Osteoporos Int (2014) 25:2445–2451 DOI 10.1007/s00198-014-2777-3
ORIGINAL ARTICLE
The incidence of a first major osteoporotic fracture in Iceland and implications for FRAX K. Siggeirsdottir & T. Aspelund & H. Johansson & E. F. Gudmundsson & B. Mogensen & B. Y. Jonsson & V. Gudnason & E. McCloskey & A. Oden & G. Sigurdsson & J. A. Kanis
Received: 31 December 2013 / Accepted: 13 June 2014 / Published online: 1 July 2014 # International Osteoporosis Foundation and National Osteoporosis Foundation 2014
Abstract Summary Based on an extensive cohort study over 25 years, the present study supports the assumption that major osteoporotic fractures can be reasonably predicted from hip fracture rates. Introduction The construct for FRAX models depends on algorithms to adjust for double counting of fracture outcomes in some models and in others, to estimate the incidence of a major fracture from hip fracture rates. The aim of the present study was to test the validity of these algorithms in a large prospective cohort. Methods The incidence of hip, clinical spine, distal forearm, and humerus fracture was determined in the prospective and ongoing population-based Reykjavik Study with follow up of
257,001 person-years. The incidence of a first major fracture was compared with the correction factors used in FRAX to adjust the incidence of several fracture outcomes for double counting. In addition, the incidence of a major osteoporotic fracture estimated from the Icelandic hip fracture rates was compared with the Malmo ratios used in FRAX. Results The adjustments necessary to account for multiple fracture outcomes were similar to those previously derived from Sweden. Additionally, incidence of a first major osteoporotic fracture was similar to that derived for FRAX models. Conclusion The findings of the present study support the algorithms used in FRAX to estimate the incidence of a first major fracture and the predictive value of hip fracture for other major fractures.
K. Siggeirsdottir (*) : T. Aspelund : E. F. Gudmundsson : V. Gudnason : G. Sigurdsson Icelandic Heart Association Research Institute, Holtasmari 1, 201 Kopavogur, Iceland e-mail: [email protected]
Keywords Fracture incidence . FRAX . Major osteoporotic fracture
T. Aspelund : B. Mogensen : V. Gudnason : G. Sigurdsson University of Iceland, Reykjavik, Iceland
Introduction
H. Johansson Centre for Bone and Arthritis Research (CBAR) at the Sahlgrenska Academy, Institute of Medicine, University of Gothenburg, Gothenburg, Sweden H. Johansson : E. McCloskey : A. Oden : J. A. Kanis (*) WHO Collaborating Centre for Metabolic Bone Diseases, University of Sheffield Medical School, Beech Hill Road, Sheffield S10 2RX, UK e-mail: [email protected] B. Mogensen : G. Sigurdsson Landspitalinn, University Hospital Reykjavík, Reykjavík, Iceland B. Y. Jonsson Department of Orthopeadics, Skane University Hospital, Malmo, Sweden
FRAX® is a computer-based algorithm (http://www.shef.ac. uk/FRAX) developed by the World Health Organization Collaborating Center for Metabolic Bone Diseases. The algorithm, intended for primary care, calculates fracture probability from readily obtained clinical risk factors (CRFs) in men and women [1, 2]. The output of FRAX is the 10-year probability of a major osteoporotic fracture (hip, clinical spine, proximal humerus, or wrist fracture) and the 10-year probability of hip fracture. Thus, information is required on the epidemiology of death and fracture in each country for which there is a FRAX model. FRAX models are presently available for 53 regions or countries. All models incorporate country-specific mortality and hip fracture risk. However, a minority of countries that have a FRAX model also have robust information on the risk
2446
of other major fractures (clinical spine, forearm, and proximal humerus). Where available or part available, these are incorporated in the models (e.g., Denmark, Iceland, Japan, Mexico, Russia, Switzerland, Sweden, UK, and US). In the absence of information, FRAX models are based on the assumption that the age- and sex-specific pattern of these fractures is similar to that observed in Sweden. Irrespectively of how fracture rates are derived, account needs to be taken of double counting. For example, studies reporting the incidence of forearm fractures may take account of double counting in the year(s) of study but rarely take account of forearm fractures that might have occurred in much earlier life. When this is accounted for, the incidence of a first forearm fracture from unadjusted population figures is overestimated by about 20 % [3]. In estimating the incidence of a major osteoporotic fracture, account needs also to be taken of double counting different fracture outcomes that are grouped as major fractures. Thus, the metric of interest is the incidence of a first major fracture whether this be at the hip, spine, proximal humerus, or forearm. Such data are available from Malmo, Sweden and mathematical correction factors derived from these data have been applied to new FRAX models where the epidemiology of all major fractures is not available. In the absence of empirical data on fracture incidence, the incidence of fractures at the spine, forearm, and proximal humerus is estimated from the age- and sex-specific pattern of fracture incidence in Malmo, Sweden [4]. Thus, in the absence of detailed information from other sources, FRAX is based on several assumptions derived from the Malmo cohort. Recently, similar detailed information has become available from Iceland [5]. The aim of the present study was to test the validity of the methods used in FRAX to derive the incidence of a major fracture.
Methods The prospective and ongoing population-based Reykjavik Study [6, 7] was initiated in 1967. The Icelandic population was then about 200,000 individuals of who half lived in the capital city Reykjavik and surroundings. The study included both men and women who were residents in the Reykjavik area on December 1st 1967. The participants were born between 1907 and 1935 and selected at random from the Icelandic National Register. The response rate was 71.8 % resulting in 18,872 participants, 9,116 men and 9,756 women who in 1967 represented 34 % of the total Icelandic population in the age range 33 to 65 years. Individuals were followed-up for a median time of 26.5 years until death or December 31st 2008, a total of 482,737 person years.
Osteoporos Int (2014) 25:2445–2451
The study was approved by the National Bioethics Committee and the Data Protection Authority in Iceland. All participants gave informed consent. The Reykjavik Study fracture registration collected all fractures of the participants from entry into the study up to December 31st 2008 [5]. All residents of Iceland have a unique personal identification number allocated at birth or when taking up residence in the country, which facilitates identity and examination of hospital records. Fractures treated on an outpatient basis in Reykjavik were uniquely referred to Landspitalinn University Hospital. Both inpatient and outpatient reports, from all hospitals in Reykjavik including different departments, e.g., trauma, radiography, and outpatients department, were manually examined and verified for fractures until 1983. From 1983, hospitals and the private radiology clinic used by the general practitioners in the Reykjavik area introduced a computerized register system including fracture diagnostic codes. All medical records for the participants, including referral letters if needed, were manually examined and verified. The medical records from the main hospitals outside Reykjavik (Akureyri and Akranes) were searched in the same way. Avulsions less than 5×6 mm, malignancy-caused fractures, and stress fractures were excluded. All fractures were registered according to the International Classification of Diseases [8] (ICD version 10 or ICD version 9). The Reykjavik Study fracture registration has been shown to have a capture rate of about 97 % for hip, forearm, and clinical spine fractures [9]. For the purpose of this report, we identified all major fractures used in the output of FRAX. Major fractures were defined as forearm (S525, S526), humerus (S422, S423), clinical spine fractures, defined as radiological findings following a complaint of back pain with or without a traumatic episode (S120, S121, S122, S220, S221, S320, T08) and hip fracture (S720, S721, S722). All major fractures of the participants, from entry into the study until December 31st 2008, were manually examined, registered, and verified as detailed above. The age- and sex-specific incidence of major fractures was determined in the cohort in 5-year age intervals from the age of 40 years. The age of 40 years was chosen since FRAX undertakes calculations from the age of 40 years. Past records were examined to exclude patients who had previously sustained a fracture at the same site to provide the incidence of a first fracture at the relevant site. From these data, the incidence of a first fracture at any one of these sites was computed, i.e., the incidence of a first major osteoporotic fracture. It is expected that the incidence of a major fracture would be less than the combined incidence of each fracture since a minority of individuals would sustain more than one type of major fracture. The difference in incidence (termed discount) was compared to that computed for Sweden [3]. For this
Osteoporos Int (2014) 25:2445–2451
2447
purpose, we applied the Swedish discounts to the Icelandic data and compared this to the empirical data derived from Iceland. Finally, we compared the incidence of a major fracture observed in Iceland with the incidence that would be predicted from the pattern of incidence in Malmo applied to the incidence of hip fracture in Iceland. This assumes that the age- and sex-specific pattern of incidence of clinical spine fracture, proximal humerus, and forearm fracture (i.e., other major fractures; OMF) and hip fracture (HF) in Reykjavik is similar to that seen in Malmo [10]. Thus at each age and sex,
therefore, HFReykjavik OMFMalmo HFMalmo
Age (years)
40–44 45–49 50–54 55–59 60–64 65–69 70–74 75–79 80–84 85–89 40+
Incidence of fracture
Incidence of first fracture
From this, the incidence of a first major osteoporotic fracture, estimated using the Malmo ratios, was compared with the empirical data from Reykjavik. The annual incidence of a first major osteoporotic fracture in men and women from Iceland by age (with 95 % confidence intervals) was compared with that estimated from the hip fracture rates in Reykjavik and the incidence of other major fractures in Malmo using the formula above. Thus for each age category and sex, OMF Reykjavik was added to the Icelandic hip fracture rates. Table 1 Annual incidence (rate/per 100,000) of major osteoporotic fracture (including those with a prior fracture at the same site)
The mean age at entry into the study was 53 years for women (9.756) (range 33–81 years) and the median follow-up time was 26.7 years with a total number of 257,001 person-years. The mean age was 52 years for men (n=9.116) (range 33– 79 years) and the median follow-up time was 25.8 years, from enrolment to December 31st 2008 or death with a total number of person-years of 225,736.
Table 1 shows the incidence for the four major fractures in men and women in 5-year age groups. Note that individuals may have sustained more than one fracture at any one of these sites. As expected, incidence at all fracture sites increased progressively with age and more markedly in women than in men. The age-specific incidence of hip fracture was 2.5-fold higher in women than in men. The respective ratios for clinical spine, distal forearm, and proximal humerus fracture were 2.8, 4.8, and 3.2.
HFReykjavik OMFReykjavik ¼ HFMalmo OMFMalmo
OMFReykjavik ¼
Results
The incidence of a first fracture at any one site is given in Table 2. As would be expected, the incidence of a first fracture was lower than the incidence of any fracture at any one site. For women, the excess of total incidence of spine fractures over first fractures was 31 % for the age group 75–89 years, compared with 23 % for the age group 60–74 years and 2 % for the age group 45–59 years. For other fractures, the excess of total incidence was between 7 and 9 % in all age groups. For men, the tendency was the same, a total incidence excess for spine fractures in the age group 75–89 years was up to 25 %, compared with 3 % for the age group 45–59 years and 16 % for the age group 60–74 years. However, the total
Spine (clinical)
Distal forearm
Humerus
Hip
Men
Women
Men
Women
Men
Women
Men
Women
0 40 73 67 82 134 216 269 500 827 106
18 41 92 176 220 354 562 856 1,262 1,448 300
50 63 85 109 133 213 198 253 346 518 126
72 214 487 738 791 944 1,024 1,035 1,306 1,438 610
0 32 43 58 49 112 121 213 191 312 65
36 58 83 143 182 285 362 595 667 863 207
0 8 13 24 71 107 260 449 871 1,758 125
0 17 52 98 137 247 504 1,010 1,623 2,917 318
2448 Table 2 Annual incidence (rate/per 100,000) of the first major osteoporotic fracture (excluding those with a prior fracture at the same site)
Osteoporos Int (2014) 25:2445–2451
Age (years)
Spine (clinical)
Distal forearm
Proximal humerus
Hip
Men
Men
Men
Men
Women
Women
Women
Women
40–44
0
18
51
72
0
36
0
0
45–49 50–54 55–59 60–64 65–69 70–74 75–79 80–84 85–89 40+
40 73 61 74 119 174 219 409 638 92
41 92 168 198 290 430 650 940 1,137 244
48 86 107 129 195 176 234 274 462 115
216 466 713 724 897 910 949 1,243 1,237 571
32 43 58 49 110 117 198 152 305 63
58 83 135 182 265 329 540 603 743 192
8 13 24 71 99 239 395 774 1,585 114
17 52 96 126 217 453 892 1,482 2,904 296
incidence for other fractures was higher in men than women or up to 15 %. Overall, the incidence of a first fracture was overestimated from the incidence unadjusted for a previous fracture. For spine, forearm, proximal humerus, and hip fracture, the overestimate was by 15, 10, 3, and 9 % for men. The respective rates for women were 23, 7, 8, and 7 %. As expected, the overestimate increased with age.
Hip fracture ratios The incidence of a first major osteoporotic fracture in Iceland is shown in Fig. 2 for men and Fig. 3 for women. Also given is the incidence of a major osteoporotic fracture estimated from
Table 3 The incidence of a major osteoporotic fracture (per 100,000 person years), by age and sex Age First at any First Ratio of first group major major to combined (years) fracture site fracture fractures
Incidence of first major fracture The combined incidence of a first-major osteoporotic fracture at any of the relevant sites is shown in Table 3. Since individuals may sustain fractures at more than one site, the incidence that is most appropriate for FRAX is the incidence of a first major fracture, also given in Table 3. As would be expected, the incidence of all major fractures overestimated the incidence of a first major fracture. As expected, the overestimate increased with age. Thus, with the knowledge of the combined incidence of a first fracture and the incidence of a first major fracture, a discount can be calculated. For example, in women aged 85–89 years, the combined incidence of any major fracture was 6,022/100,000. In comparison, incidence of a first major fracture was 4,688/100,000. Therefore, the ratio of combined major fracture to the sum was 0.78 and the discount was 22 %. The age- and sex-specific discounts are given in Table 3. Comparison with Malmo When the discounts computed for Malmo were applied to the unadjusted data from Reykjavik (i.e., the combined incidence of all major fractures), the computed incidence of a first major osteoporotic fracture was very similar to that observed in the present study (Fig. 1).
Men
40–44
Discount to combined fractures (sum)
51
51
1.00
0%
45–49 50–54 55–59 60–64 65–69 70–74 75–79 80–84 85–89 Women 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75–79 80–84
127 214 251 323 524 707 1,046 1,609 2,989 126 332 693 1,113 1,230 1,668 2,122 3,031 4,269
119 207 241 298 473 614 886 1,359 2,307 108 325 670 1,056 1,097 1,501 1,803 2,380 3,487
0.94 0.97 0.96 0.92 0.9 0.87 0.85 0.84 0.77 0.86 0.98 0.97 0.95 0.89 0.9 0.85 0.79 0.82
6% 3% 4% 8% 10 % 13 % 15 % 16 % 23 % 14 % 2% 3% 5% 11 % 10 % 15 % 21 % 18 %
85–89
6,022
4,688
0.78
22 %
Incidence is expressed as any major osteoporotic fracture (the first at any site combined) and as the first major fracture at any one site. The last column gives the corresponding discount fractions due to over counting
Osteoporos Int (2014) 25:2445–2451
2449 Incidence (/100,000/year)
Incidence (/100,000 person-years) 5,000
10000 Women
Men (calculated)
4,000
Men (observed)
1000
Women (calculated) Women (observed)
3,000 100
2,000 10 40-49
50-59
60-69
70-79
80-89
90+
Age (years)
1,000
Fig. 3 The annual incidence of a first major osteoporotic fracture in women from Iceland by age (bars and with 95 % confidence intervals). The solid circles denote the incidence of a major fracture computed from the Icelandic hip fracture rates and the Malmo ratios
0 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-90
Age group (years)
Fig. 1 The incidence of a first major osteoporotic fracture (per 100,000 person years), by age and sex observed in the present study and that computed from Malmo
the Icelandic hip fracture rates and the Malmo ratios (see methods). Overall, the incidence of a major fracture was well predicted by the use of the Malmo ratios to hip fracture incidence from Iceland. The exception in men was at the ages of 50–59 years where the use of Malmo ratios underestimated the observed incidence by 28 %. At other ages, the estimates lay within the 95 % confidence intervals of the empirical data. For women (Fig. 3), the exception was between the ages of 60–69 years where the use of Malmo ratios underestimated the observed incidence by 33 %.
Incidence (/100,000/year)
10000 Men
1000
100
10 40-49
50-59
60-69
70-79
80-89
90+
Age (years)
Fig. 2 The annual incidence of a first major osteoporotic fracture in men from Iceland by age (bars and with 95 % confidence intervals). The solid circles denote the incidence of a major fracture computed from the Icelandic hip fracture rates and the Malmo ratios
Discussion The present study permitted us to examine two assumptions that are used in the synthesis of country-specific FRAX models. Both assumptions are derived from studies of fracture incidence in Sweden. The first assumption relates to the discount rates that are applied to age- and sex-specific fracture rates to estimate the incidence of a first major fracture, i.e., to compute the incidence of a clinical spine, distal forearm, proximal humerus, or hip fracture. Data on fracture incidence extracted from registers during a limited time period commonly miss the fact that some fractures may have occurred in an individual who has already suffered a fracture previously. The problem is compounded when multiple fracture outcomes are considered. The discounts derived from Malmo and used in FRAX were mathematically derived from the incidence of each fracture site [3] but were nearly identical to those found in the present study. The result validates the approach used to determine the incidence of a first major fracture in those countries with data on the incidence of two or more fracture outcomes of relevance to FRAX. The second assumption applies to FRAX models that use locally derived hip fracture rates alone to estimate the incidence of a major osteoporotic fracture. For this purpose, it is assumed that the ratio of hip fracture incidence to other FRAX outcomes (clinical spine, distal forearm, and proximal humerus) is the same in the index country as that documented in Sweden. The ratios for Sweden were derived using national hip fracture data for Sweden and data from Malmo for the other fracture outcomes [3]. Despite a large number of studies that have examined the incidence of fractures by age and sex, there are problems in defining the pattern of fractures in different countries [10]. For example, there are differences in the population studied. Some studies have been from random samples of the general
2450
population, from self-selected populations, from accident departments, radiology departments, fracture clinics, or inpatient records. Clinical spine fractures are variously defined, even within a single study [11]. These different sampling frames give rise to large differences in the pattern of fractures reported. Moreover, several surveys do not study or report all fracture types relevant to the outcomes of FRAX, have small samples, an age range not relevant to osteoporosis, or do not include men. A further problem is that the incidence and therefore the apparent pattern of fracture may change with time-discordant samples, so that historical data may not be relevant [10]. There have been recent examples where conclusions concerning the pattern of fracture even in cohort studies have yielded misleading conclusions with further investigation [12–14]. Despite these difficulties, the available information suggests that the pattern of fractures is similar in the Western world and Australia, despite differences in incidence [3, 10]. This commonality of pattern is supported by register studies, which indicate that in those regions where hip fracture rates are high, so too is the risk of forearm fracture and spine fractures (requiring hospital admission) [4, 15, 16]. The present study using a rigorous and standardized sampling frame provided a unique opportunity to examine the relationship of different fracture outcomes. Our findings suggest that the incidence of major fractures can be reasonably predicted from the incidence of hip fracture. Thus, the pattern of osteoporotic fractures appears to be broadly similar in the Western world and suggests that the imputed rates of clinical (but not morphometric) spine, forearm, and humeral fractures used in FRAX are unlikely to be grossly over- or underestimated. The pattern of fractures elsewhere is, however, less secure. The present study has many strengths and some limitations. The study cohort is population based representing 34 % of the relevant Icelandic population born between 1907 and 1935. The study group comprised inhabitants of Reykjavik of which two thirds had migrated to Reykjavik from all parts of the country [17, 18]. It is therefore fair to assume that the cohort is representative of the Icelandic population. The study includes individual follow-up data and the fracture registration was carried out in a standardized way with manual revision of medical records and xrays based on the computerized hospital diagnosis of fractures, including those treated on outpatient basis. The registration has been validated for the fractures that we studied and shown to be 97.5 % complete for major fractures [9]. This study gives thus a rare opportunity to compare the ratio between fracture types in the population. The conclusions of the present study relate only to Iceland but support the view the burden of osteoporosis can be estimated indirectly from hip fracture rates. Notwithstanding, the assumptions when used remain
Osteoporos Int (2014) 25:2445–2451
assumptions and should not obviate the need for highquality empirical data. We conclude that the derivations used in FRAX to estimate the incidence of a first major fracture and the predictive value of hip fracture for other major fractures is supported by the findings of the present study.
Acknowledgments We thank the participants in the Reykjavik Study for their valuable contribution. Conflicts of interest None.
References 1. Kanis JA on behalf of the World Health Organization Scientific Group (2008) Assessment of osteoporosis at the primary healthcare level. Technical Report. WHO Collaborating Centre, University of Sheffield, UK. Available at http://www.shef.ac.uk/ FRAX/index.htm 2. Kanis JA, Johnell O, Oden A, Johansson H, McCloskey EV (2008) FRAX™ and the assessment of fracture probability in men and women from the UK. Osteoporos Int 19:385–397 3. Kanis JA, Johnell O, Oden A et al (2000) Long-term risk of osteoporotic fractures in Malmo. Osteoporos Int 11:669–674 4. Kanis JA, Hans D, Cooper C, The Task Force of the FRAX Initiative et al (2011) Interpretation and use of FRAX in clinical practice. Osteoporos Int 22:395–411 5. Siggeirsdottir K, Aspelund T, Jonsson BYet al. (2014) Epidemiology of fractures in Iceland and secular trends in major osteoporotic fractures 1989–2008. Osteoporos Int. Epub ahead of print July 2, with erratum Oct 22 6. Bjornsson OJ, Davidsson. D., Olafsson H, Olafsson O, Sigfusson N, Thorsteinsson T Th (1979) “Report XVIII. Health Survey in the Reykjavik Area. — Men. Stages I–III, 1967–1968, 1970–1971 and 1974–1975. Participants, Invitation, Response etc.” 7. Bjornsson GBO, Davidsson D, Kristjansson B Th, Olafsson O, Sigfusson N, Thorsteinsson Th (1982) “Report abc XXIV. Health Survey in the Reykjavik Area — Women. Stages I–III, 1968–1969, 1971–1972 and 1976–1978. Participants, Invitation, Response etc.” 8. World Health Organization (2012) International Statistical Classification of Diseases and Related Health Problems 10th Revision. http://www.who.int/classifications/icd/en/. Accessed 15 Apr 9. Siggeirsdottir K, Aspelund T, Sigurdsson G et al (2007) Inaccuracy in self-report of fractures may underestimate association with health outcomes when compared with medical record based fracture registry. Eur J Epidemiol 22:631–639 10. Kanis JA, Oden A, Johnell O, Jonsson B, de Laet C, Dawson A (2001) The burden of osteoporotic fractures: a method for setting intervention thresholds. Osteoporos Int 12:417–427 11. Bow CH, Cheung E, Cheung CL et al (2012) Ethnic difference of clinical vertebral fracture risk. Osteoporos Int 23:879–885 12. Sund R, Olliekainen L, Honkanen R, Kroger H (2013) How well hip fractures can be captured using the postal enquiry? J Bone Miner Res 28(Suppl 1):S292 13. Schousboe JT, Paudel ML, Taylor BC et al (2013) Magnitude and consequences of misclassification of incident hip fractures inlarge cohort studies: the Study of Osteoporotic Fractures and Medicare claims data. Osteoporos Int 24:801–810
Osteoporos Int (2014) 25:2445–2451 14. Lesnyak O, Ershova O, Belova K et al (2012) Epidemiology of fracture in the Russian Federation and the development of a FRAX model. Arch Osteoporos 7:67–73 15. Melton LJ (1995) Epidemiology of fractures. In: Riggs BL, Melton LJ (eds) Osteoporosis: etiology, diagnosis and management, 2nd edn. Lippincott-Raven, Philadelphia, pp 225–227 16. Johnell O, Gullberg B, Kanis JA (1997) The hospital burden of vertebral fracture. A study of national register sources. Osteoporos Int 7:138–144
2451 17. Olafsdottir E, Aspelund T, Torfadottir JE et al (2012) Early life residency associated with the risk of developing type 2 diabetes—the population-based Reykjavik Study. Laeknabladid 98:639–644 18. Lam A, Leslie W, Lix L, Yogendran M, Morin S, Majumdar S (2013) Major osteoporotic to hip fracture ratios in Canadian men and women with Swedish comparisons: a population based analysis. J Bone Miner Res. doi:10.1002/jbmr.2146
ORIGINAL ARTICLE
The incidence of a first major osteoporotic fracture in Iceland and implications for FRAX K. Siggeirsdottir & T. Aspelund & H. Johansson & E. F. Gudmundsson & B. Mogensen & B. Y. Jonsson & V. Gudnason & E. McCloskey & A. Oden & G. Sigurdsson & J. A. Kanis
Received: 31 December 2013 / Accepted: 13 June 2014 / Published online: 1 July 2014 # International Osteoporosis Foundation and National Osteoporosis Foundation 2014
Abstract Summary Based on an extensive cohort study over 25 years, the present study supports the assumption that major osteoporotic fractures can be reasonably predicted from hip fracture rates. Introduction The construct for FRAX models depends on algorithms to adjust for double counting of fracture outcomes in some models and in others, to estimate the incidence of a major fracture from hip fracture rates. The aim of the present study was to test the validity of these algorithms in a large prospective cohort. Methods The incidence of hip, clinical spine, distal forearm, and humerus fracture was determined in the prospective and ongoing population-based Reykjavik Study with follow up of
257,001 person-years. The incidence of a first major fracture was compared with the correction factors used in FRAX to adjust the incidence of several fracture outcomes for double counting. In addition, the incidence of a major osteoporotic fracture estimated from the Icelandic hip fracture rates was compared with the Malmo ratios used in FRAX. Results The adjustments necessary to account for multiple fracture outcomes were similar to those previously derived from Sweden. Additionally, incidence of a first major osteoporotic fracture was similar to that derived for FRAX models. Conclusion The findings of the present study support the algorithms used in FRAX to estimate the incidence of a first major fracture and the predictive value of hip fracture for other major fractures.
K. Siggeirsdottir (*) : T. Aspelund : E. F. Gudmundsson : V. Gudnason : G. Sigurdsson Icelandic Heart Association Research Institute, Holtasmari 1, 201 Kopavogur, Iceland e-mail: [email protected]
Keywords Fracture incidence . FRAX . Major osteoporotic fracture
T. Aspelund : B. Mogensen : V. Gudnason : G. Sigurdsson University of Iceland, Reykjavik, Iceland
Introduction
H. Johansson Centre for Bone and Arthritis Research (CBAR) at the Sahlgrenska Academy, Institute of Medicine, University of Gothenburg, Gothenburg, Sweden H. Johansson : E. McCloskey : A. Oden : J. A. Kanis (*) WHO Collaborating Centre for Metabolic Bone Diseases, University of Sheffield Medical School, Beech Hill Road, Sheffield S10 2RX, UK e-mail: [email protected] B. Mogensen : G. Sigurdsson Landspitalinn, University Hospital Reykjavík, Reykjavík, Iceland B. Y. Jonsson Department of Orthopeadics, Skane University Hospital, Malmo, Sweden
FRAX® is a computer-based algorithm (http://www.shef.ac. uk/FRAX) developed by the World Health Organization Collaborating Center for Metabolic Bone Diseases. The algorithm, intended for primary care, calculates fracture probability from readily obtained clinical risk factors (CRFs) in men and women [1, 2]. The output of FRAX is the 10-year probability of a major osteoporotic fracture (hip, clinical spine, proximal humerus, or wrist fracture) and the 10-year probability of hip fracture. Thus, information is required on the epidemiology of death and fracture in each country for which there is a FRAX model. FRAX models are presently available for 53 regions or countries. All models incorporate country-specific mortality and hip fracture risk. However, a minority of countries that have a FRAX model also have robust information on the risk
2446
of other major fractures (clinical spine, forearm, and proximal humerus). Where available or part available, these are incorporated in the models (e.g., Denmark, Iceland, Japan, Mexico, Russia, Switzerland, Sweden, UK, and US). In the absence of information, FRAX models are based on the assumption that the age- and sex-specific pattern of these fractures is similar to that observed in Sweden. Irrespectively of how fracture rates are derived, account needs to be taken of double counting. For example, studies reporting the incidence of forearm fractures may take account of double counting in the year(s) of study but rarely take account of forearm fractures that might have occurred in much earlier life. When this is accounted for, the incidence of a first forearm fracture from unadjusted population figures is overestimated by about 20 % [3]. In estimating the incidence of a major osteoporotic fracture, account needs also to be taken of double counting different fracture outcomes that are grouped as major fractures. Thus, the metric of interest is the incidence of a first major fracture whether this be at the hip, spine, proximal humerus, or forearm. Such data are available from Malmo, Sweden and mathematical correction factors derived from these data have been applied to new FRAX models where the epidemiology of all major fractures is not available. In the absence of empirical data on fracture incidence, the incidence of fractures at the spine, forearm, and proximal humerus is estimated from the age- and sex-specific pattern of fracture incidence in Malmo, Sweden [4]. Thus, in the absence of detailed information from other sources, FRAX is based on several assumptions derived from the Malmo cohort. Recently, similar detailed information has become available from Iceland [5]. The aim of the present study was to test the validity of the methods used in FRAX to derive the incidence of a major fracture.
Methods The prospective and ongoing population-based Reykjavik Study [6, 7] was initiated in 1967. The Icelandic population was then about 200,000 individuals of who half lived in the capital city Reykjavik and surroundings. The study included both men and women who were residents in the Reykjavik area on December 1st 1967. The participants were born between 1907 and 1935 and selected at random from the Icelandic National Register. The response rate was 71.8 % resulting in 18,872 participants, 9,116 men and 9,756 women who in 1967 represented 34 % of the total Icelandic population in the age range 33 to 65 years. Individuals were followed-up for a median time of 26.5 years until death or December 31st 2008, a total of 482,737 person years.
Osteoporos Int (2014) 25:2445–2451
The study was approved by the National Bioethics Committee and the Data Protection Authority in Iceland. All participants gave informed consent. The Reykjavik Study fracture registration collected all fractures of the participants from entry into the study up to December 31st 2008 [5]. All residents of Iceland have a unique personal identification number allocated at birth or when taking up residence in the country, which facilitates identity and examination of hospital records. Fractures treated on an outpatient basis in Reykjavik were uniquely referred to Landspitalinn University Hospital. Both inpatient and outpatient reports, from all hospitals in Reykjavik including different departments, e.g., trauma, radiography, and outpatients department, were manually examined and verified for fractures until 1983. From 1983, hospitals and the private radiology clinic used by the general practitioners in the Reykjavik area introduced a computerized register system including fracture diagnostic codes. All medical records for the participants, including referral letters if needed, were manually examined and verified. The medical records from the main hospitals outside Reykjavik (Akureyri and Akranes) were searched in the same way. Avulsions less than 5×6 mm, malignancy-caused fractures, and stress fractures were excluded. All fractures were registered according to the International Classification of Diseases [8] (ICD version 10 or ICD version 9). The Reykjavik Study fracture registration has been shown to have a capture rate of about 97 % for hip, forearm, and clinical spine fractures [9]. For the purpose of this report, we identified all major fractures used in the output of FRAX. Major fractures were defined as forearm (S525, S526), humerus (S422, S423), clinical spine fractures, defined as radiological findings following a complaint of back pain with or without a traumatic episode (S120, S121, S122, S220, S221, S320, T08) and hip fracture (S720, S721, S722). All major fractures of the participants, from entry into the study until December 31st 2008, were manually examined, registered, and verified as detailed above. The age- and sex-specific incidence of major fractures was determined in the cohort in 5-year age intervals from the age of 40 years. The age of 40 years was chosen since FRAX undertakes calculations from the age of 40 years. Past records were examined to exclude patients who had previously sustained a fracture at the same site to provide the incidence of a first fracture at the relevant site. From these data, the incidence of a first fracture at any one of these sites was computed, i.e., the incidence of a first major osteoporotic fracture. It is expected that the incidence of a major fracture would be less than the combined incidence of each fracture since a minority of individuals would sustain more than one type of major fracture. The difference in incidence (termed discount) was compared to that computed for Sweden [3]. For this
Osteoporos Int (2014) 25:2445–2451
2447
purpose, we applied the Swedish discounts to the Icelandic data and compared this to the empirical data derived from Iceland. Finally, we compared the incidence of a major fracture observed in Iceland with the incidence that would be predicted from the pattern of incidence in Malmo applied to the incidence of hip fracture in Iceland. This assumes that the age- and sex-specific pattern of incidence of clinical spine fracture, proximal humerus, and forearm fracture (i.e., other major fractures; OMF) and hip fracture (HF) in Reykjavik is similar to that seen in Malmo [10]. Thus at each age and sex,
therefore, HFReykjavik OMFMalmo HFMalmo
Age (years)
40–44 45–49 50–54 55–59 60–64 65–69 70–74 75–79 80–84 85–89 40+
Incidence of fracture
Incidence of first fracture
From this, the incidence of a first major osteoporotic fracture, estimated using the Malmo ratios, was compared with the empirical data from Reykjavik. The annual incidence of a first major osteoporotic fracture in men and women from Iceland by age (with 95 % confidence intervals) was compared with that estimated from the hip fracture rates in Reykjavik and the incidence of other major fractures in Malmo using the formula above. Thus for each age category and sex, OMF Reykjavik was added to the Icelandic hip fracture rates. Table 1 Annual incidence (rate/per 100,000) of major osteoporotic fracture (including those with a prior fracture at the same site)
The mean age at entry into the study was 53 years for women (9.756) (range 33–81 years) and the median follow-up time was 26.7 years with a total number of 257,001 person-years. The mean age was 52 years for men (n=9.116) (range 33– 79 years) and the median follow-up time was 25.8 years, from enrolment to December 31st 2008 or death with a total number of person-years of 225,736.
Table 1 shows the incidence for the four major fractures in men and women in 5-year age groups. Note that individuals may have sustained more than one fracture at any one of these sites. As expected, incidence at all fracture sites increased progressively with age and more markedly in women than in men. The age-specific incidence of hip fracture was 2.5-fold higher in women than in men. The respective ratios for clinical spine, distal forearm, and proximal humerus fracture were 2.8, 4.8, and 3.2.
HFReykjavik OMFReykjavik ¼ HFMalmo OMFMalmo
OMFReykjavik ¼
Results
The incidence of a first fracture at any one site is given in Table 2. As would be expected, the incidence of a first fracture was lower than the incidence of any fracture at any one site. For women, the excess of total incidence of spine fractures over first fractures was 31 % for the age group 75–89 years, compared with 23 % for the age group 60–74 years and 2 % for the age group 45–59 years. For other fractures, the excess of total incidence was between 7 and 9 % in all age groups. For men, the tendency was the same, a total incidence excess for spine fractures in the age group 75–89 years was up to 25 %, compared with 3 % for the age group 45–59 years and 16 % for the age group 60–74 years. However, the total
Spine (clinical)
Distal forearm
Humerus
Hip
Men
Women
Men
Women
Men
Women
Men
Women
0 40 73 67 82 134 216 269 500 827 106
18 41 92 176 220 354 562 856 1,262 1,448 300
50 63 85 109 133 213 198 253 346 518 126
72 214 487 738 791 944 1,024 1,035 1,306 1,438 610
0 32 43 58 49 112 121 213 191 312 65
36 58 83 143 182 285 362 595 667 863 207
0 8 13 24 71 107 260 449 871 1,758 125
0 17 52 98 137 247 504 1,010 1,623 2,917 318
2448 Table 2 Annual incidence (rate/per 100,000) of the first major osteoporotic fracture (excluding those with a prior fracture at the same site)
Osteoporos Int (2014) 25:2445–2451
Age (years)
Spine (clinical)
Distal forearm
Proximal humerus
Hip
Men
Men
Men
Men
Women
Women
Women
Women
40–44
0
18
51
72
0
36
0
0
45–49 50–54 55–59 60–64 65–69 70–74 75–79 80–84 85–89 40+
40 73 61 74 119 174 219 409 638 92
41 92 168 198 290 430 650 940 1,137 244
48 86 107 129 195 176 234 274 462 115
216 466 713 724 897 910 949 1,243 1,237 571
32 43 58 49 110 117 198 152 305 63
58 83 135 182 265 329 540 603 743 192
8 13 24 71 99 239 395 774 1,585 114
17 52 96 126 217 453 892 1,482 2,904 296
incidence for other fractures was higher in men than women or up to 15 %. Overall, the incidence of a first fracture was overestimated from the incidence unadjusted for a previous fracture. For spine, forearm, proximal humerus, and hip fracture, the overestimate was by 15, 10, 3, and 9 % for men. The respective rates for women were 23, 7, 8, and 7 %. As expected, the overestimate increased with age.
Hip fracture ratios The incidence of a first major osteoporotic fracture in Iceland is shown in Fig. 2 for men and Fig. 3 for women. Also given is the incidence of a major osteoporotic fracture estimated from
Table 3 The incidence of a major osteoporotic fracture (per 100,000 person years), by age and sex Age First at any First Ratio of first group major major to combined (years) fracture site fracture fractures
Incidence of first major fracture The combined incidence of a first-major osteoporotic fracture at any of the relevant sites is shown in Table 3. Since individuals may sustain fractures at more than one site, the incidence that is most appropriate for FRAX is the incidence of a first major fracture, also given in Table 3. As would be expected, the incidence of all major fractures overestimated the incidence of a first major fracture. As expected, the overestimate increased with age. Thus, with the knowledge of the combined incidence of a first fracture and the incidence of a first major fracture, a discount can be calculated. For example, in women aged 85–89 years, the combined incidence of any major fracture was 6,022/100,000. In comparison, incidence of a first major fracture was 4,688/100,000. Therefore, the ratio of combined major fracture to the sum was 0.78 and the discount was 22 %. The age- and sex-specific discounts are given in Table 3. Comparison with Malmo When the discounts computed for Malmo were applied to the unadjusted data from Reykjavik (i.e., the combined incidence of all major fractures), the computed incidence of a first major osteoporotic fracture was very similar to that observed in the present study (Fig. 1).
Men
40–44
Discount to combined fractures (sum)
51
51
1.00
0%
45–49 50–54 55–59 60–64 65–69 70–74 75–79 80–84 85–89 Women 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75–79 80–84
127 214 251 323 524 707 1,046 1,609 2,989 126 332 693 1,113 1,230 1,668 2,122 3,031 4,269
119 207 241 298 473 614 886 1,359 2,307 108 325 670 1,056 1,097 1,501 1,803 2,380 3,487
0.94 0.97 0.96 0.92 0.9 0.87 0.85 0.84 0.77 0.86 0.98 0.97 0.95 0.89 0.9 0.85 0.79 0.82
6% 3% 4% 8% 10 % 13 % 15 % 16 % 23 % 14 % 2% 3% 5% 11 % 10 % 15 % 21 % 18 %
85–89
6,022
4,688
0.78
22 %
Incidence is expressed as any major osteoporotic fracture (the first at any site combined) and as the first major fracture at any one site. The last column gives the corresponding discount fractions due to over counting
Osteoporos Int (2014) 25:2445–2451
2449 Incidence (/100,000/year)
Incidence (/100,000 person-years) 5,000
10000 Women
Men (calculated)
4,000
Men (observed)
1000
Women (calculated) Women (observed)
3,000 100
2,000 10 40-49
50-59
60-69
70-79
80-89
90+
Age (years)
1,000
Fig. 3 The annual incidence of a first major osteoporotic fracture in women from Iceland by age (bars and with 95 % confidence intervals). The solid circles denote the incidence of a major fracture computed from the Icelandic hip fracture rates and the Malmo ratios
0 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85-90
Age group (years)
Fig. 1 The incidence of a first major osteoporotic fracture (per 100,000 person years), by age and sex observed in the present study and that computed from Malmo
the Icelandic hip fracture rates and the Malmo ratios (see methods). Overall, the incidence of a major fracture was well predicted by the use of the Malmo ratios to hip fracture incidence from Iceland. The exception in men was at the ages of 50–59 years where the use of Malmo ratios underestimated the observed incidence by 28 %. At other ages, the estimates lay within the 95 % confidence intervals of the empirical data. For women (Fig. 3), the exception was between the ages of 60–69 years where the use of Malmo ratios underestimated the observed incidence by 33 %.
Incidence (/100,000/year)
10000 Men
1000
100
10 40-49
50-59
60-69
70-79
80-89
90+
Age (years)
Fig. 2 The annual incidence of a first major osteoporotic fracture in men from Iceland by age (bars and with 95 % confidence intervals). The solid circles denote the incidence of a major fracture computed from the Icelandic hip fracture rates and the Malmo ratios
Discussion The present study permitted us to examine two assumptions that are used in the synthesis of country-specific FRAX models. Both assumptions are derived from studies of fracture incidence in Sweden. The first assumption relates to the discount rates that are applied to age- and sex-specific fracture rates to estimate the incidence of a first major fracture, i.e., to compute the incidence of a clinical spine, distal forearm, proximal humerus, or hip fracture. Data on fracture incidence extracted from registers during a limited time period commonly miss the fact that some fractures may have occurred in an individual who has already suffered a fracture previously. The problem is compounded when multiple fracture outcomes are considered. The discounts derived from Malmo and used in FRAX were mathematically derived from the incidence of each fracture site [3] but were nearly identical to those found in the present study. The result validates the approach used to determine the incidence of a first major fracture in those countries with data on the incidence of two or more fracture outcomes of relevance to FRAX. The second assumption applies to FRAX models that use locally derived hip fracture rates alone to estimate the incidence of a major osteoporotic fracture. For this purpose, it is assumed that the ratio of hip fracture incidence to other FRAX outcomes (clinical spine, distal forearm, and proximal humerus) is the same in the index country as that documented in Sweden. The ratios for Sweden were derived using national hip fracture data for Sweden and data from Malmo for the other fracture outcomes [3]. Despite a large number of studies that have examined the incidence of fractures by age and sex, there are problems in defining the pattern of fractures in different countries [10]. For example, there are differences in the population studied. Some studies have been from random samples of the general
2450
population, from self-selected populations, from accident departments, radiology departments, fracture clinics, or inpatient records. Clinical spine fractures are variously defined, even within a single study [11]. These different sampling frames give rise to large differences in the pattern of fractures reported. Moreover, several surveys do not study or report all fracture types relevant to the outcomes of FRAX, have small samples, an age range not relevant to osteoporosis, or do not include men. A further problem is that the incidence and therefore the apparent pattern of fracture may change with time-discordant samples, so that historical data may not be relevant [10]. There have been recent examples where conclusions concerning the pattern of fracture even in cohort studies have yielded misleading conclusions with further investigation [12–14]. Despite these difficulties, the available information suggests that the pattern of fractures is similar in the Western world and Australia, despite differences in incidence [3, 10]. This commonality of pattern is supported by register studies, which indicate that in those regions where hip fracture rates are high, so too is the risk of forearm fracture and spine fractures (requiring hospital admission) [4, 15, 16]. The present study using a rigorous and standardized sampling frame provided a unique opportunity to examine the relationship of different fracture outcomes. Our findings suggest that the incidence of major fractures can be reasonably predicted from the incidence of hip fracture. Thus, the pattern of osteoporotic fractures appears to be broadly similar in the Western world and suggests that the imputed rates of clinical (but not morphometric) spine, forearm, and humeral fractures used in FRAX are unlikely to be grossly over- or underestimated. The pattern of fractures elsewhere is, however, less secure. The present study has many strengths and some limitations. The study cohort is population based representing 34 % of the relevant Icelandic population born between 1907 and 1935. The study group comprised inhabitants of Reykjavik of which two thirds had migrated to Reykjavik from all parts of the country [17, 18]. It is therefore fair to assume that the cohort is representative of the Icelandic population. The study includes individual follow-up data and the fracture registration was carried out in a standardized way with manual revision of medical records and xrays based on the computerized hospital diagnosis of fractures, including those treated on outpatient basis. The registration has been validated for the fractures that we studied and shown to be 97.5 % complete for major fractures [9]. This study gives thus a rare opportunity to compare the ratio between fracture types in the population. The conclusions of the present study relate only to Iceland but support the view the burden of osteoporosis can be estimated indirectly from hip fracture rates. Notwithstanding, the assumptions when used remain
Osteoporos Int (2014) 25:2445–2451
assumptions and should not obviate the need for highquality empirical data. We conclude that the derivations used in FRAX to estimate the incidence of a first major fracture and the predictive value of hip fracture for other major fractures is supported by the findings of the present study.
Acknowledgments We thank the participants in the Reykjavik Study for their valuable contribution. Conflicts of interest None.
References 1. Kanis JA on behalf of the World Health Organization Scientific Group (2008) Assessment of osteoporosis at the primary healthcare level. Technical Report. WHO Collaborating Centre, University of Sheffield, UK. Available at http://www.shef.ac.uk/ FRAX/index.htm 2. Kanis JA, Johnell O, Oden A, Johansson H, McCloskey EV (2008) FRAX™ and the assessment of fracture probability in men and women from the UK. Osteoporos Int 19:385–397 3. Kanis JA, Johnell O, Oden A et al (2000) Long-term risk of osteoporotic fractures in Malmo. Osteoporos Int 11:669–674 4. Kanis JA, Hans D, Cooper C, The Task Force of the FRAX Initiative et al (2011) Interpretation and use of FRAX in clinical practice. Osteoporos Int 22:395–411 5. Siggeirsdottir K, Aspelund T, Jonsson BYet al. (2014) Epidemiology of fractures in Iceland and secular trends in major osteoporotic fractures 1989–2008. Osteoporos Int. Epub ahead of print July 2, with erratum Oct 22 6. Bjornsson OJ, Davidsson. D., Olafsson H, Olafsson O, Sigfusson N, Thorsteinsson T Th (1979) “Report XVIII. Health Survey in the Reykjavik Area. — Men. Stages I–III, 1967–1968, 1970–1971 and 1974–1975. Participants, Invitation, Response etc.” 7. Bjornsson GBO, Davidsson D, Kristjansson B Th, Olafsson O, Sigfusson N, Thorsteinsson Th (1982) “Report abc XXIV. Health Survey in the Reykjavik Area — Women. Stages I–III, 1968–1969, 1971–1972 and 1976–1978. Participants, Invitation, Response etc.” 8. World Health Organization (2012) International Statistical Classification of Diseases and Related Health Problems 10th Revision. http://www.who.int/classifications/icd/en/. Accessed 15 Apr 9. Siggeirsdottir K, Aspelund T, Sigurdsson G et al (2007) Inaccuracy in self-report of fractures may underestimate association with health outcomes when compared with medical record based fracture registry. Eur J Epidemiol 22:631–639 10. Kanis JA, Oden A, Johnell O, Jonsson B, de Laet C, Dawson A (2001) The burden of osteoporotic fractures: a method for setting intervention thresholds. Osteoporos Int 12:417–427 11. Bow CH, Cheung E, Cheung CL et al (2012) Ethnic difference of clinical vertebral fracture risk. Osteoporos Int 23:879–885 12. Sund R, Olliekainen L, Honkanen R, Kroger H (2013) How well hip fractures can be captured using the postal enquiry? J Bone Miner Res 28(Suppl 1):S292 13. Schousboe JT, Paudel ML, Taylor BC et al (2013) Magnitude and consequences of misclassification of incident hip fractures inlarge cohort studies: the Study of Osteoporotic Fractures and Medicare claims data. Osteoporos Int 24:801–810
Osteoporos Int (2014) 25:2445–2451 14. Lesnyak O, Ershova O, Belova K et al (2012) Epidemiology of fracture in the Russian Federation and the development of a FRAX model. Arch Osteoporos 7:67–73 15. Melton LJ (1995) Epidemiology of fractures. In: Riggs BL, Melton LJ (eds) Osteoporosis: etiology, diagnosis and management, 2nd edn. Lippincott-Raven, Philadelphia, pp 225–227 16. Johnell O, Gullberg B, Kanis JA (1997) The hospital burden of vertebral fracture. A study of national register sources. Osteoporos Int 7:138–144
2451 17. Olafsdottir E, Aspelund T, Torfadottir JE et al (2012) Early life residency associated with the risk of developing type 2 diabetes—the population-based Reykjavik Study. Laeknabladid 98:639–644 18. Lam A, Leslie W, Lix L, Yogendran M, Morin S, Majumdar S (2013) Major osteoporotic to hip fracture ratios in Canadian men and women with Swedish comparisons: a population based analysis. J Bone Miner Res. doi:10.1002/jbmr.2146
