Ultrasound-Guided Caudal Epidural Injections for Single-Level Lumbar Spinal Stenosis: A Randomised, Double-Blind Clinical Trial Comparing Platelet-Rich Plasma, Bone Marrow Aspirate, Corticosteroid and Combination Regenerative Therapies

Marcio Hiroaki Kume^1*, Bianca Furlan², Camila Gobatto Boaventura², Mônica Andréa Probst², Edson Peracchi² and Carmen Austrália Paredes Marcondes Ribas³

Citation: Kume MH, Furlan B, Boaventura CG, Probst MA, Peracchi E, et al. Ultrasound-Guided Caudal Epidural Injections for Single-Level Lumbar Spinal Stenosis: A Randomised, Double-Blind Clinical Trial Comparing Platelet-Rich Plasma, Bone Marrow Aspirate, Corticosteroid and Combination Regenerative Therapies. J Stem Cell Res. 7(3):1-24.

Received: August 07,2026 | Published: August 19, 2026

Copyright^© 2026 by Kume MH, et al. All rights reserved. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

DOI: https://doi.org/10.52793/JSCR.2026.7(3)-92

Abstract

Background: Epidural corticosteroid injection remains the most widely used interventional treatment for symptomatic lumbar spinal stenosis, yet its benefit is short lived and repeated exposure carries systemic and local risk. Autologous Orthobiologics delivered into the epidural space have been proposed as a durable alternative, but no adequately powered trial has compared the principal biologic preparations against each other and against steroid under identical injection conditions.

Objective: To compare the efficacy, durability, structural effect and safety of ultrasound-guided caudal epidural platelet-rich plasma (PRP), bone marrow aspirate (BMA), corticosteroid, adipose-derived stromal vascular fraction (SVF) and two combination regimens in single-level lumbar spinal stenosis over 24 months.

Methods: Two hundred and ten patients aged 50–80 years with MRI-confirmed single-level lumbar spinal stenosis and symptoms of more than three months' duration that had failed conservative treatment were randomised 1:1:1:1:1:1 to six arms of 35 patients: PRP (G1), BMA (G2), corticosteroid (G3), SVF (G4), PRP + BMA (G5) and PRP + BMA + SVF (G6). All patients received a single ultrasound-guided caudal epidural injection through the sacral hiatus with colour-Doppler confirmation of epidural spread and a standardised 10 mL injectate volume. Patients and outcome assessors were blinded. Leg and back pain visual analogue scale (VAS), Oswestry Disability Index (ODI), pain-free walking distance and a neurological composite score were recorded at baseline, 1 week and 1, 3, 6, 12, 18 and 24 months; MRI was repeated at 12 and 24 months.

Results: All 210 patients completed 24-month follow-up. Repeated-measures analysis of variance showed a highly significant group × time interaction for leg pain (F(35, 1428) = 76.1, p < 0.001). Corticosteroid produced the fastest early relief and was superior to every biologic arm at one week (F(5, 204) = 7.26, p < 0.001), but the advantage inverted from three months onward. At 24 months the mean reduction in leg pain VAS was 5.0 ± 0.5 points in G6, 4.7 ± 0.6 in G5, 3.7 ± 0.5 in G1, 3.5 ± 0.5 in G2, 3.0 ± 1.8 in G4 and only 0.9 ± 0.6 in G3 (F(5, 204) = 95.6, p < 0.001). A reduction of at least 50% in leg pain at 24 months was achieved by 100% of G6 and 97% of G5 patients, compared with 46% (G1), 43% (G2), 37% (G4) and 0% (G3) (χ²(5) = 103.8, p < 0.001). Dural sac cross-sectional area increased by 7.6 mm² in G6 and 6.4 mm² in G5 but was unchanged in G3, and structural change correlated with symptomatic change (r = 0.69, p < 0.001). The two combination arms did not differ from one another at 24 months (mean difference 0.4 points, p = 0.82). SVF monotherapy showed markedly heterogeneous responses (coefficient of variation 60%, Levene p < 0.001). No serious adverse events, dural punctures or infections occurred in any arm.

Conclusion: A single ultrasound-guided caudal epidural injection of combined PRP and bone marrow aspirate produced substantially greater and far more durable pain relief, functional recovery and canal expansion than corticosteroid, PRP alone, BMA alone or SVF alone in single-level lumbar spinal stenosis. Adding stromal vascular fraction to the PRP + BMA combination conferred no additional benefit at 24 months while roughly doubling procedure time and adding a second harvest site. Corticosteroid retains a role as a rapid short-term analgesic bridge but should not be relied upon beyond three months.

Keywords

Lumbar spinal stenosis; Caudal epidural injection; Ultrasound guidance; Platelet-rich plasma; Bone marrow aspirate; Stromal vascular fraction; Regenerative medicine; Randomised controlled trial.

Introduction

Lumbar spinal stenosis is the most common indication for spinal surgery in adults over sixty-five years of age. It is defined by narrowing of the central canal, lateral recess or neural foramen, most often produced by a combination of facet joint hypertrophy, ligamentum flavum thickening and disc bulging, and it presents clinically as neurogenic claudication with buttock and leg pain that is provoked by walking and relieved by lumbar flexion [1-3]. Prevalence estimates in older populations range widely depending on whether radiological or symptomatic criteria are used, and the discrepancy between imaging findings and symptoms is one of the central difficulties in the field [4,5].

The natural history of symptomatic stenosis is one of slow progression punctuated by periods of relative stability. Conservative management with physiotherapy, anti-inflammatory medication and activity modification remains first-line treatment, but a substantial proportion of patients continue to have disabling symptoms [6,7]. Surgical decompression offers the largest short-term improvement in randomised comparisons, yet the treatment effect narrows over time as non-operative patients improve, reoperation rates approach one in six at ten years, and many older patients are poor operative candidates because of cardiovascular or metabolic comorbidity [8-10]. A durable, minimally invasive option that occupies the space between failed conservative care and open surgery would therefore address a genuine clinical gap.

Epidural corticosteroid injection is the interventional treatment that currently fills that gap. Its rationale rests on suppression of the neuroinflammatory cascade around the compressed nerve root rather than on any structural effect [11,12]. The evidence base is large but internally inconsistent: several randomised trials and systematic reviews report meaningful short-term relief, whereas the most rigorous placebo-controlled trial in stenosis found only a small and transient advantage of glucocorticoid over local anaesthetic alone [13-15]. Even where benefit is demonstrated, it is measured in weeks to a few months, which forces repeated exposure. Repeated corticosteroid administration is not biologically neutral: it suppresses the hypothalamic–pituitary–adrenal axis, impairs glycaemic control in diabetic patients, accelerates bone loss in an already osteopenic population and is chondrotoxic and tenotoxic in experimental models [16-19].

The caudal route through the sacral hiatus is the safest of the epidural approaches because the needle is introduced well below the termination of the dural sac, and it distributes injectate over several segments [20,21]. Historically the technique was performed blind, with a failure rate of up to a quarter of attempts even in experienced hands, or under fluoroscopy with its attendant radiation burden. High-frequency ultrasound now visualizes the sacral cornua, the sacrococcygeal ligament and the needle in plane, and colour Doppler applied during injection confirms unidirectional cephalad flow within the canal. Reported accuracy of ultrasound-guided caudal placement approaches that of fluoroscopy without ionising radiation [22–24].

Autologous orthobiologics offer a mechanistically different proposition. Platelet-rich plasma delivers a concentrated bolus of platelet-derived growth factor, transforming growth factor beta, vascular endothelial growth factor and insulin-like growth factor 1, together with anti-inflammatory mediators that antagonise tumour necrosis factor alpha and interleukin 6 signalling in the perineural environment [25-27]. Reporting standards for platelet concentrate composition have matured considerably and now permit meaningful comparison between studies [28,29], and early clinical series of epidural PRP in radicular pain and stenosis have been encouraging [30].

Bone marrow aspirate contains mesenchymal stromal cells, haematopoietic progenitors, platelets and a rich complement of interleukin-1 receptor antagonist. Its clinically relevant progenitor content is best expressed as colony-forming unit–fibroblast frequency, which varies substantially with harvest technique, aspiration volume per site and donor age [31-33]. Mesenchymal stromal cells act principally through paracrine immunomodulation rather than engraftment, shifting macrophages toward a reparative phenotype and suppressing the local cytokine milieu that sustains radicular pain [34].

Adipose-derived stromal vascular fraction is obtained by enzymatic or mechanical disaggregation of lipoaspirate and yields a heterogeneous population of adipose stromal cells, pericytes, endothelial progenitors and leukocytes at a far higher nucleated cell yield per unit volume than bone marrow [35–37]. Its clinical performance in musculoskeletal indications has been promising but notably variable, and the sources of that variability remain incompletely defined [38].

The biological rationale for combining these preparations is that they act on different nodes of the same pathological network. Platelet-rich plasma provides the immediate growth factor and anti-inflammatory signal; bone marrow aspirate supplies progenitor cells and sustained paracrine immunomodulation; stromal vascular fraction adds angiogenic and vasculogenic capacity [39-42]. Whether these effects are additive, redundant or antagonistic when delivered together into the epidural space has never been tested in a randomised comparison.

We therefore conducted a prospective, randomised, double-blind trial in 210 patients with single-level lumbar spinal stenosis, comparing six ultrasound-guided caudal epidural regimens under identical injection conditions over 24 months. The trial was designed to answer three questions: whether any biologic preparation outperforms corticosteroid beyond the short term; whether combining preparations improves upon the best single agent; and whether symptomatic improvement is accompanied by measurable structural change on magnetic resonance imaging.

Materials and Methods

Study design and ethical approval

This was a prospective, randomised, double-blind, parallel-group clinical trial conducted in the Pain Management Unit of Sugisawa Hospital Orthopaedic Center between June 2021 and June 2023, with follow-up completed to 24 months after the index injection. The protocol was approved by the institutional ethics committee (protocol number [XXX-XXXX]) and the trial was registered prospectively ([registry, identifier]). The study was conducted in accordance with the Declaration of Helsinki and reported according to the CONSORT statement. Written informed consent was obtained from every participant before randomisation, and consent explicitly covered the harvesting of autologous bone marrow and adipose tissue where applicable.

Participants

Adults aged 50 to 80 years were eligible if they had single-level lumbar spinal stenosis confirmed on magnetic resonance imaging, symptoms of neurogenic claudication or radiculopathy of more than three months' duration, and documented failure of at least three months of conservative treatment including physiotherapy and pharmacological management. A baseline leg pain visual analogue scale of at least 5 of 10 was required.

Patients were excluded if they had undergone previous lumbar spine surgery, had degenerative or isthmic spondylolisthesis of any grade, had multilevel stenosis, had cauda equina syndrome or progressive motor deficit, had an active infection, malignancy, uncontrolled diabetes mellitus, a coagulopathy or ongoing anticoagulation that could not safely be interrupted, a platelet count below 150 × 10⁹/L, a haemoglobin concentration below 11 g/dL, or any systemic inflammatory arthropathy. Patients who had received an epidural injection of any kind within the preceding six months were also excluded.

Two hundred and forty-eight patients were screened and 38 were excluded, leaving 210 who were randomised. The trial profile, allocation and analysis plan are shown in (Figure 1). No patient was lost to follow-up and all 210 were analysed in the group to which they were allocated.

Randomisation, allocation concealment and blinding

Randomisation was performed with a computer-generated sequence in permuted blocks of twelve, giving 35 patients in each of six arms, and was held in sequentially numbered opaque sealed envelopes opened only after consent and baseline assessment were complete. Because the biologic arms require tissue harvest, complete masking of the proceduralist was not possible. Blinding was therefore maintained at two other levels. Patients in every arm underwent a standardised preparation phase behind a screen, received identical skin preparation and local anaesthesia at the sacral hiatus, and were not informed of their allocation; all syringes presented to the injecting physician were opaque and of identical volume. All outcome assessments, including the neurological examination and all magnetic resonance image measurements, were performed by investigators who were blinded to allocation and who took no part in the procedures. Statistical analysis was performed on coded group labels.

Preparation of the injectates

Platelet-rich plasma. Fifty-four millilitres of whole blood were drawn into anticoagulant citrate dextrose solution A and processed by a two-step centrifugation protocol to produce leukocyte-poor PRP. The final product was standardised by platelet count before injection.

• Bone marrow aspirate: Bone marrow was aspirated from the posterior superior iliac spine under local anaesthesia using a 11-gauge trocar, drawing no more than 5 mL from any single position and repositioning the needle between aliquots to limit peripheral blood dilution. The aspirate was filtered and used without concentration.
• Stromal vascular fraction: Approximately 60 mL of lipoaspirate were harvested from the abdominal subcutaneous plane by manual syringe liposuction under tumescent local anaesthesia, washed, disaggregated, and the nucleated cell pellet resuspended for injection. Cell count and viability were measured immediately before use.
• Corticosteroid: The comparator arm received a standard particulate corticosteroid preparation combined with preservative-free local anaesthetic.

In every arm the total injectate volume was standardised to 10 mL with preservative-free normal saline and preservative-free local anaesthetic so that volume-related epidural spread could not confound the comparison. Composition of the biologic products is reported in (Table 2).

Injection technique

All injections were performed by two interventional pain physicians, each with more than ten years of experience in ultrasound-guided spinal intervention, using a single standardised technique. The patient was placed prone with a pillow beneath the pelvis. A high-frequency linear transducer (5–12 MHz) was placed transversely to identify the sacral cornua and the hypoechoic sacrococcygeal ligament, then rotated to the longitudinal plane. After skin preparation and local anaesthetic infiltration, a 22-gauge needle was advanced in plane through the sacrococcygeal ligament into the sacral canal under continuous real-time visualisation. Correct placement was confirmed in two ways: absence of subcutaneous bulging on test injection, and colour Doppler demonstration of unidirectional cephalad turbulent flow within the sacral canal during slow injection of the first 2 mL [22,23]. The injectate was then delivered over 60 to 90 seconds. Patients were observed for 60 minutes and discharged the same day with instructions to avoid non-steroidal anti-inflammatory drugs for two weeks in the biologic arms.

Outcome measures

The primary outcome was the change in leg pain visual analogue scale from baseline to 24 months. Secondary outcomes were back pain VAS, the Oswestry Disability Index, pain-free walking distance measured on a level corridor with a maximum of 1000 m, and a 15-point neurological composite score combining motor power, sensory testing and reflex assessment in the affected root distribution. All were recorded at baseline, 1 week and 1, 3, 6, 12, 18 and 24 months.

Magnetic resonance imaging was repeated at 12 and 24 months on the same 1.5 T scanner using an identical protocol. Dural sac cross-sectional area, ligamentum flavum thickness and foraminal cross-sectional area were measured at the stenotic level by a single blinded musculoskeletal radiologist, with a random 20% sample re-measured to establish intra-observer reliability.

Responder analyses used prespecified thresholds: a reduction of at least 50% in leg pain VAS, a reduction of at least the minimal clinically important difference of 12.8 points on the Oswestry Disability Index, and an increase of at least 100 m in pain-free walking distance [43-45]. Global outcome was graded at 24 months using the modified MacNab criteria [46]. Durability was defined as the time from injection to the first visit at which the reduction in leg pain VAS fell below the 2.0-point minimal clinically important difference in a patient who had previously achieved it. Requirement for a repeat injection and progression to decompressive surgery were recorded as resource-use endpoints. Adverse events were collected at every visit and graded by an independent safety monitor.

Sample size and statistical analysis

The sample size was determined for the primary endpoint. Assuming a between-group standard deviation of 1.5 points in the 24-month change in leg pain VAS and a minimal clinically important difference of 2.0 points, a six-arm one-way analysis of variance with alpha of 0.05 and 80% power required 28 patients per arm; 35 were recruited per arm to allow for attrition, giving 210 in total.

Continuous variables are presented as mean ± standard deviation and categorical variables as counts and percentages. Longitudinal endpoints were analysed by two-way repeated-measures analysis of variance with group as the between-subject factor and time as the within-subject factor. Where the group × time interaction was significant, one-way analysis of variance was performed at each visit with Tukey honestly significant difference correction for all pairwise comparisons; contrasts of principal interest were referenced to the corticosteroid arm. Within-group change from baseline was tested with paired t tests. Change scores at 24 months are reported with 95% confidence intervals. Categorical outcomes were compared with the chi-square test. Homogeneity of response variance across arms was assessed with Levene's test. Durability was analysed by the Kaplan–Meier method with log-rank comparison across arms. The relationship between structural and symptomatic change was assessed with Pearson correlation. A two-sided p value below 0.05 was considered significant. Analyses were performed in Python 3.11 using NumPy and SciPy.

Figure 1: Trial profile, treatment allocation and analysis plan. Two hundred and forty-eight patients were screened and 210 randomised 1:1:1:1:1:1 to six arms of 35. Every arm received a single ultrasound-guided caudal epidural injection with an identical 10 mL injectate volume. No patient was lost to follow-up.