The authors state that they have no conflicts of interest.
In the non-HRT arms of the DOPS study, 10-year fracture risk was higher at each level of T score than predicted by the Kanis algorithm. Under-reporting of fractures in registers and inclusion of HRT users are probable explanations for inappropriately low fracture risk estimates for younger women.
Introduction: International recommendations highlight the importance of absolute fracture risk in establishing intervention thresholds. The available estimates of long-term risk have been derived by combining relative risks from meta-analyses with U.S. normative BMD data and Swedish fracture incidence records. We validated the 2001 Kanis risk algorithm using incident fractures observed in untreated women in the first 10 years of the Danish Osteoporosis Prevention Study (DOPS). Comparisons were also made with the relative risks derived from a recent meta-analysis of 12 cohort studies.
Materials and Methods: We analyzed DXA of the spine and hip from 872 women who were enrolled in the non–hormone replacement therapy (HRT) arms of the study and had not received HRT, bisphosphonates, or raloxifene. We collected verified reports of fractures at each visit. We focused on fractures of the hip, spine, shoulder, and forearm to provide risks comparable with the Kanis algorithm. Accordingly, asymptomatic radiographic vertebral fractures were not included.
Results: Seventy-eight women (9%) sustained relevant fractures. The risk of fracture increased by 1.32 (95% CI, 1.02; 1.70) for each unit decrease in femoral neck T score and by 1.30 (95% CI, 1.06; 1.58) for each unit decrease in lumbar spine T score at baseline. Absolute fracture risk was higher than expected from the Kanis algorithm at all T score levels. The difference was greatest for participants in the higher range of T scores. At T = −1, the observed risk was 10.9% as opposed to an expected risk of 5.7%. Relative risk gradients were similar to those of the recent meta-analysis.
Conclusions: In healthy women, examined in the first year or two after menopause, 10-year fracture risk was higher at each level of BMD T score than expected from the model by Kanis et al. Inclusion of HRT users in the cohorts used may have led to higher BMD values and lower absolute fracture risk in the Kanis model. These longitudinal data can be used directly in estimating absolute fracture risk in untreated north European women from BMD at menopause.
Measurement of BMD served originally as a prognostic tool and a means of assessing the effects of treatment in the osteoporosis clinic. With the 1994(1) WHO working group definition of osteoporosis, reduced BMD T score gained widespread use as a diagnostic criterion. It is recognized, however, that many fractures occur in persons with normal BMD and that a given T score translates to very different fracture risk depending on the age of the patient and the presence of additional risk factors.(2) This is unfortunate, because absolute fracture risk is the relevant entity for deciding which patients to treat and which patients have sufficiently low risk of fracture to be observed without treatment. Pharmacological intervention is of course more cost-effective in patients with a high event rate, provided that the risk in question is modifiable by intervention. Absolute risk estimates applicable to the general population have not been widely available, however. In 2001, a mathematical model was devised, which combined pooled estimates of relative risk from a meta-analysis(3) with U.S. normative BMD data and Swedish fracture incidence records to provide absolute risk estimates. In the following, we refer to this model—which is chiefly based on the BMD–fracture relationship in the elderly—as the Kanis algorithm.(4) More recently, a meta-analysis has become available, which combines data from 9900 men and 29,000 women from 12 prospective cohorts(5) to tabulate relative risks of fracture by age and BMD T score. In addition, the meta-analysis contains absolute fracture risk estimates for hip fractures. In the following, we present 10-year absolute fracture risk outcomes by T score at menopause derived from a nationwide Danish study, which was initiated in 1990. These risk estimates are relevant to healthy north European women, who do not begin systemic estrogens after menopause.
MATERIALS AND METHODS
In 1990–1993, 1293 healthy, postmenopausal women were recruited to the non–hormone replacement therapy (HRT) arms of The Danish Osteoporosis Prevention Study (DOPS). The inclusion procedure has been described in detail previously.(6) Briefly, this was an open study, with a randomized (HRT or no treatment) and a nonrandomized arm (HRT or not by personal choice) and a planned duration of 20 years. Women were eligible for inclusion, provided they were 45–58 years of age and either (1) 3–24 months past last menstrual bleeding or (2) still menstruating but exhibiting perimenopausal symptoms including menstrual irregularities with a serum follicle-stimulating hormone (FSH) > 2 SD above the premenopausal mean. All participants gave their informed consent before entry in the study, which was conducted in accordance with the Helsinki II declaration and approved by the local Ethics Committees. Exclusion criteria were (1) metabolic bone disease, including osteoporosis defined as nontraumatic vertebral fractures on X-ray; (2) current estrogen use or estrogen use within the past 3 months; (3) current or past treatment with glucocorticoids >6 months; (4) current or past malignancy; (5) newly diagnosed or uncontrolled chronic disease; and (6) alcohol or drug addiction. Of the 1293 women in the non-HRT arms in the study, 207 had left the study or declined to attend the 10-year visit. In the event of severe climacteric symptoms, participants had the option of obtaining prescriptions for HRT from their general practitioners, and 197 (15.2%) received some form of HRT in the course of the study. We excluded these participants from the analysis. Finally, 17 participants who had received raloxifene or bisphosphonates were also not included in this report. The categories of excluded patients were not mutually exclusive. The study population consisted of 872 women (Table 1). The 872 participants in the analysis did not differ from the 421 nonparticipants in terms of BMD (p = 0.14 for femoral neck [FN], p = 0.66 for lumbar spine [LS]), height (p = 0.49), or weight (p = 0.64). Nonparticipants were slightly younger (12 months, p < 0.01) and had a lower menopausal age (51.8 versus 52.2 years, p < 0.01). Of the 872 participants, 302 came from the randomized arm and 570 from the self-choice arm of the DOPS study. Major risk factors for osteoporosis(7) (menopause < 45 years of age, secondary amenorrhea > 1 year, maternal hip fracture, body mass index < 19 kg/ m2, fragility fracture > 45 years, rheumatoid arthritis, chronic obstructive pulmonary disease, or immobilization > 1 month after age 45 years) were present in 168 (19.5%) of study participants.
Table Table 1.. Baseline Demographics
We measured BMD of the spine and hip using cross-calibrated QDR-1000/W and QDR-2000 densitometers. DXA was done at inclusion and after 1, 2, 3, 5, and 10 years. NHANES(8) (total hip [THIP]) and Hologic (LS) young adult reference ranges were used for calculation of T and Z scores. The in vivo short-term precision errors for BMD in the participating clinics were 1.5% (LS and THIP). Long-term stability of the equipment was assessed by daily scans of an anthropometric phantom at each center. A standardized procedure for scan acquisition and data analysis was established and followed for all scans.
We collected verified reports of incident fractures at each visit. To compare fracture risk with the risk estimated from the Kanis algorithm,(4) we included all fractures of the hip, forearm, shoulder, and spine. Asymptomatic radiographic vertebral fractures were not included, because these would not have been included in the Swedish register data from which the Kanis risk estimates were derived.
We used logistic regression analysis to model the odds of sustaining a fracture of the hip, forearm, shoulder, or spine within 10 years according to baseline T score. Separate analyses were performed for the FN and the LS regions. The odds ratios for fracture for a one unit decrease in T score were interpreted as relative risks. We used SPSS version 10.0 (SPSS, Chicago, IL, USA).
During the 10-year follow-up period, covering 8890 person-years, 78 (9%; 95% CI, 7.1–11.1%) of the participants sustained a total of 80 fractures. The fractures consisted of 8 symptomatic spine fractures, 1 hip fracture, 64 forearm fractures, and 7 fractures of the proximal humerus. Baseline T scores at both measurement sites were significantly lower in participants who experienced fractures (Fig. 1).
Logistic regression analysis (relative risk)
The relative risk of any fracture (spine, hip, forearm, or shoulder) was 1.32 (95% CI, 1.02; 1.70) for each unit decline in total hip T score and 1.30 (95% CI, 1.06–1.58) for each unit decline in spine T score (Table 2).
Table Table 2.. Observed Gradient of Risk for a 50-Year-Old Woman in the DOPS Study Compared With the 2001 Kanis Model and the 2005 12-Cohorts Meta-Analysis
Comparison with 2001 Kanis algorithm (absolute risk)
The predicted 10-year incidence of any first fracture based on age and BMD was 5.1%, but the observed incidence (9.0%) was 76% higher (p < 0.05). The difference was more marked in patients with high T scores (Fig. 2), with the two curves converging at a T score of −4.7. Specifically, the BMD–fracture relationship differed from the model in exhibiting a much higher fracture incidence at a T score level of 0 (Table 3), with a smaller increase in risk per SD decrease in BMD. The Kanis algorithm assumed a gradient of risk of 1.6 (95% CI, 1.4–1.8) for any fracture and a gradient of risk of 1.4 (95% CI, 1.3–1.6) for forearm fracture. The observed gradients of risk of 1.3 for both these estimates lie outside the CI for generic fracture prediction and just inside the CI for prediction of forearm fracture. The gradient of risk for the small number of spine and hip fractures in this young postmenopausal cohort did not differ significantly from the gradients in the Kanis model.
Table Table 3.. Tabulation of 10-Year Absolute Fracture Risk (Hip, Spine, Shoulder, or Clinical Vertebral Fracture) as a Function of Menopausal Hip BMD vs. the Expected Occurrence Using the Kanis Algorithm
Comparison with 2005 12-cohort meta-analysis (relative risk)
For relative risks, the meta-analysis cites a gradient of risk of osteoporotic fracture of 1.22 (95% CI, 1.07–1.39) for each 1 SD decline in BMD in 50-year-old women. This corresponds closely to the relative risk of 1.32 (95% CI, 1.02–1.70) observed in the present study. Absolute risk estimates for osteoporotic fractures were not stated in the 2005 meta-analysis(5) for comparison, with the exception of hip fracture risk, and the low incidence of hip fractures in our young cohort does not provide us with the opportunity to compare the absolute hip fracture estimates.
This study shows that the original Kanis algorithm could be of limited use at the time of menopause, because it underestimates absolute fracture risk. This is certainly true in healthy north European women who did not receive HRT. First, this discrepancy suggests that national Swedish fracture estimates included data from significant numbers of HRT users. Second, because the BMD–fracture relationship was expected to be steeper than that subsequently found in this study and in the 12-cohort meta-analysis,(5) too many of the fractures would be assumed to occur in the lower BMD bands. This probably accounts for the inappropriately low risk estimates in the Kanis algorithm for 50-year-old women with normal BMD. Whereas fracture risk in this age and BMD group remains low, the observed risk is around twice that expected from the Kanis estimate.
It is generally accepted that to make pharmaceutical intervention viable, it should be targeted to patients at high absolute fracture risk. Furthermore, fracture risk should be modifiable within a feasible duration of treatment. Most currently available treatments have been tested for a continuos treatment duration of at most 10–15 years. It seems reasonable, therefore, to base treatment decisions on the estimated 5- or 10-year fracture risk. Such estimates are needed, not only for the purpose of deciding whether to treat or not, but not least in devising national and international strategies for osteoporosis management.
Meta-analyses of fracture risk are extremely useful for this purpose. They collate great amounts of clinical information, which is multiplied and strengthened by inclusion of large numbers of patient-years in the analysis. They may be able to uncover trends that are overlooked in single cohort studies, which are narrower in geography and usually also in age range. Nevertheless, such cohort studies provide other valuable information. It is not necessarily a disadvantage to use a population, which is in other ways homogenous, if the point of interest is to map the BMD–fracture risk relationship. One advantage of single-cohort studies is that less extrapolation is required and that heterogeneity in how predictors and outcomes are defined can be avoided. Furthermore, fracture risk varies greatly across Europe,(9,10) and the authors of the meta-analysis appropriately note that heterogeneity was marked for fractures other than hip fractures.(5) This is not surprising because the difference in fracture risk between European countries is far from fully accounted for by BMD.(11) Finally, this study tested the validity of the Kanis and 12-cohort meta-analysis data within a very narrow age range, but one that is of particular importance in decision-making (i.e., the time of menopause).
Despite the best efforts of physicians and data managers worldwide, public health registers are inaccurate sources of fracture and trauma data, with some registers underestimating the number of fractures by >40%.(12–14) By using repeated patient interviews, we may have been able to obtain information of verifiable fractures, which may not all have come to the attention of the national databases. Unlike the original Kanis algorithm,(4) the new estimates from the DOPS study and the cohorts included in the 2005 meta-analysis(5) were based on consecutively recorded fracture events and not on register data. However, the geographical origin of the BMD and fracture data were not by any means evenly distributed across the 12 cohorts, with the Finnish Kuopio(15) study supplying the majority of the 50 year olds.
The higher incidence of fractures observed in the DOPS cohort compared with the Kanis algorithm cannot be attributed to a higher prevalence of risk factors in the former. It is even a minor limitation that the DOPS cohort is healthier than the background population, by virtue of exclusion of a small number of women with prevalent vertebral fractures or uncontrolled chronic diseases. This should not be taken to mean that the population was devoid of common risk factors. Just under 20% had one or more IOF-recognized risk factors, such as early menopause, maternal hip fracture, or diseases predisposing to osteoporosis. An obvious limitation of studies undertaken in healthy women at this age is that the number of clinical spine and in particular hip fractures is too small to test the validity of the fracture site–specific risk gradients in the 12-cohorts meta-analysis.
In conclusion, in this cohort of healthy women, examined in the first year or two after menopause, absolute fracture risk was higher at each level of BMD T score than expected from the model derived by Kanis et al. Underestimation of fractures in the applied registers and inclusion of HRT users in the cohorts used may have led to higher BMD values and lower fracture risk in the model. In contrast, the gradient of risk in this single-cohort study is close to that calculated in the very recent meta-analysis of individual BMD and fracture incidence data from 12 large observational cohort studies.(5) Arguably, the time of menopause is not the optimal time of assessment because short-term fracture risk is low and because subsequent bone loss rates will be variable. However, in recent years, concerns about the use of HRT in primary prevention of osteoporosis have led to increasing demand for guiding osteodensitometry in the first years after menopause.
These estimates can be used directly in osteoporosis clinics when advising healthy north European women who have had a BMD measurement done at menopause. Despite their relevance to the general population, it should be borne in mind that these long-term risk estimates—although higher than the Kanis algorithm—will underestimate risk in women with significant BMD-independent risk factors.
The DOPS study was financially supported by the Karen Elise Jensen foundation, Denmark, and the Danish Health Research Council SSVF.