What Proportion of Incident Radiographic Vertebral Deformities Is Clinically Diagnosed and Vice Versa?

Authors

  • Howard A Fink MD, MPH GRECC,

    Corresponding author
    1. Geriatric Research Education and Clinical Center, VA Medical Center, Minneapolis, Minnesota, USA
    2. Section of General Internal Medicine, VA Medical Center, Minneapolis, Minnesota, USA
    3. Center for Chronic Disease Outcomes Research, VA Medical Center, Minneapolis, Minnesota, USA
    4. Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis, Minnesota, USA
    • 11-G VA Medical Center, One Veterans Drive, Minneapolis, MN 55417, USA
    Search for more papers by this author
  • Donna L Milavetz,

    1. Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis, Minnesota, USA
    Search for more papers by this author
  • Lisa Palermo,

    1. Department of Epidemiology and Biostatistics, University of California, San Francisco, California, USA
    Search for more papers by this author
  • Michael C Nevitt,

    1. Department of Epidemiology and Biostatistics, University of California, San Francisco, California, USA
    Search for more papers by this author
  • Jane A Cauley,

    1. Department of Epidemiology, University of Pittsburgh, Pittsburgh, Pennsylvania, USA
    Search for more papers by this author
    • Dr Cauley received support form Eli Lilly and Company, Pfizer, and Novartis. She received an honorarium from Eli Lilly and Company, Merck, and Novartis and served as a speaker for Eli Lilly and Company and Merck. Dr Black serves as a consultant for Merck, NPS Pharmaceuticals, and Novartis and receives funding from Merck and Novartis. All other authors have no conflict of interest.

  • Harry K Genant,

    1. Department of Radiology, University of California, San Francisco, California, USA
    Search for more papers by this author
  • Dennis M Black,

    1. Department of Epidemiology and Biostatistics, University of California, San Francisco, California, USA
    Search for more papers by this author
  • Kristine E Ensrud

    1. Section of General Internal Medicine, VA Medical Center, Minneapolis, Minnesota, USA
    2. Center for Chronic Disease Outcomes Research, VA Medical Center, Minneapolis, Minnesota, USA
    3. Division of Epidemiology, School of Public Health, University of Minnesota, Minneapolis, Minnesota, USA
    Search for more papers by this author

Abstract

We prospectively examined, in a large cohort of older women, the proportion of incident radiographic vertebral deformities diagnosed as incident clinical vertebral fractures in the same women at the same vertebral level. The proportion of deformities clinically diagnosed ranged from <15% for milder deformities to nearly 30% for more severe deformities.

Introduction: The relationship between radiographic and clinical vertebral fractures is incompletely understood. No previous study has prospectively compared the agreement between incident radiographic vertebral deformities and incident community-recognized, radiographically confirmed vertebral fractures in the same women at the same vertebral level(s).

Materials and Methods: This analysis of data from the Fracture Intervention Trial included all participants who completed both baseline and at least one scheduled follow-up lateral spinal radiograph (n = 6084). Incident vertebral deformities were defined at a given vertebral level as a reduction between baseline and closeout radiographs of ≥20% and 4 mm in any vertebral height and subdivided into two severity categories. Incident clinical vertebral fractures were those reported to clinical centers by participants and confirmed by the study radiologist, who compared the community spinal radiograph with the participant's baseline study radiograph using semiquantitative methods.

Results: A total of 446 incident radiographic vertebral deformities were identified in 330 women, whereas 121 women experienced one or more confirmed incident clinical vertebral fracture. Of incident radiograpic vertebral deformities, 22.6% were also clinically diagnosed as incident vertebral fractures, with clinical diagnoses made for 28.4% of the deformities that exceeded 30% and 4 mm height loss (severe deformity) compared with 14.3% for deformities that involved ≥20% and 4 mm but <30% height loss (milder deformity). Of incident clinical vertebral fractures, 72.7% were morphometrically identified as incident deformities, most of them as severe deformities. More than 20% of incident clinical fractures were not identified as incident deformities by even the most liberal morphometric criterion used in this study.

Conclusions: Approximately one-fourth of incident radiographic vertebral deformities were clinically diagnosed as new vertebral fractures, although the proportion clinically diagnosed was increased for more severe deformities. Whereas most incident clinical vertebral fractures were identified as severe morphometric deformities, approximately one-fourth did not meet even the most liberal study criterion for morphometric deformity. Further study of factors that may explain the discordance between incident vertebral deformities and incident clinical vertebral fractures is important.

INTRODUCTION

VERTEBRAL FRACTURES have long been recognized as perhaps the most common manifestation of osteoporosis.(1) Clinically recognized vertebral fractures are associated with lower health-related quality of life,(2) an increased risk of subsequent clinical fractures,(3) and increased mortality.(4,5) Incident radiographically identified vertebral deformities, whether clinically recognized or not, are associated with an increase in back pain and disability(6) and with lower health-related quality of life.(7)

Nevertheless, defining what constitutes a vertebral fracture remains problematic. In the research setting, incident vertebral fractures are identified, irrespective of clinical symptoms, by standardized assessments of change in vertebral body height or shape on serial spine radiographs using semiquantitative visual criteria (SQ),(8) quantitative morphometric measurements (QM),(9–12) or a combination of these methods. No approach has been established as the consensus gold standard. In the clinical setting, vertebral fractures are usually identified when a patient presents with back pain and a spinal radiograph is interpreted as showing a fracture of a vertebral body.(1) However, it is impossible in many cases to determine whether this finding constitutes a new or recent vertebral fracture, particularly when no prior spinal radiograph is available. Accurate clinical diagnosis of a new vertebral fracture may be confounded by the high population prevalence of back pain from other causes, by changes in vertebral morphology that are longstanding or caused by nonfracture processes, or by nonstandardized interpretation of spinal radiographs.

Some have suggested that newly identified radiographic changes in vertebral shape or height should be referred to as “vertebral deformities” and that the term “vertebral fracture” should refer only to incident vertebral deformities that are associated with clinical symptoms.(13) Such a classification scheme may be supported by findings in several studies that most radiographically detected changes in vertebral morphology do not come to clinical attention.(10,14) To our knowledge, however, no published study has prospectively compared the agreement between incident radiographic vertebral deformities and incident community-recognized, radiographically confirmed clinical vertebral fractures in the same women at the same vertebral level(s).

Studies examining the association of incident vertebral deformities with clinical validation criteria have found that deformity definitions requiring a greater loss of vertebral height or that use more stringent criteria (both of which yield a lower incidence of deformities) are more strongly associated with concurrent loss of stature and increased back pain(11) and are more strongly related to known risk factors for vertebral fracture.(11,12) Conversely, milder deformities may not be associated with clinical symptoms,(15) and in some cases, may represent measurement errors caused by the inherent variability in vertebral height measurement.(16) Whereas it seems plausible that more severe new deformities, which are more likely to cause symptoms or height loss, are also more likely to be diagnosed clinically, this has not been studied.

Therefore, to more precisely determine the proportion of incident radiographic vertebral deformities in older women that also are clinically diagnosed, we used data from the Fracture Intervention Trial (FIT). Using systematically collected FIT data on both incident clinical vertebral fractures and incident radiographic vertebral deformities, we calculated the proportion of incident vertebral deformities that were diagnosed as incident clinical vertebral fractures and the proportion of incident clinical vertebral fractures that were identified as new radiographic vertebral deformities. We evaluated the impact on the relationship between these entities of using alternative radiographic criteria for the identification of new deformities.

MATERIALS AND METHODS

Study design and participants

This study is a secondary analysis of data collected in the FIT, a randomized, double-blind, placebo-controlled multicenter trial designed to determine whether alendronate treatment reduced fracture incidence in women with low femoral neck BMD. The trial was conducted at 11 clinical centers in the United States and included 6459 community-dwelling postmenopausal women 55-81 years of age with a femoral neck BMD of 0.68 g/cm2 or less (T score of −1.6 or less, as measured by DXA with Hologic QDR 2000 densitometers; Hologic, Bedford, MA, USA). At baseline, women were assigned to the vertebral deformity study arm if they had at least one prevalent vertebral deformity noted on their baseline radiograph (n = 2027) or to the clinical fracture study arm if no vertebral deformities were noted (n = 4432).

Participants were recruited principally from mass mailings. Women were excluded from participating in the trial if they had unstable major medical conditions, recent use of medications affecting bone turnover such as estrogen, secondary causes of osteoporosis, significant active upper gastrointestinal ulcer disease, or chronic dyspepsia treated on a daily basis with medication. The institutional review boards at all participating centers approved the study protocol. Written informed consent was obtained from all participants. FIT participants were randomly assigned to double-blind treatment with alendronate (5 mg/day for 2 years and 10 mg/day thereafter) or placebo. Follow-up visits were scheduled at 3, 6, and 12 months and every 6 months thereafter. Participants were contacted by telephone at alternating 3-month intervals to inquire about adverse experiences, including fractures. The study design and methods including inclusion/exclusion criteria have been outlined in detail elsewhere.(17)

The present analyses included all 6084 women (94% of randomized participants) who completed a baseline and at least one scheduled follow-up lateral spinal radiograph.

Ascertainment of radiographic vertebral deformities

Lateral thoracic and lumbar spinal radiographs were obtained in study clinics in accordance with standard guidelines(16) at study baseline and several scheduled follow-up time-points, at 24 and 36 months for subjects in the vertebral deformity study arm, and at 48 months for subjects in the clinical fracture study arm.

The determination of a prevalent vertebral deformity at baseline was made at the coordinating center using a combination of semiquantitative assessment(8) and vertebral morphometry by digitization.(17) Radiographs classified semiquantitatively as having a mild or greater deformity were assessed by vertebral morphometry, in which vertebral heights measured on an electronic digitizing board were used to determine, for each vertebra, its height ratios (anterior to posterior height, middle to posterior height, and the ratio of its posterior height to the posterior heights of the adjacent vertebrae). A prevalent vertebral deformity was considered present if any of these vertebral height ratios was more than 3 SD below the mean population norm for that vertebral level.(18)

Incident radiographic vertebral deformities were identified morphometrically by comparing the baseline and closeout spinal radiographs. For these analyses, incident vertebral deformities were primarily defined as a reduction between baseline and closeout radiographs, at a given vertebral level, of 20% or more in either the anterior, middle, or posterior vertebral height(s) with at least a 4-mm decrease in the corresponding vertebral height(s). Incident vertebral deformities then were subdivided by severity into those with at least 20% but <30% loss in any vertebral height and those with at least 30% loss in any vertebral height. In a secondary analysis, the definition of incident vertebral deformity was expanded to include vertebrae with a height loss between baseline and closeout of at least 15%.19;20 All assessments were blinded to study arm and treatment assignment.

Ascertainment of clinical vertebral fractures

Throughout the duration of the trial, participants were queried about new fractures every 3 months, during in-person interviews at semiannual clinic visits and by telephone interviews at alternating 3-month intervals. Clinical vertebral fractures were those that were diagnosed in the community setting by local physicians, reported to the clinical centers by the participants, and radiographically confirmed at the study coordinating center by the study radiologist. A copy of the radiograph obtained by the participant's community physician was sent to the study coordinating center and compared with the baseline study spinal radiograph by the study radiologist, who confirmed the presence or absence of an incident clinical vertebral fracture using a standard semiquantitative method.(8) Specifically, vertebrae were visually inspected without direct vertebral measurement and were graded as normal, mildly deformed (20-25% reduction in any vertebral height and 10-20% reduction in vertebral area), moderately deformed (25-40% reduction in any vertebral height and 20-40% reduction in vertebral area), or severely deformed (40% reduction in any vertebral height and area). For our analyses, a vertebra was considered to have an incident clinical fracture if it was graded as mildly deformed or worse and the deformity was a higher grade than was present in the baseline study radiograph. The date of clinical vertebral fracture occurrence was considered the date the community radiograph was performed, regardless of the timing of onset of any symptoms that may have prompted evaluation. All confirmed incident clinical vertebral fractures that occurred after the completion of a participant's baseline study spinal radiograph and before the completion of her final scheduled study spinal radiograph were considered for primary analyses. All assessments were blinded to study arm and treatment.

Statistical analyses

The proportion of incident morphometric vertebral deformities that were also diagnosed as clinical vertebral fractures at the corresponding vertebral level was calculated. First, this was performed for all incident morphometric deformities considered as a group and then separately for each severity category of incident deformity. Likewise, the proportion of incident clinical vertebral fractures that also were identified as incident morphometric vertebral deformities at the same vertebral level was calculated. This was performed for all incident morphometric deformities considered as a group, after which the proportion of clinical fractures that fell into each morphometric deformity severity category was determined. Comparisons of the proportion of incident morphometric vertebral deformities that were identified and confirmed as clinical vertebral fractures were performed between different deformity severity categories using χ2 tests.

Because individual women may have had multiple incident morphometric deformities and/or incident clinical vertebral fractures, we calculated the proportion of women with incident vertebral deformities that were also diagnosed with incident clinical vertebral fracture(s) at the same vertebral level(s) and vice versa. All comparisons were conducted within women for all incident morphometric deformities considered as a group and separately for each severity category of incident deformity. Again, comparisons of the proportion of women with incident morphometric vertebral deformities that were identified and confirmed as clinical vertebral fractures were made between different categories of deformity severity using χ2 tests.

In a sensitivity analysis, women who had confirmed incident clinical vertebral fractures but were excluded from the primary analyses, because they either were not diagnosed until after completion of their final scheduled study spinal radiograph (n = 12) or they did not complete a final study spinal radiograph (n = 12), were considered to have an incident clinical vertebral fracture and an incident morphometric vertebral deformity at the same vertebral level. In addition, the primary analyses above were repeated after stratification of women by the presence or absence of prevalent morphometric vertebral deformities at the baseline exam.

RESULTS

Of the 6459 women enrolled in FIT, 6084 women completed both baseline and at least one scheduled follow-up lateral spinal radiographs (mean, 3.8 ± 0.54 years between baseline and final examinations; range, 1.9-4.8 years). Baseline characteristics of these 6084 women are shown in Table 1 and did not significantly differ from those of the overall cohort (data not shown). Of 379 women not completing closeout radiographs, 122 (32%) died before the closeout visit.

Table Table 1.. Baseline Characteristics of 6084 Participants*
original image

Among the 6084 women eligible for these analyses, 330 (5.4%) were identified as having at least one new radiographic deformity involving a loss of at least 20% and 4 mm in any vertebral height between baseline and their final scheduled follow-up lateral spinal radiograph. A total of 446 incident radiographic deformities were identified in these 330 women. Using a criterion to define an incident vertebral deformity of any vertebral height loss of at least 15% between the baseline and final scheduled follow-up spinal radiograph,(19,20) 453 women had a total of 659 incident radiographic vertebral deformities.

Also, among the 6084 eligible women, 170 (2.8%) self-reported a clinical vertebral fracture after completing their baseline spinal radiograph. Of these 170 women, 10 were excluded from the primary analyses for having no community spinal radiograph performed and 39 because their community diagnosed clinical vertebral fracture was not confirmed when centrally reviewed. Of the remaining 121 women whose community diagnosed clinical vertebral fractures were centrally confirmed, 12 were not diagnosed in the community until after completion of the study closeout spinal radiograph. In 12 additional cases, no final scheduled follow-up spinal radiograph was completed. This left 97 women with 139 confirmed incident clinical vertebral fractures that were included in the primary analyses. For women with confirmed fractures, back pain was the presenting complaint in ∼90% of cases.

Approximately one-fourth (101/446 = 22.6%) of all incident radiographic vertebral deformities involving a loss of at least 20% and 4 mm in any vertebral height were clinically diagnosed as vertebral fractures at the same vertebral level (Table 2). More severe incident deformities were more often clinically diagnosed. Clinical diagnoses were made for 28.4% of the deformities of at least 30% and 4-mm height loss compared with 14.3% clinical diagnoses for deformities involving at least 20% and 4-mm but <30% height loss (p = 0.0005 for comparison between deformity severity categories). Only 4.7% of deformities that involved at least 15% but <20% loss in vertebral height were clinically diagnosed. Results were similar when calculated at the individual participant level in that 23.0% of women with incident deformities also had an incident vertebral fracture clinically diagnosed at the same vertebral level. Furthermore, 28.9% of women with deformities that involved at least 30% and 4-mm height loss were also clinically diagnosed with a vertebral fracture at the same vertebral level compared with only 11.6% clinical diagnoses in women with deformities that involved at least 20% and 4-mm but <30% height loss (p = 0.0004 for comparison between deformity severity categories).

Table Table 2.. Proportion of Incident Morphometric Vertebral Deformities That Were Clinically Diagnosed as Incident Vertebral Fractures at the Same Vertebral Level
original image

Considering all confirmed incident clinical vertebral fractures, 72.7% were also identified by morphometry at the same vertebral level (Table 3). Morphometry identified a severe deformity involving at least 30% and 4-mm height loss at the same vertebral level for 54.0% of incident clinical vertebral fractures and identified a deformity involving at least 20% and 4-mm but <30% height loss for 18.7% of clinical vertebral fractures. Results were similar when calculated at the individual participant level in that 78.3% of women with clinical vertebral fractures also had a morphometric vertebral deformity at the same vertebral level; 64.9% of these women with clinical fractures were identified as having a severe deformity at the same vertebral level, whereas 13.4% were identified with a milder deformity.

Table Table 3.. Proportion of Incident Clinically Diagnosed Vertebral Fractures With an Incident Morphometric Deformity at the Same Vertebral Level
original image

Concordance between clinical and morphometric radiographic fractures was slightly higher in a sensitivity analysis that included additional women with confirmed clinical vertebral fractures that either were diagnosed after completion of their final scheduled study spinal radiograph or who had not completed a final study spinal radiograph. When these women were considered to have had both an incident clinical vertebral fracture and an incident morphometric vertebral deformity at the corresponding vertebral level, overall, 26.6% of incident deformities were clinically diagnosed, and 76.7% of clinical fractures were also identified by mor phometry. Findings were similar to the primary results reported above in analyses stratified by the presence or absence of prevalent morphometric vertebral deformities at the baseline exam (data not shown).

DISCUSSION

In this large prospective study of postmenopausal women with low BMD, we found that approximately one-fourth of incident radiographic vertebral deformities that involved a loss of 20% and at least 4 mm in any vertebral height were clinically diagnosed as vertebral fractures at the same vertebral level. The proportion of deformities that were clinically diagnosed ranged from nearly 30% for the most severe deformities to ∼15% for milder deformities. Furthermore, <5% of deformities that involved at least 15% but <20% vertebral height loss were clinically diagnosed. Conversely, of incident clinically diagnosed vertebral fractures, ∼75% were also identified as incident morphometric deformities at the same vertebral level, the majority of which were severe deformities. One-fourth of incident clinical fractures were not identified as incident deformities in this study by even the most liberal morphometric criteria used.

Our data are consistent with those from previous studies correlating more severe deformities with increased clinical symptoms and physical findings possibly attributable to incident clinical vertebral fractures.(10,11,15) They also are consistent with findings from a study in which 40 of 50 patients clinically diagnosed with vertebral fractures, confirmed retrospectively based on investigator review of written radiographic reports, were determined to have severe deformities (i.e., any vertebral height ratio below expected by >4 SD) at the same vertebral level that had been considered fractured in the clinical setting.(1)

Our overall estimate of the proportion of incident vertebral deformities that comes to medical attention seems modestly lower than that reported in the literature. Using two cohorts of older women from Rochester, MN, Cooper et al.(1) indirectly estimated that one-third of incident morphometric vertebral deformities are clinically diagnosed. They estimated an incident deformity rate from data on age-stratified prevalent vertebral deformities, defined as any vertebral height ratio below expected by >3 SD,(10) and estimated the incident clinical vertebral fracture rate from a retrospective review of medical records, including radiology reports. Using data from a multisite prospective cohort study, Nevitt et al.(6) reported that 27% of postmenopausal women with an incident morphometric vertebral deformity also had a clinical vertebral fracture. Incident deformities were defined by a decrease in any vertebral height by at least 20% and 4 mm compared with the baseline height, whereas incident clinical vertebral fractures were ascertained when subjects self-reported a new diagnosis of vertebral fracture that was confirmed by a corresponding community radiology report that noted a vertebral fracture. Spinal radiographs were not obtained to confirm a new clinical fracture. In the Multiple Outcomes of Raloxifene Evaluation (MORE) randomized trial, 33% of postmenopausal women with new radiographic vertebral deformities were defined as having an incident clinical vertebral fracture.(21,22) Incident vertebral deformities were defined by an increase of at least one semiquantitative grade from baseline, as rated by two independent assessments, and/or a decrease in any vertebral height by at least 20% and 4 mm compared with baseline height (adjudicated deformities must have been confirmed by at least two of these three assessments). Clinical vertebral fractures were considered to have occurred when subjects reported signs or symptoms suggestive of vertebral fracture (e.g., back pain) at any research clinic visit that subsequently were corroborated by a spinal radiograph: either a radiograph scheduled in all subjects per study protocol or an additional radiograph acquired because of suspicion for vertebral fracture. However, there was no standardization across study sites with respect to which signs and symptoms triggered acquisition of these unscheduled radiographs. Radiographic criteria for incident clinical vertebral fractures were the same as those for incident deformities and, by definition, all clinical vertebral fractures were considered incident vertebral deformities.

Differences between our findings and those reported by Cooper et al.(1) and Nevitt et al.(6) in part may be because these earlier studies did not compare baseline and follow-up radiographs to confirm a new clinical fracture but relied solely on diagnoses rendered by community physicians, who did not follow standardized criteria. This increased the probability in these studies that some prevalent vertebral deformities were mistaken for incident vertebral fractures. Differences between our findings and those derived from MORE study data may be related to differences in methods of ascertainment of both incident vertebral deformities and incident clinical vertebral fractures. Our study defined incident vertebral deformities based solely on morphometric criteria, whereas incident deformities could have been identified in the MORE study based on semiquantitative criteria with or without meeting morphometric criteria. With respect to incident clinical vertebral fractures, in the MORE study, these were not community-diagnosed, but instead were identified in the research clinic based on a research protocol that may have had a different threshold for obtaining interim spinal radiographs than exists in a community setting. Furthermore, MORE subjects who underwent scheduled spinal radiographs during routine research clinic visits were defined as having a clinical vertebral fracture if the radiograph contained an incident morphometric deformity and, at the time of the visit, the subject also had reported signs or symptoms suggestive of vertebral fracture. This methodology is likely to have classified some deformities that would not have come to attention in the community as clinical vertebral fractures. Whereas our results may represent the most precise contemporary estimation of agreement between incident vertebral deformities and incident vertebral fractures detected in the clinical setting, MORE data suggest that even with baseline radiographs and a standardized detection system it may be difficult to exceed clinical detection of more than one-third of incident vertebral deformities.

Results from this study highlight a number of important issues. On initial examination, study findings suggest a relatively straightforward trade-off between the stringency of the criteria used to define incident vertebral deformities and the agreement between deformities and incident clinical vertebral fractures. For example, using a strict definition of incident vertebral deformity increased the proportion of deformities that were identified as clinical fractures, but decreased the proportion of all incident clinical vertebral fractures identified by morphometry. Analogously, using a more liberal definition of incident vertebral deformity decreased the proportion of deformities identified as clinical fractures but increased the proportion of all incident clinical fractures identified by morphometry. However, other study findings show the more complex relationship between incident deformities and incident clinical fractures. First, whereas expansion of the primary definition for incident vertebral deformity to include those with a 15% or greater loss in vertebral height would have further increased the proportion of clinical fractures identified by morphometry, very few of these deformities were clinically diagnosed, and some may have been false-positive morphometric fractures attributable to measurement error.(17,23) Second, approximately one-fourth of clinical fractures did not meet even the most liberal morphometric criterion for an incident deformity. This finding could be attributable in part to nondetection by quantitative morphometry of mild to moderate incident fractures that may be more easily identified using semiquantitative methods.(24) Alternatively, some of these cases may have arisen because of inconsistencies between radiographs obtained in study centers versus community settings and between semiquantitative and morphometric measurement techniques. Third, although the majority of even the most severe incident morphometric deformities were not clinically diagnosed in this study, it is likely that a substantial portion of these undiagnosed deformities were associated with increased back pain and back-related disability.(6) Therefore, it is possible that a change in clinical algorithms could have increased the clinical diagnosis rate and subsequently increased the agreement between incident deformities and incident clinical fractures above that observed in this study. However, findings from the MORE study suggest that the impact of efforts to aggressively increase ascertainment of clinical fractures may be limited.

Our study has several important strengths. The morphometric approach used for defining incident vertebral deformities has previously been validated in relation to several clinical predictors and consequences of deformities,(11) such that identified deformities are likely to have been clinically relevant endpoints. With respect to the clinically reported vertebral fractures, central review of all community spinal radiographs, including comparison with the participants' baseline radiograph, ensured that these reported fractures were, in fact, incident fractures. In addition, ascertainment of incident clinical vertebral fractures was likely to have been sensitive. FIT participants were contacted every 3 months to inquire about fractures and were instructed to contact their clinical center immediately after being diagnosed with a new fracture. The requirement in this study that incident deformities and fractures must have occurred at the same vertebral level to be considered concordant should have increased the precision of our estimate on the agreement between these entities by avoiding the comparison of discrepant events. Finally, because loss to follow-up in FIT was very low—94% of all enrolled FIT subjects completed both baseline and closeout spinal radiographs—study results should be internally valid.

However, our study has a number of limitations. Our participants are community-dwelling white women and therefore the results may not be generalizable to women of other races, nursing home residents, and men. In addition, as discussed above, because there is no gold standard for defining clinical vertebral fractures, it is possible that the concordance between incident vertebral deformities and incident clinical vertebral fractures observed in this study is in part sensitive to differences in diagnosing these fractures in community practice. For example, changes in clinical algorithms could increase the clinical vertebral fracture diagnosis rate and the concordance between incident vertebral deformities and incident clinical vertebral fractures above that observed in this study. Alternatively, because our participants were enrolled in a randomized trial of fracture prevention, they may have been more likely to obtain clinical evaluation in the community for symptoms suggestive of vertebral fracture in comparison with the general population. To the extent that this occurred, our findings regarding the number of incident vertebral deformities coming to medical attention may exceed what would be expected outside a research setting. It also is possible that errors in the fracture adjudication and morphometry procedures could have misclassified either incident vertebral deformities or incident clinical vertebral fractures with respect to vertebral level. This could have resulted in fracture-level analyses underestimating the proportion of incident clinical fractures identified as incident vertebral deformities as well as the proportion of incident deformities clinically recognized as incident fractures. However, fracture-level results differed only slightly from those at the individual subject level, suggesting that this is not a major issue. Concordance between incident vertebral deformities and incident clinical vertebral fractures also may have been limited by differences between morphometric and semiquantitative measurement techniques. However, using morphometric measurements of nonstandardized community radiographs to confirm incident clinical fractures probably would have introduced substantial errors into vertebral height change measurements, and possibly caused misclassification of fractures. Use of semiquantitative assessments to define both incident vertebral deformities and incident clinical vertebral fractures may have improved their concordance, but this is unknown as these data were not collected in FIT. Another possible limitation of this study is that diagnosis of some incident clinical fractures appeared to have occurred after the final scheduled follow-up spinal radiograph was completed. Because these clinical fractures were not included in the study primary analyses, this could have led to an underestimation of the proportion of incident vertebral deformities that are clinically diagnosed as vertebral fractures. However, in a sensitivity analysis that included all confirmed clinical fractures, concordance between clinical fractures and incident radiographic deformities was only modestly improved. Finally, because the FIT study collected spinal radiographs for morphometry only at baseline and several scheduled time-points, it was not possible to determine the timing of the incident deformities, knowledge of which may have helped more precisely correlate them to the clinical vertebral fractures.

In conclusion, in this study, only about one-fourth of incident radiographic vertebral deformities were clinically diagnosed as new vertebral fractures, although the proportion clinically diagnosed was increased for more severe deformities. Whereas the majority of incident clinical vertebral fractures were identified as severe morphometric deformities, approximately one-fourth did not meet even the most liberal morphometric deformity criterion. Further study of the factors that may explain the discordance between incident vertebral deformities and incident clinical vertebral fractures is important.

Acknowledgements

This study was supported by funding from Merck Research Laboratories.

Ancillary