Intervertebral disc degeneration associated with vertebral marrow fat, assessed using quantitative magnetic resonance imaging

This cross-sectional cohort study was approved by the Ethics Committee of our hospital. All subjects provided written informed consent before examination. A total of 104 male and female volunteers of different age groups (20–29 years, 30–39 years, 40–49 years, 50–59 years, 60–70 years) with chronic lower-back pain between August 2018 and June 2019 were recruited in this study. Among them, 47 were male (mean age 44.6 years) and 56 were female (mean age 44.3 years). The inclusion criteria included the following: (1) male or female volunteers with chronic lower-back pain (lasted for more than 3 months), which was based on Alleva et al.’s study [24]; (2) underwent routine MRI lumbar spine examination, IDEAL-IQ, and T2 mapping imaging with satisfactory image quality. Exclusion criteria included the following: (1) medical history of diseases, such as a tumor, spondylolisthesis, lumbar scoliosis, lumbar trauma or fracture, surgery, spinal infectious diseases, hemopathy, and metabolic bone disease; (2) incomplete imaging data or poor image quality.

MRI was performed using a 3T system (GE Discovery MR750 3.0T, America) with a dedicated spine coil. The lumbar spine scanning sequence included routine T1WI and T2WI sequence, IDEAL-IQ, and T2 mapping sequence. The following imaging parameters were used: (1) Sagittal T1WI (TR/TE = 809/11 ms), T2WI (TR/TE = 3093/120 ms), transversal T2WI (TR/TE = 3500/120 ms), slice thickness 3 mm, intersection gap 0.4, FOV 320 × 320 mm, matrix 132 × 132, the number of excitations (NEX) 2; (2) Sagittal IDEAL-IQ:TR 10.5 ms, minimum TE 1.2 ms, maximum TE 6.5 ms, FOV 300.0 mm × 300.0 mm, matrix 132 × 132, layer thickness 5.0 mm, acquisition time 168 s; (3) transversal T2 mapping TR 718 ms, TE1-TE8 time 9.7 ms, 19.4 ms, 29.1 ms, 38.7 ms, 48.4 ms, 58.1 ms, 67.8 ms, and 77.5 ms, layer thickness 3 mm, layer spacing 1 mm, NEX 1.

All routine MR images were used to identify pre-existing abnormalities of the lumbar vertebrae, and T2WI was used to assess the Pfirrmann classification [25]; the specific classification standards are listed in Table 1; we excluded 23 Pfirrmann V IVDs. All vertebral bone marrow fat fractions (FF) were measured along with the remaining 497 Pfirrmann I–IV IVDs’ T2 values on the AW4.6 workstation. All MRI images were analyzed by two experienced radiologists in clinical diagnostic and spine MRI research, MLH, and JYY. We loaded FatFrac images in the REFOMAT system and measured the FF value of the vertebral bone marrow (VFF). The L1-S1 vertebral body was mapped manually at the three central layers, to avoid the endplate and basivertebral foramen. To avoid bias, the average value of the three measurements was considered (Fig. 1). We defined the fat fraction of adjacent superior vertebral bodies as SFF and the fat fraction of adjacent inferior vertebral bodies as IFF. The Functool post-processing was also performed on the T2 mapping images. We indicated the round region of interesting (ROI) within the anterior annulus fibrosus (AAF) to measure the T2 value and copied the ROI onto the nucleus pulposus (NP) and posterior annulus fibrosus (PAF) to measure the T2 values, respectively (Fig. 1). The T2 values of the NP, AAF, and PAF of L1/2-L5/S1 discs were all measured. The final Pfirrmann results were obtained by consensus of the two experts, and the average value of the two expert measurements was taken for the quantitative result.

Agreement between the two experts’ Pfirrmann gradings was evaluated using Cohen’s kappa (κ) statistic. The strength classification based on kappa values was as follows: excellent (κ = 0.8–1.0); substantial (κ = 0.6–0.80); moderate (κ = 0.4–0.6); fair (κ = 0.2–0.4); and slight (κ = 0.0–0.2). A Bland–Altman plot was used to analyze the quantitative measurements between the two experts, by plotting the mean of the two measurements against their difference with a 95% limits of agreement (LOA) (= mean difference ± 1.96 × SD of the difference). Bivariate correlations of the T2 value, Pfirrmann I–V in IVD, and FF in the adjacent vertebra were measured using the Pearson product-moment two-tailed correlation coefficient analysis. The T2 values and the adjacent SFF/IFF value differences within the five Pfirrmann groups, five age groups, and five lumbar levels were examined using the Kruskal–Wallis test [26]. Analyses, where the overall test was significant, and pairwise comparisons were carried out using the Mann–Whitney U test while controlling for type I error using the Bonferroni adjustment. Data were represented as boxplots. All statistical analyses were performed using the SPSS version 23 (IBM, Chicago, IL, USA). p < 0.05 was considered statistically significant.

In total, 520 IVDs were analyzed from the 104 volunteers. The mean body mass index (BMI) was 23.2 ± 3.2 kg/m² (ranged 16.2–33.2), and there were 196 (38.7%), 140 (26.9%), 95 (18.3%), 66 (12.7%), and 23 (4.4%) IVDs in each of the Pfirrmann classification from grade I to V, respectively (Table 2). The remaining 497 IVDs of Pfirrmann I–IV were quantitatively analyzed.

Pfirrmann classification of κ = 0.83 demonstrated an excellent agreement, while Bland–Altman analysis of all paired measurements revealed a mean difference (± SD, 95% limits of agreement) of − 0.3 ms (± 2.3 ms, − 4.8 to 4.1 ms) for NP; − 0.4 ms (± 0.9 ms, − 0.5 to − 0.3 ms) for AAF; − 0.6 ms (± 0.8 ms, − 0.7 to − 0.5 ms) for PAF; − 0.4 ms (± 0.8 ms, − 0.5 to − 0.3 mmHg) for SFF; and − 0.4 ms (± 0.9 ms, − 0.5 to − 0.3 mmHg) for IFF as shown in Fig. 2.

For all 497 discs, the T2 value of the NP/AAF/PAF significantly correlated with not only the SFF but also the IFF (p < 0.001, Fig. 3). The NP T2 was negatively correlated with adjacent VFF (p < 0.001, Fig. 4). The Pfirrmann classification I–V correlated positively with the adjacent SFF and IFF (p < 0.01) (Fig. 5).

There was no statistical difference between SFF and IFF adjacent to all the IVDs (p > 0.05). However, only the T2 values of NP significantly inversely correlated with the adjacent VFF in Pfirrmann classification degrees II–III but not in grades IV–V (Fig. 3). When comparing II–III and IV parameters, significant differences were observed in the NP T2 values and SFF, but none was found in the AAF T2, PAF T2, adjacent IFF, ages, gender, and the BMI parameters (Tables 3 and 4). NP T2 values decreased gradually from Pfirrmann I to IV grades, while FF values increased in Pfirrmann I–III and decreased in Pfirrmann IV (Fig. 6).

Quantitative MRI characteristics were assessed for age groups: 20–29 years, 30–39 years, 40–49 years, 50–59 years, and 60–70 years (Fig. 7). T2 values of NP were weakly correlated with the adjacent VFF values in the age group 40–49 years (r = 0.246). The VFF value increased with age while the NP T2 value decreased after the age of 30. After the age of 50, VFF increases more rapidly. The NP T2 value and VFF of adjacent vertebrae bone were significantly different between the age groups 30–39 and 40–49, and the age groups 50–59 and 60–70 (p < 0.001) (Table 5).

We observed a relationship not only between the NP T2 value and adjacent VFF but also between the Pfirrmann grading and adjacent VFF in L1/2-L5/S1 levels, which was statistically significant except for the L5/S1 NP T2 value and the S1 VFF. However, the AAF and PAF did not significantly correlate to VFF (Figs. 8 and 9). T2 value of NP, AAF, and PAF and its adjacent SFF were significantly difference from those of L1/2 to L5/S1 (Table 6). Among the 104 volunteers, the number of IVDD (Pfirrmann II–V IVD) occurring alone in one segment was 8 and contained the following: L1/2 (n = 0), L2/3 (n = 0), L4/5 (n = 4), L3/4 (n = 2), and L5/S1 (n = 2).