Format

Send to

Choose Destination
Osteoporos Int. 2007 Jan;18(1):35-43. Epub 2006 Sep 2.

Use of clinical risk factors to identify postmenopausal women with vertebral fractures.

Author information

1
Department of Clinical Science at South Bristol, University of Bristol, Bristol, BS2 8HW, UK. Jon.Tobias@bristol.ac.uk

Abstract

INTRODUCTION AND HYPOTHESIS:

Previous studies have been unable to identify risk factors for prevalent vertebral fractures (VF), which are suitable for use in selection strategies intended to target high-risk sub-groups for diagnostic assessment. However, these studies generally consisted of large epidemiology surveys based on questionnaires and were only able to evaluate a limited number of risk factors. Here, we investigated whether a stronger relationship exists with prevalent VF when conventional risk factors are combined with additional information obtained from detailed one-to-one assessment.

METHODS:

Women aged 65-75 registered at four geographically distinct GP practices were invited to participate (n=1,518), of whom 540 attended for assessment as follows: a questionnaire asking about risk factors for osteoporosis such as height loss compared to age 25 and history of non-vertebral fracture (NVF), the get-up-and-go test, Margolis back pain score, measurement of wall-tragus and rib-pelvis distances, and BMD as measured by the distal forearm BMD. A lateral thoraco-lumbar spine X-ray was obtained, which was subsequently scored for the presence of significant vertebral deformities.

RESULTS:

Of the 509 subjects who underwent spinal radiographs, 37 (7.3%) were found to have one or more VF. Following logistic regression analysis, the four most predictive clinical risk factors for prevalent VF were: height loss (P=0.006), past NVF (P=0.004), history of back pain (P=0.075) and age (P=0.05). BMD was also significantly associated with prevalent VF (P=0.002), but its inclusion did not affect associations with other variables. Factors elicited from detailed one-to-one assessment were not related to the risk of one or more prevalent VFs. The area under ROC curves derived from these regressions, which suggested that models for prevalent VF had modest predictive accuracy, were as follows: 0.68 (BMD), 0.74 (four clinical risk factors above) and 0.78 (clinical risk factors + BMD). Analyses were repeated in relation to the subgroup of 13 patients with two or more VFs, which revealed that in this instance, the Margolis back pain score and rib-pelvis distance were associated with the presence of multiple VFs (P=0.022 and 0.026, respectively). Moreover, the predictive value as reflected by the ROC curve area was improved: 0.80 (BMD), 0.88 (the four most predictive clinical risk factors consisting of the height loss, past NVF, Margolis back pain score and rib-pelvis distance) and 0.91 (clinical risk factors + BMD).

CONCLUSIONS:

Evaluation of additional risk factors from detailed one-to-one assessment does not improve the predictive value of risk factors for one or more prevalent vertebral deformities in postmenopausal women. However, the use of factors such as the Margolis back pain score and rib-pelvis distance may be helpful in identifying postmenopausal women at high risk of multiple prevalent VFs.

PMID:
16951907
DOI:
10.1007/s00198-006-0209-8
[Indexed for MEDLINE]

Supplemental Content

Full text links

Icon for Springer
Loading ...
Support Center