ParticipantsEligibility criteria for participants

The study population consisted of male and female adult individuals, aged between 18 and 80 years of age, with traumatic spinal cord injury due to complete or incomplete C4-T12 for more than three months. The participants also included those who had been diagnosed with neuropathic pain, with a pain score ≥4.0 according to the Visual Analogue Scale (VAS features a 10 cm line where individuals mark their level of pain, ranging from 0.0, indicating no pain, to 10.0 indicating maximum pain. Results were recorded in centimetres to one decimal place corresponding to a 100-point scale), during the last week before the randomisation date, and those who had been stably treated at a stable dose for at least the last month with pregabalin in the range of 150–300 mg/day, which was maintained at the same dose until the end of the study. Participants could not be treated with opiates or cannabinoids, but those who had been treated with stable doses of other neuroactive drugs (antidepressants, anticonvulsants, antispastic, and similar medicines), at a stable dose, at least during the last month, could also be recruited (this medication was kept stable). SCI individuals were excluded from the trial if they had been treated with opiates and/or cannabinoids or if they had a history of alcohol or drug abuse within 6 months before the screen, and participants who had psychiatric disorders or moderate or severe cognitive impairment were also excluded. In addition, SCI individuals were excluded if they had severe arterial hypo or hypertension (a blood pressure outside of 90–160 for systolic pressure and 50–115 for diastolic pressure), if they were pregnant or lactating, showed evidence of significant liver or kidney disease, or if they had clinically significant diseases recorded in their medical history or detected at the time of the physical examination and/or clinically significant laboratory analyses (haematology, biochemistry, and urinalysis) or ECG.

Data were collected at seven different hospitals in Spain: 1) National Hospital for Paraplegics of Toledo; 2) Vall d’Hebron University Hospital, Barcelona; 3) A Coruña University Hospital Complex; 4) Virgen de las Nieves University Hospital, Granada; 5) Institut Guttmann, Barcelona; 6) Hospital Los Madroños, Brunete, Madrid; and 7) Virgen del Rocío University Hospital, Sevilla. All centres had extensive experience in SCI management.

Trial design

The design of the clinical trial is described in Fig. 1 and is a phase IIA randomised, double-blind, placebo-controlled, parallel-group, and multicentric trial, carried out between 2019 and 2022, involving individuals with neuropathic pain due to spinal cord injury (EudraCT number: 2018-004792-13/ Clinicaltrials.gov: NCT04148573). The allocation ratio was 1:1:1:1 to a placebo, NFX88 1.05 g, 2.1 g, 4.2 g/day. The treatment duration was 12 weeks, and the whole duration of the study was ~17 weeks divided as follows: a screening period (one week) including one visit (visit at screening-VS), a treatment period (12 weeks) including 4 visits (V1, V2, V3, V4/End of Treatment, EoT) and a follow-up period (4 weeks) that ended on the fifth visit (V5/End of Study, EoS).

Fig. 1: Clinical trial design.

NFX88 treatment started after Visit 1 (V1). V2, V3, and V4 (EoT, end of treatment) were carried out after 1, 2, and 3 months of treatment, respectively. During the last month, participants did not receive NFX88, and a final follow-up visit (V5, EoS, end of study) was carried out to determine the effect of NFX88 withdrawal.

This study was approved by the Institutional Review Board (Regional Committee of Ethics for Research with Medicines, CEIm, of the Community of Madrid: ref EC 06/19) and by the Spanish Drug Agency (AEMPS). All participants provided written informed consent before participating in the trial. This study was carried out in compliance with applicable regulations and guidelines and following the principles of the Declaration of Helsinki and the International Conference on Harmonization Good Clinical Practice guidelines. The CT was fully monitored by a professional Contract Research Organization (CRO) and all data presented here were reported after the final report had been sealed.

The study consisted of a 1-week baseline screening period, followed by the participants being randomly assigned one of the doses of NFX88 or the placebo. Each subject was administered four tablets of the study drug or the placebo three times a day over 90 days (from V1 to V4, see Fig. 1). Participants with 1.05 g/day doses received 1 tablet of 350 mg + 3 placebo tablets three times per day, 2.1 g/day group received 2 tablets of 350 mg + 2 placebo tablets, three times per day, participants allocated at 4.2 g/day group received 4 tablets of 350 mg, three times per day and the placebo group received 4 placebo tablets, 3 times per day. Visits occurred on the first day of treatment (V1), 30 days later (V2), 60 days later (V3), and 90 days later (V4/EoT). A follow-up visit occurred 4 weeks after treatment completion (V5/EoS) and only the participants who had received at least 75% of the planned treatment were included in subsequent analyses.

Outcomes (pre-specified primary and secondary outcome measures)

Safety and efficacy variables were collected during the study. The study chronogram is shown in Fig. 2. The outcome to analyse the safety and tolerability of NFX88 in SCI individuals was based on the incidence, severity, and causality of any adverse events (AEs); changes in vital sign parameters and the physical examination, changes in laboratory values, changes in ECGs, and changes detected through Modified Ashworth Scale (MAS) (e.g. to monitor spasticity worsening) and the International Standards for Neurological Classification of Spinal Cord Injury (ISNCSCI) (e.g. to monitor neurological worsening) scores.

Fig. 2: Chronogram.

D: days of treatment with respect to the treatment initiation (D1, day 1). For other details, see text.

The following information was recorded as efficacy variables: Visual Analogue Scale (VAS), PainDETECT Questionnaire (PD-Q) [26, 27], and Patient Global Impression of Changes (PGIC).

Blinding

Identification of placebo versus NFX88 treatment was blind. SCI individuals were randomly assigned to receive the two drugs in a double-blind model such that neither the investigator nor the participant knew which combination was being administered. Blinding of either medication was ensured as the packaging of both NFX88 and the placebo were identical. Personnel involved in conducting the study did not have access to the randomisation code before the blind trail was officially broken. The study participants were notified through the informed consent form that they would be receiving a blind treatment; that is, they would not be informed about which drug (Investigational or Comparator) was being administered. The treatments assigned to each participant were kept blind to the study team until database lock. In the event that an individual subject had a medical emergency or became pregnant, where knowledge about the treatment being administered was critical to the person’s health, the blind for that subject was broken by the researcher. Unblinding was only granted if knowledge of the study treatment was essential for the appropriate clinical management or welfare of the participants. No unblinding occurred in this study.

Blinding the analytical process

The statistician was not blinded to participant allocation. To avoid bias, the analysis was carried out in several phases, where true allocation was only added during the final phase. These phases were:

Phase 1—data pre-processing and cleaning.

Phase 2—the programming of models, tables, and graphs (the statistician wrote the code for all statistical models, tables, and graphs using the data-2 dataset).

Phase 3—running the code on unblinded data (data managers provided the true allocation variable to the statistician who incorporated it into the clean data-2 dataset, which generated the data-3 dataset. The code produced in Phase 2 was run in data-3 and a report was automatically produced with the unblinded results of the analysis).

Phase 4—an unplanned exploratory efficacy analysis was performed on all unblinded data.

Randomisation

All eligible participants fulfilling all inclusion criteria and no one of the exclusion criteria were randomly assigned to the NFX88 or placebo treatment arms in a 1:1:1:1 ratio. Due to the small number of SCI individual recruited in each group, a unique, centralized with no stratification (centres or participants), and randomised list was generated. The randomisation was done by creating blocks of multiples of 4 (so 4, 8, 12….). The size of the blocks was kept secret by the data manager until the end of the study to minimize the chance of a person involved in the study guessing the allocation of the next participant. All SCI individuals participating in this study were not allowed to be randomised in this study again.

Statistical methodsSample size calculation

As this is a Phase IIA CT, the sample size was determined to demonstrate safety. The power of this trial lies in the probability of detecting cases of AEs caused by the drug used in the treatment arms. This power depends on the true (unobservable) risks of those AEs (the higher the risk the higher the chance of detecting cases). The sample size was determined to guarantee a power of 80% for detecting any AE that has an underlying risk of 5% of happening in the clinical trial participants during the treatment (Table 1). Eleven subjects per arm, from a total of 44 participants, was estimated to be enough to evaluate safety. Based on this calculation, groupings of 15 participants per arm were planned. In Table 1, we show the sample size calculation for guaranteeing a power of 80% to detect any AE that has an underlying risk of 5% of happening in the participants during the treatment.

Table 1 Sample size estimation.Statistical methods used to compare groups for primary and secondary outcomes Safety assessment (main objective)

The population used to test safety included all participants who were randomised, including those who left the study for different reasons. For this analysis, each AE was coded as a binary variable (Present/Absent) for each participant. AEs defined as “unsafe levels” of laboratory parameters or clinical outcomes (vital signs, ECGs, ISNCSCI, and MAS scales) were also coded as binary. The clinical researcher was the person to assess whether the AEs were related to the treatment or not. Several binary variables were calculated if the same AE occurred to varying degrees of severity. Tables with counts and proportions of each AE in each arm were compiled. For the intervention arms, exact confidence intervals for the proportion of each AE were estimated. A comparison of the risk of each AE between arms was done using Fisher’s Exact test. To increase power, the participants from the three intervention doses were analysed together in one intervention arm. If some AEs were found to be relatively common, a logistic regression model was built to examine if there was a dose-response effect on the probability of such an adverse event happening. A variable with the count of repetitions was created for cases where it was considered appropriate to compare AEs occurring in an individual more than once. For the intervention arms, confidence intervals for the rates were calculated and the rates between arms were compared using the Poisson regression model. Drop-out participants, those leaving the study before the follow-up visit and replaced by other individuals, were also included in the analysis of safety. In this analysis, there is a potential for bias if the total time the participant is in the trial (and the total amount of drug taken) varies considerably in each trial arm (because of differences in the number of participants and the time at which they abandoned the trial). In situations like these, a sensitivity analysis was done considering individual time within the study by using rates of AEs over person-time in the study in Poisson models.

Efficacy assessment (secondary objective)

The Phase IIA CT presented here was specifically focused on determining safety profiles through the use of an oral tablet formulation and on testing preliminary efficacy at different dosage levels. A group of participants was administered the drug at different doses to determine the recommended dose for phase 2B and 3 clinical studies. In this trial, the power analysis was focused on determining safety, therefore, assessing efficacy should be considered exploratory. The efficacy analysis included all participants who were randomised, had at least 75% of treatment compliance and had completed the appropriate VAS, PD-Q, or PGIC questionnaires. The variable VAS used to analyse efficacy was coded as a continuous non-parametric variable. PD-Q was coded as being continuous (from 0 to 38) but also as a categorical variable with three possible categories (10). The three possible categories were low, intermediate/uncertain, and high probability of neuropathic component of the pain. PGIC was analysed as a categorical variable recoded (due to the small sample size) in three categories: improved (a score between 1 and 3 values), no change (a score of 4), and worse (score between 5 and 7).

The preplanned/predeclared first exploratory efficacy endpoint was the change in VAS (principal efficacy variable) from baseline to V4/EoT. The difference between V1 and V4/EoT, in the different groups, was analysed using the Kruskal–Wallis test. Due to the exploratory nature of the Phase IIA CT, we included additional analyses to the statistical plan after analysing the results.

Further analyses of the VAS were performed, using a Friedman test (Conover’s Test was included for post hoc analysis), on the four groups (doses) that were separately analysed to evaluate the time course of the effects of NFX88. After a visual inspection of the data, three-time points were included in the Friedman test (V1, V3, and V4/EoT). Due to the exploratory nature of the analysis, the same data obtained for the four groups were also compared directly using the Wilcoxon Test (V1 vs. V3, V1 vs. V4/EoT). Our a priori hypothesis was that NFX88 would reduce neuropathic pain, so when we compared V1 vs. V3 and V1 vs. V4/EoT, we reported the p-value of the post hoc analysis considering the hypothesis of V1 different from V3 and V4/EoT (two tails) and V1 higher than V3 and V4/EoT (one tail). For all the analyses, the exact p-value is indicated to provide a reference for the robustness of these exploratory analyses.

To further explore the efficacy at different doses, the number of responders at V4/EoT using different definitions was reported. Due to the exploratory nature of our efficacy analyses, we defined the responders in four different ways: 1) participants with a ≥ 30% reduction in pain (based on VAS); 2) participants with a ≥ 50% reduction in pain (based on VAS); 3) participants with a ≥ 1.5-point reduction in pain according to the VAS; and 4) participants with a ≥ 2-point reduction in pain according to the VAS. These categorizations (≥30%, ≥50%, ≥1.5-point, and 2-point reduction in pain according to the VAS) were arbitrarily decided to allow data presentation in individuals with more or less intense pain relief. Moreover, we descriptively report the number of drop-out participants which can be considered a measure of treatment failure (being determined by AEs, treatment discomfort, and lack of efficacy) and the PGIC for all groups. These variables (responders vs. non-responders, drop-outs vs. participants that completed the treatment, and PGIC) were separately analysed using Chi-squared analysis.

We explored the use of PD-Q (secondary efficacy variable) to test NFX88 analgesic effects in two different ways: 1) the score (assuming that the higher the score the more intense is the neuropathic pain); 2) the change in the neuropathic component of pain. The differences found in the PD-Q scores between V1 and V4/EoT were analysed using a Kruskal-Wallis test. This is an exploratory valuation as PD-Q is not the appropriate scale to test pain severity. The PD-Q was also analysed as a categorical variable with the following cut-offs: a score of ≤ 12 indicates that pain is unlikely to have a neuropathic component, a score of ≥ 19 suggests that pain is likely to have a neuropathic component, and a score between these values (13–18) indicates that the result is uncertain. The main idea with this analysis is that NFX88 is theoretically useful only for neuropathic pain. So, if the number of individuals with pain that is likely to have a neuropathic component decrease, we can indirectly consider that the drug is effective on reducing at least this component of pain.

Stratification based on the initial PD-Q classification

The PD-Q quantifies the probability that the pain described by the participants is neuropathic in origin. According to its mechanism of action, NFX88 should be more efficient in treating this type of participants. Therefore, we explored its efficacy in a subgroup of participants where the probability of experiencing pain with a neuropathic component was greater than 90% (PD-Q ≥ 19). In this subgroup, the difference between the VAS scores obtained at the end-of-treatment visit (V4) and the first visit (V1) for the four treatment arms (placebo, NFX88 1.05 g/day; NFX88 2.1 g/day NFX88 4.2 g/day), the time course of the four treatment arms, the quantification of responders and the PGIC were determined.