What the Early Examination Can—and Cannot—Predict

After securing the airway, circulation, cord perfusion, temperature, thromboembolic protection, bowel and bladder management, skin integrity, and early rehabilitation access, the next bedside task is answering the question patients and families ask almost immediately: “What will come back?” The responsible answer begins by separating neurologic impairment from function and function from quality of life. AIS conversion is not synonymous with hand use, independent transfers, walking, continence, employment, or a life the patient considers worthwhile. Evidence available through August 13, 2026 supports no single acute variable as a sufficiently reliable stand-alone predictor of those outcomes. The Neurocritical Care Society guideline identified initial injury severity, neurological level of injury, and conventional MRI pathology as only moderately reliable predictors of later AIS conversion, and the Dutch Clinical Prediction Rule as a moderately reliable model for one-year independent ambulation; it identified no other reliable or moderately reliable acute predictors of mortality or broad functional outcome. In the guideline’s terminology, “moderately reliable” means that individual predictors should be combined with other credible findings and the clinical context, and that even a concordant forecast should be communicated as “likely” with substantial uncertainty—not as destiny (Mahanes et al., 2024).

The foundation is a technically valid ISNCSCI examination. ASIA released the ninth-edition standards in 2026, so institutions should use current materials and record the edition employed; this matters because nearly all prognostic cohorts were classified under earlier editions, most commonly the 2019 revision (American Spinal Injury Association, 2026; Rupp et al., 2021). The neurological level is not the vertebral fracture level or the MRI epicenter. It is the most caudal segment with normal sensory and motor function bilaterally and, operationally, the most rostral of the right sensory, left sensory, right motor, and left motor levels. AIS A means no sensory or motor function in S4–S5; it means neurologically complete by the sacral-sparing definition, not histologic transection and not proof that no axons cross the lesion. AIS B requires sacral sensory sparing but no motor function more than three levels below the ipsilateral motor level on either side. AIS C and D are motor-incomplete injuries, established by voluntary anal contraction or by sensory incompleteness together with motor preservation more than three levels below the ipsilateral motor level; C versus D depends on whether fewer than half or at least half of the key muscles below the single neurological level have grade 3 or greater strength. AIS E denotes a normal ISNCSCI examination in someone who previously had SCI deficits; it does not mean that pain, dysautonomia, sexual dysfunction, fatigue, fine-motor impairment, or every participation limitation has resolved.

The anorectal examination is therefore prognostically decisive, not ceremonial. Light touch and pinprick at S4–S5, deep anal pressure, and voluntary anal contraction must be documented separately. Deep anal pressure is conscious pressure perception, not merely discomfort produced by the examination, and voluntary contraction must be distinguished from a reflex contraction. When the examination otherwise suggests AIS B, non-key muscles below the motor level should be tested because preserved voluntary movement sufficiently caudal to the motor level changes the classification to motor incomplete. Absence of the bulbocavernosus reflex during spinal shock does not itself make the injury AIS A and is not a substitute for testing sacral sparing.

The first examination should occur as soon as feasible because it establishes baseline status, identifies deterioration, and documents the phenotype before and after major interventions, but it should not automatically become the definitive prognostic examination. Sedation, intoxication, delirium, traumatic brain injury, hypothermia, pain, long-bone or shoulder fractures, plexus or peripheral nerve injury, pre-existing neuropathy, immobilization, language barriers, and ongoing resuscitation can all create false weakness or unreliable sensory findings. A muscle that cannot be tested because of a non-SCI condition should be recorded as not testable with the reason, not converted reflexively to zero.

The Neurocritical Care Society recommends deferring prognostic use of the neurological level and AIS grade until approximately 72 hours; an examination then is generally more predictive than an admission examination, and previous guidance favored a comprehensive examination between days 3 and 7. This does not mean waiting 72 hours to examine the patient—it means repeating the examination after confounders have been reduced and attaching the forecast to the more dependable result. Exact time after injury, relationship to decompression, analgesic and sedative exposure, examiner confidence, upper- and lower-extremity motor scores, sensory subscores, sacral findings, neurological level, AIS grade, and zones of partial preservation should travel together in the record. A one-grade change that appears when pain resolves or a borderline muscle moves from 2 to 3 should be confirmed before it is interpreted as biological recovery.

The zone of partial preservation adds information that the single AIS letter discards. Under the revised framework, a motor ZPP can be recorded on a side when voluntary anal contraction is absent, and a sensory ZPP can be recorded on a side lacking the relevant sacral sensory sparing. Its caudal extent should be documented rather than summarized as “some movement below the level” (Schuld et al., 2024). Recent registry evidence makes this clinically consequential: among initially AIS A patients with follow-up data, a motor ZPP extending more than three levels below the motor level—a phenotype resembling motor-incomplete injury despite absent sacral sparing—was associated with subsequent conversion to motor-incomplete status in approximately 43% to 54%, compared with about 13% among those without that pattern, depending on the motor-level definition used. This is an association from a follow-up subset, not permission to relabel the initial injury or promise conversion, but it is a strong reason not to reduce AIS A to a binary statement of “nothing preserved” (Kirshblum et al., 2025).

Recovery is also time-dependent. Across historical cohorts, the velocity of motor and AIS improvement is greatest during the first three months, most recovery occurs within six to nine months, and smaller gains may continue through 12–18 months or later. Early improvement can reflect resolution of edema, hemorrhagic mass effect, ischemia, ionic conduction block, and neuronal or axonal “stunning,” followed by remyelination, synaptic reweighting, sprouting, activity-dependent plasticity, strengthening, and compensation. Therefore, early AIS conversion is not proof of axonal regeneration, and the absence of dramatic conversion during the first week is not proof that later functional gains are impossible (Kirshblum et al., 2021).

Initial severity nonetheless changes probability. In the guideline synthesis, fewer than one third of patients initially classified AIS A improved by at least one AIS grade at one year, while AIS conversion was four- to eightfold more likely after an initial AIS B–D injury than after AIS A. In one 931-patient motor-complete cohort, approximately 21% of those initially AIS A improved compared with 69% initially AIS B. Those figures are population estimates affected by examination timing, injury level, treatment era, attrition, and case mix; they should anchor counseling, not be presented as an individualized percentage. Neurological level also modifies the chance of AIS conversion: lumbar and thoracolumbar or conus-region injuries convert more often than cervical or thoracic injuries in available cohorts. At the conus, recovery of roots and mixed cord–root anatomy complicates comparison with a purely thoracic cord lesion. Older age is associated on average with less recovery and greater functional burden in several datasets, but age alone is not a reliable predictor of acute mortality, long-term mortality, one-year function, or 12- to 24-month AIS conversion; there is no defensible bedside age threshold beyond which meaningful recovery should be declared impossible.

Functional prognosis must then move beyond the AIS letter. An AIS A-to-B conversion may add sacral sensation without useful limb motor recovery, while an apparently modest segmental motor gain in tetraplegia may transform independence. C5 elbow flexion supports feeding strategies; C6 wrist extension enables tenodesis and changes transfers and device use; C7 elbow extension can materially improve pressure relief and transfers; C8 finger flexion and T1 intrinsic hand function reshape grasp and dexterity. Conversely, an AIS C-to-D conversion can occur because several muscles cross grade 3 without necessarily producing safe community walking. AIS is an ordinal classification, not a linear functional scale.

Serial upper- and lower-extremity motor scores should therefore be paired with patient-centered measures: GRASSP or another detailed upper-limb assessment in tetraplegia, SCIM for self-care, respiration, sphincter management and mobility, and WISCI II plus walking speed or endurance when ambulation emerges. The target outcome should be chosen with the patient rather than assumed. Surveys summarized in the guideline found arm and hand recovery to be the highest priority among people with tetraplegia, sexual function the highest among those with paraplegia, and bowel and bladder recovery highly valued by both groups. Walking is important, but it is not a universal proxy for recovery.

When walking is the specific question, the Dutch Clinical Prediction Rule is the best-supported early model. It combines age dichotomized at 65 years, L3 quadriceps and S1 gastrocnemius motor scores, and light-touch sensation in the L3 and S1 dermatomes, using an examination obtained within 15 days. The resulting score ranges from −10 to 40 and estimates the probability of independent indoor walking at one year, with or without an assistive device. The original and several external-validation cohorts showed excellent discrimination, often with an area under the curve above 0.95 (van Middendorp et al., 2011). That performance does not make the output a personal certainty. Calibration varies across settings, performance is less secure in some AIS B/C subgroups, the age term is an interaction within the model rather than evidence that age alone determines recovery, and “independent walking” does not necessarily mean community distance, normal speed, freedom from falls, or preference for walking over wheelchair mobility. It predicts neither hand function nor continence, ventilation, total SCIM independence, participation, or quality of life. The result should be described exactly as the guideline advises: an objective estimate subject to considerable uncertainty.

MRI supplies a different layer of information by depicting the injured substrate rather than the patient’s performance. Intramedullary hemorrhage, longer edema or lesion length, greater axial tissue destruction, and persistent compression generally associate with worse neurological recovery. The absence of pathological cord findings is a moderately reliable favorable predictor of AIS conversion, although that recommendation is weak and supported by low-certainty evidence. Published thresholds such as a particular edema length or percentage compression arose from individual cohorts and should not be universalized. MRI changes with time, field strength, sequence, positioning, surgery, and decompression; a preoperative image and a postoperative image are not interchangeable prognostic tests. Signal abnormality is not histology, and even dramatic hemorrhage or apparent discontinuity cannot by itself prove that no functional axons remain.

Midsagittal “tissue bridges”—preserved anterior and posterior cord tissue spanning the lesion—are a promising refinement. In a multicenter retrospective cohort of 227 people with cervical SCI, bridge width measured on T2 imaging at approximately three to four weeks added information about three- and 12-month sensorimotor recovery beyond clinical variables (Pfyffer et al., 2024). Because measurement occurred weeks after injury and model-derived cut points require prospective validation, tissue bridges are best treated as evidence of spared neural substrate and a potential model component, not as a stand-alone day-one threshold for treatment limitation. Diffusion imaging, magnetization-transfer techniques, spectroscopy, and functional MRI remain research tools rather than validated individual prognostic tests.

Neurophysiology can be helpful when the behavioral examination is limited or when one wants evidence of conduction across the lesion, but timing and confounders matter. Motor-evoked potentials, somatosensory-evoked potentials, and nerve-conduction studies are influenced by anesthesia and sedatives, temperature, edema, peripheral nerve injury, technical quality, and the muscles and pathways sampled. In a 224-patient cervical SCI study, neurophysiology obtained within 40 days improved a clinical model’s discrimination for complete functional recovery from an area under the curve of 0.936 to 0.956; that is a statistically meaningful but clinically modest incremental gain, obtained far too late and under too variable conditions to make an absent ultra-early response synonymous with irreversibility (Hupp et al., 2018).

Blood and CSF biomarkers are similarly promising but not ready to adjudicate an individual future. Neurofilament light and GFAP concentrations correlate with tissue injury, initial severity, and later AIS conversion at the cohort level, but overlap remains substantial and no standardized threshold has been externally validated for bedside counseling or withdrawal decisions (Stukas et al., 2023). A 2026 scoping review confirms the broader pattern: many clinical, imaging, electrophysiological, biomarker, and machine-learning models have been proposed, but external validation, calibration, transportability, consistent outcome definitions, and evidence of clinical utility remain inadequate for most (Boyles et al., 2026).

The most defensible bedside forecast is therefore conditional, endpoint-specific, time-stamped, and repeatedly updated. A useful formulation is:

“On the reliable examination at 72 hours, this is a neurologically complete thoracic injury with no sacral sparing; the MRI also shows intramedullary hemorrhage. Those findings make motor recovery and independent walking less likely than after an incomplete injury, but neither proves anatomic transection or makes recovery impossible. Most neurological change occurs over months, particularly the first three, and we will revise the forecast using serial examinations.”

For motor-incomplete cervical injury, one might instead say:

“Preserved motor function below the injury gives a meaningfully greater probability of additional recovery, but today’s examination cannot determine community walking or hand independence; the distribution and trajectory of individual muscle recovery will be more informative than the AIS letter alone.”

The discussion should include the most likely course, credible better and worse courses, expected near-term dependence, and realistic avenues to autonomy through neurological recovery, rehabilitation, assistive technology, environmental adaptation, reconstructive procedures, and caregiver training. Prognostic statements should specify whether they refer to AIS conversion at 6–12 months, independent indoor walking at one year, a particular hand function, ventilator liberation, or global independence; these endpoints cannot be substituted for one another. They should also be explicitly conditional on continued treatment, because decompression, prevention of secondary injury, complications, rehabilitation exposure, and practice-dependent plasticity influence the observed outcome and were incompletely captured in most prognostic datasets.

Finally, no AIS grade, MRI appearance, biomarker, or walking model should independently drive withdrawal of life-sustaining treatment. Premature pessimism can create a self-fulfilling prophecy through treatment limitation, loss of rehabilitation engagement, and the mistaken assumption that severe disability precludes acceptable quality of life. The neurointensivist’s job is not to withhold a difficult forecast; it is to distinguish what is likely from what is known, what remains biologically possible from what is probable, and neurological recovery from the many ways a person can regain agency.

At the bedside, the practical synthesis is to obtain and repeat a meticulous ISNCSCI examination, state and correct its confounders, preserve modality-level sacral and ZPP data, combine severity, level, MRI, and—when relevant—the DCPR, measure outcomes that matter to the individual, and revise the forecast as the trajectory declares itself.

References