Methylprednisolone in acute traumatic spinal cord injury: case-matched outcomes from the NASCIS2 and Sygen historical spinal cord injury studies with contemporary statistical analysis

Fred H. GeislerDepartment of Medical Imaging, College of Medicine at the University of Saskatchewan, Saskatoon, Saskatchewan;

Search for other papers by Fred H. Geisler in
jns
Google Scholar
PubMed
Close
 MD, PhD
,
Ali MoghaddamjouDivision of Neurosurgery, Department of Surgery, University of Toronto and Spinal Program, Toronto Western Hospital, University Health Network, Toronto, Ontario, Canada; and

Search for other papers by Ali Moghaddamjou in
jns
Google Scholar
PubMed
Close
 MD
,
Jamie R. F. WilsonDepartment of Neurosurgery, College of Medicine, University of Nebraska Medical Center, Omaha, Nebraska

Search for other papers by Jamie R. F. Wilson in
jns
Google Scholar
PubMed
Close
 MD, MSc
, and
Michael G. FehlingsDivision of Neurosurgery, Department of Surgery, University of Toronto and Spinal Program, Toronto Western Hospital, University Health Network, Toronto, Ontario, Canada; and

Search for other papers by Michael G. Fehlings in
jns
Google Scholar
PubMed
Close
 MD, PhD
Open access

OBJECTIVE

Methylprednisolone (MP) to treat acute traumatic spinal cord injury (ATSCI) remains controversial since the release of the second National Acute Spinal Cord Injury Study (NASCIS2) in 1990. As two historical studies, NASCIS2 and Sygen in ATSCI, used identical MP dosages, it was possible to construct a new case-level pooled ATSCI data set satisfying contemporary criteria and able to clarify the effect of MP.

METHODS

The new pooled data set was first modernized by excluding patients with injury levels caudal to T10, lower-extremity American Spinal Injury Association (ASIA) motor scores (LEMSs) ≥ 46, Glasgow Coma Scale scores ≤ 11, and age < 15 or > 75 years, and then standardized to the ASIA grading and scoring format. A new updated NASCIS2 data set from this pooled data set contained 31.6% fewer patients than the 1990 NASCIS2 data set.

RESULTS

In the new pooled data set, recovery of LEMSs from baseline to 26 weeks, the primary outcome variable, was separated statistically into five different injury severity cohorts (p < 0.0001). The severity cohorts contained groups with severe floor (62.9%) and ceiling (10.7%) effects, which do not contribute to drug effects. The new NASCIS2 data set duplicated the p value for MP versus placebo in the sub-subgroup analysis of MP initiated ≤ 8 hours (the subgroup) and recovery of motor function on only the right side of the body (a further subgroup within the ≤ 8-hour subgroup), presented as the positive MP effect in the original NASCIS2 reporting. However, current statistical interpretation considers results seen only in post hoc sub-subgroups, without multi-test corrections, to be random effects without clinical significance. The combined case-level pooled data set from the NASCIS2 and Sygen studies increased the MP group from 106 to 431 patients, creating a new MP combined group. This new data set served as a surrogate for a contemporary MP study and found that administration of MP did not enhance ASIA motor score improvement in the lower extremities at 26 weeks. Secondary analysis of descending ASIA motor and sensory cervical neurological levels in cervical ATSCI patients at 26 weeks also found no MP drug effect.

CONCLUSIONS

Analysis of both the new updated NASCIS2 data set and the new case-matched pooled data set from two historical ATSCI studies revealed that administration of MP after spinal cord injury did not demonstrate any enhancement in neurological recovery at 26 weeks. The results of this analysis warrant review by clinical guideline groups.

ABBREVIATIONS

ASIA = American Spinal Injury Association; ATSCI = acute traumatic spinal cord injury; CMI = complete motor injury; LEMS = lower-extremity ASIA motor score; MP = methylprednisolone; MP-C = MP-N and MP-S combined; MP-N = MP patients from the NASCIS2 study; MP-S = identically dosed MP patients from the Sygen ATSCI study; NASCIS2 = second National Acute Spinal Cord Injury Study; NEJM = The New England Journal of Medicine ; NIH = National Institutes of Health.

OBJECTIVE

Methylprednisolone (MP) to treat acute traumatic spinal cord injury (ATSCI) remains controversial since the release of the second National Acute Spinal Cord Injury Study (NASCIS2) in 1990. As two historical studies, NASCIS2 and Sygen in ATSCI, used identical MP dosages, it was possible to construct a new case-level pooled ATSCI data set satisfying contemporary criteria and able to clarify the effect of MP.

METHODS

The new pooled data set was first modernized by excluding patients with injury levels caudal to T10, lower-extremity American Spinal Injury Association (ASIA) motor scores (LEMSs) ≥ 46, Glasgow Coma Scale scores ≤ 11, and age < 15 or > 75 years, and then standardized to the ASIA grading and scoring format. A new updated NASCIS2 data set from this pooled data set contained 31.6% fewer patients than the 1990 NASCIS2 data set.

RESULTS

In the new pooled data set, recovery of LEMSs from baseline to 26 weeks, the primary outcome variable, was separated statistically into five different injury severity cohorts (p < 0.0001). The severity cohorts contained groups with severe floor (62.9%) and ceiling (10.7%) effects, which do not contribute to drug effects. The new NASCIS2 data set duplicated the p value for MP versus placebo in the sub-subgroup analysis of MP initiated ≤ 8 hours (the subgroup) and recovery of motor function on only the right side of the body (a further subgroup within the ≤ 8-hour subgroup), presented as the positive MP effect in the original NASCIS2 reporting. However, current statistical interpretation considers results seen only in post hoc sub-subgroups, without multi-test corrections, to be random effects without clinical significance. The combined case-level pooled data set from the NASCIS2 and Sygen studies increased the MP group from 106 to 431 patients, creating a new MP combined group. This new data set served as a surrogate for a contemporary MP study and found that administration of MP did not enhance ASIA motor score improvement in the lower extremities at 26 weeks. Secondary analysis of descending ASIA motor and sensory cervical neurological levels in cervical ATSCI patients at 26 weeks also found no MP drug effect.

CONCLUSIONS

Analysis of both the new updated NASCIS2 data set and the new case-matched pooled data set from two historical ATSCI studies revealed that administration of MP after spinal cord injury did not demonstrate any enhancement in neurological recovery at 26 weeks. The results of this analysis warrant review by clinical guideline groups.

In Brief

Methylprednisolone (MP) in acute traumatic spinal cord injury (ATSCI) remains controversial. The second National Acute Spinal Cord Injury Study (NASCIS2) and Sygen studies used identical MP dosages, which allowed for construction of a case-level pooled data set. The original 1990 NASCIS2 study had a large percentage of patients without an ATSCI, and the positive results reported were only in a sub-subgroup. No MP drug effect was noted in the combined data set, removing the rationale for the use of MP in ATSCI.

Various opinions, questions, and controversies regarding methylprednisolone (MP) as a treatment for acute traumatic spinal cord injury (ATSCI) have persisted for 30 years. The second National Acute Spinal Cord Injury Study (NASCIS2) initially released its results at a news conference and was reported on the front page of the New York Times13 in 1990. Two weeks later, NASCIS2 sent a National Institutes of Health (NIH) “Clinical Alert”4 directly to 19,000 doctors and hospitals with dosing information on MP for ATSCI, and 4 weeks after that, a report was published in The New England Journal of Medicine (NEJM).5 Despite the NEJM reporting positive results only in a sub-subgroup analysis of MP initiated ≤ 8 hours from injury (the subgroup) and recovery of motor function on only the right side of the body (a further subgroup within the ≤ 8-hour subgroup of patients), the NASCIS2 MP protocol quickly became a standard of care for all ATSCIs. The NASCIS2 MP dosing was included in the 7th edition of the ATLS Advanced Trauma Life Support Program for Doctors protocol in 20046 before being removed from the 8th edition in 2008 over safety concerns.79 The decades since then have seen major international institutions such as the Congress of Neurological Surgeons, the American Association of Neurological Surgeons, AO Spine, UpToDate, and the Cochrane Library produce conflicting guidelines and recommendations regarding the use of MP for ATSCI.1014 An application for US FDA approval for MP in ATSCI has never been made, and its use remains off-label in the US. Several meta-analyses15,16 and systematic reviews14,15 have reported no significant benefit from MP administration, leading some to discontinue MP use for ATSCI. The non–science-based use of MP in ATSCI has been documented internationally,1719 and some have considered its efficacy to be a philosophical question.20

Because MP therapy after ATSCI has known risks,21 no prospective, adequately powered, randomized study of MP versus placebo has occurred in the past 30 years, and none is anticipated in the foreseeable future. Retrospective reexamination of historical data sets appears to be the best opportunity to investigate the efficacy of MP versus placebo in a larger ATSCI patient cohort. This paper presents an expanded analysis of the original NASCIS2 study data set,5 created by combining the randomized MP patients from the NASCIS2 study (MP-N) with the identically dosed MP patients from the Sygen ATSCI study (MP-S).2224 The data pooling was considered appropriate because patients in both studies received identical dosing of MP and were provided similar medical and surgical care for ATSCI. Additionally, it was anticipated that questions raised by others about the completeness of NASCIS2 reporting could be addressed.10,2530 The new pooled ATSCI database has a greater number of MP patients than does the NASCIS2 study and can serve as a surrogate for a contemporary randomized MP study of ATSCI once modern ATSCI definitions and exclusion criteria are applied.

Methods

The NASCIS2 and Sygen ATSCI studies were pooled using the new case-matched technique, championed by the NEJM and others as superior to meta-analyses in revealing new insights from decades-old databases that are combined.31,32 Initially, a new data set of 1247 subjects was prepared by case matching at the neurological examination level the data sets from two historical ATSCI studies (487 subjects from NASCIS25 and 760 from the Sygen ATSCIS2224). Both studies used a randomized, prospective, blinded protocol with FDA (Sygen) or NIH (NASCIS2) oversight of data quality. Both studies were conducted in multiple major North American ATSCI treatment facilities.5,24 All patients were enrolled from 1985 to 1997, with the NASCIS2 study entering patients a few years earlier than the Sygen study.

The Sygen study was historically block randomized to minimize imbalances between treatment groups, anatomical regions, and entrance American Spinal Injury Association (ASIA) grades. The NASCIS2 study did not use block randomization. MP was used in the MP-N group of the NASCIS2 study and all patients in the Sygen study at identical dosing of a 30-mg/kg bolus followed by 5.4 mg/kg/hr for 23 hours. The time from injury to initiation of MP was notably less in the Sygen study than in the NASCIS2 study (median times of 2.267 and 7.875 hours, respectively). Both studies presented uniformly low complication rates in the original publications.5,2224

The modernization of the pooled neurological examination source data of the NASCIS2 and Sygen data sets required five additional steps for matching and updating. 1) The motor scores were updated to match current ASIA standards33,34 by applying motor group conversions to the NASCIS2 data set (see Table 1). 2) To match baseline examination times, a new baseline study examination variable for the NASCIS2 data set was constructed using the 48-hour examination (n = 324) or the 72-hour (n = 3), 24-hour (n = 2), and then initial emergency department (n = 4) examination in that order of priority to approximate the Sygen mean time of 55 hours. 3) Bony levels were determined for the 29 NASCIS2 patients lacking them after neurological scores were added to the ASIA clinical worksheet.35 4) Missing motor score values were assigned a value of zero. 5) Contemporary exclusion criteria were applied.

TABLE 1.

Myotome conversions applied to the NASCIS2 data set

ASIA-ISNCSCI Key Muscles on WorksheetMotor Muscle Used in the NASCIS2 Data Set for Each ASIA-ISNCSCI Key Muscle
Elbow flexors, C5Biceps (C5, C6)*

  If missing, use deltoid (C5)*
Wrist extensors, C6Extensor carpi radialis longus (C6, C7)*

  If missing, use biceps (C5, C6)*
Elbow extensors, C7Triceps (C7)*

  If missing, use extensor carpi radialis longus (C6, C7)*

   If also missing, use extensor digitorum (C7, C8)*
Finger flexors, C81st dorsal interosseus (C8, T1)*

  If missing, use opponens pollicis (C8, T1)*

   If also missing, use extensor digitorum (C7, C8)*
Finger abductors (little finger), T1Opponens pollicis (C8, T1)*

  If missing, use 1st dorsal interosseus (C8, T1)*
Hip flexors, L2Iliopsoas (L1, L2, L3)*

  If missing, use quadriceps femoris (L2, L3, L4)*
Knee extensors, L3Quadriceps femoris (L2, L3, L4)*

  If missing, use iliopsoas (L1, L2, L3)*
Ankle dorsiflexors, L4Tibialis anterior (L4, L5)*

  If missing, use quadriceps femoris (L2, L3, L4)*
Long toe extensors, L5Extensor hallucis longus (L5, S1)*

  If missing, use tibialis anterior (L4, L5)*

   If also missing, use peroneus longus & brevis (L5, S1)*
Ankle plantar flexors, S1Gastrocnemius (S1, S2)*

  If missing, use hamstring muscles (L4, L5, S1, S2)*

   If also missing, use peroneus longus & brevis (L5, S1)*

ISNCSCI = International Standards for Neurological Classification of Spinal Cord Injury. The assignment of the motor function measured in the NASCIS2 data set to a similar myotome muscle listed as a key muscle on the ASIA-ISNCSCI motor scoring worksheet.35 The primary muscle measured by NASCIS2 is listed first. In the few instances in which the first choice for motor group was missing from the original NASCIS2 data set, a second choice and, in some cases, a third choice, was used.

The myotome(s) of each specific motor group tested in the original NASCIS2 data set.

Contemporary exclusion criteria were developed from ATSCI studies admitting patients within the last decade, as listed on the ClinicalTrials.gov registry (https://clinicaltrials.gov) and reported by Fawcett et al.,36 to make this new case-matched pooled data set match modern study designs. Exclusion criteria and the number of patients excluded from each study were 1) missing neurological examination for the baseline lower-extremity ASIA motor score (LEMS; n = 2 NASCIS2 and n = 8 Sygen) or baseline LEMS showing no motor deficit (n = 7 NASCIS2 and n = 0 Sygen); 2) head injury patients with a Glasgow Coma Scale score ≤ 11 (n = 17 NASCIS2 and n = 0 Sygen); 3) age < 15 or > 75 years (n = 14 NASCIS2 and n = 9 Sygen); 4) bony injuries caudal to T10 (n = 77 NASCIS2 and n = 2 Sygen); and 5) baseline LEMS of 46–50 out of 50 (n = 37 NASCIS2 and n = 0 Sygen).

The five patient exclusion criteria above resulted in the removal of 173 patients (n = 154, 31.6% of NASCIS2 patients and n = 19, 2.5% of Sygen patients). The bony injuries caudal to T10, the cauda equina injuries, will be presented in a subsequent article. The bony level distribution separated by study and drug treatment group for the MP-N and MP-S combined (MP-C) and placebo patients in the pooled data set is shown in Fig. 1.

FIG. 1.
FIG. 1.

Graph showing the neurological injury level for the placebo (n = 113) and MP-C (n = 431) treatment groups in the new pooled data set. The MP-C patients are displayed both in total and stratified by the ATSCI study (ATSCIS) in which they were included. The incidence of ATSCI was greatest at the C4, C5, and C6 cervical levels, with a smaller increase in the midthoracic area.

The present study used a 26-week follow-up time as the primary outcome time point because it was the primary outcome time of the Sygen study2224 and the follow-up time point in the NEJM publication of NASCIS25 results. The primary outcome variable was the total LEMS improvement from baseline to the 26-week follow-up. The fractional improvement in LEMS at 26 weeks is an alternate presentation of the primary outcome variable,37 useful for separating the injury severity cohorts, and is defined as follows: 26-week fractional improvement in LEMS = ([26-week LEMS] − [baseline LEMS])/(50 − [baseline LEMS]).

The patient distribution in the study, study groups, timing of study drug initiation, and injury severity cohorts from the case-matched complete data set are shown in Fig. 2. The new NASCIS2 data set analysis uses the MP-N, naloxone, and placebo drug groups from only original NASCIS2 patients. The new pooled MP-C versus placebo data set analysis uses the MP-N and MP-S combined (MP-C) and placebo groups of the pooled case-matched new data set.

FIG. 2.
FIG. 2.

Chart showing the number of placebo and the MP-N and MP-S patients at 26 weeks by injury severity classification group, divided into three different times from spinal cord injury (SCI) to initiation of the study drug to highlight the earlier dosages given to the MP-S group. Each injury severity cohort is classified as having a severe floor effect, measurement of drug effect improvement possible, or a severe ceiling effect. All injury severity groups except CMI included only a small and imbalanced number of patients from the NASCIS2 study. Note that the addition of the MP-S patients into the pooled data set to create the MP-C group significantly increased the number of MP patients available for analysis. *If 26-week follow-up data were missing, the 52-week follow-up examination was used. **The unknown time from injury to drug bolus was due to unknown injury time. ***MP dosing of a 30-mg/kg bolus followed by 5.4 mg/kg/hr for 23 hours.

The analysis of the new NASCIS2 database also included the total motor improvement subgroups (by time) and sub-subgroups (motor recovery of the right and left sides of the body) at 26 weeks, calculated to match the analysis presented originally by NASCIS2.

The complete motor injury (CMI) group remained with a score of 0 from baseline to 26 weeks and represented 62.9% of all patients in the new pooled data set. To enable inclusion of most of the CMI patients in the drug-effect analysis of the new pooled data set, secondary outcome variables of improvement in motor, pinprick, and light touch cervical function were analyzed for those with C4–7 bony injury levels (n = 451).

It is notable that this new combined case-matched ATSCI database is the largest prospectively collected or constructed database for the analysis of MP treatment in ATSCI (113 placebo and 431 MP-C patients). Statistical analysis was performed using SAS JMP version 16.2.0 (SAS Institute) and Stata version 17 (StataCorp LLC).

The new pooled data set of 1074 patients (113 placebo, 114 naloxone, 431 MP-C, and 416 MP+Sygen) in Fig. 2 included cohorts of ATSCI injury severity with highly different recovery potentials. The five injury severity grades used in the paper are expanded from the ASIA injury severity classifications, as follows:33,34

  1. 1)CMI (baseline ASIA grade A or B with no improvement at either 26 or 52 weeks)
  2. 2)A+ (baseline ASIA grade A with a baseline LEMS of 0 and improvement at 26 or 52 weeks)
  3. 3)B+ (baseline ASIA grade B with improvement, and baseline ASIA grade C with a baseline LEMS of 1–3)
  4. 4)C4–15 (baseline ASIA grade C with a baseline LEMS of 4–15)
  5. 5)
    1. a)C16–25 (baseline ASIA grade C with a baseline LEMS of 16–25)
    2. b)C26–45 (baseline ASIA grade C with a baseline LEMS of 26–45)

The fractional improvement in the 26-week LEMS for the injury severity grades of all five drug groups from both pooled studies is plotted in Fig. 3 with the individual points scattered. Each of the five injury severity grades has highly statistically significant different recovery patterns from the others (all at p < 0.0001) and hence requires stratification.

FIG. 3.
FIG. 3.

Using the complete pooled database of n = 1074, a one-way analysis of improvement (Imp) in fractional LEMS from baseline to 26 weeks by injury severity grade with dithered points, mean diamonds, mean error bars, and comparison circles is displayed. The mean points of the injury severity improvements are connected as a visual aid to highlight the monotonic increasing difference between the C16–25 and C26–45 groups, which have the same ceiling effect. The comparison circles are a visual representation of the boundaries of the p = 0.05 paired Student t-test; when there is no overlapping, a statistical difference exists. The CMI, A+, B+, C04–15, and C16–25 baseline injury severity cohort groups had highly statistically significant different recovery patterns from one another (p < 0.0001), while the concentric nature of the two injury severity cohorts corresponding to C16–25 and C26–45 for baseline LEMS indicates no difference in their recovery potentials. The CMI group has a severe floor effect and the C16–25 and C26–45 groups a severe ceiling effect, which prevent their use for measuring drug effect differences.

Three distinct regions requiring separate statistical analysis became apparent: 1) floor effect, 2) measurable recovery, and 3) ceiling effect. The C16–25 and C26–45 injury severity groups (5a and 5b as above) had similar high mean outcomes and are presented separately to show the large baseline range of ceiling effect for injuries with baseline LEMS scores of C16–45. Creating the groups of CMI and severity grades A+ and B+ can only be done after the follow-up is known at 26 and/or 52 weeks. The severe floor effect, the CMI cohort, had a baseline LEMS of 0, made no recovery, and the score remained 0 at the 26-week follow-up, and hence does not contribute to the recovery in the fractional LEMS at 26-weeks. The ceiling effect patients also do not contribute to the recovery of the fractional LEMS at 26 weeks, but for a different reason: they have full or nearly full recovery at 6 months. As the 26-week LEMS is at the ceiling, only the baseline randomness determines the recovery in fractional LEMS at 26 weeks.

The severe floor and ceiling effect injury grades cannot contribute to measurement of drug effect. This is a fundamental limitation of all ATSCI studies. The patients contributing to recovery differences are the A+, B+, and C4–15 injury severity groups (26.4% of all patients in these two pooled studies, see Fig. 2), a factor useful for sample size calculations in future ATSCI studies. For the pooled data set of the two studies, the statistical analysis is separately performed on each of the five injury severity cohorts so that patients with highly statistically significant different motor recoveries are compared with patients in the same injury severity cohort. This removes the significant confounding effect of an imbalance in sample allocation between the treatment groups.

Results

Figure 2 lists the ATSCI patient demographics for the three drug treatment groups in the new NASCIS2 data set and the two drug treatment groups from the Sygen data set. The figure is further divided into injury severity cohorts grouped into three regions: floor effect, injury severity determining recovery, and ceiling effect, along with the number of patients in the MP-N, MP-S, and placebo groups in each injury severity group. The demographics for both the new NASCIS2 data set and the new pooled database are presented. Inspection of the three drug groups of the new NASCIS2 database—placebo, naloxone, and MP-N—reveals that each injury severity group, except for CMI, only included a small number of unbalanced patients in the drug treatment groups. After the floor and ceiling effect patients are removed, the sample size that determines drug comparison in lower-extremity motor function recovery is disappointingly small. For instance, the sample sizes in the new NASCIS2 subgroups of ≤ 8-hour injury (0- to 3-hour and 3- to 8-hour groups combined) to the start of drug initiation, which was reported to have a positive drug effect on the right side of the body, were 9 for the placebo group and 9 for the MP-N group (the sum of the A+, B+, and C4–15 groups).

Figure 4 is a complete compilation of subgroup and sub-subgroup analyses of the original 1990 NASCIS2 published results and is a natural consequence of using the post hoc sub-subgroup of the right side of the body only and ≤ 8-hour injury to the start of drug initiation. This figure displays the new NASCIS2 data set results in the following three groupings: 1) primary analyses: all MP-N patients versus all placebo patients, and all naloxone patients versus all placebo patients (2 main tests); 2) subgroup by time (0 to ≤ 8 and > 8 to 24 hours from ATSCI to initiation of the study drug) for the two main groups (4 additional tests); and 3) sub-subgroups by 8 potential anatomical regions for each main group and subgroup (32 additional tests). Thus, there are 38 tests when the main tests, subgroup, and sub-subgroup tests are summed. With this many multiple tests, the threshold for statistical significance of any single test if the p value of 0.05 is equally divided is 0.05/38 = 0.00132.

FIG. 4.
FIG. 4.

Independently calculated p values for the main tests, subgroups, and sub-subgroups of the new NASCIS2 data set. Estimates for the p value correction for multiple tests are listed to the right of the p values but are not used as correction for the listed p values. The two main tests were statistically negative for both MP-N and naloxone compared with the placebo group. Even the time subgroups were negative, with one interesting p value (p = 0.0511) for MP-N having worse recovery than placebo in the 8- to 24-hour group. The sub-subgroups are separated by upper extremity (UE) versus lower extremity (LE), right versus left side of the body, and ≤ 8 hours versus > 8 hours from injury to initiation of the study drug. Of note is that the sub-subgroups of ≤ 8 hours for right upper extremity and right body side had calculated p values < 0.05. After applying a correction for multiple tests, no test approached statistical significance. *No correction for multiple tests was included in the listed p value. **Multiple test correction was calculated for equal p value distribution.

Figure 5A is an analysis of MP-N or naloxone drug effect in the subgroups of ≤ 8 hours and for the sub-subgroups of right and left sides of the body. There are another 6 drug treatment groups for 8–24 hours that are not graphed. It is notable that neither the main test nor any subgroup by time test was significant individually except for the MP-N versus placebo in the 8- to 24-hour time subgroup favoring placebo (i.e., a harmful effect on neurological recovery for MP-N) (Fig. 4).

FIG. 5.
FIG. 5.

One-way ANOVA of motor recovery from baseline to 6 months for sub-subgroup by right and left sides of the body, based on the new NASCIS2 data set. A: Patient injury severities. B: Severities after removal of floor and ceiling effect patients. The one-way ANOVA of improvement in total LEMS at 26 weeks in the subgroup of patients with drug treatment initiated within ≤ 8 hours of injury and in the sub-subgroups separated by body side tests are presented two ways. This is a partial analysis presentation of the large list of sub-subgroups in Fig. 4. Note the continued decrease in the placebo mean recovery of the right side of the body in the ≤ 8-hour time sub-subgroup, which leads to a calculated right side of the body placebo versus MP-N result that meets the threshold of p < 0.05, uncorrected for multiple tests. BL = baseline; Std Err Diff = standard error difference.

When MP-N is compared with placebo for only the right side of the body, a p value of 0.0364 is calculated with no multiple test corrections; this p value is like the same sub-subgroup comparison in the 1990 reported study results. Visual inspection of Fig. 5 shows that the MP-N average recovery value is only mildly elevated relative to the average of the other treatment groups, while the placebo value for the right side of the body is decreased by 60%. Thus, the calculated drug effect, the difference between the MP-N group and the placebo group, arose mostly from the low recovery in the placebo group. Furthermore, if the highest patient in the MP-N group or the lowest patient in the placebo group is removed from the analysis, even the uncorrected p value threshold of 0.05 is not obtained. Thus, the fragility analysis38 for the new NASCIS2 data set is the removal of one patient from either MP-N or placebo. It is notable that the same analysis on the left side of the body did not meet the p < 0.05 uncorrected threshold.

Figure 5B is the same statistical analysis with the floor and ceiling patients excluded. For these patients, the ≤ 8-hour sub-subgrouping produces different means but similar statistical analyses as when the floor and ceiling effect patients are included. The lack of symmetry between the right and left sides of the body on both graphs in Fig. 5 is evidence for a random statistical effect causing less recovery in the right body side placebo group.

The primary outcome variable for the case-matched pooled data set was the improvement in the total LEMS, which is the sum of the left and right body measurements. Combining both sides to assess neurological improvement is the norm for all contemporary studies, and testing for a one-sided body drug effect is unique to the NASCIS2 studies.

The outcome at 26 weeks (improvement in LEMS) presented for each analysis classification group and the subset of placebo and MP-C patients with ≤ 8 hours from injury to initiation of study drug is shown in Fig. 6. The CMI group is not plotted on this graph, as all patients in this group have baseline and 26-week LEMSs of 0. Groups with fewer than 5 patients were considered too small to present statistical results and were not plotted.

FIG. 6.
FIG. 6.

Bar graph showing improvement in LEMS at 26 weeks by injury severity group for 1) all patients in the new data set; 2) placebo and MP-C groups; and 3) the subsets of placebo and MP-C patients ≤ 8 hours from injury to initiation of the study drug. The all-patients bar for each analysis group is the same as in Fig. 3; the bars are all p < 0.0001, statistically different from one another, and serve as reference values for the separate drug treatment groups. The only placebo group in the ≤ 8-hour placebo versus MP-C comparison is the A+ group, as it was the only one passing the threshold for analysis of more than 5 patients. Error bars are 95% CI.

In Fig. 6, the bars corresponding to all patients in each injury severity group are all statistically different from one another at p < 0.0001 and serve as reference values for the separate drug treatment groups. The placebo and MP-C bars compare the two drug treatment groups, and no statistically significant difference or trends were noted. Additionally, it was visually apparent that the C16–45 group numerically favored placebo, while the B+ and C4–15 groups numerically favored MP-C. For the ≤ 8-hour placebo versus ≤ 8-hour MP-C comparison, the A+ group taking placebo was the only injury severity drug treatment group that included more than 5 patients, and the A+ placebo group had numerically greater recovery than the group taking MP. For the other injury severity groups without placebo comparison, an inspection notes the overlapping of the 95% mean error bars of the all-patient, placebo, and MP-C groups with the ≤ 8-hour MP-C groups. The fraction of recoveries in the baseline LEMS = 0 group (number of patients with ASIA grade A+/number of patients with ASIA grade A in the CMI group) did not demonstrate an MP drug effect (p = 0.331, Fisher’s exact test, two-tailed). Refiguring all the above calculations using the 52-week outcomes confirmed the lack of significant MP effects. Note that although each all-patient grouping in each injury severity classification group was highly statistically different from every other (p < 0.0001), none of the MP-C treatment groups had a statistically significant difference from its corresponding placebo in the same injury severity group.

Figure 7 is the secondary analysis on the C4–7 bony injury CMI patients for the improvement in neurological function of ASIA score from baseline to 26 weeks for upper-extremity motor, pinprick, and light touch. The bars for the all-patient recoveries for each neurological function are highly statistically significantly different from one another (p < 0.0001), with light touch recovering more numerically than pinprick score, and pinprick recovering more numerically than the upper-extremity motor score. There were no significant differences in the three neurological measurements between MP-C and placebo. The same tests on the ≤ 8-hour subgroups also showed no significant MP-C versus placebo drug effect.

FIG. 7.
FIG. 7.

Bar graph showing the C4–7 bony injury in CMI patients, assessing neurological improvement by evaluating descending levels from baseline to 26 weeks for motor, pinprick, and light touch in the cervical levels. The recovery amounts of these neurological functions for all patients are all statistically significantly different from one another (p < 0.0001), but MP provided no significant difference in recovery in any placebo versus MP-C analysis of any of the three neurological functions. Error bars are 95% CI.

Discussion

The usefulness and correct application of subgroup analysis in randomized clinical studies has been debated and characterized, and consensus on the major aspects has been achieved. Yusuf et al.39 and Alosh et al.40 noted that the main statistical test is a better test for true effect than are subgroup tests, and, furthermore, the investigation of subgroup tests requires the main test to be positive and is used only to clarify which demographic groups have differences. The necessity to prospectively specify subgroup analysis in a randomized clinical trial was emphasized by Assmann et al.41 in 2000. Subgroup analysis in clinical trials was surveyed in several major medical journals by Moreira et al. in 2001 and found that published subgroup analyses were often harmful to treatment recommendations.46 Lagakos42 calculated the probability of having one false positive when 30 subgroups were surveyed to be > 80% without correction for multiple tests (alpha sharing not considered).

The original NASCIS2 data set and analysis have three issues that do not conform with modern thought. First, the NASCIS2 data set contains a large percentage of patients (31.6%) who do not meet the modern ASIA definition of spinal cord injury. Second, in the original NASCIS2 analysis, ATSCI patients with different initial injury severities and outcomes were grouped together without either a stratification or case-matching step, and the large fraction of patients with severe floor or ceiling effects was not considered in the analysis presentation. The third issue was the use of subgroup and sub-subgroup analyses, when the main and subgroup tests were negative, coupled with no corrections for multiple tests. The NASCIS2 report, by substituting sub-subgroups for the main groups, converted a prospective randomized study into a post hoc analysis. The sub-subgroup reported was 1) neurological function only on the right side of the body and 2) ≤ 8 hours from injury to drug initiation and was used as a substitute for the main test of all patients and all entry times. Guidelines for subgroup analysis are available in the NEJM43 and on the US FDA website.44,45 The reported results of the NASCIS2 study in the New York Times,13 via an NIH “Clinical Alert,”4 and then in the NEJM5 in 1990 did not follow contemporary guidelines or the philosophy of subgroup analysis references. The NASCIS2 results of 1990 are thus currently considered an obsolete reporting and statistical analysis.

In general, removing patients who do not actually have the disease being studied from a statistical analysis should favor any positive statistical result for patients with the disease of interest. Also, increasing the total number of patients with the disease of interest, such as by adding the MP-S group to the MP-N group in the case-matched pooled group, increases the statistical power. The MP initiation time was also significantly earlier in the MP-S group, which should have been more effective if the MP drug effect were initiation time dependent. Thus, there are three factors built into this pooled case-matched database analysis that favor finding a statistically significant effect for MP in ATSCI. Although these three biases make any statistically positive results subject to skepticism and criticism, finding no statistical significance, as in the present study, is less controversial.

Use of the case-matched approach, injury severity stratification, and patient exclusions in this paper is not possible in a meta-analysis, which only considers results and groupings already created in the original presentations. Separating the new data set into the five injury severity cohorts was considered essential before investigating for drug effects and was only possible with both the new NASCIS2 database and the case-matched merger of the two historical studies.

The analysis of the new NASCIS2 data set did verify the main statistical calculation presented by NASCIS2 in 1990—a statistical effect at the p < 0.05 threshold, presented without a multiple test correction, in a post hoc analysis of the sub-subgroup 1) ≤ 8 hours from injury to study drug initiation and 2) right side of the body only. Meanwhile, our investigation on both new data sets found no MP drug effect in either the main or subgroup analysis. We now know that significance was only found on the right side of the body and not the left, making the clinical interpretation of a drug effect clinically ambiguous at best, and in fact highly suggestive of a random event, as it contradicts known clinical medicine. As the main test was not statistically significant, current analysis techniques would not allow statistical presentation of subgroup or sub-subgroup tests as a substitute for the main test. The current statistical interpretation for a possible MP drug effect in the new NASCIS2 data set is exactly opposite of that presented in the 1990 press conference,1 the NIH “Clinical Alert,”4 and the NEJM article.7

Neither the analysis of the new NASCIS2 data set nor that of the new case-matched pooled study database are reanalyses of the original NASCIS2 data set.5 Rather, both are new analyses on new databases mimicking contemporary ATSCI research protocols and present the answers clinicians need today to manage patients with ATSCI.

Long nerve fiber tracts were assessed by analyzing the changes in lower-extremity function (Fig. 6), intrinsically weighted to recovery in the long nerve fiber tracts. The recovery of gray matter neuron pools was evaluated by examining the cervical CMI patients, who can only recover neurological function by changes at and directly caudal to the cervical injury zone. Their recovery is measured by an increase in the number of useful motor levels and by descending ASIA levels of pinprick and light touch (Fig. 7).

Although significant changes have occurred in emergency medical service protocols as well as medical, surgical, and ICU care in the 30 years since the NASCIS2 and Sygen studies enrolled patients, this historical spinal cord injury sample, in which 212 (23.9%) of 888 patients recovered lower-extremity motor function from baseline to follow-up, provides a large group to analyze for differences in recovery.

The new case level combined data set failed to demonstrate statistical significance of drug effect in the primary or time subgroup analysis for recovery of motor function in the lower extremities (Fig. 6) or in the secondary analysis of improvement in cervical motor, pinprick, or light touch in cervical CMI (Fig. 7). This new case-matched database of two historical studies, NASCIS2 and Sygen, serves as a surrogate for a contemporary clinical ATSCI study comparing MP against placebo and thus provides guidance for physicians today.

Conclusions

It is not surprising that, in the more than 30 years since publication of the NASCIS2 and Sygen studies, the definition of ATSCI, statistical analysis techniques and interpretations, and publishing standards have changed. A new, case-matched data set pooled from the NASCIS2 and Sygen databases was used to investigate the effect of MP on ATSCI with contemporary clinical scales and exclusion criteria as well as a greatly increased number of patients in the MP-C treatment group.

This analysis provides evidence that the positive post hoc sub-subgroup results initially reported in 1990 by NASCIS2 resulted from randomization imbalance in the injury severity groups and inclusion of patients not currently considered as having ATSCI. Furthermore, current statistical guidelines and statistical interpretation do not permit the substitution of a positive sub-subgroup investigation for negative main and subgroup tests. Modern statistical analysis and interpretation of the new data sets, both with all the MP and placebo patients in the two studies and also with only the NASCIS2 patients, found no evidence for MP drug effect in ATSCI. The results of this analysis warrant review by clinical guideline groups.

Acknowledgments

We acknowledge Lois A. Polatnick, MD, for editorial assistance with the manuscript and William P. Coleman, PhD, for statistical advice.

Disclosures

The authors report no conflict of interest concerning the materials or methods used in this study or the findings specified in this paper.

Author Contributions

Conception and design: Geisler, Moghaddamjou. Acquisition of data: Geisler. Analysis and interpretation of data: all authors. Drafting the article: Geisler, Wilson. Critically revising the article: Geisler, Wilson, Fehlings. Reviewed submitted version of manuscript: Geisler, Wilson, Fehlings. Approved the final version of the manuscript on behalf of all authors: Geisler. Statistical analysis: Geisler, Moghaddamjou. Administrative/technical/material support: Geisler, Wilson, Fehlings. Study supervision: Geisler.

Supplemental Information

Online-Only Content

Supplemental material is available with the online version of the article.

References

  • 1

    Leary WE. Treatment is said to reduce disability from spinal injury. New York Times. March 31, 1990:1. https://timesmachine.nytimes.com/timesmachine/1990/03/31/078590.html?pageNumber=1

    • Search Google Scholar
    • Export Citation
  • 2

    Leary WE. Delay seen in publicizing spinal drug. New York Times. April 10, 1990:5.https://timesmachine.nytimes.com/timesmachine/1990/04/10/651990.html?pageNumber=45

    • Search Google Scholar
    • Export Citation
  • 3

    Associated Press. Spine treatment to be publicized. New York Times. April 12, 1990:21.https://timesmachine.nytimes.com/timesmachine/1990/04/12/161790.html?pageNumber=21

    • Search Google Scholar
    • Export Citation
  • 4

    Clinical Alert—New Treatment for Acute Spinal Cord Injury. NIH, National Institute of Neurological Disorders and Stroke; 1990.

  • 5

    Bracken MB, Shepard MJ, Collins WF, et al. A randomized, controlled trial of methylprednisolone or naloxone in the treatment of acute spinal-cord injury. Results of the Second National Acute Spinal Cord Injury Study. N Engl J Med. 1990;322(20):14051411.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 6

    ATLS Advanced Trauma Life Support Program for Doctors. 7th ed. American College of Surgeons; 2004.

  • 7

    Ito Y, Sugimoto Y, Tomioka M, Kai N, Tanaka M. Does high dose methylprednisolone sodium succinate really improve neurological status in patient with acute cervical cord injury? A prospective study about neurological recovery and early complications. Spine (Phila Pa 1976). 2009;34(20):21212124.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8

    Matsumoto T, Tamaki T, Kawakami M, Yoshida M, Ando M, Yamada H. Early complications of high-dose methylprednisolone sodium succinate treatment in the follow-up of acute cervical spinal cord injury. Spine (Phila Pa 1976). 2001; 26(4):426430.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 9

    Kortbeek JB, Al Turki SA, Ali J, et al. Advanced trauma life support,. 8th edition, the evidence for change. J Trauma. 2008;64(6):16381650.

  • 10

    Hadley MN, Walters BC, Grabb PA, et al. Pharmacological therapy after acute cervical spinal cord injury. Neurosurgery. 2002;50(3 suppl):S63S72.

  • 11

    Hurlbert RJ, Hadley MN, Walters BC, et al. Pharmacological therapy for acute spinal cord injury. Neurosurgery. 2013;72(suppl 2):93105.

  • 12

    Bracken MB. Steroids for acute spinal cord injury. Cochrane Database Syst Rev. 2012;1(1):CD001046.

  • 13

    Hansebout RR, Kachur E. Acute traumatic spinal cord injury. UpToDate. Accessed December 19, 2022. https://www.uptodate.com/contents/acute-traumatic-spinal-cord-injury

    • Search Google Scholar
    • Export Citation
  • 14

    Fehlings MG, Wilson JR, Harrop JS, et al. Efficacy and safety of methylprednisolone sodium succinate in acute spinal cord injury: a systematic review. Global Spine J. 2017;7(3 suppl):116S137S.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 15

    Evaniew N, Belley-Côté EP, Fallah N, Noonan VK, Rivers CS, Dvorak MF. Methylprednisolone for the treatment of patients with acute spinal cord injuries: a systematic review and meta-analysis. J Neurotrauma. 2016;33(5):468481.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 16

    Liu Z, Yang Y, He L, et al. High-dose methylprednisolone for acute traumatic spinal cord injury: a meta-analysis. Neurology. 2019;93(9):e841e850.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 17

    Falavigna A, Quadros FW, Teles AR, et al. Worldwide steroid prescription for acute spinal cord injury. Global Spine J. 2018;8(3):303310.

  • 18

    Teles AR, Cabrera J, Riew KD, Falavigna A. Steroid use for acute spinal cord injury in Latin America: a potentially dangerous practice guided by fear of lawsuit. World Neurosurg. 2016;88:342349.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 19

    Adamczak SE, Hoh DJ. Steroids and spinal cord injury—a global dilemma. World Neurosurg. 2016;90:641643.

  • 20

    Bowers CA, Kundu B, Hawryluk GW. Methylprednisolone for acute spinal cord injury: an increasingly philosophical debate. Neural Regen Res. 2016;11(6):882885.

  • 21

    Lenzer J, Brownlee S. Methyprednisolone and acute spinal cord injury: redux. Rapid response. BMJ. 2008;336:532.

  • 22

    Geisler FH, Coleman WP, Grieco G, Poonian D. The Sygen multicenter acute spinal cord injury study. Spine (Phila Pa 1976). 2001;26(24 suppl):S87S98.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 23

    Geisler FH, Coleman WP, Grieco G, Poonian D. Measurements and recovery patterns in a multicenter study of acute spinal cord injury. Spine (Phila Pa 1976). 2001;26(24 suppl):S68S86.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 24

    Geisler FH, Coleman WP, Grieco G, Poonian D. Recruitment and early treatment in a multicenter study of acute spinal cord injury. Spine (Phila Pa 1976). 2001;26(24 suppl):S58S67.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 25

    Hanigan WC, Anderson RJ. Commentary on NASCIS-2. J Spinal Disord. 1992;5(1):125133.

  • 26

    Nesathurai S. Steroids and spinal cord injury: revisiting the NASCIS 2 and NASCIS 3 trials. J Trauma. 1998;45(6):10881093.

  • 27

    Coleman WP, Benzel D, Cahill DW, et al. A critical appraisal of the reporting of the National Acute Spinal Cord Injury Studies (II and III) of methylprednisolone in acute spinal cord injury. J Spinal Disord. 2000;13(3):185199.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 28

    Hurlbert RJ. Methylprednisolone for acute spinal cord injury: an inappropriate standard of care. J Neurosurg. 2000;93(1 suppl):17.

  • 29

    Short DJ, El Masry WS, Jones PW. High dose methylprednisolone in the management of acute spinal cord injury—a systematic review from a clinical perspective. Spinal Cord. 2000;38(5):273286.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 30

    Hurlbert RJ. The role of steroids in acute spinal cord injury: an evidence-based analysis. Spine (Phila Pa 1976). 2001;26(24 suppl):S39S46.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 31

    Drazen JM, Morrissey S, Malina D, Hamel MB, Campion EW. The importance—and the complexities—of EMSCI. N Engl J Med. 2016;375(12):11821183.

  • 32

    Warren E. Strengthening research through data sharing. N Engl J Med. 2016;375(5):401403.

  • 33

    EMSCI ISNCSCI calculator. Accessed December 19, 2022. http://ais.emsci.org

  • 34

    Kirshblum S, Snider B, Rupp R, Read MS. Updates of the International Standards for Neurologic Classification of Spinal Cord Injury: 2015 and 2019. Phys Med Rehabil Clin N Am. 2020;31(3):319330.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 35

    International Standards for Neurological Classification of Spinal Cord Injury (ISNCSI) Worksheet. American Spinal Injury Association. Accessed December 19, 2022. https://asia-spinalinjury.org/wp-content/uploads/2019/10/ASIA-ISCOS-Worksheet_10.2019_PRINT-Page-1-2.pdf

    • Search Google Scholar
    • Export Citation
  • 36

    Fawcett JW, Curt A, Steeves JD, et al. Guidelines for the conduct of clinical trials for spinal cord injury as developed by the ICCP panel: spontaneous recovery after spinal cord injury and statistical power needed for therapeutic clinical trials. Spinal Cord. 2007;45(3):190205.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 37

    Bracken MB. Pharmacological treatment of acute spinal cord injury: current status and future prospects. Paraplegia. 1992;30(2):102107.

    • Search Google Scholar
    • Export Citation
  • 38

    Muthu S, Ramakrishnan E. Fragility analysis of statistically significant outcomes of randomized control trials in spine surgery: a systematic review. Spine (Phila Pa 1976). 2021;46(3):198208.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 39

    Yusuf S, Wittes J, Probstfield J, Tyroler HA. Analysis and interpretation of treatment effects in subgroups of patients in randomized clinical trials. JAMA. 1991;266(1):9398.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 40

    Alosh M, Fritsch K, Huque M, et al. Statistical considerations on subgroup analysis in clinical trials. Stat Biopharm Res. 2015;7(4):286303.

  • 41

    Assmann SF, Pocock SJ, Enos LE, Kasten LE. Subgroup analysis and other (mis)uses of baseline data in clinical trials. Lancet. 2000;355(9209):10641069.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 42

    Lagakos SW. The challenge of subgroup analyses—reporting without distorting. N Engl J Med. 2006;354(16):16671669.

  • 43

    Wang R, Lagakos SW, Ware JH, Hunter DJ, Drazen JM. Statistics in medicine—reporting of subgroup analyses in clinical trials. N Engl J Med. 2007;357(21):21892194.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 44

    Guideline for the Format and Content of the Clinical and Statistical Sections of an Application. Food and Drug Administration, Center for Drug Evaluation and Research; 1988.Accessed December 19, 2022. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/format-and-content-clinical-and-statistical-sections-application

    • Search Google Scholar
    • Export Citation
  • 45

    Collection, Analysis, and Availability of Demographic Subgroup Data for FDA-Approved Medical Products August 2013. Department of Health and Human Services, Food and Drug Administration; 2013.Accessed December 19, 2022. https://www.fda.gov/files/about%20fda/published/Collection--Analysis--and-Availability-of-Demographic-Subgroup-Data-for-FDA-Approved-Medical-Products.pdf

    • Search Google Scholar
    • Export Citation
  • 46

    Moreira ED Jr, Stein Z, Susser E. Reporting on methods of subgroup analysis in clinical trials: a survey of four scientific journals. Braz J Med Biol Res. 2001;34(11):14411446.

    • Crossref
    • Search Google Scholar
    • Export Citation

Supplementary Materials

  • Collapse
  • Expand
  • View in gallery
    FIG. 1.

    Graph showing the neurological injury level for the placebo (n = 113) and MP-C (n = 431) treatment groups in the new pooled data set. The MP-C patients are displayed both in total and stratified by the ATSCI study (ATSCIS) in which they were included. The incidence of ATSCI was greatest at the C4, C5, and C6 cervical levels, with a smaller increase in the midthoracic area.

  • View in gallery
    FIG. 2.

    Chart showing the number of placebo and the MP-N and MP-S patients at 26 weeks by injury severity classification group, divided into three different times from spinal cord injury (SCI) to initiation of the study drug to highlight the earlier dosages given to the MP-S group. Each injury severity cohort is classified as having a severe floor effect, measurement of drug effect improvement possible, or a severe ceiling effect. All injury severity groups except CMI included only a small and imbalanced number of patients from the NASCIS2 study. Note that the addition of the MP-S patients into the pooled data set to create the MP-C group significantly increased the number of MP patients available for analysis. *If 26-week follow-up data were missing, the 52-week follow-up examination was used. **The unknown time from injury to drug bolus was due to unknown injury time. ***MP dosing of a 30-mg/kg bolus followed by 5.4 mg/kg/hr for 23 hours.

  • View in gallery
    FIG. 3.

    Using the complete pooled database of n = 1074, a one-way analysis of improvement (Imp) in fractional LEMS from baseline to 26 weeks by injury severity grade with dithered points, mean diamonds, mean error bars, and comparison circles is displayed. The mean points of the injury severity improvements are connected as a visual aid to highlight the monotonic increasing difference between the C16–25 and C26–45 groups, which have the same ceiling effect. The comparison circles are a visual representation of the boundaries of the p = 0.05 paired Student t-test; when there is no overlapping, a statistical difference exists. The CMI, A+, B+, C04–15, and C16–25 baseline injury severity cohort groups had highly statistically significant different recovery patterns from one another (p < 0.0001), while the concentric nature of the two injury severity cohorts corresponding to C16–25 and C26–45 for baseline LEMS indicates no difference in their recovery potentials. The CMI group has a severe floor effect and the C16–25 and C26–45 groups a severe ceiling effect, which prevent their use for measuring drug effect differences.

  • View in gallery
    FIG. 4.

    Independently calculated p values for the main tests, subgroups, and sub-subgroups of the new NASCIS2 data set. Estimates for the p value correction for multiple tests are listed to the right of the p values but are not used as correction for the listed p values. The two main tests were statistically negative for both MP-N and naloxone compared with the placebo group. Even the time subgroups were negative, with one interesting p value (p = 0.0511) for MP-N having worse recovery than placebo in the 8- to 24-hour group. The sub-subgroups are separated by upper extremity (UE) versus lower extremity (LE), right versus left side of the body, and ≤ 8 hours versus > 8 hours from injury to initiation of the study drug. Of note is that the sub-subgroups of ≤ 8 hours for right upper extremity and right body side had calculated p values < 0.05. After applying a correction for multiple tests, no test approached statistical significance. *No correction for multiple tests was included in the listed p value. **Multiple test correction was calculated for equal p value distribution.

  • View in gallery
    FIG. 5.

    One-way ANOVA of motor recovery from baseline to 6 months for sub-subgroup by right and left sides of the body, based on the new NASCIS2 data set. A: Patient injury severities. B: Severities after removal of floor and ceiling effect patients. The one-way ANOVA of improvement in total LEMS at 26 weeks in the subgroup of patients with drug treatment initiated within ≤ 8 hours of injury and in the sub-subgroups separated by body side tests are presented two ways. This is a partial analysis presentation of the large list of sub-subgroups in Fig. 4. Note the continued decrease in the placebo mean recovery of the right side of the body in the ≤ 8-hour time sub-subgroup, which leads to a calculated right side of the body placebo versus MP-N result that meets the threshold of p < 0.05, uncorrected for multiple tests. BL = baseline; Std Err Diff = standard error difference.

  • View in gallery
    FIG. 6.

    Bar graph showing improvement in LEMS at 26 weeks by injury severity group for 1) all patients in the new data set; 2) placebo and MP-C groups; and 3) the subsets of placebo and MP-C patients ≤ 8 hours from injury to initiation of the study drug. The all-patients bar for each analysis group is the same as in Fig. 3; the bars are all p < 0.0001, statistically different from one another, and serve as reference values for the separate drug treatment groups. The only placebo group in the ≤ 8-hour placebo versus MP-C comparison is the A+ group, as it was the only one passing the threshold for analysis of more than 5 patients. Error bars are 95% CI.

  • View in gallery
    FIG. 7.

    Bar graph showing the C4–7 bony injury in CMI patients, assessing neurological improvement by evaluating descending levels from baseline to 26 weeks for motor, pinprick, and light touch in the cervical levels. The recovery amounts of these neurological functions for all patients are all statistically significantly different from one another (p < 0.0001), but MP provided no significant difference in recovery in any placebo versus MP-C analysis of any of the three neurological functions. Error bars are 95% CI.

  • 1

    Leary WE. Treatment is said to reduce disability from spinal injury. New York Times. March 31, 1990:1. https://timesmachine.nytimes.com/timesmachine/1990/03/31/078590.html?pageNumber=1

    • Search Google Scholar
    • Export Citation
  • 2

    Leary WE. Delay seen in publicizing spinal drug. New York Times. April 10, 1990:5.https://timesmachine.nytimes.com/timesmachine/1990/04/10/651990.html?pageNumber=45

    • Search Google Scholar
    • Export Citation
  • 3

    Associated Press. Spine treatment to be publicized. New York Times. April 12, 1990:21.https://timesmachine.nytimes.com/timesmachine/1990/04/12/161790.html?pageNumber=21

    • Search Google Scholar
    • Export Citation
  • 4

    Clinical Alert—New Treatment for Acute Spinal Cord Injury. NIH, National Institute of Neurological Disorders and Stroke; 1990.

  • 5

    Bracken MB, Shepard MJ, Collins WF, et al. A randomized, controlled trial of methylprednisolone or naloxone in the treatment of acute spinal-cord injury. Results of the Second National Acute Spinal Cord Injury Study. N Engl J Med. 1990;322(20):14051411.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 6

    ATLS Advanced Trauma Life Support Program for Doctors. 7th ed. American College of Surgeons; 2004.

  • 7

    Ito Y, Sugimoto Y, Tomioka M, Kai N, Tanaka M. Does high dose methylprednisolone sodium succinate really improve neurological status in patient with acute cervical cord injury? A prospective study about neurological recovery and early complications. Spine (Phila Pa 1976). 2009;34(20):21212124.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 8

    Matsumoto T, Tamaki T, Kawakami M, Yoshida M, Ando M, Yamada H. Early complications of high-dose methylprednisolone sodium succinate treatment in the follow-up of acute cervical spinal cord injury. Spine (Phila Pa 1976). 2001; 26(4):426430.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 9

    Kortbeek JB, Al Turki SA, Ali J, et al. Advanced trauma life support,. 8th edition, the evidence for change. J Trauma. 2008;64(6):16381650.

  • 10

    Hadley MN, Walters BC, Grabb PA, et al. Pharmacological therapy after acute cervical spinal cord injury. Neurosurgery. 2002;50(3 suppl):S63S72.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 11

    Hurlbert RJ, Hadley MN, Walters BC, et al. Pharmacological therapy for acute spinal cord injury. Neurosurgery. 2013;72(suppl 2):93105.

  • 12

    Bracken MB. Steroids for acute spinal cord injury. Cochrane Database Syst Rev. 2012;1(1):CD001046.

  • 13

    Hansebout RR, Kachur E. Acute traumatic spinal cord injury. UpToDate. Accessed December 19, 2022. https://www.uptodate.com/contents/acute-traumatic-spinal-cord-injury

    • Search Google Scholar
    • Export Citation
  • 14

    Fehlings MG, Wilson JR, Harrop JS, et al. Efficacy and safety of methylprednisolone sodium succinate in acute spinal cord injury: a systematic review. Global Spine J. 2017;7(3 suppl):116S137S.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 15

    Evaniew N, Belley-Côté EP, Fallah N, Noonan VK, Rivers CS, Dvorak MF. Methylprednisolone for the treatment of patients with acute spinal cord injuries: a systematic review and meta-analysis. J Neurotrauma. 2016;33(5):468481.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 16

    Liu Z, Yang Y, He L, et al. High-dose methylprednisolone for acute traumatic spinal cord injury: a meta-analysis. Neurology. 2019;93(9):e841e850.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 17

    Falavigna A, Quadros FW, Teles AR, et al. Worldwide steroid prescription for acute spinal cord injury. Global Spine J. 2018;8(3):303310.

  • 18

    Teles AR, Cabrera J, Riew KD, Falavigna A. Steroid use for acute spinal cord injury in Latin America: a potentially dangerous practice guided by fear of lawsuit. World Neurosurg. 2016;88:342349.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 19

    Adamczak SE, Hoh DJ. Steroids and spinal cord injury—a global dilemma. World Neurosurg. 2016;90:641643.

  • 20

    Bowers CA, Kundu B, Hawryluk GW. Methylprednisolone for acute spinal cord injury: an increasingly philosophical debate. Neural Regen Res. 2016;11(6):882885.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 21

    Lenzer J, Brownlee S. Methyprednisolone and acute spinal cord injury: redux. Rapid response. BMJ. 2008;336:532.

  • 22

    Geisler FH, Coleman WP, Grieco G, Poonian D. The Sygen multicenter acute spinal cord injury study. Spine (Phila Pa 1976). 2001;26(24 suppl):S87S98.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 23

    Geisler FH, Coleman WP, Grieco G, Poonian D. Measurements and recovery patterns in a multicenter study of acute spinal cord injury. Spine (Phila Pa 1976). 2001;26(24 suppl):S68S86.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 24

    Geisler FH, Coleman WP, Grieco G, Poonian D. Recruitment and early treatment in a multicenter study of acute spinal cord injury. Spine (Phila Pa 1976). 2001;26(24 suppl):S58S67.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 25

    Hanigan WC, Anderson RJ. Commentary on NASCIS-2. J Spinal Disord. 1992;5(1):125133.

  • 26

    Nesathurai S. Steroids and spinal cord injury: revisiting the NASCIS 2 and NASCIS 3 trials. J Trauma. 1998;45(6):10881093.

  • 27

    Coleman WP, Benzel D, Cahill DW, et al. A critical appraisal of the reporting of the National Acute Spinal Cord Injury Studies (II and III) of methylprednisolone in acute spinal cord injury. J Spinal Disord. 2000;13(3):185199.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 28

    Hurlbert RJ. Methylprednisolone for acute spinal cord injury: an inappropriate standard of care. J Neurosurg. 2000;93(1 suppl):17.

  • 29

    Short DJ, El Masry WS, Jones PW. High dose methylprednisolone in the management of acute spinal cord injury—a systematic review from a clinical perspective. Spinal Cord. 2000;38(5):273286.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 30

    Hurlbert RJ. The role of steroids in acute spinal cord injury: an evidence-based analysis. Spine (Phila Pa 1976). 2001;26(24 suppl):S39S46.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 31

    Drazen JM, Morrissey S, Malina D, Hamel MB, Campion EW. The importance—and the complexities—of EMSCI. N Engl J Med. 2016;375(12):11821183.

  • 32

    Warren E. Strengthening research through data sharing. N Engl J Med. 2016;375(5):401403.

  • 33

    EMSCI ISNCSCI calculator. Accessed December 19, 2022. http://ais.emsci.org

  • 34

    Kirshblum S, Snider B, Rupp R, Read MS. Updates of the International Standards for Neurologic Classification of Spinal Cord Injury: 2015 and 2019. Phys Med Rehabil Clin N Am. 2020;31(3):319330.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 35

    International Standards for Neurological Classification of Spinal Cord Injury (ISNCSI) Worksheet. American Spinal Injury Association. Accessed December 19, 2022. https://asia-spinalinjury.org/wp-content/uploads/2019/10/ASIA-ISCOS-Worksheet_10.2019_PRINT-Page-1-2.pdf

    • Search Google Scholar
    • Export Citation
  • 36

    Fawcett JW, Curt A, Steeves JD, et al. Guidelines for the conduct of clinical trials for spinal cord injury as developed by the ICCP panel: spontaneous recovery after spinal cord injury and statistical power needed for therapeutic clinical trials. Spinal Cord. 2007;45(3):190205.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 37

    Bracken MB. Pharmacological treatment of acute spinal cord injury: current status and future prospects. Paraplegia. 1992;30(2):102107.

    • Search Google Scholar
    • Export Citation
  • 38

    Muthu S, Ramakrishnan E. Fragility analysis of statistically significant outcomes of randomized control trials in spine surgery: a systematic review. Spine (Phila Pa 1976). 2021;46(3):198208.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 39

    Yusuf S, Wittes J, Probstfield J, Tyroler HA. Analysis and interpretation of treatment effects in subgroups of patients in randomized clinical trials. JAMA. 1991;266(1):9398.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 40

    Alosh M, Fritsch K, Huque M, et al. Statistical considerations on subgroup analysis in clinical trials. Stat Biopharm Res. 2015;7(4):286303.

  • 41

    Assmann SF, Pocock SJ, Enos LE, Kasten LE. Subgroup analysis and other (mis)uses of baseline data in clinical trials. Lancet. 2000;355(9209):10641069.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 42

    Lagakos SW. The challenge of subgroup analyses—reporting without distorting. N Engl J Med. 2006;354(16):16671669.

  • 43

    Wang R, Lagakos SW, Ware JH, Hunter DJ, Drazen JM. Statistics in medicine—reporting of subgroup analyses in clinical trials. N Engl J Med. 2007;357(21):21892194.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • 44

    Guideline for the Format and Content of the Clinical and Statistical Sections of an Application. Food and Drug Administration, Center for Drug Evaluation and Research; 1988.Accessed December 19, 2022. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/format-and-content-clinical-and-statistical-sections-application

    • Search Google Scholar
    • Export Citation
  • 45

    Collection, Analysis, and Availability of Demographic Subgroup Data for FDA-Approved Medical Products August 2013. Department of Health and Human Services, Food and Drug Administration; 2013.Accessed December 19, 2022. https://www.fda.gov/files/about%20fda/published/Collection--Analysis--and-Availability-of-Demographic-Subgroup-Data-for-FDA-Approved-Medical-Products.pdf

    • Search Google Scholar
    • Export Citation
  • 46

    Moreira ED Jr, Stein Z, Susser E. Reporting on methods of subgroup analysis in clinical trials: a survey of four scientific journals. Braz J Med Biol Res. 2001;34(11):14411446.

    • Crossref
    • Search Google Scholar
    • Export Citation

Metrics

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 1344 1344 1344
PDF Downloads 1214 1214 1214
EPUB Downloads 0 0 0