Repeat Low-Trauma Fractures Occur Frequently Among Men and Women Who Have Osteopenic BMD

Abstract

Fracture risk assessment based solely on BMD has limitations. Additional risk factors include the presence of a previous low-trauma fracture. We sought to quantify the fracture burden attributable to first versus repeat fracture. We studied 2179 men and 5269 women, 50–90 yr of age, participating in the Canadian Multicentre Osteoporosis Study (CaMos). We included all low-trauma fractures that occurred over 8 yr of follow-up and classified these as either first or repeat clinical low-trauma fracture based on lifetime fracture history. Analyses were further stratified by sex, age, BMD risk categories (normal, osteopenia, osteoporosis), and vertebral deformity status. There were 128 fractures in men and 577 fractures in women. About 25% of fractures in men and 40% in women were repeat fractures. Just over one half of first fractures occurred in those with osteopenic BMD (58% in men, 54% in women). Just under one half of repeat fractures also occurred in those with osteopenic BMD (42% in men, 47% in women). The incidence of repeat fracture was, in most cases, nearly double, but sometimes nearly quadruple, the incidence of first fracture within a given BMD risk category in both men and women. Repeat fractures contribute substantially to overall fracture burden, and the contribution is independent of BMD. Furthermore, those with a combination of prior low-trauma fracture and another risk factor were at especially high risk of future fracture.

INTRODUCTION

There were approximately nine million osteoporotic fractures worldwide in 2000, and close to one in five of these were hip fractures.⁽¹⁾ Osteoporotic fractures contribute to increased mortality,⁽²⁾ decreased quality of life,⁽³⁾ and substantial direct and indirect costs.⁽⁴⁾ One approach to reduce the burden of fractures is primary prevention (i.e., to identify men and women at high risk and intervene before the fracture occurs). Traditionally, fracture risk is assessed using BMD determined by DXA. The standard categories of fracture risk are “normal” BMD (T-score is −1 or greater), osteopenic BMD (T-score between −2.5 and −1), and osteoporotic BMD (T-scores of −2.5 or less), and these categories, respectively, represent low, medium, and high risk of fracture. The major limitation with a classification system based on BMD T-scores is that, whereas the fracture rate is highest among those with osteoporotic BMD, most fragility fractures occur among patients who have a BMD T-score in the osteopenic range, simply because the prevalence of osteopenia is much higher than the prevalence of osteoporosis.^(5–8) As a result, recent research has focused on identifying factors other than BMD that contribute to fracture risk among those with higher BMD T-scores. This would enable the clinician to identify high-risk patients with greater sensitivity.

Having sustained a fracture is a strong and consistent risk factor for subsequent fractures.^(7,9,10) However, the incidence of fracture reported in most cohort studies excludes repeat fractures for the same individual within the study period. Excluding these fractures may bias estimates of fracture burden away from population estimates. In particular, such studies will underestimate the percentage of fractures that are repeat fractures. Indeed, fractures are more likely to occur in close succession, with the highest risk of fracture occurring just after a previous fracture.^(11,12) Accurate estimates of fracture burden should take into account repeat clinical fracture outcomes.

The aims of our study were to quantify the fracture burden attributable to first versus repeat fracture, both with and without consideration of BMD; to ascertain potential differences between men and women; and to identify whether specific risk groups exist in which the potential for targeted intervention might better impact overall fracture incidence. Answering these questions are critical steps in developing better strategies for fracture prevention.

MATERIALS AND METHODS

Subjects

We included men and women participating in an ongoing cohort study, the Canadian Multicentre Osteoporosis Study (CaMos), who were between 50 and 90 yr of age during the first 8 yr, who had follow-up fracture data, and who had a baseline BMD measurement. There were a total of 9423 participants at baseline, 8525 met the age criterion, and 7448 were included in the study.

The methodological details of CaMos have been described elsewhere.⁽¹³⁾ Briefly, eligible participants were at least 25 yr of age at the start of the study, lived within a 50-km radius of one of nine Canadian cities (St. John’s, Halifax, Quebec City, Toronto, Hamilton, Kingston, Saskatoon, Calgary, and Vancouver), and were able to converse in English, French, or Chinese (Toronto and Vancouver). Households were randomly selected from a list of residential phone numbers, and participants were randomly selected from eligible household members using standard protocol. Of those selected, 42% agreed to participate and had a baseline interview. Ethics approval was granted through McGill University and the appropriate ethics review boards for each participating center.

Data collection

All participants were given a standardized interviewer-administered questionnaire (CaMos questionnaire 1995) at baseline, which determined demographics, general health, nutrition, medication use, and medical history. The questionnaire was designed to capture detailed information about risk factors for fractures including information about prior fractures and, as such, assessed all previous fractures (fracture site, date, and circumstances), family history of osteoporosis/fracture, and falls in the past month. All participants were also given the Medical Outcomes Trust 36-Item Health Survey (SF-36).⁽¹⁴⁾ All participants had a baseline clinical assessment that included measurement of height, weight, and BMD. Lateral lumbar and thoracic spine X-rays were performed in all subjects who were ≥50 yr of age. Vertebral deformities were assessed from X-rays by a trained technologist using vertebral morphometry.⁽¹⁵⁾

Follow-up visits were scheduled in the third year (1998–1999) for those between 40 and 60 yr old and in the fifth year (2000–2001) for all participants. The follow-up visits included an interviewer-administered questionnaire together with measured height, weight, and BMD. In years that participants did not come to a study center, a self-administered fracture questionnaire was mailed out. In this study, we included fractures that occurred before the eighth annual follow-up (2003–2004).

BMD

BMD was measured at the lumbar spine (L₁–L₄), femoral neck, trochanter, Ward’s triangle, and total hip. Seven centers used Hologic densitometers and two used Lunar densitometers. T-scores were based on published reference standards for Canadians.⁽¹⁶⁾ All Lunar measurements were converted to equivalent Hologic values using standard reference equations.⁽¹⁷⁾ All BMD values were calibrated using a European spine phantom circulated between study centers. A more detailed description of BMD quality control appears elsewhere.⁽¹⁸⁾

Fracture assessment

Self-reported incident fractures were identified by yearly postal questionnaire or at the scheduled interview (year 3 and year 5). Confirmation and further information concerning the fracture was gathered using a structured interview that included items on date, fracture site, circumstances leading to fracture, X-ray report (if obtainable), and medical treatment. Fractures that occurred without trauma (such as in a motor vehicle accident) or from a fall of standing height or less were considered to be low-trauma fracture. Low-trauma fractures, classified by skeletal site, were included in our analyses with the exception of any fractures of the skull, face, hands, or feet. To insure that there were no duplicate events in the database, all repeat fractures of the same skeletal site and all multiple fractures were assessed for possible replication using X-ray and/or medical reports.

Statistical methods

We performed separate analyses for men and women. The date of study entry was the latest of baseline interview or the participant’s 50th birthday. The date of study exit was the earliest of eighth annual follow-up, the participant’s 90th birthday, loss to follow-up, or death. We considered age as a categorical variable with age groups separated into 5-yr age bands. We compared continuous variables for eligible subjects with and without missing data with a t-test (p = 0.05) assuming unequal variances.

We considered two categories of fracture: first low-trauma fracture after age 40 and repeat low-trauma fracture after age 40. Those who had a low-trauma fracture before baseline but after age 40 were at risk for repeat fracture from study entry. Those who had a low-trauma fracture after study entry were at risk for repeat fracture any time after the first event. We used a 10-yr lag between starting age for fracture history and cohort entry to insure that all fractures classified as first fractures were at least 10 yr after any previous fracture. We also divided fractures into categories based on the minimum of either femoral neck or lumbar spine BMD T-score: −1 or greater was normal, between −1 and −2.5 was osteopenia, and −2.5 or less was osteoporosis. To determine both the individual and joint relationship of BMD and prior fracture to incident fracture, we generated six risk categories: (1) first low-trauma fracture after age 40 and normal BMD; (2) first low-trauma fracture after age 40 and osteopenic BMD; (3) first low-trauma fracture after age 40 and osteoporotic BMD; (4) repeat low-trauma fracture after age 40 and normal BMD; (5) repeat low-trauma fracture after age 40 and osteopenic BMD; and (6) repeat low-trauma fracture after age 40 and osteoporotic BMD.

We used age category weights derived from the 2001 Canadian census to make age-standardized estimates. We estimated the distribution of fractures cases by risk category. We also estimated the underlying baseline prevalence of each risk category. We calculated the adjusted rate ratios by indirect standardization using the category of first fracture and normal BMD as the reference category. We also calculated the population attributable rate percent (PAR%) based on the adjusted rate ratios and the case prevalence.⁽¹⁹⁾ PAR% estimates the hypothetical fracture reduction attainable assuming the fracture rates in the higher-risk category were reduced to the rates of the reference category. All above estimates were for a hypothetical population with age distribution equivalent to the 50- to 90-yr-old 2001 Canadian census population and age-stratified parameters derived from the study sample.

Analysis was performed using Stata Version 9.2.

RESULTS

The baseline characteristics of the 2179 men and 5269 women in the study sample are shown in Table 1. The study population excluded 51 men and 83 women because they had no follow-up data and 245 men and 698 women because they did not have baseline BMD. Men who met the age criterion but were excluded were on average 8.2 yr older (95% CI: 6.9, 9.5), 3.5 kg lighter (95% CI: 1.4, 5.6), and in worse health (i.e., had SF-36 physical scores that were on average 4.2 points [95% CI: 2.9, 5.5] lower than those in the study). They were also more likely to have had a prior low-trauma clinical fracture than those in the study (10.8% versus 6.2%). Women who met the age criterion but were excluded were on average 7.4 yr older (95% CI: 6.8, 8.0), 2.7 kg lighter (95% CI: 1.7, 3.7), and in worse health (i.e., had SF-36 physical scores that were on average 5.5 points [95% CI: 4.7, 6.3] lower than those in the study]. They were also more likely to have had a prior low-trauma clinical fracture than those in the study (17.9% versus 12.5%).

TABLE 1.

Baseline Characteristics of Eligible CaMos Participants With Baseline BMD and Follow-Up Fracture Data

	Men (N = 2179)		Women (N = 5269)
	Mean	SD	Mean	SD
Age (yr)	62.9	10.4	64.0	10.0
Weight (kg)	81.8	13.4	69.0	13.5
Total hip BMD (g/cm2)	1.00	0.15	0.85	0.14
Lumbar spine BMD (g/cm2)	1.05	0.17	0.93	0.17
SF-36 physical	48.7	9.2	46.8	10.2
SF-36 mental	54.5	7.9	53.2	8.8
	n	(%)	n	(%)
Prevalent low-trauma clinical fracture after age 40	135	6.2	657	12.5
Previous falls (last month)	134	6.1	339	6.4
Current, oral or inhaled, corticosteroid use	232	10.6	736	14.0
Current antiresorptive use	6	0.3	1576	29.9
Current tobacco use	360	16.5	707	13.4
Current alcohol use	1604	73.7	2977	56.6
Vertebral deformity
Normal	1074	49.3	2776	52.7
Grade 1	238	10.9	608	11.5
Grade 2+	124	5.7	403	7.6
Missing	743	34.1	1,482	28.1

The cohort at risk for first fracture had 2044 men with mean follow-up of 5.8 yr and 4612 women with mean follow-up of 6.0 yr; there were 95 fractures in men and 344 fractures in women. The cohort at risk for repeat fracture had 230 men with mean follow-up of 4.8 yr and 1001 women with mean follow-up of 5.1 yr; there were 33 fractures in men and 233 fractures in women. The age-adjusted distribution of the six risk categories among low-trauma fracture cases for men and women is shown in Fig. 1. An estimated 25% of fractures in men and 40% in women were repeat fractures. Most first fractures occurred among those with osteopenic BMD in both men and women. In contrast, repeat fractures were evenly divided between all three BMD categories in men and evenly divided between those with osteopenic BMD and osteoporotic BMD in women. Osteoporosis, as defined by T-scores, contributed to 21% of the total fracture burden in men and 39% of the total fracture burden in women. Osteopenia, on the other hand, contributed to slightly more than one half the estimated total fracture burden (54% among men and 51% among women).

FIG. 1.

Estimated age-standardized distribution of incident low-trauma fractures by risk category (BMD and prior fracture) for men and women 50–90 yr of age.

The estimated age-adjusted prevalence of the six fracture risk categories at baseline is shown in Table 2. For a population with the sex-specific age distribution of the 2001 Canadian census population between 50 and 90, an estimated 4.6% of men and 9.3% women would have a history of low-trauma fracture after the age of 40. An additional 3.8% of men and 9.6% of women would have osteoporotic BMD, but with no history of low-trauma fracture.

TABLE 2.

Estimated Standardized^* Baseline Prevalence of Risk Categories Among Men and Women 50–90 yr of Age

BMD category	Prior low-trauma fracture (age 40+)	Prevalence percentage (95% CI)
		Men	Women
Osteoporosis	Yes	0.6 (0.3–0.8)	2.9 (2.5–3.4)
Osteopenia	Yes	2.6 (2.0–3.2)	5.0 (4.1–5.9)
Normal	Yes	1.4 (0.9–1.8)	1.4 (1.1–1.7)
Osteoporosis	No	3.8 (3.0–4.5)	9.5 (8.8–10.2)
Osteopenia	No	39.0 (35.2–42.8)	43.7 (40.9–46.5)
Normal	No	52.8 (49.0–56.6)	37.5 (34.8–40.3)

^*Baseline age strata weighted so that age distribution is equivalent to sex-specific Canadian Census 2001 population 50–90 yr of age.

The estimated incidence of first and repeat fractures by age group is shown in Fig. 2.

FIG. 2.

Incidence of first and repeat low-trauma fracture in men and women by age group.

The incidence of both first and repeat fractures increased with increasing age in both men and women. The magnitude of this increase was roughly 60–70% per decade and similar for first and repeat fractures. Second, the incidence of repeat fracture was markedly higher than incidence of first fracture within each 10-yr age group. In all age groups and in both men and women, the incidence of repeat fractures was at least double the incidence of first fractures.

The estimated incidence of first and repeat fractures by BMD category is shown in Fig. 3. There was a clear trend for first fractures to increase from the normal BMD category to the osteoporotic BMD category in both men and women. For repeat fractures, the same was true, except that there was an unexpectedly high incidence of repeat fracture among men with normal BMD. The incidence of first and repeat fracture was roughly similar among women with normal BMD, but in all other subgroups, the incidence of repeat fracture was two to four times higher than the incidence of first fracture.

FIG. 3.

Incidence of first and repeat low-trauma fracture in men and women by BMD category.

The estimated incidence of first and repeat fractures by vertebral deformity category is shown in Fig. 4. The figure shows that vertebral deformity and previous low-trauma fracture are independent predictors of fracture. In both men and women, the incidence of first fracture was lower in those with out vertebral deformity and was highest among those with grade 2 or higher vertebral deformity. We observed the same pattern among those with repeat fractures, except that there was an unexpectedly high incidence of repeat fracture among men with no vertebral deformity. The incidence of first and repeat fracture was roughly similar among men with grade 1 vertebral deformity, but in all other subgroups, the incidence of repeat fracture was three to four times higher than the incidence of first fracture.

FIG. 4.

Incidence of first and repeat low-trauma fracture in men and women by vertebral deformity status.

The estimated adjusted rate ratio (RR) and PAR% for each risk category other than the reference category is shown in Table 3. The RRs were highest among those with both osteoporotic BMD and prior fracture for both men (RR = 14.6, 95% CI: 6.3–33.8) and women (RR = 9.2, 95% CI: 4.9–17.2). Women with normal BMD and prior low trauma fracture were noted to have only a marginally excess number of low-trauma fractures, and the resulting CIs were inconclusive. We note that prior fracture is a risk factor independent of BMD, because a separate calculation shows that the RRs for those with prior fracture compared with those without is 1.9 (95% CI: 1.3–2.8) for men and 2.3 (95% CI: 1.9–2.7) for women after adjusting for both age and BMD.

TABLE 3.

Adjusted Rate Ratio and Population Attributable Rate Percent (PAR %) for Low-Trauma Fracture Among Men and Women 50–90^* yr of Age by Risk Category

BMD category	Prior low-trauma fracture (age 40+)	Adjusted rate ratio (95% CI)		PAR %
		Men	Women	Men	Women
Osteoporosis	Yes	14.6 (6.3–33.8)	9.2 (4.9–17.2)	7.7	18.0
Osteopenia	Yes	4.1 (2.0–8.7)	5.5 (3.4–8.8)	7.9	15.3
Normal	Yes	6.4 (2.6–15.4)	1.2 (0.4–3.4)	5.3	0.2
Osteoporosis	No	5.4 (2.5–11.7)	3.6 (2.1–6.2)	10.4	13.5
Osteopenia	No	2.4 (1.4–4.0)	2.1 (1.4–3.0)	25.5	16.7
Normal	No	Reference	Reference	—	—

^*Rate ratios were calculated using indirect standardization.

For men, those with osteopenia and no prior fracture had a PAR% = 25.5%, in contrast to those in other risk groups, where the estimated PAR% ranged from 5.3% to 10.4%. For women, the PAR% was evenly distributed among four risk groups, with PAR% ranging from 13.5% to 18.0% in all risk groups with osteopenic or osteoporotic BMD. PAR% is a measure based on both the percentage of people within a specified group and incidence of the outcome within the specified group. In the case of osteopenic men without prior fracture, the high PAR% is largely caused by the high percentage of men within this risk group, resulting in a high number of fractures despite having only slightly increased incidence or risk as shown by the adjusted rate ratio.

DISCUSSION

We found that repeat clinically symptomatic low-trauma fractures constitute one quarter of the overall fracture burden in men and just under one half of the fracture burden in women and that the contribution to fracture burden was independent of BMD. The fact that a large percentage of all fractures occur among a readily identifiable population (i.e., those with previous low-trauma fractures highlights the importance of assessing factors other than BMD to assess fracture risk). The overemphasis of BMD in the clinical setting may explain the persistence of a significant care-gap (i.e., low percentage of osteoporosis diagnosis and therapy among those who have fragility fracture).^(20–22) Because repeat fractures are such a large part of the overall fracture burden, further work is necessary to reduce this care-gap.

We also found that more than one half of the fractures among both men and women occur in those with osteopenia, a finding consistent with previous studies performed among postmenopausal women.^(5–8) Our work further shows that those with BMD T scores between −1 and −2.5 comprise the largest group among those with first clinical low-trauma fracture. Furthermore, those with BMD T-scores between −1 and −2.5 and those with BMD T-scores ≤ −2.5 comprise nearly equal proportions among those with repeat fractures. The implications of these findings are not only that first low-trauma fractures often occur before BMD falls below −2.5 but also that women and men with BMD in the osteopenic range are at high risk of repeat low-trauma fractures. This finding reinforces the notion that, whereas BMD in the osteoporotic range is a good indicator of fracture risk in someone who has not yet had a fragility fracture, a prior fragility fracture confirms the presence of skeletal fragility and is indicative of clinically manifest osteoporosis, regardless of the underlying BMD. It also indicates that further assessment is necessary for fracture prevention among those with osteopenic BMD.

We noted substantial differences in fracture burden between men and women. Our study showed that men and women with osteoporotic BMD have a similar rate of first fracture and repeat fracture, a finding consistent with the results of other studies.⁽²³⁾ However, the prevalence of this risk factor in men is less than one half of the prevalence in women, as previously reported,⁽¹⁶⁾ and, as a result, the percentage of fracture cases having osteoporotic BMD is approximately one half as much in men as it is in women. The differences in distribution of risk factor categories have important clinical implications for fracture prevention strategies. Among women, the majority of all fractures occur among those with either osteoporotic BMD or prior fracture (i.e., those considered at high risk under most risk assessments). As seen from the PAR%, targeted intervention for women with prior fracture and/or osteoporotic BMD has the potential to substantially change overall fracture rate.

In contrast, among men, more than one half of all fractures occur in those with normal to osteopenic BMD and no prior fracture. A potential explanation for this finding includes the fact that, among men, there may be factors, such as falls or specific lifestyle behaviors, associated with fractures that are not associated with BMD. This hypothesis requires formal testing. Of clinical note, this group of men has been considered to be at low fracture risk on the basis of BMD and has not been targeted for fracture prevention strategies. As seen from the PAR%, fracture prevention strategies not including lower-risk men (i.e., osteopenic men without prior fracture) will have limited impact because of the large percentage of fracture cases within this group. The extensive research on osteoporosis that has been based on postmenopausal women may therefore not be applicable for men, and further work is necessary to identify subgroups of osteopenic men at high risk of fracture and to develop cost-effective interventions for these subgroups.

We have shown that the incidence of repeat fracture is almost always at least double the incidence of first fracture within risk strata based on age and BMD. These results are largely consistent with previous results. Based on first and second fracture within the cohort, Center et al.⁽¹²⁾ showed that that the incidence of repeat fracture was estimated to be from 1.65 to 4.32 times the incidence of first fracture within a given age category. Based on the first fracture within the cohort, Pasco et al.⁽⁷⁾ showed that the incidence of fracture among those with prevalent fracture was double the risk of incidence of those without prevalent fracture, after adjustment for both age and BMD in a study of postmenopausal women. A meta-analysis based on multiple cohorts including the CaMos cohort reported the incidence of fracture among those with prevalent fracture was between 1.83 and 2.03 times the incidence of those without prevalent fracture within each age category and slightly more modest associations after adjusting for BMD.⁽²⁴⁾

We found that the presence of vertebral deformity was a risk factor for fracture and that the incidence of repeat clinical fracture was at least double the incidence of first fracture within risk strata based on vertebral deformity. It has been noted that the strongest associations between prior fracture and subsequent fracture are between fractures at the same skeletal site, likely because of having both independent and common risk factors.⁽²⁵⁾ This association between prior fracture and repeat fracture is most notable for vertebral fracture and results in a series or cascade of repeat events.⁽²⁶⁾ It has also been observed that both the number of vertebral deformities⁽²⁷⁾ and the number of previous fractures⁽²⁸⁾ are predictive of future fracture. We found that risk associated with prevalent vertebral deformity and clinical low-trauma fracture for future clinical fracture is especially high.

The strengths of our study include the fact that we were able to measure BMD and assess lifetime fracture history and incident fracture (by site and circumstances) in a randomly selected population, which enhances the generalizability of our findings. Our study is limited by the observational study design. There were notable differences in several risk factors for fracture between the study population and those excluded from the study. The study population was on average in better health and thus our results might underestimate the burden attributable to repeat fracture and low BMD. Age differences and any difference attributable to age would not bias the study. We also included only clinically recognized vertebral fracture in the definition of low-trauma fracture. Those with incident clinical fracture may in fact have one or more unrecognized vertebral deformities.⁽²⁹⁾ If vertebral deformities were counted as low-trauma fracture, it would result in an even higher burden of repeat fractures. The use of antiresorptive therapy was not constant over the six risk groups, because those in high-risk groups were more likely to start bisphosphonate therapy. The resulting effect would be a modest attenuation of the increased risk among those with osteoporosis and/or prior fracture. Whereas this also notably impacts the PAR%, it does so in a way that reflects diminishing potential of treatment because of the fact that many of those in high-risk groups are already treated.

In summary, we found that repeat fractures contribute substantially to the overall fracture burden. We also found that the fragility fractures may occur early in the natural course of osteoporosis, and there may be repeat fracture without a BMD T-score below −2.5. We also noted a different pattern of fracture burden in men and women that should be taken into account when developing strategies for fracture prevention.