Injury to Time of Marked Recovery as a New Acute Spinal Cord Injury Outcome Measurement for Statistical Analysis of Interventions and Drug Treatments

Abstract

No pharmacologic agent has demonstrated long-term neurological recovery following acute traumatic spinal cord injury (ATSCI). Thus, consideration of alternative outcome indicators is reasonable. A new and novel outcome variable, time to recovery, is introduced and compared to long-term recovery, examining nine variables for a two-grade improvement (Marked Recovery [MR]). The updated historical US FDA IND SYGEN ATSCI database (n = 760 patients) was used, as it is unique in having sufficient neurological examinations (4, 8, 16, 26, and 52 weeks) to allow the analysis. The time to MR was defined as the geometric mean between the time to first MR and the preceding examination with no MR. The SYGEN group exhibited the maximum fractional MR disparity at 8 weeks (p = 0.0248), coinciding with the discontinuation of the SYGEN study drug treatment. In the model fit analysis using long-term MR outcomes (26–52 weeks), significant variables were ASIA Injury Score (AIS) baseline grade (p < 0.0001), cervical versus thoracic injury (p = 0.0334), and direct admission versus transfer to spinal cord injury center (p = 0.0269). In the model fit analysis using time to MR, significant variables were AIS baseline grade (p < 0.0001), SYGEN versus placebo (p = 0.0227), age <30 years old (p = 0.0248), and early surgery ≤72 h (p = 0.0350). For time to MR, the combined effects of SYGEN and early surgery were additive, with a total decrease of 40.87 days (p = 0.0041) compared with placebo and late surgery. The time to MR is a new and novel metric for evaluating differences in recovery, as demonstrated by finding significant statistical relationships in several variables. The mechanism of action of a shortening of time to MR between groups after ATSCI is an augmentation of the body’s inherent neurological healing process. Clinically, this results in shorter rehabilitation time.

Introduction

Acute Traumatic Spinal Cord Injury (ATSCI) causes varying degrees of neurological impairment, with the most severely injured left with permanent disability and significantly reduced life expectancy.^1,2 ATSCI peaks in younger adults 18–35 and in those over 65 years old and causes significant socioeconomic impact.^2,3 Despite multiple ATSCI clinical drug studies published in the last 35 years, none has resulted in Food and Drug Administration (FDA)-approved drug treatments. Possible reasons include ineffective drugs, insufficient precision in clinical measurement tools, or inadequate assessment of neurological recovery.

Previously reported clinical trials for ATSCI only tested long-term improvement from baseline. ⁴ Table 1 lists these clinical trials since 1992. The SYGEN ATSCI study represents Class I Medical Evidence ⁵ and is the only ATSCI study that collected clinical exams at intermediate time points, allowing a time to recovery analysis.

Table 1.

Characteristics of Prospective Randomized Studies Investigating Traumatic Acute Spinal Cord Injury with References

Study	Study groups, sample size	Steroids for both groups	Injury to follow-up in weeks						Sample size at baseline of ASIA As, ASIA Bs	Sample size at 6 or 12 months of ASIA a_Imp, B _Imp
			4	8	12	16	26	52
Bracken et al.10,40	MP 24 h, n = 162 Naloxone 24 h, n = 154 Placebo 24 h, n = 171	MP N N	○	○	●	○	●	●	132, Unknown 125, Unknown 132, Unknown	7,6 6,14 7,10
Bracken et al.41,42	MP 48 h, n = 167 MP and tirilazad 48 h, n = 166 MP 24 h, n = 166	Y Y Y	○	○	○	○	●	●	Unknown	Unknown
Geisler et al.6–8	SYGEN, n = 430 P, n = 330	Y Y	●	●	○	●	●	●	235, 71 195, 49	36*, 41* 30*, 28*
Kwon et al. 43	CSF drainage, n = 11 No CSF drainage, n = 11	N	○	○	○	○	●	○	6, 4 6, 2	0*, 2* 0*, 1*
Casha et al. 44	Minocycline, n = 27 Placebo, n = 25	N	○	○	●	○	●	●	6, 0 6, 2	0*, 0* 0*, 1*
Alibai et al. 45	EPO, n = 15 Placebo, n = 15	Y Y	●	○	○	○	●	○	8, 0 7, 2	1*, 0* 1*, 1*
Rahimi-Movaghar et al. 46	Early surgery, n = 16 Delayed surgery, n = 19	Y Y	○	○	○	○	●	●	7, 1 9, 5	1*, 0* 1*, 2*
Costa et al. 47	EPO, n = 11 MP, n = 8	—	●	●	●	○	○	○	10, 1 6, 2	1*, 0* 0*, 1*
Meshkini et al. 48	Riluzole, n = 30 No Riluzole, n = 30	Unknown	○	○	●	○	○	○	11, 9 12, 9	4, 5* 1, 5*
Nagoshi et al. 49	Hepatocyte growth factor, n = 26 Placebo, n = 17	Unknown	●	●	●	●	●	○	15, 11 16, 1	2*, 6* 2*, 0*
Chikuda et al. 29	Early surgery, n = 37 Delayed surgery, n = 33	Y Y	○	○	●	○	●	●	0, 0 0, 0	0, 0 0, 0
Ganjeifar et al. 50	EPO, n = 27 No EPO, n = 27	Y Y	○	○	○	○	●	●	22, 2 23, 2	15, 1* 17, 1*
Koda et al. 51	G-CSF, n = 43 Placebo, n = 45	N	○	○	●	○	●	●	1, 12 0, 11	0*, 6* 0*, 6*
Kumarasamy et al. 52	Riluzole, n = 23 No Riluzole, n = 20	N	○	○	●	○	●	●	0, 4 0, 9	0, 4 0, 9
Saini et al. 53	BMSC, n = 14 Placebo, n = 13	N	○	○	○	○	●	○	Unknown	Unknown
Fehlings et al. 54	Riluzole, n = 96 Placebo, n = 97	N	○	○	○	○	●	○	49, 19 52, 19	7*, 10* 7*, 10*

ASIA A_Imp are patients with Baseline ASIA A grade that had neurological improvement in the lower extremities at 6 or 12 Month follow-up.

ASIA B_Imp are patients with Baseline ASIA B grade that had neurological motor improvement in the lower extremities at 6 or 12 Months follow-up.

*Estimated fraction of historical ASIA A_Imp and historical ASIA B_Imp.

ASIA, American Spinal Injury Association score; Adm, admission; BMSC, bone marrow derived stem cells; EPO, erythropoietin; G-CSF, granulocyte-colony stimulating factor; GM, ganglioside; MP, methylprednisolone; N, no; n, number of patients; N/A, not available; P, placebo; Wk, week; Y, yes.

The SYGEN ATSCI study found no statistically significant drug effect of SYGEN compared with placebo on testing the fraction of Marked Recovery (MR) in long-term follow-up.^6–8 A fourth publication in 2023 combined the SYGEN ATSCI study’s placebo group and NASCIS2 datasets ⁹ and found no methylprednisolone effect on ATSCI. The MR comparison of the SYGEN group with placebo group with time did show a statistically significant faster recovery in the original 2001 publication (Figs. 4–9, 12, and 13 ⁶ ), a positive result not fully explored in the prior publications. The objective of the current study is to model fit nine variables to assess their statistical significance in both the SYGEN ATSCI study’s primary outcome—the fraction of long-term MR—and its secondary outcome—time to MR recovery as a new outcome variable. Figure 1 conceptually shows the contrasting recovery patterns for the two approaches.

FIG. 1.

Two distinct types of group recovery after ATSCI are conceptually presented in this figure. Both outcomes require different outcome variables to evaluate fit variables for a statistical effect. On the right is a representation of a difference in long-term recovery between treatment groups; this is the analysis performed by all previous ATSCI studies, including the 2001 papers on the SYGEN ATSCI study. On the left is a representation of faster recovery between treatment groups, which has been referred to in the literature as a left shift of the recovery curve. The time to MR is a new outcome variable introduced and analyzed in this paper to assess the time to recovery. This assessment was (1) a prespecified secondary outcome in the SYGEN study; (2) the reason for collection of extra examinations in the Recovery Period (4, 8, 16 weeks), and (3) reported to have a beneficial clinical effect in FDA reports and the original SYGEN ATSCI study publications with an analysis like that presented in Figure 4 of this paper. ATSCI, acute traumatic spinal cord injury; FDA, Food and Drug Administration; MR, Marked Recovery.

The U.S. FDA has approved several drugs with labeling that supports accelerated recovery-related outcomes. Examples include (1) repository corticotropin injection (Acthar Gel) for acute exacerbations of multiple sclerosis; (2) filgrastim and sargramostim for marrow suppression related to chemotherapy, leukemia, or transplant, where approval was based on reducing the duration of neutropenia and accelerating hematological/myeloid recovery; and (3) selected antivirals, such as oseltamivir for influenza and acyclovir, valacyclovir, and famciclovir for herpesvirus infections. In all these treatments, approval was based on shorter time to symptom improvement or lesion healing. Thus, FDA-recognized acceleration of recovery is clearly established in multiple sclerosis relapses, after cancer therapy or transplantation, and in certain acute viral infections. The purpose of the present study is to demonstrate that speed to recovery is also an important variable for evaluating treatments in ATSCI.

Methods

The FDA IND SYGEN ATSCI study^6–8 is a prospective, placebo-controlled, double-blinded, block-randomized, multi-center drug trial comparing SYGEN (monosialoganglioside sodium) with placebo. The block randomized variables were (1) baseline ASIA AIS grade grouped as A, B, or C and D; (2) cervical or thoracic anatomical location of SCI; and (3) age (<30 versus ≥30 years old). The 760 patients (430 SYGEN and 330 placebo) were enrolled in 28 neurotrauma centers in North America between Q2 1992 and Q1 1997. Inclusion criteria were: (1) Major spinal cord injury with a neurological deficit in one lower extremity having a total ASIA motor score ≤ 15/25; (2) Spinal cord injury at or rostral to T10 without a spinal cord transection; (3) Methylprednisolone given within 8 h of injury per NIH NASCIS2 dosing recommendations;^9,10 (4) Baseline neurological exam and loading dose of study medication (SYGEN vs. placebo) administered after the 24-h methylprednisolone infusion and before 72 h from injury (mean 54.9 h). After the initial bolus dose of SYGEN, a daily treatment period of 56 days was begun. The 2-month dosing period for SYGEN was based on published animal and human studies,^11–14 the prior FDA IND ATSCI study, ¹⁵ and the safety profile history of SYGEN, which included an observational study on 25,422 patients treated with GM-1. ¹⁶ SYGEN is currently marketed in multiple countries worldwide (TRB CHEMEDICA SA, Switzerland), although not in the USA or Canada.

All neurological exams were performed by trained physical therapists over 20–30 min, testing every myotome and dermatome of the ASIA data collection tool. Follow-up neurological exams were planned at 4, 8, 16, 26, and 52 weeks after the ATSCI. Details of enrollment, demographics, and long-term follow-up comparisons were presented in the original 2001 series of three articles,^6–8 and the modernization of the ASIA AIS grades used in this study was described in a fourth publication. ⁹

The definition matrix of MR versus No MR is presented in Figure 2. The MR definition is the same as the two-grade ASIA Injury Score (AIS) improvement^17–19 for baseline grades of A, B, and C. The associated modified Benzel scale for follow-up expands the baseline AIS D grade into three ordinal groupings according to walking ability. The placebo group had 111 MR and 219 No MR patients, and the SYGEN group had 144 MR and 286 No MR patients. There were 42 deaths, and 43 patients were lost to follow-up without 26 or 52 week neurological examinations. All patients who died or had not achieved MR before being lost to follow-up were classified as No MR patients. The percentage of patients with MR in the SYGEN ATSCI study was highly dependent (p < 0.0001) on baseline ASIA AIS grade: AIS A, n = 70 of 461 (15.2%); AIS B, n = 59 of 140 (42.1%); and AIS C and D, 126 of 159 (79.25%).

FIG. 2.

The definition matrix Panel (A) is a visual representation of the definitions of Marked Recovery (MR) and No Marked Recovery (No MR) regions for differing baselines and follow-up examinations. The specific definitions for the contemporary ASIA AIS grades were used for baseline categorization and are listed in Panel (B). The modified Benzel Classification (MBC) was used in follow-up examinations; definitions are listed in Panel (C). The ASIA AIS grades of A, B, and C mirror the MBC grades of 1, 2, and 3. The ASIA AIS grade of D is separated into the MBC grades of 4, 5, and 6 based on walking ability. The use of the MBC rather than the ASIA AIS grade for follow-up allows a potential two-grade improvement for the ASIA AIS C patients with significant recovery but does not require the recovery to be complete. The MBC also allowed a two-grade improvement for the ASIA AIS D patients, whereas if the ASIA AIS grade were used in follow-up a two-grade improvement would be impossible. The determination of MR or No MR in (A) converts the different baselines and follow-up grades into a single binary variable for each patient. Patients who died or were lost to long-term follow-up without a recorded MR were classified as No MR. Thus, all 760 patients who entered the study have a MR or No MR designation.

The protocol time intervals, along with the actual times of follow-up, are shown in Figure 3. Although the mean times are reflective of the protocol time intervals, there is a standard deviation about the mean time of approximately 10% with outliers.

FIG. 3.

The distribution of actual injury times to follow-up examinations for each protocol follow-up time interval is presented. The top five histograms show actual times of injury to follow-up for each planned follow-up interval (4, 8, 16, 26, and 52 weeks). The table below lists descriptive statistics for the five-time distributions. The standard deviation for each protocol interval is between 9.3% and 13.0% of its mean, representing a sizable error that can be easily corrected by using the exact values. Although time intervals were used in Figure 2, the exact times of the clinical examinations were used in the calculation of the estimated time to MR. The use of the exact times of examinations provides the more precise presentation of the time to MR analysis described in this paper.

The time of MR is between the first exam where MR is noted and the prior examination with NO MR. The protocol clinical examination times of data collection for this study were in an almost perfect geometric series at 4, 8, 16, 26, and 52 weeks from injury. Thus, the clinical examinations are almost exactly evenly spaced after a logarithmic transformation. The time of the MR can be estimated within this geometric series to be the geometric mean of two times: injury to first MR and injury to the prior examination with NO MR noted. This places the estimate of time from injury to MR to be equally spaced between the clinic exam times after a logarithmic transformation and preserves the geometric series nature of the collected data points. The geometric mean can be calculated using two mathematical formulas, as listed below. The square root of the two numbers multiplied together is easier to calculate, whereas the formula on the far right below visualizes that the geometric mean is the average of the two numbers after a logarithmic transformation.

$$\mathit{Geometric}\mathit{Mean}=\sqrt{A\ast B}={10}^{\left\{\frac{\mathit{log}10\left(A\right)+\mathit{log}10\left(B\right)}{2}\right\}}$$

where A = time from ATSCI to the clinical examination with the first MR and

B = time from ATSCI to the clinical examination before A with NO MR.

Both A and B are positive numbers, as the ATSCI time is set to zero.

The estimate of the time from injury to MR calculated by using the geometric mean produces one number for each patient that can be used as an outcome variable in multivariate analysis. The actual times of the clinical examinations were used in the geometric mean calculations to estimate the time to MR. Figure 3 shows the distributions of times around the protocol specified times; the same distributions cause the spread of points noticed in Figure 7.

FIG. 7.

Panel (A) is a comparison of time to MR for the SYGEN versus the Placebo group. The points are dithered, and green diamond overlays show the mean and 95% confidence intervals from a One-way Anova analysis. Mean values for the distributions are listed at the top of the graphs. All 255 patients with MR before 12 months were included in this analysis. The SYGEN group had a 21.19-day faster mean time to first MR (91.95 days for Placebo vs. 70.76 days for SYGEN, p = 0.0024). Panel (B) is the cumulative distribution function (CDF) for (A), which graphically presents regions of different recovery rates. Panel (C) separates SYGEN versus Placebo groups by surgical timing of prompt (≤72 h) versus delayed (>72 h) or no surgery to create four groups for comparison. The SYGEN with prompt surgery group has the fastest time to first marked recovery. The time difference to MR for Group 1 (Sygen and prompt surgery) versus Group 4 (placebo and delayed surgery) was 40.87 days, p = 0.0041 and for Group 1 (SYGEN and prompt surgery) versus Group 3 (placebo and prompt surgery) was 24.24 days, p = 0.0230. These comparisons verify that the beneficial effects of SYGEN and prompt surgery are additive for the outcome variable of time to MR and that SYGEN can improve the time to MR for patients with prompt surgery. Panel (D) is the CDF for (C), which graphically presents regions of different recovery rates.

Model fits of both outcome variables, the fraction of long-term MR conversions and the estimated time from injury to MR, were performed using the same nine model fit variables. These nine tested model effects are in three groups: (1) The SYGEN ATSCI study’s primary variable (SYGEN versus placebo); (2) Block randomized variables (ASIA AIS baseline grade, cervical versus thoracic, and age < versus ≥30 years old); and (3) other variables (gender, direct admission versus transfer, start of methylprednisolone, surgical timing, and mean blood pressure threshold). Nonsignificant model fit variables were sequentially eliminated until only significant (p < 0.05) ones remained.

Primary and secondary endpoints were prospectively specified before data collection started. Because no prespecified multiplicity-adjustment strategy was included in the protocol or statistical analysis plan, hypothesis tests for the secondary end-point were performed using nominal two-sided p values and 95% confidence intervals and were interpreted as supportive rather than confirmatory. In the absence of a hierarchical testing procedure or other alpha-allocation framework, statistical inference for the secondary end-point was not used to establish formal trial success, and any findings for secondary outcomes were considered exploratory, with appropriate caution regarding inflation of the family-wise type 1 error rate. JMP® Version 19.1.2 was used for descriptive statistics, analysis of variance (ANOVA), life distribution, and model fitting analysis.

Results and Discussion

Time interval MR evaluation

Using the fraction of MR patients in each of the time intervals to evaluate the long-term outcome (Fig. 4) revealed that the SYGEN group experienced earlier MR compared with the placebo group at all time points (p = 0.0327, Fisher’s Exact, two-tailed). The maximum difference in recovery rate was at the 8-week follow-up (p = 0.0248, Fisher’s Exact, two-tailed), coinciding with the end of the 56-day SYGEN drug therapy. No difference between the SYGEN and placebo groups was noted in the 6- and 12-month fractions of MR. However, this graph discloses that the fraction of SYGEN MR at 8 weeks was equivalent to the fraction of placebo MR at 16 weeks, consistent with a faster recovery time of 8 weeks for SYGEN.

FIG. 4.

The cumulative fraction of Marked Recovery (MR) for the SYGEN and Placebo groups at the protocol follow-up time intervals is bar graphed in this figure. The drug treatment period of 8-weeks is signified by a banner extending to the 8 week time point, which was also the time of maximum SYGEN-Placebo significance. The SYGEN group maintained a larger cumulative MR fraction than the placebo group until the 26- and 52-week comparisons, when they were essentially equivalent. The 4-, 8-, and 16-week time intervals were labeled as the Recovery Period. The 26- and 52-week time points are labeled as the long-term outcome. The recovery of the SYGEN group preceded the placebo group throughout the first 16 weeks, which signifies an earlier recovery pattern for SYGEN compared with Placebo. The SYGEN and Placebo groups were tested across the time course overall with a 2 by 6 Fisher’s Exact Test, two-tailed (p = 0.0327), for the five first MR conversion times combined with No MR. In addition, the cumulative fraction of the SYGEN and placebo groups at the 8 week follow-up had the largest group difference, with a p = 0.0248 (Fisher’s Exact Test, two-tailed). Both tests signify a statistically significant faster recovery for the SYGEN patients compared with the placebo patients.

Model fit results for MR fraction at long-term follow-up

Figure 5 has three tables of model fit using a nominal logistic model for the binary outcome variable of MR versus No MR in long-term follow-up. Figure 5A is the starting model fit with all nine initial variables. Figure 5B is an intermediate model fit with four variables remaining after successive elimination of most of the nonsignificant ones. Figure 5C is the model fit after successive elimination of all of the nonsignificant variables; the remaining three statistically significant model fit variables are (1) ASIA baseline AIS grade of A, B, or C and D (p < 0.0001); (2) cervical versus thoracic anatomical region (p = 0.0334); and (3) direct admission to a specialized SCI center versus initial care at a community hospital with subsequent transfer (p = 0.0269).

FIG. 5.

Long-term recovery for the fraction of MR was investigated with a nominal logistic fit model for the nine candidate fit variables. Panel (A) is the starting model fit of parameters and output. After sequential elimination of nonsignificant variables, an intermediate model fit of four variables is provided in Panel (B). Panel (C) is the final model fit for only the statistically significant parameters: (1) ASIA AIS baseline grade of A, B, and C and D (p < 0.0001) [the C and D groups were combined because of similarity in this model fit], (2) cervical versus thoracic anatomical SCI (p = 0.0334), and (3) direct admission versus transfer to an SCI study facility (p = 0.0269). SCI, spinal cord injury.

The first two model fit variables, baseline AIS grade and the anatomical injury region, need to be balanced in analyses of long-term MR fraction from ATSCI. The choice of hospital for initial care can be made to optimize the fraction of long-term MR. A subgroup analysis on the baseline AIS grades and anatomical regions disclosed that the overall statistical significance of direct admission was driven by the cervical injuries in the baseline AIS A group (MR 13.02% [25 of 192] in the transfer group versus 26.83% [33 of 123] in the direct admission to SCI center group; Fisher’s Exact two-tailed, p = 0.0027).

Model fit results for time from injury to estimated MR

Figure 6 has three tables of model fit using a standard least square effect leverage model for the continuous outcome variable of time to MR. Figure 6A is the starting model fit with all nine initial model fit variables; Figure 6B is an intermediate model fit with the five variables remaining after successive elimination of most of the nonsignificant ones, and Figure 6C is the model fit with the remaining four statistically significant variables after successive elimination of all the nonsignificant ones. The significant variables were (1) SYGEN drug group (p = 0.0227); (2) ASIA baseline AIS grades of C and D, with a faster recovery compared with the A and B (p < 0.0001); (3) Age group < 30 years old (p = 0.0248); and (4) Spinal surgery ≤ 72 h (p = 0.0350).

FIG. 6.

A standard least squares effect leverage model fit was utilized for evaluating the time to MR as an assessment of the time to marked recovery. The time to MR was estimated as the geometric mean of the time to first marked recovery and the time of the prior exam without MR. Panel (A) is the starting model fit of parameters and output with the nine candidate fit variables. After sequential elimination of nonsignificant variables, an intermediate model fit of five variables is provided in Panel (B). Panel (C) is the final model fit for only the statistically significant parameters: (1) ASIA AIS baseline grade of A and B and C and D (p < 0.0001), [the A and B and the C and D groups were combined because of similarity in this model fit]; (2) Sygen versus placebo (p = 0.0227); (3) age < 30 years old (p = 0.0248); and (4) Injury to surgery ≤72 h (p = 0.0350).

SYGEN versus placebo

SYGEN versus placebo was a block randomized variable, and the model fit (Fig. 6C) of the time to MR for the SYGEN group was decreased from 91.95 to 70.76 days (21.19 days, p = 0.0227, a 23.05% decrease) compared with the placebo group.

The distributions of the SYGEN and placebo groups of time to MR are directly compared in Figure 7A. This direct comparison yielded a difference of 21.19 days, with a significant p value (p = 0.0024). The difference in p values is because other model fit variables were not included in this direct comparison. The similarity between the model fit and the direct comparison demonstrates the success of the block randomization protocol. Figure 7B is the Cumulative Distribution Function (CDF).

Baseline ASIA AIS grade

Baseline ASIA AIS grade was a block-randomized variable and was the most significant factor in determining outcome for both greater long-term MR recovery and the time to MR recovery. It was the only model fit variable that was highly statistically significant (p < 0.0001) for both model fit outcome variables (Figs. 5C and 6C).

Cervical versus thoracic injuries in neurological recovery

Cervical versus thoracic anatomical area was a block-randomized variable, and the model fit for improved MR fraction showed a larger fraction of MR in the cervical group in long-term follow-up, statistically significant at p = 0.0334 (Fig. 5C).

Age < 30 versus ≥30 years old

Age < 30 versus ≥30 years old was a block-randomized variable. The age < 30-year-old group demonstrated a statistically significant beneficial effect compared with ≥30-year-old patients with the outcome variable of time to MR. The model fit of time to MR for the age < 30-year-old group was decreased by 13.73 days (86.39–72.66 days, 15.89%) (Fig. 6C).

Methylprednisolone

Much has been said and written about the NASCIS2 methylprednisolone (MPSS) recommendation regarding treatment for ATSCI since 1990, when the NIH sent its methylprednisolone protocol to all ERs in the US before the publication of the initial scientific paper ¹⁰ in the New England Journal of Medicine. Recently, the NASCIS2 study and its statistical analysis have been discredited with a reanalysis of the original NASCIS2 dataset in combination with the SYGEN dataset. ⁹ In addition, methylprednisolone has only been examined in the literature for differences at long-term follow-up, but never as a drug that altered the time to recovery. Animal experiments have indicated that the beneficial effect of methylprednisolone is greater when it is administered closer to the ATSCI.

The time from injury to initiation of the methylprednisolone loading dose was converted into a binary variable split at the median of 2.333 h to search for a time-dependent effect of methylprednisolone. The new binary variable of early versus later MPSS administration was used in the model fit. There was no statistically significant difference between the two groups. Thus, this study provides additional evidence that MPSS does not augment recovery after ATSCI for either long-term fraction of MR or time to MR outcomes (Figs. 5A and 6A, p = 0.88 for long-term MR and p = 0.75 for time to MR).

Gender (male versus female)

There was no statistically significant difference between the two genders for either of the outcome variables (Fig. 5A and B, p = 0.64 for long-term MR and p = 0.36 for time to MR).

Early surgery ≤72 h from injury

Timing of surgical intervention after ATSCI has been debated for two decades, with several contemporary recommendations in the ATSCI literature of ≤24 h,^20–24 although Sheldon et al., ²⁵ in a 2025 review article, separated the early versus late grouping at 96 h, a notably divergent approach. Despite uncertainty in the literature^26–28 on timing and efficacy, prompt surgical decompression and stabilization have become standard treatments after ATSCI.^21,29–31 For instance, the most cited article reporting improvement in the <24 h time from injury to surgery is the STASCIS study ²¹ in 2012, referenced over a thousand times and described as a landmark paper, which is the basis of several clinical guidelines. However, this study was an unrandomized study and only examined injury to surgery time at one time point < 24 h. The baseline AIS grade effect on the fraction of two-grade improvements in the STASCIS and the present study’s databases were both highly significant, p < 0.0001. To compare the STASCIS paper with the results of this paper, one must first correct the imbalance in the baseline AIS groups left uncorrected in the STASCIS study report. When the Cochran–Mantel–Haenszel test is applied to correct it, the two AIS grade improvement test changes from the reported value of p = 0.036 to the insignificant p = 0.1727. Thus, the STASCIS and the present study actually both have the same not-significant statistical conclusion for the fraction of 2 grade improvers in long-term outcome. The STASCIS study did not collect neurological examinations during the recovery period, so it cannot be compared with the current study’s significant results regarding surgical timing for the outcome variable of time to MR.

Dividing the surgery group in the first month approximately in half according to injury to surgery time of ≤72 h versus surgery >72 h or no surgery (n = 293, 467, respectively) ensured a balanced number of patients in the tested groups. The ≤3 day timing for early surgery is supported by: (1) No SCI animal evidence that severely compressed spinal cord tissue decompressed after more than a few minutes survives to recover; and (2) The authors’ belief that any positive benefits of early surgery arise from bony stabilization preventing reinjury of the spinal cord from unstable fracture motion ³² or from the medical benefits that accompany early post-injury mobilization. The outcome variable of time to MR demonstrated that the early surgery group, ≤72 h from the ATSCI injury, had a statistically significant beneficial effect. The model fit of time to MR for the Early Surgery ≤72 h from ATSCI Injury group was statistically significant with a p = 0.0350 for the decrease of 17.01 days (86.32–69.31 days, 19.71% reduction) (Fig. 6C).

Direct admission to SCI center versus transfer to SCI center

Specialized ATSCI centers were established to improve outcomes by reducing time from injury to initiation of definitive care. In the SYGEN ATSCI study, patients were enrolled in 28 specialized ATSCI centers in North America as both direct admissions and as transfers from community hospitals. The time from injury to direct admission (n = 288) at the SYGEN study facilities was 1.49 h mean and 0.967 h median, and for the transfer patients (n = 455), it was 9.16 h mean and 5.55 h median (p < 0.0001). The 17 patients not included in this timing demographic were missing injury or admission times.

The direct admission patients had a significant statistical advantage in the model fit fraction of MR in long-term follow-up (p = 0.0269) (Fig. 5C). A subgroup analysis of direct admission patients revealed that cervical injuries drove this significance. Thirty-three of 123 direct admission cervical baseline AIS A patients had MR (26.83%), while 25 of 192 transfer admissions first seen at a community hospital had MR (13.02%); the difference between these two MR fractions is statistically significant (p = 0.0027, Fisher’s Exact Test, two-tailed).

Mean blood pressure (MBP) augmentation to ≥85 mm Hg after ATSCI

In the literature, blood pressure augmentation after ATSCI engenders evolving opinions. The largest prospective cohort (nonrandomized) study on this subject is the 19-center EAST study ³³ (2021–2023, n = 222 with baseline and discharge ASIA), which found no association between adherence to mean arterial pressure (MAP) >85 mm Hg or duration of MAP augmentation and ASIA improvement. A recent systematic review and meta-analysis ³⁴ found that lower BP thresholds were associated with worse functional outcomes up to one year and supports the rationale for MAP augmentation. A 2025 randomized study (n = 92) did not find a difference in outcome with MAP augmentation (>85 to 90mmHg). ³⁵ The most recent clinical guideline on MAP management after acute spinal cord injury, listed in the 2024 Global Spine Journal Clinical Practice Guideline, ³⁶ recommended augmentation to >85 mm Hg.

Blood pressure augmentation in ATSCI patients was included as part of the SYGEN ATSCI study initial management protocol and discussed at the clinical investigators’ meetings. Blood pressure was measured at the initiation of the study medication (55.4 ± 13.9 h from injury) as part of the study dataset and was a planned secondary variable. To investigate the augmentation of BP in the SYGEN ATSCI dataset, a new binary variable was created for MBP of <85 or ≥85 mm Hg.

There was no statistically significant difference between the patients with or without mean blood pressure augmentation to ≥85 mm Hg around 48 h from ATSCI in either outcome model fit analysis (Figs. 5A and 6A with p = 0.91 for long-term MR and p = 0.11 for time to MR). Thus, this study does not support augmentation of MBP in the early post-injury period.

Interaction of SYGEN and timing of surgery

The two model fit variables for time to MR of SYGEN versus placebo and prompt versus deferred or no surgery are choices that can be decided after the ATSCI. Both had statistically significant effects, as presented above. Thus, the interaction of these two effects could be additive or equivalent, i.e., does using either alone yield the same decrease in time to MR as using both, or does the combination result in an even greater decrease in time to MR? Figure 7C examines the relationship by subdividing the patients into four groups: (1) SYGEN and prompt surgery ≤72 h; (2) SYGEN and surgery >72 h or no surgery; (3) placebo and prompt surgery ≤72 h, (4) placebo and surgery >72 h or no surgery. Figure 7D is the CDF of 7C. Visual inspection of Figure 7C and D concludes that the times to MR for groups 2 and 3 are midway between group 1 and group 4, which have a statistically significant difference of 40.87 days (98.69 days for group 4 and 57.82 days for group 1, p = 0.0041, decrease of 41.4%).

Thus, the effects of SYGEN drug treatment and prompt surgery ≤72 h demonstrated that the beneficial effects of the time to MR of each were additive. The SYGEN beneficial effect was also statistically significant for SYGEN versus placebo in a subgroup analysis comparing only those patients with prompt surgery. In this subgroup, the difference in time to MR was 24.24 days (82.06 days for group 3 and 57.82 days for group 1, p = 0.0230, a decrease of 29.5%). Thus, SYGEN demonstrates statistical significance for the reduction of time to MR in all study patients as well as in the subgroup of patients with prompt time to surgery. These comparisons verify that the beneficial effects of SYGEN and prompt surgery are additive for the outcome variable of time to MR.

Table 2 is a comparison of two statistical analysis techniques evaluating Time to MR: 2A) is the Life Distribution and Window Interval Censored MR times using a Weibull Distribution Model, and 2B) is the Geometric Mean Estimate conclusions of MR Conversion Time used in one-way Anova Analysis, as presented in Figure 7. Note the similarity of the two different statistical analyses for the group means, group mean differences from the fastest MR group, and the p values of comparison from the fastest MR group to the other groups. This comparison table provides statistical evidence for the equivalence of the two statistical analysis methods on this dataset and provides confidence in the use of the geometric mean to estimate the time to MR.

Table 2.

Comparison of Two Statistical Analysis Evaluating Time to Marked Recovery (MR)

A) Time to MR evaluated with Life Distribution and Window Interval Censored using a Weibull Distribution Model
Groups for comparison	Mean	Standard error	n*	Mean difference between SYGEN and placebo	p-value*
SYGEN	71.33 Days	5.40 Days	144
Placebo	92.20 Days	6.89 Days	111	20.87 Days	0.0180

Note the similarity of the two different statistical analyses for the group means, group mean differences from the fastest MR group, and the p-values of comparison of the fastest MR group to the other groups. This comparison shows the statistical equivalence of the two statistical analyses methods in this dataset.

p < 0.05 were considered significant and highlighted in bold format.

*t-test, unequal variances (Welsh-Satterthwaite), two-tailed.

The Life Distribution with Window Interval Censoring statistical analysis is useful for parametric analysis of time-to-event data; however, its results are inherently sensitive to the model fit of the failure-time distribution model used. For this dataset, the Weibull model was the best and was also a good visual fit. Life Distribution analysis also has limited capability to handle complex covariate structures and is not suitable for multiple covariate analysis. Assigning an estimate for the time of conversion to MR with the geometric average allows use of the more powerful multivariate fitting techniques, as shown in Figure 6 to test multiple covariate analyses.

Change of medical care with time for ATSCI patients

The MR timing differences reported in this article require additional verification to achieve recommended status, as these are secondary analyses. The direct application of the results to contemporary SCI populations may not be appropriate because of changes in medical practice since this 30-year-old dataset was collected. This analysis serves as a guide for future ATSCI studies for data collection and statistical analysis planning. Medical treatment has evolved in the last 30 years and generally been regarded as beneficial to the patient. However, actual studies of improved neurological outcome with newer treatments after ATSCI are lacking. The literature can be examined for differences in outcome from ATSCI with time by comparing the two AIS grade improvement rates of baseline ASIA AIS grade A patients in both a historical series (the SYGEN ATSCI study, 1992–1997) and a more recent study (STASCIS ²¹ study, 2012). For all ASIA AIS baseline A patients, the SYGEN study had 70 of 461 (15.18%) and the STASCIS study had 11 of 71 (15.5%) improving 2 AIS grades (p = 1.000, Fisher’s Exact Test, two-tailed). Comparing the groups in each study with the best outcomes, 37 of 182 (20.3%) of the direct admission patients in the SYGEN study and 8 of 44 (18.2%) of the early surgery group (presumably also direct admission) in the STASCIS study improved two AIS grades (p = 0.836, Fisher’s Exact Test, two-tailed). Although caution is needed in comparing any two non-randomized studies, this analysis does not support enhancement in neurological outcome with changes in medical practice in the 20 years between the SYGEN and STASCIS studies.

Understanding the historical context of the SYGEN ATSCI study helps explain its almost identical outcome findings to the STASCIS ²¹ study. The SYGEN ATSCI study was initiated at the beginning of the era when both early surgery and initial correction of ATSCI hypotension were considered as potentially beneficial. The study therefore collected variables related to timing of surgery and initial hemodynamic parameters in its dataset. In addition, these concepts were repeatedly discussed at multiple investigators’ meetings before the start of the patient entry phase and during the multiyear patient entry phase of the study. The NASCIS II,^9,10 an ATSCI study that entered patients a few years prior to SYGEN, predated the era of these opinions; in all three treatment groups, a total of only 20 of 223 (8.97%) baseline ASIA AIS As improved in long-term follow-up. The fractional improvement in the NASCIS II ATSCI study was statistically inferior compared with all patients in the SYGEN study (p = 0.0293, Fisher’s Exact Test, two-tailed).

Considering the two-grade improvement fractions of the baseline ASIA AIS As in these three ATSCI studies together leads to the conclusion that the increase in fraction of improvements occurred in medical care of ATSCI patients between the time of the NASCIS II and the SYGEN ATSCI studies (1988–1992). Although caution is needed in comparing any two non-randomized studies, this analysis does not support enhancement in neurological outcome with changes in medical practice in the 20 years between the SYGEN and STASCIS studies.

Rehabilitation after ATSCI

Rehabilitation treatments are also believed to affect the duration and extent of recovery, and details will require future studies. The current statistical exploration evaluates only the acute management and neurological follow-up to 12 months and does not analyze modern rehabilitation measurement scales, as they did not exist during the data collection period. The patients all received best rehabilitation therapy available at the time of study. The SYGEN versus placebo group assignment remained double blinded until the last patient had completed long-term follow-up to minimize bias from all medical treatments, including rehabilitation therapies.

This study did not collect data to analyze rehabilitation duration regarding goals of functional independence or enhancement of quality of life. The SYGEN effect presented in this paper is a shortening of the time to reach the MR milestone. Figure 1 diagrammed two effects: (1) faster recovery and (2) larger fraction of recovery. Attaining MR at an earlier stage appears to be an initial process that positively impacts patients by reducing rehabilitation time and achieving functional objectives sooner.

Conclusion

The fact that three model fit variables yielded statistical significance in the model fit of the time to MR but were negative in the model fit of the long-term fraction of MR proves that the new time to MR outcome is useful in assessing new aspects of the recovery process after ATSCI. A potential action for model fit variables having faster recovery of MR but no greater fractional recovery long-term would be a pharmaceutical enhancement of the speed of the body’s inherent neurological healing process. Thus, we recommend the time to MR should be included in future ATSCI studies. We recommend expanding the number of neurological assessments in recovery to 1, 2, 4, 8, and 16 weeks to more completely assess the recovery phase of ATSCI and to collect 26 and 52-week data for long-term evaluations to allow assessing both the time to MR and the fraction of long-term MR.

While direct admission to a specialized SCI center is generally believed to lead to better outcomes, this conclusion has not been previously proven in a randomized prospective ATSCI study. ³⁷ Prior retrospective and registry data have showed improved outcomes for patients with ATSCI admitted to designated trauma centers compared with non-trauma centers. ³⁸ Consequently, the 2021 national guideline recommends triaging patients to the “highest-level trauma center available within the regional system.” ³⁹ Our study documented that interhospital transfers delay care and yield worse long-term outcomes. The direct admission subgroup of baseline AIS grade A patients had a 26.83% two-AIS grade long-term MR improvement rate in this study, which is double the MR rate of both transferred patients in this study and the rate reported by other ATSCI clinical series. We also recommend direct admission to SCI facilities because of the higher rate of recovery in the ASIA AIS A patients. Specifically, the results of this study indicate that when ATSCI is diagnosed at emergency medical technician contact in the field, these patients should be preferentially taken to SCI centers and not the closest community hospitals.

The statistically significant difference between SYGEN and Placebo for time to MR of 21.19 days (p = 0.0024) is the first reported positive result in any FDA IND study showing improvement in outcome after ATSCI. Achieving MR 3 weeks sooner with SYGEN was beneficial to the patient by itself and also enabled cost saving by decreasing the rehabilitation time by about 3 weeks. The shortened recovery time to MR observed during this trial was the first positive drug effect in any FDA IND ATSCI trial.

We also consider achieving MR approximately 17 days sooner because of Early Surgery to be beneficial to the patient, as the rehabilitation time would be decreased about 2½ weeks. When the SYGEN group with early surgery is compared with the patients who received Placebo and did not get early surgery, the difference is 40.87 days (98.69 reduced to 57.82 days, p = 0.0041), showing that these two effects are statistically additive (Fig. 7C). This relationship strengthens the premise that time to MR is a useful outcome to examine in ATSCI datasets.

Significant effects were observed for the prespecified secondary end-point; however, because no multiplicity-adjustment procedure was prospectively specified, these findings should be considered supportive or hypothesis-generating rather than confirmatory. We recommend that future ATSCI studies: (1) use block randomization to minimize baseline imbalances and employ statistical analyses that account for remaining imbalances in baseline AIS grades, (2) collect data on direct admission versus transfer from community hospital, (3) do multiple neurological examinations between the initial exam and 6 month exam to allow assessment of the speed of recovery, and (4) perform prompt surgery in the first few days after the ATSCI.

Journal of Neurotrauma transparency, rigor and reproducibility statement for clinical studies papers statement

The SYGEN ATSCI study was a US FDA IND study (https://www.fda.gov) that entered 760 patients from 28 Neurotrauma Centers in North America between Q2 1992 and Q1 1997 with registration and analytic plan prespecified as required by the US FDA IND process. The SYGEN ATSCI study was a block randomized, double-blinded clinical trial design to test SYGEN versus placebo drug. The current paper introduces a new outcome variable, defined as the time from ATSCI to marked recovery (a two-grade improvement from baseline) to expand on the preplanned secondary outcome assessment of the time to recovery. This new variable quantifies the recovery time, allowing statistical analysis of group recovery time differences. For the SYGEN ATSCI study, a simple analysis of time to recovery statistically favoring SYGEN over Placebo was included in the final report submitted to the US FDA as well as in the initial publications in 2001.^6–8 Details of the screening and study admission process are detailed in these initial publications. In the current study, the data analysis was performed using JMP®, 19.1.2 software (SAS Institute Inc., Cary, NC, USA). Several tests and patient groupings demonstrated consistency of the statistically significant variables, minimizing the possibility that the conclusions were by chance. We are unaware of any planned studies to duplicate the results presented in this paper. The raw data of the SYGEN ATSCI study have been shared with multiple other SCI researchers in the past directly by Fred H. Geisler, M.D., Ph.D., principal investigator, (fgeisler@gmail.com). The current paper will be published under a Creative Commons Open Access license, and upon publication will be freely available at https://www.liebertpub.com/loi/neu.

Authors’ Contributions

F.H.G.: Conceptualization, methodology, formal analysis, writing—original draft, and visualization. L.J.W.: Writing—review and editing and visualization. J.R.F.W.: Writing—review and editing and visualization.

Data Availability Statement

Raw data will be shared on request to authors.

Informed Consent/Patient Consent

Not relevant; no patient interactions were involved in this new statistical presentation of old data.

Abbreviations Used

AIS

ASIA Injury Score

ANOVA

analysis of variance

ASIA

American Spinal Injury Association

ATSCI

acute traumatic spinal cord injury

BMSC

bone marrow derived stem cells

BP

blood pressure

CDF

cumulative distribution function

CSF

cerebrospinal fluid

EANS

European Association of Neurosurgical Societies

EAST

Eastern Association for the Surgery of Trauma

EPO

erythropoietin

ER

emergency room

FDA

Food and Drug Administration

FU

follow-up

G-CSF

granulocyte-colony stimulating factor

GM

ganglioside

HBO

hyperbaric oxygen

IND

investigational new drug

JMP

statistical software by SAS Institute Inc.

LE

lower extremity

MAP

mean arterial pressure

MBC

modified Benzel Classification

MBP

mean blood pressure

MP

methylprednisolone

MPSS

methylprednisolone sodium succinate

MR

Marked Recovery

NASCIS2

Second National Acute Spinal Cord Injury Study

NIH

National Institutes of Health

RISCIS

Riluzole in Acute Spinal Cord Injury Study

SAS

SAS Institute Inc.

SCI

spinal cord injury

STASCIS

Surgical Timing in Acute Spinal Cord Injury Study

SYGEN

monosialoganglioside sodium (study drug and trial name)

WSES

World Society of Emergency Surgery

Cite this article as: Geisler FH, Weisbrod LJ, Wilson JR F. Injury to time of marked recovery as a new acute spinal cord injury outcome measurement for statistical analysis of interventions and drug treatments, Neurotrauma Reports, 2026, 7, 2689288X261469045, doi: 10.1177/2689288X261469045.