Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You and Your Team.

Learn More →

Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability Among Middle-aged, Community-Dwelling Women

Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability... Key Points Question Are lumbar spine IMPORTANCE Previous studies, using mostly cross-sectional data, provide conflicting evidence of radiographic changes associated with an association between lumbar spine radiographic changes and the severity of back pain–related severity of back pain–related disability disability. Such conflicting evidence may be associated with widely unnecessary diagnostic imaging among middle-aged, community- of the lumbar spine. dwelling women? Findings In this population-based OBJECTIVE To examine both cross-sectional and longitudinal associations between lumbar spine cohort study of women from the UK, radiographic changes and the severity of back pain–related disability among middle-aged, community- there was no evidence to support an dwelling women. association between a higher number of lumbar segments with radiographic DESIGN, SETTING, AND PARTICIPANTS This population-based prospective cohort study used data changes (Kellgren-Lawrence grade, from the Chingford 1000 Women Study. Analyses included data collected from year 6 (1994-1996; osteophytes, and disc space narrowing) physical activity was measured), year 9 (1997-1999; treated as baseline), and year 15 (2003-2005), and more severe back pain–related with a total length of follow-up for longitudinal analyses of 6 years. Data were analyzed from April 17 disability in cross-sectional (650 to November 3, 2020. women) or longitudinal (443 women) analyses. EXPOSURES Primary exposure was lumbar spine radiographic changes, defined using the Kellgren- Lawrence (K-L) grade. Secondary exposures were defined using presence of osteophytes and disc Meaning The findings suggest that the space narrowing. The composite score combined the number of lumbar spine segments with definite changes detected on lumbar changes detected on radiographic images (ie, radiographic changes) (K-L grade2, which means at radiographs provide limited value for least definite osteophyte and possible narrowing of disc space are present; osteophyte and disc decision-making regarding back pain space narrowing grade1, which means at least mild or definite changes are present). management in this population. MAIN OUTCOMES AND MEASURES Self-reported back pain–related disability measured in years 9 Supplemental content and 15 assessed by the St Thomas disability questionnaire. Author affiliations and article information are listed at the end of this article. RESULTS Among 650 women (mean [SD] age, 61.3 [5.9] years) in cross-sectional analyses and 443 women (mean [SD] age, 60.6 [6.0] years) in longitudinal analyses, there was no evidence to support an association between higher number of lumbar segments with radiographic changes (K-L grade, osteophytes, and disc space narrowing) and more severe back pain–related disability (eg, cross- sectional analyses using the K-L grade; 1 segment vs 0 segment: adjusted odds ratio, 1.22 [95% CI, 0.76-1.96]). No interactions were found of an association between lumbar spine radiographic changes and the severity of back pain–specific disability with age, body mass index, or smoking status. CONCLUSIONS AND RELEVANCE In this cohort of middle-aged, community-dwelling women, there was no evidence to support an association between a higher number of lumbar segments with radiographic changes (K-L grade, osteophytes, and disc space narrowing) and more severe back (continued) Open Access. This is an open access article distributed under the terms of the CC-BY License. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 1/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability Abstract (continued) pain–related disability cross-sectionally or over time. These findings provide further evidence against routinely using diagnostic imaging of the lumbar spine. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 Introduction Low back pain (LBP) is a highly prevalent condition in the general population worldwide and has been the leading cause of disability for nearly 3 decades, according to the Global Burden of Disease Study 1,2 2017. Among all musculoskeletal problems, LBP is also the most common reason for patients to seek primary care. Current guidelines for treatment of LBP do not recommend routinely using diagnostic imaging, except when patients either present with severe, progressive neurologic deficits or with signs or symptoms indicative of a serious or specific underlying condition (eg, fracture or 4-6 cancer). Nonetheless, diagnostic imaging is still widely used in clinical practice for LBP, with a recent meta-analysis indicating that more than 15% of patients in primary care and approximately 25% in emergency care receive a referral for simple imaging (mainly radiograph). Moreover, nearly 10% of patients with LBP in primary care and emergency care also receive a referral for complex imaging (mainly computed tomography scan and magnetic resonance imaging). There has also been a 53% relative increase in referrals for complex imaging from 1995 to 2017, with no change observed during that period for the rate of referrals for simple imaging. Unnecessary diagnostic imaging not only wastes limited medical resources but is also associated with poorer health outcomes, such as iatrogenic disease from techniques that use ionizing radiation. In addition, patients who undergo unnecessary diagnostic imaging might be labeled with a pseudodisease, which may be associated with unnecessary subsequent interventions that may have adverse effects. Possible explanations for the unwarranted prevalence of imaging referrals for LBP are (1) the patient’s expectation that imaging results could provide valuable information on the cause and, consequently, the appropriate management of her or his condition and (2) the clinician’s desire to 10-12 reassure the patient of the absence of any underlying pathologic condition. Previous studies have 9,13 confirmed that imaging does not improve clinical outcomes for patients with LBP. However, the definition of normal or abnormal imaging diagnostic findings is still debatable. Currently, the presence of osteophytes and disc space narrowing are the most frequent changes detected on radiographs (hereafter referred to as radiographic changes) that may be indicative of spinal pathologic conditions, and the Kellgren-Lawrence (K-L) grade is the tool commonly used to assess the severity of osteoarthritis. A study including elderly women who lived in rural South Korea showed a positive association of the presence of osteophytes (grade2), disc space narrowing (grade2), and K-L grade (grade2) with the severity of disability, measured by a validated Korean version of the Oswestry Disability Index. However, a study conducted in Sri Lanka including patients with LBP concluded that neither disc space narrowing nor the presence of osteophytes was associated with the severity of disability (also measured with the Modified Oswestry Disability Index). Both studies are cross-sectional and failed to adjust their analyses for important confounders, including smoking status, level of participation in physical activity, and medication use. Past studies have also failed to identify whether the number of affected lumbar segments is associated with the severity of back pain–related disability. Therefore, the role of radiographic findings as a potential prognostic factor of the clinical course of LBP is still unclear and needs to be fully explored in population-based cohort studies. Previous studies of radiographic changes in knee osteoarthritis have indicated that the presence of osteophytes may be used to diagnose the condition and that the presence of joint space narrowing may be used to assess both the diagnosis 17,18 and the progression of osteoarthritis. This is still to be elucidated among patients with LBP. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 2/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability The aim of this study was to examine both cross-sectional and longitudinal associations between lumbar radiographic changes and the severity of back pain–related disability among middle- aged, community-dwelling women using composite scores that combined the number of segments and type of changes in terms of K-L grade, disc space narrowing, and osteophytes. We hypothesized that a higher number of segments with lumbar radiographic changes would be associated with more severe back pain–related disability. Methods Study Design, Data Sources, and Study Population From an age and sex register of a large practice of more than 11 000 patients in Chingford in east London, UK, all 1353 women in the age range of 45 to 64 years were invited to participate in a population study assessing musculoskeletal diseases. A total of 1003 women were examined between 1989 and 1991 (year 1; baseline visit for original cohort); 6 died, 66 had moved away, and 278 refused to participate or did not respond. All the women lived within 8 km (5 miles) of the general practice, and 98% of the women were white. Women from this general practice are similar to the UK general population in terms of weight, height, and body mass index (BMI). In the data analyses, and given their availability, we included data on physical activity collected in year 6 (1994- 1996 [ie, prebaseline]) and imaging data, all other covariates, and the outcome for cross-sectional analyses collected in year 9 (1997-1999 [ie, baseline for our study]). The outcome for longitudinal analyses was obtained in year 15 (2003-2005). We followed the Strengthening the Reporting of 19,20 Observational Studies in Epidemiology (STROBE) reporting guideline. The Waltham Forest and Redbridge local research ethics committee approved the study, and all participants provided written informed consent to participate in the study. Exposures Lateral lumbar spine radiographs at year 9 were taken by 1 radiographer, centered on the L3 vertebra, with the participants in the left lateral recumbent position. A single trained observer (a rheumatologist) blinded to patient identity and chronologic order read all of the radiographs. Within- observer variation was assessed by test-retest analysis of 40 randomly selected radiographs from the study. Good within-observer reproducibility (κ = 0.78-0.89) was found. At each lumbar spine segment (L1-L2, L2-L3, L3-L4, and L4-L5), disc space narrowing and osteophytes (both anterior and posterior) were assessed through the semiquantitative method reported by Lane et al, with grade 0 corresponding to normal, grade 1 to mild narrowing and osteophytes, grade 2 to moderate narrowing and osteophytes, and grade 3 to severe narrowing and osteophytes. The Kellgren- Lawrence (K-L) grade was summarized as grade 0 indicating normal; grade 1 indicating doubtful narrowing of disc space and possible osteophytic lipping; grade 2 indicating definite osteophyte and possible narrowing of disc space; grade 3 indicating moderate multiple osteophytes, definite narrowing of disc space, some sclerosis, and possible deformity of bone contour; and grade 4 indicating large osteophytes, marked narrowing of disc space, severe sclerosis, and definite deformity of bone contour. Considering that the number of lumbar spine segments with radiographic changes detected and the various types of radiographic changes might be associated with the results, we generated 3 composite scores: a K-L grade–based score, an osteophyte grade–based score, and a disc space narrowing grade–based score; at each segment, a binary exposure variable of 1 (K-L grade2, which means at least definite osteophyte and possible narrowing of disc space are present; disc space narrowing and osteophyte grade1, which means at least mild or definite changes are present) vs 0 (K-L grade 0 or 1; disc space narrowing and osteophyte grade 0) was used. The composite score was then calculated as the final L1-L2 score + L2-L3 score + L3-L4 score + L4-L5 score, with values ranging from 0 to 4 (where 0 indicates no lumbar spine segments with radiographic changes detected and 4 indicates 4 lumbar spine segments with radiographic changes detected). The K-L grade–based JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 3/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability score was defined as the primary exposure. Osteophyte and disc space narrowing grade–based scores were set as secondary exposures. Outcomes Back pain–related disability was assessed at year 9 and year 15 using a back pain questionnaire (St Thomas disability questionnaire), which correlated well with the Oswestry Disability Questionnaire (r =0.77; P < .001). The outcome was defined by questions at 2 levels. At the first level, women were asked whether they had any back pain for at least 1 day at any time in the last 12 months. At the second level, those who answered yes to the first-level question were asked 8 questions related to the disability due to back pain (corresponding to the previous year’s status): walking around the house; standing for 15 minutes; getting up from a low chair; getting out of a bath; getting in and out of a car; going up and down stairs; putting on socks, stockings, or tights; and cutting toenails. Each question was summarized as grade 0 indicating no difficulty, grade 1 indicating difficult but possible, and grade 2 indicating impossible. We built a composite score based on the aforementioned 8 questions; values ranged from 0 to 16, with higher values corresponding to more severe disability. We assumed the composite score as 0 if women answered “no” to the first-level question. In case of missing data for any of the 8 questions, we kept the data if the women responded to at least 6 questions and calculated the composite score as [(total score)/(number of questions answered)] × 8. Covariates Causal diagram through DAGitty, version 3.0 was used to choose the minimal sufficient adjustment sets for estimating the total association of the exposure with the outcome. Age, BMI, smoking status, back pain status, bisphosphonate use, and physical activity were included in the final model (details in eFigure 1 in the Supplement). All covariates, except physical activity, which was measured in year 6, were measured in year 9. Statistical Analysis Data were analyzed from April 17 to November 3, 2020. Owing to the skewed distribution of back pain–related disability (eFigure 2 in the Supplement), ordinal logistic regression, which holds a proportional odds assumption, was performed. Considering that physical activity was measured at a different time point (ie, year 6) compared with other covariates (ie, year 9) and with potential measurement error, we established a stepped modeling framework: step 1, unadjusted analyses; step 2, analyses adjusted for age, BMI, back pain status, bisphosphonate use status, and smoking status (additionally adjusted for year 9 back pain–related disability for the longitudinal analysis); and step 3, analyses further adjusted for physical activity. Separate analyses were conducted for cross-sectional and longitudinal data. For the longitudinal analyses, data on lumbar spine radiographic changes collected in year 9 were treated as the exposure, and back pain–related disability data collected in year 15 were treated as the outcome. In addition to the confounders mentioned, data on back pain–related disability collected in year 9 were included in the longitudinal analysis as a strong prognostic factor to adjust. Based on the recommendation from Modern Epidemiology, the exposures were modeled as unordered categorical variables and trend test. The proportion of missing data in each covariate is provided in Table 1. Missing data were handled through multiple imputation, which holds a missing-at-random assumption. The assumption was graphically tested (eFigure 3 in the Supplement). No additional variables were used; all covariates in the minimal sufficient adjustment sets were used. Flexible additive models with 10 imputed data sets were used. We did not impute data for the exposure variables. The relative risk was presented as adjusted proportional odds ratios (ORs) with 95% CIs. Extensive sensitivity analyses were performed (eAppendix 1 in the Supplement). All statistical analyses were performed with rms, Hmisc, and tidyverse packages in R, version 3.6.2 (R Group for Statistical Computing). Details of the statistical methods are provided in eAppendix 2 in the Supplement. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 4/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability Table 1. Baseline Characteristics of Study Participants Participants, No. (%) No. of segments Characteristic 012 3 4 Whole cohort Cross-sectional (n = 650) No. 154 142 140 118 96 650 Age, mean (SD), y 59.7 (5.5) 60.3 (5.7) 60.8 (5.8) 62.8 (5.7) 64.5 (5.8) 61.3 (5.9) Missing 1 (0.6) 2 (1.4) 4 (2.9) 2 (1.7) 1 (1.0) 10 (1.5) Smoking status Never 105 (68.2) 84 (59.2) 89 (63.6) 71 (60.2) 61 (63.5) 410 (63.1) Current 18 (11.7) 25 (17.6) 20 (14.3) 20 (16.9) 15 (15.6) 98 (15.1) Ex-smoker 28 (18.2) 31 (21.8) 25 (17.9) 23 (19.5) 18 (18.8) 125 (19.2) Missing 3 (1.9) 2 (1.4) 6 (4.3) 4 (3.4) 2 (2.1) 17 (2.6) BMI, mean (SD) 26.6 (4.2) 26.8 (5.1) 27.2 (5.0) 26.9 (4.4) 27.6 (5.4) 27.0 (4.8) Missing 1 (0.6) 3 (2.1) 4 (2.9) 4 (3.4) 1 (1.0) 13 (2.0) Back pain status Yes 51 (33.1) 46 (32.4) 39 (27.9) 42 (35.6) 32 (33.3) 210 (32.3) No 103 (66.9) 96 (67.6) 101 (72.1) 76 (64.4) 64 (66.7) 440 (67.7) Bisphosphonate use status Yes 5 (3.2) 3 (2.1) 4 (2.9) 1 (0.8) 1 (1.0) 14 (2.2) No 68 (44.2) 62 (43.7) 71 (50.7) 55 (46.6) 44 (45.8) 300 (46.2) Missing 81 (52.6) 77 (54.2) 65 (46.4) 62 (52.5) 51 (53.1) 336 (51.7) Physical activity Walking, km/wk <0.8 14 (9.1) 6 (4.2) 12 (8.6) 7 (5.9) 13 (13.5) 52 (8.0) 0.8 to <8.1 75 (48.7) 75 (52.8) 60 (42.9) 70 (59.3) 43 (44.8) 323 (49.7) 8.1 to <16.1 42 (27.3) 36 (25.4) 41 (29.3) 30 (25.4) 17 (17.7) 166 (25.5) ≥16.1 17 (11.0) 19 (13.4) 21 (15.0) 5 (4.2) 16 (16.7) 78 (12.0) Missing 6 (3.9) 6 (4.2) 6 (4.3) 6 (5.1) 7 (7.3) 31 (4.8) Job Sedentary 8 (5.2) 8 (5.6) 4 (2.9) 4 (3.4) 0 24 (3.7) Sedentary plus 23 (14.9) 14 (9.9) 16 (11.4) 13 (11.0) 8 (8.3) 74 (11.4) occasional exercise 0.5 Sedentary plus 0.5 88 (57.1) 77 (54.2) 87 (62.1) 73 (61.9) 59 (61.5) 384 (59.1) active (or active housework [eg, daily dusting or vacuuming]) Predominantly manual, 27 (17.5) 30 (21.1) 22 (15.7) 17 (14.4) 15 (15.6) 111 (17.1) active all day Missing 8 (5.2) 13 (9.2) 11 (7.9) 11 (9.3) 14 (14.6) 57 (8.8) Sport None 80 (51.9) 78 (54.9) 89 (63.6) 70 (59.3) 60 (62.5) 377 (58.0) Golf, bowling, 22 (14.3) 26 (18.3) 15 (10.7) 15 (12.7) 8 (8.3) 86 (13.2) badminton, cycling, or swimming, 1 h/wk 2 h/wk of Golf, 30 (19.5) 23 (16.2) 22 (15.7) 19 (16.1) 11 (11.5) 105 (16.2) bowling, badminton, cycling, or swimming or1hof staying fit, aerobics, or squash ≥2 h/wk of Staying fit, 17 (11.0) 9 (6.3) 8 (5.7) 8 (6.8) 11 (11.5) 53 (8.2) aerobics, or squash Missing 5 (3.2) 6 (4.2) 6 (4.3) 6 (5.1) 6 (6.3) 29 (4.5) Disability, year 9, 0 (0.0-5.8) 0 (0.0-6.0) 0 (0.0-3.3) 0 (0.0-4.9) 0 (0.0-3.3) 0 (0.0-5.0) median (IQR) Longitudinal (n = 443) No. 112 100 97 76 58 443 Age, mean (SD), y 59.1 (5.3) 59.5 (5.7) 60.5 (6.3) 62.6 (6.1) 62.8 (5.9) 60.6 (6.0) Missing 1 (0.9) 2 (2.0) 3 (3.1) 0 0 6 (1.4) (continued) JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 5/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability Table 1. Baseline Characteristics of Study Participants (continued) Participants, No. (%) No. of segments Characteristic 012 3 4 Whole cohort Smoking status Never 77 (68.8) 63 (63.0) 64 (66.0) 44 (57.9) 38 (65.5) 286 (64.6) Current 13 (11.6) 14 (14.0) 11 (11.3) 16 (21.1) 9 (15.5) 63 (14.2) Ex-smoker 19 (17.0) 21 (21.0) 18 (18.6) 15 (19.7) 11 (19.0) 84 (19.0) Missing 3 (2.7) 2 (2.0) 4 (4.1) 1 (1.3) 0 10 (2.3) BMI, mean (SD) 26.4 (4.3) 26.7 (4.9) 27.2 (5.2) 26.5 (3.7) 28.3 (5.3) 26.9 (4.7) Missing 1 (0.9) 3 (3.0) 3 (3.1) 2 (2.6) 0 9 (2.0) Back pain status Yes 40 (35.7) 32 (32.0) 24 (24.7) 26 (34.2) 23 (39.7) 145 (32.7) No 72 (64.3) 68 (68.0) 73 (75.3) 50 (65.8) 35 (60.3) 298 (67.3) Bisphosphonate use status Yes 4 (3.6) 3 (3.0) 2 (2.1) 1 (1.3) 1 (1.7) 11 (2.5) No 43 (38.4) 43 (43.0) 55 (56.7) 35 (46.1) 26 (44.8) 202 (45.6) Missing 65 (58.0) 54 (54.0) 40 (41.2) 40 (52.6) 31 (53.4) 230 (51.9) Physical activity Walking, km/wk <0.8 11 (9.8) 5 (5.0) 12 (12.4) 5 (6.6) 9 (15.5) 42 (9.5) 0.8 to <8.1 54 (48.2) 50 (50.0) 42 (43.3) 46 (60.5) 29 (50.0) 221 (49.9) 8.1 to <16.1 31 (27.7) 27 (27.0) 22 (22.7) 20 (26.3) 8 (13.8) 108 (24.4) ≥16.1 10 (8.9) 16 (16.0) 17 (17.5) 3 (3.9) 10 (17.2) 56 (12.6) Missing 6 (5.4) 2 (2.0) 4 (4.1) 2 (2.6) 2 (3.5) 16 (3.6) Job Sedentary 5 (4.5) 5 (5.0) 4 (4.1) 3 (3.9) 0 17 (3.8) Sedentary plus 15 (13.4) 13 (13.0) 10 (10.3) 9 (11.8) 6 (10.3) 53 (12.0) occasional exercise 0.5 Sedentary plus 0.5 65 (58.0) 55 (55.0) 62 (63.9) 49 (64.5) 37 (63.8) 268 (60.5) active (or active housework [eg, daily dusting or vacuuming]) Predominantly manual, 21 (18.8) 24 (24.0) 14 (14.4) 11 (14.5) 10 (17.2) 80 (18.1) active all day Missing 6 (5.4) 3 (3.0) 7 (7.2) 4 (5.3) 5 (8.6) 25 (5.6) Sport None 60 (53.6) 52 (52.0) 57 (58.8) 45 (59.2) 38 (65.5) 252 (56.9) Golf, bowling, 14 (12.5) 21 (21.0) 15 (15.5) 8 (10.5) 4 (6.9) 62 (14.0) badminton, cycling, or swimming, 1 h/wk 2 h/wk of Previous 22 (19.6) 19 (19.0) 16 (16.5) 17 (22.4) 6 (10.3) 80 (18.1) or1hof staying fit, aerobics, or squash Abbreviations: BMI, body mass index (calculated as ≥2 h/wk of Staying fit, 11 (9.8) 6 (6.0) 4 (4.1) 4 (5.3) 8 (13.8) 33 (7.4) aerobics, or squash weight in kilograms divided by height in meters Missing 5 (4.5) 2 (2.0) 5 (5.2) 2 (2.6) 2 (3.4) 16 (3.6) squared); IQR, interquartile range. Disability, year 15, 0 (0.0-5.0) 0 (0.0-5.1) 0 (0.0-4.0) 0 (0.0-2.5) 0 (0.0-4.8) 0 (0.0-4.8) Number of segments of lumbar spine radiographic median (IQR) changes (based on Kellgren-Lawrence grade). Results Participant Characteristics A total of 650 women (mean [SD] age, 61.3 [5.9] years) were included in cross-sectional analyses, and a total of 443 women (mean [SD] age, 60.6 [6.0] years) were included in longitudinal analyses (Table 1; Figure). Most study participants were classified as either never smokers or ex-smokers (Table 1). The median score of back pain–related disability was 0 (interquartile range, 0.0-5.0 in cross-sectional analyses and 0.0-4.8 in longitudinal analyses) in both cross-sectional and longitudinal analyses. The distribution of each lumbar spine radiographic change at each lumbar spine segment JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 6/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability is listed in eTable 1 in the Supplement. Redundancy analyses were performed to assess whether 1 exposure could be estimated from any 2 other exposures, at each lumbar spine segment. No exposure was redundant (eTable 2 in the Supplement). K-L Grade–Based Score Using a multivariable ordinal logistic regression model, we found that women who had 1 or more segments with lumbar spine radiographic K-L grade–based changes were not statistically more likely to report more disability compared with women with no observed changes in both cross-sectional (eg, 1 segment vs 0 segments; step 2 model; OR, 1.22 [95% CI, 0.76-1.96]) and longitudinal analyses (Table 2). When further adjustment was made for physical activity, the results were similar (OR, 1.19 [95% CI, 0.73-1.93]). No evidence was found to support a linear or nonlinear trend between number of segments with lumbar spine radiographic changes and the severity of back pain–related disability. Osteophyte Grade–Based Score and Disc Space Narrowing Grade–Based Score For osteophyte grade–based score, no statistically significant association was found between the number of lumbar segments with radiographic changes and the severity of back pain–related disability in both cross-sectional (eg, 1 segment vs 0 segment; step 2 model; OR, 0.83 [95% CI, 0.57- 1.22]) and longitudinal analyses (Table 3). Similar results were observed when further adjustments were made in the models to account for participation in physical activity in both cross-sectional and longitudinal analyses. In the longitudinal analysis, a greater number of affected segments were linearly associated with less severe back pain–related disability (step 2 model). For the disc space narrowing grade–based score, no statistically significant association was observed between the number of lumbar segments with radiographic changes and the severity of back pain–related disability in both cross-sectional (eg, 1 segment vs 0 segment; step 2 model; OR, 1.43 [95% CI, 0.78-2.61]) and longitudinal analyses (Table 4). Results were similar when we further adjusted for physical activity (OR, 1.41 [95% CI, 0.77-2.60]). For the cross-sectional analyses, a higher Figure. Flowchart of Study Participants 1003 Patients in the Chingford 1000 Women Study 828 Participants followed up in year 9 655 Participants followed up in year 15 178 Excluded 212 Excluded 159 Without radiographic data 100 Without radiographic data 6 Missing value in 4 Missing value in radiographic data radiographic data 13 Without outcome data 108 Without outcome data 650 Included in cross-sectional main analysis 443 Included in longitudinal main analysis Sensitivity analysis with different sample size: Sensitivity analysis with different sample size: 210 With back pain 145 With back pain 365 With severe disease per K-L grade–based score 260 With severe disease per K-L grade–based score 283 With severe disease per osteophyte grade–based score 206 With severe disease per osteophyte grade–based score 126 With severe disease per disc space narrowing grade–based score 87 With severe disease per disc space narrowing grade–based score A total of 95.0% women (421 of 443) in the longitudinal analysis are also in the cross-sectional analysis. Besides all data mentioned in the flowchart, we also collected the data from year 6 (1994-1996) because it was when 1 necessary covariate (ie, physical activity) was measured. K-L indicates Kellgren-Lawrence. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 7/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability number of segments of lumbar spine radiographic characteristics were nonlinearly associated with less severe back pain–related disability (step 2 model). Exploratory and Sensitivity Analyses We did not find interactions with age, BMI, or smoking status (eTable 3 in the Supplement) between lumbar spine radiographic changes and the severity of back pain–related disability. Overall, our results remained similar under extensive sensitivity analyses (eAppendices 3-8 and eTables 4-25 in the Supplement). All E-values are listed in eTable 26 in the Supplement. Discussion Key Results In this cohort of middle-aged women from Chingford in east London, UK, no evidence was found to support an association between a higher number of segments with lumbar radiographic changes (K-L grade, osteophyte, and disc space narrowing) and more severe back pain–related disability. Our Table 2. Association Between K-L Grade–Based Score and Severity of Back Pain–Related Disability K-L grade–based score, OR (95% CI) P value for trend Variable 0 Segments 1 Segment 2 Segments 3 Segments 4 Segments Linear model Nonlinear model Cross-sectional, year 9 (n = 650) Women, No. (%) 154 (23.7) 142 (21.8) 140 (21.5) 118 (18.2) 96 (14.8) NA NA Unadjusted 1 [Reference] 1.19 (0.75-1.89) 0.84 (0.52-1.36) 0.90 (0.54-1.49) 0.81 (0.47-1.38) .24 .78 Multivariable adjusted 1 [Reference] 1.22 (0.76-1.96) 0.84 (0.51-1.38) 0.92 (0.54-1.56) 0.89 (0.50-1.57) .39 .91 Further adjusted for 1 [Reference] 1.19 (0.73-1.93) 0.81 (0.49-1.35) 0.87 (0.51-1.49) 0.85 (0.48-1.50) .30 .96 physical activity Longitudinal, year 15 (n = 443) Women, No. (%) 112 (25.3) 100 (22.6) 97 (21.9) 76 (17.2) 58 (13.1) NA NA Unadjusted 1 [Reference] 1.22 (0.70-2.11) 0.95 (0.53-1.68) 0.77 (0.41-1.44) 1.05 (0.55-2.02) .57 .92 Multivariable adjusted 1 [Reference] 1.06 (0.57-1.96) 0.94 (0.50-1.76) 0.69 (0.34-1.38) 0.83 (0.40-1.72) .20 .79 Further adjusted for 1 [Reference] 1.08 (0.58-2.02) 0.91 (0.48-1.72) 0.67 (0.33-1.38) 0.80 (0.38-1.68) .17 .80 physical activity Abbreviations: K-L, Kellgren-Lawrence; NA, not applicable; OR, odds ratio. Adjusted for age, body mass index, smoking status, back pain status, and bisphosphonate use. Adjusted for age, body mass index, smoking status, back pain status, bisphosphonate use, and year 9 back pain–related disability. Table 3. Association Between Osteophyte Grade–Based Score and the Severity of Back Pain–Related Disability Osteophyte grade–based score, OR (95% CI) P value for trend Variable 0 Segments 1 Segment 2 Segments 3 Segments 4 Segments Linear model Nonlinear model Cross-sectional, year 9 (n = 650) Women, No. (%) 258 (39.7) 226 (34.8) 102 (15.7) 44 (6.8) 20 (3.1) NA NA Unadjusted 1 [Reference] 0.80 (0.55-1.16) 0.81 (0.50-1.32) 0.53 (0.25-1.12) 0.81 (0.31-2.14) .12 .59 Multivariable adjusted 1 [Reference] 0.83 (0.57-1.22) 0.78 (0.47-1.30) 0.58 (0.27-1.26) 1.03 (0.37-2.85) .25 .42 Further adjusted for 1 [Reference] 0.82 (0.56-1.21) 0.79 (0.47-1.32) 0.60 (0.28-1.31) 0.98 (0.35-2.73) .26 .46 physical activity Longitudinal, year 15 (n = 443) Women, No. (%) 192 (43.3) 157 (35.4) 67 (15.1) 19 (4.3) 8 (1.8) NA NA Unadjusted 1 [Reference] 0.71 (0.46-1.10) 0.60 (0.33-1.08) 0.41 (0.13-1.27) 0.49 (0.10-2.40) .02 .65 Multivariable adjusted 1 [Reference] 0.76 (0.47-1.24) 0.53 (0.28-1.02) 0.49 (0.14-1.70) 0.31 (0.06-1.72) .01 .75 Further adjusted for 1 [Reference] 0.76 (0.46-1.24) 0.52 (0.27-1.03) 0.50 (0.14-1.74) 0.33 (0.06-1.79) .01 .68 physical activity Abbreviations: NA, not applicable; OR, odds ratio. Adjusted for age, body mass index, smoking status, back pain status, and bisphosphonate use. Adjusted for age, body mass index, smoking status, back pain status, bisphosphonate use, and year 9 back pain–related disability. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 8/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability results remained unchanged after including potential interactions with important confounders, such as age, BMI, and smoking status, and after extensive sensitivity analyses. Comparison With Previous Studies For K-L results, our findings contradict those of Lee et al, who found that K-L grades were significantly associated with the Oswestry Disability Index. The main reason for such a discrepancy in results may be the design features of the study by Lee et al, which only included cross-sectional analyses with insufficient adjustment for important confounders (eg, physical activity, smoking status, and BMI). For osteophyte results, the results from the cross-sectional analyses are consistent with a previous cross-sectional study by Perera et al, who identified that the presence of osteophytes on radiographs was not associated with physical disability measured with the Oswestry Disability Index. However, results from our longitudinal analyses indicated that a greater number of affected segments were linearly associated with less severe back pain–related disability. One possible explanation is the biomechanical stability provided by spinal osteophytes, which have been proven to increase spinal resistance in compression. Another explanation is that the results simply reflect large numbers of analyses completed, which need future studies to verify. For disc space narrowing 16 15 results, our results are similar to those of Perera et al but different than the findings of Lee et al. Such differences could also be associated with the methodological limitations already described in the study conducted by Lee et al. Overall, the prevalence of lumbar spine radiographic findings in our study is similar to that of previous studies that indicated that many imaging-based spinal radiographic changes are likely part of normal, asymptomatic aging. Strengths and Limitations Our study has several strengths. To our knowledge, this is the first study to create a composite score that reflects the overall association of lumbar spine radiographic changes (ie, number of affected segments and severity of changes) with the severity of back pain–related disability. We used population-based data that contain a long-term follow-up with good recruitment and retention rates, and multiple potential confounders were measured. We also overcame some methodological limitations from previous studies; we included a cohort study design and incorporated a systematic way to select and control confounders, we have repeated measures of back pain–related disability that allow us to adjust for baseline disability, we assessed the potential interaction term, and we performed extensive sensitivity analyses to evaluate the robustness of the results. Limitations also need to be considered. First, the Chingford 1000 Women Study included middle-aged women in a specific area of the UK. We must exercise caution when generalizing the Table 4. Association Between Disc Space Narrowing Grade–Based Score and the Severity of Back Pain–Related Disability Disc space narrowing grade–based score, OR (95% CI) P value for trend Variable 0 Segments 1 Segment 2 Segments 3 Segments 4 Segments Linear model Nonlinear model Cross-sectional, year 9 (n = 650) Women, No. (%) 100 (15.4) 107 (16.5) 147 (22.6) 131 (20.2) 165 (25.4) NA NA Unadjusted 1 [Reference] 1.33 (0.74-2.37) 1.30 (0.75-2.24) 1.33 (0.77-2.31) 0.98 (0.57-1.69) .75 .11 Multivariable adjusted 1 [Reference] 1.43 (0.78-2.61) 1.56 (0.88-2.76) 1.44 (0.81-2.57) 1.07 (0.60-1.92) .94 .04 Further adjusted for 1 [Reference] 1.41 (0.77-2.60) 1.56 (0.87-2.77) 1.45 (0.81-2.59) 1.04 (0.57-1.87) .86 .03 physical activity Longitudinal, year 15 (n = 443) Women, No. (%) 70 (15.8) 84 (19.0) 102 (23.0) 88 (19.9) 99 (22.3) NA NA Unadjusted 1 [Reference] 0.91 (0.47-1.75) 0.78 (0.41-1.49) 1.07 (0.56-2.05) 1.24 (0.67-2.29) .34 .25 Multivariable adjusted 1 [Reference] 0.72 (0.34-1.53) 0.74 (0.36-1.52) 1.06 (0.52-2.20) 1.26 (0.62-2.57) .18 .18 Further adjusted for 1 [Reference] 0.68 (0.31-1.48) 0.67 (0.31-1.43) 1.12 (0.53-2.34) 1.33 (0.63-2.80) .13 .12 physical activity Abbreviations: NA, not applicable; OR, odds ratio. Adjusted for age, body mass index, smoking status, back pain status, and bisphosphonate use. Adjusted for age, body mass index, smoking status, back pain status, bisphosphonate use, and year 9 back pain–related disability. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 9/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability results to men, other age groups, other racial/ethnic groups, or other countries. Second, as with most studies, there is the potential for residual confounding (eg, participation in physical activity was measured 3 years before baseline). Third, the labeling of the images may have introduced potential bias in our results, given that there was only 1 observer and that lumbar spine levels were decided by the clinical experience. Fourth, owing to data unavailability, we could not establish whether there was any association between other radiologic changes, including spondylolisthesis or vertebral body height (ie, osteoporotic fractures), and severity of functional limitation. Fifth, although our outcome correlated well with the Oswestry Disability Questionnaire, it lacked strict validation. Sixth, although we aimed to focus on LBP–related disability, we only had a back pain variable, which might be slightly different from LBP. Implications for Practice and Research Clinicians may use the results of this study to educate patients and their colleagues that lumbar radiographic findings cannot provide prognostic information on back pain–related disability, further adding to the evidence supporting the urge to reduce unnecessary imaging referrals. Future studies should include participants of both sexes and larger sample sizes and should include multiple centers to increase external validity. The association between the findings of complex imaging (eg, computed tomography scans, magnetic resonance imaging, or nuclear bone scans) and symptom severity in people with LBP needs to be further explored, considering the increasing use of such imaging. Conclusions In this cohort of middle-aged, community-dwelling women, there was no evidence to support an association between a higher number of lumbar segments with radiographic changes (K-L grade, osteophytes, and disc space narrowing) and more severe back pain–related disability cross- sectionally or over time. The findings suggest that the changes detected on lumbar radiographs provide limited value for decision-making regarding back pain management in this population. ARTICLE INFORMATION Accepted for Publication: March 24, 2021. Published: May 20, 2021. doi:10.1001/jamanetworkopen.2021.10715 Open Access: This is an open access article distributed under the terms of the CC-BY License.©2021ChenLetal. JAMA Network Open. Corresponding Author: Lingxiao Chen, MBBS, MMed, Institute of Bone and Joint Research, The Kolling Institute, Northern Clinical School, Faculty of Medicine and Health, University of Sydney, Sydney, NSW 2064, Australia (lche4036 @uni.sydney.edu.au). Author Affiliations: Institute of Bone and Joint Research, The Kolling Institute, Northern Clinical School, Faculty of Medicine and Health, University of Sydney, Sydney, New South Wales, Australia (Chen, M. L. Ferreira); Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, Oxford, United Kingdom (Perera, Radojčić, Arden); Department of Allied Health Sciences, Faculty of Medicine, University of Colombo, Colombo, Sri Lanka (Perera); Centre for Sport, Exercise and Osteoarthritis Research Versus Arthritis, University of Oxford, Oxford, United Kingdom (Radojčić); School of Health Sciences, Faculty of Medicine and Health, University of Sydney, Sydney, New South Wales, Australia (Beckenkamp, P. H. Ferreira); Department of Twin Research and Genetic Epidemiology, King’s College London, London, United Kingdom (Hart, Spector); Medical Research Council Environmental Epidemiology Unit, University of Southampton, Southampton, United Kingdom (Arden). Author Contributions: Drs Chen and Perera had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Arden and M. L. Ferreira contributed equally as co–senior authors. Concept and design: Chen, Perera, Radojčić, Beckenkamp, P.H. Ferreira, Spector, Arden, M. L. Ferreira. Acquisition, analysis, or interpretation of data: Chen, Perera, Beckenkamp, Hart, Arden, M. L. Ferreira. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 10/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability Drafting of the manuscript: Chen, Beckenkamp, Spector, Arden. Critical revision of the manuscript for important intellectual content: Chen, Perera, Radojčić, Beckenkamp, P. H. Ferreira, Hart, Arden, M. L. Ferreira. Statistical analysis: Chen, Spector. Obtained funding: Hart. Administrative, technical, or material support: Chen, Perera, Hart. Supervision: Beckenkamp, P. H. Ferreira, Arden, M. L. Ferreira. Conflict of Interest Disclosures: Dr Spector reported serving as a scientific consultant for Zoe Global Ltd. Dr Arden reported receiving personal fees from Pfizer/Lilly and Bristows LLP and grants from Merck outside the submitted work. No other disclosures were reported. Funding/Support: Arthritis Research UK (now Versus Arthritis) and the Oxford NIHR Musculoskeletal Biomedical Research Unit provided funding support to the study. Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. Additional Contributions: We would like to thank all the participants of the Chingford 1000 Women Study; Alan Hakim, MA, Rheumatology, Platinum Medical Centre, The Wellington Hospital, research coordination; Maxine Daniels, BSc(Hons), University of Oxford, research coordination; and Alison Turner, BSc, NIHR Musculoskeletal Biomedical Research Unit, Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, development of data management system, for their time and dedication. We would like to thank Julie Damnjanović, BSc, Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, for her valued support to the team and the study data management. None of the individuals named here were compensated for their contributions. REFERENCES 1. GBD 2017 Disease and Injury Incidence and Prevalence Collaborators. Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet. 2018;392(10159):1789-1858. doi:10. 1016/S0140-6736(18)32279-7 2. Hartvigsen J, Hancock MJ, Kongsted A, et al; Lancet Low Back Pain Series Working Group. What low back pain is and why we need to pay attention. Lancet. 2018;391(10137):2356-2367. doi:10.1016/S0140-6736(18)30480-X 3. Jordan KP, Kadam UT, Hayward R, Porcheret M, Young C, Croft P. Annual consultation prevalence of regional musculoskeletal problems in primary care: an observational study. BMC Musculoskelet Disord. 2010;11:144. doi:10. 1186/1471-2474-11-144 4. Chou R, Qaseem A, Owens DK, Shekelle P; Clinical Guidelines Committee of the American College of Physicians. Diagnostic imaging for low back pain: advice for high-value health care from the American College of Physicians. Ann Intern Med. 2011;154(3):181-189. doi:10.7326/0003-4819-154-3-201102010-00008 5. Foster NE, Anema JR, Cherkin D, et al; Lancet Low Back Pain Series Working Group. Prevention and treatment of low back pain: evidence, challenges, and promising directions. Lancet. 2018;391(10137):2368-2383. doi:10. 1016/S0140-6736(18)30489-6 6. National Institute of Health and Care Excellence (NICE); National Guideline Centre. Low Back Pain and Sciatica in Over 16s: Assessment And Management. NICE Guideline No. 59. National Institute for Health and Care Excellence; 2016. 7. Downie A, Hancock M, Jenkins H, et al. How common is imaging for low back pain in primary and emergency care? systematic review and meta-analysis of over 4 million imaging requests across 21 years. Br J Sports Med. 2020;54(11):642-651. doi:10.1136/bjsports-2018-100087 8. Wall BF, Kendall GM, Edwards AA, Bouffler S, Muirhead CR, Meara JR. What are the risks from medical X-rays and other low dose radiation? Br J Radiol. 2006;79(940):285-294. doi:10.1259/bjr/55733882 9. Chou R, Fu R, Carrino JA, Deyo RA. Imaging strategies for low-back pain: systematic review and meta-analysis. Lancet. 2009;373(9662):463-472. doi:10.1016/S0140-6736(09)60172-0 10. Rhodes LA, McPhillips-Tangum CA, Markham C, Klenk R. The power of the visible: the meaning of diagnostic tests in chronic back pain. . 1999;48(9):1189-1203. doi:10.1016/s0277-9536(98)00418-3 11. Schers H, Wensing M, Huijsmans Z, van Tulder M, Grol R. Implementation barriers for general practice guidelines on low back pain: a qualitative study. Spine (Phila Pa 1976). 2001;26(15):E348-E353. doi:10.1097/ 00007632-200108010-00013 JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 11/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability 12. Taylor S, Bishop A. Patient and public beliefs about the role of imaging in the management of non-specific low back pain: a scoping review. Physiotherapy. 2020;107:224-233. doi:10.1016/j.physio.2019.08.014 13. Jarvik JG, Gold LS, Comstock BA, et al. Association of early imaging for back pain with clinical outcomes in older adults. JAMA. 2015;313(11):1143-1153. doi:10.1001/jama.2015.1871 14. Kellgren JH, Lawrence JS. Radiological assessment of osteo-arthrosis. Ann Rheum Dis. 1957;16(4):494-502. doi:10.1136/ard.16.4.494 15. Lee SY, Cho NH, Jung YO, Seo YI, Kim HA. Prevalence and risk factors for lumbar spondylosis and its association with low back pain among rural Korean residents. J Korean Neurosurg Soc. 2017;60(1):67-74. doi:10.3340/jkns. 2016.0505.007 16. Perera RS, Dissanayake PH, Senarath U, Wijayaratne LS, Karunanayake AL, Dissanayake VHW. Associations between disc space narrowing, anterior osteophytes and disability in chronic mechanical low back pain: a cross sectional study. BMC Musculoskelet Disord. 2017;18(1):193. doi:10.1186/s12891-017-1562-9 17. Bartlett SJ, Ling SM, Mayo NE, Scott SC, Bingham CO III. Identifying common trajectories of joint space narrowing over two years in knee osteoarthritis. Arthritis Care Res (Hoboken). 2011;63(12):1722-1728. doi:10.1002/ acr.20614 18. Felson DT, Gale DR, Elon Gale M, et al. Osteophytes and progression of knee osteoarthritis. Rheumatology (Oxford). 2005;44(1):100-104. doi:10.1093/rheumatology/keh411 19. Vandenbroucke JP, von Elm E, Altman DG, et al; STROBE initiative. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. Ann Intern Med. 2007;147(8): W163-94. doi:10.7326/0003-4819-147-8-200710160-00010-w1 20. von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP; STROBE Initiative. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann Intern Med. 2007;147(8):573-577. doi:10.7326/0003-4819-147-8- 200710160-00010 21. Hassett G, Hart DJ, Manek NJ, Doyle DV, Spector TD. Risk factors for progression of lumbar spine disc degeneration: the Chingford Study. Arthritis Rheum. 2003;48(11):3112-3117. doi:10.1002/art.11321 22. Lane NE, Nevitt MC, Genant HK, Hochberg MC. Reliability of new indices of radiographic osteoarthritis of the hand and hip and lumbar disc degeneration. J Rheumatol. 1993;20(11):1911-1918. 23. Co YY, Eaton S, Maxwell MW. The relationship between the St. Thomas and Oswestry Disability Scores and the severity of low back pain. J Manipulative Physiol Ther. 1993;16(1):14-18. 24. DAGitty.net. DAGitty—draw and analyze causal diagrams. Accessed April 7, 2021. http://www.dagitty.net/ 25. Textor J, van der Zander B, Gilthorpe MS, Liśkiewicz M, Ellison GT. Robust causal inference using directed acyclic graphs: the R package “dagitty”. Int J Epidemiol. 2016;45(6):1887-1894. 26. Harrell Jr FE. Regression Modeling Strategies: With Applications to Linear Models, Logistic and Ordinal Regression, and Survival Analysis. Springer; 2015. doi:10.1007/978-3-319-19425-7 27. Brookhart MA, Schneeweiss S, Rothman KJ, Glynn RJ, Avorn J, Stürmer T. Variable selection for propensity score models. Am J Epidemiol. 2006;163(12):1149-1156. doi:10.1093/aje/kwj149 28. Rothman KJ, Greenland S, Lash TL. Modern Epidemiology. Lippincott Williams & Wilkins; 2008. 29. Harrell Jr FE. Package ‘Hmisc.’ CRAN2018. Accessed July 22, 2020. https://cran.uib.no/web/packages/Hmisc/ Hmisc.pdf 30. Al-Rawahi M, Luo J, Pollintine P, Dolan P, Adams MA. Mechanical function of vertebral body osteophytes, as revealed by experiments on cadaveric spines. Spine (Phila Pa 1976). 2011;36(10):770-777. doi:10.1097/BRS. 0b013e3181df1a70 31. Brinjikji W, Luetmer PH, Comstock B, et al. Systematic literature review of imaging features of spinal degeneration in asymptomatic populations. AJNR Am J Neuroradiol. 2015;36(4):811-816. doi:10.3174/ajnr.A4173 SUPPLEMENT. eFigure 1. Causal Diagram eFigure 2. Histogram of Outcome Distribution eFigure 3. Missing Data Pattern eAppendix 1. Methods for Exploratory and Sensitivity Analyses eAppendix 2. Details of Statistical Methods eTable 1. Distribution of Lumbar Spine Radiographic Changes at Each Lumbar Spine Segment eTable 2. Redundancy Analysis of Exposures eTable 3. Interaction with Age, BMI, or Smoking Status JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 12/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability eAppendix 3. Sensitivity Analyses Results: Change on the Cut-Off Points of Exposures eTable 4. Kellgren-Lawrence Grade–Based Score eTable 5. Osteophytes Grade–Based Score eTable 6. Disc Space Narrowing Grade–Based Score eAppendix 4. Sensitivity Analyses Results: Total Original Score of Exposures eTable 7. Kellgren-Lawrence Grade eTable 8. Osteophytes Grade eTable 9. Disc Space Narrowing Grade eAppendix 5. Sensitivity Analyses Results: Disease Severity eTable 10. Kellgren-Lawrence Grade–Based Score eTable 11. Osteophytes Grade–Based Score eTable 12. Disc Space Narrowing Grade–Based Score eAppendix 6. Sensitivity Analyses Results: Restricting to Women With Back Pain eTable 13. Kellgren-Lawrence Grade–Based Score eTable 14. Osteophytes Grade–Based Score eTable 15. Disc Space Narrowing Grade–Based Score eAppendix 7. Sensitivity Analyses Results: Changing the Model to cloglog Link Function eTable 16. Kellgren-Lawrence Grade–Based Score eTable 17. Osteophytes Grade–Based Score eTable 18. Disc Space Narrowing Grade–Based Score eAppendix 8. Sensitivity Analyses Results: Change the Model to Linear Regression eTable 19. Kellgren-Lawrence Grade–Based Score eTable 20. Osteophytes Grade–Based Score eTable 21. Disc Space Narrowing Grade–Based Score eTable 22. Additionally Adjusted for Pain Medication and Depression eTable 23. Kellgren-Lawrence Grade–Based Score eTable 24. Osteophytes Grade–Based Score eTable 25. Disc Space Narrowing Grade–Based Score eTable 26. E-Value JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 13/13 Supplementary Online Content Chen L, Perera RS, MR, et al. Association of lumbar spine radiographic changes with severity of back pain related disability among middle-aged, community- dwelling women. JAMA Netw Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 eFigure 1. Causal Diagram eFigure 2. Histogram of Outcome Distribution eFigure 3. Missing Data Pattern eAppendix 1. Methods for Exploratory and Sensitivity Analyses eAppendix 2. Details of Statistical Methods eTable 1. Distribution of Lumbar Spine Radiographic Changes at Each Lumbar Spine Segment eTable 2. Redundancy Analysis of Exposures eTable 3. Interaction with Age, BMI, or Smoking Status eAppendix 3. Sensitivity Analyses Results: Change on the Cut-Off Points of Exposures eTable 4. Kellgren-Lawrence Grade Based Score eTable 5. Osteophytes Grade Based Score eTable 6. Disc Space Narrowing Grade Based Score eAppendix 4. Sensitivity Analyses Results: Total Original Score of Exposures eTable 7. Kellgren-Lawrence Grade eTable 8. Osteophytes Grade eTable 9. Disc Space Narrowing Grade eAppendix 5. Sensitivity Analyses Results: Disease Severity eTable 10. Kellgren-Lawrence Grade Based Score eTable 11. Osteophytes Grade Based Score eTable 12. Disc Space Narrowing Grade Based Score eAppendix 6. Sensitivity Analyses Results: Restricting to Women With Back Pain eTable 13. Kellgren-Lawrence Grade Based Score eTable 14. Osteophytes Grade Based Score eTable 15. Disc Space Narrowing Grade Based Score eAppendix 7. Sensitivity Analyses Results: Changing the Model to cloglog Link Function eTable 16. Kellgren-Lawrence Grade Based Score eTable 17. Osteophytes Grade Based Score eTable 18. Disc Space Narrowing Grade Based Score eAppendix 8. Sensitivity Analyses Results: Change the Model to Linear Regression eTable 19. Kellgren-Lawrence Grade Based Score eTable 20. Osteophytes Grade Based Score eTable 21. Disc Space Narrowing Grade Based Score eTable 22. Additionally Adjusted for Pain Medication and Depression eTable 23. Kellgren-Lawrence Grade Based Score eTable 24. Osteophytes Grade Based Score eTable 25. Disc Space Narrowing Grade Based Score eTable 26. E-Value © 2021 Chen L et al. JAMA Network Open. This supplementary material has been provided by the authors to give readers additional information about their work. © 2021 Chen L et al. JAMA Network Open. eFigure 1. Causal Diagram © 2021 Chen L et al. JAMA Network Open. DAGitty is a browser-based environment for creating, editing, and analyzing causal diagrams (also known as directed acyclic graphs or causal Bayesian networks). The focus is on the use of causal diagrams for minimizing bias in empirical studies in epidemiology and o ther disciplines. We used two steps to choose the final covariates included in the model. At first, we pre-specified a lot of potential confounders based on clinical knowledge. The second step included searching previous literature for well accepted confounders and then drawing the causal diagram to find the minimal sufficient adjustment sets based on the established evidence. Initially selected covariates: age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, blood pressure, serious operations/illnesses, quality of life, history of fall, pain-killer usage status, anticholinergic medication usage status, antidepressant medication usage status, and angiotensin-converting enzyme inhibitors usage status. Minimal sufficient adjustment sets: age (as a continuous variable), BMI (as a continuous variable), smoking status (never, current, and ex - smoker), back pain status (yes or no), bisphosphonates use (yes or no), and physical activity (as three domains: walking, job and spor t; Walking [per week]: <0.5 miles, 0.5-5 miles, 5-10 miles, and 10+ miles. Job: sedentary, sedentary + occasional exercise, 0.5 sedentary + 0.5 active [or active housework, e.g., daily dust/hoover]), and predominantly manual, active all day. Sport: none, 1 hour per week golf, bowls, badminton, cycling or swimming, 2 hours previous or 1 hour keep-fit, aerobics, squash, and 2 hours + keep-fit, aerobics, squash). © 2021 Chen L et al. JAMA Network Open. eFigure 2. Histogram of Outcome Distribution A. Cross-sectional part. B. Longitudinal part. © 2021 Chen L et al. JAMA Network Open. eFigure 3. Missing Data Pattern A. Cross-sectional part. B. Longitudinal part. © 2021 Chen L et al. JAMA Network Open. eAppendix 1. Methods for Exploratory and Sensitivity Analyses Exploratory analyses We examined whether the association between lumbar spine radiographic changes and the severity of back pain-related disability differed by age, BMI or smoking status through testing of multiplicative interactions using WALD statistics. Sensitivity analyses 1) To assess the validity of our cut-off points for defining lumbar spine radiographic changes, we changed the cut-off moderate-to- 2) We changed the composite score by using the total original score, which is a sum score from the original grade score at each segment. For example, the K/L grade based on total original score ranged from 0-16 (at each segment 0-4). 3) We built a new composite score based on the disease severity: for disc space narrowing and osteophyte based score, grade 0 vs grade 2 and 3; for K/L grade based score, grade 0 and 1 vs grade 3 and 4. 4) Considering the potential heterogeneity of the population, we restricted our analyses to women with back pain. 5) Considering the potential model misspecification issue, we changed it to linear model and ordinal logistic regression with cloglog link function. 6) To explore the potential influence from unmeasured confounding, E-value was calculated. 7) As one reviewer suggested, we additionally adjusted pain medication and depression considering they might affect the results as potential strong prognostic factors. © 2021 Chen L et al. JAMA Network Open. eAppendix 2. Details of Statistical Methods Ordinal logistic regression: Liu et al. indicated that ordinal regression models are robust for continuous outcomes, especially when the distributions of continuous responses are skewed (Stat Med. 2017 Nov 30;36(27):4316-4335). Our continuous outcomes were skewed so that we chose ordinal logistic regression. Stepped modelling framework: If all covariates were measured at the same time, two models were sufficient: step 1: unadjusted analyses; step 2: analyses adjusted for age, BMI, back pain status, physical activity, bisphosphonates usage status, and smoking status (additionally adjusted for Year 9 back pain-related disability for the longitudinal analysis). The reason for the inclusion of Year 9 back pain-related disability for the longitudinal analysis is that it is a strong prognostic factor for Year 15 back pain-related disability. As VanderWeele et al. indicated, it is often important to control, whenever possible, for the outcome at or prior to the time of the baseline exposure assessment so that confounding control assumption is as plausible as possible (Statistical Science 35.3 (2020): 437-466.). In our study, among all selected covariates, physical activity was measured at Year 6 and others were measured at Year 9 (baseline at our study). Considering physical activity could change between Year 6 and Year 9, we built three models: step 1: unadjusted analyses; step 2: analyses adjusted for age, BMI, back pain status, bisphosphonates usage status, and smoking status (additionally adjusted for Year 9 back pain-related disability for the longitudinal analysis); step 3: analyses further adjusted for physical activity. Exposure modelling: For each type of radiographic changes, our exposure has five values: 0, 1, 2, 3, and 4. The value equals the number of lumbar spine segments affected by radiographic © 2021 Chen L et al. JAMA Network Open. changes. We could consider the exposure as the unordered categorical variable. In this case, we set 0 as the reference level and obtained the estimate by comparing other values with 0. We could also consider 0-4 as the continuous variable. As a continuous variable, we tested the linear trend by modelling the exposure as continuous variable and reported the P-value (whether the regression coefficient of the exposure variable equalled zero); we also tested the non-linear trend by adding the quartic term to the previous model and reported the P-value (whether the added quartic term could improve performance of the previous model through analysis of variance). Multiple imputation: It is a general approach to handle missing data in epidemiological and clinical research (BMJ 2009;338:b2393). It includes two steps: step 1: to create multiple copies of the dataset with the missing data replaced by imputed values; step 2: to fit the model to each of the imputed datasets and then calculate the final estimate by combing the estimate from each Rubin, Donald B. Multiple imputation for nonresponse in surveys. Vol. 81. John Wiley & Sons, 2004.). © 2021 Chen L et al. JAMA Network Open. eTable 1. Distribution of Lumbar Spine Radiographic Changes at Each Lumbar Spine Segment Cross-section (n=650) Longitudinal (n=443) Kellgren-Lawrence grade L1-L2 grade 0 229 160 grade 1 199 140 grade 2 94 57 grade 3 96 68 grade 4 32 18 L2-L3 grade 0 160 121 grade 1 183 126 grade 2 126 75 grade 3 137 92 grade 4 44 L3-L4 grade 0 102 72 grade 1 202 143 grade 2 146 93 grade 3 171 119 grade 4 29 16 L4-L5 grade 0 135 89 grade 1 230 167 grade 2 88 61 grade 3 116 75 grade 4 81 51 Disc space narrowing L1-L2 grade 0 311 219 grade 1 295 196 grade 2 29 23 grade 3 15 5 L2-L3 grade 0 300 216 grade 1 296 197 grade 2 40 22 grade 3 14 8 L3-L4 grade 0 276 198 grade 1 330 222 grade 2 31 15 grade 3 13 8 L4-L5 grade 0 259 191 grade 1 280 182 © 2021 Chen L et al. JAMA Network Open. grade 2 67 38 grade 3 44 32 Osteophytes L1-L2 grade 0 551 383 grade 1 92 58 grade 2 7 2 grade 3 0 0 L2-L3 grade 0 509 368 grade 1 129 71 grade 2 10 4 grade 3 2 0 L3-L4 grade 0 468 335 grade 1 161 98 grade 2 18 10 grade 3 3 0 L4-L5 grade 0 430 306 grade 1 191 118 grade 2 27 18 grade 3 2 1 © 2021 Chen L et al. JAMA Network Open. eTable 2. Redundancy Analysis of Exposures R-squared with which each variable can be predicted from all other variables. R-squared cut-off: 0.75. R-squared Cross-sectional L1-L2 Osteophytes 0.157 Disc space narrowing 0.662 Kellgren-Lawrence grade 0.644 L2-L3 Osteophytes 0.235 Disc space narrowing 0.649 Kellgren-Lawrence grade 0.633 L3-L4 Osteophytes 0.198 Disc space narrowing 0.659 Kellgren-Lawrence grade 0.647 L4-L5 Osteophytes 0.165 Disc space narrowing 0.721 Kellgren-Lawrence grade 0.717 Longitudinal L1-L2 Osteophytes 0.020 Disc space narrowing 0.574 Kellgren-Lawrence grade 0.574 L2-L3 Osteophytes 0.195 Disc space narrowing 0.616 Kellgren-Lawrence grade 0.603 L3-L4 Osteophytes 0.147 Disc space narrowing 0.657 Kellgren-Lawrence grade 0.657 L4-L5 Osteophytes 0.149 Disc space narrowing 0.725 Kellgren-Lawrence grade 0.727 © 2021 Chen L et al. JAMA Network Open. eTable 3. Interaction with Age, BMI, or Smoking Status Cross-sectional (P-value) Longitudinal (P-value) Kellgren-Lawrence grade based score Age 0.84 0.61 BMI 0.62 0.23 Smoke status 0.18 0.66 Osteophytes grade based score Age 0.61 0.96 BMI 0.45 0.50 Smoke status 0.63 0.62 Disc space narrowing grade based score Age 0.28 0.65 BMI 0.88 0.14 Smoke status 0.41 0.20 © 2021 Chen L et al. JAMA Network Open. eAppendix 3. Sensitivity Analyses Results: Change on the Cut-Off Points of Exposures (corresponding to the first sensitivity analysis; the cut- off points changed -to- ). eTable 4. Kellgren-Lawrence Grade Based Score Variables K/L grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 296 (45.5) 153 (23.5) 99 (15.2) 53 (8.2) 49 (7.5) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.05 (0.70, 1.58) 1.24 (0.78, 1.98) 1.09 (0.60, 1.98) 1.02 (0.53, 1.93) 0.66 0.75 Multivariable adjusted 1 (reference) 1.07 (0.70, 1.64) 1.31 (0.80, 2.13) 1.15 (0.61, 2.16) 1.32 (0.67, 2.62) 0.29 0.78 Further adjusted for 1 (reference) 1.06 (0.69, 1.63) 1.25 (0.76, 2.04) 1.12 (0.60, 2.12) 1.32 (0.66, 2.64) 0.34 0.89 physical activity Longitudinal (Year 15, n=443) Number of women (%) 200 (45.1) 110 (24.8) 68 (15.3) 38 (8.6) 27 (6.1) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.16 (0.71, 1.88) 1.14 (0.64, 2.03) 1.34 (0.68, 2.66) 1.39 (0.62, 3.14) 0.29 0.57 Multivariable adjusted 1 (reference) 1.10 (0.64, 1.88) 1.05 (0.54, 2.02) 1.38 (0.65, 2.95) 1.12 (0.45, 2.77) 0.56 0.87 Further adjusted for 1 (reference) 1.08 (0.62, 1.86) 1.00 (0.51, 1.97) 1.46 (0.66, 3.19) 1.15 (0.46, 2.90) 0.53 0.98 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 5. Osteophytes Grade Based Score Variables Osteophytes grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 596 (91.7) 45 (6.9) 4 (0.6) 4 (0.6) 1 (0.2) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.89 (0.47, 1.69) 0.56 (0.06, 5.02) N/A N/A 0.16 0.45 Multivariable adjusted 1 (reference) 0.98 (0.51, 1.89) 0.53 (0.06, 4.89) N/A N/A 0.31 0.33 Further adjusted for 1 (reference) 1.00 (0.52, 1.97) 0.62 (0.07, 5.99) N/A N/A 0.35 0.33 physical activity Longitudinal (Year 15, n=443) Number of women (%) 411 (92.8) 30 (6.8) 1 (0.2) 1 (0.2) 0 (0.0) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.55 (0.23, 1.28) N/A N/A N/A 0.10 0.37 Multivariable adjusted 1 (reference) 0.62 (0.24, 1.56) N/A N/A N/A 0.18 0.85 Further adjusted for 1 (reference) 0.64 (0.25, 1.62) N/A N/A N/A 0.18 0.88 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 6. Disc Space Narrowing Grade Based Score Variables Disc space narrowing grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 480 (73.8) 114 (17.5) 37 (5.7) 11 (1.7) 8 (1.2) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.39 (0.92, 2.11) 1.01 (0.51, 2.01) 0.75 (0.20, 2.77) 0.33 (0.04, 2.72) 0.87 0.18 Multivariable adjusted 1 (reference) 1.44 (0.93, 2.22) 1.01 (0.49, 2.06) 0.88 (0.23, 3.42) 0.46 (0.05, 3.81) 0.88 0.11 Further adjusted for 1 (reference) 1.45 (0.93, 2.25) 1.00 (0.48, 2.06) 0.95 (0.24, 3.76) 0.46 (0.05, 3.94) 0.84 0.12 physical activity Longitudinal (Year 15, n=443) Number of women (%) 332 (74.9) 82 (18.5) 21 (4.7) 5 (1.1) 3 (0.7) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.35 (0.83, 2.22) 1.75 (0.77, 3.97) 0.45 (0.05, 3.88) N/A 0.64 0.13 Multivariable adjusted 1 (reference) 1.28 (0.73, 2.23) 1.63 (0.67, 3.98) 0.61 (0.06, 6.65) N/A 0.93 0.06 Further adjusted for 1 (reference) 1.41 (0.80, 2.49) 1.68 (0.68, 4.12) 0.67 (0.06, 7.59) N/A 0.72 0.05 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eAppendix 4. Sensitivity Analyses Results: Total Original Score of Exposures (corresponding to the second sensitivity analysis; the total original score which is a sum score from the original grade score at each segment). eTable 7. Kellgren-Lawrence Grade Linear model (effect estimate with its 95% confidence interval) Non-linear model (p-value) Cross-sectional (Year 9, n=650) Unadjusted -0.12 (-0.34, 0.11) 0.26 Multivariable adjusted -0.07 (-0.31, 0.17) 0.19 Further adjusted for physical activity -0.09 (-0.33, 0.16) 0.15 Longitudinal (Year 15, n=443) Unadjusted -0.04 (-0.32, 0.23) 0.11 Multivariable adjusted -0.10 (-0.41, 0.21) 0.25 Further adjusted for physical activity -0.11 (-0.43, 0.20) 0.13 Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 8. Osteophytes Grade Linear model Non-linear model Cross-sectional (Year 9, n=650) Unadjusted -0.12 (-0.26, 0.01) 1.00 Multivariable adjusted -0.10 (-0.25, 0.05) 0.70 Further adjusted for physical activity -0.10 (-0.24, 0.05) 0.71 Longitudinal (Year 15, n=443) Unadjusted -0.25 (-0.44, -0.05) 0.79 Multivariable adjusted -0.26 (-0.48, -0.05) 0.99 Further adjusted for physical activity -0.26 (-0.48, -0.04) 0.90 Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 9. Disc Space Narrowing Grade Linear model Non-linear model Cross-sectional (Year 9, n=650) Unadjusted -0.07 (-0.30, 0.16) 0.06 Multivariable adjusted -0.03 (-0.29, 0.22) 0.07 Further adjusted for physical activity -0.03 (-0.29, 0.22) 0.09 Longitudinal (Year 15, n=443) Unadjusted 0.16 (-0.13, 0.44) 0.98 Multivariable adjusted 0.20 (-0.13, 0.52) 0.94 Further adjusted for physical activity 0.25 (-0.08, 0.59) 1.00 Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eAppendix 5. Sensitivity Analyses Results: Disease Severity (corresponding to the third sensitivity analysis; a new composite score based on the disease severity). eTable 10. Kellgren-Lawrence Grade Based Score Variables K/L grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=365) Number of women (%) 154 (42.2) 77 (21.1) 47 (12.9) 38 (10.4) 49 (13.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.92 (0.52, 1.63) 1.27 (0.66, 2.44) 1.14 (0.56, 2.32) 0.92 (0.47, 1.82) 0.89 0.63 Multivariable adjusted 1 (reference) 1.06 (0.59, 1.92) 1.33 (0.67, 2.65) 1.26 (0.58, 2.70) 1.34 (0.63, 2.84) 0.34 0.83 Further adjusted for 1 (reference) 1.03 (0.56, 1.88) 1.33 (0.66, 2.66) 1.31 (0.60, 2.89) 1.35 (0.63, 2.88) 0.31 0.87 physical activity Longitudinal (Year 15, n=260) Number of women (%) 112 (43.1) 60 (23.1) 34 (13.1) 27 (10.4) 27 (10.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.31 (0.70, 2.48) 1.49 (0.70, 3.17) 1.15 (0.50, 2.67) 1.25 (0.53, 2.91) 0.51 0.42 Multivariable adjusted 1 (reference) 1.32 (0.65, 2.67) 1.25 (0.52, 2.99) 1.37 (0.54, 3.50) 1.12 (0.42, 2.94) 0.61 0.47 Further adjusted for 1 (reference) 1.36 (0.66, 2.82) 1.18 (0.47, 2.92) 1.71 (0.63, 4.67) 1.31 (0.47, 3.65) 0.40 0.57 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 11. Osteophytes Grade Based Score Variables Osteophytes grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=283) Number of women (%) 258 (91.2) 20 (7.1) 1 (0.4) 3 (1.1) 1 (0.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.56 (0.20, 1.56) 2.29 (0.14, 37.06) N/A N/A 0.14 0.71 Multivariable adjusted 1 (reference) 0.49 (0.17, 1.39) 1.02 (0.06, 17.96) N/A N/A 0.13 0.94 Further adjusted for 1 (reference) 0.50 (0.17, 1.46) 0.80 (0.04, 16.41) N/A N/A 0.11 0.88 physical activity Longitudinal (Year 15, n=206) Number of women (%) 192 (93.2) 12 (5.8) 1 (0.5) 1 (0.5) 0 (0.0) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.56 (0.15, 2.09) N/A N/A N/A 0.20 0.89 Multivariable adjusted 1 (reference) 0.87 (0.19, 4.00) N/A N/A N/A 0.49 0.88 Further adjusted for 1 (reference) 1.41 (0.24, 8.33) N/A N/A N/A 0.71 0.87 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 12. Disc Space Narrowing Grade Based Score Variables Disc space narrowing grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=126) Number of women (%) 100 (79.4) 11 (8.7) 5 (4.0) 2 (1.6) 8 (6.3) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.69 (0.46, 6.24) 1.45 (0.26, 8.15) N/A 0.38 (0.04, 3.20) 0.44 0.25 Multivariable adjusted 1 (reference) 1.33 (0.32, 5.50) 2.03 (0.30, 13.90) N/A 0.53 (0.06, 5.14) N/A N/A Further adjusted for 1 (reference) 1.20 (0.26, 5.56) 3.25 (0.37, 28.69) N/A 0.59 (0.05, 6.89) N/A N/A physical activity Longitudinal (Year 15, n=87) Number of women (%) 70 (80.5) 9 (10.3) 4 (4.6) 1 (1.1) 3 (3.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.15 (0.27, 4.99) 0.53 (0.06, 4.98) 1.95 (0.12, 32.49) N/A 0.39 0.47 Multivariable adjusted 1 (reference) 1.76 (0.17, 18.44) 0.54 (0.03, 9.08) 0.02 (0.0001, N/A 0.33 0.19 3.28) Further adjusted for 1 (reference) 0.69 (0.09, 5.38) 0.06 (0.002, 1.38) 8.07 (0.12, 54.29) N/A 0.16 0.89 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eAppendix 6. Sensitivity Analyses Results: Restricting to Women With Back Pain (corresponding to the fourth sensitivity analysis; the potential heterogeneity of the population). eTable 13. Kellgren-Lawrence Grade Based Score Variables K/L grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=210) Number of women (%) 51 (24.3) 46 (21.9) 39 (18.6) 42 (20.0) 32 (15.2) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.60 (0.26, 1.38) 0.73 (0.31, 1.71) 0.66 (0.29, 1.54) 0.53 (0.21, 1.33) 0.23 0.74 Multivariable adjusted 1 (reference) 0.54 (0.15, 1.95) 0.29 (0.07, 1.17) 0.77 (0.21, 2.86) 0.21 (0.04, 1.13) 0.14 0.85 Further adjusted for 1 (reference) 0.74 (0.18, 3.10) 0.23 (0.05, 1.07) 0.76 (0.17, 3.37) 0.24 (0.04, 1.44) 0.14 0.66 physical activity Longitudinal (Year 15, n=145) Number of women (%) 40 (27.6) 32 (22.1) 24 (16.6) 26 (17.9) 23 (15.9) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.05 (0.40, 2.74) 1.16 (0.42, 3.26) 0.43 (0.14, 1.38) 1.43 (0.52, 3.94) 0.93 0.49 Multivariable adjusted 1 (reference) 0.75 (0.24, 2.36) 1.93 (0.60, 6.15) 0.56 (0.15, 2.12) 1.81 (0.54, 6.10) 0.43 0.62 Further adjusted for 1 (reference) 0.53 (0.15, 1.85) 1.57 (0.42, 5.78) 0.41 (0.09, 1.75) 1.92 (0.52, 7.06) 0.53 0.32 physical activity Adjusted for age, BMI, smoke status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 14. Osteophytes Grade Based Score Variables Osteophytes grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=210) Number of women (%) 81 (38.6) 76 (36.2) 32 (15.2) 15 (7.1) 6 (2.9) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.70 (0.36, 1.34) 0.88 (0.38, 2.02) 0.60 (0.18, 2.03) 0.35 (0.04, 3.14) 0.27 0.98 Multivariable adjusted 1 (reference) 1.05 (0.04, 2.76) 1.65 (0.48, 5.62) 0.74 (0.11, 4.83) N/A 0.79 0.24 Further adjusted for 1 (reference) 0.97 (0.35, 2.73) 1.49 (0.41, 5.41) 1.08 (0.16, 7.34) N/A 0.88 0.34 physical activity Longitudinal (Year 15, n=145) Number of women (%) 56 (38.6) 59 (40.7) 20 (13.8) 8 (5.5) 2 (1.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.77 (0.37, 1.63) 0.64 (0.22, 1.87) 0.55 (0.10, 2.88) N/A 0.18 0.74 Multivariable adjusted 1 (reference) 0.91 (0.39, 2.16) 0.68 (0.20, 2.25) 0.67 (0.10, 4.68) N/A 0.39 0.66 Further adjusted for 1 (reference) 0.90 (0.36, 2.25) 0.39 (0.10, 1.51) 0.58 (0.08, 4.29) N/A 0.18 0.65 physical activity Adjusted for age, BMI, smoke status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 15. Disc Space Narrowing Grade Based Score Variables Disc space narrowing grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=210) Number of women (%) 25 (11.9) 27 (12.9) 45 (21.4) 41 (19.5) 72 (34.3) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.94 (0.34, 2.64) 0.40 (0.15, 1.10) 0.55 (0.20, 1.47) 0.39 (0.15, 0.99) 0.03 0.58 Multivariable adjusted 1 (reference) 0.81 (0.18, 3.55) 0.29 (0.06, 1.34) 0.55 (0.10, 2.93) 0.32 (0.07, 1.35) 0.10 0.53 Further adjusted for 1 (reference) 0.43 (0.08, 2.28) 0.23 (0.04, 1.23) 0.39 (0.06, 2.38) 0.26 (0.05, 1.30) 0.19 0.32 physical activity Longitudinal (Year 15, n=145) Number of women (%) 14 (9.7) 27 (18.6) 31 (21.4) 28 (19.3) 45 (31.0) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.50 (0.15, 1.68) 0.22 (0.06, 0.79) 0.25 (0.07, 0.91) 0.83 (0.28, 2.45) 0.87 0.002 Multivariable adjusted 1 (reference) 0.60 (0.15, 2.39) 0.43 (0.09, 2.00) 0.55 (0.12, 2.45) 1.47 (0.38, 5.75) 0.23 0.04 Further adjusted for 1 (reference) 0.55 (0.11, 2.66) 0.29 (0.05, 1.61) 0.53 (0.10, 2.80) 1.28 (0.27, 6.07) 0.25 0.03 physical activity Adjusted for age, BMI, smoke status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eAppendix 7. Sensitivity Analyses Results: Changing the Model to cloglog Link Function (corresponding to the fifth sensitivity analysis; the potential model misspecification). eTable 16. Kellgren-Lawrence Grade Based Score Variables K/L grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 154 (23.7) 142 (21.8) 140 (21.5) 118 (18.2) 96 (14.8) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.16 (0.79, 1.70) 0.86 (0.57, 1.30) 0.89 (0.58, 1.37) 0.85 (0.53, 1.34) 0.24 0.82 Multivariable adjusted 1 (reference) 1.20 (0.81, 1.77) 0.86 (0.57, 1.30) 0.93 (0.60, 1.44) 0.91 (0.56, 1.46) 0.36 0.93 Further adjusted for 1 (reference) 1.18 (0.80, 1.75) 0.84 (0.55, 1.28) 0.89 (0.57, 1.39) 0.89 (0.55, 1.44) 0.30 0.99 physical activity Longitudinal (Year 15, n=443) Number of women (%) 112 (25.3) 100 (22.6) 97 (21.9) 76 (17.2) 58 (13.1) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.18 (0.75, 1.87) 0.95 (0.59, 1.55) 0.81 (0.48, 1.40) 1.07 (0.62, 1.84) 0.63 0.90 Multivariable adjusted 1 (reference) 0.89 (0.54, 1.46) 0.91 (0.55, 1.53) 0.74 (0.42, 1.31) 0.84 (0.47, 1.52) 0.28 0.60 Further adjusted for 1 (reference) 0.86 (0.52, 1.43) 0.89 (0.53, 1.49) 0.72 (0.40, 1.28) 0.83 (0.45, 1.50) 0.25 0.54 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 17. Osteophytes Grade Based Score Variables Osteophytes grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 258 (39.7) 226 (34.8) 102 (15.7) 44 (6.8) 20 (3.1) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.84 (0.62, 1.15) 0.84 (0.56, 1.26) 0.59 (0.31, 1.13) 0.83 (0.36, 1.90) 0.12 0.64 Multivariable adjusted 1 (reference) 0.87 (0.63, 1.19) 0.83 (0.55, 1.27) 0.66 (0.34, 1.30) 0.99 (0.42, 2.32) 0.29 0.50 Further adjusted for 1 (reference) 0.86 (0.63, 1.19) 0.85 (0.56, 1.31) 0.68 (0.35, 1.34) 0.97 (0.41, 2.30) 0.32 0.55 physical activity Longitudinal (Year 15, n=443) Number of women (%) 192 (43.3) 157 (35.4) 67 (15.1) 19 (4.3) 8 (1.8) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.77 (0.53, 1.11) 0.68 (0.41, 1.13) 0.49 (0.18, 1.34) 0.58 (0.14, 2.37) 0.04 0.71 Multivariable adjusted 1 (reference) 0.91 (0.62, 1.33) 0.68 (0.40, 1.16) 0.52 (0.18, 1.45) 0.48 (0.11, 2.06) 0.03 0.97 Further adjusted for 1 (reference) 0.94 (0.63, 1.39) 0.63 (0.36, 1.10) 0.55 (0.19, 1.57) 0.54 (0.12, 2.33) 0.03 0.93 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 18. Disc Space Narrowing Grade Based Score Variables Disc space narrowing grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 100 (15.4) 107 (16.5) 147 (22.6) 131 (20.2) 165 (25.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.28 (0.79, 2.10) 1.26 (0.79, 1.99) 1.31 (0.82, 2.10) 0.99 (0.62, 1.58) 0.80 0.10 Multivariable adjusted 1 (reference) 1.36 (0.83, 2.25) 1.49 (0.92, 2.39) 1.42 (0.88, 2.30) 1.10 (0.67, 1.81) 0.87 0.04 Further adjusted for 1 (reference) 1.33 (0.80, 2.21) 1.47 (0.91, 2.38) 1.40 (0.86, 2.27) 1.06 (0.65, 1.75) 0.98 0.03 physical activity Longitudinal (Year 15, n=443) Number of women (%) 70 (15.8) 84 (19.0) 102 (23.0) 88 (19.9) 99 (22.3) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.90 (0.52, 1.57) 0.79 (0.46, 1.36) 1.01 (0.59, 1.73) 1.15 (0.69, 1.93) 0.42 0.21 Multivariable adjusted 1 (reference) 0.86 (0.48, 1.54) 0.89 (0.50, 1.59) 1.08 (0.61, 1.90) 1.37 (0.78, 2.39) 0.13 0.23 Further adjusted for 1 (reference) 0.86 (0.46, 1.59) 0.84 (0.46, 1.54) 1.16 (0.64, 2.10) 1.51 (0.84, 2.73) 0.08 0.14 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eAppendix 8. Sensitivity Analyses Results: Change the Model to Linear Regression (corresponding to the fifth sensitivity analysis; the potential model misspecification). eTable 19. Kellgren-Lawrence Grade Based Score Variables K/L grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 154 (23.7) 142 (21.8) 140 (21.5) 118 (18.2) 96 (14.8) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.42 (0.59, 3.40) 0.86 (0.36, 2.06) 0.99 (0.40, 2.50) 0.65 (0.25, 1.74) 0.31 0.49 Multivariable adjusted 1 (reference) 1.44 (0.61, 3.42) 0.88 (0.37, 2.09) 1.01 (0.40, 2.55) 0.73 (0.27, 1.99) 0.43 0.55 Further adjusted for 1 (reference) 1.40 (0.59, 3.36) 0.81 (0.34, 1.94) 0.96 (0.38, 2.42) 0.69 (0.25, 1.88) 0.35 0.61 physical activity Longitudinal (Year 15, n=443) Number of women (%) 112 (25.3) 100 (22.6) 97 (21.9) 76 (17.2) 58 (13.1) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.41 (0.48, 4.09) 0.91 (0.31, 2.68) 0.56 (0.18, 1.77) 0.89 (0.25, 3.13) 0.38 0.93 Multivariable adjusted 1 (reference) 0.93 (0.36, 2.44) 1.01 (0.38, 2.67) 0.60 (0.21, 1.74) 0.54 (0.17, 1.75) 0.14 0.89 Further adjusted for 1 (reference) 0.92 (0.35, 2.43) 1.00 (0.38, 2.68) 0.59 (0.20, 1.72) 0.54 (0.16, 1.74) 0.13 0.87 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 20. Osteophytes Grade Based Score Variables Osteophytes grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 258 (39.7) 226 (34.8) 102 (15.7) 44 (6.8) 20 (3.1) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.64 (0.32, 1.27) 0.64 (0.26, 1.54) 0.35 (0.26, 1.20) 0.68 (0.12, 3.90) 0.11 0.47 Multivariable adjusted 1 (reference) 0.68 (0.35, 1.34) 0.61 (0.25, 1.48) 0.42 (0.12, 1.46) 0.92 (0.16, 5.43) 0.22 0.36 Further adjusted for 1 (reference) 0.67 (0.34, 1.32) 0.63 (0.26, 1.55) 0.43 (0.12, 1.52) 0.84 (0.14, 5.01) 0.21 0.41 physical activity Longitudinal (Year 15, n=443) Number of women (%) 192 (43.3) 157 (35.4) 67 (15.1) 19 (4.3) 8 (1.8) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.46 (0.20, 1.05) 0.28 (0.09, 0.83) 0.20 (0.03, 1.25) 0.18 (0.01, 2.96) 0.004 0.47 Multivariable adjusted 1 (reference) 0.58 (0.28, 1.24) 0.27 (0.10, 0.74) 0.32 (0.06, 1.74) 0.12 (0.01, 1.41) 0.002 0.61 Further adjusted for 1 (reference) 0.59 (0.27, 1.26) 0.27 (0.10, 0.76) 0.32 (0.06, 1.76) 0.13 (0.01, 1.63) 0.002 0.57 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 21. Disc Space Narrowing Grade Based Score Variables Disc space narrowing grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 100 (15.4) 107 (16.5) 147 (22.6) 131 (20.2) 165 (25.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.59 (0.56, 4.52) 1.68 (0.63, 4.45) 1.43 (0.53, 3.90) 0.92 (0.35, 2.39) 0.62 0.11 Multivariable adjusted 1 (reference) 1.70 (0.60, 4.82) 2.07 (0.78, 5.46) 1.52 (0.56, 4.12) 0.98 (0.36, 2.63) 0.69 0.05 Further adjusted for 1 (reference) 1.61 (0.56, 4.61) 2.00 (0.75, 5.34) 1.50 (0.55, 4.10) 0.94 (0.35, 2.56) 0.65 0.05 physical activity Longitudinal (Year 15, n=443) Number of women (%) 70 (15.8) 84 (19.0) 102 (23.0) 88 (19.9) 99 (22.3) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.96 (0.27, 3.36) 0.75 (0.23, 2.51) 1.55 (0.45, 5.36) 1.83 (0.55, 6.15) 0.18 0.36 Multivariable adjusted 1 (reference) 0.64 (0.20, 2.01) 0.67 (0.22, 2.03) 1.22 (0.39, 3.77) 1.35 (0.43, 4.31) 0.14 0.22 Further adjusted for 1 (reference) 0.60 (0.19, 1.95) 0.63 (0.20, 1.92) 1.19 (0.38, 3.72) 1.39 (0.43, 4.49) 0.14 0.18 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 22. Additionally Adjusted for Pain Medication and Depression (corresponding to the seventh sensitivity analysis; potential strong prognostic factors). Cross-sectional (n=650) 0 segment (n=154) 1 segment (n=142) 2 segments (n=140) 3 segments (n=118) 4 segments (n=96) Whole (n=650) Pain Medication Yes 10 (6.5) 7 (4.9) 16 (11.4) 8 (6.8) 10 (10.4) 51 (7.8) No 68 (44.2) 60 (42.3) 59 (42.1) 53 (44.9) 37 (38.5) 277 (42.6) Missing 76 (49.4) 75 (52.8) 65 (46.4) 57 (48.3) 49 (51.1) 322 (49.5) Depression Yes 0 (0.0) 1 (0.7) 0 (0.0) 1 (0.8) 1 (1.0) 3 (0.5) No 154 (100.0) 141 (99.3) 140 (100.0) 117 (99.2) 95 (99.0) 647 (99.5) Longitudinal (n=443) 0 segment (n=112) 1 segment (n=100) 2 segments (n=97) 3 segments (n=76) 4 segments (n=58) Whole (n=443) Pain Medication Yes 10 (8.9) 6 (6.0) 9 (9.3) 4 (5.3) 6 (10.3) 35 (7.9) No 42 (37.5) 42 (42.0) 48 (49.5) 35 (46.1) 24 (41.4) 191 (43.1) © 2021 Chen L et al. JAMA Network Open. Missing 60 (53.6) 52 (52.0) 40 (41.2) 37 (48.7) 28 (48.3) 217 (49.0) Depression Yes 0 (0.0) 1 (1.0) 1 (1.0) 1 (1.3) 1 (1.7) 3 (0.7) No 112 (100.0) 99 (99.0) 96 (99.0) 75 (98.7) 57 (98.3) 443 (99.3) The columns except the first correspond to the number of segments of Lumbar Spine Radiographic Changes (Kellgren-Lawrence grade based). Data are present as number (percentage) of participants unless otherwise indicated. Women reported current medication use in an open field question within the medical history questionnaire. Data on use of non-opioid and opioid analgesics, defined based on Anatomical Therapeutic Chemical codes M01 and N02, were extracted from this question from Year 9. The details are: With specific name With opioid involved Dihydrocodeine; Dextromoramide; Tramadol; Codeine; Morphine; Paracetamol and dextropropoxyphene; Paracetamol and Codeine; Paracetamol and dihydrocodeine Without opioid involved Indomethacin; Ibuprofen; Diclofenac; Etodolac; Fenbufen; Flurbiprofen; Fenoprofen; Mefenamic acid; Naproxen; Piroxicam; Ketoprofen; Movelat; Glucosamine; Feverfew; Paracetamol Without specific name NSAID; Anti-inflammatory; Analgesics; Painkillers Depression was defined by text response. From Year 1 to Year 4, women were asked the question: Serious operations/illnesses: Other? From Year 8 to Year 9, women were asked the question: Any major illnesses or operations? If the participant reported depression in at least one year (Year 1 to 9, our baseline is Year 9), we defined the value of this covariate as yes. © 2021 Chen L et al. JAMA Network Open. eTable 23. Kellgren-Lawrence Grade Based Score Variables K/L grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 154 (23.7) 142 (21.8) 140 (21.5) 118 (18.2) 96 (14.8) Odds ratio (95% confidence interval) Additional adjusted 1 (reference) 1.21 (0.61, 2.42) 0.85 (0.42, 1.75) 1.25 (0.60, 2.63) 0.87 (0.38, 2.02) 0.84 0.76 Longitudinal (Year 15, n=443) Number of women (%) 112 (25.3) 100 (22.6) 97 (21.9) 76 (17.2) 58 (13.1) Odds ratio (95% confidence interval) Additional adjusted 1 (reference) 1.10 (0.59, 2.04) 0.92 (0.49, 1.74) 0.66 (0.33, 1.33) 0.78 (0.37, 1.62) 0.14 0.86 Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, pain medication and depression. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, pain medication, depression and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 24. Osteophytes Grade Based Score Variables Osteophyte grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 154 (23.7) 142 (21.8) 140 (21.5) 118 (18.2) 96 (14.8) Odds ratio (95% confidence interval) Additional adjusted 1 (reference) 1.17 (0.67, 2.05) 1.25 (0.63, 2.48) 0.66 (0.22, 1.99) 2.46 (0.70, 8.56) 0.52 0.86 Longitudinal (Year 15, n=443) Number of women (%) 112 (25.3) 100 (22.6) 97 (21.9) 76 (17.2) 58 (13.1) Odds ratio (95% confidence interval) Additional adjusted 1 (reference) 0.87 (0.44, 1.73) 0.54 (0.22, 1.32) 0.29 (0.05, 1.63) 0.38 (0.06, 2.40) 0.04 0.87 Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, pain medication and depression. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, pain medication, depression and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 25. Disc Space Narrowing Grade Based Score Variables Disc space narrowing grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 154 (23.7) 142 (21.8) 140 (21.5) 118 (18.2) 96 (14.8) Odds ratio (95% confidence interval) Additional adjusted 1 (reference) 1.42 (0.58, 3.45) 1.40 (0.59, 3.32) 1.48 (0.62, 3.54) 1.20 (0.50, 2.86) 0.86 0.34 Longitudinal (Year 15, n=443) Number of women (%) 112 (25.3) 100 (22.6) 97 (21.9) 76 (17.2) 58 (13.1) Odds ratio (95% confidence interval) Additional adjusted 1 (reference) 0.62 (0.21, 1.83) 0.45 (0.15, 1.35) 0.87 (0.29, 2.55) 1.07 (0.38, 3.02) 0.49 0.10 Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, pain medication and depression. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, pain medication, depression and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 26. E-value (corresponding to the sixth sensitivity analysis; potential influence from unmeasured confounding). Variable Compared with 0 segment 1 segment 2 segments 3 segments 4 segments Cross-sectional Kellgren-Lawrence grade based score Odds ratio (95%CI) 1.22 (0.76, 1.96) 0.84 (0.51, 1.38) 0.92 (0.54,1.56) 0.89 (0.50, 1.57) E-value (lower Bound) 1.44 (1) 1.41 (1) 1.25 (1) 1.31 (1) Osteophytes grade based score Odds ratio (95%CI) 0.83 (0.57, 1.22) 0.78 (0.47, 1.30) 0.58 (0.27, 1.26) 1.03 (0.37, 2.85) E-value (lower Bound) 1.43 (1) 1.52 (1) 1.95 (1) 1.14 (1) Disc space narrowing grade based score Odds ratio (95%CI) 1.43 (0.78, 2.61) 1.56 (0.88, 2.76) 1.44 (0.81, 2.57) 1.07 (0.60, 1.92) E-value (lower Bound) 1.68 (1) 1.81 (1) 1.69 (1) 1.22 (1) Longitudinal Kellgren-Lawrence grade based score Odds ratio (95%CI) 1.06 (0.57, 1.96) 0.94 (0.50, 1.76) 0.69 (0.34, 1.38) 0.83 (0.40, 1.72) E-value (lower Bound) 1.20 (1) 1.21 (1) 1.70 (1) 1.43 (1) Osteophytes grade based score Odds ratio (95%CI) 0.76 (0.47, 1.24) 0.53 (0.28, 1.02) 0.49 (0.14, 1.70) 0.31 (0.06, 1.72) E-value (lower Bound) 1.56 (1) 2.09 (1.21) 2.21 (1) 2.99 (1) Disc space narrowing grade based score Odds ratio (95%CI) 0.72 (0.34, 1.53) 0.74 (0.36, 1.52) 1.06 (0.52, 2.20) 1.26 (0.62, 2.57) E-value (lower Bound) 1.64 (1) 1.60 (1) 1.20 (1) 1.42 (1) We calculated E-value through Online Calculator (https://mmathur.shinyapps.io/evalue/) based on results from step 2 of the stepped modelling framework. Explanation: for an unmeasured confounder to explain the OR estimate of 1.22, the unmeasured confounder would have to be associated with both the exposure and the outcome by 1.44-fold above and beyond the measured confounders. © 2021 Chen L et al. JAMA Network Open. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png JAMA Network Open American Medical Association

Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability Among Middle-aged, Community-Dwelling Women

Loading next page...
 
/lp/american-medical-association/association-of-lumbar-spine-radiographic-changes-with-severity-of-back-C77oAN8w3n
Publisher
American Medical Association
Copyright
Copyright 2021 Chen L et al. JAMA Network Open.
eISSN
2574-3805
DOI
10.1001/jamanetworkopen.2021.10715
Publisher site
See Article on Publisher Site

Abstract

Key Points Question Are lumbar spine IMPORTANCE Previous studies, using mostly cross-sectional data, provide conflicting evidence of radiographic changes associated with an association between lumbar spine radiographic changes and the severity of back pain–related severity of back pain–related disability disability. Such conflicting evidence may be associated with widely unnecessary diagnostic imaging among middle-aged, community- of the lumbar spine. dwelling women? Findings In this population-based OBJECTIVE To examine both cross-sectional and longitudinal associations between lumbar spine cohort study of women from the UK, radiographic changes and the severity of back pain–related disability among middle-aged, community- there was no evidence to support an dwelling women. association between a higher number of lumbar segments with radiographic DESIGN, SETTING, AND PARTICIPANTS This population-based prospective cohort study used data changes (Kellgren-Lawrence grade, from the Chingford 1000 Women Study. Analyses included data collected from year 6 (1994-1996; osteophytes, and disc space narrowing) physical activity was measured), year 9 (1997-1999; treated as baseline), and year 15 (2003-2005), and more severe back pain–related with a total length of follow-up for longitudinal analyses of 6 years. Data were analyzed from April 17 disability in cross-sectional (650 to November 3, 2020. women) or longitudinal (443 women) analyses. EXPOSURES Primary exposure was lumbar spine radiographic changes, defined using the Kellgren- Lawrence (K-L) grade. Secondary exposures were defined using presence of osteophytes and disc Meaning The findings suggest that the space narrowing. The composite score combined the number of lumbar spine segments with definite changes detected on lumbar changes detected on radiographic images (ie, radiographic changes) (K-L grade2, which means at radiographs provide limited value for least definite osteophyte and possible narrowing of disc space are present; osteophyte and disc decision-making regarding back pain space narrowing grade1, which means at least mild or definite changes are present). management in this population. MAIN OUTCOMES AND MEASURES Self-reported back pain–related disability measured in years 9 Supplemental content and 15 assessed by the St Thomas disability questionnaire. Author affiliations and article information are listed at the end of this article. RESULTS Among 650 women (mean [SD] age, 61.3 [5.9] years) in cross-sectional analyses and 443 women (mean [SD] age, 60.6 [6.0] years) in longitudinal analyses, there was no evidence to support an association between higher number of lumbar segments with radiographic changes (K-L grade, osteophytes, and disc space narrowing) and more severe back pain–related disability (eg, cross- sectional analyses using the K-L grade; 1 segment vs 0 segment: adjusted odds ratio, 1.22 [95% CI, 0.76-1.96]). No interactions were found of an association between lumbar spine radiographic changes and the severity of back pain–specific disability with age, body mass index, or smoking status. CONCLUSIONS AND RELEVANCE In this cohort of middle-aged, community-dwelling women, there was no evidence to support an association between a higher number of lumbar segments with radiographic changes (K-L grade, osteophytes, and disc space narrowing) and more severe back (continued) Open Access. This is an open access article distributed under the terms of the CC-BY License. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 1/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability Abstract (continued) pain–related disability cross-sectionally or over time. These findings provide further evidence against routinely using diagnostic imaging of the lumbar spine. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 Introduction Low back pain (LBP) is a highly prevalent condition in the general population worldwide and has been the leading cause of disability for nearly 3 decades, according to the Global Burden of Disease Study 1,2 2017. Among all musculoskeletal problems, LBP is also the most common reason for patients to seek primary care. Current guidelines for treatment of LBP do not recommend routinely using diagnostic imaging, except when patients either present with severe, progressive neurologic deficits or with signs or symptoms indicative of a serious or specific underlying condition (eg, fracture or 4-6 cancer). Nonetheless, diagnostic imaging is still widely used in clinical practice for LBP, with a recent meta-analysis indicating that more than 15% of patients in primary care and approximately 25% in emergency care receive a referral for simple imaging (mainly radiograph). Moreover, nearly 10% of patients with LBP in primary care and emergency care also receive a referral for complex imaging (mainly computed tomography scan and magnetic resonance imaging). There has also been a 53% relative increase in referrals for complex imaging from 1995 to 2017, with no change observed during that period for the rate of referrals for simple imaging. Unnecessary diagnostic imaging not only wastes limited medical resources but is also associated with poorer health outcomes, such as iatrogenic disease from techniques that use ionizing radiation. In addition, patients who undergo unnecessary diagnostic imaging might be labeled with a pseudodisease, which may be associated with unnecessary subsequent interventions that may have adverse effects. Possible explanations for the unwarranted prevalence of imaging referrals for LBP are (1) the patient’s expectation that imaging results could provide valuable information on the cause and, consequently, the appropriate management of her or his condition and (2) the clinician’s desire to 10-12 reassure the patient of the absence of any underlying pathologic condition. Previous studies have 9,13 confirmed that imaging does not improve clinical outcomes for patients with LBP. However, the definition of normal or abnormal imaging diagnostic findings is still debatable. Currently, the presence of osteophytes and disc space narrowing are the most frequent changes detected on radiographs (hereafter referred to as radiographic changes) that may be indicative of spinal pathologic conditions, and the Kellgren-Lawrence (K-L) grade is the tool commonly used to assess the severity of osteoarthritis. A study including elderly women who lived in rural South Korea showed a positive association of the presence of osteophytes (grade2), disc space narrowing (grade2), and K-L grade (grade2) with the severity of disability, measured by a validated Korean version of the Oswestry Disability Index. However, a study conducted in Sri Lanka including patients with LBP concluded that neither disc space narrowing nor the presence of osteophytes was associated with the severity of disability (also measured with the Modified Oswestry Disability Index). Both studies are cross-sectional and failed to adjust their analyses for important confounders, including smoking status, level of participation in physical activity, and medication use. Past studies have also failed to identify whether the number of affected lumbar segments is associated with the severity of back pain–related disability. Therefore, the role of radiographic findings as a potential prognostic factor of the clinical course of LBP is still unclear and needs to be fully explored in population-based cohort studies. Previous studies of radiographic changes in knee osteoarthritis have indicated that the presence of osteophytes may be used to diagnose the condition and that the presence of joint space narrowing may be used to assess both the diagnosis 17,18 and the progression of osteoarthritis. This is still to be elucidated among patients with LBP. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 2/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability The aim of this study was to examine both cross-sectional and longitudinal associations between lumbar radiographic changes and the severity of back pain–related disability among middle- aged, community-dwelling women using composite scores that combined the number of segments and type of changes in terms of K-L grade, disc space narrowing, and osteophytes. We hypothesized that a higher number of segments with lumbar radiographic changes would be associated with more severe back pain–related disability. Methods Study Design, Data Sources, and Study Population From an age and sex register of a large practice of more than 11 000 patients in Chingford in east London, UK, all 1353 women in the age range of 45 to 64 years were invited to participate in a population study assessing musculoskeletal diseases. A total of 1003 women were examined between 1989 and 1991 (year 1; baseline visit for original cohort); 6 died, 66 had moved away, and 278 refused to participate or did not respond. All the women lived within 8 km (5 miles) of the general practice, and 98% of the women were white. Women from this general practice are similar to the UK general population in terms of weight, height, and body mass index (BMI). In the data analyses, and given their availability, we included data on physical activity collected in year 6 (1994- 1996 [ie, prebaseline]) and imaging data, all other covariates, and the outcome for cross-sectional analyses collected in year 9 (1997-1999 [ie, baseline for our study]). The outcome for longitudinal analyses was obtained in year 15 (2003-2005). We followed the Strengthening the Reporting of 19,20 Observational Studies in Epidemiology (STROBE) reporting guideline. The Waltham Forest and Redbridge local research ethics committee approved the study, and all participants provided written informed consent to participate in the study. Exposures Lateral lumbar spine radiographs at year 9 were taken by 1 radiographer, centered on the L3 vertebra, with the participants in the left lateral recumbent position. A single trained observer (a rheumatologist) blinded to patient identity and chronologic order read all of the radiographs. Within- observer variation was assessed by test-retest analysis of 40 randomly selected radiographs from the study. Good within-observer reproducibility (κ = 0.78-0.89) was found. At each lumbar spine segment (L1-L2, L2-L3, L3-L4, and L4-L5), disc space narrowing and osteophytes (both anterior and posterior) were assessed through the semiquantitative method reported by Lane et al, with grade 0 corresponding to normal, grade 1 to mild narrowing and osteophytes, grade 2 to moderate narrowing and osteophytes, and grade 3 to severe narrowing and osteophytes. The Kellgren- Lawrence (K-L) grade was summarized as grade 0 indicating normal; grade 1 indicating doubtful narrowing of disc space and possible osteophytic lipping; grade 2 indicating definite osteophyte and possible narrowing of disc space; grade 3 indicating moderate multiple osteophytes, definite narrowing of disc space, some sclerosis, and possible deformity of bone contour; and grade 4 indicating large osteophytes, marked narrowing of disc space, severe sclerosis, and definite deformity of bone contour. Considering that the number of lumbar spine segments with radiographic changes detected and the various types of radiographic changes might be associated with the results, we generated 3 composite scores: a K-L grade–based score, an osteophyte grade–based score, and a disc space narrowing grade–based score; at each segment, a binary exposure variable of 1 (K-L grade2, which means at least definite osteophyte and possible narrowing of disc space are present; disc space narrowing and osteophyte grade1, which means at least mild or definite changes are present) vs 0 (K-L grade 0 or 1; disc space narrowing and osteophyte grade 0) was used. The composite score was then calculated as the final L1-L2 score + L2-L3 score + L3-L4 score + L4-L5 score, with values ranging from 0 to 4 (where 0 indicates no lumbar spine segments with radiographic changes detected and 4 indicates 4 lumbar spine segments with radiographic changes detected). The K-L grade–based JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 3/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability score was defined as the primary exposure. Osteophyte and disc space narrowing grade–based scores were set as secondary exposures. Outcomes Back pain–related disability was assessed at year 9 and year 15 using a back pain questionnaire (St Thomas disability questionnaire), which correlated well with the Oswestry Disability Questionnaire (r =0.77; P < .001). The outcome was defined by questions at 2 levels. At the first level, women were asked whether they had any back pain for at least 1 day at any time in the last 12 months. At the second level, those who answered yes to the first-level question were asked 8 questions related to the disability due to back pain (corresponding to the previous year’s status): walking around the house; standing for 15 minutes; getting up from a low chair; getting out of a bath; getting in and out of a car; going up and down stairs; putting on socks, stockings, or tights; and cutting toenails. Each question was summarized as grade 0 indicating no difficulty, grade 1 indicating difficult but possible, and grade 2 indicating impossible. We built a composite score based on the aforementioned 8 questions; values ranged from 0 to 16, with higher values corresponding to more severe disability. We assumed the composite score as 0 if women answered “no” to the first-level question. In case of missing data for any of the 8 questions, we kept the data if the women responded to at least 6 questions and calculated the composite score as [(total score)/(number of questions answered)] × 8. Covariates Causal diagram through DAGitty, version 3.0 was used to choose the minimal sufficient adjustment sets for estimating the total association of the exposure with the outcome. Age, BMI, smoking status, back pain status, bisphosphonate use, and physical activity were included in the final model (details in eFigure 1 in the Supplement). All covariates, except physical activity, which was measured in year 6, were measured in year 9. Statistical Analysis Data were analyzed from April 17 to November 3, 2020. Owing to the skewed distribution of back pain–related disability (eFigure 2 in the Supplement), ordinal logistic regression, which holds a proportional odds assumption, was performed. Considering that physical activity was measured at a different time point (ie, year 6) compared with other covariates (ie, year 9) and with potential measurement error, we established a stepped modeling framework: step 1, unadjusted analyses; step 2, analyses adjusted for age, BMI, back pain status, bisphosphonate use status, and smoking status (additionally adjusted for year 9 back pain–related disability for the longitudinal analysis); and step 3, analyses further adjusted for physical activity. Separate analyses were conducted for cross-sectional and longitudinal data. For the longitudinal analyses, data on lumbar spine radiographic changes collected in year 9 were treated as the exposure, and back pain–related disability data collected in year 15 were treated as the outcome. In addition to the confounders mentioned, data on back pain–related disability collected in year 9 were included in the longitudinal analysis as a strong prognostic factor to adjust. Based on the recommendation from Modern Epidemiology, the exposures were modeled as unordered categorical variables and trend test. The proportion of missing data in each covariate is provided in Table 1. Missing data were handled through multiple imputation, which holds a missing-at-random assumption. The assumption was graphically tested (eFigure 3 in the Supplement). No additional variables were used; all covariates in the minimal sufficient adjustment sets were used. Flexible additive models with 10 imputed data sets were used. We did not impute data for the exposure variables. The relative risk was presented as adjusted proportional odds ratios (ORs) with 95% CIs. Extensive sensitivity analyses were performed (eAppendix 1 in the Supplement). All statistical analyses were performed with rms, Hmisc, and tidyverse packages in R, version 3.6.2 (R Group for Statistical Computing). Details of the statistical methods are provided in eAppendix 2 in the Supplement. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 4/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability Table 1. Baseline Characteristics of Study Participants Participants, No. (%) No. of segments Characteristic 012 3 4 Whole cohort Cross-sectional (n = 650) No. 154 142 140 118 96 650 Age, mean (SD), y 59.7 (5.5) 60.3 (5.7) 60.8 (5.8) 62.8 (5.7) 64.5 (5.8) 61.3 (5.9) Missing 1 (0.6) 2 (1.4) 4 (2.9) 2 (1.7) 1 (1.0) 10 (1.5) Smoking status Never 105 (68.2) 84 (59.2) 89 (63.6) 71 (60.2) 61 (63.5) 410 (63.1) Current 18 (11.7) 25 (17.6) 20 (14.3) 20 (16.9) 15 (15.6) 98 (15.1) Ex-smoker 28 (18.2) 31 (21.8) 25 (17.9) 23 (19.5) 18 (18.8) 125 (19.2) Missing 3 (1.9) 2 (1.4) 6 (4.3) 4 (3.4) 2 (2.1) 17 (2.6) BMI, mean (SD) 26.6 (4.2) 26.8 (5.1) 27.2 (5.0) 26.9 (4.4) 27.6 (5.4) 27.0 (4.8) Missing 1 (0.6) 3 (2.1) 4 (2.9) 4 (3.4) 1 (1.0) 13 (2.0) Back pain status Yes 51 (33.1) 46 (32.4) 39 (27.9) 42 (35.6) 32 (33.3) 210 (32.3) No 103 (66.9) 96 (67.6) 101 (72.1) 76 (64.4) 64 (66.7) 440 (67.7) Bisphosphonate use status Yes 5 (3.2) 3 (2.1) 4 (2.9) 1 (0.8) 1 (1.0) 14 (2.2) No 68 (44.2) 62 (43.7) 71 (50.7) 55 (46.6) 44 (45.8) 300 (46.2) Missing 81 (52.6) 77 (54.2) 65 (46.4) 62 (52.5) 51 (53.1) 336 (51.7) Physical activity Walking, km/wk <0.8 14 (9.1) 6 (4.2) 12 (8.6) 7 (5.9) 13 (13.5) 52 (8.0) 0.8 to <8.1 75 (48.7) 75 (52.8) 60 (42.9) 70 (59.3) 43 (44.8) 323 (49.7) 8.1 to <16.1 42 (27.3) 36 (25.4) 41 (29.3) 30 (25.4) 17 (17.7) 166 (25.5) ≥16.1 17 (11.0) 19 (13.4) 21 (15.0) 5 (4.2) 16 (16.7) 78 (12.0) Missing 6 (3.9) 6 (4.2) 6 (4.3) 6 (5.1) 7 (7.3) 31 (4.8) Job Sedentary 8 (5.2) 8 (5.6) 4 (2.9) 4 (3.4) 0 24 (3.7) Sedentary plus 23 (14.9) 14 (9.9) 16 (11.4) 13 (11.0) 8 (8.3) 74 (11.4) occasional exercise 0.5 Sedentary plus 0.5 88 (57.1) 77 (54.2) 87 (62.1) 73 (61.9) 59 (61.5) 384 (59.1) active (or active housework [eg, daily dusting or vacuuming]) Predominantly manual, 27 (17.5) 30 (21.1) 22 (15.7) 17 (14.4) 15 (15.6) 111 (17.1) active all day Missing 8 (5.2) 13 (9.2) 11 (7.9) 11 (9.3) 14 (14.6) 57 (8.8) Sport None 80 (51.9) 78 (54.9) 89 (63.6) 70 (59.3) 60 (62.5) 377 (58.0) Golf, bowling, 22 (14.3) 26 (18.3) 15 (10.7) 15 (12.7) 8 (8.3) 86 (13.2) badminton, cycling, or swimming, 1 h/wk 2 h/wk of Golf, 30 (19.5) 23 (16.2) 22 (15.7) 19 (16.1) 11 (11.5) 105 (16.2) bowling, badminton, cycling, or swimming or1hof staying fit, aerobics, or squash ≥2 h/wk of Staying fit, 17 (11.0) 9 (6.3) 8 (5.7) 8 (6.8) 11 (11.5) 53 (8.2) aerobics, or squash Missing 5 (3.2) 6 (4.2) 6 (4.3) 6 (5.1) 6 (6.3) 29 (4.5) Disability, year 9, 0 (0.0-5.8) 0 (0.0-6.0) 0 (0.0-3.3) 0 (0.0-4.9) 0 (0.0-3.3) 0 (0.0-5.0) median (IQR) Longitudinal (n = 443) No. 112 100 97 76 58 443 Age, mean (SD), y 59.1 (5.3) 59.5 (5.7) 60.5 (6.3) 62.6 (6.1) 62.8 (5.9) 60.6 (6.0) Missing 1 (0.9) 2 (2.0) 3 (3.1) 0 0 6 (1.4) (continued) JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 5/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability Table 1. Baseline Characteristics of Study Participants (continued) Participants, No. (%) No. of segments Characteristic 012 3 4 Whole cohort Smoking status Never 77 (68.8) 63 (63.0) 64 (66.0) 44 (57.9) 38 (65.5) 286 (64.6) Current 13 (11.6) 14 (14.0) 11 (11.3) 16 (21.1) 9 (15.5) 63 (14.2) Ex-smoker 19 (17.0) 21 (21.0) 18 (18.6) 15 (19.7) 11 (19.0) 84 (19.0) Missing 3 (2.7) 2 (2.0) 4 (4.1) 1 (1.3) 0 10 (2.3) BMI, mean (SD) 26.4 (4.3) 26.7 (4.9) 27.2 (5.2) 26.5 (3.7) 28.3 (5.3) 26.9 (4.7) Missing 1 (0.9) 3 (3.0) 3 (3.1) 2 (2.6) 0 9 (2.0) Back pain status Yes 40 (35.7) 32 (32.0) 24 (24.7) 26 (34.2) 23 (39.7) 145 (32.7) No 72 (64.3) 68 (68.0) 73 (75.3) 50 (65.8) 35 (60.3) 298 (67.3) Bisphosphonate use status Yes 4 (3.6) 3 (3.0) 2 (2.1) 1 (1.3) 1 (1.7) 11 (2.5) No 43 (38.4) 43 (43.0) 55 (56.7) 35 (46.1) 26 (44.8) 202 (45.6) Missing 65 (58.0) 54 (54.0) 40 (41.2) 40 (52.6) 31 (53.4) 230 (51.9) Physical activity Walking, km/wk <0.8 11 (9.8) 5 (5.0) 12 (12.4) 5 (6.6) 9 (15.5) 42 (9.5) 0.8 to <8.1 54 (48.2) 50 (50.0) 42 (43.3) 46 (60.5) 29 (50.0) 221 (49.9) 8.1 to <16.1 31 (27.7) 27 (27.0) 22 (22.7) 20 (26.3) 8 (13.8) 108 (24.4) ≥16.1 10 (8.9) 16 (16.0) 17 (17.5) 3 (3.9) 10 (17.2) 56 (12.6) Missing 6 (5.4) 2 (2.0) 4 (4.1) 2 (2.6) 2 (3.5) 16 (3.6) Job Sedentary 5 (4.5) 5 (5.0) 4 (4.1) 3 (3.9) 0 17 (3.8) Sedentary plus 15 (13.4) 13 (13.0) 10 (10.3) 9 (11.8) 6 (10.3) 53 (12.0) occasional exercise 0.5 Sedentary plus 0.5 65 (58.0) 55 (55.0) 62 (63.9) 49 (64.5) 37 (63.8) 268 (60.5) active (or active housework [eg, daily dusting or vacuuming]) Predominantly manual, 21 (18.8) 24 (24.0) 14 (14.4) 11 (14.5) 10 (17.2) 80 (18.1) active all day Missing 6 (5.4) 3 (3.0) 7 (7.2) 4 (5.3) 5 (8.6) 25 (5.6) Sport None 60 (53.6) 52 (52.0) 57 (58.8) 45 (59.2) 38 (65.5) 252 (56.9) Golf, bowling, 14 (12.5) 21 (21.0) 15 (15.5) 8 (10.5) 4 (6.9) 62 (14.0) badminton, cycling, or swimming, 1 h/wk 2 h/wk of Previous 22 (19.6) 19 (19.0) 16 (16.5) 17 (22.4) 6 (10.3) 80 (18.1) or1hof staying fit, aerobics, or squash Abbreviations: BMI, body mass index (calculated as ≥2 h/wk of Staying fit, 11 (9.8) 6 (6.0) 4 (4.1) 4 (5.3) 8 (13.8) 33 (7.4) aerobics, or squash weight in kilograms divided by height in meters Missing 5 (4.5) 2 (2.0) 5 (5.2) 2 (2.6) 2 (3.4) 16 (3.6) squared); IQR, interquartile range. Disability, year 15, 0 (0.0-5.0) 0 (0.0-5.1) 0 (0.0-4.0) 0 (0.0-2.5) 0 (0.0-4.8) 0 (0.0-4.8) Number of segments of lumbar spine radiographic median (IQR) changes (based on Kellgren-Lawrence grade). Results Participant Characteristics A total of 650 women (mean [SD] age, 61.3 [5.9] years) were included in cross-sectional analyses, and a total of 443 women (mean [SD] age, 60.6 [6.0] years) were included in longitudinal analyses (Table 1; Figure). Most study participants were classified as either never smokers or ex-smokers (Table 1). The median score of back pain–related disability was 0 (interquartile range, 0.0-5.0 in cross-sectional analyses and 0.0-4.8 in longitudinal analyses) in both cross-sectional and longitudinal analyses. The distribution of each lumbar spine radiographic change at each lumbar spine segment JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 6/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability is listed in eTable 1 in the Supplement. Redundancy analyses were performed to assess whether 1 exposure could be estimated from any 2 other exposures, at each lumbar spine segment. No exposure was redundant (eTable 2 in the Supplement). K-L Grade–Based Score Using a multivariable ordinal logistic regression model, we found that women who had 1 or more segments with lumbar spine radiographic K-L grade–based changes were not statistically more likely to report more disability compared with women with no observed changes in both cross-sectional (eg, 1 segment vs 0 segments; step 2 model; OR, 1.22 [95% CI, 0.76-1.96]) and longitudinal analyses (Table 2). When further adjustment was made for physical activity, the results were similar (OR, 1.19 [95% CI, 0.73-1.93]). No evidence was found to support a linear or nonlinear trend between number of segments with lumbar spine radiographic changes and the severity of back pain–related disability. Osteophyte Grade–Based Score and Disc Space Narrowing Grade–Based Score For osteophyte grade–based score, no statistically significant association was found between the number of lumbar segments with radiographic changes and the severity of back pain–related disability in both cross-sectional (eg, 1 segment vs 0 segment; step 2 model; OR, 0.83 [95% CI, 0.57- 1.22]) and longitudinal analyses (Table 3). Similar results were observed when further adjustments were made in the models to account for participation in physical activity in both cross-sectional and longitudinal analyses. In the longitudinal analysis, a greater number of affected segments were linearly associated with less severe back pain–related disability (step 2 model). For the disc space narrowing grade–based score, no statistically significant association was observed between the number of lumbar segments with radiographic changes and the severity of back pain–related disability in both cross-sectional (eg, 1 segment vs 0 segment; step 2 model; OR, 1.43 [95% CI, 0.78-2.61]) and longitudinal analyses (Table 4). Results were similar when we further adjusted for physical activity (OR, 1.41 [95% CI, 0.77-2.60]). For the cross-sectional analyses, a higher Figure. Flowchart of Study Participants 1003 Patients in the Chingford 1000 Women Study 828 Participants followed up in year 9 655 Participants followed up in year 15 178 Excluded 212 Excluded 159 Without radiographic data 100 Without radiographic data 6 Missing value in 4 Missing value in radiographic data radiographic data 13 Without outcome data 108 Without outcome data 650 Included in cross-sectional main analysis 443 Included in longitudinal main analysis Sensitivity analysis with different sample size: Sensitivity analysis with different sample size: 210 With back pain 145 With back pain 365 With severe disease per K-L grade–based score 260 With severe disease per K-L grade–based score 283 With severe disease per osteophyte grade–based score 206 With severe disease per osteophyte grade–based score 126 With severe disease per disc space narrowing grade–based score 87 With severe disease per disc space narrowing grade–based score A total of 95.0% women (421 of 443) in the longitudinal analysis are also in the cross-sectional analysis. Besides all data mentioned in the flowchart, we also collected the data from year 6 (1994-1996) because it was when 1 necessary covariate (ie, physical activity) was measured. K-L indicates Kellgren-Lawrence. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 7/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability number of segments of lumbar spine radiographic characteristics were nonlinearly associated with less severe back pain–related disability (step 2 model). Exploratory and Sensitivity Analyses We did not find interactions with age, BMI, or smoking status (eTable 3 in the Supplement) between lumbar spine radiographic changes and the severity of back pain–related disability. Overall, our results remained similar under extensive sensitivity analyses (eAppendices 3-8 and eTables 4-25 in the Supplement). All E-values are listed in eTable 26 in the Supplement. Discussion Key Results In this cohort of middle-aged women from Chingford in east London, UK, no evidence was found to support an association between a higher number of segments with lumbar radiographic changes (K-L grade, osteophyte, and disc space narrowing) and more severe back pain–related disability. Our Table 2. Association Between K-L Grade–Based Score and Severity of Back Pain–Related Disability K-L grade–based score, OR (95% CI) P value for trend Variable 0 Segments 1 Segment 2 Segments 3 Segments 4 Segments Linear model Nonlinear model Cross-sectional, year 9 (n = 650) Women, No. (%) 154 (23.7) 142 (21.8) 140 (21.5) 118 (18.2) 96 (14.8) NA NA Unadjusted 1 [Reference] 1.19 (0.75-1.89) 0.84 (0.52-1.36) 0.90 (0.54-1.49) 0.81 (0.47-1.38) .24 .78 Multivariable adjusted 1 [Reference] 1.22 (0.76-1.96) 0.84 (0.51-1.38) 0.92 (0.54-1.56) 0.89 (0.50-1.57) .39 .91 Further adjusted for 1 [Reference] 1.19 (0.73-1.93) 0.81 (0.49-1.35) 0.87 (0.51-1.49) 0.85 (0.48-1.50) .30 .96 physical activity Longitudinal, year 15 (n = 443) Women, No. (%) 112 (25.3) 100 (22.6) 97 (21.9) 76 (17.2) 58 (13.1) NA NA Unadjusted 1 [Reference] 1.22 (0.70-2.11) 0.95 (0.53-1.68) 0.77 (0.41-1.44) 1.05 (0.55-2.02) .57 .92 Multivariable adjusted 1 [Reference] 1.06 (0.57-1.96) 0.94 (0.50-1.76) 0.69 (0.34-1.38) 0.83 (0.40-1.72) .20 .79 Further adjusted for 1 [Reference] 1.08 (0.58-2.02) 0.91 (0.48-1.72) 0.67 (0.33-1.38) 0.80 (0.38-1.68) .17 .80 physical activity Abbreviations: K-L, Kellgren-Lawrence; NA, not applicable; OR, odds ratio. Adjusted for age, body mass index, smoking status, back pain status, and bisphosphonate use. Adjusted for age, body mass index, smoking status, back pain status, bisphosphonate use, and year 9 back pain–related disability. Table 3. Association Between Osteophyte Grade–Based Score and the Severity of Back Pain–Related Disability Osteophyte grade–based score, OR (95% CI) P value for trend Variable 0 Segments 1 Segment 2 Segments 3 Segments 4 Segments Linear model Nonlinear model Cross-sectional, year 9 (n = 650) Women, No. (%) 258 (39.7) 226 (34.8) 102 (15.7) 44 (6.8) 20 (3.1) NA NA Unadjusted 1 [Reference] 0.80 (0.55-1.16) 0.81 (0.50-1.32) 0.53 (0.25-1.12) 0.81 (0.31-2.14) .12 .59 Multivariable adjusted 1 [Reference] 0.83 (0.57-1.22) 0.78 (0.47-1.30) 0.58 (0.27-1.26) 1.03 (0.37-2.85) .25 .42 Further adjusted for 1 [Reference] 0.82 (0.56-1.21) 0.79 (0.47-1.32) 0.60 (0.28-1.31) 0.98 (0.35-2.73) .26 .46 physical activity Longitudinal, year 15 (n = 443) Women, No. (%) 192 (43.3) 157 (35.4) 67 (15.1) 19 (4.3) 8 (1.8) NA NA Unadjusted 1 [Reference] 0.71 (0.46-1.10) 0.60 (0.33-1.08) 0.41 (0.13-1.27) 0.49 (0.10-2.40) .02 .65 Multivariable adjusted 1 [Reference] 0.76 (0.47-1.24) 0.53 (0.28-1.02) 0.49 (0.14-1.70) 0.31 (0.06-1.72) .01 .75 Further adjusted for 1 [Reference] 0.76 (0.46-1.24) 0.52 (0.27-1.03) 0.50 (0.14-1.74) 0.33 (0.06-1.79) .01 .68 physical activity Abbreviations: NA, not applicable; OR, odds ratio. Adjusted for age, body mass index, smoking status, back pain status, and bisphosphonate use. Adjusted for age, body mass index, smoking status, back pain status, bisphosphonate use, and year 9 back pain–related disability. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 8/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability results remained unchanged after including potential interactions with important confounders, such as age, BMI, and smoking status, and after extensive sensitivity analyses. Comparison With Previous Studies For K-L results, our findings contradict those of Lee et al, who found that K-L grades were significantly associated with the Oswestry Disability Index. The main reason for such a discrepancy in results may be the design features of the study by Lee et al, which only included cross-sectional analyses with insufficient adjustment for important confounders (eg, physical activity, smoking status, and BMI). For osteophyte results, the results from the cross-sectional analyses are consistent with a previous cross-sectional study by Perera et al, who identified that the presence of osteophytes on radiographs was not associated with physical disability measured with the Oswestry Disability Index. However, results from our longitudinal analyses indicated that a greater number of affected segments were linearly associated with less severe back pain–related disability. One possible explanation is the biomechanical stability provided by spinal osteophytes, which have been proven to increase spinal resistance in compression. Another explanation is that the results simply reflect large numbers of analyses completed, which need future studies to verify. For disc space narrowing 16 15 results, our results are similar to those of Perera et al but different than the findings of Lee et al. Such differences could also be associated with the methodological limitations already described in the study conducted by Lee et al. Overall, the prevalence of lumbar spine radiographic findings in our study is similar to that of previous studies that indicated that many imaging-based spinal radiographic changes are likely part of normal, asymptomatic aging. Strengths and Limitations Our study has several strengths. To our knowledge, this is the first study to create a composite score that reflects the overall association of lumbar spine radiographic changes (ie, number of affected segments and severity of changes) with the severity of back pain–related disability. We used population-based data that contain a long-term follow-up with good recruitment and retention rates, and multiple potential confounders were measured. We also overcame some methodological limitations from previous studies; we included a cohort study design and incorporated a systematic way to select and control confounders, we have repeated measures of back pain–related disability that allow us to adjust for baseline disability, we assessed the potential interaction term, and we performed extensive sensitivity analyses to evaluate the robustness of the results. Limitations also need to be considered. First, the Chingford 1000 Women Study included middle-aged women in a specific area of the UK. We must exercise caution when generalizing the Table 4. Association Between Disc Space Narrowing Grade–Based Score and the Severity of Back Pain–Related Disability Disc space narrowing grade–based score, OR (95% CI) P value for trend Variable 0 Segments 1 Segment 2 Segments 3 Segments 4 Segments Linear model Nonlinear model Cross-sectional, year 9 (n = 650) Women, No. (%) 100 (15.4) 107 (16.5) 147 (22.6) 131 (20.2) 165 (25.4) NA NA Unadjusted 1 [Reference] 1.33 (0.74-2.37) 1.30 (0.75-2.24) 1.33 (0.77-2.31) 0.98 (0.57-1.69) .75 .11 Multivariable adjusted 1 [Reference] 1.43 (0.78-2.61) 1.56 (0.88-2.76) 1.44 (0.81-2.57) 1.07 (0.60-1.92) .94 .04 Further adjusted for 1 [Reference] 1.41 (0.77-2.60) 1.56 (0.87-2.77) 1.45 (0.81-2.59) 1.04 (0.57-1.87) .86 .03 physical activity Longitudinal, year 15 (n = 443) Women, No. (%) 70 (15.8) 84 (19.0) 102 (23.0) 88 (19.9) 99 (22.3) NA NA Unadjusted 1 [Reference] 0.91 (0.47-1.75) 0.78 (0.41-1.49) 1.07 (0.56-2.05) 1.24 (0.67-2.29) .34 .25 Multivariable adjusted 1 [Reference] 0.72 (0.34-1.53) 0.74 (0.36-1.52) 1.06 (0.52-2.20) 1.26 (0.62-2.57) .18 .18 Further adjusted for 1 [Reference] 0.68 (0.31-1.48) 0.67 (0.31-1.43) 1.12 (0.53-2.34) 1.33 (0.63-2.80) .13 .12 physical activity Abbreviations: NA, not applicable; OR, odds ratio. Adjusted for age, body mass index, smoking status, back pain status, and bisphosphonate use. Adjusted for age, body mass index, smoking status, back pain status, bisphosphonate use, and year 9 back pain–related disability. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 9/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability results to men, other age groups, other racial/ethnic groups, or other countries. Second, as with most studies, there is the potential for residual confounding (eg, participation in physical activity was measured 3 years before baseline). Third, the labeling of the images may have introduced potential bias in our results, given that there was only 1 observer and that lumbar spine levels were decided by the clinical experience. Fourth, owing to data unavailability, we could not establish whether there was any association between other radiologic changes, including spondylolisthesis or vertebral body height (ie, osteoporotic fractures), and severity of functional limitation. Fifth, although our outcome correlated well with the Oswestry Disability Questionnaire, it lacked strict validation. Sixth, although we aimed to focus on LBP–related disability, we only had a back pain variable, which might be slightly different from LBP. Implications for Practice and Research Clinicians may use the results of this study to educate patients and their colleagues that lumbar radiographic findings cannot provide prognostic information on back pain–related disability, further adding to the evidence supporting the urge to reduce unnecessary imaging referrals. Future studies should include participants of both sexes and larger sample sizes and should include multiple centers to increase external validity. The association between the findings of complex imaging (eg, computed tomography scans, magnetic resonance imaging, or nuclear bone scans) and symptom severity in people with LBP needs to be further explored, considering the increasing use of such imaging. Conclusions In this cohort of middle-aged, community-dwelling women, there was no evidence to support an association between a higher number of lumbar segments with radiographic changes (K-L grade, osteophytes, and disc space narrowing) and more severe back pain–related disability cross- sectionally or over time. The findings suggest that the changes detected on lumbar radiographs provide limited value for decision-making regarding back pain management in this population. ARTICLE INFORMATION Accepted for Publication: March 24, 2021. Published: May 20, 2021. doi:10.1001/jamanetworkopen.2021.10715 Open Access: This is an open access article distributed under the terms of the CC-BY License.©2021ChenLetal. JAMA Network Open. Corresponding Author: Lingxiao Chen, MBBS, MMed, Institute of Bone and Joint Research, The Kolling Institute, Northern Clinical School, Faculty of Medicine and Health, University of Sydney, Sydney, NSW 2064, Australia (lche4036 @uni.sydney.edu.au). Author Affiliations: Institute of Bone and Joint Research, The Kolling Institute, Northern Clinical School, Faculty of Medicine and Health, University of Sydney, Sydney, New South Wales, Australia (Chen, M. L. Ferreira); Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, Oxford, United Kingdom (Perera, Radojčić, Arden); Department of Allied Health Sciences, Faculty of Medicine, University of Colombo, Colombo, Sri Lanka (Perera); Centre for Sport, Exercise and Osteoarthritis Research Versus Arthritis, University of Oxford, Oxford, United Kingdom (Radojčić); School of Health Sciences, Faculty of Medicine and Health, University of Sydney, Sydney, New South Wales, Australia (Beckenkamp, P. H. Ferreira); Department of Twin Research and Genetic Epidemiology, King’s College London, London, United Kingdom (Hart, Spector); Medical Research Council Environmental Epidemiology Unit, University of Southampton, Southampton, United Kingdom (Arden). Author Contributions: Drs Chen and Perera had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Arden and M. L. Ferreira contributed equally as co–senior authors. Concept and design: Chen, Perera, Radojčić, Beckenkamp, P.H. Ferreira, Spector, Arden, M. L. Ferreira. Acquisition, analysis, or interpretation of data: Chen, Perera, Beckenkamp, Hart, Arden, M. L. Ferreira. JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 10/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability Drafting of the manuscript: Chen, Beckenkamp, Spector, Arden. Critical revision of the manuscript for important intellectual content: Chen, Perera, Radojčić, Beckenkamp, P. H. Ferreira, Hart, Arden, M. L. Ferreira. Statistical analysis: Chen, Spector. Obtained funding: Hart. Administrative, technical, or material support: Chen, Perera, Hart. Supervision: Beckenkamp, P. H. Ferreira, Arden, M. L. Ferreira. Conflict of Interest Disclosures: Dr Spector reported serving as a scientific consultant for Zoe Global Ltd. Dr Arden reported receiving personal fees from Pfizer/Lilly and Bristows LLP and grants from Merck outside the submitted work. No other disclosures were reported. Funding/Support: Arthritis Research UK (now Versus Arthritis) and the Oxford NIHR Musculoskeletal Biomedical Research Unit provided funding support to the study. Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. Additional Contributions: We would like to thank all the participants of the Chingford 1000 Women Study; Alan Hakim, MA, Rheumatology, Platinum Medical Centre, The Wellington Hospital, research coordination; Maxine Daniels, BSc(Hons), University of Oxford, research coordination; and Alison Turner, BSc, NIHR Musculoskeletal Biomedical Research Unit, Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, development of data management system, for their time and dedication. We would like to thank Julie Damnjanović, BSc, Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, for her valued support to the team and the study data management. None of the individuals named here were compensated for their contributions. REFERENCES 1. GBD 2017 Disease and Injury Incidence and Prevalence Collaborators. Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990-2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet. 2018;392(10159):1789-1858. doi:10. 1016/S0140-6736(18)32279-7 2. Hartvigsen J, Hancock MJ, Kongsted A, et al; Lancet Low Back Pain Series Working Group. What low back pain is and why we need to pay attention. Lancet. 2018;391(10137):2356-2367. doi:10.1016/S0140-6736(18)30480-X 3. Jordan KP, Kadam UT, Hayward R, Porcheret M, Young C, Croft P. Annual consultation prevalence of regional musculoskeletal problems in primary care: an observational study. BMC Musculoskelet Disord. 2010;11:144. doi:10. 1186/1471-2474-11-144 4. Chou R, Qaseem A, Owens DK, Shekelle P; Clinical Guidelines Committee of the American College of Physicians. Diagnostic imaging for low back pain: advice for high-value health care from the American College of Physicians. Ann Intern Med. 2011;154(3):181-189. doi:10.7326/0003-4819-154-3-201102010-00008 5. Foster NE, Anema JR, Cherkin D, et al; Lancet Low Back Pain Series Working Group. Prevention and treatment of low back pain: evidence, challenges, and promising directions. Lancet. 2018;391(10137):2368-2383. doi:10. 1016/S0140-6736(18)30489-6 6. National Institute of Health and Care Excellence (NICE); National Guideline Centre. Low Back Pain and Sciatica in Over 16s: Assessment And Management. NICE Guideline No. 59. National Institute for Health and Care Excellence; 2016. 7. Downie A, Hancock M, Jenkins H, et al. How common is imaging for low back pain in primary and emergency care? systematic review and meta-analysis of over 4 million imaging requests across 21 years. Br J Sports Med. 2020;54(11):642-651. doi:10.1136/bjsports-2018-100087 8. Wall BF, Kendall GM, Edwards AA, Bouffler S, Muirhead CR, Meara JR. What are the risks from medical X-rays and other low dose radiation? Br J Radiol. 2006;79(940):285-294. doi:10.1259/bjr/55733882 9. Chou R, Fu R, Carrino JA, Deyo RA. Imaging strategies for low-back pain: systematic review and meta-analysis. Lancet. 2009;373(9662):463-472. doi:10.1016/S0140-6736(09)60172-0 10. Rhodes LA, McPhillips-Tangum CA, Markham C, Klenk R. The power of the visible: the meaning of diagnostic tests in chronic back pain. . 1999;48(9):1189-1203. doi:10.1016/s0277-9536(98)00418-3 11. Schers H, Wensing M, Huijsmans Z, van Tulder M, Grol R. Implementation barriers for general practice guidelines on low back pain: a qualitative study. Spine (Phila Pa 1976). 2001;26(15):E348-E353. doi:10.1097/ 00007632-200108010-00013 JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 11/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability 12. Taylor S, Bishop A. Patient and public beliefs about the role of imaging in the management of non-specific low back pain: a scoping review. Physiotherapy. 2020;107:224-233. doi:10.1016/j.physio.2019.08.014 13. Jarvik JG, Gold LS, Comstock BA, et al. Association of early imaging for back pain with clinical outcomes in older adults. JAMA. 2015;313(11):1143-1153. doi:10.1001/jama.2015.1871 14. Kellgren JH, Lawrence JS. Radiological assessment of osteo-arthrosis. Ann Rheum Dis. 1957;16(4):494-502. doi:10.1136/ard.16.4.494 15. Lee SY, Cho NH, Jung YO, Seo YI, Kim HA. Prevalence and risk factors for lumbar spondylosis and its association with low back pain among rural Korean residents. J Korean Neurosurg Soc. 2017;60(1):67-74. doi:10.3340/jkns. 2016.0505.007 16. Perera RS, Dissanayake PH, Senarath U, Wijayaratne LS, Karunanayake AL, Dissanayake VHW. Associations between disc space narrowing, anterior osteophytes and disability in chronic mechanical low back pain: a cross sectional study. BMC Musculoskelet Disord. 2017;18(1):193. doi:10.1186/s12891-017-1562-9 17. Bartlett SJ, Ling SM, Mayo NE, Scott SC, Bingham CO III. Identifying common trajectories of joint space narrowing over two years in knee osteoarthritis. Arthritis Care Res (Hoboken). 2011;63(12):1722-1728. doi:10.1002/ acr.20614 18. Felson DT, Gale DR, Elon Gale M, et al. Osteophytes and progression of knee osteoarthritis. Rheumatology (Oxford). 2005;44(1):100-104. doi:10.1093/rheumatology/keh411 19. Vandenbroucke JP, von Elm E, Altman DG, et al; STROBE initiative. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. Ann Intern Med. 2007;147(8): W163-94. doi:10.7326/0003-4819-147-8-200710160-00010-w1 20. von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP; STROBE Initiative. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann Intern Med. 2007;147(8):573-577. doi:10.7326/0003-4819-147-8- 200710160-00010 21. Hassett G, Hart DJ, Manek NJ, Doyle DV, Spector TD. Risk factors for progression of lumbar spine disc degeneration: the Chingford Study. Arthritis Rheum. 2003;48(11):3112-3117. doi:10.1002/art.11321 22. Lane NE, Nevitt MC, Genant HK, Hochberg MC. Reliability of new indices of radiographic osteoarthritis of the hand and hip and lumbar disc degeneration. J Rheumatol. 1993;20(11):1911-1918. 23. Co YY, Eaton S, Maxwell MW. The relationship between the St. Thomas and Oswestry Disability Scores and the severity of low back pain. J Manipulative Physiol Ther. 1993;16(1):14-18. 24. DAGitty.net. DAGitty—draw and analyze causal diagrams. Accessed April 7, 2021. http://www.dagitty.net/ 25. Textor J, van der Zander B, Gilthorpe MS, Liśkiewicz M, Ellison GT. Robust causal inference using directed acyclic graphs: the R package “dagitty”. Int J Epidemiol. 2016;45(6):1887-1894. 26. Harrell Jr FE. Regression Modeling Strategies: With Applications to Linear Models, Logistic and Ordinal Regression, and Survival Analysis. Springer; 2015. doi:10.1007/978-3-319-19425-7 27. Brookhart MA, Schneeweiss S, Rothman KJ, Glynn RJ, Avorn J, Stürmer T. Variable selection for propensity score models. Am J Epidemiol. 2006;163(12):1149-1156. doi:10.1093/aje/kwj149 28. Rothman KJ, Greenland S, Lash TL. Modern Epidemiology. Lippincott Williams & Wilkins; 2008. 29. Harrell Jr FE. Package ‘Hmisc.’ CRAN2018. Accessed July 22, 2020. https://cran.uib.no/web/packages/Hmisc/ Hmisc.pdf 30. Al-Rawahi M, Luo J, Pollintine P, Dolan P, Adams MA. Mechanical function of vertebral body osteophytes, as revealed by experiments on cadaveric spines. Spine (Phila Pa 1976). 2011;36(10):770-777. doi:10.1097/BRS. 0b013e3181df1a70 31. Brinjikji W, Luetmer PH, Comstock B, et al. Systematic literature review of imaging features of spinal degeneration in asymptomatic populations. AJNR Am J Neuroradiol. 2015;36(4):811-816. doi:10.3174/ajnr.A4173 SUPPLEMENT. eFigure 1. Causal Diagram eFigure 2. Histogram of Outcome Distribution eFigure 3. Missing Data Pattern eAppendix 1. Methods for Exploratory and Sensitivity Analyses eAppendix 2. Details of Statistical Methods eTable 1. Distribution of Lumbar Spine Radiographic Changes at Each Lumbar Spine Segment eTable 2. Redundancy Analysis of Exposures eTable 3. Interaction with Age, BMI, or Smoking Status JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 12/13 JAMA Network Open | Imaging Association of Lumbar Spine Radiographic Changes With Severity of Back Pain–Related Disability eAppendix 3. Sensitivity Analyses Results: Change on the Cut-Off Points of Exposures eTable 4. Kellgren-Lawrence Grade–Based Score eTable 5. Osteophytes Grade–Based Score eTable 6. Disc Space Narrowing Grade–Based Score eAppendix 4. Sensitivity Analyses Results: Total Original Score of Exposures eTable 7. Kellgren-Lawrence Grade eTable 8. Osteophytes Grade eTable 9. Disc Space Narrowing Grade eAppendix 5. Sensitivity Analyses Results: Disease Severity eTable 10. Kellgren-Lawrence Grade–Based Score eTable 11. Osteophytes Grade–Based Score eTable 12. Disc Space Narrowing Grade–Based Score eAppendix 6. Sensitivity Analyses Results: Restricting to Women With Back Pain eTable 13. Kellgren-Lawrence Grade–Based Score eTable 14. Osteophytes Grade–Based Score eTable 15. Disc Space Narrowing Grade–Based Score eAppendix 7. Sensitivity Analyses Results: Changing the Model to cloglog Link Function eTable 16. Kellgren-Lawrence Grade–Based Score eTable 17. Osteophytes Grade–Based Score eTable 18. Disc Space Narrowing Grade–Based Score eAppendix 8. Sensitivity Analyses Results: Change the Model to Linear Regression eTable 19. Kellgren-Lawrence Grade–Based Score eTable 20. Osteophytes Grade–Based Score eTable 21. Disc Space Narrowing Grade–Based Score eTable 22. Additionally Adjusted for Pain Medication and Depression eTable 23. Kellgren-Lawrence Grade–Based Score eTable 24. Osteophytes Grade–Based Score eTable 25. Disc Space Narrowing Grade–Based Score eTable 26. E-Value JAMA Network Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 (Reprinted) May 20, 2021 13/13 Supplementary Online Content Chen L, Perera RS, MR, et al. Association of lumbar spine radiographic changes with severity of back pain related disability among middle-aged, community- dwelling women. JAMA Netw Open. 2021;4(5):e2110715. doi:10.1001/jamanetworkopen.2021.10715 eFigure 1. Causal Diagram eFigure 2. Histogram of Outcome Distribution eFigure 3. Missing Data Pattern eAppendix 1. Methods for Exploratory and Sensitivity Analyses eAppendix 2. Details of Statistical Methods eTable 1. Distribution of Lumbar Spine Radiographic Changes at Each Lumbar Spine Segment eTable 2. Redundancy Analysis of Exposures eTable 3. Interaction with Age, BMI, or Smoking Status eAppendix 3. Sensitivity Analyses Results: Change on the Cut-Off Points of Exposures eTable 4. Kellgren-Lawrence Grade Based Score eTable 5. Osteophytes Grade Based Score eTable 6. Disc Space Narrowing Grade Based Score eAppendix 4. Sensitivity Analyses Results: Total Original Score of Exposures eTable 7. Kellgren-Lawrence Grade eTable 8. Osteophytes Grade eTable 9. Disc Space Narrowing Grade eAppendix 5. Sensitivity Analyses Results: Disease Severity eTable 10. Kellgren-Lawrence Grade Based Score eTable 11. Osteophytes Grade Based Score eTable 12. Disc Space Narrowing Grade Based Score eAppendix 6. Sensitivity Analyses Results: Restricting to Women With Back Pain eTable 13. Kellgren-Lawrence Grade Based Score eTable 14. Osteophytes Grade Based Score eTable 15. Disc Space Narrowing Grade Based Score eAppendix 7. Sensitivity Analyses Results: Changing the Model to cloglog Link Function eTable 16. Kellgren-Lawrence Grade Based Score eTable 17. Osteophytes Grade Based Score eTable 18. Disc Space Narrowing Grade Based Score eAppendix 8. Sensitivity Analyses Results: Change the Model to Linear Regression eTable 19. Kellgren-Lawrence Grade Based Score eTable 20. Osteophytes Grade Based Score eTable 21. Disc Space Narrowing Grade Based Score eTable 22. Additionally Adjusted for Pain Medication and Depression eTable 23. Kellgren-Lawrence Grade Based Score eTable 24. Osteophytes Grade Based Score eTable 25. Disc Space Narrowing Grade Based Score eTable 26. E-Value © 2021 Chen L et al. JAMA Network Open. This supplementary material has been provided by the authors to give readers additional information about their work. © 2021 Chen L et al. JAMA Network Open. eFigure 1. Causal Diagram © 2021 Chen L et al. JAMA Network Open. DAGitty is a browser-based environment for creating, editing, and analyzing causal diagrams (also known as directed acyclic graphs or causal Bayesian networks). The focus is on the use of causal diagrams for minimizing bias in empirical studies in epidemiology and o ther disciplines. We used two steps to choose the final covariates included in the model. At first, we pre-specified a lot of potential confounders based on clinical knowledge. The second step included searching previous literature for well accepted confounders and then drawing the causal diagram to find the minimal sufficient adjustment sets based on the established evidence. Initially selected covariates: age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, blood pressure, serious operations/illnesses, quality of life, history of fall, pain-killer usage status, anticholinergic medication usage status, antidepressant medication usage status, and angiotensin-converting enzyme inhibitors usage status. Minimal sufficient adjustment sets: age (as a continuous variable), BMI (as a continuous variable), smoking status (never, current, and ex - smoker), back pain status (yes or no), bisphosphonates use (yes or no), and physical activity (as three domains: walking, job and spor t; Walking [per week]: <0.5 miles, 0.5-5 miles, 5-10 miles, and 10+ miles. Job: sedentary, sedentary + occasional exercise, 0.5 sedentary + 0.5 active [or active housework, e.g., daily dust/hoover]), and predominantly manual, active all day. Sport: none, 1 hour per week golf, bowls, badminton, cycling or swimming, 2 hours previous or 1 hour keep-fit, aerobics, squash, and 2 hours + keep-fit, aerobics, squash). © 2021 Chen L et al. JAMA Network Open. eFigure 2. Histogram of Outcome Distribution A. Cross-sectional part. B. Longitudinal part. © 2021 Chen L et al. JAMA Network Open. eFigure 3. Missing Data Pattern A. Cross-sectional part. B. Longitudinal part. © 2021 Chen L et al. JAMA Network Open. eAppendix 1. Methods for Exploratory and Sensitivity Analyses Exploratory analyses We examined whether the association between lumbar spine radiographic changes and the severity of back pain-related disability differed by age, BMI or smoking status through testing of multiplicative interactions using WALD statistics. Sensitivity analyses 1) To assess the validity of our cut-off points for defining lumbar spine radiographic changes, we changed the cut-off moderate-to- 2) We changed the composite score by using the total original score, which is a sum score from the original grade score at each segment. For example, the K/L grade based on total original score ranged from 0-16 (at each segment 0-4). 3) We built a new composite score based on the disease severity: for disc space narrowing and osteophyte based score, grade 0 vs grade 2 and 3; for K/L grade based score, grade 0 and 1 vs grade 3 and 4. 4) Considering the potential heterogeneity of the population, we restricted our analyses to women with back pain. 5) Considering the potential model misspecification issue, we changed it to linear model and ordinal logistic regression with cloglog link function. 6) To explore the potential influence from unmeasured confounding, E-value was calculated. 7) As one reviewer suggested, we additionally adjusted pain medication and depression considering they might affect the results as potential strong prognostic factors. © 2021 Chen L et al. JAMA Network Open. eAppendix 2. Details of Statistical Methods Ordinal logistic regression: Liu et al. indicated that ordinal regression models are robust for continuous outcomes, especially when the distributions of continuous responses are skewed (Stat Med. 2017 Nov 30;36(27):4316-4335). Our continuous outcomes were skewed so that we chose ordinal logistic regression. Stepped modelling framework: If all covariates were measured at the same time, two models were sufficient: step 1: unadjusted analyses; step 2: analyses adjusted for age, BMI, back pain status, physical activity, bisphosphonates usage status, and smoking status (additionally adjusted for Year 9 back pain-related disability for the longitudinal analysis). The reason for the inclusion of Year 9 back pain-related disability for the longitudinal analysis is that it is a strong prognostic factor for Year 15 back pain-related disability. As VanderWeele et al. indicated, it is often important to control, whenever possible, for the outcome at or prior to the time of the baseline exposure assessment so that confounding control assumption is as plausible as possible (Statistical Science 35.3 (2020): 437-466.). In our study, among all selected covariates, physical activity was measured at Year 6 and others were measured at Year 9 (baseline at our study). Considering physical activity could change between Year 6 and Year 9, we built three models: step 1: unadjusted analyses; step 2: analyses adjusted for age, BMI, back pain status, bisphosphonates usage status, and smoking status (additionally adjusted for Year 9 back pain-related disability for the longitudinal analysis); step 3: analyses further adjusted for physical activity. Exposure modelling: For each type of radiographic changes, our exposure has five values: 0, 1, 2, 3, and 4. The value equals the number of lumbar spine segments affected by radiographic © 2021 Chen L et al. JAMA Network Open. changes. We could consider the exposure as the unordered categorical variable. In this case, we set 0 as the reference level and obtained the estimate by comparing other values with 0. We could also consider 0-4 as the continuous variable. As a continuous variable, we tested the linear trend by modelling the exposure as continuous variable and reported the P-value (whether the regression coefficient of the exposure variable equalled zero); we also tested the non-linear trend by adding the quartic term to the previous model and reported the P-value (whether the added quartic term could improve performance of the previous model through analysis of variance). Multiple imputation: It is a general approach to handle missing data in epidemiological and clinical research (BMJ 2009;338:b2393). It includes two steps: step 1: to create multiple copies of the dataset with the missing data replaced by imputed values; step 2: to fit the model to each of the imputed datasets and then calculate the final estimate by combing the estimate from each Rubin, Donald B. Multiple imputation for nonresponse in surveys. Vol. 81. John Wiley & Sons, 2004.). © 2021 Chen L et al. JAMA Network Open. eTable 1. Distribution of Lumbar Spine Radiographic Changes at Each Lumbar Spine Segment Cross-section (n=650) Longitudinal (n=443) Kellgren-Lawrence grade L1-L2 grade 0 229 160 grade 1 199 140 grade 2 94 57 grade 3 96 68 grade 4 32 18 L2-L3 grade 0 160 121 grade 1 183 126 grade 2 126 75 grade 3 137 92 grade 4 44 L3-L4 grade 0 102 72 grade 1 202 143 grade 2 146 93 grade 3 171 119 grade 4 29 16 L4-L5 grade 0 135 89 grade 1 230 167 grade 2 88 61 grade 3 116 75 grade 4 81 51 Disc space narrowing L1-L2 grade 0 311 219 grade 1 295 196 grade 2 29 23 grade 3 15 5 L2-L3 grade 0 300 216 grade 1 296 197 grade 2 40 22 grade 3 14 8 L3-L4 grade 0 276 198 grade 1 330 222 grade 2 31 15 grade 3 13 8 L4-L5 grade 0 259 191 grade 1 280 182 © 2021 Chen L et al. JAMA Network Open. grade 2 67 38 grade 3 44 32 Osteophytes L1-L2 grade 0 551 383 grade 1 92 58 grade 2 7 2 grade 3 0 0 L2-L3 grade 0 509 368 grade 1 129 71 grade 2 10 4 grade 3 2 0 L3-L4 grade 0 468 335 grade 1 161 98 grade 2 18 10 grade 3 3 0 L4-L5 grade 0 430 306 grade 1 191 118 grade 2 27 18 grade 3 2 1 © 2021 Chen L et al. JAMA Network Open. eTable 2. Redundancy Analysis of Exposures R-squared with which each variable can be predicted from all other variables. R-squared cut-off: 0.75. R-squared Cross-sectional L1-L2 Osteophytes 0.157 Disc space narrowing 0.662 Kellgren-Lawrence grade 0.644 L2-L3 Osteophytes 0.235 Disc space narrowing 0.649 Kellgren-Lawrence grade 0.633 L3-L4 Osteophytes 0.198 Disc space narrowing 0.659 Kellgren-Lawrence grade 0.647 L4-L5 Osteophytes 0.165 Disc space narrowing 0.721 Kellgren-Lawrence grade 0.717 Longitudinal L1-L2 Osteophytes 0.020 Disc space narrowing 0.574 Kellgren-Lawrence grade 0.574 L2-L3 Osteophytes 0.195 Disc space narrowing 0.616 Kellgren-Lawrence grade 0.603 L3-L4 Osteophytes 0.147 Disc space narrowing 0.657 Kellgren-Lawrence grade 0.657 L4-L5 Osteophytes 0.149 Disc space narrowing 0.725 Kellgren-Lawrence grade 0.727 © 2021 Chen L et al. JAMA Network Open. eTable 3. Interaction with Age, BMI, or Smoking Status Cross-sectional (P-value) Longitudinal (P-value) Kellgren-Lawrence grade based score Age 0.84 0.61 BMI 0.62 0.23 Smoke status 0.18 0.66 Osteophytes grade based score Age 0.61 0.96 BMI 0.45 0.50 Smoke status 0.63 0.62 Disc space narrowing grade based score Age 0.28 0.65 BMI 0.88 0.14 Smoke status 0.41 0.20 © 2021 Chen L et al. JAMA Network Open. eAppendix 3. Sensitivity Analyses Results: Change on the Cut-Off Points of Exposures (corresponding to the first sensitivity analysis; the cut- off points changed -to- ). eTable 4. Kellgren-Lawrence Grade Based Score Variables K/L grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 296 (45.5) 153 (23.5) 99 (15.2) 53 (8.2) 49 (7.5) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.05 (0.70, 1.58) 1.24 (0.78, 1.98) 1.09 (0.60, 1.98) 1.02 (0.53, 1.93) 0.66 0.75 Multivariable adjusted 1 (reference) 1.07 (0.70, 1.64) 1.31 (0.80, 2.13) 1.15 (0.61, 2.16) 1.32 (0.67, 2.62) 0.29 0.78 Further adjusted for 1 (reference) 1.06 (0.69, 1.63) 1.25 (0.76, 2.04) 1.12 (0.60, 2.12) 1.32 (0.66, 2.64) 0.34 0.89 physical activity Longitudinal (Year 15, n=443) Number of women (%) 200 (45.1) 110 (24.8) 68 (15.3) 38 (8.6) 27 (6.1) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.16 (0.71, 1.88) 1.14 (0.64, 2.03) 1.34 (0.68, 2.66) 1.39 (0.62, 3.14) 0.29 0.57 Multivariable adjusted 1 (reference) 1.10 (0.64, 1.88) 1.05 (0.54, 2.02) 1.38 (0.65, 2.95) 1.12 (0.45, 2.77) 0.56 0.87 Further adjusted for 1 (reference) 1.08 (0.62, 1.86) 1.00 (0.51, 1.97) 1.46 (0.66, 3.19) 1.15 (0.46, 2.90) 0.53 0.98 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 5. Osteophytes Grade Based Score Variables Osteophytes grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 596 (91.7) 45 (6.9) 4 (0.6) 4 (0.6) 1 (0.2) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.89 (0.47, 1.69) 0.56 (0.06, 5.02) N/A N/A 0.16 0.45 Multivariable adjusted 1 (reference) 0.98 (0.51, 1.89) 0.53 (0.06, 4.89) N/A N/A 0.31 0.33 Further adjusted for 1 (reference) 1.00 (0.52, 1.97) 0.62 (0.07, 5.99) N/A N/A 0.35 0.33 physical activity Longitudinal (Year 15, n=443) Number of women (%) 411 (92.8) 30 (6.8) 1 (0.2) 1 (0.2) 0 (0.0) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.55 (0.23, 1.28) N/A N/A N/A 0.10 0.37 Multivariable adjusted 1 (reference) 0.62 (0.24, 1.56) N/A N/A N/A 0.18 0.85 Further adjusted for 1 (reference) 0.64 (0.25, 1.62) N/A N/A N/A 0.18 0.88 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 6. Disc Space Narrowing Grade Based Score Variables Disc space narrowing grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 480 (73.8) 114 (17.5) 37 (5.7) 11 (1.7) 8 (1.2) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.39 (0.92, 2.11) 1.01 (0.51, 2.01) 0.75 (0.20, 2.77) 0.33 (0.04, 2.72) 0.87 0.18 Multivariable adjusted 1 (reference) 1.44 (0.93, 2.22) 1.01 (0.49, 2.06) 0.88 (0.23, 3.42) 0.46 (0.05, 3.81) 0.88 0.11 Further adjusted for 1 (reference) 1.45 (0.93, 2.25) 1.00 (0.48, 2.06) 0.95 (0.24, 3.76) 0.46 (0.05, 3.94) 0.84 0.12 physical activity Longitudinal (Year 15, n=443) Number of women (%) 332 (74.9) 82 (18.5) 21 (4.7) 5 (1.1) 3 (0.7) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.35 (0.83, 2.22) 1.75 (0.77, 3.97) 0.45 (0.05, 3.88) N/A 0.64 0.13 Multivariable adjusted 1 (reference) 1.28 (0.73, 2.23) 1.63 (0.67, 3.98) 0.61 (0.06, 6.65) N/A 0.93 0.06 Further adjusted for 1 (reference) 1.41 (0.80, 2.49) 1.68 (0.68, 4.12) 0.67 (0.06, 7.59) N/A 0.72 0.05 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eAppendix 4. Sensitivity Analyses Results: Total Original Score of Exposures (corresponding to the second sensitivity analysis; the total original score which is a sum score from the original grade score at each segment). eTable 7. Kellgren-Lawrence Grade Linear model (effect estimate with its 95% confidence interval) Non-linear model (p-value) Cross-sectional (Year 9, n=650) Unadjusted -0.12 (-0.34, 0.11) 0.26 Multivariable adjusted -0.07 (-0.31, 0.17) 0.19 Further adjusted for physical activity -0.09 (-0.33, 0.16) 0.15 Longitudinal (Year 15, n=443) Unadjusted -0.04 (-0.32, 0.23) 0.11 Multivariable adjusted -0.10 (-0.41, 0.21) 0.25 Further adjusted for physical activity -0.11 (-0.43, 0.20) 0.13 Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 8. Osteophytes Grade Linear model Non-linear model Cross-sectional (Year 9, n=650) Unadjusted -0.12 (-0.26, 0.01) 1.00 Multivariable adjusted -0.10 (-0.25, 0.05) 0.70 Further adjusted for physical activity -0.10 (-0.24, 0.05) 0.71 Longitudinal (Year 15, n=443) Unadjusted -0.25 (-0.44, -0.05) 0.79 Multivariable adjusted -0.26 (-0.48, -0.05) 0.99 Further adjusted for physical activity -0.26 (-0.48, -0.04) 0.90 Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 9. Disc Space Narrowing Grade Linear model Non-linear model Cross-sectional (Year 9, n=650) Unadjusted -0.07 (-0.30, 0.16) 0.06 Multivariable adjusted -0.03 (-0.29, 0.22) 0.07 Further adjusted for physical activity -0.03 (-0.29, 0.22) 0.09 Longitudinal (Year 15, n=443) Unadjusted 0.16 (-0.13, 0.44) 0.98 Multivariable adjusted 0.20 (-0.13, 0.52) 0.94 Further adjusted for physical activity 0.25 (-0.08, 0.59) 1.00 Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eAppendix 5. Sensitivity Analyses Results: Disease Severity (corresponding to the third sensitivity analysis; a new composite score based on the disease severity). eTable 10. Kellgren-Lawrence Grade Based Score Variables K/L grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=365) Number of women (%) 154 (42.2) 77 (21.1) 47 (12.9) 38 (10.4) 49 (13.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.92 (0.52, 1.63) 1.27 (0.66, 2.44) 1.14 (0.56, 2.32) 0.92 (0.47, 1.82) 0.89 0.63 Multivariable adjusted 1 (reference) 1.06 (0.59, 1.92) 1.33 (0.67, 2.65) 1.26 (0.58, 2.70) 1.34 (0.63, 2.84) 0.34 0.83 Further adjusted for 1 (reference) 1.03 (0.56, 1.88) 1.33 (0.66, 2.66) 1.31 (0.60, 2.89) 1.35 (0.63, 2.88) 0.31 0.87 physical activity Longitudinal (Year 15, n=260) Number of women (%) 112 (43.1) 60 (23.1) 34 (13.1) 27 (10.4) 27 (10.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.31 (0.70, 2.48) 1.49 (0.70, 3.17) 1.15 (0.50, 2.67) 1.25 (0.53, 2.91) 0.51 0.42 Multivariable adjusted 1 (reference) 1.32 (0.65, 2.67) 1.25 (0.52, 2.99) 1.37 (0.54, 3.50) 1.12 (0.42, 2.94) 0.61 0.47 Further adjusted for 1 (reference) 1.36 (0.66, 2.82) 1.18 (0.47, 2.92) 1.71 (0.63, 4.67) 1.31 (0.47, 3.65) 0.40 0.57 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 11. Osteophytes Grade Based Score Variables Osteophytes grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=283) Number of women (%) 258 (91.2) 20 (7.1) 1 (0.4) 3 (1.1) 1 (0.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.56 (0.20, 1.56) 2.29 (0.14, 37.06) N/A N/A 0.14 0.71 Multivariable adjusted 1 (reference) 0.49 (0.17, 1.39) 1.02 (0.06, 17.96) N/A N/A 0.13 0.94 Further adjusted for 1 (reference) 0.50 (0.17, 1.46) 0.80 (0.04, 16.41) N/A N/A 0.11 0.88 physical activity Longitudinal (Year 15, n=206) Number of women (%) 192 (93.2) 12 (5.8) 1 (0.5) 1 (0.5) 0 (0.0) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.56 (0.15, 2.09) N/A N/A N/A 0.20 0.89 Multivariable adjusted 1 (reference) 0.87 (0.19, 4.00) N/A N/A N/A 0.49 0.88 Further adjusted for 1 (reference) 1.41 (0.24, 8.33) N/A N/A N/A 0.71 0.87 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 12. Disc Space Narrowing Grade Based Score Variables Disc space narrowing grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=126) Number of women (%) 100 (79.4) 11 (8.7) 5 (4.0) 2 (1.6) 8 (6.3) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.69 (0.46, 6.24) 1.45 (0.26, 8.15) N/A 0.38 (0.04, 3.20) 0.44 0.25 Multivariable adjusted 1 (reference) 1.33 (0.32, 5.50) 2.03 (0.30, 13.90) N/A 0.53 (0.06, 5.14) N/A N/A Further adjusted for 1 (reference) 1.20 (0.26, 5.56) 3.25 (0.37, 28.69) N/A 0.59 (0.05, 6.89) N/A N/A physical activity Longitudinal (Year 15, n=87) Number of women (%) 70 (80.5) 9 (10.3) 4 (4.6) 1 (1.1) 3 (3.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.15 (0.27, 4.99) 0.53 (0.06, 4.98) 1.95 (0.12, 32.49) N/A 0.39 0.47 Multivariable adjusted 1 (reference) 1.76 (0.17, 18.44) 0.54 (0.03, 9.08) 0.02 (0.0001, N/A 0.33 0.19 3.28) Further adjusted for 1 (reference) 0.69 (0.09, 5.38) 0.06 (0.002, 1.38) 8.07 (0.12, 54.29) N/A 0.16 0.89 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eAppendix 6. Sensitivity Analyses Results: Restricting to Women With Back Pain (corresponding to the fourth sensitivity analysis; the potential heterogeneity of the population). eTable 13. Kellgren-Lawrence Grade Based Score Variables K/L grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=210) Number of women (%) 51 (24.3) 46 (21.9) 39 (18.6) 42 (20.0) 32 (15.2) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.60 (0.26, 1.38) 0.73 (0.31, 1.71) 0.66 (0.29, 1.54) 0.53 (0.21, 1.33) 0.23 0.74 Multivariable adjusted 1 (reference) 0.54 (0.15, 1.95) 0.29 (0.07, 1.17) 0.77 (0.21, 2.86) 0.21 (0.04, 1.13) 0.14 0.85 Further adjusted for 1 (reference) 0.74 (0.18, 3.10) 0.23 (0.05, 1.07) 0.76 (0.17, 3.37) 0.24 (0.04, 1.44) 0.14 0.66 physical activity Longitudinal (Year 15, n=145) Number of women (%) 40 (27.6) 32 (22.1) 24 (16.6) 26 (17.9) 23 (15.9) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.05 (0.40, 2.74) 1.16 (0.42, 3.26) 0.43 (0.14, 1.38) 1.43 (0.52, 3.94) 0.93 0.49 Multivariable adjusted 1 (reference) 0.75 (0.24, 2.36) 1.93 (0.60, 6.15) 0.56 (0.15, 2.12) 1.81 (0.54, 6.10) 0.43 0.62 Further adjusted for 1 (reference) 0.53 (0.15, 1.85) 1.57 (0.42, 5.78) 0.41 (0.09, 1.75) 1.92 (0.52, 7.06) 0.53 0.32 physical activity Adjusted for age, BMI, smoke status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 14. Osteophytes Grade Based Score Variables Osteophytes grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=210) Number of women (%) 81 (38.6) 76 (36.2) 32 (15.2) 15 (7.1) 6 (2.9) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.70 (0.36, 1.34) 0.88 (0.38, 2.02) 0.60 (0.18, 2.03) 0.35 (0.04, 3.14) 0.27 0.98 Multivariable adjusted 1 (reference) 1.05 (0.04, 2.76) 1.65 (0.48, 5.62) 0.74 (0.11, 4.83) N/A 0.79 0.24 Further adjusted for 1 (reference) 0.97 (0.35, 2.73) 1.49 (0.41, 5.41) 1.08 (0.16, 7.34) N/A 0.88 0.34 physical activity Longitudinal (Year 15, n=145) Number of women (%) 56 (38.6) 59 (40.7) 20 (13.8) 8 (5.5) 2 (1.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.77 (0.37, 1.63) 0.64 (0.22, 1.87) 0.55 (0.10, 2.88) N/A 0.18 0.74 Multivariable adjusted 1 (reference) 0.91 (0.39, 2.16) 0.68 (0.20, 2.25) 0.67 (0.10, 4.68) N/A 0.39 0.66 Further adjusted for 1 (reference) 0.90 (0.36, 2.25) 0.39 (0.10, 1.51) 0.58 (0.08, 4.29) N/A 0.18 0.65 physical activity Adjusted for age, BMI, smoke status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 15. Disc Space Narrowing Grade Based Score Variables Disc space narrowing grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=210) Number of women (%) 25 (11.9) 27 (12.9) 45 (21.4) 41 (19.5) 72 (34.3) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.94 (0.34, 2.64) 0.40 (0.15, 1.10) 0.55 (0.20, 1.47) 0.39 (0.15, 0.99) 0.03 0.58 Multivariable adjusted 1 (reference) 0.81 (0.18, 3.55) 0.29 (0.06, 1.34) 0.55 (0.10, 2.93) 0.32 (0.07, 1.35) 0.10 0.53 Further adjusted for 1 (reference) 0.43 (0.08, 2.28) 0.23 (0.04, 1.23) 0.39 (0.06, 2.38) 0.26 (0.05, 1.30) 0.19 0.32 physical activity Longitudinal (Year 15, n=145) Number of women (%) 14 (9.7) 27 (18.6) 31 (21.4) 28 (19.3) 45 (31.0) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.50 (0.15, 1.68) 0.22 (0.06, 0.79) 0.25 (0.07, 0.91) 0.83 (0.28, 2.45) 0.87 0.002 Multivariable adjusted 1 (reference) 0.60 (0.15, 2.39) 0.43 (0.09, 2.00) 0.55 (0.12, 2.45) 1.47 (0.38, 5.75) 0.23 0.04 Further adjusted for 1 (reference) 0.55 (0.11, 2.66) 0.29 (0.05, 1.61) 0.53 (0.10, 2.80) 1.28 (0.27, 6.07) 0.25 0.03 physical activity Adjusted for age, BMI, smoke status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eAppendix 7. Sensitivity Analyses Results: Changing the Model to cloglog Link Function (corresponding to the fifth sensitivity analysis; the potential model misspecification). eTable 16. Kellgren-Lawrence Grade Based Score Variables K/L grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 154 (23.7) 142 (21.8) 140 (21.5) 118 (18.2) 96 (14.8) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.16 (0.79, 1.70) 0.86 (0.57, 1.30) 0.89 (0.58, 1.37) 0.85 (0.53, 1.34) 0.24 0.82 Multivariable adjusted 1 (reference) 1.20 (0.81, 1.77) 0.86 (0.57, 1.30) 0.93 (0.60, 1.44) 0.91 (0.56, 1.46) 0.36 0.93 Further adjusted for 1 (reference) 1.18 (0.80, 1.75) 0.84 (0.55, 1.28) 0.89 (0.57, 1.39) 0.89 (0.55, 1.44) 0.30 0.99 physical activity Longitudinal (Year 15, n=443) Number of women (%) 112 (25.3) 100 (22.6) 97 (21.9) 76 (17.2) 58 (13.1) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.18 (0.75, 1.87) 0.95 (0.59, 1.55) 0.81 (0.48, 1.40) 1.07 (0.62, 1.84) 0.63 0.90 Multivariable adjusted 1 (reference) 0.89 (0.54, 1.46) 0.91 (0.55, 1.53) 0.74 (0.42, 1.31) 0.84 (0.47, 1.52) 0.28 0.60 Further adjusted for 1 (reference) 0.86 (0.52, 1.43) 0.89 (0.53, 1.49) 0.72 (0.40, 1.28) 0.83 (0.45, 1.50) 0.25 0.54 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 17. Osteophytes Grade Based Score Variables Osteophytes grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 258 (39.7) 226 (34.8) 102 (15.7) 44 (6.8) 20 (3.1) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.84 (0.62, 1.15) 0.84 (0.56, 1.26) 0.59 (0.31, 1.13) 0.83 (0.36, 1.90) 0.12 0.64 Multivariable adjusted 1 (reference) 0.87 (0.63, 1.19) 0.83 (0.55, 1.27) 0.66 (0.34, 1.30) 0.99 (0.42, 2.32) 0.29 0.50 Further adjusted for 1 (reference) 0.86 (0.63, 1.19) 0.85 (0.56, 1.31) 0.68 (0.35, 1.34) 0.97 (0.41, 2.30) 0.32 0.55 physical activity Longitudinal (Year 15, n=443) Number of women (%) 192 (43.3) 157 (35.4) 67 (15.1) 19 (4.3) 8 (1.8) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.77 (0.53, 1.11) 0.68 (0.41, 1.13) 0.49 (0.18, 1.34) 0.58 (0.14, 2.37) 0.04 0.71 Multivariable adjusted 1 (reference) 0.91 (0.62, 1.33) 0.68 (0.40, 1.16) 0.52 (0.18, 1.45) 0.48 (0.11, 2.06) 0.03 0.97 Further adjusted for 1 (reference) 0.94 (0.63, 1.39) 0.63 (0.36, 1.10) 0.55 (0.19, 1.57) 0.54 (0.12, 2.33) 0.03 0.93 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 18. Disc Space Narrowing Grade Based Score Variables Disc space narrowing grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 100 (15.4) 107 (16.5) 147 (22.6) 131 (20.2) 165 (25.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.28 (0.79, 2.10) 1.26 (0.79, 1.99) 1.31 (0.82, 2.10) 0.99 (0.62, 1.58) 0.80 0.10 Multivariable adjusted 1 (reference) 1.36 (0.83, 2.25) 1.49 (0.92, 2.39) 1.42 (0.88, 2.30) 1.10 (0.67, 1.81) 0.87 0.04 Further adjusted for 1 (reference) 1.33 (0.80, 2.21) 1.47 (0.91, 2.38) 1.40 (0.86, 2.27) 1.06 (0.65, 1.75) 0.98 0.03 physical activity Longitudinal (Year 15, n=443) Number of women (%) 70 (15.8) 84 (19.0) 102 (23.0) 88 (19.9) 99 (22.3) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.90 (0.52, 1.57) 0.79 (0.46, 1.36) 1.01 (0.59, 1.73) 1.15 (0.69, 1.93) 0.42 0.21 Multivariable adjusted 1 (reference) 0.86 (0.48, 1.54) 0.89 (0.50, 1.59) 1.08 (0.61, 1.90) 1.37 (0.78, 2.39) 0.13 0.23 Further adjusted for 1 (reference) 0.86 (0.46, 1.59) 0.84 (0.46, 1.54) 1.16 (0.64, 2.10) 1.51 (0.84, 2.73) 0.08 0.14 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eAppendix 8. Sensitivity Analyses Results: Change the Model to Linear Regression (corresponding to the fifth sensitivity analysis; the potential model misspecification). eTable 19. Kellgren-Lawrence Grade Based Score Variables K/L grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 154 (23.7) 142 (21.8) 140 (21.5) 118 (18.2) 96 (14.8) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.42 (0.59, 3.40) 0.86 (0.36, 2.06) 0.99 (0.40, 2.50) 0.65 (0.25, 1.74) 0.31 0.49 Multivariable adjusted 1 (reference) 1.44 (0.61, 3.42) 0.88 (0.37, 2.09) 1.01 (0.40, 2.55) 0.73 (0.27, 1.99) 0.43 0.55 Further adjusted for 1 (reference) 1.40 (0.59, 3.36) 0.81 (0.34, 1.94) 0.96 (0.38, 2.42) 0.69 (0.25, 1.88) 0.35 0.61 physical activity Longitudinal (Year 15, n=443) Number of women (%) 112 (25.3) 100 (22.6) 97 (21.9) 76 (17.2) 58 (13.1) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.41 (0.48, 4.09) 0.91 (0.31, 2.68) 0.56 (0.18, 1.77) 0.89 (0.25, 3.13) 0.38 0.93 Multivariable adjusted 1 (reference) 0.93 (0.36, 2.44) 1.01 (0.38, 2.67) 0.60 (0.21, 1.74) 0.54 (0.17, 1.75) 0.14 0.89 Further adjusted for 1 (reference) 0.92 (0.35, 2.43) 1.00 (0.38, 2.68) 0.59 (0.20, 1.72) 0.54 (0.16, 1.74) 0.13 0.87 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 20. Osteophytes Grade Based Score Variables Osteophytes grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 258 (39.7) 226 (34.8) 102 (15.7) 44 (6.8) 20 (3.1) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.64 (0.32, 1.27) 0.64 (0.26, 1.54) 0.35 (0.26, 1.20) 0.68 (0.12, 3.90) 0.11 0.47 Multivariable adjusted 1 (reference) 0.68 (0.35, 1.34) 0.61 (0.25, 1.48) 0.42 (0.12, 1.46) 0.92 (0.16, 5.43) 0.22 0.36 Further adjusted for 1 (reference) 0.67 (0.34, 1.32) 0.63 (0.26, 1.55) 0.43 (0.12, 1.52) 0.84 (0.14, 5.01) 0.21 0.41 physical activity Longitudinal (Year 15, n=443) Number of women (%) 192 (43.3) 157 (35.4) 67 (15.1) 19 (4.3) 8 (1.8) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.46 (0.20, 1.05) 0.28 (0.09, 0.83) 0.20 (0.03, 1.25) 0.18 (0.01, 2.96) 0.004 0.47 Multivariable adjusted 1 (reference) 0.58 (0.28, 1.24) 0.27 (0.10, 0.74) 0.32 (0.06, 1.74) 0.12 (0.01, 1.41) 0.002 0.61 Further adjusted for 1 (reference) 0.59 (0.27, 1.26) 0.27 (0.10, 0.76) 0.32 (0.06, 1.76) 0.13 (0.01, 1.63) 0.002 0.57 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 21. Disc Space Narrowing Grade Based Score Variables Disc space narrowing grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 100 (15.4) 107 (16.5) 147 (22.6) 131 (20.2) 165 (25.4) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 1.59 (0.56, 4.52) 1.68 (0.63, 4.45) 1.43 (0.53, 3.90) 0.92 (0.35, 2.39) 0.62 0.11 Multivariable adjusted 1 (reference) 1.70 (0.60, 4.82) 2.07 (0.78, 5.46) 1.52 (0.56, 4.12) 0.98 (0.36, 2.63) 0.69 0.05 Further adjusted for 1 (reference) 1.61 (0.56, 4.61) 2.00 (0.75, 5.34) 1.50 (0.55, 4.10) 0.94 (0.35, 2.56) 0.65 0.05 physical activity Longitudinal (Year 15, n=443) Number of women (%) 70 (15.8) 84 (19.0) 102 (23.0) 88 (19.9) 99 (22.3) Odds ratio (95% confidence interval) Unadjusted 1 (reference) 0.96 (0.27, 3.36) 0.75 (0.23, 2.51) 1.55 (0.45, 5.36) 1.83 (0.55, 6.15) 0.18 0.36 Multivariable adjusted 1 (reference) 0.64 (0.20, 2.01) 0.67 (0.22, 2.03) 1.22 (0.39, 3.77) 1.35 (0.43, 4.31) 0.14 0.22 Further adjusted for 1 (reference) 0.60 (0.19, 1.95) 0.63 (0.20, 1.92) 1.19 (0.38, 3.72) 1.39 (0.43, 4.49) 0.14 0.18 physical activity Adjusted for age, BMI, smoke status, back pain status and bisphosphonates usage status. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 22. Additionally Adjusted for Pain Medication and Depression (corresponding to the seventh sensitivity analysis; potential strong prognostic factors). Cross-sectional (n=650) 0 segment (n=154) 1 segment (n=142) 2 segments (n=140) 3 segments (n=118) 4 segments (n=96) Whole (n=650) Pain Medication Yes 10 (6.5) 7 (4.9) 16 (11.4) 8 (6.8) 10 (10.4) 51 (7.8) No 68 (44.2) 60 (42.3) 59 (42.1) 53 (44.9) 37 (38.5) 277 (42.6) Missing 76 (49.4) 75 (52.8) 65 (46.4) 57 (48.3) 49 (51.1) 322 (49.5) Depression Yes 0 (0.0) 1 (0.7) 0 (0.0) 1 (0.8) 1 (1.0) 3 (0.5) No 154 (100.0) 141 (99.3) 140 (100.0) 117 (99.2) 95 (99.0) 647 (99.5) Longitudinal (n=443) 0 segment (n=112) 1 segment (n=100) 2 segments (n=97) 3 segments (n=76) 4 segments (n=58) Whole (n=443) Pain Medication Yes 10 (8.9) 6 (6.0) 9 (9.3) 4 (5.3) 6 (10.3) 35 (7.9) No 42 (37.5) 42 (42.0) 48 (49.5) 35 (46.1) 24 (41.4) 191 (43.1) © 2021 Chen L et al. JAMA Network Open. Missing 60 (53.6) 52 (52.0) 40 (41.2) 37 (48.7) 28 (48.3) 217 (49.0) Depression Yes 0 (0.0) 1 (1.0) 1 (1.0) 1 (1.3) 1 (1.7) 3 (0.7) No 112 (100.0) 99 (99.0) 96 (99.0) 75 (98.7) 57 (98.3) 443 (99.3) The columns except the first correspond to the number of segments of Lumbar Spine Radiographic Changes (Kellgren-Lawrence grade based). Data are present as number (percentage) of participants unless otherwise indicated. Women reported current medication use in an open field question within the medical history questionnaire. Data on use of non-opioid and opioid analgesics, defined based on Anatomical Therapeutic Chemical codes M01 and N02, were extracted from this question from Year 9. The details are: With specific name With opioid involved Dihydrocodeine; Dextromoramide; Tramadol; Codeine; Morphine; Paracetamol and dextropropoxyphene; Paracetamol and Codeine; Paracetamol and dihydrocodeine Without opioid involved Indomethacin; Ibuprofen; Diclofenac; Etodolac; Fenbufen; Flurbiprofen; Fenoprofen; Mefenamic acid; Naproxen; Piroxicam; Ketoprofen; Movelat; Glucosamine; Feverfew; Paracetamol Without specific name NSAID; Anti-inflammatory; Analgesics; Painkillers Depression was defined by text response. From Year 1 to Year 4, women were asked the question: Serious operations/illnesses: Other? From Year 8 to Year 9, women were asked the question: Any major illnesses or operations? If the participant reported depression in at least one year (Year 1 to 9, our baseline is Year 9), we defined the value of this covariate as yes. © 2021 Chen L et al. JAMA Network Open. eTable 23. Kellgren-Lawrence Grade Based Score Variables K/L grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 154 (23.7) 142 (21.8) 140 (21.5) 118 (18.2) 96 (14.8) Odds ratio (95% confidence interval) Additional adjusted 1 (reference) 1.21 (0.61, 2.42) 0.85 (0.42, 1.75) 1.25 (0.60, 2.63) 0.87 (0.38, 2.02) 0.84 0.76 Longitudinal (Year 15, n=443) Number of women (%) 112 (25.3) 100 (22.6) 97 (21.9) 76 (17.2) 58 (13.1) Odds ratio (95% confidence interval) Additional adjusted 1 (reference) 1.10 (0.59, 2.04) 0.92 (0.49, 1.74) 0.66 (0.33, 1.33) 0.78 (0.37, 1.62) 0.14 0.86 Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, pain medication and depression. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, pain medication, depression and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 24. Osteophytes Grade Based Score Variables Osteophyte grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 154 (23.7) 142 (21.8) 140 (21.5) 118 (18.2) 96 (14.8) Odds ratio (95% confidence interval) Additional adjusted 1 (reference) 1.17 (0.67, 2.05) 1.25 (0.63, 2.48) 0.66 (0.22, 1.99) 2.46 (0.70, 8.56) 0.52 0.86 Longitudinal (Year 15, n=443) Number of women (%) 112 (25.3) 100 (22.6) 97 (21.9) 76 (17.2) 58 (13.1) Odds ratio (95% confidence interval) Additional adjusted 1 (reference) 0.87 (0.44, 1.73) 0.54 (0.22, 1.32) 0.29 (0.05, 1.63) 0.38 (0.06, 2.40) 0.04 0.87 Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, pain medication and depression. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, pain medication, depression and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 25. Disc Space Narrowing Grade Based Score Variables Disc space narrowing grade based score P for trend 0 segment 1 segment 2 segments 3 segments 4 segments Linear model Non-linear model Cross-sectional (Year 9, n=650) Number of women (%) 154 (23.7) 142 (21.8) 140 (21.5) 118 (18.2) 96 (14.8) Odds ratio (95% confidence interval) Additional adjusted 1 (reference) 1.42 (0.58, 3.45) 1.40 (0.59, 3.32) 1.48 (0.62, 3.54) 1.20 (0.50, 2.86) 0.86 0.34 Longitudinal (Year 15, n=443) Number of women (%) 112 (25.3) 100 (22.6) 97 (21.9) 76 (17.2) 58 (13.1) Odds ratio (95% confidence interval) Additional adjusted 1 (reference) 0.62 (0.21, 1.83) 0.45 (0.15, 1.35) 0.87 (0.29, 2.55) 1.07 (0.38, 3.02) 0.49 0.10 Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, pain medication and depression. Adjusted for age, BMI, smoke status, back pain status, bisphosphonates usage status, physical activity, pain medication, depression and Year 9 back pain-related disability. © 2021 Chen L et al. JAMA Network Open. eTable 26. E-value (corresponding to the sixth sensitivity analysis; potential influence from unmeasured confounding). Variable Compared with 0 segment 1 segment 2 segments 3 segments 4 segments Cross-sectional Kellgren-Lawrence grade based score Odds ratio (95%CI) 1.22 (0.76, 1.96) 0.84 (0.51, 1.38) 0.92 (0.54,1.56) 0.89 (0.50, 1.57) E-value (lower Bound) 1.44 (1) 1.41 (1) 1.25 (1) 1.31 (1) Osteophytes grade based score Odds ratio (95%CI) 0.83 (0.57, 1.22) 0.78 (0.47, 1.30) 0.58 (0.27, 1.26) 1.03 (0.37, 2.85) E-value (lower Bound) 1.43 (1) 1.52 (1) 1.95 (1) 1.14 (1) Disc space narrowing grade based score Odds ratio (95%CI) 1.43 (0.78, 2.61) 1.56 (0.88, 2.76) 1.44 (0.81, 2.57) 1.07 (0.60, 1.92) E-value (lower Bound) 1.68 (1) 1.81 (1) 1.69 (1) 1.22 (1) Longitudinal Kellgren-Lawrence grade based score Odds ratio (95%CI) 1.06 (0.57, 1.96) 0.94 (0.50, 1.76) 0.69 (0.34, 1.38) 0.83 (0.40, 1.72) E-value (lower Bound) 1.20 (1) 1.21 (1) 1.70 (1) 1.43 (1) Osteophytes grade based score Odds ratio (95%CI) 0.76 (0.47, 1.24) 0.53 (0.28, 1.02) 0.49 (0.14, 1.70) 0.31 (0.06, 1.72) E-value (lower Bound) 1.56 (1) 2.09 (1.21) 2.21 (1) 2.99 (1) Disc space narrowing grade based score Odds ratio (95%CI) 0.72 (0.34, 1.53) 0.74 (0.36, 1.52) 1.06 (0.52, 2.20) 1.26 (0.62, 2.57) E-value (lower Bound) 1.64 (1) 1.60 (1) 1.20 (1) 1.42 (1) We calculated E-value through Online Calculator (https://mmathur.shinyapps.io/evalue/) based on results from step 2 of the stepped modelling framework. Explanation: for an unmeasured confounder to explain the OR estimate of 1.22, the unmeasured confounder would have to be associated with both the exposure and the outcome by 1.44-fold above and beyond the measured confounders. © 2021 Chen L et al. JAMA Network Open.

Journal

JAMA Network OpenAmerican Medical Association

Published: May 20, 2021

References