Association of Lumbar MRI Findings with Current and Future... : Spine

• Updated on: 2023-06-17
• Read original article here

Low back pain (LBP) is the leading cause of years lived with disability worldwide resulting in $US135billion health care spending each year in the United States and >€40billion in Germany. Most patients with low back pain (LBP) have nonspecific LBP, a term signifying that a pathoanatomical source cannot be specified. Most clinical practice guidelines recommend that patients with nonspecific LBP should not receive imaging or further diagnostic work-up. Critics of the term argue that it does not provide a biological basis for the pain nor guide clinical management and research should be undertaken to assist subgrouping of patients into homogenous subgroups that share a common pathoanatomical source.

Magnetic resonance imaging (MRI) of the spine is able to detect structural abnormalities that are potential sources of LBP. Some studies report that MRI degenerative findings are more common in individuals with current LBP and so potentially represent a way to subgroup nonspecific LBP. Other studies report that MRI findings are common in symptomatic as well as asymptomatic individuals and are likely part of normal aging. The association is presently uncertain as most studies are cross-sectional; longitudinal studies are sparse and mostly based on small convenience samples. There are no large cohort studies where population-based samples of participants have been imaged and their LBP status tracked over time. Previous longitudinal studies did not investigate associations between MRI findings and LBP separately for those with and without LBP at baseline in whom the predictive value of MRI findings may be different.

The aim of this study is to investigate the relationship between lumbar degenerative MRI findings and LBP in a large population-based cohort study. Specific aims were to investigate the relationship between MRI findings and LBP severity at baseline and how baseline MRI findings predict severity of LBP six years later, in those with and without LBP at baseline. Predictors were single MRI findings and the number of different MRI findings.

The Study of Health in Pomerania (SHIP) is a population-based cohort study conducted in West Pomerania, a region in the northeast of Germany. SHIP comprises two separate cohorts SHIP and SHIP-TREND. Detailed information on the SHIP study design, sampling, and quality assurance is described elsewhere. The study protocol was approved by the local Ethics Committee of the Medical University of Greifswald.

The present analysis is based on data from SHIP-2 (2008–2012) and SHIP-TREND-0 (2008–2012) and the respective follow-up examinations SHIP-3 (2014–2016) and SHIP-TREND-1 (2016–2019). The net sample comprising SHIP-2 and SHIP-Trend-0 consists of 6753 participants (Figure 1). Of these 3369 persons underwent MRI examination (1183 SHIP-2; 2186 SHIP-TREND-0). A total of 3384 participants did not receive an MRI invitation due to cost and time limitations, not being willing to undergo whole-body MRI examination or contraindications (e.g., metallic implants, physical constraints, or claustrophobia). A total of 2415 participants had complete information on all MRI findings and LBP severity at baseline. Of these 1819 had follow-up LBP severity information. The mean time between baseline (SHIP-2/SHIP-TREND-0) and follow-up (SHIP-3/SHIP-TREND-1) was 6.41 (SD 1.19) years.

Whole-body MRI imaging was obtained using a 1.5-Tesla system (Magnetom-Avanto; Siemens Medical Solutions, Erlangen, Germany). The complete imaging protocols have been described previously. Standardised native whole-body MRI examinations were performed by trained technicians including a detailed imaging of the spine. The present study analysed sagittal T1 and T2 weighted images (voxel size of 1.1 × 1.1 × 4 mm) as well as coronary turbo inversion recovery magnitude sequences. Before imaging analysis intra- and inter-reader reliability was established (kappa value ≥0.8) through double prereading of lumbar images of 150 random SHIP participants. Trained radiologists evaluated the lumbar images and had no access on other participant's information. Reporting was performed using a standardised protocol in a picture archiving and communication system (IMPAX ES 5.2, AGFA Healthcare, Mortsel, Belgium). The presence of each MRI finding was assessed at each level of the lumbar spine (L1/L2 to sacrum) and then recoded for analysis as described in Supplemental Data File Table S1, https://links.lww.com/BRS/B790. MRI findings were coded as presence of (yes/no):

Information on imaging findings at each lumbar spine level was combined to define the presence of the respective finding at any level for each participant (yes/no). If participants had Modic change type 1 in combination with type 2 or 3 at any spinal level this was categorized as type 1. The sum of different MRI findings for each participant was calculated and categorized as “1,” “2,” “3,” “4,” and “5 or more findings.”

In a computer-assisted personal interview at baseline and follow-up, participants were asked “Have you had back pain within the last 3 months?" Furthermore, average back pain was assessed on a 0 to 10 numerical rating scale using German versions of the graded back pain items by Von Korff “How strong was your back pain within the last three months, if 0 = no pain and 10 = strongest imaginable pain?”. Disability during the last three months was assessed by asking participants to rate their “Interference due to back pain within the last 3 months, if 0 = no restriction and 10 = no activity possible?.” The study outcome, LBP severity, was calculated as the mean of average back pain and disability. Age, sex and duration of education (<10 years, 10 years, >10 years) were assessed at baseline.

Percentages, means, and corresponding confidence intervals (CIs) were calculated for all participants and stratified by baseline pain. Due to the skewed distribution of LBP severity with a high proportion of zeros, cross-sectional associations of MRI findings, and LBP severity were analyzed using two-part models.

Longitudinal models included all cases with complete information and were weighted using inverse probability weights to address potential selection bias. Weights accounted for sex, age, occupational status, health care consultation, medication use during last 7 days, body mass index, distance between place of residence and MRI study center, mental and physical health. Multivariable, linear mixed-effects regression models (level 1: timepoints, level 2: participants) were fitted to estimate the longitudinal association between MRI findings and LBP severity. In the next step, longitudinal analyses were conducted using separate models for participants with LBP (linear mixed-effects regression models) and without LBP (general linear models) at baseline. All models were adjusted for age, sex, and educational level. Analyses were conducted using STATA 14.2 (2016; StataCorp LP, College Station, TX).

In addition to complete case analyses, full information maximum likelihood estimates were used to address missing data using Mplus (Version 8.4).

Participant characteristics are presented in Table 1. At baseline 59.5% of participants had LBP during the last 3 months with a mean LBP severity of 4.1 points on a 0 to 10 scale. The proportion of MRI findings ranged from 0.9% for spondylolisthesis to 57.6% for disc degeneration per participant. Of all participants, 23.6% had no MRI findings, 25.1% had one, 21.2% had two, 13.0% had three, 8.8% had four, and 8.3% had five or more different MRI findings. All of the MRI findings were present among persons with and without LBP, with slightly higher rates in those with LBP. 77.8% of persons with pain at baseline and 74.4% of persons without pain at baseline had a least one MRI finding. Percentages stratified for age quintiles and pain at baseline are presented in Figure 2 and Table 2. A higher proportion of MRI findings was found in older age groups: 50% of persons younger than 41 years, approximately 80% of persons aged 50 to 59 years, and >90% of persons aged 70 and older had at least one MRI finding.

The cross-sectional associations between MRI findings and LBP severity at baseline are presented in Table 3. All mean effects were small, beta values ranged from 0.06 (95% CI −0.09 to 0.22) for high-intensity zone to 0.83 (95% CI −0.19 to 1.84) for spondylolisthesis. With the exception of high-intensity zone, spondylolisthesis, and Modic change type 2 or 3, the presence of all MRI findings was statistically significantly associated with greater LBP severity. The number of different MRI findings was weakly associated with LBP severity. Compared to participants with no MRI finding those with five or more MRI findings had 0.84 points (95% CI 0.50 to 1.17) higher LBP severity.

The longitudinal associations between MRI findings and LBP severity among all participants are presented in Table 4. Most mean effects were small with inconsistent directions ranging from −0.15 (95% CI −0.57 to 0.28) for disc height loss to 0.56 (95% CI −0.09 to 1.21) for spinal canal stenosis. Persons with spondylolisthesis at baseline had lower mean LBP severity at follow-up compared to persons without spondylolisthesis (−2.09; 95% CI −4.60 to 0.42), but the result was associated with a high degree of uncertainty. Most associations were not statistically significant, except for the presence of Schmorl lesions, which were associated with slightly lower LBP severity at follow-up (−0.54; 95% CI −0.84 to −0.24). The number of different MRI findings at baseline was not significantly associated with LBP severity at follow-up.

The longitudinal associations between different types of MRI findings and change in LBP severity stratified for presence of baseline pain are presented in Figure 3. In all participants, irrespective of the presence of LBP at baseline, associations between most MRI findings and LBP severity at follow-up were small (<1 point on 0–10 scale). Among persons with no LBP at baseline the presence of Schmorl lesions (−0.53; 95% CI −0.86 to −0.19) was statistically significantly associated with LBP severity after 6 years. Among persons with LBP at baseline disc degeneration was statistically significantly associated with higher LBP severity (−0.47; 95% CI 0.03–0.91). The presence of spondylolisthesis appeared to be associated with moderately greater future pain in those with no baseline pain (1.11; 95% CI −0.06 to 2.28) and moderately less future pain in those with baseline pain (−1.66; 95% CI −4.24 to 0.91).

The longitudinal associations between the number of MRI findings and LBP severity stratified for presence of baseline pain are presented in Figure 4. Among persons without LBP at baseline the presence of five or more MRI findings was associated with moderately greater future pain (1.21; 95% CI 0.04–2.37). Among persons with LBP at baseline effect sizes were very small and not statistically significant throughout.

Results persisted in sensitivity analyses using full information maximum likelihood estimation based on the complete data set including participants with missing values (see Supplemental Data File, Table S2, https://links.lww.com/BRS/B790 and S3, https://links.lww.com/BRS/B790).

This population-based cohort study examined the associations between lumbar degenerative MRI findings with current and future LBP. MRI findings were present in people with and without pain at baseline. A larger proportion of MRI findings was found in older age groups. The associations between MRI findings and LBP severity at baseline were mostly statistically significant but small with effect sizes typically less than half a unit on a 0 to 10 pain severity scale. Longitudinally, most associations were of negligible size, pointing to a low overall relevance of MRI findings for predicting future LBP in the general population. This does not exclude the possibility of relevant associations in population subgroups. Spondylolisthesis was one MRI finding that appeared to be associated with greater future pain severity in those with no baseline pain but, less future pain in those with baseline pain. Among persons without LBP at baseline the presence of multiple MRI findings (five or more) was associated with moderately greater future pain (1.21; 95% CI 0.04–2.37). No such association was evident in those with baseline pain.

To our knowledge, this is the first large population-based study investigating longitudinal relationships between lumbar MRI findings and LBP, where analyses were conducted for the whole cohort and separately for those with and without baseline pain. We used standardized interviews for assessment of LBP severity, standardised classification for common lumbar MRI findings and took steps to ensure quality of the MRI readings. Sensitivity analyses comparing multilevel linear regression models and full information maximum likelihood models support the robustness of our results. The study focused on common degenerative findings, but the sample is not suitable to detect red flag pathologies like metastases or fracture due to their low prevalence in a population-based sample. LBP was assessed as pain in the back and we assumed pragmatically that this refers to LBP, but we cannot exclude pain in other spinal regions. Patients were asked about pain in the last three months at baseline and follow-up to minimize recall bias. Given the high temporal variability of pain severity this has to be kept in mind when interpreting our results, as the average follow-up period was 6.4 years. It is possible that the predictive relationship may be different over shorter or longer follow-up periods. Information on any interventions received for LBP during the observation period such as medication use or surgical interventions was not available and may have altered the course of clinical symptoms. We assume that the number of participants, who underwent surgery during the observational period is low and expect no considerable effect on our findings. SHIP participants received no information regarding the MRI findings of their spine, which could otherwise potentially have resulted in more intensive treatment or surgery.

Studies on the association of MRI degenerative findings and LBP are very heterogenous regarding design, for example, sample source, duration of follow-up and pain assessment. Thus, direct comparison of our findings with previous studies is challenging. We found a larger proportion of MRI findings in older age groups in symptomatic as well as asymptomatic persons, which is in line with previous studies. Boden et al reported MRI findings (comprising disc herniation, spinal canal stenosis, disc degeneration, or bulging discs) in one-third of asymptomatic persons aged 20 to 80 years including four different MRI findings. In our study 74.4% of persons without pain at baseline had a least one MRI finding (comprising nine different MRI findings). A meta-analysis of cross-sectional case–control studies (14 studies, n = 3097) including mostly young and athletic persons reported clinically relevant higher prevalence of disc degeneration, Modic change, and disc herniation in symptomatic patients. However, we observed very small, clinically unimportant cross-sectional associations (Table 3). This might be due to differences of the study design, sample characteristics (e.g., age range), or definition of LBP. The few previous longitudinal studies are mostly based on small samples with present pain and typically found no clinically relevant association between single MRI findings and LBP at follow-up. The follow-up periods of these studies ranged from 1 to 13 years. One study investigating a population of adults without current LBP found that spondylolisthesis and disc degeneration were associated with increased risk of a future episode of LBP. These results are somewhat consistent with our findings that in people without baseline pain, disc degeneration and spondylolisthesis were associated with higher average pain. However, the Hancock study included participants recently recovered (<3 months) from an episode of LBP and the outcome was time to a recurrence, whereas in our study the pain free at baseline group may or may not have experienced LBP at any previous time in their life. Borenstein et al investigated the association of disc herniation, spinal canal stenosis, disc bulge and disc degeneration, and development of LBP among persons without baseline pain. The results are in line with our findings as the reported MRI findings were not predictive of the development of LBP. Only a few previous studies have investigated the relationship between the number of MRI findings and future LBP. In patients with no baseline pain Hancock et al found that individuals with three or more MRI findings were at greater risk of future recurrence of LBP over a period of 1 year. McNee et al reported no association between the number of MRI findings and LBP after 18 months among persons with a history of LBP. This is in accordance with our findings that participants with no baseline pain and five or more MRI findings had greater average pain severity scores and no or marginal association was found among persons with LBP at baseline.

Strengths of this work include the large sample size, the wide range of covered MRI findings, and the long follow-up period. Yet, given its limitations, inferences on the clinical relevance of our findings should only be made in the context of other evidence. According to guideline recommendations and previous findings from randomized controlled trials clinicians need to be cautious about drawing etiological assumptions on the causes and the prognosis of LBP solely based on the MRI findings in the absence of specific pathologies such as a fractures, infections or cancer (red flags), given the lack of strong and consistent relationships with pain severity. Our findings from a large population-based cohort that the MRI degenerative findings we examined do not have clinically important association with LBP add to recommendations that emphasize a cautious use of MRI in back pain patients.

Future research should include more extensive clinical data and attempt to better understand the pathoanatomical source for LBP by subgrouping patients into homogenous groups. Although the MRI findings we assessed seem unsuitable to define those subgroups, we cannot rule out that other imaging approaches offer a different potential. Our findings also suggest that the relationship between MRI findings and LBP may vary significantly (e.g., spondylolisthesis) depending on the population included (e.g., no present pain, current pain, never had pain, young vs. old) and future investigation in more homogenous populations may reveal stronger associations and advance understanding of the contribution of pathoanatomy to LBP.

The lumbar degenerative MRI findings we tested have either small or no association with current or future LBP on the population-level supporting current guidelines recommending restrictive imaging for LBP. Future studies should include continuous clinical data at multiple time points to better reflect the natural course of LBP. The relevance of lumbar degenerative MRI findings for the development of LBP is unclear due to few cohort studies with limited sample size where general-population samples of participants are imaged and their LBP status is tracked over time. This cohort study found only small associations between MRI findings such as disc degeneration or herniation and present or future LBP. Future research aiming to identify subgroups of patients of LBP could evaluate use of more advanced imaging protocols in combination with nonimaging clinical findings.