Bone remodeling, not inflammation, as the predominant pathology in modic type 1 lesions of the lumbar spine

Abstract

Purpose

Lesions of the vertebral endplates (Modic changes) are associated with low back pain (LBP), and different pathological processes, including inflammation, have been proposed as causative. This study aimed to explore the relationship between [¹⁸F]FDG uptake (inflammation) and [¹⁸F]NaF uptake (bone remodeling) and patient-reported back pain as well as conventional MRI detected endplate changes.

Methods

Participants were selected from an IRB-approved study, creating 2 arms: [¹⁸F]NaF (n = 11) and [¹⁸F]FDG (n = 11). Scans were performed on a 3.0 T PET-MRI scanner. An MSK radiologist rated endplates changes using Modic classification and a relative scale: None; only edema; edema > fat; edema=fat; edema< fat; only fat. Radiotracer uptake (SUVmax) in endplate-adjacent bone was measured using ROIs. VAS scores (0–10) for back pain were recorded before imaging. Generalized estimating equations assessed associations between pain and radiotracer uptake as outcomes and radiologic findings as predictors, adjusted for age, sex, and body mass index (BMI).

Results

The study included 220 endplates from 22 patients (110 per arm). Overall, pain was positively associated with edematous lesions (beta 0.21, P = .024) but not with fatty changes (beta 0.10, P = .567). In the [¹⁸F]NaF arm, tracer uptake was associated with edematous lesions (beta 1.79, P = .001) and with pain (beta 0.18, P = .007). In the [¹⁸F]FDG arm, tracer uptake was negatively associated with edematous lesions (beta −0.11, P = .013) and showed no significant association with pain (beta −1.50, P = .064).

Conclusion

Our study shows that [¹⁸F]NaF-PET-MRI detected bone-remodeling is more closely linked to painful vertebral end-plate degeneration than inflammatory [¹⁸F]FDG signal.

Degenerative endplate lesions (Modic changes, MC), are a known structural correlate of low back pain, but the pathobiology underpinning these changes remains incompletely understood. Our study quantified the uptake of [18F]FDG, a marker of inflammation, and [18F]NaF, a marker of bone remodeling, on PET-MRI in a prospectively recruited cohort of 22 patients. We found a significant association of [18F]NaF, but not [18F]FDG with presence of edema within lesions as well as with pain.

Introduction

According to the 2019 National Health Interview Survey (NHIS), published by the Centers for Disease Control and Prevention (CDC) in 2021, the back emerged as the most frequently reported site of pain among adults; prevalence increased with age, peaking at 45.6% among individuals aged 65 and older.¹ In 2016, low back pain (LBP), together with neck pain, incurred an estimated $134.5 billion in healthcare cost. Apart from economic concerns, LBP is also considered a major contributor to the rise in opioid use disorders.^2,3

Clinically, LBP is typically categorized as either axial or radicular, each with different imaging pathologies that are more or less closely associated with presence of clinical symptoms⁴. While radicular pain often results from nerve compression caused by intervertebral disk herniation or spinal canal or foraminal stenosis⁵, axial LBP commonly arises from structures such as the intervertebral disks, facet joints, or sacroiliac joints and may refer to adjacent areas like the hips or thighs.⁶

Among the imaging lesions with the most widely accepted association with axial backpain are structural changes of the vertebral endplates, commonly referred to as Modic changes.^7–9 The original Modic classification recognizes 3 types of changes, classified based on their appearance on MRI—type 1, characterized by increased fluid signal; type 2, characterized by increased fat signal; and type 3, representing sclerosis with low signal across all sequences. Type 1 changes, which may alternatively be described as endplate adjacent bone marrow edema, have been shown to be associated with ingrowth of blood vessels and fibrous tissue in these regions, inspiring the term fibro-vascular change.¹⁰ The strongest association with back pain has been established for edematous, or type 1, changes.¹¹

A promising technological tool to further the understanding of biological processes within these lesions are modern hybrid positron emission tomography—magnetic resonance imaging (PET-MRI) systems,¹² which allow the simultaneous detailed visualization of soft tissue, discs, and bone, with specific tracers that offer insight into local cell activity. Fluorodeoxyglucose ([¹⁸F]FDG) PET, for example, is a widely used marker of glucose metabolism and is highly sensitive to areas of acute cellular response, indicating inflammation.^13,14 Sodium Fluoride ([¹⁸F]NaF) PET imaging is a method for quantifying bone remodeling through osteoblast activity,¹⁵ which has gained traction in musculoskeletal imaging in recent years.¹⁶

Several previous studies have attempted to use advanced metabolic imaging to further the understanding of biological endplate changes. Russo et al. showed increased activity on bone SPECT/CT in Modic changes but did not differentiate between different types of Modic changes¹⁷ and Dimitrou et al. demonstrated focal increases in [¹⁸F]NaF uptake on PET-MRI in Modic type 1 lesions, but did not specifically investigate the relationship between bone remodeling and pain in the presence of these lesions.¹⁸

The aim of this study was to investigate the association between clinical imaging characteristics and both [¹⁸F]NaF and [¹⁸F]FDG PET signal of vertebral endplate lesions, and to further investigate the association between pain and these imaging findings. Our hypothesis at the outset of this study was that edematous lesions show high [¹⁸F]FDG signal, while fatty and sclerotic lesions and endplates with erosions show high [¹⁸F]NaF signal.

Materials and methods

Participants

All participants in this analysis were prospectively recruited for an observational study focused on the development of advanced imaging technologies for low back pain (LBP) at our institution, which is conducted as part of the BACPAC consortium (19). Prior to study commencement, the institutional review board (IRB) approved of the study (IRB 19–29744), and all participants gave written, informed consent prior to study inclusion.

BACPAC).¹⁹ In short, participants with low back pain (duration of at least 3 months; pain quality axial, radicular [including radiculopathy], or mixed) were prospectively recruited via treating physicians. Exclusion criteria were age <18 years, inability to provide written informed consent, pregnancy or current breast-feeding, recent (<12 months) history of fracture, tumor or operation of the spine, rheumatic disease (eg, axial spondyloarthritis), active malignancy, contraindication for MR imaging; furthermore, for patients scheduled to receive PET-imaging, contraindication for tracer injection, and specifically for [¹⁸F]FDG-application, known diabetes mellitus. All participants received comprehensive spinal MRI and if clinically indicated, image guided pain therapy (nerve root block or facet joint infiltration), after acquisition of PET-imaging.

Excluded from this ancillary analysis were all participants who did not receive PET imaging—a schematic of subject selection is given as Figure 1. Benign osseous lesions such typical hemangiomas and Schmorl’s nodes did not constitute exclusion criteria; no instances of atypical hemangiomas or lesions suspect of malignancy were observed. As shown there, included participants were divided into 2 study arms, depending on primary pain profile, which determined the administered radiotracer: Those with primary radiating LBP received [¹⁸F]FDG-PET MRI, whereas those with primary axial back pain received [¹⁸F]NaF-PET MRI. Each participant was required to complete a self-administered pain assessment using the Visual Analog Scale (VAS) at baseline. The VAS asks individuals to rate their current pain levels in various areas of the body on a scale from 0 to 10, where 0 represents no pain, 5 indicates moderate pain, and 10 signifies extreme pain.

Figure 1.

Participant selection and basic demographics. Abbreviations: BMI = Body mass index; IQR = interquartile range; LBP = Low back pain; PET = Positron emission tomography; VAS = Visual analogue scale. ^aActivated facet joint arthropathy/endplate disease was defined as presence of either contrast enhancement or edema around the joint or disc space.

Image acquisition

All patients underwent imaging using a GE 3.0 T whole-body PET/MRI (GE Healthcare, Waukesha, WI) scanner after providing informed consent for study participation. The clinical spinal MRI protocol included sagittal T1-weighted 2D fast spin-echo (FSE) and sagittal T2-weighted FSE with and without fat suppression, axial T1-weighted FSE, and axial T2-wieghted FSE, (for MR parameters, see Table 1).

Table 1.

MR Sequence parameters.

	Sagittal T2-weighted FSE with fat saturation	Sagittal T2-weighted FSE without fat saturation	Sagittal T1-weighted FSE	Axial T2-weighted FSE	Axial T1-weighted FSE
TE	60 ms	60 ms	9.2 ms	60 ms	9.4 ms
TR	7660 ms	8216 ms	583 ms	3992 ms	739 ms
Echo train length	18	20	6	18	6
Field of View	26 × 26 cm	26 × 26 cm	26 × 26 cm	18 × 18 cm	18 × 18 cm
Acquisition Matrix	448 × 320	448 × 320	416 × 224	340 × 224	340 × 224
Slice thickness	3 mm	3 mm	3 mm	4 mm	4 mm
Slice gap	0 mm	0 mm	0 mm	1 mm	1 mm
Scan Time	3 min 20 sec	3 min 10 sec	2 min 40 sec	4 min	3 min

Abbreviations: FSE fast spin echo; TE time to echo; TR time to repetition.

All sequences were acquired a bandwidth of ±62.5 kHz and without additional averages.

The administered dose of [¹⁸F]NaF for PET-MRI was weight-adjusted, averaging at 2.98 (±0.09) mCi. Participants in the [¹⁸F]FDG group received an intravenous injection dose of 4.92 (±0.18) mCi depending on their weight. This was followed by a 45-minute seated rest for the radiotracer uptake before they proceeded to the scanner for the imaging experiment. The standardized uptake value (SUV) images were derived from the source static PET images reconstructed with the ordered subset expectation maximation (OSEM²⁰) algorithm that the vendor provided. The PET signal acquisition was performed for the first 20 minutes of the [¹⁸F]FDG PET-MRI scan and for the first 25 minutes of the [¹⁸F]FDG PET-MRI scan. The signals acquired during the whole acquisition window were used to generate a single static PET image set.

Image analysis

Semi-quantitative MRI assessment

A board-certified radiologist (V.K., 7 years of experience) evaluated all endplates of the disc spaces L1/2 to L5/S1 in all participants, independently of each other, and blinded to PET imaging as well as clinical endplate lesions. For evaluating endplate lesions the well-established Modic classification system was used, Modic type 1 for lesions containing increased fluid signal, Modic type 2 for fatty lesions and Modic type 3 for sclerotic lesions, exhibiting low signal across all MR sequences. To account for the mixed appearance of these lesions, an additional assessment of these lesions was formulated, that aimed at the relative assessment of endplate adjacent bone marrow edema: 0 for no lesion, 1 for predominantly fluid, 2 for more fluid than fat, 3 for an equal proportion of fat and fluid, 4 for more fat than fluid, and 5 for predominantly fat. Additionally, endplate erosions were assessed on a scale from 0 to 3, where 0 indicates no erosion, 1 represents mild erosion, 2 moderate erosion, and 3 severe erosions. A pictorial guide to image assessments is given in Figure 2. As the focus of this analysis was the exploration of the association of metabolic characteristics of endplate changes, we refrained from including detailed data on other potential pain generators, such as disc pathologies or facet joint arthropathy.

Figure 2.

MRI assessment of endplate lesions. White boxes in (a) indicate area that is shown as magnification in panel (b–d). a = sagittal T2 FSE. b = sagittal T2 FSE, magnified. c = sagittal T1 FSE, magnified. d = sagittal T2 FSE with fat saturation, magnified. 1: Participant with Modic 2 lesion, classified as “more fat than edema” (= 4), because of slight edematous signal in the superior endplate of L4 (white arrow heads); also, presence of subtle erosions ( = 1) in the same location (black arrowheads). 2: Participant with Modic type 3 lesion, classified as “more edema than fat” ( = 2), because of marked edematous rim (white arrowheads) around the lesion; also, presence of marked erosions (=3) in the superior endplate of L4 (black arrowheads).

Quantitative PET assessment

Maximum radiotracer uptake (SUVmax) in the disc-adjacent vertebral endplate was evaluated for each endplate of every patient by a board-certified radiologist (V.K.). Tracer uptake was measured spanning from the lower endplate of L1 to the upper endplate of S1, using elliptical ROIs applied to transverse plane images. A visualization of measurements is given as Figure 3.

Figure 3.

Quantitative PET measurements. Row 1: Oblique-axial Na[¹⁸F]F-T2w fusion image (1a), manually angulated parallel to the superior endplate of L3 (dotted line), with elliptical ROI. Sagittal fusion image (1b) and sagittal T2w image (1c) are given for reference. Row 2: oblique-axial [¹⁸F]FDG-T2w fusion image (2a), manually angulated parallel to the superior endplate of L3 (dotted line), with elliptical ROI. Sagittal fusion image (2b) and sagittal T2w image (2c) are given for reference.

Inter- and intrareader agreement

To calculate inter-reader agreement, all semi-quantitative and quantitative endplate assessments were also performed by a second board-certified radiologist (K.Z., 9 years of experience), although only the first readers assessments were used for analysis. For intra-reader agreement, a random sample of 60 endplates (30 from each group) were reassessed by the primary reader (V.K.) after a wash-out period of 4 weeks.

Statistical analysis

All analyses were performed using Python Version 3.11.8. As units of observation, individual endplates were chosen. To account for lack of independence between multiple measurements within one subject, generalized estimating equations (GEE) were employed to examine the relationships between radiotracer uptake, radiological findings, and reported pain levels. To enhance the robustness of the analysis, adjustments were made for potential confounding variables, including patient age, sex, and body mass index (BMI). As this constitutes an exploratory ancillary analysis of the described study cohort, we did not perform a formal sample size estimation, but rather a post-hoc power-analysis. This was performed separately for each group using the observed effect sizes, standard deviations, and cluster structure (11 subjects with 10 repeated measurements each). Based on the estimated β coefficients for the association between tracer uptake and pain from the GEE models and assuming an exchangeable working correlation structure, the NaF group showed >80% power to detect the observed association, whereas the FDG group was underpowered (<60%) to detect the estimated effect size at α = 0.05.

Inter- and intra-reader agreement for semi-quantitative scoring was assessed using (linear) weighted Cohen’s kappa, while agreement for quantitative measurements was assessed with intra-class correlation coefficients (ICC; 2-way mixed effects model [absolute agreement]). Only results with 2-sided P values < .05 were considered statistically significant.

Results

Study subjects

A total of 22 participants, 11 in each study arm, were included in this ancillary analysis. Basic demographic characteristics for each group are given in Figure 1. As this figure shows, both groups contained more men than women, but the distribution was similar in both study arms (P > .999). There was no significant difference in mean age (P = .403), BMI (P = .614), pain intensity at baseline (p-0.188) or duration of pain (P = .199) between both groups. Using a conversion coefficient of 0.019 mSv/MBq²¹ for [¹⁸F]FDG scans, a mean radiation exposure of 3.46 (±0.12) mSv was calculated. For Na[¹⁸F]F the calculated exposure was 2.64 (±0.08) mSv, using a conversion coefficient of 0.024 mSv/MBq.²²

Descriptive imaging findings

Semi-quantitative MRI findings

A total of 220 endplates from 22 patients were included in the study, with each study arm consisting of 110 endplates from 11 patients. Table 2 details the frequency of each lesion type per endplate for the whole cohort. Modic lesions, regardless of type, were found in 30.9% (34/100) and 25.5% (28/110) of endplates in the [¹⁸F]NaF and [¹⁸F]FDG group, respectively (P = .454); Type 1 lesions were more common in the [¹⁸F]FDG group (13.6%; 15/110) than in the [¹⁸F]NaF group (4.5%; 5/110). The mixed lesion scoring method showed lesions with more edema than fat in 13.6% (15/110) of the endplates in the [¹⁸F]NaF group and 11.8% (13/110) of endplates in the [¹⁸F]FDG group (P = .840). Endplate erosions were found in similar proportions in both groups, with 22.7% (25/110) in the [¹⁸F]NaF group and 19.1% (21/110) in the [¹⁸F]FDG group (P = .619). Inter-reader agreement was good to excellent with a weighted Cohen’s kappa of 0.66 (95% CI 0.42–0.84), 0.89 (95% CI 0.54–0.90), and 0.84 (95% CI 0.59–0.95) for Modic classification, relative bone marrow endplate changes and erosions, respectively.

Table 2.

Descriptive results per endplate.

Endplate	Modic changes [%, n]			Mixed endplate lesion scoring [%, n]			Erosions [%, n]	SUVMAX (mean, SD)
	Type 1	Type 2	Type 3	Edema>fat	Edema=fat	Edema<fat		NaF	FDG
L1 inf	9.1% (2/22)	4.5% (1/22)	0.0% (0/22)	9.1% (2/22)	4.5% (1/22)	4.5% (1/22)	9.1% (2/22)	12.41 (3.60)	2.90 (0.81)
L2 sup	13.6% (3/22)	13.6% (3/22)	0.0% (0/22)	22.7% (5/22)	0.0% (0/22)	13.6% (3/22)	9.1% (2/22)	12.87 (3.83)	2.86 (0.84)
L2 inf	4.5% (1/22)	13.6% (3/22)	0.0% (0/22)	9.1% (2/22)	4.5% (1/22)	9.1% (2/22)	9.1% (2/22)	11.76 (3.92)	2.83 (0.73)
L3 sup	13.6% (3/22)	18.2% (4/22)	0.0% (0/22)	13.6% (3/22)	9.1% (2/22)	18.2% (4/22)	9.1% (2/22)	12.14 (3.39)	2.69 (0.84)
L3 inf	0.0% (0/22)	18.2% (4/22)	4.5% (1/22)	9.1% (2/22)	4.5% (1/22)	13.6% (3/22)	13.6% (3/22)	13.92 (3.61)	2.64 (0.62)
L4 sup	0.0% (0/22)	22.7% (5/22)	4.5% (1/22)	13.6% (3/22)	4.5% (1/22)	18.2% (4/22)	9.1% (2/22)	11.31 (2.94)	2.67 (0.74)
L4 inf	4.5% (1/22)	27.3% (6/22)	0.0% (0/22)	0.0% (0/22)	9.1% (2/22)	22.7% (5/22)	13.6% (3/22)	12.11 (4.05)	2.62 (0.69)
L5 sup	13.6% (3/22)	31.8% (7/22)	0.0% (0/22)	9.1% (2/22)	9.1% (2/22)	31.8% (7/22)	13.6% (3/22)	10.34 (3.81)	2.94 (0.93)
L5 inf	18.2% (4/22)	18.2% (4/22)	0.0% (0/22)	18.2% (4/22)	0.0% (0/22)	22.7% (5/22)	13.6% (3/22)	10.78 (3.16)	2.86 (0.71)
S1 sup	13.6% (3/22)	13.6% (3/22)	0.0% (0/22)	18.2% (5/22)	0.0% (0/22)	13.6% (3/22)	9.1% (2/22)	10.69 (3.13)	2.92 (1.25)

Abbreviations: Inf, inferior; Sup, superior.

Clinical MRI results for entire cohort, PET results separate for each study arm.

Quantitative PET findings

Overall mean tracer uptake (SUV_MAX) for [¹⁸F]NaF was 11.83 (±3.57) across all endplates, and 11.70 (±2.61) in the subset of endplates without any MR imaging abnormality. For [¹⁸F]FDG, mean SUV_MAX across all endplates was 2.22 (±0.52) and 2.26 (±0.54) when excluding all endplates with any MR imaging abnormality. Mean SUV_MAX per endplate is given in Table 2—there was no conspicuous trend regarding uptake per level.

Inter- and intrareader agreement

Inter-reader agreement for endplate bone marrow edema, Modic lesions and endplate erosions was excellent with Cohen’s kappa of 0.89 (95% CI 0.78, 0.91), 0.80 (95% CI 0.69, 0.85), and 0.88 (95% CI 0.75, 0.89), respectively. For quantitative assessments, agreement was excellent with an ICC of 0.98 (95% CI 0.98, 0.99). Intra-reader agreement for semi-quantitative assessments was similarly high: 0.92 (95% CI 0.65, 0.91) for endplate bone marrow edema, 0.78 (95% CI 0.58, 0.90) for Modic lesions and 0.86 (95% CI 0.64, 0.93) for endplate erosions. Intra-reader agreement for quantitative assessments was 0.97 (95% CI 0.95, 0.98).

Association between endplate lesions and tracer uptake

Unadjusted mean SUVmax for [¹⁸F]NaF was 11.79 (±2.69) in endplates without endplate lesion, 20.77 (±1.92) in endplates with Modic type 1 lesions, 9.82 (±2.99) in endplates with Modic type 2 lesions, and 18.53 (±0.66) in Modic type 3 lesions. In our GEE analysis, tracer uptake showed a significant positive association with bone marrow edema (beta 1.79, 95% CI 0.75–2.84, P = .001), while no significant relationship was observed with endplate fat lesions (beta −0.46, 95% CI −0.94 to 0.02, P = .059) or endplate erosions (beta 0.58, 95% CI −0.53 to 1.70, P = .059). See Figure 4 for clinical imaging examples, demonstrating the increased uptake in edema-rich lesions.

Figure 4.

Clinical imaging examples. Row 1: Participant with radiating LBP and edematous endplate lesions in the L5/S1 disc space (black arrows), identified by high signal on fat-saturated T2 (1a) and low signal on T1 (1b) weighted images. Note the lack of signal increase on [¹⁸F]FDG-T1w fusion image (1c). Row 2: Participant with axial LBP and edematous endplate lesions in the L2/3 disc space (white arrows), also identified by high signal on fat-saturated T2 (2a) and low signal on T1 (2b) weighted images. Note the marked signal increase on Na[¹⁸F]F-T1w fusion image (2c).

Conversely, in the [¹⁸F]FDG study arm, tracer uptake was negatively associated with bone marrow edema (beta −0.11, 95% CI −0.19 to −0.02, P = .013), fat lesions (beta −0.21, 95% CI −0.32 to −0.09, P < 0.001) and erosions (beta −0.51, 95% CI -0.58–0.10, P = .005). Unadjusted mean SUVmax for [¹⁸F]FDG was 2.89 (±0.86) in endplates without endplate lesion, 2.73 (±0.55) in endplates with Modic type 1 lesions and 2.25 (±0.40) in endplates with Modic type 2 lesions; no Modic type 3 lesions were observed in this study arm. Refer also to Figure 4 for clinical imaging examples of edematous lesions with marked lack of [¹⁸F]FDG uptake.

Association between imaging findings and pain

In the overall cohort, pain demonstrated a positive association with relative edema content of endplate lesions (beta 0.21, 95% CI 0.03–0.38, P = .024), whereas no significant relationship was identified between pain and relative fat content of endplate lesions (beta 0.10, 95% CI −0.23 to 0.42, P = .567). Endplate erosions were not associated with pain (beta 0.08, 95% CI −0.34 to 0.51, P = .706). Additionally, [¹⁸F]NaF uptake was positively correlated with pain (beta 0.18, 95% CI 0.05–0.31, P = .007), independent of the presence of bone marrow edema. However, no significant association was found between [¹⁸F]FDG uptake and pain (beta −1.50, 95% CI −3.09 to 0.09, P = .064), independent of bone marrow edema.

Discussion

Our study is the first to demonstrate the association between MRI lesions of the vertebral endplates, pain, and simultaneously acquired biological information from PET, in a prospective cohort of LBP patients. We found a significant association between edematous lesions and bone remodeling, quantified by Na[¹⁸F]F, but not with [¹⁸F]FDG, thus challenging the notion that these imaging lesions represent florid inflammation. We furthermore showed a positive association between bone remodeling and pain but could not detect such an association for inflammation indicated by [¹⁸F]FDG uptake.

Pain that is attributed to the vertebral endplates, that is, vertebrogenic pain, is understood to derive from defects in the cartilage endplate,²³ which then trigger a tissue response in the adjacent bone that involves the ingrowth of nerve fibers, local attraction of TNF-immunoreactive cells²⁴ and bone remodeling.¹⁰ On imaging this form of endplate degeneration is associated with signal changes on MRI, which were first described by Modic et al.:⁷ generally, Modic Type 1 changes exhibit MR signal properties that can be described by the non-specific umbrella term “bone marrow edema,” which just describes an increase of water signal relative to normal bone marrow, with the assumption of a pathological change. This is characterized by a concomitant increase in T2 signal and decrease in T1 signal. These changes can be seen in various pathologies, including fractures, avascular necrosis, inflammation (both auto-immune and septic) and tumor.^25,26 Modic type 1 changes are rare in asymptomatic individuals, found in approximately 3%,¹¹ and our findings further corroborate this with a significant positive association between relative edema content of endplate lesions and pain. For Modic type 2 changes, which are characterized by an increased fat signal, the evidence linking them to pain is less unequivocal.^11,27 In our study, increased relative fat content of endplate lesions was not associated with pain.

Inflammation in bone marrow changes, particularly in Modic Type 1, is supported by findings of elevated inflammatory markers like interleukin 1 (IL-1)²⁸ and interleukin 6 (IL-6), along with increased C-reactive protein (CRP) levels.²⁹ These markers point to active inflammatory activity and potential autoimmune interactions with nucleus pulposus cells, leading to localized tissue changes.³⁰ These data informed our original hypothesis, that FDG-uptake, which is increased in the presence of acute inflammation,^31,32 should be elevated in active, for example, edematous endplate lesions, but our results refuted this emphatically, showing a negative association between degree of edema on imaging and FDG uptake. Our findings are in line with a previous investigation by Albert et al, who also did not find increased uptake on [¹⁸F]FDG-PET-CT in individuals with Modic changes on MRI.³³ Even though our findings do not refute the presence of inflammation in these lesions, they do show that the edema shown on MR imaging should not be equated with inflammation, as it is commonly done in rheumatic diseases, such as axial spondyloarthritis.³⁴ More likely, these edematous areas contain a heterogeneous mixture of pathological processes, of which inflammation is only one, which is further supported by the limited success of biologic drugs in the management of LBP.³⁵

The other biological process in the endplates that has been described in histological studies and was the second focus of our investigation, is bone turnover.³⁰ On imaging, bony changes are visualized mainly in Modic Type 3 lesions, where the bone becomes sclerotic and stabilized, with increased volume and thickness, signifying a more quiescent remodeling stage.³⁶ Histological studies showed a decreased osteoid surface to bone surface ratio in Modic Type 2 lesions compared to type 1 and 3,³⁶ which is in line with our finding, that relative fat content of endplate lesions was not associated with Na[¹⁸F]F uptake.

Overall, the combined multiparametric PET-MRI signatures of edematous endplate lesions share some similarities with bone marrow edema like lesions (BMELLs) observed in subchondral bone in osteoarthritis—lesions have high T2 signal on conventional MR sequences, are associated with pain [ref], show increased Na[¹⁸F]F-uptake on PET³⁷ but no elevated uptake on [¹⁸F]FDG PET.³⁸ Even though this analogy can only ever be partial, as the cartilage endplate of the vertebral body differs from the hyaline cartilage of the peripheral joints, considering these parallels may still be instructive. The most pertinent common denominator in both lesions may be that at least in a subset of cases, the edema observed on imaging is best understood as a bone bruise²⁶ in response to diminished force dissipating capacities of shock absorbing soft tissues (cartilage, menisci, intervertebral discs) because of degeneration.

In our analysis we deliberately performed two different scorings of endplate bone marrow lesions in parallel: The traditional Modic scoring and a specifically developed scoring method that grades fat and edema content relative to each other. This approach was chosen because of the limited inter-reader agreement for endplate lesions, which has been cited as low as kappa 0.52,³⁹ which may in part be caused by the difficulty to classify lesions that have characteristics of both Modic 1 and Modic 2 lesions. Our proposed scoring methods showed a higher inter reader agreement than the Modic scoring (kappa 0.89 vs 0.66) but fails to account for sclerosis of the endplate (Modic 3 lesions) and does not assess the size of the lesion. These findings highlight the necessity of a more quantitative approach in future research of these kinds of lesions.⁴⁰

While our study employed standard clinical MR sequences, including T1-weighted spin-echo and fat saturated T2-weighted spin-echo sequences, further insights into pathological endplate changes may be gleaned by applying more advanced and quantitative imaging techniques in these patients. Ultrashort-echo-time (UTE) MRI permits direct visualization of the thin cartilage endplate and shows promise for quantifying early micro-structural damage.⁴¹ Furthermore, compositional disc biomarkers such as T1rho,⁴² when integrated with molecular PET tracers may improve our understanding of early endplate degeneration, before unequivocal Modic type changes become apparent.

Our study has methodical limitations that need to be addressed. Participants in the study arms had different pain phenotypes (axial vs radiating LBP), this it is not clear whether similar imaging findings carried similar significance in both groups—a design with application of both tracers in the same participant would be of high scientific value but might face ethical hurdles because of from radiation exposure. Furthermore, the cross-sectional nature of our data prevents causal inference or assessment of lesion evolution over time. While our manual SUV_max measurements yielded excellent inter‐reader agreement, future work may benefit from automated segmentation tools or alternative metrics such as SUV_peak to further minimize operator‐dependent variability. Furthermore, only a limited number of study subjects could be examined with this sophisticated and resource-intensive imaging technique, which likely contributed to the non-significant results observed in the FDG study arm, which was underpowered (<60%), as laid out in the methods section. Furthermore, the limited sample size prohibited more elaborate subgroup analyses, such as an exploration of sex-specific differences, which is warranted in light of the know differences in both immunosenescence and bone biology between men and women. Lastly, our study design attributes reported pain only to structural bone changes, while it is well established that other pain generators, such as myofascial trigger points may contribute to patient reported pain.

Our findings indicate that, within the limits of this small prospective study, vertebral endplates affected by Modic 1 like changes exhibit increased Na[^18F]F uptake, whereas [^18F]FDG uptake is low or absent. These results support the hypothesis that bone remodeling activity is a key component of painful Modic 1 lesions, but causal confirmation will require longitudinal imaging and, ideally, histopathological correlation. Future studies with sufficient sample size, longitudinal design, randomized tracer allocation, and improved quantification of tracer uptake utilizing fully automized segmentation and exploring methods of normalization of uptake are needed to further develop our understanding of pain generation in these lesions.

Author contributions

The authors have made substantial contributions to the following sections: Conception and design (V.K., K.Z., C.C., S.M.). Analysis and interpretation of the data (V.K., K.Z., C.C., M.H., E.B., D.Y., T.M.L., S.M.). Collection and assembly of data (V.K., M.H., E.B., K.Z.). Drafting of the article (V.K., S.M.). Statistical expertise (V.K., K.Z., S.M.). Critical revision of the article for important intellectual content (V.K., K.Z., C.C., M.H., E.B., D.Y., T.M.L., S.M.). Final approval of the article (V.K., K.Z., C.C., M.H., E.B., D.Y., T.M.L., S.M.). Virginie Kreutzinger takes responsibility for the integrity of the work as a whole from inception to the finished article.

Funding

This project is supported by the National Institutes of Health (NIH) grant UH3AR076724.

Conflicts of interest: The authors declare no conflict of interest.