Federal funds were received in support of this work. No benefits in any form have been or will be received from a commercial party related directly or indirectly to the subject of this manuscript.

Supported by the Department of Veterans Affairs-ERIC Grant and in part by Grants HS-08194, HS-094990 from the Agency for Healthcare Research and Quality and P60-AR48093-01 from National Institute for Arthritis and Musculoskeletal and Skin Diseases.

From the Departments of *Radiology, †Health Services, ‡Medicine, §Biostatistics, ¶Epidemiology and ∥Neurological Surgery; the **Center for Cost and Outcomes Research, University of Washington, Seattle, Washington; and ††Veterans Affairs Puget Sound Health Care System and Seattle Epidemiologic Research and Information Center (ERIC), Seattle, Washington.

We performed a prospective cohort study of 131 Veteran Affairs subjects who were without low back pain at baseline, obtaining lumbar spine magnetic resonance images at baseline and 3 years. Depression had the largest hazard ratio of any baseline predictor of incident back pain.

Conclusion. Depression is an important predictor of new LBP, with MRI findings likely less important. New imaging findings have a low incidence; disc extrusions and nerve root contact may be the most important of these findings.

Results. After 3 years, 131 subjects were contacted, and 123 had repeat MRI. The 3-year incidence of pain was 67% (88 of 131). Depression had the largest hazard ratio (2.3, 95% CI = 1.2–4.4) of any baseline predictor of inci-dent back pain. Among baseline imaging findings, central spinal stenosis and nerve root contact had the highest, though nonsignificant, hazard ratios. We did not find an association between new LBP and type 1 endplate changes, disc degeneration, annular tears, or facet degeneration. The incidence of new MRI findings was low, with the most common new finding being disc signal loss in 11 (9%) subjects. All five subjects with new disc extrusions and all four subjects with new nerve root impingement had new pain.

Methods. We randomly selected 148 Veterans Affairs out-patients (aged 35 to 70) without LBP in the past 4 months. We compared baseline and 3-year lumbar spine MRI. Using data collected every 4 months, we developed a prediction model of back pain-free survival.

Summary of Background Data. Few prospective studies have examined clinical and anatomic risk factors for the development of LBP, or the incidence of new imaging findings and their relationship to back pain onset.

Objective. To determine predictors of new LBP as well as the 3-year incidence of magnetic resonance imaging (MRI) findings.

ArticlePlus

Click on the links below to access all the ArticlePlus for this article.

Please note that ArticlePlus files may launch a viewer application outside of your web browser.

Researchers have questioned the clinical importance of many spine imaging findings for nearly 60 years.1–13 Findings such as disc height loss and disc bulges are common in individuals without low back pain (LBP). As imaging techniques advance, our ability to accurately depict anatomy improves, yet paradoxically, confusion regarding the clinical importance of some anatomic findings increases.6,8,9,12–14

We previously reported a cohort of subjects with little or no back pain and no sciatica at the time of magnetic resonance imaging (MRI). Some reported prior back pain, allowing us to identify imaging findings associated with earlier symptoms.15 Most imaging findings fell into one of five categories: (1) common findings with little relationship to either aging or previous back pain (e.g., annular tears, disc protrusions); (2) common findings that were associated with increased age, but not with prior symptoms (e.g., disc bulges, facet joint degeneration, endplate changes, mild spondylolisthesis); (3) common findings related both to aging and previous LBP (e.g., decreased disc signal on T2-weighted images, decreased disc height); (4) rare findings unrelated to age, but strongly associated with previous back pain (disc extrusions and nerve root contact); and (5) moderate or severe stenosis, which was related to previous back pain, mild current symptoms and aging.

We sought to determine the 3-year follow-up incidence of LBP in this cohort and identify its risk factors. We also aimed to describe incident imaging findings and their relationship to onset of new symptoms.

Materials and Methods

Cohort Assembly.

We previously described details of the cohort assembly.15 Briefly, we used electronic records to randomly sample patients from four clinics at the Veterans Affairs Puget Sound Health Care System, Seattle Division, stratifying by age, with half of the subjects aged 35 to 52 years and the remainder aged 53 to 70.

After excluding patients with ICD-9 diagnostic and procedure codes related to LBP or lumbar surgery, we contacted the remaining patients in random order. We excluded subjects with any history of back surgery, chymopapain injections, discography, acute lumbar trauma, fibromyalgia, peripheral neuropathy, or serious comorbid disease that could affect 3-year survival. Eligible subjects completed a back pain bothersomeness questionnaire and a modified Roland Disability Scale.16 We excluded patients with any sciatica or LBP more than “mildly bothersome” in the previous 4 months, or a modified Roland score ≥3.

All imaging was free, and we reimbursed subjects $55. The University of Washington Institutional Review Board approved the study and all subjects gave written informed consent.

Baseline and Outcome Measures.

Participants completed a baseline questionnaire that included demographic information, comorbidity, back pain-specific functional status, general functional status, and health-related quality of life.

Our primary measure of back pain and sciatica was the pain frequency index (PFI).17 Subjects rated the frequency of the following four symptoms on a 1 (none of the time) to 6 (all of the time) point scale: (A) low back or buttock pain; (B) sciatic leg pain; (C) numbness or tingling in the leg, foot, or groin; and (D) weakness in leg or foot. The aggregate index scores ranged from 4 to 24. We defined an incident case as a subject who rated pain frequency for low back or buttock pain as more than “some of the time” (>2), or any of the other three symptoms as more than “none” (>1).

We used the modified Roland Disability Scale to assess the functional impact of LBP. Scores range from 0 (no dysfunction) to 23 (most severe functional impairment).

We assessed health-related quality-of-life with the SF-12,18,19 a 12-item version of the SF-36 general health status questionnaire. The SF-12 provides physical and mental component scores, which range from 0 (poor health) to 100 (good health) and are normalized so that a score of 50 equals the mean score in the US general population.

Subjects completed a comorbidity questionnaire17 that relies on self-reported conditions. Subjects classified their previous history of LBP as none previously, 1 to 5 episodes, or >5 episodes.

We contacted subjects every 4 months by telephone, except at 36 months, when they returned for a repeat MRI. We asked all outcome questions at 12, 24, and 36 months. At all other follow-ups, subjects were given the full battery of questions only if they scored back pain >2 on the pain frequency questionnaire.

Imaging.

We used a Philips 1.5 Tesla system to perform all MRIs, obtaining sagittal and axial T1- and T2-weighted images through each of the five lumbar intervertebral disc levels. One of two attending neuroradiologists, both senior members of the American Society of Neuroradiology and with clinical and academic expertise in lumbar spine imaging, interpreted all MRI images (J.G.J., D.R.H.). Other than knowing subjects were asymptomatic at baseline, clinical information was concealed from the radiologists.

The radiologists independently interpreted a sample of the examinations, and agreement between readers for each finding was calculated using the unweighted κ statistic for dichotomous variables and the weighted κ statistic for ordinal variables.20,21 The radiologists discussed disagreements and reached a consensus interpretation. We repeated this process until agreement was substantial, after which the radiologists each interpreted half of the remaining images.

The same radiologist interpreted the baseline and 3-year images to avoid inter-reader variability. The radiologists reviewed both the baseline images and their interpretations when reading the follow-up images, again to minimize variability other than true anatomic changes.

Subjects were not informed of MRI findings, unless a finding required immediate medical attention, because “labeling” patients with MRI diagnoses might have altered their behavior, sensitized them to trivial symptoms, or patients might have amplified symptoms that occurred.

At each lumbar disc level, the radiologists recorded the presence or absence of anatomic findings that are postulated to cause back pain or sciatica. Whenever possible, the radiologists used published diagnostic definitions.8,22–25 Disc herniations were classified as protrusions or extrusions.26 Endplate changes were classified as type 1 (edema), type 2 (fat), and type 3 (sclerosis) according to the scheme of Modic et al.27 See Appendix for a complete description of imaging variables.

Data Analysis.

We used Microsoft Access and SPSS28 for data management and analysis, respectively. We applied the χ2 test for trend (linear by linear) for ordinal variables, Yates continuity correction for dichotomous variables, and Fisher exact test for sparse contingency tables.29

We used a proportional hazards model to investigate the relationship between the incident LBP and potential risk factors. Clinical covariates included in our primary model were age, gender, history of prior LBP, smoking history, body mass index (BMI), and depression. Depression was ascertained with a checklist adapted from previous studies, using the following 3 questions: (1) Do you have depression? (2) Are you being treated for depression? (3) Does depression limit your activities? Baseline imaging covariates tested in our model were bulge, protrusion, extrusion, disc height loss, disc signal loss, annular tear, type 1 endplate findings,27 central stenosis, facet degeneration, and nerve root contact.

Role of the Funding Source.

This study was funded primarily by the Department of Veterans Affairs-Epidemiologic Research and Information Center Grant, with additional support for the investigators time from the National Institutes of Health (National Institute for Arthritis and Musculoskeletal and Skin Diseases). Neither the Veterans Affairs nor NIH had any role other than funding.

Results

Subject Characteristics

Of 148 enrolled subjects, 123 returned for repeat MRI after 3 years. We contacted eight additional subjects by telephone, making the 3-year follow-up rate 88.5% (131 of 148). We contacted 94 subjects (64%) at all 9 follow-up intervals. The follow-up rates at the 4-month time-points varied between 84.5% at 12 months and 98.6% at 4 months, with a mean of 90.7%. Because of occasional missing data, the sample size in each of the subsequent analyses varies slightly.

Compared with subjects available for follow-up, the 17 subjects without follow-up were more likely to be disabled for nonback pain causes [7 of 17 (41%) versus 21 of 131 (16%), P = 0.02, χ2], have been hospitalized for back pain [2 of 17 (12%) versus 1 of 131 (1%), P = 0.04, Fisher exact test], and have heart disease [5 of 17 (29%) versus 13 of 131 (10%), P = 0.04, Fisher exact test]. The mean age at enrollment of subjects available for follow-up was 54 (9.9 SD) years and the majority of subjects were white males, reflecting the population at this Veterans Affairs hospital (Table 1). Forty-seven percent (69 of 148) of the subjects had never experienced LBP and 16% (23 of 148) had had more than five episodes. All 21 subjects with self-identified depression at baseline were either being treated for depression (19 of 21) or said that their activities were limited by depression (15 of 21). Moreover, subjects with self-identified depression had significantly lower mean SF-12 mental health component summary scores than subjects without depression [56.5 (4.2 SD) versus 38.2 (12.9 SD), P < 0.001]. Additional baseline findings can be found in a previous publication.15

LBP and Back-Related Functional Status

For our primary analysis, we defined incident back pain as a score of >2 for low back pain or >1 for leg pain, numbness or weakness on the 6-point PFI at any time during the 3-year follow-up. Thus, our definition of incident back pain also includes radicular pain. Correlations between the symptom frequency index and other measures of back symptoms and function were strong (Pearson correlation coefficient between the pain frequency index and Roland disability scale at 3 years was 0.78, P < 0.01).

Sixty-seven percent of subjects (88 of 131) had incident LBP over 3 years, with 58 subjects (44%) having pain that was at least moderately bothersome. These proportions were not significantly different for the subjects <53 versus ≥53, so we report only data for the entire cohort. The SF-12 was relatively stable over time, but there was deterioration in the back-specific pain and functional status measures.

Incident MRI Findings (n = 123)

New imaging findings were infrequent. The most common new finding after 3 years was disc signal loss in 11 of 123 (9%; Table 2). No subjects developed MRI evidence of cancer or infection.

There were 114 discs in 67 subjects that had decreased height at both baseline and follow-up, and an additional 8 discs in 6 subjects that had lost height after 3 years. At baseline, 3 of these 6 subjects had demonstrated height loss at other levels.

Five subjects had a disc that changed from normal to bulging, 8 had a disc that changed from normal to protrusion, and 1 had a disc that changed from bulging to protrusion. Five subjects had a disc that changed from normal to extrusion, and 4 of these had new nerve root contact with these disc extrusions (Figure 1).

Some imaging findings became less severe or even normal. Four discs in 3 subjects changed from bulging to normal. Two subjects had protruded discs that transformed to bulges, and one of these subjects had a 10% retrolisthesis that reverted to 0% (Figure 2).

There were 8 levels in 7 subjects that changed from normal endplates at baseline to type 1 at 3 years, and 3 levels in 3 subjects that changed from normal to type 2. There was only 1 level that changed from type 1 to type 2 at 3 years. All 42 discs with type 2 endplates, and the one disc with a type 3 endplate at baseline remained stable over 3 years.

Risk of Incident LBP with Imaging and Clinical Factors

We used Cox regression models to determine which baseline factors predicted new onset of back pain during follow-up. Separate models of individual baseline imaging variables controlling only for age and a multivariate model that combined clinical factors and the three baseline imaging findings with the largest univariate hazard ratios and clinical factors produced similar findings. Therefore we report only the results of the multivariate model (Table 3). Self-identified depression was the strongest predictor of subsequent back pain, with a greater hazard ratio (HR) than any imaging finding (HR = 2.3, 95% CI = 1.2–4.4). Surprisingly, disc protrusions were associated with a lower risk of subsequent back pain (HR = 0.5, 95% CI = 0.3–0.9). Nerve root contact (HR = 1.9, 95% CI = 0.6–5.8) and central stenosis (HR = 1.8, 95% CI = 0.8–4.1) had the largest hazard ratios among baseline imaging findings, and they were associated with incident back pain in the expected directionbut not statistically significant. Figure 3 depicts the survival curve generated by this Cox regression, comparing groups with and without baseline depression.