The Association of Radiographically Detected Vertebral Fractures with Back Pain and Function: A Prospective Study

Methods

Participants

Study patients were participants in the Study of Osteoporotic Fractures, a cohort recruited from population-based listings in four U.S. metropolitan areas. Details of the design of this study are published elsewhere [17]. Lateral spine radiographs were obtained for 9677 white women between the ages of 65 and 99 years (median age, 70 years) who underwent baseline examination between 1986 and 1988. Repeated spinal radiographs suitable for morphometry were obtained for 7223 women (75% of the original cohort) at a follow-up clinic visit held an average of 3.7 years (range, 1.3 to 5.1 years) later. All participants gave informed consent.

Vertebral Morphometry

Lateral radiographs of the thoracic and lumbar spine were obtained in accordance with current guidelines [18]. Quantitative vertebral morphometry was performed using six-point digitization as described elsewhere [3, 19] to calculate the anterior (H_a), mid- (H_m), and posterior (H_p) height for each vertebral body from T4 to L4. A system of triage of radiographs, described elsewhere [3, 20], was used to reduce the number of radiographs requiring morphometric measurements. Briefly, trained technicians separated sets of radiographs into normal, uncertain, or probably fractured groups on the basis of a limited semiquantitative grading scheme that categorized women by the most abnormal vertebral level [20]. Uncertain grades were further categorized by the study radiologist as normal or probably fractured. Morphometry was done on the radiograph pairs that were categorized as probably fractured (42%). In a random sample of 503 women whose radiographs were triaged and then digitized, triage missed no incident fractures according to the study definition.

Definition of Vertebral Fracture

A vertebra was classified as having a prevalent fracture on the baseline radiograph if any of the following ratios were more than 3 SDs (>4 SDs for severe fractures) below the normal mean for that vertebral level: (H_a/H_p), (H_m/H_p), or a combination of (H_π/H [_π] ± 1) and (H_ai/H_ai ± 1) [3, 21]. A new (incident) fracture was identified if any of the three vertebral heights (H_a, H_m, or H_p) on follow-up radiographs decreased by 20% or more and by at least 4 mm compared with the baseline height. Incident fractures identified by morphometry were reviewed by a radiologist to exclude imaging artifacts or such conditions as osteophytosis and Scheuermann disease; 7% of vertebrae meeting the morphometric criteria for incident fracture were reclassified as not fractured.

Incident Clinical Fractures

We used previously described methods [22] to assess the occurrence of clinical fractures of any bone during follow-up. Women were considered to have a clinical vertebral fracture if they reported a new diagnosis of spinal fracture and a clinical radiology report confirmed that a vertebral fracture was present.

Measurements of Pain, Disability, and Limited Activity

We evaluated outcome measures by using a previously described questionnaire [7, 23] that asked about back pain and back-related disability in the past 12 months and the number of days of limited activity due to back pain. The questionnaire was administered at baseline and at three annual follow-up contacts held before assessment of vertebral fractures. The third follow-up contact coincided with follow-up radiography.

Back pain was assessed on scales of frequency (0, never or rarely; 1, some of the time; 2, most of the time; or 3, all of the time) and severity (0, no pain; 1, mild pain; 2, moderate pain; or 3, severe pain). The two pain questions had high internal consistency (Cronbach α = 0.81) and were summed for a total score that could range from 0 to 6. We defined clinically significant back pain as pain that was experienced “most” or “all of the time” or pain that was “moderate” or “severe.” Women without significant back pain at baseline were considered to have increased back pain if clinically significant pain had developed between any follow-up contacts. For women with clinically significant back pain at baseline, increased back pain was defined as an increase in total pain score of at least two points. Both types of increase had a similar association with incident fractures and thus were combined for a single outcome.

Back-related disability was assessed with questions about the degree of difficulty (0, no difficulty; 1, some difficulty; 2, much difficulty; or 3, unable to perform activity) in six activities of daily living that involved the back (bending down to pick up light-weight objects, lifting a 10-pound object from the floor, reaching for objects just above the head, putting on socks or stockings, getting in and out of an automobile, and standing for 2 hours). These measures were combined in a back-related disability score ranging from 0 to 18. As reported elsewhere [7], this scale has high internal consistency (Cronbach α = 0.82) and is highly correlated (Spearman r = 0.73) with a more extensive instrument used to assess disability caused by low back pain [24]. We defined clinically significant disability as “much difficulty” or “unable” in one or more of the six activities. Women without significant disability at baseline were considered to have increased disability if clinically significant disability had developed between any follow-up contacts. For women with clinically significant disability at baseline, increased disability was defined as an increase in disability score of at least three points. Both types of increase had a similar association with incident fractures and thus were combined for a single outcome.

We also asked participants if they had limited their activities because of back pain since the last contact; if the answer was yes, we asked for the number of days they had stayed in bed and the number of days on which activity was limited (not including days in bed) because of back pain. Questions were adapted from previous surveys [25, 26]. For all follow-up contacts, we summed the number of days of bed rest and, in a separate measure, the number of days of limited activity; we then divided these numbers by the total years of follow-up to estimate the average number of affected days per year.

Other Measurements

The baseline questionnaire assessed potential confounding factors that may be associated with the risk for incident vertebral fracture and with back pain or disability, including smoking (current or past smoker); “inactivity,” defined as walking less than one block daily (yes or no); a previous physician diagnosis of osteoporosis or spinal fracture (yes or no); current use of estrogen (yes or no); hip pain in the past 12 months (yes or no); and height at 25 years of age. At the baseline examination, we assessed height and weight and calculated body mass index (kg/m²). We assessed grip strength by using an isometric dynamometer (Jamar Hydraulic Hand Dynamometer, JA Preston, Jackson, Mississippi) at baseline and at the follow-up examination and calculated change in grip strength between the two measurements. A random sample of 16% of baseline spine radiographs was assessed for spinal disc degeneration by using previously published methods [27].

Statistical Analysis

Unless otherwise indicated, analyses were done separately in groups stratified by the presence of one or more baseline prevalent vertebral fractures. Descriptive and bivariate associations were assessed by using the t-test for continuous variables and the chi-square test for dichotomous variables. The association between incident vertebral fractures and dichotomous outcomes (increased back pain and increased back disability) was analyzed with logistic regression techniques.

We analyzed the association of incident vertebral fracture with days of bed rest and days of limited activity per year by using Poisson regression. The distribution of days of bed rest (mean ±SD, 0.44 ± 5.15) and limited-activity days (16.3 ± 53.7) indicate that considerable overdispersion is present. Poisson regression allowing for this overdispersion provides a good estimation and inferential scheme [28]. To correct CIs and P values for overdispersion of the number of days, we estimated an overdispersion (scale) parameter as the deviance divided by degrees of freedom for number of days of bed rest (3.36) and number of limited-activity days (17.1). We examined the model's goodness of fit by using both deviance and Pearson chi-square criteria; for both statistics and both outcomes, model adequacy pertained. The β estimates from Poisson regression are interpretable as relative risks on the underlying event rate [28]. To estimate the number of additional days of bed rest and limited-activity days due to incident vertebral fractures for 1 year, we calculated predicted values for the number of days from Poisson models with a dichotomous variable for incident vertebral fractures set to 0 or 1 and took the difference as the excess number of days [28]. Values of all other variables were fixed at their mean values in the study sample. Examination of each outcome revealed that the principal cause of overdispersion was an excess of zeroes. To assess the effect of this on our results, we fit logistic regression models using 0 and more than 0 indicators for days of bed rest and limited-activity days and found that the association of incident fractures with these outcomes was consistent with the association in the Poisson models. All multivariate models included adjustments for clinic site, age, body mass index, decline in grip strength, incident nonspinal fractures, a previous physician diagnosis of osteoporosis, inactivity, current smoking, and current use of estrogen. Additional adjustment for height loss from age 25 years to baseline, baseline hip pain, and radiographic measures of spinal disc degeneration did not change our results.

Results

Incident Vertebral Fractures

Baseline prevalent fractures were present in 1416 (20%) of the 7223 women studied. During a mean (±SD) 3.7 ± 0.4 years of follow-up, incident vertebral fractures occurred in 5.1% of the cohort (annual risk, 1.4% [95% CI, 1.1% to 1.7%]) but were about four times more frequent (n = 193 [13.8%]) in women with a prevalent vertebral fracture at baseline than in those without a baseline fracture (n = 178 [3.0%]). Women with one or more incident vertebral fractures had an average of 1.8 ± 0.9 new fractures. Women with a new vertebral fracture during the study were older and thinner, had a greater decline in grip strength, and had lost more height than women with no new fracture (Table 1).

Increases in Back Pain and Disability

Among women with no prevalent or incident vertebral fractures, 23% had an increase in back pain and 15% had an increase in disability during follow-up compared with baseline levels. After adjustment for covariates, women with a first incident fracture had a 2.4 times higher risk for increased back pain and women with a recurrent fracture (that is, women with a baseline prevalent fracture who had a new fracture) had a twofold higher risk for increased back pain compared with women without a new fracture during follow-up (Table 2). The association between one or more incident vertebral fractures and increases in back-related disability during follow-up was similar to the results for back pain (Table 2). The risk for increased back pain did not differ substantially by the number of incident fractures. In contrast, multiple new fractures had a substantially greater effect on the risk for increased disability than did a single new fracture.

Days Spent in Bed and Limited-Activity Days

Of women with no prevalent or incident vertebral fractures, 4% had at least 1 day of bed rest per year and 13% had 7 or more days of limited activity due to back pain per year. After adjustment for covariates, the annual rate of days of bed rest was 9 times higher in women with a first incident fracture and 8 times higher in those with a recurrent incident fracture than it was in women without a new fracture. The annual rate of limited-activity days was about twice as high in women with a first incident fracture or those with a recurrent incident fracture (Table 3). A woman with a first vertebral fracture had 11 additional limited-activity days and 1.7 additional days of bed rest, whereas a woman with a recurrent fracture had 10 additional days of limited activity and 1.1 additional days of bed rest per year compared with women with no incident fracture. Two or more new fractures were associated with an additional 32 limited-activity days and 5.3 days of bed rest per year in women without a baseline fracture and 14 limited-activity days and 2.3 days of bed rest per year in those with a baseline fracture.

Incident Clinical Spine Fractures Compared with Incident Morphometric Fractures

Clinical spine fractures during 3.7 years of follow-up were confirmed in 101 (27%) of women who had an incident morphometric fracture; these women had a high risk for increased back-related disability (Table 4). In addition, women with new morphometric fractures who did not report a clinical spinal fracture had about a twofold higher risk for increased back pain, disability, at least 1 day of bed rest, and at least 7 limited-activity days. A new vertebral fracture occurred in 11% of women with increased back disability, 15% of those with at least 7 limited-activity days, and 22% of those with at least 1 day of bed rest per year.

Baseline Prevalent Fracture and Clinical Impact

Women with vertebral fractures at baseline had a high level of baseline back problems (Table 1). However, even severe baseline prevalent fractures (≥ 4 SDs) were only modestly associated with increased back pain (odds ratio, 1.1 [CI, 0.9 to 1.4]), increased disability (odds ratio, 1.2 [CI, 1.0 to 1.4]), at least 1 day of bed rest per year (odds ratio, 1.2 [CI, 0.9 to 1.7]), and 7 or more limited-activity days per year (odds ratio, 1.2 [CI, 1.0 to 1.4]) during 3.7 years of follow-up. Among women with baseline prevalent fractures, only those who also had an incident fracture had an increased risk for back pain, disability, and functional limitation during the final 12 months of follow-up compared with women without any vertebral fractures (Table 5).

Discussion

We found that both first and recurrent radio-graphically detected incident vertebral fractures in elderly white women were associated with substantial increases in back pain and back-related disability compared with levels before fracture. In addition, women with incident fractures had about 1 to 2 additional days of bed rest and 10 additional days of limited activity due to back pain per year during almost 4 years of follow-up. Each additional new vertebral fracture was associated with a further increase in functional limitation. However, even severe baseline prevalent fractures were not associated with increased back pain or functional limitation during follow-up in women who did not have a subsequent new fracture.

Slightly more than two thirds of new radiographically detected vertebral fractures were not diagnosed clinically, a finding consistent with those of previous studies [15, 16]. New fractures that did not come to medical attention were nevertheless associated with a two- to threefold increase in back pain and functional limitation. Studies that include only clinically diagnosed fractures will underestimate the impact of vertebral fractures in the population. Our results support the use of radiographically defined incident vertebral fractures as study end points and suggest that such fractures have important implications for function and quality of life. Our findings indicate that 10% to 20% of older women who have increased functional limitation due to back pain have had a new vertebral fracture and that these fractures are often unrecognized by the patient and her physician. Because effective treatment is available to decrease the risk for subsequent fractures in women with radiographically detected vertebral fractures [15, 29], an evaluation for osteoporosis may be indicated in older women who have new or worsening back pain and disability.

Our estimate of an additional 10 days per year of limited activity due to back pain during almost 4 years of follow-up in women with new fractures is similar to the estimated increase in annual days of limited activity for patients with diabetes (15 days), ischemic heart disease (15 days), and arthritis and rheumatism (7 days) [30]. Women with no baseline vertebral fractures and 2 or more new fractures had more than 5 additional days of bed rest and more than 30 additional days of limited activity per year. The number of limited-activity days and days of bed rest are global measures of physical disability [25] that are responsive to the occurrence of illness or injury [26]. Our results suggest that limited-activity days and, in particular, days of bed rest that are attributed to back pain are responsive to the occurrence of vertebral fractures and should be considered for use as outcome measures in clinical trials of osteoporosis treatment.

Our study has several important strengths. It was large enough to separately assess the effect of first and recurrent vertebral fractures in a general population sample. It is also the first to describe the substantial impact of the first occurrence of radio-graphically detected fractures. Previous reports studied clinically identified patients with vertebral fractures [31-34] or combined persons with and without prevalent fracture [8, 13]. We assessed back pain and function annually during the same follow-up period in which new vertebral fractures were assessed, and we defined outcomes in terms of changes in our measures during follow-up. This design enhanced our ability to detect the effect of new fractures against a high background level of other back problems in older women. Our definition of incident fractures is similar to that used in recent clinical trials of osteoporosis treatment [15, 35, 36].

Reports of patients presenting with clinically recognized vertebral fractures suggest that acute bone pain is usually self-limited and resolves within several months [12, 37]. However, in a previous cross-sectional analysis of women in the Study of Osteoporotic Fractures [7], we found that severe (≥ 4 SDs) prevalent fractures were associated with chronic back problems. In addition, a more recent study by Huang and colleagues [13] suggests that increased back pain and disability from new fractures may persist for several years in many women. In our study, even women with a severe baseline prevalent fracture did not have increased levels of back pain and disability 3 to 4 years later unless a new fracture had occurred in the interim. Thus, chronic back problems in women with vertebral osteoporosis may result, at least in part, from the continued occurrence of new vertebral fractures. Treatments that decrease the risk for recurring fractures should reduce chronic back problems in women with osteoporosis.

Several aspects of our methods deserve comment. We studied only white women, a group with an especially high risk for vertebral osteoporosis; our results may not apply to men and to other ethnic groups. We cannot determine when the incident morphometric or clinical vertebral fractures occurred; this limited our ability to examine the temporal relation between new fractures and outcomes. If decline in grip strength is caused by new vertebral fractures, adjustment for this decline could have caused us to underestimate the association of new fractures with outcomes. Our questions about back pain used an ordinal response format with a limited range and may be insensitive to small changes or exhibit ceiling effects in women with severe pain at baseline. We did not assess certain psychosocial or quality-of-life outcomes, such as fear of falling or depression, which may be adversely affected by vertebral fracture [38]. Nevertheless, back pain and difficulty with physical function involving the back are rated important domains of quality of life by women with osteoporosis [32]. Outcome measures analyzed for our study are based on self-report; direct assessment of physical function may identify additional, independent areas of impairment [33].

In conclusion, back problems are common in older white women without vertebral fractures. Nevertheless, both first and recurrent incident vertebral fractures substantially increase back pain and functional limitation due to back pain above these high background levels. The association of prevalent fractures with chronic back pain and disability may be largely explained by the fourfold increased risk for new vertebral fractures in women who already have a fracture. Because each additional new vertebral fracture results in a further decrease in function, prevention of recurrent fractures done using proven therapies should reduce the burden of back-related disability in women with spinal osteoporosis. Our results suggest that most new vertebral fractures, even painful ones, remain unrecognized by women and their physicians. Women with new vertebral fractures that do not come to medical attention may benefit from effective treatments for osteoporosis that can reduce the risk for future fractures, back pain, and disability. Ten percent to 20% of elderly women with increased functional limitation due to back pain may have had a new vertebral fracture. An evaluation for vertebral osteoporosis may be warranted in such women.

From the University of California, San Francisco, San Francisco, California; Kaiser Permanente Medical Care Program, Oakland, California; and Minneapolis Veterans Affairs Medical Center and University of Minnesota School of Public Health, Minneapolis, Minnesota.

Dr. Ettinger: Division of Research, Kaiser Permanente Medical Care Program, 3505 Broadway, Oakland, CA 94611.

Dr. Ensrud: General Medicine (111-0), Minneapolis Veterans Affairs Medical Center, One Veterans Drive, Minneapolis, MN 55417.

Dr. Genant: Department of Radiology, Box 0628, University of California, San Francisco, CA 94143.