic dimensions. Journal of Neurology, 273(10), Article 604. https://doi.org/10.1007/s00415-026-14134-z

Image Credits: AI Generated

DOI: 10.1007/s00415-026-14134-z

Keywords: amyotrophic lateral sclerosis, motor phenotype, prognosis, population-based registry, survival analysis, motor neurons, disease staging, OPM classification, flail-arm syndrome, disease propagation, Cox model, clinical trials