Skip to main content

Foundational Trials

NASCIS III

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Administration of Methylprednisolone for 24 or 48 Hours or Tirilazad Mesylate for 48 Hours in the Treatment of Acute Spinal Cord Injury: Results of the Third National Acute Spinal Cord Injury Randomized Controlled Trial
  • Acronym: NASCIS III
  • Year: 1997
  • Journal published in: JAMA
  • Citation: Bracken MB, Shepard MJ, Holford TR, et al. Administration of Methylprednisolone for 24 or 48 Hours or Tirilazad Mesylate for 48 Hours in the Treatment of Acute Spinal Cord Injury: Results of the Third National Acute Spinal Cord Injury Randomized Controlled Trial. JAMA. 1997;277(20):1597–1604.

Context & Rationale

  • Background
    • Secondary injury cascades after traumatic spinal cord injury (SCI) include lipid peroxidation, inflammatory amplification, and microvascular dysfunction; these were hypothesised to be modifiable in the early post-injury window.
    • NASCIS II popularised a 24-hour high-dose methylprednisolone sodium succinate (MPSS) regimen (30 mg/kg bolus then 5.4 mg/kg/h infusion) when initiated within 8 hours, but the evidentiary basis rested heavily on time-to-treatment subgroup analyses and generated sustained methodological criticism.
    • Tirilazad mesylate (a 21-aminosteroid “lazaroid” free-radical scavenger) was developed as a non-glucocorticoid strategy to target lipid peroxidation, potentially avoiding steroid-related adverse effects.
    • The optimal duration of high-dose MPSS after the initial bolus, and whether duration should vary by time to treatment within the 8-hour window, remained uncertain.
  • Research Question/Hypothesis
    • In patients with acute traumatic SCI treated within 8 hours, does (a) extending MPSS infusion from 24 to 48 hours, or (b) substituting tirilazad mesylate for the MPSS maintenance phase, improve neurological and functional recovery versus the 24-hour MPSS regimen?
  • Why This Matters
    • Even modest improvements in motor recovery after SCI could translate into major lifetime gains in independence and societal participation.
    • High-dose steroids carry important risks (infective, gastrointestinal, metabolic); duration-dependent harm was plausible, making comparative duration trials high stakes.
    • NASCIS III sought to refine (or de-escalate) a therapy already influencing practice and medicolegal norms, and to test a mechanistically motivated alternative (tirilazad).

Design & Methods

  • Research Question: Among patients with acute traumatic SCI treated within 8 hours, does 48-hour MPSS or 48-hour tirilazad (versus 24-hour MPSS) improve neurological and functional outcomes?
  • Study Type: Multicentre, randomised, double-blind, controlled, three-arm trial conducted in 16 North American SCI centres.
  • Population:
    • Setting: Participating SCI centres; enrolment required rapid transfer and protocol initiation in the hyperacute window.
    • Key inclusion criteria: Age ≥14 years; acute traumatic SCI with neurological deficit; clinical diagnosis within 8 hours; randomised within 6 hours to allow protocol start within 8 hours.
    • Key exclusion criteria: Gunshot wounds; body weight >109 kg; pregnancy; “illegal immigrants” or indicted criminals; comorbid illness likely to influence neurological assessment/outcomes; inability to provide consent/participate in follow-up (including language barriers and chronic steroid use were reported reasons for non-enrolment).
  • Intervention:
    • Common to all groups (pre-randomisation): MPSS 30 mg/kg IV bolus (with additional bolus if the initial dose was <20 mg/kg; reported as uncommon).
    • 48-hour MPSS protocol: Continuous MPSS infusion at 5.4 mg/kg/h for 47 hours (following the bolus).
    • 48-hour tirilazad protocol: Tirilazad 2.5 mg/kg IV every 6 hours for 48 hours (administered in a masked infusion system alongside placebo(s) as required for blinding).
    • Timing: Maintenance protocol initiation within 3 hours of the bolus.
  • Comparison:
    • 24-hour MPSS protocol: Continuous MPSS infusion at 5.4 mg/kg/h for 23 hours (following the bolus), with matched placebo(s) for non-assigned agents.
  • Blinding: Double-blind; MPSS and tirilazad (or matching placebos) were administered in separate masked preparations; randomisation codes were stored centrally and were reported as not broken during the trial.
  • Statistics: Powered to detect a ≥5-point difference in mean motor change score with α=0.05 and β=0.20 (80% power), requiring 150 patients per group (total 450); planned sample size 485, with 499 randomised. Primary analyses used ANCOVA adjusting for baseline motor score and included protocol-by-time parameters; analyses were reported for a modified intention-to-treat cohort treated within 8 hours and for a per-protocol “compilers” cohort.
  • Follow-Up Period: Neurological and functional assessments at 6 weeks (42–49 days) and 6 months (180–210 days); survival follow-up to approximately 6 months.

Key Results

This trial was not stopped early. Enrolment proceeded to the planned sample size range (499 randomised; 485 planned).

Outcome 48 h protocol 24 h MPSS Effect p value / 95% CI Notes
Change in motor score at 6 weeks (treated within 8 h; modified ITT) 11.8 (48 h MPSS) 9.0 Not reported (ANCOVA comparison) P=0.09 ANCOVA adjusted for baseline motor score; comparison 48 h MPSS vs 24 h MPSS
Change in motor score at 6 months (treated within 8 h; modified ITT) 16.8 (48 h MPSS) 13.4 Not reported (ANCOVA comparison) P=0.07 Borderline result; primary inference in publication did not meet P<0.05 threshold overall
Change in motor score at 6 weeks (3–8 h subgroup; modified ITT) 12.5 (48 h MPSS) 7.6 Not reported (ANCOVA comparison) P=0.04 Subgroup defined by time to bolus; key driver of the trial’s clinical message
Change in motor score at 6 months (3–8 h subgroup; modified ITT) 17.6 (48 h MPSS) 11.2 Not reported (ANCOVA comparison) P=0.01 Signal confined to this subgroup; no benefit in the <3 h subgroup
Total FIM score at 6 months (treated within 8 h; modified ITT) 103.3 (48 h MPSS) 99.1 Not reported (ANCOVA comparison) P=0.08 Functional significance of small absolute differences debated
Improvement by ≥1 FIM functional category at 6 months (treated within 8 h; modified ITT) 67.2% (48 h MPSS) 53.0% RR 1.28 95% CI 1.05 to 1.57; P=0.02 Binary categorisation of functional change; reported as relative risk
Severe pneumonia (treated within 8 h) 5.8% (48 h MPSS) 2.6% Not reported P=0.02 P value reported for between-protocol comparison; infections numerically highest with 48 h MPSS
Severe sepsis (treated within 8 h) 2.6% (48 h MPSS) 0.6% Not reported P=0.07 Numerical excess with 48 h MPSS; trial-level power for harms was limited
Change in motor score at 6 months (treated within 8 h; modified ITT) 13.2 (48 h tirilazad) 13.4 Not reported (ANCOVA comparison) P=0.58 Motor recovery with tirilazad was comparable to 24 h MPSS
Improvement by ≥1 FIM functional category at 6 months (treated within 8 h; modified ITT) 58.5% (48 h tirilazad) 53.0% RR 1.10 95% CI 0.90 to 1.35; P=0.35 No evidence of functional superiority versus 24 h MPSS
  • Across all patients treated within 8 hours, extending MPSS to 48 hours did not achieve conventional statistical significance for motor recovery versus 24 hours (P=0.09 at 6 weeks; P=0.07 at 6 months).
  • The principal “positive” finding was confined to the 3–8 hour subgroup (motor change at 6 months: 17.6 with 48 h MPSS vs 11.2 with 24 h MPSS; P=0.01), creating a subgroup-driven interpretation pathway.
  • 48 h MPSS was associated with higher severe pneumonia (5.8% vs 2.6%), reinforcing concerns that any small neurological signal must be weighed against clinically important infectious harm.

Internal Validity

  • Randomisation and allocation: Blocked randomisation (blocks of 9) within centre; central pharmacy prepared masked trial kits, supporting allocation concealment.
  • Dropout/exclusions (post-randomisation): 499 randomised (166/167/166); administration errors occurred (7/14/9 across 24 h MPSS/48 h tirilazad/48 h MPSS); infusion stopped due to complications in 0/3/2; follow-up neurological exams were available in most survivors (6-week exams: 154/157/154; 6-month exams: 145/150/149), with deaths balanced (9 per group by 6 months).
  • Performance/detection bias: Double blinding with matched infusions reduced bias; however, neurological examination-based outcomes are operator-dependent and susceptible to measurement error, particularly in the early post-injury period.
  • Protocol adherence: Among patients treated within 8 hours, reported receipt of >97% of assigned MPSS dose in both MPSS protocols, and 94.3% of assigned dose in the tirilazad protocol; non-adherence and early discontinuation reduce between-group separation.
  • Baseline characteristics: Broadly similar demographics and injury patterns, but important imbalances existed, including lower baseline expanded motor score in the tirilazad group (26.5 vs 33.9 vs 30.5) and lower diastolic blood pressure (71.6 vs 76.0 vs 76.8 mm Hg), raising concern for prognostic imbalance despite adjustment.
  • Heterogeneity: Mixed injury severity (approximately half complete) and level (cervical predominance); such heterogeneity is typical for SCI trials but can dilute treatment effects and complicate subgroup inference.
  • Timing: Mean time from injury to bolus was approximately 3.1–3.4 hours; analyses were emphasised by a ≤3 vs 3–8 hour split, and the clinical conclusions depended heavily on this time-stratified framing.
  • Dose: The 48-hour MPSS regimen substantially increased total steroid exposure versus 24 hours (same bolus, longer high-dose infusion), plausibly increasing immunosuppressive and metabolic harms.
  • Separation of the variable of interest: All patients received a pre-randomisation MPSS bolus; separation therefore concerned the maintenance phase (23 vs 47 hours of infusion) and/or substitution with tirilazad. Open-label bolus prior to randomisation was common (68.1%–74.2%), which may have introduced additional variability in early exposure.
  • Key delivery aspects: Rapid enrolment and protocol initiation were achieved in a substantial proportion, but excluding patients who did not receive study drug within 8 hours produced a modified intention-to-treat population and risks post-randomisation selection effects.
  • Adjunctive therapy use: Not reported in sufficient detail to assess whether co-interventions were balanced (e.g., haemodynamic augmentation, surgical timing, infection prophylaxis).
  • Outcome assessment: Outcomes were prespecified and clinically relevant (motor score change; Functional Independence Measure), but the translation of small motor score differences into patient-valued function is not linear.
  • Statistical rigour: The trial met its target sample size for the hypothesised effect size, but inference was complicated by modified intention-to-treat definitions, multiple subgroup comparisons, and reliance on per-protocol (“compilers”) results for stronger P values.

Conclusion on Internal Validity: Overall, internal validity is moderate: randomisation and blinding were strong, follow-up was relatively complete, and dosing separation was substantial; however, modified intention-to-treat exclusions, baseline imbalances, and subgroup-driven inference meaningfully limit confidence in the causal interpretation of the “positive” signal.

External Validity

  • Population representativeness: Predominantly young, male, blunt-trauma SCI patients in high-resource centres; penetrating injuries (gunshot) were excluded and early consent/transfer was required.
  • Applicability: Findings are most applicable to closed traumatic SCI managed in systems capable of protocol initiation within 8 hours; generalisability to delayed presentation, penetrating injury, very high body weight, or settings with limited ICU/trauma resources is limited.
  • Contemporary relevance: Changes in modern SCI care (prehospital systems, surgical timing, critical care practices, infection prevention) may modify both baseline prognosis and the risk profile of high-dose steroids.

Conclusion on External Validity: External validity is moderate for hyperacute, closed traumatic SCI in high-income trauma systems, but limited for patients outside the early treatment window or with excluded injury mechanisms and comorbidity profiles.

Strengths & Limitations

  • Strengths:
    • Large multicentre, double-blind randomised design with central drug preparation and strong masking procedures.
    • Prospectively defined dosing regimens with high reported adherence among those treated within the protocol window.
    • High follow-up completeness among survivors, with structured neurological and functional outcome assessments at clinically relevant time points.
  • Limitations:
    • Common pre-randomisation MPSS bolus to all groups limited the ability to assess “steroids vs no steroids” and may have attenuated between-group differences.
    • Primary conclusions were driven by time-to-treatment subgroup analyses, with a cutpoint that may have been data-influenced and without robust control for multiplicity.
    • Modified intention-to-treat exclusions (patients not treated within 8 hours) and the use of per-protocol (“compilers”) analyses risk bias.
    • Baseline imbalances (notably lower baseline motor score in the tirilazad group) complicate interpretation despite adjustment.
    • Adverse event detection was reported, but the trial was not powered for harms; nevertheless, infection signals were clinically important.

Interpretation & Why It Matters

  • Clinical implications
    • NASCIS III did not provide compelling overall evidence that extending MPSS to 48 hours improves neurological recovery compared with 24 hours when all patients treated within 8 hours are analysed.
    • The trial’s perceived “practice-changing” message emerged from subgroup comparisons and must be interpreted with caution given multiplicity and the fragility of borderline P values.
    • Given the observed infection signal with 48-hour MPSS, and later guideline shifts, routine prolonged high-dose MPSS should not be considered evidence-based standard care in acute SCI.

Controversies & Subsequent Evidence

  • Subgroup-driven inference (time-to-treatment): The overall modified intention-to-treat analyses did not show statistically significant superiority of 48-hour MPSS over 24-hour MPSS, yet the paper’s clinical conclusion relied heavily on improved outcomes in the 3–8 hour subgroup; critiques emphasised that this effectively constructed a clinical rule (“≤3 hours: 24 h; 3–8 hours: 48 h”) from within-trial subgroup contrasts rather than from the overall randomised comparison.12
  • Data-informed subgroup boundary and absence of a robust interaction framework: The 3-hour split was justified in the trial report as the modal time to bolus administration; published critique argued that privileging a boundary derived from the observed timing distribution, combined with multiple subgroup contrasts, inflates false-positive risk and should not drive practice recommendations without clear evidence of a treatment-by-time interaction.124
  • Multiplicity across arms, outcomes, and timepoints: NASCIS III compared three protocols, evaluated outcomes at both 6 weeks and 6 months, and highlighted subgroup contrasts; critical appraisal argued that this multiplicity makes isolated P values near 0.05 statistically fragile and increases the probability of chance findings being interpreted as clinically actionable evidence.12
  • Post-randomisation exclusions and “compilers” analyses: Key efficacy analyses were presented for a “treated within 8 hours” modified intention-to-treat cohort and, separately, a per-protocol (“compilers”) cohort; stronger P values in the compilers analysis were criticised as vulnerable to bias from post-randomisation selection and non-adherence, particularly when overall effects were borderline.24
  • Clinical meaningfulness vs statistical signal: Published critiques argued that motor score differences of the magnitude observed (even when statistically significant in the 3–8 hour subgroup) have uncertain translation into patient-valued function and may be sensitive to modelling choices; categorised functional outcomes (e.g., “improved by ≥1 functional category”) were viewed as potentially magnifying small differences in continuous endpoints.12
  • Benefit–harm imbalance and the “standard of care” claim: NASCIS III demonstrated higher severe pneumonia (and numerically higher severe sepsis) with 48-hour MPSS, and subsequent published critiques stated that MPSS could not be considered a standard of care (and that 48-hour therapy should not be recommended) given weak overall efficacy signals and credible mechanisms for clinically important harm.31
  • Adoption, medicolegal drivers, and deimplementation: MPSS became embedded in practice and medicolegal discourse despite contested efficacy; surveys documented widespread prescribing even when clinicians did not believe benefit was established, with medicolegal justification prominent—highlighting how subgroup-driven interpretations can be implemented and then only slowly reversed as evidence and guidance evolve.5
  • Guideline reversal and modern positioning: Evidence-based neurosurgical guidelines later recommended against MPSS for acute SCI, consolidating a shift away from “standard of care” framing and supporting deimplementation of routine high-dose steroid protocols (including prolonged courses).6
  • Subsequent evidence synthesis and persistent heterogeneity: Later systematic review work underpinning subsequent guideline efforts concluded that any neurological benefit of MPSS (particularly the NASCIS II 24-hour regimen initiated within 8 hours) is at most small and must be weighed against complications and patient preferences, helping explain ongoing international variation rather than uniform re-adoption.7

Summary

  • NASCIS III randomised 499 patients with acute traumatic SCI to 24-hour MPSS, 48-hour MPSS, or 48-hour tirilazad, with a common pre-randomisation MPSS bolus.
  • In the overall analysis of patients treated within 8 hours, 48-hour MPSS did not show statistically significant improvement in motor recovery versus 24-hour MPSS (P=0.07 at 6 months).
  • The trial’s influential “positive” message depended on a time-to-treatment subgroup (3–8 hours), where 48-hour MPSS showed greater motor score change (17.6 vs 11.2 at 6 months; P=0.01), alongside higher severe pneumonia (5.8% vs 2.6%).
  • Methodological critiques focused on subgroup interpretation, multiplicity, and post-randomisation exclusions, arguing that MPSS (particularly 48-hour regimens) should not be considered standard care.
  • Subsequent evidence-based guidelines recommended against MPSS in acute SCI, driving deimplementation and leaving a legacy of persistent debate over small, time-dependent neurological signals versus meaningful harm.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • NASCIS III is historically important for illustrating how borderline overall results and time-window subgroup analyses can strongly influence practice—even when later evidence synthesis and guideline appraisal support deimplementation.
  • When citing NASCIS-era interventions in modern protocols or education, explicitly separate (a) neurological “signal” size, (b) patient-centred functional meaning, and (c) complication rates; these dimensions diverge in high-dose steroid trials.

Overall Takeaway

NASCIS III is “landmark” primarily for how it shaped (and later reshaped) practice: the overall trial did not show clear benefit of 48-hour MPSS over 24-hour MPSS, yet time-window subgroup findings drove widespread uptake of prolonged steroids. The prolonged-regimen harm signal, methodological critique, and subsequent guideline reversal supported broad deimplementation of prolonged high-dose MPSS and left NASCIS III as a canonical case study in the risks of subgroup-driven clinical inference.

Overall Summary

  • NASCIS III’s overall results did not show clear superiority of 48-hour MPSS over 24-hour MPSS, yet practice was heavily shaped by time-to-treatment subgroup findings.
  • Prolonged high-dose MPSS carried clinically important infectious signals (notably severe pneumonia), shifting the benefit–harm balance unfavourably.
  • Later critiques and evidence-based guidelines reframed MPSS away from “standard of care”, driving deimplementation and ongoing heterogeneity.

Bibliography