Recent advances in the diagnosis and prognosis of ALS

Motor symptoms and phenotypic heterogeneity

ALS was historically considered a relatively uniform disease of progressive painless voluntary muscle weakness.¹¹ Studies over the past decades have redefined ALS as a complex disorder with significant heterogeneity in clinical presentation in site of disease onset and distribution of upper and lower neuron motor signs (Figure 1A, Table 1). Recognizing these multiple heterogeneous ALS phenotypes facilitates earlier diagnosis and informs prognosis.⁸ The Australian National Motor Neuron Disease (1,677 ALS patients)⁹ and Italian Piemonte and Valle d’Aosta (2,839 ALS patients)^5,10 have documented this heterogeneity in ALS phenotypes, which also correlate with median survival (Figure 1B–C). Uniformly, bulbar onset patients are at a greater risk for FTD than other phenotypes.⁵ Additionally, less common ALS phenotypes exist, e.g., hemiplegic (Table 1).¹¹ Furthermore, phenotypes correlate with timing of certain treatments. In the Australian registry, feeding tube placement secondary to dysphagia occurs earlier in bulbar versus spinal onset patients,⁹ as also reported in a European tertiary care ALS cohort.¹²

Thus, phenotyping is based on clinical criteria, such as site of disease onset and distribution of upper and lower motor signs.⁸ Additional relevant clinical variables aid disease classification and can provide prognostic guidance,¹³ such as age, sex, family history, progression rate, genetic profile and presence of cognitive impairment and other non-motor symptoms (see sections below).

Non-motor ALS symptoms

The concept of ALS as a pure motor disease is now abandoned. In fact, executive dysfunction in 50% and FTD in 15% of ALS patients has been known for decades. Executive dysfunction is evaluated by a suite of neuropsychological tests (Table 2)¹⁴ and FTD in ALS is diagnosed by the revised Strong criteria.¹⁵ The most characteristic cognitive changes in ALS include impaired language function¹⁶ and executive function deficits involving working memory, inhibition, set shifting, and fluency, whereas memory and spatial function are typically spared.¹⁷ ALS patients also experience social and cognitive decline, including apathy, disinhibition, irritability, loss of sympathy/empathy, perseveration, reduced concern for hygiene, and changes in eating habits. Similar clinical patterns are present in FTD.¹⁷ Additionally, many ALS patients suffer from anxiety, depression, and sleep disorders.¹⁸

Executive dysfunction is a negative prognostic indicator, and if present tends to worsen over time.¹⁹ Cognitive impairment can later manifest even in patients who appear to be cognitively spared at diagnosis¹⁹ and appears to be in part related to the worsening of motor function.¹⁷ Thus, there is a growing need to incorporate an evaluation of cognitive function into ALS diagnosis and ongoing management. These behavioral changes can also frustrate family members and caregivers and prevent the patient from accepting medical recommendations, emphasizing the importance of addressing care preferences early in disease.²⁰ These cognitive and behavioral symptoms are accompanied by structural changes to the brain in extra-motor domains (see “Brain and spinal cord imaging” section).

ALS genes influence clinical phenotype

The discovery of mutant SOD1 in a subset of ALS patients in 1993 suggested a potential genetic etiology, which could enhance our understanding of disease risk factors and pathophysiology as well as identify therapeutic targets.²¹ This possibility was strengthened in 2011 by the discovery of C9orf72 repeat expansions in a larger proportion of individuals, both with and without family history of ALS.²² The genetic architecture of ALS and nuances of familial versus sporadic ALS are fully detailed in the accompanying research-focused review. Over 40 ALS genes are identified to date, which account for approximately 15% of cases. Thus, genetic testing is a growing, albeit non-uniform, component of ALS management. As the cost of genetic profiling drops, we anticipate earlier and broader adoption for ALS patients. First, detecting known pathogenic ALS variants could complement and bolster diagnosis achieved by diagnostic criteria (see “ALS diagnosis” section). Second, though most mutations converge on a typical ALS phenotype, there are important prognostic implications for certain mutant genes linked to unique features (Table 3). For example, ALS2, DCTN1, MATR3, OPTN, and SETX mutations are associated with slower clinical trajectories than classical ALS, information valuable to patients and their families. Furthermore, routine genetic profiling could move past the present and inadequate stratification of patients into sporadic or familial ALS, by uniformly identifying specific gene mutations. Additionally, genetic profiling promotes precision medicine for managing ALS patients²³ and clinical trial stratification for targeted therapeutics, e.g., gene therapies (see accompanying review). Therefore, a genetic profile could potentially facilitate diagnosis, prognosis, and treatment for patients harboring known ALS genetic variants.

Established diagnostic criteria

ALS diagnostic criteria date back to the original El Escorial and later the revised El Escorial (Airlie House) and Awaji criteria. They rate the degree of diagnostic “certainty by clinical assessment alone” from possible to probable to definite ALS, based on the number of affected segments combined with clinical and/or electrophysiological findings.^24–26 The El Escorial classification provides prognostic information since, for instance, definite ALS progresses faster.¹³ Although approaches that score the certainty of diagnosis solely by clinical assessment are reasonable (i.e., possible ALS), they delay diagnosis and confuse patients, their families, and clinicians, who misinterpret these terms as meaning the diagnosis is unlikely or incorrect.²⁷ In reality, nearly all patients diagnosed as possible ALS progress and ultimately die from ALS.

Emerging diagnostic criteria

To address these limitations, an international consensus group reconsidered criteria to improve the diagnostic process for ALS, particularly in the early stages of disease, when clinical symptoms are minimal.²⁸ Recognizing the broad properties of the disease, the Gold Coast criteria define ALS by:

• Progressive motor impairment, documented by history or repeated clinical assessment, preceded by normal motor function.

• Upper and lower motor neuron dysfunction in at least one body region, or lower motor neuron dysfunction in at least two body regions.

• Investigative findings that exclude alternative diseases.

Adopting these simplified criteria for ALS abandons the previous diagnostic categories of possible, probable, and definite. The advent of these new criteria facilitates diagnosing ALS early and definitively. An Australian study found Gold Coast criteria diagnostic sensitivity (92%) was maintained irrespective of functional status, disease duration, or disease onset site and was generally similar to the revised El Escorial (88.6%) and Awaji criteria (90.3%); however, Gold Coast criteria were more sensitive and specific, including for progressive muscular atrophy and for excluding primary lateral sclerosis as a form of ALS, the latter of which meets possible ALS by the revised El Escorial and Awaji criteria.²⁹ This finding was validated in a five-center European study, which similarly found consistent and improved sensitivity of Gold Coast criteria across ALS phenotypes, due to greater sensitivity for identifying progressive muscular atrophy.³⁰ Lastly, a Chinese study corroborated the greater sensitivity of the Gold Coast against the revised El Escorial and Awaji criteria,³¹ suggesting diagnostic utility would be maintained in racially diverse ALS populations.

Importantly, the Gold Coast criteria were marginally less specific, which clinicians should bear in mind as they monitor their patients’ disease course. However, overall, we anticipate the new Gold Coast criteria will facilitate ALS diagnosis and dispel uncertainty and confusion for patients and their families.

Clinical overlap of ALS with Other Neurodegenerative Disorders

ALS is a multifaceted disease with remarkable phenotypic heterogeneity of motor and non-motor features, as described in previous sections and more fully in our accompanying review. This complexity contributes, in part, to the difficulty of diagnosing ALS, which is rendered more challenging by clinical overlap with other more common neurological and neuromuscular diseases (Table 3). Additionally, C9orf72 repeat expansions, the most common ALS mutations in populations of European descent, are among the strongest determinants of FTD. However, the clinical phenotypes present as a continuum from pure ALS, to ALS-FTD, to pure FTD, sometimes even within the same pedigree. Further complicating the situation, C9orf72 repeat expansions are associated with movement disorders such as parkinsonism, essential tremor, and myoclonus,³² in addition to cognitive impairments. The presence of these conditions in patients may present as atypical ALS, which could contribute to a more difficult and lengthy diagnosis process. Therefore, awareness of additional manifestations of an ALS mutation could facilitate early diagnosis. Additionally, C9orf72 repeat expansions are the most frequent cause of Huntington’s disease (HD) phenocopies, patients with the classical HD phenotype but lacking characteristic huntingtin (HTT) repeat expansions and inclusions.³³ Conversely, ALS patients may harbor HTT repeat expansions simultaneously with TDP-43 inclusions,³⁴ underscoring the complexity of genotype-phenotype relationships. Understanding the spectrum of clinical presentations and overlap arising from mutations will expedite ALS diagnosis. Finally, ALS aggregates with neuropsychiatric illnesses, such as psychosis and suicide.³⁵ ALS and schizophrenia share a risk gene, GLT8D1,³⁶ as well as polygenic risk.³⁷ Therefore, one can appreciate the full spectrum of neurodegenerative and neuropsychiatric conditions that may be present an ALS patient’s family history.

Established prognostic methods

Nearly every ALS patient asks a series of questions, which usually end with “How much time do I have left?” Access to reliable prognostic methods allows physicians to give patients and their families evidence-based answers. Despite important limitations,³⁸ the ALS community currently relies on the ALS functional rating score-revised (ALSFRS-R),³⁹ a scoring system that monitors the rate of disease progression. ALSFRS-R score changes do not necessarily reflect improvement in disease; for instance, symptom management (e.g., treating sialorrhea) or medical decisions (e.g., discontinuing non-invasive ventilation) impact ALSFRS-R score, even though there is no change in the patient’s underlying disease. The ALSFRS-R multi-dimensionality limits its clinical usefulness, especially in clinical trials,⁴⁰ as well as its lack of responsiveness during plateau periods, which makes it hard to discern treatment effects in trials.⁴¹ Clinicians also derive prognostic value from respiratory tests, such as forced vital capacity (FVC);⁴² indeed, FVC is a predictive parameter in the ENCALS model (see below).

Emerging prognostic methods

Neurofilaments

Neurofilaments are neuronal cytoskeletal proteins that control neuron shape. Two markers exist, phosphorylated neurofilament heavy chain (pNFH) in cerebrospinal fluid (CSF) and neurofilament light chain (NFL) in plasma, serum, or CSF. Elevated pNFH and NFL correlate with ALS versus controls.⁵⁰ NFL levels also rise in presymptomatic individuals harboring an ALS gene one year before phenoconversion.⁵¹ Higher pNFH and NFL levels in ALS patients correlate with more aggressive disease and shorter survival, but are of low prognostic value.^50,52 Incorporating baseline NFL into mixed effects models of ALSFRS-R slopes may lower the number of ALS trial participants needed.⁵³ However, elevated neurofilaments are characteristic of neurodegenerative diseases generally⁵⁴ though they may still be relatively diagnostic of ALS;⁵⁵ thus, overall, neurofilaments remain of uncertain diagnostic and prognostic use alone, but could add value when combined with other methods.

Brain and spinal cord imaging

Functional and structural brain imaging is a rapidly growing field in ALS⁵⁶ with considerable recent progress with the advent of multisite ALS imaging protocols,⁵⁷ studies indicating feasibility for early diagnosis⁵⁸ and possibility of prognosis,^59,60 and for insight into pathogenesis, e.g., quantifying brain atrophy and connectomics (connections between brain regions). Spinal cord MRI is widely used to rule out diagnostic considerations other than ALS,⁴⁹ but more advanced diagnostic⁶¹ and prognostic⁶² applications are emerging.⁶³

MRI of ALS patients assesses tissue appearance, brain structure volumes, and diffusivity, among other factors (Appendix Table and Figure). Routine MRI does not identify persons with ALS; findings, if present, may be higher corticospinal tract (CST) and corpus callosum intensity in ALS patients.⁶⁴ A hypo-intensity of the cortical band along the precentral gyrus, called the “motor band sign”, may be characteristic of ALS and can be detected by routine susceptibility-weighted images.⁶⁵ However, advanced MRI analyses generate deeper insight using post image processing, e.g., assessing brain volumes by mapping brain regions, versus established clinical standards. Advanced MRI of ALS patients indicates, to variable degrees, atrophy in the precentral gyri, posterior cingulate cortex, thalamus, caudate, pallidum, putamen, hippocampus, and amygdala.⁶⁶ Additional MRI techniques include diffusion tensor imaging (DTI) and diffusion weighted imaging (DWI), which focus on white matter tracts. Studies consistently report changes to CST, corticopontine tract, corticorubral tract, corticostriatal pathway, and corpus callosum.^66,67 Diffusion kurtosis, an DTI adjunct, is a newer, more sensitive neuroimaging technique of white matter abnormalities, which may more accurately identify ALS patients.⁶⁸ White matter changes represent the earliest ALS findings, followed by gray matter changes.⁶⁹ Spinal cord findings in ALS suggest a drop in CST magnetization transfer ratio and potential DTI changes, although progressive atrophy and cross-sectional area may be the most accurate biomarkers.⁶³

The complexity of ALS pathology advocates multimodal MRI, which combines multiple advanced MRI techniques. Multimodal MRI of both brain volume and white matter integrity has 85·7% sensitivity and 78·4% accuracy for discriminating ALS from control scans.⁷⁰ A multi-site Italian study evaluated global and lobar connectivity in ALS using DTI, fractional anisotropy (white matter track integrity measure), and resting state functional MRI.⁷¹ The results found widespread connectomics dysfunction with early degeneration of brain motor regions followed by breakdown in functional connections, leading to cognitive decline.⁷¹ Multimodal longitudinal MRI can monitor spatiotemporal spread in ALS via the brain connectome and potentially serve as a disease biomarker.⁶⁷ Finally, quantitative susceptibility mapping MRI measures iron accumulation in motor cortex,⁷² which can be coupled with white matter assessments (DTI, DWI, diffusion kurtosis) to identify early track changes associated with metal toxicity in ALS. Similarly, multimodal MRI of spinal cord in ALS participants leveraged fractional anisotropy, magnetization transfer ratio, and cross-sectional area to build a survival prediction model.⁶²

PET imaging is another modality that may facilitate ALS diagnosis and prognosis (Appendix Table). [¹⁸F]-FDG PET to measure glucose metabolism reported hypometabolism in the frontal cortex and hypermetabolism in the temporal cortex, cerebellum, and brainstem in a two-site study of ALS patients.⁷³ PET [¹¹C]-PBR28 brain uptake, a surrogate of microglial activation, is elevated in the bilateral precentral and paracentral gyri of ALS patients versus controls, and colocalizes with cortical thinning, as assessed by integrated MRI imaging,⁷⁴ but may not correlate with clinical progression.⁷⁴ Integrating spinal cord with brain in [¹⁸F]-FDG PET allows differentiation of ALS from ALS mimics.⁷⁵

Overall, tremendous progress has been made in advanced brain MRI and PET along with more emerging advanced spinal cord imaging applications, which feasibly could improve diagnosis^58,61 and prognosis.^59,60,62 Although we anticipate imaging will be useful as an adjunct to existing methods, additional research is required to evaluate how to integrate imaging into eventual clinical care and at the single patient level. Furthermore, most imaging studies focused on imaging in ALS versus controls; however, future studies will need to include patients with ALS mimic disorders to better evaluate sensitivity and specificity.^58,75

Spectral electroencephalogram (EEG) mapping and magnetoencephalography

Electrophysiological measures are another technique to assess brain networks. High-density spectral EEG mapping measures the coherence of certain frequency bands between brain regions, generating a functional measure of brain connectivity in ALS.^76,77 EEG changes occur to brain connectivity in both motor and non-motor systems, confirming ALS is not a pure motor disease, in agreement with MRI connectomics.⁷⁷ Magnetoencephalography shows brain networks become increasingly connected during ALS progression, indicating a dysfunctional, modified brain topology.⁷⁸ These findings are significant because reorganization of brain connections could potentially predict disease spread.⁶⁷ ALS connectomics studies are needed coupling multimodal MRI, high-density spectral EEG, and magnetoencephalography to further understand how brain structural changes and corresponding connectivity changes associate with ALS symptomatology and disease course. EEG and magnetoencephalography connectomics are very novel techniques not presently in clinical use and their potential as diagnostic and prognostic tools remains unknown.

Hyperexcitability

Excessive cortical excitability, i.e., hyperexcitability, in ALS is increasingly recognised as a pathophysiological mechanism of the neurodegenerative cascade.⁷⁹ Clinically, hyperexcitability manifests as fasciculations combined with upper motor neuron features of increased tone and hyperreflexia.⁸⁰ Hyperexcitability is linked to excitotoxicity from excessive glutamate receptor activity at the synaptic cleft, leading to motor neuron death.^23,81 Cortical motor neuronal hyperexcitability can be captured by transcranial magnetic stimulation (TMS) techniques.⁸² A TMS coil is placed over the motor cortex and responses are recorded from the contralateral hand in the abductor pollicis brevis muscle. Short interval intracortical inhibition (SICI) and short interval intracortical facilitation (SICF) are extracted and represent interneuron function.

There is presymptomatic decrease in SICI and increase in SICF in ALS.⁸³ TMS detects cortical hyperexcitability across a range of ALS phenotypes and differentiates ALS from non-ALS disorders with sensitivity (73·21%) and specificity (80·88%) at early disease stages.⁸⁴ TMS distinguishes ALS, with cortical hyperexcitability predominance, from PLS, with cortical inexcitability predominance.⁸⁵ TMS can also investigate pathologic spread using hyperexcitability as a surrogate by recording responses to TMS at the tibialis anterior in addition to the abductor pollicis brevis. Analysis of ALS patients demonstrates heterogeneity in cortical dysfunction by body region; cortical hyperexcitability predominates in the upper limbs, whereas cortical inexcitability predominates in the lower limbs versus controls.⁸⁶ Furthermore, cortical hyperexcitability correlates with the clinically affected body region; ALS patients exhibit focal asymmetry at the onset site early in disease, but widespread hyperexcitability alterations in late stages.⁸⁷ Cortical motor hyperexcitability may also detect cognitive dysfunction and cortical resting motor threshold distinguishes pure ALS, ALS-FTD, and pure FTD.⁸⁸

The role of TMS in prognosis is less established compared to diagnosis. A recent longitudinal study of suspected ALS participants found cortical hyperexcitability increases with longer disease duration, indicating a potential link to ALS progression.⁸⁹ Cortical inexcitability may predict a poorer clinical trajectory in ALS with inexcitability in all four limbs correlating with younger age, lower limb onset, greater extent of functional disability, and more rapid disease progression.⁹⁰ Thus, cortical hyperexcitability may improve our ability to predict clinical outcomes. It could also serve as a biomarker for drug activity, e.g., in clinical trials of ezogabine, an activator of voltage gated potassium channels.⁹¹

Presently, TMS is not in clinical use, although it does appear to offer some diagnostic and prognostic utility and likely will be informative as an adjunct to pre-existing methods. However, future research will determine the full clinical and research potential of TMS in ALS, and determine whether this novel, electrophysiological assessment will become a fully accepted disease biomarker.

Machine learning in ALS

ALS is a highly heterogenous syndrome of various genetic and unknown etiologies converging on a “typical” ALS phenotype with diverse clinical presentations. Machine learning approaches can analyze large datasets (e.g., clinical, demographic, electrophysiological, imaging, morphology) in an agnostic, data-driven manner to develop diagnostic and prognostic models.⁹² Tang et al. employed unsupervised clustering of clinical data encompassing 8,000 patients, 3 million records, and 200 clinical features over 12 months from the Patient Data Pooled Resource Open-Access ALS Clinical Trials Database archive.⁹³ Unsupervised clustering yielded four consistent computable phenotypes, defined by slope change in ALSFRS-R, with over 95% diagnostic accuracy, based on multivariate features. Deep learning modeling, a form of machine learning, was used for prognosis, and predicted ALS patient survival in this cohort when incorporating TDP-43 aggregation and morphology and MRI connectivity data with clinical characteristics.⁶⁷ Further research will determine whether machine learning may unlock a way forward for diagnosing and prognosticating for ALS patients at the individual level by integrating multi-domain information into diagnostic and prediction models.

Summary

Overall, most of the recent novel diagnostic and prognostic ALS tests are limited to the research setting. Further studies are needed to determine whether these approaches will be useful in a clinical real-world setting. This will entail studies enrolling participants with diseases mimicking ALS and longitudinal studies against validated prognostic scales to evaluate their potential for improved ALS diagnosis (sensitivity/specificity) and prognosis. Additionally, it will be necessary to determine how to apply findings made from large cohort studies to individual patient diagnosis and prognosis. Until more specific and sensitive tests are developed, ALS diagnosis will remain an integrative and iterative process reliant on clinical history, physical examination, and/or confirmatory clinical electrodiagnostic tests.